Postal Rate Commission
Submitted 4/8/2005 12:24 pm
Filing ID: 43446
Accepted 4/8/2005
USPS-T-7
BEFORE THE
POSTAL RATE COMMISSION
WASHINGTON, D.C. 20268-0001
POSTAL RATE AND FEE CHANGES
PURSUANT TO PUBLIC LAW 108-18
DIRECT TESTIMONY
OF
THOMAS E. THRESS
ON BEHALF OF THE
UNITED STATES POSTAL SERVICE
Docket No. R2005-1
Table of Contents
AUTOBIOGRAPHICAL SKETCH .................................................................................... 1
PURPOSE AND SCOPE OF TESTIMONY..................................................................... 2
I. Introduction .............................................................................................................. 3
A. General Outline of Testimony and Supporting Material, and Relationship to Other
Testimonies ................................................................................................................. 3
B. Overview of My Testimony ................................................................................... 4
1. Centerpiece of Postal Service Volume Forecasts ............................................. 4
2. Volume Forecasting Methodology..................................................................... 5
3. Econometric Methodology................................................................................. 6
C.
Summary of Final Results ................................................................................. 7
D.
Comparison with Previous Methodology ......................................................... 13
II. Analysis of Factors Affecting Mail Volume ............................................................. 15
A. General Overview............................................................................................... 15
1. Division of Mail for Estimation Purposes ......................................................... 15
2. Sources of Information used in Modeling Demand Equations......................... 17
a. General Demand Equation Methodology .................................................... 17
b. Postal Volume Data..................................................................................... 18
c. Factors Affecting Demand ........................................................................... 18
i.
Price......................................................................................................... 18
ii. Macroeconomic Variables........................................................................ 21
(a)
Retail Sales ...................................................................................... 21
(b)
Mail-Order Retail Sales..................................................................... 22
(c)
Employment...................................................................................... 22
(d)
Total Real Investment....................................................................... 23
iii. The Internet and Electronic Diversion ...................................................... 23
(a)
History of Electronic Diversion in Postal Service Demand Equations23
(b)
Relationship of Mail Volume to the Internet ...................................... 25
(c)
Internet Variables Considered Here.................................................. 26
(i) Consumption Expenditures on Internet Service Providers................ 26
(ii) Internet Experience .......................................................................... 29
(iii) Number of Broadband Subscribers ................................................. 30
(iv) Average Time Spent Online ............................................................ 31
(v) Measures of Direct Substitution ....................................................... 32
(vi) Internet Advertising Expenditures.................................................... 33
iv.
Time Trends ......................................................................................... 33
v. Other Variables........................................................................................ 35
vi.
Seasonality........................................................................................... 36
3. Forecasting Philosophy................................................................................... 40
B. First-Class Mail................................................................................................... 43
1. General Overview ........................................................................................... 43
2. History of First-Class Mail Volume .................................................................. 44
3. Trends in First-Class Mail Volume .................................................................. 48
4. Shifts Between First-Class Single-Piece and Workshared Letters Due to
Changes in Worksharing Discounts ....................................................................... 50
5.
First-Class Single-Piece Letters...................................................................... 53
a. Factors Affecting First-Class Single-Piece Letters Volume ......................... 53
b. Econometric Demand Equation................................................................... 57
i.
Summary of Demand Equation................................................................ 57
ii. Box-Cox Transformation of Internet Experience Variable ........................ 58
6. First-Class Workshared Letters....................................................................... 61
a. Factors Affecting First-Class Workshared Letters Volume .......................... 61
b. Econometric Demand Equation................................................................... 64
7. Understanding First-Class Letters Price Elasticities........................................ 67
8. First-Class Cards ............................................................................................ 70
a. Volume History ............................................................................................ 70
b. Factors Affecting First-Class Cards Volume................................................ 71
i.
Economy.................................................................................................. 71
ii. Internet..................................................................................................... 72
iii. Prices....................................................................................................... 72
iv.
Summary of Demand Equation Specification ....................................... 73
c. Econometric Demand Equation................................................................... 76
C.
Standard Mail.................................................................................................. 79
1. Overview of Direct-Mail Advertising ................................................................ 79
2. Advertising Decisions and Their Impact on Mail Volume ................................ 81
a. How Much to Invest in Advertising .............................................................. 81
b. Which Advertising Media to Use.................................................................. 83
i.
Price of Direct-Mail Advertising................................................................ 83
(a)
Postage Costs .................................................................................. 83
(b)
Paper and Printing Costs.................................................................. 83
(c)
Technological Costs ......................................................................... 83
ii. Competing Advertising Media .................................................................. 85
iii. Relationship between the Internet and Direct-Mail Advertising................ 86
c. How to Send Mail-Based Advertising .......................................................... 88
3. Final Equation Specifications for Standard Mail.............................................. 90
a. Overview of Standard Mail Subclasses ....................................................... 90
b. Standard Regular Mail................................................................................. 93
i.
Factors Affecting Standard Regular Mail Volume .................................... 93
ii. Econometric Demand Equation ............................................................... 97
c. Standard Enhanced Carrier Route (ECR) ................................................. 100
i.
Factors Affecting Standard ECR Mail Volume ....................................... 100
ii. Econometric Demand Equation ............................................................. 103
d. Standard Bulk Nonprofit ............................................................................ 106
i.
Factors Affecting Standard Bulk Nonprofit Mail Volume ........................ 106
ii. Econometric Demand Equation ............................................................. 110
D.
Expedited Delivery Services ......................................................................... 113
1. General Overview ......................................................................................... 113
2. Factors Affecting Express Mail Volume ....................................................... 115
a. Demand for Overnight Delivery Services................................................... 115
b. Demand for Express Mail as Overnight Delivery Service of Choice .......... 116
3. Demand Equation for Express Mail............................................................... 116
a. Factors Affecting Express Mail Volume .................................................... 116
b. Econometric Demand Equation................................................................. 120
E. Package Delivery Services ............................................................................... 122
1. Overview of Ground Package Delivery Market ............................................. 122
2. Final Demand Equations............................................................................... 125
a. Priority Mail................................................................................................ 127
i.
Factors Affecting Priority Mail ................................................................ 127
ii. Econometric Demand Equation ............................................................. 133
b. Parcel Post ................................................................................................ 137
i.
General Overview .................................................................................. 137
ii. Non-Destination-Entry Parcel Post Mail................................................. 141
(a)
Factors Affecting Non-Destination Entry Parcel Post Mail .............. 141
(b)
Econometric Demand Equation ...................................................... 144
iii. Destination Entry Parcel Post ................................................................ 147
(a)
Factors Affecting Destination Entry Parcel Post Mail...................... 147
(b)
Econometric Demand Equation ...................................................... 150
c. Other Package Services............................................................................ 152
i.
Overview of Bound Printed Matter and Media Mail................................ 152
(a)
History of Bound Printed Matter and Media Mail Subclasses ......... 152
(b)
Factors affecting Demand for Bound Printed Matter and Media Mail
156
ii. Bound Printed Matter Demand Equation ............................................... 156
(a)
Factors Affecting the Bound Printed Matter Demand Equation ...... 156
(b)
Econometric Demand Equation ...................................................... 160
iii. Media and Library Rate Mail Demand Equation .................................... 162
(a)
Factors Affecting the Media and Library Rate Demand Equation ... 162
(b)
Econometric Demand Equation ...................................................... 164
F. Periodicals Mail ................................................................................................ 167
1. General Overview ......................................................................................... 167
2. Factors Affecting Demand for Periodicals Mail ............................................. 170
3. Regular Rate................................................................................................. 174
a. Factors Affecting Periodical Regular Rate Volume.................................... 174
b. Econometric Demand Equation................................................................. 177
4. Preferred Periodicals Subclasses ................................................................. 179
a. Overview ................................................................................................... 179
b. Within-County............................................................................................ 180
i.
Factors Affecting Periodical Within-County Mail Volume ....................... 180
ii. Econometric Demand Equation ............................................................. 183
c. Nonprofit and Classroom Mail ................................................................... 186
i.
Factors Affecting Periodical Nonprofit and Classroom Mail Volume ...... 186
ii. Econometric Demand Equation ............................................................. 189
G.
Other Mail Categories ................................................................................... 191
1. Mailgrams ..................................................................................................... 191
2. Postal Penalty Mail ....................................................................................... 194
3. Free for the Blind and Handicapped Mail...................................................... 197
H.
Special Services ........................................................................................... 200
1.
2.
General Overview ......................................................................................... 200
Registered Mail ............................................................................................. 201
a. Factors Affecting Registered Mail Volume................................................. 201
b. Econometric Demand Equation................................................................. 204
3. Postal Insurance ........................................................................................... 207
a. Factors Affecting Insured Mail Volume ...................................................... 207
b. Econometric Demand Equation................................................................. 211
4. Certified Mail ................................................................................................. 213
a. Factors Affecting Certified Mail Volume .................................................... 213
b. Econometric Demand Equation................................................................. 215
5. Collect on Delivery (COD) Mail ..................................................................... 218
a. Factors Affecting COD Mail Volume.......................................................... 218
b. Econometric Demand Equation................................................................. 220
6. Return Receipts ............................................................................................ 223
a. Factors Affecting Return Receipts Volume................................................ 223
b. Econometric Demand Equation................................................................. 225
7. Money Orders ............................................................................................... 228
a. History of Money Orders Volume .............................................................. 228
b. Factors Affecting Money Orders Volume................................................... 230
c. Econometric Demand Equation................................................................. 232
8. Delivery and Signature Confirmation ............................................................ 235
a. Factors Affecting Delivery and Signature Confirmation ............................. 235
b. Econometric Demand Equation................................................................. 239
9. Post Office Boxes ......................................................................................... 241
a. Estimating the Number of Post Office Boxes............................................. 241
b. Factors Affecting Post Office Boxes .......................................................... 243
c. Econometric Demand Equation................................................................. 245
10.
Stamped Cards ......................................................................................... 248
a. Estimating Stamped Cards Volume........................................................... 248
b. Factors Affecting Stamped Cards Volume................................................. 249
c. Econometric Demand Equation................................................................. 252
III.
Econometric Demand Equation Methodology .................................................. 254
A. Functional Form of the Equation....................................................................... 254
B. Data Used in Modeling Demand Equations ...................................................... 254
C.
Basic Ordinary Least Squares Model............................................................ 257
D.
Adjustments to the Basic Ordinary Least Squares Model ............................. 257
1. Introduction of Outside Restrictions into OLS Estimation.............................. 257
2. Multicollinearity ............................................................................................. 259
a. Shiller Smoothness Priors ......................................................................... 260
b. Special Note on Price Lags ....................................................................... 263
c. Slutsky-Schultz Symmetry Condition......................................................... 264
3. Autocorrelation.............................................................................................. 266
E. Summary of Demand Equation to be Estimated............................................... 269
F. Step-by-Step Examples.................................................................................... 270
1. First-Class Workshared Letters..................................................................... 270
a. Basic Demand Equation ............................................................................ 270
b.
c.
d.
Seasonal Variables ................................................................................... 271
Restriction Matrices................................................................................... 272
Final Econometric Estimates ..................................................................... 273
2. Standard Regular Mail .................................................................................. 273
a. Basic Demand Equation ............................................................................ 273
b. Restriction Matrices................................................................................... 274
c. Final Econometric Estimates ..................................................................... 275
IV.
Volume Forecasting Methodology .................................................................... 277
A. Volume Forecasting Equation........................................................................... 277
B. Base Period Used in Forecasting ..................................................................... 278
C.
Projection Factors ......................................................................................... 278
1. Rate Effect Multiplier ..................................................................................... 278
2. Non-Rate Effect Multiplier ............................................................................. 279
3. Forecasts of Non-Rate Variables .................................................................. 280
a. Mechanical Non-Rate Variables ................................................................ 280
b. Macroeconomic Variables Forecasted by Global Insight........................... 281
c. Non-Rate Variables Forecasted Here ....................................................... 281
i.
Mail-Order Retail Sales.......................................................................... 281
ii. Total Advertising Expenditures .............................................................. 286
iii. Prices of Newspaper and Direct-Mail Advertising .................................. 291
iv.
Average Delivery Days, Priority Mail .................................................. 293
v. Measures of Electronic Diversion Used Here ........................................ 296
(a)
Consumption Expenditures on Internet Service Providers.............. 297
(b)
Number of Broadband Subscribers................................................. 301
(c)
Internet Advertising Expenditures ................................................... 303
4. Composite Multiplier ..................................................................................... 307
V. Forecasts of Component Mail Categories ............................................................ 309
A. Overview .......................................................................................................... 309
B. Presort and Automation Shares........................................................................ 309
1. Basic Theory of Consumer Worksharing ...................................................... 309
2. Derivation of Basic Share Equation .............................................................. 311
3. Modeling Consumers’ Use of Worksharing Options...................................... 312
a. Modeling User-Cost Distributions .............................................................. 313
i.
Issue 1: Population of Potential Worksharers ....................................... 313
ii. Issue 2: Negative User Costs ............................................................... 314
iii. Issue 3: Non-Integrability of Normal p.d.f.............................................. 315
iv.
Resolution of Issues ........................................................................... 315
b. Changes in the User-Cost Distribution over Time ..................................... 316
i.
Changes in the Type of Distribution....................................................... 317
ii. Changes in the Standard Deviation of the Distribution .......................... 318
iii. Changes in the Ceiling of the Distribution .............................................. 318
iv.
Changes in the Mean of the Distribution ............................................ 319
c. Opportunity Costs...................................................................................... 319
d. Empirical Problem to be Solved to Model Use of Worksharing ................. 321
4. Econometric Share Equations....................................................................... 321
5. Share Forecast Equations............................................................................. 324
6.
Presort and Automation Share Equations by Mail Subclass ......................... 324
a. First-Class Workshared Letters ................................................................. 324
i.
Automation Basic Letters ....................................................................... 325
ii. Automation Basic Flats .......................................................................... 325
iii. Automation 3-Digit Letters ..................................................................... 325
iv.
Automation 5-Digit Letters.................................................................. 326
v. Automation 3- and 5-Digit Flats ............................................................. 327
vi.
Automation Carrier-Route Letters ...................................................... 327
b. First-Class Cards....................................................................................... 328
i.
Nonautomation Presort .......................................................................... 328
ii. Automation Basic................................................................................... 328
iii. Automation 3-Digit ................................................................................. 329
iv.
Automation 5-Digit.............................................................................. 329
v. Automation Carrier-Route ...................................................................... 330
c. Standard Regular Mail............................................................................... 330
i.
Automation Basic Letters ....................................................................... 330
ii. Automation Basic Flats .......................................................................... 330
iii. Automation 3-Digit Letters ..................................................................... 331
iv.
Automation 5-Digit Letters.................................................................. 331
v. Automation 3/5-Digit Flats...................................................................... 332
d. Standard Nonprofit Mail............................................................................. 333
i.
Automation Basic Letters ....................................................................... 333
ii. Automation Basic Flats .......................................................................... 333
iii. Automation 3-Digit Letters ..................................................................... 333
iv.
Automation 5-Digit Letters.................................................................. 334
v. Automation 3/5-Digit Flats...................................................................... 334
7. Final Presort and Automation Share Forecasts ............................................ 335
C.
Standard Regular Letters versus Non-Letters.............................................. 344
ATTACHMENT A: QUARTERLY VOLUME FORECASTS
Library References Associated with This Testimony:
LR-K-63: DATA USED IN DEMAND ANALYSIS AND VOLUME FORECASTING
LR-K-64: DEMAND ANALYSIS ECONOMETRIC MATERIALS
LR-K-65: DEMAND ANALYSIS ECONOMETRIC CHOICE TRAIL
LR-K-66: VOLUME FORECASTING MATERIALS
USPS-T-7
1
1
2
3
4
DIRECT TESTIMONY
OF
THOMAS E. THRESS
5
AUTOBIOGRAPHICAL SKETCH
6
7
My name is Thomas E. Thress. I am a Vice-President at RCF Economic and
8
Financial Consulting, Inc., where I have been employed since 1992. As a Vice
9
President at RCF, I have major responsibilities in RCF’s forecasting, econometric, and
10
quantitative analysis activities. I have had primary responsibility for the econometric
11
analysis underlying Dr. George Tolley’s volume forecasting testimony since Docket No.
12
R94-1. In addition, I was responsible for the development of the share equation
13
methodology used by the Postal Service since MC95-1, as well as the classification shift
14
matrix construction used in Dr. Tolley’s volume forecasting testimony in MC95-1 and
15
MC96-2 to shift mail into the new categories proposed under classification reform.
16
I testified regarding the demand equations underlying the volume forecasts for all
17
mail categories except for Priority and Express Mail in Docket Nos. R97-1, R2000-1,
18
and R2001-1. Prior to that, I appeared as a rebuttal witness for the Postal Service in
19
Docket No. MC95-1, and submitted written testimony for the Postal Service in Docket
20
No. MC97-2.
21
I completed my Master’s Degree in Economics in 1992 at the University of Chicago.
22
I received a B.A. in Economics and a B.S. in Mathematics from Valparaiso University in
23
1990.
USPS-T-7
2
1
PURPOSE AND SCOPE OF TESTIMONY
2
The purpose of this testimony is to model the demand for domestic mail volume, to
3
identify and quantify the factors which affect mail volumes, and to project these factors
4
through the Test Period for the purposes of developing a set of volume forecasts. The
5
work presented in this testimony is intimately connected to and is best read in concert
6
with the testimony of Peter Bernstein (USPS-T-8).
USPS-T-7
3
1
2
3
4
5
I. Introduction
A. General Outline of Testimony and Supporting Material, and Relationship to
Other Testimonies
My testimony is divided into five sections. Section I (this section) provides a brief
6
overview of the demand equation estimation and volume forecasting methodology and
7
summarizes the final results. Section II provides a detailed analysis of each of the 27
8
mail categories and special services considered in my testimony. The econometric
9
methodology used in developing this work is described in detail in Section III below.
10
Section IV describes the volume forecasting methodology used here in some detail, with
11
a particular emphasis on the forecasts of the explanatory variables developed here.
12
Finally, Section V presents forecasts of some mail categories at a finer level of detail
13
than those presented in Section II, including the shares of First-Class and Standard Mail
14
that are expected to take advantage of automation discounts and the relative shares of
15
Standard Regular letters and non-letters.
16
My direct testimony is supported by four library references. The first of these, LR-K-
17
63 presents the data used in the development of my testimony. Data are presented
18
historically as well as for the forecast period of interest. The sources of the data are
19
detailed and any adjustments made to the data by me prior to their use are documented
20
in that library reference. The second library reference supporting this testimony, LR-K-
21
64, documents the econometric results presented here as well as the data and
22
programs necessary to replicate these results. Library reference LR-K-65 presents
23
some intermediate econometric results which were used in the development of my
24
testimony. Finally, library reference LR-K-66 presents the spreadsheets which were
25
used to make the before-rates and after-rates volume forecasts for this case and
26
provides a step-by-step example of the volume forecasting process.
USPS-T-7
4
1
Inputs from postal sources used by me in my testimony include RWP, ODIS, and
2
billing determinant data from the Postal Service, as well as after-rates prices from the
3
pricing witnesses. Because of the central role of volume forecasts in ratemaking, my
4
outputs may be employed by virtually any witness whose testimony touches upon the
5
test year. The most prominent users, however, are witnesses Waterbury (USPS-T-10)
6
and Kay (USPS-T-18) for purposes of developing test year costs, witness Tayman
7
(USPS-T-6) for the revenue requirement, and witnesses Robinson (USPS-T-27) and
8
Taufique (USPS-T-28) for purposes of rate policy, rate design, and revenue estimation.
9
10
11
B. Overview of My Testimony
1. Centerpiece of Postal Service Volume Forecasts
The centerpiece of the volume forecasts presented in this case is a series of
12
equations that attempt to relate the volume of mail to a number of factors, referred to
13
here as explanatory variables. The basic format of these equations is as follows:
Vt = a·x1te1·x2te2·…·xnten·εt
(Equation 1)
14
15
16
where Vt is volume at time t, x1 to xn are explanatory variables, e1 to en are elasticities
17
associated with these variables, and εt represents the residual, or unexplained, factor(s)
18
affecting mail volume.
19
20
21
22
23
24
25
26
Equations along the line of Equation 1 are developed here for 27 categories of
domestic mail and special services:
First-Class Mail
First-Class Single-Piece Letters, Flats, and Irregular Packages and
Parcels (IPPs)
First-Class Workshared Letters, Flats, and IPPs
First-Class Cards
USPS-T-7
5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Standard Mail
Standard Regular Mail
Standard Enhanced Carrier Route (ECR) Mail
Standard Bulk Nonprofit (Nonprofit and Nonprofit ECR) Mail
Expedited Delivery Services
Express Mail
Package Delivery Services
Priority Mail
Single-Piece (i.e., non-destination entry) Parcel Post
Bulk (i.e., destination entry) Parcel Post
Bound Printed Matter
Media and Library Rate Mail
Periodicals Mail
Periodicals Regular Rate Mail
Periodicals Within-County Mail
Periodicals Nonprofit and Classroom Mail
Other Mail Categories
Mailgrams
Postal Penalty Mail
Free for the Blind and Handicapped Mail
Special Services
Registered Mail
Postal Insurance
Certified Mail
COD (Collect-on-Delivery or Cash-on-Delivery)
Return Receipts
Money Orders
Delivery and Signature Confirmation
Post Office Boxes
Stamped Cards
2. Volume Forecasting Methodology
38
As noted above, the basic forecasting equation employed here, Equation 1, relates
39
volume at time t to a series of explanatory variables according to the following formula:
40
41
Vt = a·x1te1·x2te2·…·xnten·εt
(Equation 1)
USPS-T-7
6
1
Equation 1 is assumed to hold both historically as well as into the forecast period.
2
Of particular interest, Equation 1 is assumed to hold over the most recent time period,
3
called the Base Period. That is,
4
5
6
7
8
9
10
11
VB = a·x1Be1·x2Be2·…·xnBen·εB
(Equation 1B)
Dividing Equation 1 by Equation 1B, for forecast time period t and multiplying both
sides by VB yields the following equation:
Vt = VB·[x1t/x1B]e1·[x2t/x2B]e2·…· [xnt/xnB]en·[εt/εB]
(Equation 2)
Equation 2 forms the basis for the Postal Service’s volume forecasts. The Postal
12
Service’s volume forecasting methodology is sometimes referred to as a base-volume
13
forecasting methodology. The logic of this name can be seen quite readily in Equation
14
2. Using this forecasting methodology, volume at time t (Vt) is projected to be equal to
15
volume in the base period (VB) times a series of multipliers which reflect the extent to
16
which the explanatory variables have changed from the base period to time t. The
17
volume forecasting methodology is described in more detail in Section IV below.
18
This testimony, then, is concerned with properly identifying explanatory variables, xit,
19
which affect mail volume; forecasting these explanatory variables as necessary; and
20
estimating a set of elasticities, ei, to be used in forecasting.
21
3. Econometric Methodology
22
The centerpiece of the econometric work presented here is a set of demand
23
equations for the 27 domestic mail categories and special services identified above.
24
These demand equations take the form of Equation 1 above. In general, variables
25
which are believed to influence the demand for mail volume are introduced into an
26
econometric equation as a quarterly time series in which the elasticity of mail volume
USPS-T-7
7
1
with respect to the particular variable is estimated using a Generalized Least Squares
2
estimation procedure. The complete econometric methodology is described in detail in
3
Section III below. The explanatory variables considered here include Postal prices,
4
other input prices (e.g., printing, paper), prices of competing goods (e.g., non-direct-mail
5
advertising, UPS, FedEx), measures of macroeconomic activity (e.g., retail sales,
6
employment, investment), measures of potential electronic substitutes for the mail (e.g.,
7
Internet usage, Internet advertising expenditures), time trends, seasonal variables, and
8
other variables as needed.
9
The specific equations which are ultimately used to make mail volume forecasts are
10
described in detail in Section II below.
11
C. Summary of Final Results
12
The end product of this work is, of course, Test Year volume forecasts with and
13
without the rate increase being requested by the Postal Service. Table 1 below
14
presents Test Year before- and after-rates volume forecasts. Projected average annual
15
growth rates over the next two Government Fiscal Years (2004 – 2006) are also shown,
16
along with historical average annual growth rates over the most recent two Government
17
Fiscal Years (2002 – 2004).
18
Long-run Postal own-price elasticities are also shown in Table 1. These measure
19
the percentage change in volume that is expected as a result of a one percent change
20
in the price of that category of mail, holding all other things constant. In most cases, the
21
full long-run impact of prices is not expected to be felt for several quarters after a rate
22
increase. In these cases, the effect of the Postal Service’s proposed rate increases on
23
volume will not be fully realized until at least partway through the Test Year.
USPS-T-7
8
1
Complete detailed quarterly and annual volume forecasts with and without the rate
2
increase being requested by the Postal Service for 2005, 2006, and 2007 are presented
3
in Attachment A at the conclusion of my testimony.
USPS-T-7
9
T able 1
Summary o f M ail Vo lume, Go vernment F iscal Year 2002 - Go vernment F iscal Year 2006
(millions of pieces)
Before-Rates Volume Forecast
Avg. Annual
After-Rates Volume Forecast
Avg. Annual
Avg. Annual
2002GFY
2004GFY
Growth Rate
2006GFY
Growth Rate
Postage Own-Price
2006GFY
Growth Rate
(actual)
(actual)
(02 – 04)
(forecast)
(04 – 06)
Elasticity
(forecast)
(04 – 06)
F IR ST - C LA SS M A IL
F irst -C lass Letters & F lats
96,911.342
92,495.564
- 2.30%
91,324.157
- 0.64%
90,346.014
- 1.17%
-- Single-Piece
49,253.266
45,161.746
-4.24%
42,987.742
-2.44%
-0.175
42,459.296
-3.04%
-- Workshared 1
47,658.076
47,333.818
-0.34%
48,336.414
1.05%
-0.329
47,886.718
0.58%
First-Class Cards
5,467.290
5,430.832
-0.33%
5,544.356
1.04%
-0.376
5,463.895
0.30%
102,378.632
97,926.396
- 2.20%
96,868.513
- 0.54%
95,809.909
-1.09%
-2.95%
T OT A L F IR ST - C LA SS M A IL
Priority M ail
998.151
848.633
-7.79%
842.705
-0.35%
-1.004
799.324
Express M ail
61.280
54.123
-6.02%
52.945
-1.09%
-1.470
50.388
-3.51%
2.757
1.648
-22.68%
1.359
-9.18%
1.359
-9.18%
849.911
760.020
-5.44%
743.285
-1.11%
-0.235
753.578
-0.42%
2,052.033
1,913.176
-3.44%
1,896.987
-0.42%
-0.237
1,879.593
-0.88%
-0.193
M ailgrams
P ER IOD IC A LS M A IL
Within County
Nonprofit & Classroom
Regular Rate
T OT A L P ER IOD IC A L M A IL
6,787.814
6,462.075
-2.43%
6,438.348
-0.18%
9,689.758
9,135.272
- 2.90%
9,078.621
- 0.31%
73,224.143
6,416.651
-0.35%
9,049.822
-0.47%
ST A N D A R D M A IL
R egular R ate B ulk
81,121.684
5.25%
90,314.679
5.51%
88,665.738
4.55%
Regular1
43,552.691
50,776.236
7.97%
56,985.773
5.94%
-0.267
56,478.638
5.47%
Enhanced Carrier-Route
29,671.452
30,345.448
1.13%
33,328.906
4.80%
-1.093
32,187.100
2.99%
14,006.494
14,441.837
1.54%
15,502.729
3.61%
-0.319
15,418.326
3.33%
87,230.637
95,563.521
4.67%
105,817.408
5.23%
104,084.064
4.36%
Nonprofit Rate Bulk
1
T OT A L ST A N D A R D M A IL
USPS-T-7
10
Table 1 (continued)
Summary o f M ail Volume, Government Fiscal Year 2002 - Government Fiscal Year 2006
(millions of pieces)
Before-Rates Volume Forecast
Avg. Annual
After-Rates Volume Forecast
Avg. Annual
Avg. Annual
2002GFY
2004GFY
Growth Rate
2006GFY
Growth Rate
Postage Own-Price
2006GFY
Growth Rate
(actual)
(actual)
(02 – 04)
(forecast)
(04 – 06)
Elasticity
(forecast)
(04 – 06)
336.448
-5.36%
P ACKAGE SERVICES
P arcel P ost
372.591
375.618
0.41%
354.061
-2.91%
Non-Destination Entry
108.625
109.963
0.61%
116.209
2.80%
-0.382
114.911
2.23%
Destination Entry
263.966
265.655
0.32%
237.852
-5.38%
-1.351
221.536
-8.68%
Bound Printed Matter
507.702
553.666
4.43%
598.339
3.96%
-0.604
605.996
4.62%
Media & Library Rate Mail
194.793
202.645
2.00%
209.679
1.72%
-0.796
208.348
1.40%
1,075.087
1,131.928
2.61%
1,162.079
1.32%
1,150.792
0.83%
424.929
529.326
11.61%
666.538
12.21%
666.538
12.21%
56.821
71.082
11.85%
75.317
2.94%
75.317
2.94%
201,918.051 205,261.930
0.82%
214,565.484
2.24%
211,687.513
1.55%
TOTAL P ACKAGE SERVICES M AIL
Postal Penalty
Free-for-the-Blind
TOTAL DOM ESTIC M AIL
DOM ESTIC SP ECIAL SERVICES
Registry
6.277
5.009
-10.67%
3.990
-10.75%
-0.099
3.738
-13.60%
58.516
51.514
-6.17%
35.903
-16.52%
-0.230
35.366
-17.14%
283.468
273.701
-1.74%
282.145
1.53%
-0.183
278.811
0.93%
2.282
1.905
-8.63%
1.693
-5.74%
-0.592
1.673
-6.28%
Return Receipts
249.436
238.544
-2.21%
250.973
2.57%
-0.180
245.970
1.54%
Money Orders
216.867
187.211
-7.09%
181.567
-1.52%
-0.604
179.939
-1.96%
-0.466
Insurance
Certified
Collect-on-Delivery
Delivery & Signature Confirmation
TOTAL DOM ESTIC SP ECIAL SERVICES
Post Office Boxes
Stamped Cards
282.952
1,099.798
599.341
1,357.225
45.54%
724.011
9.91%
11.09%
1,480.281
4.44%
15.816
15.325
-1.56%
16.100
2.50%
234.432
96.807
-35.74%
90.352
-3.39%
-0.608
695.440
7.72%
1,440.938
3.04%
15.573
0.81%
89.429
-3.89%
1 I note here my understanding that the Before-Rates and After-Rates Test Year forecasts of First-Class Mail and Standard Mail shown in this Table are subsequently adjusted by witness Taufique.
Specifically, in both instances, the Test Year First-Class workshared letter volume is increased by 75.805 million pieces, and the Test Year Standard Regular volume is decreased by the same amount.
1
Please refer to the testimony of witness Taufique (USPS-T-28) for full details and an explanation of this adjustment.
USPS-T-7
11
1
In general, mail volume is projected to grow somewhat faster over the next two
2
years than it has over the past two years. The past few years have been quite hard on
3
mail volume, of course, as a combination of factors, including a recession, two rate
4
increases, the September 11th terrorist attacks followed by several cases of anthrax in
5
the mail, and increasing competition from electronic and other alternatives, have cut into
6
mail volumes over this time period. While some of these factors are expected to turn
7
around – the economy most notably – other factors, such as the Internet and increasing
8
competition, are continuing concerns.
9
Table 2 below shows the estimated impact of various factors on total domestic mail
10
volume over the past ten years as well as the projected impact of these factors over the
11
next three years.
USPS-T-7
12
Ta ble 2
Estim ate d Im pa ct of Factors Affecting Total Dom estic Mail Volume , 1994 – 2007
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.21%
1.13%
1.21%
1.18%
1.20%
1.38%
1.23%
1.30%
1.31%
1.22%
Macroeconomic
Factors
1.84%
0.61%
1.53%
1.65%
1.78%
1.77%
-0.06%
-1.11%
-0.16%
1.47%
Trends
-0.24%
-0.19%
-0.15%
-0.14%
-0.13%
-0.02%
0.03%
-0.03%
-0.49%
-0.54%
Internet
-1.11%
-1.31%
-1.35%
-1.21%
-1.23%
-1.47%
-1.43%
-1.35%
-1.33%
-1.36%
Competitor
Prices
0.65%
1.28%
0.70%
0.71%
0.62%
0.43%
0.61%
0.65%
0.48%
0.48%
Input
Prices
0.02%
-0.08%
-0.11%
0.14%
0.11%
0.03%
0.11%
0.00%
0.17%
0.21%
Postage
Prices
-1.95%
-2.31%
1.03%
0.37%
-0.85%
-0.75%
-2.04%
-1.90%
-2.17%
-0.50%
Inflation1
1.93%
1.65%
1.30%
1.12%
1.06%
1.41%
1.42%
1.07%
1.19%
1.03%
Other
0.11%
-0.03%
-0.55%
-0.18%
-0.04%
0.83%
-0.03%
-0.74%
0.78%
-0.06%
Total
2.43%
0.69%
3.62%
3.67%
2.51%
3.63%
-0.20%
-2.15%
-0.27%
1.93%
1994 - 2004
Total
Avg per Year
13.07%
1.24%
9.66%
0.93%
-1.88%
-0.19%
-12.40%
-1.32%
6.82%
0.66%
0.62%
0.06%
-10.59%
-1.11%
13.98%
1.32%
0.09%
0.01%
16.83%
1.57%
2001 - 2004
Total
Avg per Year
3.87%
1.27%
0.18%
0.06%
-1.06%
-0.35%
-3.99%
-1.35%
1.63%
0.54%
0.38%
0.13%
-4.52%
-1.53%
3.33%
1.10%
-0.03%
-0.01%
-0.53%
-0.18%
2005
2006
2007
1.24%
1.12%
1.11%
1.74%
0.54%
0.42%
-0.48%
-0.47%
-0.51%
-1.27%
-1.18%
-1.09%
0.18%
0.53%
0.48%
0.11%
-0.01%
0.03%
0.00%
0.00%
0.00%
1.08%
0.92%
0.92%
0.69%
-0.24%
-0.06%
3.30%
1.19%
1.28%
2004 - 2007
Total
Avg per Year
3.51%
1.16%
2.72%
0.90%
-1.45%
-0.49%
-3.50%
-1.18%
1.19%
0.40%
0.13%
0.04%
0.00%
0.00%
2.95%
0.97%
0.38%
0.13%
5.87%
1.92%
1/ In this and other tables that estimate the impact of prices on postal volumes, I use the implicit price def lator f or personal consumption expenditures as my measure of inf lation. From 1994 to 2004, inf lation measured
in this w ay averaged 1.9 percent per year. From 2004 to 2007, inf lation measured in this w ay is projected to increase by 1.9 percent per year. In contrast, using the consumer price index (CPI) instead,
1
inf lation w ould have averaged 2.4 percent per year f rom 1994 through 2004 and w ould be projected to average 1.8 percent per year f rom 2004 through 2007.
USPS-T-7
13
As can be seen in the last column of Table 2, the last four years have been
1
2
somewhat difficult ones for the Postal Service in terms of domestic mail volume growth.
3
There are two principal reasons for this difficulty as outlined in Table 2. First, the
4
economy slumped during this time. From 2000 – 2003, the net effect of the economy
5
on domestic mail volume was negative, as compared with 1.5 – 2 percent annual
6
growth from the economy through the late 1990s. The other factor which hurt mail
7
volumes from 2000 – 2003 was increases in Postal prices. Two omnibus rate cases,
8
R2000-1 and R2001-1, were enacted over this time period. In addition, the Internet had
9
a negative influence on mail volume, and is expected to operate negatively for the
10
foreseeable future.
11
However, the Postal Service has benefited (in terms of volume) from stable postage
12
rates for nearly three years now (since June, 2002).1 Global Insight projections indicate
13
that the economy will be somewhat stronger than in recent years (although not as
14
strong as it was in the late 1990s).
A major risk factor moving forward is the continuation and potential acceleration of
15
16
negative time trends which began to emerge in 2002 and 2003. The most notable of
17
these affect First-Class letters, although other mail categories may also be at risk. The
18
factors underlying these trends are analyzed in the discussion of the relevant mail
19
categories below.
20
D.
21
There have been no general methodological changes in the areas described in my
22
testimony since Docket No. R2001-1. The work presented by Dr. Tolley and me (and
1
Comparison with Previous Methodology
Price changes have been found to affect the demand for some mail categories with a lag of as many as
four quarters. Because of this, the R2001-1 rate change, which took place in June of 2002 (2002Q4) still
had some effect on the change in mail volumes from FY 2003 to FY 2004 for some categories of mail.
The mathematical explanation for this is given in Section III below.
USPS-T-7
14
1
Dr. Musgrave) in that case was directly adopted by the Commission. Therefore, in
2
terms of material methodological differences between the volume forecasting
3
presentation I am sponsoring in this case, and that relied upon by the Commission in
4
the last case, there are none.
5
As stated above, the general methodology employed here is to identify the factors
6
which affect mail volume and to quantify the impact of these factors on mail volume. As
7
part of this methodology, therefore, several new explanatory variables have been
8
introduced in one or more of the demand equations presented here, including, for
9
example, private employment, gross private domestic investment, and the number of
10
broadband subscribers. In addition, several minor changes have been made which
11
could be considered methodological in nature. These include the introduction of
12
additional seasonal variables because of the change in the timing of Postal Service
13
quarters as well as the projection of changes in the proportion of Standard Regular mail
14
that is letter-sized. Each of these changes is described in detail in the relevant sections
15
of my testimony below.
USPS-T-7
15
1
2
3
4
II. Analysis of Factors Affecting Mail Volume
A. General Overview
1. Division of Mail for Estimation Purposes
The demand for mail is not limited to a single demand based upon a single purpose.
5
Rather, mail demand is expected to differ across mailers, due, at least in part, to
6
differences in the purpose of the mail. Mail is a commodity in many economic markets,
7
in the sense that it satisfies a number of unique roles and purposes. For example, mail
8
can be used for personal correspondence, for bill-sending and bill-paying, for
9
advertising, for delivery of newspapers and magazines, and for delivery of other types of
10
11
goods.
Mail can be divided into five broad categories, based on the purpose of the mail:
12
(i) Correspondence and Transactions
13
(ii) Direct-mail Advertising
14
(iii) Expedited Delivery Services
15
(iii) Package Delivery Services
16
(v) Periodicals
17
Correspondence and Transactions mail is mail sent for the purpose of establishing
18
or maintaining a relationship. This mail may be sent between households (e.g., letters,
19
greeting cards), between households and non-households (e.g., orders, bills, bill
20
payments, financial statements), or between non-households (e.g., invoices, bill
21
payments). For the purposes of my testimony, Correspondence and Transactions are
22
equated to First-Class Mail. Not all First-Class Mail would properly be considered
23
Correspondence and Transactions based on this breakdown of mail. For example,
24
there is a significant amount of direct-mail advertising that is sent as First-Class Mail.
25
Data limitations effectively prevent us from separating out this portion of First-Class
USPS-T-7
16
1
Mail, however. Hence, this mail is combined with the rest of First-Class Mail. The
2
distinctions made within First-Class Mail and the final demand equations associated
3
with this type of mail are developed and presented in section B. below.
4
Direct-mail advertising is mail sent by businesses or other organizations for the
5
purpose of advertising goods or services. Over 90 percent of Standard Mail falls within
6
this category. As noted above, some portion of First-Class Mail is also direct-mail
7
advertising. It is difficult, if not impossible, however, to develop a useable time series of
8
First-Class advertising mail volume given available data sources. Hence, this category
9
of mail is included with the rest of First-Class Mail for modeling purposes. Standard
10
Mail volume is modeled in section C. below using a model of direct-mail advertising.
11
Expedited delivery services are mail that is sent with greater-than-usual urgency.
12
This could include Correspondence and Transactions, as described above, or Package
13
Delivery Services, as described below. The defining characteristic of this mail is the
14
speed of delivery. This refers to the Express Mail subclass here. This type of mail is
15
modeled and discussed in section D. below.
16
Package delivery services refer to the non-expedited delivery of goods which would
17
not fall into one of the other categories listed above (e.g., mail-order deliveries, books).
18
This corresponds roughly to the Priority Mail subclass as well as the Package Services
19
class of mail. These categories of mail are modeled and discussed in section E. below.
20
Periodicals are magazines, newspapers, journals, and newsletters sent on a periodic
21
basis through the mail. This corresponds to the Postal Service’s Periodicals class. As
22
with other types of mail, the correspondence between the Periodicals mail market and
23
the Periodicals mail class may not be exact. For purposes of estimating demand
24
equations, given the data available from the RPW (Revenue, Pieces, and Weight)
25
system, however, this distinction is useful and sufficient. The distinctions within
USPS-T-7
17
1
Periodicals Mail and the final demand equations associated with this type of mail are
2
developed and presented in section F. below.
3
Other categories of mail are discussed briefly in section G. below, including
4
Mailgrams, Postal Penalty mail, and Free for the Blind and Handicapped Mail. Special
5
Services are discussed in section H.
6
7
8
2. Sources of Information used in Modeling Demand Equations
a. General Demand Equation Methodology
Demand equations relate the demand for some good, in this case, mail volume, to
9
variables that are believed to influence demand. The general form of the demand
10
equations to be estimated expresses mail volume as a function of price, economic
11
conditions, and other variables which are believed to influence mail volume:
Vt = f(pt, Yt, etc.)
12
13
14
Conventionally, when economists discuss the impact of explanatory variables on the
15
demand for a particular good or service, the measure used to describe this impact is the
16
concept of “elasticity.” The elasticity of demand for a good, i, with respect to some
17
explanatory variable, x, is equal to the percentage change in the quantity demanded of
18
good i resulting from a one percent change in x. Mathematically, the elasticity of Vt with
19
respect to some variable, xt, is defined as follows:
20
21
22
23
where the t subscript denotes the time period for which the elasticity is being calculated.
24
The goal in modeling demand equations can thus be stated as identifying all relevant
25
etVx = [∂Vt / ∂xt]•[xt / Vt]
factors affecting demand and calculating elasticities with respect to these factors.
USPS-T-7
18
1
2
b. Postal Volume Data
The primary source of information on mail volumes is the Postal Service’s quarterly
3
RPW reports. These data serve as the dependent variables in the demand equations
4
developed and described in my testimony and as the base volumes from which volume
5
forecasts are made.
6
Through Postal Fiscal Year 2003, the Postal Service divided its Fiscal Year into
7
thirteen four-week accounting periods. This resulted in years which were only 364 days
8
long and which, consequently, moved over time relative to the Gregorian calendar.
9
Starting in 2004, the Postal Service switched to a more traditional, twelve-month
10
calendar. Postal Fiscal Years now run from October 1 (of the preceding calendar year)
11
through September 30. Postal Service volume data have been restated to correspond
12
to these (Gregorian) quarters dating back to Government Fiscal Year (GFY) 2000. Data
13
prior to GFY 2000 are still dated by Postal quarters, resulting in 364-day years. The
14
econometric methodologies used to deal with the impact of the Postal calendar are
15
detailed below.
c. Factors Affecting Demand
16
17
18
19
20
The starting point for traditional micro-economic theory is a demand equation that
21
relates quantity demanded to price. Quantity demanded is inversely related to price.
22
That is, if the price of a good were increased, the volume consumed of that good would
23
be expected to decline, all other things being equal.
24
25
i. Price
This fundamental relationship of price to quantity is modeled in the Postal Service’s
demand equations by including the price of postage in each of the demand equations
USPS-T-7
19
1
estimated by the Postal Service (with the exception of the demand equations associated
2
with Mailgrams, Postal penalty mail, and Free for the Blind and Handicapped Mail).
3
The Postal prices entered into these demand equations are calculated as weighted
4
averages of the various rates within each particular category of mail. For example, the
5
price of First-Class single-piece letters is a weighted average of the single-piece letters
6
rate (37 cents), the additional ounce rate (23 cents), and the nonstandard surcharge (11
7
cents). Product-by-product billing determinants provide the components of the market
8
baskets which are used as weights in developing these price measures.
9
Experience indicates that mailers may not react immediately to changes in Postal
10
rates. For some types of mail it may take up to a year for the full effect of changes in
11
Postal rates to influence mail volumes. To account for the possibility of a lagged
12
reaction to changes in Postal prices on the demand for certain types of mail, the Postal
13
price may be entered into the demand equations lagged by up to four quarters.
14
The price of postage is not the only price paid by most mailers to send a good or
15
service through the mail. For those cases where the non-Postal price of mail is
16
significant and for which a reliable time series of non-Postal prices is available, these
17
prices are also included explicitly in the demand equations used to explain mail volume.
18
For example, the price of paper is included as an explanatory variable in the demand
19
equations for Periodicals Mail, since paper is an important input in the production of
20
newspapers and magazines.
21
Prices of competing goods are also included in many of the Postal Service’s demand
22
equations. For example, the average price of UPS and FedEx Ground service is a
23
variable in the Priority Mail and Parcel Post equations, while Federal Express average
24
price is used in the Express Mail forecast.
USPS-T-7
20
1
Finally, several demand equations include cross-price measures with other Postal
2
products, such as single-piece and First-Class workshared letters and Bound Printed
3
Matter and Media Mail. In most cases, cross-price variables enter the equation in the
4
same way as the own-price variables, i.e., as a measure of the average price of the
5
product. In a few cases, however, cross-price variables may be measured in relative
6
terms. For example, the single-piece and First-Class workshared letters equations each
7
include the average worksharing discount, which measures the average difference
8
between the prices of single-piece and workshared mail. Similarly, the First-Class
9
workshared letters and Standard Regular demand equations include the average
10
difference between the price of a First-Class workshared letter and a similarly prepared
11
letter sent as Standard Regular mail.
12
Another non-price measure of price substitution is used in the case of First-Class
13
cards and Standard Regular mail, where the variable used in the Postal Service’s
14
demand equations is a measure of the percentage of Standard Regular letter mail for
15
which First-Class cards rates are less expensive. Cases like these are used when a
16
simple price comparison does not adequately reflect the decisions which mailers face in
17
choosing from among alternate mail categories.
18
19
All prices are expressed in real 2000 dollars. The implicit price deflator for personal
consumption expenditures is used to deflate the prices.
20
In general, when the Postal Service refers to price elasticities, the reference is to
21
long-run price elasticities. The long-run price elasticity of mail category i with respect to
22
the price of mail category j is equal to the sum of the coefficients on the current and
23
lagged price of mail category j. The long-run price elasticity therefore reflects the
24
cumulative impact of price on mail volume after allowing time for all of the lag effects to
25
be felt.
USPS-T-7
21
1
2
ii. Macroeconomic Variables
With the exception of price, the most basic economic factor affecting consumption at
3
a theoretical level is income. As incomes rise, consumers are able to consume more.
4
This is generally true of Postal Services. That is, mail volumes tend to rise during
5
periods of strong economic growth and stagnate or decline during recessions. To
6
model this relationship, the demand equations presented here include several
7
macroeconomic variables which relate mail volumes to general economic conditions.
8
Four macroeconomic variables are used in the demand equations presented here:
9
retail sales, mail-order retail sales, total employment, and investment. These data are
10
compiled by the United States government and are obtained by the Postal Service from
11
Global Insight.
12
The specific variable choices are made on an equation-by-equation basis. The
13
decision process in choosing macroeconomic variables includes an effort to develop
14
equations which are both theoretically correct as well as empirically robust. These
15
decisions are described in more detail in the specific discussions associated with the
16
particular demand equations in Section II above.
17
18
19
The macroeconomic variables used in this case are discussed briefly below.
(a) Retail Sales
Total retail sales are included as an explanatory variable in the demand equations
20
associated with First-Class workshared letters, Priority Mail, Standard Regular,
21
Standard Enhanced Carrier Route, and Standard bulk nonprofit mail.
22
In the case of Standard Mail (and, to some extent, First-Class workshared letters),
23
retail sales were chosen based on a theory of direct-mail advertising which posits that
24
the level of advertising will be chosen as a function of expected sales. Hence, there is
25
presumed to be a close relationship between direct-mail advertising volume and the
USPS-T-7
22
1
level of retail sales. This presumption has been borne out by the empirical results
2
presented in Section II below.
3
The volume of Priority Mail consists largely of the delivery of products bought by
4
either the sender or the recipient of the mail. Hence, much of this volume is derived
5
from retail sales.
6
7
(b) Mail-Order Retail Sales
As with Priority Mail, Package Services mail volumes consist largely of the delivery
8
of products bought by the sender or recipient of the mail so that this type of mail volume
9
also derives almost directly from retail sales. More specifically, package delivery
10
services are a function of mail-order retail sales, that is, sales of goods which are
11
delivered to the consumer. Hence, mail-order retail sales (which include sales identified
12
as “electronic shopping”) are included directly in the demand equations for Bound
13
Printed Matter and Media and Library Rate Mail to reflect this direct relationship
14
between mail-order retail sales and these mail volumes.
15
Total retail sales, as opposed to mail-order retail sales, were used in the Priority Mail
16
equation, mostly for empirical reasons. The relationship of package delivery services
17
and the economy is discussed in more detail in section D. below.
18
19
(c) Employment
Total private employment is included in several of the demand equations used in this
20
case, including First-Class single-piece letters, Express Mail, Periodicals mail, money
21
orders, and Post Office Boxes. Employment is an excellent measure of the overall level
22
of business activity in the economy. In many cases, mail volume is not affected by the
23
dollar value of economic transactions, so much as by the number of such transactions.
24
For example, the number of credit card bills one receives does not necessarily go up as
25
the total amount charged per card goes up. While variables like retail sales may be
USPS-T-7
23
1
good measures of the total dollar amount of economic activity (e.g., the total amount
2
charged per credit card), employment appears to be a better measure of the number of
3
business transactions (e.g., number of credit card bills received).
4
Ultimately, the choice of which macroeconomic variables to use in the demand
5
equations discussed here was largely an empirical decision. In those cases where
6
employment is used as an economic variable in the Postal demand equations, its
7
inclusion clearly improved the econometric fit for these equations.
8
9
(d) Total Real Investment
Advertising can be viewed as a type of business investment. As such, direct-mail
10
advertising volume could be affected by the same factors which drive investment in
11
general. To reflect this relationship, real gross private domestic investment is included
12
as an explanatory variable in the demand equations for Standard Regular and Standard
13
Enhanced Carrier Route (ECR) mail used in this case.
14
iii. The Internet and Electronic Diversion
15
16
17
18
(a) History of Electronic Diversion in Postal Service Demand
Equations
Perhaps the most important factor affecting the Postal Service’s mail volume
19
forecasts over the past two to three years is the threat, both realized and potential, of
20
electronic diversion of mail. E-mail has emerged as a potent substitute for personal
21
letters and business correspondence; bills can be paid online; and now, some
22
consumers are even beginning to receive bills and statements through the Internet
23
rather than through the mail. Many magazines and newspapers now have an on-line
24
edition in addition to their print editions. Expenditures on Internet advertising have
25
doubled over the past five years, while online shopping has experienced even greater
USPS-T-7
24
1
growth. Understanding the emergence of the Internet and its role vis-à-vis the mail is
2
critical in understanding mail volume, both today and in the future.
3
When a new factor emerges which affects the demand for a product it is usually
4
difficult initially to find a variable that can adequately explain this effect and is amenable
5
to inclusion in an econometric equation. This was certainly the case with the impact of
6
electronic diversion and the Internet on mail volume. As recently as the R2000−1 rate
7
case, for example, the Internet was not explicitly included as an explanatory variable in
8
any of the mail demand equations used for forecasting. Even at that time, however, it
9
was understood that some electronic diversion was occurring. In my R2000−1
10
testimony, for example, I estimated that the time trends in the First-Class letters
11
equations implied that electronic diversion had reduced First-Class letters volume by 6 –
12
7 billion pieces from 1988 through 1999.
13
As the Internet has grown, both in size and in importance, more and more variables
14
which measure various aspects of Internet usage have become available. One of the
15
first variables to be made available on a regular basis was consumption expenditures
16
on Internet Service Providers (ISPs). These data are compiled by the Bureau of
17
Economic Analysis (BEA) and are a component of the National Income and Product
18
Accounts (NIPA). Although these data are unpublished (by the BEA), they are reported
19
by Global Insight, from whom we obtain them.
20
Consumption expenditures on Internet Service Providers have many desirable
21
properties which make it well suited for use in an econometric equation. Foremost
22
among these properties is the fact that a consistent measure of this variable exists by
23
month from January, 1988 to the present. Since Internet usage was minimal in 1988,
24
we have almost a complete history of Internet expenditures.
USPS-T-7
25
1
As reported by Global Insight, consumption expenditures on Internet Service
2
Providers grew from $12 million (annualized) in January, 1988 to $8 billion by the end of
3
1999. This series has continued to grow, reaching an annualized level of $17.7 billion
4
by November, 2004. In R2001-1, this variable was introduced in my testimony as an
5
explanatory variable in several of the Postal Service’s demand equations, including
6
First-Class single-piece letters and First-Class cards. At that time, ISP consumption
7
was found to explain a decline of approximately 4 billion First-Class letters from 1988
8
through 1999, and an additional loss of 3.5 billion First-Class single-piece letters from
9
1999 through 2001. It was not claimed, either then or now, that this variable fully
10
explained all electronic diversion. In fact, the First-Class single-piece letters equation
11
presented in R2001-1 also included a negative time trend, which explained an additional
12
loss of more than 10 billion single-piece letters from 1988 through 1999, much of which
13
might also have been lost to the Internet or other electronic substitutes for the mail.
14
15
16
17
18
(b) Relationship of Mail Volume to the Internet
There are two general dimensions to the Internet as a possible substitute for mail
volume: the breadth of Internet usage and the depth of Internet usage.
The breadth of Internet usage simply refers to the number of people online. As more
19
people use the Internet, there are simply more people for whom the Internet is available
20
as a substitute for the mail.
21
The depth of Internet usage refers to the number of things which an individual does
22
on the Internet. As the depth of Internet usage increases for a particular person, the
23
number of activities for which the Internet may substitute for mail may increase, thereby
24
increasing the overall level of substitution of the Internet for mail volume, even in the
25
absence of an increase in the number of Internet users.
USPS-T-7
26
1
Historically, increases in both the breadth and depth of Internet usage have
2
adversely affected mail volumes. The breadth of Internet usage, as measured by
3
consumption expenditures on Internet Service Providers, for example, has increased by
4
nearly 150,000 percent since 1988, and has doubled since early 2000. At the same
5
time, the depth of Internet usage has also increased dramatically. For example, the
6
average time spent online by a typical Internet user has also doubled over a similar time
7
period. That is, there are now twice as many users and each user is now spending
8
twice as much time online now as compared to early 2000. Another example of the
9
increasing depth of Internet usage that directly affects mail volume is that, since 2000,
10
the percentage of people who pay at least some of their bills online has nearly
11
quadrupled.
12
The breadth and depth of Internet usage have both been important in understanding
13
the impact of the Internet on mail volumes historically. However, moving forward, the
14
depth of Internet usage is a much more important consideration. The reason for this is
15
that the breadth of Internet usage has a natural ceiling. Eventually, everybody who
16
would ever obtain Internet access will actually have Internet access. At that point, the
17
only source of increasing electronic diversion of the mail will be an increasing depth of
18
Internet usage. Hence, in searching for possible Internet variables to include in the
19
econometric equations presented here, it is important that they not only measure the
20
breadth of Internet usage in the United States, but that they reflect this potential for
21
increasing depth.
22
23
24
25
26
(c) Internet Variables Considered Here
(i) Consumption Expenditures on Internet Service Providers
As noted above, Global Insight provides a measure of consumption expenditures on
Internet Service Providers (ISP) that is compiled by the Bureau of Economic Analysis.
USPS-T-7
27
1
In addition to ISP consumption, Global Insight reports a price index associated with
2
these expenditures. This price index fell dramatically in the early days of the Internet,
3
from a level of 157.1 in January, 1988 to a low value of 89.4 in January, 1995, a decline
4
of 43 percent, as Internet usage became much more affordable over this time period.
5
More recently, the price index for ISP consumption has exhibited no discernible trend
6
over the past several years.
7
Dividing total ISP consumption expenditures by the ISP price index yields a measure
8
of the volume of ISP usage by consumers. The units on this variable will be the quantity
9
of Internet usage in the year 2000 (the year for which the price index was set equal to
10
100). Based on other estimates of Internet usage, one can restate such a variable in
11
terms of the percentage of the U.S. population which has access to an Internet Service
12
Provider at any given time.
13
Table 3 below summarizes the level of Internet usage implied by this variable. The
14
ISP consumption expenditures data in Table 3 are divided by the price index shown in
15
the second column to create a measure of Internet volume. From other sources, the
16
level of Internet penetration is estimated to be approximately 60 percent for 2002. The
17
implied Internet penetration for year Y is then estimated by multiplying the ratio of ISP
18
volume per adult in year Y to ISP volume per adult in 2002 times 60 percent.
19
For example, ISP volume in 2004 was approximately equal to 16.78 billion. Total
20
adult population for 2004 was approximately 204.28 million people. Hence, total ISP
21
volume per adult was approximately equal to 82.16 in 2004. ISP volume per adult in
22
2002 was estimated from these data to be equal to 62.96. The 2004 value then
23
represents a 30.5 percent increase from 2002. Hence, the implied Internet penetration
24
in 2004 is approximately 30.5 percent greater than the Internet penetration in 2002, or
25
78.3 percent (60%•1.305).
USPS-T-7
28
Ta ble 3
Estima te d Pe rce nta ge of Ame ricans w ith Inte rne t Acce ss
1
2
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
ISP Consumption
Expenditures
Price Index
(billions of dollars)
(divided by 100)
$0.02
1.591
$0.03
1.515
$0.07
1.439
$0.15
1.331
$0.25
1.226
$0.37
1.123
$0.82
0.988
$1.73
0.931
$2.79
0.956
$3.60
0.995
$4.42
1.063
$6.77
1.038
$9.48
1.001
$11.93
1.018
$12.88
1.027
$14.90
1.036
$16.96
1.011
Implied Internet
Penetration
0.06%
0.12%
0.27%
0.62%
1.10%
1.76%
4.39%
9.68%
15.03%
18.38%
20.92%
32.41%
46.50%
56.79%
60.00%
67.91%
78.29%
Adjusting the ISP consumption expenditure data in this way, one can infer that the
3
percentage of the population with Internet access was virtually zero in 1988. From
4
there, it reached 10 percent in 1995, passed 50 percent by 2001, and has continued to
5
grow, to approximately 78 percent by 2004.
6
Going forward, Table 3 highlights one drawback to using either ISP consumption or
7
ISP volume as a measure of the Internet’s impact on mail. If the price index associated
8
with ISPs remains relatively constant, there is only so much more growth possible in
9
ISP consumption expenditures. While the numbers in the final column of Table 3 are
10
only approximations, nevertheless the level of Internet penetration obviously cannot
11
exceed 100 percent of the population, and, in fact, is unlikely to ever reach that level.
12
That is, consumption expenditures on Internet Service Providers are primarily a
13
measure of the breadth of Internet usage but provide little or no information on the
USPS-T-7
29
1
depth of such usage. Yet, as discussed above, it is the depth of Internet usage which
2
would be expected to have the much greater impact on mail volumes in the future.
3
(ii) Internet Experience
4
As consumers gain experience online, the number of activities for which people are
5
comfortable using the Internet increases. One simple measure of the depth of Internet
6
usage, then, is the total length of time users have been on the Internet. As explained
7
above, the volume of ISP consumption represents a measure of the number of people
8
online at a particular time. Hence, one can use the ISP volume series to calculate the
9
total experience of the online population at any point in time.
10
This may best be shown by an example. Suppose that in year 1 there are 100
11
people online, in year 2, there are 200, and in year 3, there are 250 people online. If we
12
assume that all of the people online in year 1 were still online in years 2 and 3, then, in
13
year 2 there are 100 people with two years’ experience (Year 1 and Year 2) and 100
14
people (200 – 100) with one year of experience (the new users in Year 2). So, total
15
experience in Year 2 is equal to 100*2 + 100*1 = 300. In Year 3, there are now 100
16
people with 3 years’ experience, 100 people with 2 years’ experience, and 50 people
17
with 1 year of experience. Total experience in Year 3 is equal to 100*3 + 100*2 + 50*1
18
= 550. Generalizing, total experience in year X is equal to total experience in year X–1
19
plus the total number of users in year X, or, equivalently, total experience in year X is
20
equal to the sum of the online variable from year 1 to year X.
21
The same thing can be done here with ISP Volume. That is, we construct a variable,
22
which I call Internet Experience, which is equal to the sum of ISP Volume since 1988.
23
This variable measures the total experience of the online population. Internet
24
Experience is used as an explanatory variable in several of the demand equations
25
presented here, most notably in the First-Class single-piece letters equation.
USPS-T-7
30
1
(iii)
Number of Broadband Subscribers
2
Several measures of Internet activity exist in at least a somewhat usable form in
3
addition to ISP consumption. Several of these variables were investigated as possible
4
explanatory variables in mail-volume demand equations.
5
The one such variable that is actually used in some of the demand equations here –
6
specifically First-Class workshared letters and Periodicals mail – is the number of
7
broadband subscribers. For certain types of Internet activities, the limiting factor which
8
may delay their adoption is not Internet experience so much as a technological
9
limitation. That is, certain things may not have been possible to do via the Internet in
10
1995 or may have been too time-consuming to do with regular Internet connections.
11
Over time, as Internet connection speeds increase, the Internet becomes a more
12
feasible substitute for more things.
13
The number of broadband subscribers is reported quarterly by Leichtman Research
14
Group (LRG). Data from LRG exist quarterly dating back to 2001, with annual year-end
15
data available for 1998, 1999, and 2000. A consistent quarterly time series was
16
constructed from these data which assumed that the number of broadband subscribers
17
was equal to zero prior to 1997, growing fairly smoothly from 1997 through 2000 based
18
upon the year-end data provided by LRG for 1998, 1999, and 2000. From 2001
19
onward, the LRG data are used directly.
20
I am not asserting here that the use of broadband Internet access leads directly to a
21
proportional decrease in mail volume. Rather, I am suggesting that the historical
22
pattern of the adoption of broadband Internet access has mirrored electronic
23
substitution out of certain types of mail. In some cases, mail loss may be a direct result
24
of the use of broadband. For example, higher-speed connections, which allow for faster
25
downloads of graphical images, may make online magazines a more attractive
USPS-T-7
31
1
alternative to Periodicals mail. In other cases, however, it may simply be the case that
2
the adoption of these technologies is occurring along a similar time path. This similarity
3
may be more than coincidental, of course, and may be the result of common
4
technological advancements. Recent increases in electronic bill presentment may have
5
aspects of both of these factors. That is, while higher-speed connections may make it
6
more feasible to receive bills and statements online, it is also the case that the
7
technology which allows for such things has also developed more or less over this same
8
time period. This distinction is important to understand when developing forecasts of
9
these variables for use in making volume forecasts. While, to a certain extent, the
10
question is, “How many households do we expect to have broadband Internet access in
11
2006?” the more relevant question to us is, “How much mail is expected to be diverted
12
due to electronic substitution by 2006?” This issue is developed more fully in Section IV
13
below.
14
15
(iv)
Average Time Spent Online
One measure that, in theory at least, should be an excellent measure of the depth of
16
Internet usage is average time spent online within a certain time period. Nielsen
17
NetRatings tracks such a series monthly. This series was investigated, but ultimately
18
two problems rendered it largely unsuitable for my purposes. First, the data only go
19
back to the late 1990s. But, of course, the Internet was being used prior to then.
20
Hence, it would have been necessary to impute some earlier historical data. The other,
21
ultimately more serious, problem, however, is that the amount of time spent online tends
22
to be heavily influenced by news events. For example, time online spiked upward
23
following the September 11th terrorist attacks. Likewise, time online spiked upward in
24
the weeks just before and during the initial invasion of Iraq. Because these increases
25
have little or nothing to do with the mail, they reduce the explanatory power of this
USPS-T-7
32
1
variable in helping to understand the impact of the Internet on mail volume. That is,
2
while long-run changes in time spent online probably reflect changes in the depth of
3
Internet usage, short-run changes in time spent online are less reflective of this.
4
5
(v) Measures of Direct Substitution
Finally, several measures of direct substitution of mail were investigated. For
6
example, the Household Diary Study provides estimates of the number and percentage
7
of bills paid by electronic means (and, more specifically, by computer). The National
8
Automated Clearing House Association (NACHA) also provides historical data on
9
electronic data and fund transfers between businesses and between consumers and
10
businesses. These series tended to be less robust within the econometric demand
11
equations. I think that this is because electronic diversion of the mail is a very
12
generalized risk. That is, almost any transaction that is conducted by mail could instead
13
be conducted via some electronic alternative. By focusing exclusively on a specific
14
aspect of this diversion, such as electronic bill-paying, however, these variables tend to
15
make poor proxies for other types of electronic diversion. On the other hand, while such
16
variables were not found to work particularly well as explanatory variables in
17
econometric equations, variables of this sort are extremely valuable as a way to check
18
the reasonableness of the econometric estimates and to gain additional non-
19
econometric insights which can be valuable in making volume forecasts.
USPS-T-7
33
(vi)
1
2
Internet Advertising Expenditures
Standard Mail faces direct competition from the Internet for limited advertising
3
dollars. Hence, Internet advertising expenditures can be seen as a direct competitor for
4
direct-mail advertising. The Interactive Advertising Bureau reports total Internet
5
advertising expenditures on a quarterly basis, as compiled by PricewaterhouseCoopers.
6
Unfortunately, this measure of Internet advertising expenditures has some
7
drawbacks, which limit its effectiveness as a true measure of potential substitution
8
between the Internet and Standard Mail. For example, this measure of Internet
9
advertising does not include e-mail advertising by non-Internet companies, which would
10
11
seem to represent the closest Internet analog to direct-mail advertising.
Nevertheless, this variable is included in the Standard equations to the extent that it
12
works. In this case, this means that Internet advertising expenditures are included in
13
the Standard Enhanced Carrier Route (ECR) equation. Internet advertising
14
expenditures are entered into the Standard ECR equation as a percentage of total
15
advertising expenditures. This is done to ensure that the focus is on changes to the use
16
of the Internet as an advertising medium, as opposed to more general changes to the
17
overall level of advertising, which are measured through other variables in the Standard
18
ECR equation.
19
The specific measures of Internet usage and electronic diversion used in specific
20
demand equations are explained in the descriptions of specific demand equations
21
below.
22
iv. Time Trends
23
It is always desirable to be able to explain the behavior of a variable that is being
24
estimated econometrically as a function of other observable variables. Occasionally,
25
however, the behavior of a variable is due to factors that do not easily lend themselves
USPS-T-7
34
1
to capture within a time series variable suitable for inclusion in an econometric
2
experiment. It is not uncommon for such phenomena to be modeled in part through the
3
use of trend variables.
4
Given that trend variables are needed within particular demand equations, an
5
equally important question becomes what forms these trend variables ought to take.
6
7
8
9
10
11
12
13
14
A trend is a trend is a trend
But the question is, will it end?
Will it alter its course
Through some unforeseen force,
And come to a premature end?
Sir Alec Cairncross
It is not sufficient to merely plug linear time trends into all of one’s econometric
15
equations and project these trends to continue unabated throughout the forecast period.
16
Rather, it is important to evaluate each equation individually and determine the
17
appropriate trend specification for each equation, if any.
18
The Periodicals equations, the Standard Regular and ECR equations, and most of
19
the special service equations include linear time trends to account for long-run trends in
20
the volumes of these types of mail, for which either economic sources have not been
21
found or which are most readily modeled by a simple trend variable.
22
The Priority and Express Mail equations include linear time trends over their full
23
sample, which have modest positive coefficients, as well as linear time trends over only
24
the last three years, which have much stronger negative coefficients.
25
Several other mail categories, including single-piece and First-Class workshared
26
letters, destination-entry Parcel Post, and money orders also include negative time
27
trends over shorter time periods, typically the last two to four years, to reflect recent
28
changes in the markets for these mail categories.
USPS-T-7
35
1
2
All of these trends are discussed in detail in the discussions of the specific demand
equations used in this case.
v. Other Variables
3
4
Other variables are included in the demand equations as events warrant. Most of
5
these variables take the form of simple dummy variables. For example, certain
6
equations include dummy variables for some rate or classification changes that are
7
inadequately modeled by the price indices used here.
8
9
10
11
One example is classification reform, MC95-1, which is modeled by a dummy
variable in the First-Class and Standard equations because certain rule changes are not
adequately modeled by the simple fixed-weight price indices.
A second example of this type of variable is a dummy variable for the general UPS
12
strike which occurred in August, 1997. This strike dummy is included in the demand
13
equations for Priority Mail, Express Mail, Parcel Post, insurance, and COD.
14
A final example is a dummy variable for the quarter immediately following
15
September 11, 2001, and the subsequent anthrax attacks on the mail. These events
16
served to temporarily lower mail volume in several mail categories, including Express
17
Mail, Standard Regular, Standard bulk nonprofit, Parcel Post, and several special
18
services.
USPS-T-7
36
1
vi. Seasonality
2
The volume data used in modeling the demand for mail are quarterly. Until recently,
3
the Postal Service reported data using a 52-week Postal calendar composed of 13, 28-
4
day accounting periods. Because the 52-week Postal year was only 364 days long, the
5
beginning of the Postal year, as well as the beginning of each Postal quarter, shifted
6
over time relative to the traditional Gregorian calendar. Specifically, the Postal calendar
7
lost five days every four years relative to the Gregorian calendar.
8
For example, prior to 1983, Christmas Day fell in the first quarter of the Postal year
9
(which began in the previous Fall). After 1983, however, Christmas Day fell within the
10
second Postal quarter. Between Postal Fiscal Year 1983 (PFY 1983) and PFY 1999
11
(the last year for which Postal quarterly data are used here), the second Postal quarter
12
gained the 20 days immediately preceding Christmas (December 5 through December
13
24) which are among the Postal Service’s heaviest days in terms of mail volume. Not
14
surprisingly, therefore, the relative volumes of mail in Postal Quarter 1 and Postal
15
Quarter 2 changed over this time period for most mail categories, as Christmas-related
16
mailings shifted from the first Postal quarter to the second Postal quarter, solely
17
because of the effect of the Postal Service’s moving calendar.
18
This shift created a difficulty in modeling the seasonal pattern of mail volume using
19
traditional econometric techniques, such as simple quarterly dummy variables. If the
20
seasonal pattern of mail volume was due to seasonal variations within the Gregorian
21
calendar (e.g., Christmas), then the perceived seasonal pattern across Postal quarters
22
may not have been constant over time, even if the true seasonal pattern across periods
23
of the Gregorian calendar was constant over time.
24
25
Consequently, the seasonal variables used in the econometric demand equations
developed in my testimony were defined to correspond to constant time periods in the
USPS-T-7
37
1
Gregorian calendar. Defining seasons in this way turned the moving Postal calendar
2
into an advantage, because it allowed one to isolate more than just four seasons, even
3
with simple quarterly data.
4
For any given quarter, the value of each seasonal variable was set equal to the
5
proportion of delivery days within the quarter that fell within the season of interest. An
6
example of the construction of one of these variables may be instructive. Consider, for
7
example, the values of one of Postal Service’s seasonal variables, November 1 -
8
December 10, for the four quarters of 1999.
9
Postal 1999Q1 spanned the time period from September 12, 1998, through
10
December 4, 1998, and included a total of 69 Postal delivery days (12 weeks @ 6
11
delivery days per week less three holidays: Columbus Day, Veterans’ Day, and
12
Thanksgiving). The period from November 1, 1998, through December 4, 1998, fell
13
within the season of November 1 - December 10 as well as 1999PQ1. This time period
14
encompassed a total of 27 delivery days (34 total days less 2 Holidays and 5 Sundays).
15
Hence, the seasonal variable November 1 - December 10 has a value equal to (27/69)
16
in 1999PQ1.
17
Postal 1999Q2 spanned the time period from December 5, 1998, through February
18
26, 1999, and contained 68 delivery days (12 weeks @ 6 delivery days per week minus
19
Christmas, New Year’s Day, Martin Luther King’s Birthday, and Presidents’ Day). The
20
period from December 5, 1998, through December 10, 1998, fell within the season of
21
November 1 - December 10 as well as 1999PQ2. This time period encompasses a total
22
of 5 delivery days (6 days less 1 Sunday). Hence, the seasonal variable November 1 -
23
December 10 has a value equal to (5/68) in 1999PQ2.
USPS-T-7
38
1
None of the seasonal variable, November 1 - December 10, falls within either the
2
third or fourth Postal quarters. Hence, this variable has a value of zero in both
3
1999PQ3 and 1999PQ4.
4
In theory, variables constructed in this way could be used for quarters which are
5
defined in any way, including true Gregorian quarters. In fact, however, for reasons that
6
are not entirely clear, but may include difficulties with the way in which daily volumes
7
are estimated and combined into quarterly data by the Postal Service, these variables
8
do not, in general, do as well in explaining the seasonal pattern of mail volume by
9
Gregorian quarter since 2000. To supplement these variables, therefore, simple
10
quarterly dummy variables for the four Gregorian quarters are also included in the
11
Postal Service’s demand equations. These variables are equal to zero for all Postal
12
quarters and for three of the four Gregorian quarters, and equal to one for Gregorian
13
quarters during the quarter of interest.
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
A total of 22 seasonal variables are used in estimating the demand equations in this
case. These seasons correspond to the following periods of the Gregorian calendar:
September 1 - 15
September 16 - 30
October
November 1 - December 10
December 11 – 12
December 13 – 15
December 16 – 17
December 18 – 19
December 20 – 21
December 22 – 23
December 24
December 25 – 31
January 2 - February 28/29
March 1 - March 31
April 1 - 15
April 16 - May 31
June
USPS-T-7
39
1
2
3
4
5
6
7
July 1 - August 31
Dummy for Gregorian Quarter 1 (October – December, since GFY 2000)
Dummy for Gregorian Quarter 2 (January – March, since GFY 2000)
Dummy for Gregorian Quarter 3 (April – June, since GFY 2000)
Dummy for Gregorian Quarter 4 (July – September, since GFY 2000)
These 22 seasonal variables are used to model the seasonal pattern of mail
8
volumes econometrically. Twenty-one of the 22 seasonal variables are included in each
9
econometric equation. The excluded seasonal variable is the variable covering the
10
period from July 1 through August 31, the effect of which is captured implicitly within the
11
constant term. The coefficients on the 21 included seasonal variables are estimated
12
along with the other econometric parameters as described below.
13
By making the months of July and August the excluded variable, the dummy
14
variables for all four Gregorian quarters are thereby included in the demand equations.
15
This results in there being an implicit dummy variable included in the Postal Service’s
16
demand equations equal to one concurrent with the change from Postal to Gregorian
17
quarters (between 1999 and 2000) reflected in the sum of the four Gregorian quarterly
18
dummies. To eliminate this potential difficulty, therefore, the sum of the coefficients on
19
the four Gregorian dummy variables is constrained to be equal to zero.
20
In an effort to maximize the explanatory power of the seasonal variables, taking into
21
account the cost of including these variables, in terms of degrees of freedom, the
22
coefficients on adjoining seasons that are similar in sign and magnitude are constrained
23
to be equal. These constraints across seasons are done on an equation-by-equation
24
basis. The criterion used for this constraining process is generally to minimize the
25
mean-squared error of the equation (which is equal to the sum of squared residuals
26
divided by degrees of freedom). As part of these constraints, the seasonal variables
27
spanning the period from December 11 – 17 were generally constrained to have the
USPS-T-7
40
1
same coefficient, as were the seasonal variables spanning the period from December
2
18 – 24, although these two restrictions were not necessarily absolute. For equations
3
estimated over sufficiently short sample periods (starting in 1990 or later), some of the
4
adjoining seasonal variables had to be constrained to avoid problems of perfect
5
multicollinearity.
6
Restrictions across seasonal coefficients were not applied to the four Gregorian
7
quarterly dummies, aside from the restriction that their coefficients sum to zero as noted
8
above.
9
The estimated effects of the 21 seasonal variables are combined into a seasonal
10
index, which can be arrayed by Postal quarter to observe the quarterly seasonal pattern
11
and to understand how this seasonal pattern changed over time prior to 2000 as a result
12
of the moving Postal calendar. Since 2000, of course, this seasonal index is generally
13
constant for a given quarter each year, although changes in the number of Sundays
14
within a given quarter and the existence of Leap Years lead to some modest year-to-
15
year changes. This seasonal index forms the basis for the seasonal multipliers used in
16
making volume forecasts.
17
18
3. Forecasting Philosophy
The forecasting philosophy espoused here, which has been advocated by Dr. Tolley
19
and the Postal Service in earlier rate cases, is that the past is the best predictor of the
20
future. The goal of this testimony is to identify the factors which have caused mail
21
volume to change historically and to quantify the exact relationship between these
22
factors and mail volume as best as possible.
USPS-T-7
41
1
Once this is done, the first step in making volume forecasts is to project these
2
explanatory variables forward. For most of the macroeconomic variables used here,
3
this is done by relying upon the forecasts of Global Insight. For certain other variables,
4
such as Postal prices, time trends, and seasonal variables, projections are relatively
5
straightforward. In many other cases, however, these forecasts are made here in my
6
testimony. These forecasts are described in detail in Section IV below.
7
Given a series of forecasted explanatory variables and estimated elasticities, it is
8
possible to make an initial, purely mechanical, volume forecast. The methodology that
9
is used in this case to do so was described briefly above and is discussed in more detail
10
in Section IV below.
11
However, the method just described is only the first step in developing the volume
12
forecasts used in this case. At this point, the forecasts are checked for reasonableness.
13
In particular, additional information is investigated at this time. The types of additional
14
information that should be considered at this point fall into two general categories.
15
First, historical changes in mail volume that were not fully explained by the demand
16
equations presented here are analyzed with an eye toward understanding the factors
17
underlying these changes. Typically, the sources of such changes are not amenable to
18
direct inclusion in an econometric equation. Even if this is the case, however, it may be
19
possible, and is certainly desirable, to include such factors in making volume forecasts
20
going forward.
USPS-T-7
42
1
Second, there may be factors which have not (or have only minimally) affected mail
2
volumes historically but which are expected to affect mail volumes in the forecast
3
period. For example, there may be a threat, which is as yet unrealized, that an
4
alternative to a specific type of mail may arise. This was, for example, the case for
5
many years with electronic bill presentment and First-Class workshared mail. For
6
years, it was known that e-mail and the Internet represented a potential alternative to
7
First-Class Mail for receiving bills and statements. This potentiality began to become a
8
reality within the past two to three years. As this happened, it was necessary to
9
augment the strict econometric volume forecast of First-Class workshared letters with
10
additional information regarding the potential impact of electronic bill-presentment on
11
First-Class Mail volumes. By now, the recent events which have affected First-Class
12
Mail volumes have been operating for a sufficiently long time that we are able to
13
represent these with time trends. When influences are newer, however, non-
14
econometric adjustments will still be required.
15
Only when one has endeavored to understand fully all of the factors which may
16
affect mail volumes in the forecast period, whether these factors are amenable to
17
econometric analysis or not, and incorporated them as best as one can in the volume
18
forecasts, should one have full confidence in the resulting volume forecasts.
USPS-T-7
43
1
2
3
B. First-Class Mail
1. General Overview
First-Class Mail is a very heterogeneous class of mail. First-Class Mail includes a
4
wide variety of mail sent by a wide variety of mailers for a wide variety of purposes.
5
This mail can be divided into various substreams of mail based on several possible
6
criteria, including the content of the mail-piece (e.g., bills, statements, advertising, and
7
personal correspondence), the sender of the mail-piece (e.g., households versus
8
businesses versus government), or the recipient of the mail-piece (e.g., households
9
versus business versus government).
10
First-Class Mail can be broadly divided into two categories of mail: Individual
11
Correspondence, consisting of household-generated mail, and nonhousehold-generated
12
mail sent a few pieces at a time; and Bulk Transactions, consisting of nonhousehold-
13
generated mail sent in bulk. Relating these two categories of First-Class Mail to rate
14
categories, Individual Correspondence mail may be thought of as being approximately
15
equivalent to First-Class single-piece Mail, while Bulk Transactions mail could be
16
viewed as comparable to First-Class workshared Mail.
17
First-Class Mail is divided into two subclasses on the basis of the shape of the mail:
18
First-Class letters, flats, and IPPs (referred to here simply as First-Class letters); and
19
First-Class cards. Each of these two subclasses is further divided between single-piece
20
and workshared mail. In the case of First-Class cards, however, single-piece and First-
21
Class workshared cards volumes are relatively small and are somewhat unstable over
22
time (particularly workshared cards volume). Therefore, it is difficult to obtain good
23
separate econometric estimates for these two categories of First-Class cards. Hence,
24
First-Class Mail is divided into three categories for estimation purposes: First-Class
25
single-piece letters, First-Class workshared letters, and First-Class cards.
USPS-T-7
44
1
2. History of First-Class Mail Volume
2
Annual First-Class Mail volumes are shown in Table 4 below from 1970 through
3
2004. Percentage changes in First-Class Mail volume from 1971 through 2004 are
4
shown in Table 5 below. Tables 4 and 5 show volumes and growth rates for First-Class
5
single-piece letters, First-Class workshared letters, total First-Class letters, total cards,
6
and total First-Class Mail.
USPS-T-7
45
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
Ta ble 4
First-Cla ss Ma il Volume
(millions of pieces)
First-Class Letters
W orkshared
Total
Single-Piece
47,948.327
0.000
47,948.327
48,570.710
0.000
48,570.710
48,354.241
0.000
48,354.241
50,221.593
0.000
50,221.593
50,787.976
0.000
50,787.976
50,084.451
0.000
50,084.451
49,901.384
238.357
50,139.741
49,492.374
1,849.630
51,342.004
50,734.338
2,821.012
53,555.350
50,807.539
4,863.929
55,671.468
50,618.035
6,832.491
57,450.526
49,818.740
8,813.451
58,632.191
48,396.758
11,071.710
59,468.468
47,917.584
13,291.101
61,208.685
50,186.055
15,295.783
65,481.838
51,647.961
17,564.241
69,212.202
52,817.615
19,865.834
72,683.449
53,710.714
21,594.965
75,305.679
55,125.410
24,702.497
79,827.907
55,447.005
26,055.896
81,502.901
56,481.808
27,810.521
84,292.329
55,854.437
29,062.843
84,917.280
54,020.849
31,507.301
85,528.150
55,531.933
31,840.566
87,372.500
55,361.085
34,043.032
89,404.117
53,527.014
37,387.685
90,914.699
53,848.090
37,998.282
91,846.372
54,504.037
38,648.275
93,152.312
53,936.203
40,421.136
94,357.339
53,412.621
42,684.840
96,097.461
52,369.535
45,675.520
98,045.055
50,945.540
47,074.794
98,020.334
49,253.266
47,658.076
96,911.342
46,557.786
47,287.971
93,845.757
45,161.746
47,333.818
92,495.564
First-Class
Cards
2,467.193
2,323.332
2,246.015
2,296.859
2,223.231
2,087.810
2,141.717
2,225.694
2,366.851
2,338.233
2,354.084
2,447.242
2,557.463
2,779.720
2,840.292
2,963.320
3,176.779
3,317.527
4,111.434
4,200.790
4,875.731
5,117.145
4,536.837
4,524.523
4,640.952
4,816.931
4,926.662
5,318.337
5,489.394
5,267.824
5,480.714
5,500.088
5,467.290
5,213.099
5,430.832
First-Class
Mail
50,415.520
50,894.042
50,600.256
52,518.452
53,011.207
52,172.261
52,281.458
53,567.698
55,922.201
58,009.701
59,804.610
61,079.433
62,025.931
63,988.405
68,322.130
72,175.522
75,860.228
78,623.206
83,939.341
85,703.691
89,168.060
90,034.425
90,064.987
91,897.023
94,045.069
95,731.630
96,773.034
98,470.649
99,846.733
101,365.286
103,525.768
103,520.422
102,378.632
99,058.856
97,926.396
note: Data show n are f or Postal Fiscal Y ears through 1999, by Government Fiscal Y ears 2000 - 2004
USPS-T-7
46
Ta ble 5
Pe rce ntange Cha nge in First-Cla ss Ma il Volum e
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
2
First-Class Letters
Single-Piece
W orkshared
1.30%
-0.45%
3.86%
1.13%
-1.39%
-0.37%
-0.82%
675.99%
2.51%
52.52%
0.14%
72.42%
-0.37%
40.47%
-1.58%
28.99%
-2.85%
25.62%
-0.99%
20.05%
4.73%
15.08%
2.91%
14.83%
2.26%
13.10%
1.69%
8.70%
2.63%
14.39%
0.58%
5.48%
1.87%
6.73%
-1.11%
4.50%
-3.28%
8.41%
2.80%
1.06%
-0.31%
6.92%
-3.31%
9.82%
0.60%
1.63%
1.22%
1.71%
-1.04%
4.59%
-0.97%
5.60%
-2.32%
6.07%
-2.72%
3.06%
-3.32%
1.24%
-5.47%
-0.78%
-3.00%
0.10%
Total
1.30%
-0.45%
3.86%
1.13%
-1.39%
0.11%
2.40%
4.31%
3.95%
3.20%
2.06%
1.43%
2.93%
6.98%
5.70%
5.02%
3.61%
6.01%
2.10%
3.42%
0.74%
0.72%
2.16%
2.33%
1.69%
1.02%
1.42%
1.29%
1.84%
1.41%
-0.03%
-1.13%
-3.16%
-1.44%
First-Class
Cards
-5.83%
-3.33%
2.26%
-3.21%
-6.09%
2.58%
3.92%
6.34%
-1.21%
0.68%
3.96%
4.50%
8.69%
2.18%
4.33%
7.20%
4.43%
23.93%
2.17%
16.07%
4.95%
-11.34%
-0.27%
2.57%
3.79%
2.28%
7.95%
3.22%
-4.04%
3.08%
0.35%
-0.60%
-4.65%
4.18%
First-Class
Mail
0.95%
-0.58%
3.79%
0.94%
-1.58%
0.21%
2.46%
4.40%
3.73%
3.09%
2.13%
1.55%
3.16%
6.77%
5.64%
5.11%
3.64%
6.76%
2.10%
4.04%
0.97%
0.03%
2.03%
2.34%
1.79%
1.09%
1.75%
1.40%
1.52%
1.50%
-0.01%
-1.10%
-3.24%
-1.14%
note: Data show n are f or Postal Fiscal Y ears through 2000, by Government Fiscal Y ears 2001 - 2004
From 1970 through 2000, First-Class Mail volume grew at an average annual rate of
3
approximately 2.4 percent. Within this time period, growth was strongest through the
4
1980s (4.0 percent per year), slowing to 1.7 percent per year for the 1990s.
USPS-T-7
47
1
Growth in total First-Class letters volume has been concentrated in First-Class
2
workshared letters for a long time. In fact, First-Class single-piece letters volume
3
actually peaked in 1990 at 56.5 billion pieces. For the decade of the 1990s, First-Class
4
single-piece letters volume fell by just under one percent per year, while workshared
5
letters volume grew by more than 5 percent per year.
6
Both single-piece and First-Class workshared letters have experienced breaks from
7
these historical trends in recent years. In the case of single-piece letters, this break
8
came in 2000. From 1990 through 1999, First-Class single-piece letters volume was
9
fairly stable, falling at a modest average rate of 0.6 percent per year. The years 1998
10
and 1999 were fairly typical years in this respect, with First-Class single-piece letters
11
volume declining by about one percent per year in each of these two years. In PFY
12
2000, however, First-Class single-piece letters volume declined by more than 2.3
13
percent, the largest annual decline since 1995. This decline was followed by four years
14
of even greater losses, including a 5.5 percent decline in GFY 2003, the largest annual
15
percentage decline in First-Class single-piece letters since at least 1970. Overall, from
16
1999 through 2004, First-Class single-piece letters volume declined by 16 percent. This
17
represents an average annual decline of 3.4 percent. The average annual decline over
18
this five-year period exceeds the percentage decline in any single year prior to 2000.
19
While largely unprecedented, the decline in First-Class single-piece letters volume
20
over this time period was not entirely unexpected. For example, the R2001-1 after-rates
21
volume forecast projected a 4.8 percent decline in First-Class single-piece letters
22
volume from GFY 2002 to GFY 2003.
23
The break in the historical trend for First-Class workshared letters, on the other
24
hand, was much more unanticipated. From the introduction of worksharing discounts in
25
July, 1976 through the third quarter of 2002, First-Class workshared letters volume was
USPS-T-7
48
1
less in a quarter than in the same quarter the previous year a total of four times –
2
1989Q2, 1990Q1, 1996Q4, and 1997Q1. The last two of these were the result of
3
classification reform (MC95-1), when the presort non-automation discount was reduced
4
from 4.6 cents to 2.5 cents, and were not entirely unexpected. From 1997 (the first full
5
year after MC95-1) through 2001, First-Class workshared letters volume grew at an
6
average annual rate of 5 percent.
7
Then, almost without warning, the growth rate of First-Class workshared letters
8
volume fell sharply, from 4.68 percent in 2002Q1 to 1.34 percent in 2002Q2 to 0.25
9
percent in 2002Q3. From there, First-Class workshared letters volume declined over
10
the same period the previous year for four straight quarters from 2002Q4 through
11
2003Q3 and six of eight quarters overall through 2004Q3.
12
13
14
These recent trends in First-Class letter volume and the ways in which they are dealt
econometrically are discussed in the next section.
3. Trends in First-Class Mail Volume
15
Certainly, one of the most significant factors affecting First-Class Mail volume in
16
recent years is the increasing use of the Internet and electronic media as alternatives to
17
the Postal Service. E-mail has emerged as a potent substitute for personal letters, bills
18
can be paid online, and now, some consumers are even beginning to receive bills and
19
statements through the Internet rather than through the mail. Understanding the
20
emergence of the Internet and its role vis-à-vis the mail is critical, therefore, in
21
understanding First-Class Mail volume, both today and in the future.
22
The treatment of the Internet and electronic diversion within the demand equations
23
presented here was discussed above and is also discussed in more detail in Peter
24
Bernstein’s testimony in this case (USPS-T-8). As outlined there, the First-Class single-
25
piece letters equation includes a measure of Internet Experience as an explanatory
USPS-T-7
49
1
variable, while the First-Class workshared letters equation includes the number of
2
broadband subscribers.
3
Although these Internet variables do a good job of explaining much of the history of
4
First-Class letters volume, recent First-Class letters volumes have fallen even more
5
than can be explained directly by these Internet variables. In particular, both single-
6
piece and First-Class workshared letters volumes declined more sharply from 2002
7
through 2004 than would have been expected, given the growth in the Internet and the
8
condition of the economy at that time.
9
To reflect this, it was decided to include time trends starting in 2002Q4 (which began
10
on July 1, 2002) in the single-piece and First-Class workshared letters equations used
11
in this case. The terrorist attacks of September 11, 2001, occurred about three weeks
12
before the start of Government Fiscal Year 2002 on October 1. Shortly thereafter,
13
several people died as a result of anthrax-laced letters. This occurred during the course
14
of the first two months of GFY 2002. This threat to the Postal Service could have
15
prompted some people to seek alternatives to the Postal Service. A rate increase,
16
which increased the price of a single-piece letter from 34 cents to 37 cents, then took
17
effect on June 30, 2002 (the last day of 2002Q3). Coupled with the bioterrorism scare
18
nine months earlier, this rate increase could have further accelerated people’s desire to
19
find alternatives to the Postal Service.
20
In many cases, new technologies can have a snowball effect. For example, if more
21
people demand the ability to receive statements online, more banks will offer such a
22
service. As the number of banks offering the service increases, the number of people
23
demanding such a service may grow, et cetera. Hence, an increasing desire to find
24
electronic alternatives to the Postal Service in late 2001 and early 2002 could have
USPS-T-7
50
1
started a trend toward increasing electronic substitution that could continue into the
2
foreseeable future.
3
The exact impact of these variables on single-piece and First-Class workshared
4
letters volumes, both historically and over the next three years, is presented and
5
discussed in more detail in the specific discussions of these mail categories which
6
follow.
7
8
9
10
4. Shifts Between First-Class Single-Piece and Workshared Letters Due to
Changes in Worksharing Discounts
Shifts between single-piece and First-Class workshared letters due to changes in
11
price are modeled through the inclusion of the average worksharing First-Class letters
12
discount in the demand equations for both single-piece and First-Class workshared
13
letters. The discount is used here, rather than the price, to reflect the nature of the
14
decision being made by mailers, which is whether to workshare or not, as opposed to a
15
decision of whether to send the mail or not.
16
Holding all other factors constant, the total volume leaving First-Class single-piece
17
letters due solely to changes in worksharing discounts should be exactly equal to the
18
volume entering First-Class workshared letters. Mathematically, this is a restriction that
19
(∂Vsp/∂dws) = -(∂Vws/∂dws)
(II.1)
20
where Vsp is the volume of First-Class single-piece letters, Vws is the volume of First-
21
Class workshared letters, and dws is the worksharing discount. Given the log-log
22
functional form used throughout my testimony,
23
(∂Vsp/∂dws) = βsp·(Vsp/dws)
24
(∂Vws/∂dws) = βws·(Vws/dws)
(II.2)
25
where βsp is the coefficient on the worksharing discount in the single-piece letters
26
equation, which is equivalent to the elasticity with respect to the worksharing discount in
USPS-T-7
51
1
the single-piece letters equation, and βws is the coefficient on the worksharing discount
2
in the worksharing letters equation, which is equivalent to the elasticity with respect to
3
the worksharing discount in the worksharing letters equation.
4
5
Combining these results and canceling out the dws from both sides of the equation,
we get that
βws = -βsp / (Vws/Vsp)
6
7
(II.3)
The ratio (Vws/Vsp) varies over time. This implies that βsp and/or βws also varies over
8
time. In fact, (Vws/Vsp) has grown over time. Hence, the value of βsp must also have
9
grown over time relative to βws. Mathematically, this could be accomplished either
10
through an increase in the value of βsp over time or by a decline in the value of βws over
11
time (or both).
12
The latter of these options, a decline in the value of βws over time, seems more
13
plausible. As more and more mail shifts from single-piece into workshared, the volume
14
of mail left in single-piece letters that could possibly shift as a result of subsequent
15
increases in the worksharing discount decreases. Hence, one might reasonably expect
16
the percentage increase in worksharing letters volume due to changes in the
17
worksharing discount to decline as the ratio of workshared to single-piece letters
18
increases.
19
The demand equations used here are log-log equations of the following form:
20
Ln(Vsp) = a + βsp·Ln(dws) + ...
21
Ln(Vws) = a + βws·Ln(dws) + ...
22
23
24
25
(II.4)
Using equation II.3, the latter of these equations can be restated as follows:
Ln(Vws) = a - βsp·[Ln(dws) / (Vws/Vsp)] + ...
(II.5)
The worksharing discount is divided by the ratio of workshared to single-piece letters
in the workshared letters equation. Using this specification, the coefficient on this
USPS-T-7
52
1
variable from the workshared letters equation, [Ln(dws) / (Vws/Vsp)], is equal to the
2
negative of the discount elasticity in the single-piece letters equation. Hence, the
3
coefficient from the workshared letters equation can be used as a (stochastic) constraint
4
in the single-piece letters equation. That is, the value of βsp is freely estimated in
5
equation II.5. This value of βsp is then used as a stochastic constraint in the First-Class
6
single-piece letters equation.
7
There is, however, one problem with equation II.5 above: the volume of workshared
8
letters (Vws) is expressed, in part, as a function of the volume of workshared letters. To
9
solve this problem, the discount was not divided by the true ratio of workshared to
10
single-piece letters in the demand equation actually used here. Instead, the discount
11
was divided by a fitted ratio of workshared to single-piece letters. This fitted value was
12
constructed by fitting the following equation:
13
Ln(V’ws / V’sp) = a + b0·dMC95 + b1·t + b2·t2
(II.6)
14
where V’ws and V’sp are seasonally adjusted volumes of workshared and single-piece
15
letters, respectively, dMC95 is a dummy variable equal to zero prior to MC95-1 and equal
16
to one after MC95-1, t is a time trend, and t2 is the time trend squared. Equation II.6
17
was estimated using a sample period from 1993Q1 through 2005Q1.
18
The fitted values of equation II.6 were as follows (t-statistics in parentheses):
19
20
21
22
Ln(V’ws / V’sp) = -1.748494 – 0.121290·dMC95 + 0.034237·t - 0.000136·t2
(13.75)
(6.318)
(8.447)
(4.733)
Besides allowing the elasticity with respect to the worksharing discount to change
23
over time in the First-Class workshared letters equation, another feature of equation II.5
24
is that the term [Ln(dws) / (Vws/Vsp)] has the effect of introducing a time trend into the
25
workshared letters equation through the (Vws/Vsp) term. Specifically, from equation II.6,
26
the trend term, (0.034237·t - 0.000136·t2), is incorporated into the First-Class
USPS-T-7
53
1
workshared letters demand equation. This aspect of this variable provides a measure,
2
then, of the growth in the volume of First-Class workshared letters vis-à-vis First-Class
3
single-piece letters volume. The trend aspect of this variable and its impact on First-
4
Class workshared letters volume are discussed in more detail below in the discussion of
5
the First-Class workshared letters equation.
6
5. First-Class Single-Piece Letters
7
a. Factors Affecting First-Class Single-Piece Letters Volume
8
First-Class single-piece letters volume was found to be affected by the following
9
10
11
12
13
14
15
variables:
Employment
The Internet
Recent Trends
Price of First-Class Letters
The effect of these variables on First-Class single-piece letters volume over the past
16
ten years is shown in Table 6 on the next page. Table 6 also shows the projected
17
impacts of these variables through GFY 2007.
18
The Test Year before-rates volume forecast for First-Class single-piece letters is
19
42,987.742 million pieces, a 4.8 percent decline from GFY 2004. The Postal Service’s
20
proposed rates in this case are predicted to reduce the Test Year volume of First-Class
21
single-piece letters by 1.2 percent, for a Test Year after-rates volume forecast for First-
22
Class single-piece letters of 42,459.296 million.
USPS-T-7
54
Other
-0.77%
1.99%
-0.29%
-0.41%
0.30%
-1.10%
0.53%
-0.39%
-0.52%
-0.12%
Total Change
in Volum e
-3.31%
0.60%
1.22%
-1.04%
-0.97%
-1.95%
-2.72%
-3.32%
-5.47%
-3.00%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.19%
1.13%
1.20%
1.16%
1.18%
1.36%
1.21%
1.29%
1.29%
1.20%
1994 - 2004
Total
Avg per Year
12.90%
1.22%
1.79%
0.18%
-2.04%
-0.21%
-33.93%
-4.06%
-3.37%
-0.34%
1.80%
0.18%
5.42%
0.53%
6.64%
0.65%
-0.82%
-0.08%
-18.42%
-2.02%
2001 - 2004
Total
Avg per Year
3.83%
1.26%
-2.28%
-0.77%
-2.04%
-0.69%
-11.77%
-4.09%
-1.61%
-0.54%
-0.74%
-0.25%
1.51%
0.50%
3.04%
1.00%
-1.03%
-0.34%
-11.35%
-3.94%
Forecas t
1.21%
1.10%
1.09%
0.18%
0.22%
0.04%
-1.09%
-1.09%
-1.10%
-3.73%
-3.54%
-3.41%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.58%
0.47%
0.54%
-0.34%
0.51%
0.24%
1.28%
-0.52%
0.00%
-1.99%
-2.89%
-2.62%
3.44%
1.13%
0.45%
0.15%
-3.25%
-1.09%
-10.30%
-3.56%
0.00%
0.00%
0.00%
0.00%
1.60%
0.53%
0.41%
0.14%
0.75%
0.25%
-7.31%
-2.50%
After-Rates Volum e Forecas t
2005
1.21%
2006
1.10%
2007
1.09%
0.18%
0.22%
0.04%
-1.09%
-1.09%
-1.10%
-3.73%
-3.54%
-3.41%
0.00%
-0.71%
-0.18%
0.00%
-0.53%
0.00%
0.58%
0.47%
0.54%
-0.34%
0.51%
0.24%
1.28%
-0.52%
0.00%
-1.99%
-4.08%
-2.80%
0.45%
0.15%
-3.25%
-1.09%
-10.30%
-3.56%
-0.89%
-0.30%
-0.53%
-0.18%
1.60%
0.53%
0.41%
0.14%
0.75%
0.25%
-8.61%
-2.96%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 6
Estimated Impact of Factors Affecting First-Class Single-Piece Letters Volume, 1994 – 2007
Pos tal Prices
Other Factors
Em ploym ent
Trends
Internet
S-P Letters
WS Letters
Inflation Econom etric
1.07%
0.00%
-3.47%
-0.67%
-0.60%
0.65%
-0.68%
0.50%
0.00%
-4.32%
-0.68%
0.33%
0.56%
1.22%
0.65%
0.00%
-4.30%
0.00%
3.47%
0.57%
0.08%
0.71%
0.00%
-3.69%
0.00%
0.00%
0.31%
0.96%
0.59%
0.00%
-3.77%
-0.13%
-0.30%
0.33%
0.91%
0.54%
0.00%
-4.16%
-0.15%
-0.17%
0.73%
1.08%
0.03%
0.00%
-4.62%
-0.18%
-0.15%
0.65%
-0.11%
-1.11%
-0.07%
-4.38%
-0.47%
-0.08%
0.41%
1.55%
-0.83%
-0.88%
-3.95%
-1.14%
-0.66%
0.54%
0.64%
-0.36%
-1.10%
-3.93%
0.00%
0.00%
0.54%
0.81%
2004 - 2007
Total
Avg per Year
3.44%
1.13%
USPS-T-7
55
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 6 above.
5
The effect of the economy on First-Class single-piece letters volume is modeled
6
through the inclusion of employment as an explanatory variable in the single-piece
7
letters demand equation. The relationship between the economy and single-piece
8
letters volume has lessened over time. This is reflected here in the fact that the
9
elasticity of single-piece letters volume with respect to employment has declined over
10
time.
11
In 1995, the elasticity of First-Class single-piece letters with respect to employment
12
was equal to 0.45, meaning a 10 percent increase in employment could be expected to
13
lead to an increase in single-piece letters volume of 4.5 percent. By 2005, it is
14
estimated that this elasticity has fallen to 0.36.
15
The First-Class single-piece letters equation includes a time trend starting in
16
2002Q4. The rationale for this variable was discussed above. Since its inception in late
17
2002, this variable has explained a loss of slightly more than one percent per year of
18
First-Class single-piece letters volume.
19
The impact of the Internet on First-Class single-piece letters volume is measured by
20
including Internet Experience in the First-Class single-piece letters equation. The
21
calculation of Internet Experience was described above. The Internet has had a very
22
strong negative effect on First-Class single-piece letters volume, explaining annual
23
losses that have averaged 4 percent per year for nearly a decade. The negative impact
24
of the Internet is actually forecasted to ameliorate somewhat as the breadth of Internet
25
usage reaches saturation. Nevertheless, the expected impact of the Internet on First-
USPS-T-7
56
1
Class single-piece letters volume is projected to remain formidable over the next three
2
years.
3
The own-price elasticity of First-Class single-piece letters is calculated to be equal to
4
-0.175 (t-statistic of -2.176). The impact of the price of First-Class workshared letters on
5
single-piece letters volume is measured through the inclusion of the average
6
worksharing discount in both the single-piece and First-Class workshared letters
7
equations. The average discount has a current value of 7.9 cents. The estimated
8
discount elasticity for single-piece letters is equal to -0.102 (t-statistic of -5.562). Hence,
9
for example, a one cent increase in the average worksharing discount would be
10
expected to cause approximately a 1.2 percent reduction in First-Class single-piece
11
letters volume.
12
The impact of the worksharing discount on First-Class single-piece letters volume is
13
constrained from the First-Class workshared letters equation. The process by which
14
this is done is described above.
15
The Postal price impacts shown in Table 6 above are the result of changes in
16
nominal prices. Prices enter the demand equations developed here in real terms,
17
however. The impact of inflation reported in Table 6 measures the impact that a change
18
in real Postal prices, in the absence of nominal rate changes, has on the volume of
19
First-Class single-piece letters mail.
20
Other econometric variables include mainly seasonal variables. A more detailed
21
look at the econometric demand equation for First-Class single-piece letters follows.
USPS-T-7
57
b. Econometric Demand Equation
1
i. Summary of Demand Equation
2
3
The demand equation for First-Class single-piece letters in this case models First-
4
Class single-piece letters volume per adult per delivery day as a function of the
5
following explanatory variables:
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
·
Seasonal Variables
·
Total private employment lagged one quarter
·
Total private employment interacted with a time trend (also lagged one
quarter)
·
Time trend starting in 2002Q4, equal to zero before then, increasing by one
each quarter since 2002Q4
·
Internet Experience, constructed from data on consumption expenditures on
Internet Service Providers as described above
·
Dummy variable equal to one starting 1988Q1
This variable reflects when government mail began to be distributed by class of mail.
Prior to 1988, mail sent by the Federal government was not included in total First-Class
Mail measures. Mail volumes since 1988 include government mail.
·
Dummy variable equal to one starting 1993Q1
This variable reflects a change in the Postal Service’s method for constructing RPW
volumes. Prior to 1993, First-Class workshared letters volume was estimated using the
same sampling methodology as First-Class single-piece letters. Since 1993, First-Class
workshared letters volume is measured directly from mailing statement data.
·
Dummy variable for MC95-1, which took effect in 1996Q4
·
Average worksharing discount for First-Class letters
·
Current and one lag of the price of First-Class single-piece letters
As noted above, the coefficient on the average worksharing discount is constrained
from the First-Class workshared letters equation. In addition, the treatment of the
USPS-T-7
58
1
Internet Experience variable is somewhat complicated. This latter issue is discussed
2
next.
3
ii. Box-Cox Transformation of Internet Experience Variable
4
The natural log is taken of most of the macroeconomic and price variables used in
5
estimating the demand equations presented here. In the case of Internet Experience,
6
however, it was not possible to take the natural logarithm because Internet Experience
7
has a value equal to zero prior to 1988.
8
9
10
11
12
Instead, in this case, a Box-Cox transformation was performed on the Internet
Experience variable. That is, Internet Experience entered the demand equations in the
following way:
Ln(Volume) = a + ... + bI · [Internet Experience]γ + ...
(II.7)
A value of γ equal to one would be equivalent to entering Internet Experience directly
13
in the demand equation, and would mean that a given unit increase in the level of
14
Internet Experience would lead to the same percentage decrease in mail volume. For
15
example, an increase in Internet Experience from 0.5 to 0.6 would have the same effect
16
as an increase from 1.7 to 1.8.
17
A value of γ approximately equal to zero would be equivalent to entering the natural
18
logarithm of Internet Experience in the demand equation, and would mean that a given
19
percentage increase in the level of Internet Experience would lead to the same
20
percentage decrease in mail volume. For example, an increase in Internet Experience
21
from 0.5 to 0.6 would have the same effect as an increase from 1.7 to 2.04.
USPS-T-7
59
1
The value of γ was estimated from the First-Class single-piece letters equation using
2
nonlinear least squares. The estimated value of γ used here is equal to 0.326, which
3
had a t-statistic associated with it of 13.38. This value of γ was also applied to the
4
Internet Experience variable in the Free for the Blind and Handicapped equation. A
5
value of γ equal to 0.326 means that an increase in Internet Experience from 1.56 to
6
2.67, the approximate change projected from 2004 to 2007, would have the same effect
7
as an increase from 0.82 to 1.56, the change from the end of 2001 to mid-2004. Hence,
8
the impact of the Internet on First-Class single-piece letters from 2004 to 2007 is
9
estimated to be comparable in magnitude to the impact of the Internet on First-Class
10
single-piece letters from the end of 2001 to mid-2004.
11
Details of the econometric demand equation are shown in Table 7 below. A detailed
12
description of the econometric methodologies used to obtain these results can be found
13
in Section III below.
USPS-T-7
60
1
2
TABLE 7
ECONOMETRIC DEMAND EQUATION FOR FIRST-CLASS SINGLE-PIECE LETTERS
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.175
-2.176
-0.046
-0.400
Current
Lag 1
-0.129
-1.164
Average Worksharing Discount
-0.102
-5.562
Total Employment
Constant
0.673
5.794
Times Trend
-0.0023
-2.546
Internet Experience
Box-Cox Coefficient
0.326
13.38
Coefficient
-0.491
-15.11
Time Trend since 2002Q4
-0.0028
-1.572
Dummy since 1988Q1
0.018
1.744
Dummy since 1993Q1
0.008
0.926
Dummy for MC95-1
0.069
5.406
Seasonal Coefficients
September 1 – 15
-0.539
-1.689
September 16 – 30
-0.230
-2.789
October 1 – December 10
0.088
1.852
December 11 – 19
0.324
2.744
December 20 – 24
-0.179
-1.007
December 25 – February
0.086
1.770
March
-0.149
-2.023
April 1 – 15
0.538
1.591
April 16 – May
-0.195
-1.668
Quarter 1 (October – December)
-0.002
-0.227
Quarter 2 (January – March)
-0.005
-0.333
Quarter 3 (April – June)
-0.037
-1.897
Quarter 4 (July – September)
0.045
3.282
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.102280
Quarter 2 (January – March)
1.002287
Quarter 3 (April – June)
0.965528
Quarter 4 (July – September)
0.931507
REGRESSION DIAGNOSTICS
Sample Period
1983Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
66
Mean-Squared Error
0.000288
Adjusted R-Squared
0.986
USPS-T-7
61
1
6. First-Class Workshared Letters
2
a. Factors Affecting First-Class Workshared Letters Volume
3
First-Class workshared letters volume was found to be affected by the following
4
5
6
7
8
9
10
variables:
Retail Sales
The Internet
Recent Trends
Prices of First-Class Letters and Standard Regular Mail
The effect of these variables on First-Class workshared letters volume over the past
11
ten years is shown in Table 8 on the next page. Table 8 also shows the projected
12
impacts of these variables through GFY 2007.
13
The Test Year before-rates volume forecast for First-Class workshared letters is
14
48,336.414 million pieces, a 2.1 percent increase from GFY 2004. The Postal Service’s
15
proposed rates in this case are predicted to reduce the Test Year volume of First-Class
16
workshared letters by 0.9 percent, for a Test Year after-rates volume forecast for First-
17
Class workshared letters of 47,886.718 million.
USPS-T-7
62
Other
2.14%
-1.54%
1.39%
-1.57%
-0.15%
0.81%
0.69%
-0.73%
0.12%
-1.08%
Total Change
in Volum e
-3.31%
0.60%
1.22%
-1.04%
-0.97%
-1.95%
-2.72%
-3.32%
-5.47%
-3.00%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.24%
1.15%
1.20%
1.19%
1.21%
1.40%
1.24%
1.33%
1.30%
1.21%
1994 - 2004
Total
Avg per Year
13.20%
1.25%
11.08%
1.06%
-3.11%
-0.32%
-4.08%
-0.42%
-12.12%
-1.28%
-1.98%
-0.20%
37.46%
3.23%
0.49%
0.05%
0.00%
0.00%
-18.42%
-2.02%
2001 - 2004
Total
Avg per Year
3.89%
1.28%
2.28%
0.76%
-3.11%
-1.05%
-3.43%
-1.16%
-3.48%
-1.17%
-0.95%
-0.32%
6.95%
2.26%
0.62%
0.21%
-1.69%
-0.57%
-11.35%
-3.94%
Forecas t
1.24%
1.11%
1.10%
1.59%
0.37%
0.52%
-1.71%
-1.69%
-1.69%
-1.58%
-1.64%
-1.60%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
1.85%
1.58%
1.43%
-0.25%
0.01%
-0.31%
2.13%
-0.78%
0.00%
3.24%
-1.09%
-0.61%
3.49%
1.15%
2.49%
0.82%
-5.00%
-1.70%
-4.74%
-1.61%
0.00%
0.00%
0.00%
0.00%
4.94%
1.62%
-0.56%
-0.19%
1.33%
0.44%
1.50%
0.50%
After-Rates Volum e Forecas t
2005
1.24%
2006
1.11%
2007
1.10%
1.59%
0.37%
0.52%
-1.71%
-1.69%
-1.69%
-1.58%
-1.64%
-1.60%
0.00%
-0.42%
-0.86%
0.00%
-0.51%
0.00%
1.85%
1.58%
1.43%
-0.25%
0.01%
-0.31%
2.13%
-0.78%
0.00%
3.24%
-2.01%
-1.46%
2.49%
0.82%
-5.00%
-1.70%
-4.74%
-1.61%
-1.28%
-0.43%
-0.51%
-0.17%
4.94%
1.62%
-0.56%
-0.19%
1.33%
0.44%
-0.31%
-0.10%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 8
Estimated Impact of Factors Affecting First-Class W orkshared Letters Volume, 1994 – 2007
Pos tal Prices
Other Factors
Retail Sales
Trends
Internet
Firs t-Clas s
Standard
Inflation Econom etric
1.41%
0.00%
0.00%
-0.99%
-0.28%
5.85%
0.21%
0.87%
0.00%
0.00%
-2.70%
-0.16%
4.55%
-0.39%
0.99%
0.00%
0.00%
-5.07%
-0.21%
3.04%
0.56%
0.97%
0.00%
0.00%
0.64%
0.00%
3.23%
0.10%
2.20%
0.00%
-0.03%
-0.21%
-0.42%
3.17%
-0.23%
2.12%
0.00%
-0.13%
-0.42%
-0.22%
2.98%
0.31%
-0.24%
0.00%
-0.51%
-0.45%
0.25%
2.78%
-0.69%
0.23%
-0.11%
-0.96%
-0.97%
-0.12%
2.31%
0.31%
0.34%
-1.36%
-1.15%
-1.69%
-0.83%
2.65%
-0.09%
1.71%
-1.68%
-1.36%
-0.86%
0.00%
1.84%
0.40%
2004 - 2007
Total
Avg per Year
3.49%
1.15%
USPS-T-7
63
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 8 above.
5
First-Class workshared letters have a retail sales elasticity of 0.459 (t-statistic of
6
2.703), meaning that a 10 percent increase in retail sales will lead to a 4.59 percent
7
increase in the volume of First-Class workshared letters.
8
The First-Class workshared letters equation includes two trend variables. The
9
combined impact of these two variables is shown in Table 8. The first trend is the fitted
10
ratio of workshared to single-piece letters. The theoretical basis for and construction of
11
this variable were described above in section 4. The inclusion of this trend is roughly
12
comparable to including a trend and a trend-squared term in the First-Class workshared
13
letters equation. Econometrically, this variable explained an increase in First-Class
14
workshared letters volume of more than 28 percent from 1994 through 2004. The
15
annual impact of this trend has fallen considerably over this time period due to the
16
inclusion of the trend-squared term, from 4.7 percent in 1995 to 1.2 percent in 2004.
17
Moving forward, the impact of this trend is projected to continue to decline to less than
18
one percent per year by GFY 2006 and GFY 2007.
19
The First-Class workshared letters equation also includes a negative time trend
20
beginning in 2002Q4. The rationale for this variable was discussed above. Since its
21
inception, this variable has explained a loss of approximately 1.7 percent per year of
22
First-Class workshared letters volume.
23
The impact of the Internet on First-Class workshared letters volume is measured by
24
including the number of broadband subscribers (lagged one year) in the First-Class
25
workshared letters equation. The impact of the Internet on workshared letters volume
USPS-T-7
64
1
has been gradually increasing over time from −0.5 percent between 2000 and 2001 to
2
−1.4 percent from 2003 to 2004. The impact of the Internet is expected to continue to
3
grow through the Test Year with the Internet projected to explain a 1.6 percent decline
4
in First-Class workshared letters volume from 2005 to 2006.
5
The own-price elasticity of First-Class workshared letters was calculated to be equal
6
to -0.329 (t-statistic of -2.179). In addition, the average worksharing discount for First-
7
Class letters and the average discount for Standard Regular letters also help to explain
8
First-Class workshared letters volume. These variables are described in more detail in
9
section b. below.
10
The Postal price impacts shown in Table 8 above are the result of changes in
11
nominal prices. Prices enter the demand equations developed here in real terms,
12
however. The impact of inflation reported in Table 8 measures the impact that a change
13
in real Postal prices, in the absence of nominal rate changes, has on the volume of
14
First-Class workshared letters mail.
15
Other econometric variables include mainly seasonal variables. A more detailed
16
look at the econometric demand equation for First-Class workshared letters follows.
b. Econometric Demand Equation
17
18
The demand equation for First-Class workshared letters in this case models First-
19
Class workshared letters volume per adult per delivery day as a function of the following
20
explanatory variables:
21
22
23
24
25
26
27
28
·
Seasonal Variables
·
Retail Sales
·
Time trend starting in 2002Q4, equal to zero before then, increasing by one
each quarter since 2002Q4
·
Number of broadband subscribers lagged one year
USPS-T-7
65
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
This variable enters the workshared letters equation unlogged. A Box-Cox
transformation, as described in the discussion of single-piece letters above, was
investigated here. The resulting Box-Cox coefficient was found to be insignificantly
different from one. Therefore, the number of broadband subscribers was simply
entered directly into the First-Class workshared letters equation.
·
Dummy variable equal to one starting 1993Q1
This variable reflects a change in the Postal Service’s method for constructing RPW
volumes. Prior to 1993, First-Class workshared letters volume was estimated using the
same sampling methodology as First-Class single-piece letters. Since 1993, First-Class
workshared letters volume has been measured directly from mailing statement data.
·
Dummy variable for MC95-1, which took effect in 1996Q4
·
Average discount for Standard Regular letters
This variable is equal to the average cost savings for a typical Standard Regular letter
versus being sent as a First-Class workshared letter. The coefficient on this variable is
constrained from the Standard Regular demand equation using the Slutsky-Schultz
symmetry condition.
·
Average worksharing discount for First-Class letters
The average worksharing discount is divided by the fitted ratio of First-Class
workshared letters volume to single-piece letters volume. The rationale for this as well
as the calculation of this fitted ratio was described above.
·
Current and four lags of the price of First-Class workshared letters
31
Details of the econometric demand equation are shown in Table 9 below. A detailed
32
description of the econometric methodologies used to obtain these results can be found
33
in Section III below.
USPS-T-7
66
1
2
TABLE 9
ECONOMETRIC DEMAND EQUATION FOR FIRST-CLASS WORKSHARED LETTERS
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-2.179
-0.329
-0.128
-0.728
Current
Lag 1
-0.040
-0.217
Lag 2
-0.002
-0.012
-0.043
-0.269
Lag 3
Lag 4
-0.116
-0.996
Avg. First-Class Worksharing Discount
0.108
5.554
Avg. Standard Regular Letters Discount
-0.097
-2.856
(relative to First-Class)
Retail Sales
0.459
2.703
Number of Broadband subscribers
Box-Cox Coefficient
1.000
(N/A)
Coefficient
-1.261
-2.065
Time Trend since 2002Q4
-0.0043
-1.553
Dummy since 1993Q1
-0.069
-4.550
Dummy for MC95-1
-0.068
-5.013
Seasonal Coefficients
September 1 – December 10
0.157
0.546
December 11 – 31
0.462
1.610
January – May
0.122
0.421
June
0.229
0.262
Quarter 1 (October – December)
-0.063
-8.747
Quarter 2 (January – March)
0.070
12.63
Quarter 3 (April – June)
-0.054
-0.275
Quarter 4 (July – September)
0.047
0.239
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.023760
Quarter 2 (January – March)
1.053845
Quarter 3 (April – June)
0.965235
Quarter 4 (July – September)
0.959176
REGRESSION DIAGNOSTICS
Sample Period
1991Q1 – 2005Q1
Autocorrelation Coefficients
AR-4: -0.367
Degrees of Freedom
32
Mean-Squared Error
0.000173
Adjusted R-Squared
0.984
USPS-T-7
67
1
7. Understanding First-Class Letters Price Elasticities
2
The own-price elasticities of First-Class single-piece and workshared letters cited
3
above measure the effect of a change in these prices holding everything else constant.
4
Mathematically, what this means is the effect of a change in these prices holding all
5
other variables in the equation constant. That is, the First-Class single-piece letters’
6
own-price elasticity represents the impact on First-Class single-piece letters volume of a
7
change in the single-piece letters price, holding the worksharing discount constant. This
8
is not, however, the impact of a change in the single-piece letters price holding the
9
workshared letters price constant, since changing the single-piece letters price while
10
holding the workshared letters price constant would, of course, change the worksharing
11
discount.
12
The “own-price elasticity” of First-Class single-piece letters, holding the price of First-
13
Class workshared letters constant, is not -0.175, but is, instead, equal to -0.175 plus the
14
impact of the change in the workshared letters discount on single-piece letters volume.
15
Similarly, the “own-price elasticity” of First-Class workshared letters, holding the price of
16
First-Class single-piece letters constant, is not -0.329, but is equal to -0.329 plus the
17
impact of the resulting change in the workshared letters discount.
18
Table 10 below shows the levels of the single-piece and First-Class workshared
19
letters price indices as well as the average worksharing discount used in this case.
20
These prices are constructed as weighted averages of the various rate elements
21
associated with First-Class letters. The weights used are calculated from 2004 billing
22
determinants provided by the Postal Service. These prices are then divided by the
23
implicit price deflator for personal consumption expenditures to express them in 2000
24
dollars.
USPS-T-7
68
1
2
GFY
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Ta ble 10
First-Cla ss Lette rs Price Indice s
(2000 dollars)
Single-Piece W orkshared
Discount
$0.398850
$0.281950
$0.037749
$0.406809
$0.278930
$0.044408
$0.450793
$0.312003
$0.049991
$0.433297
$0.300695
$0.047728
$0.416568
$0.284008
$0.050964
$0.422214
$0.284520
$0.058212
$0.420701
$0.281673
$0.061657
$0.407323
$0.272716
$0.059696
$0.418891
$0.282248
$0.058692
$0.426540
$0.289028
$0.057306
$0.404447
$0.274057
$0.054338
$0.427352
$0.291483
$0.055873
$0.435604
$0.297980
$0.056249
$0.425027
$0.290700
$0.054928
$0.416512
$0.284876
$0.053827
$0.432904
$0.299569
$0.056389
$0.432608
$0.296718
$0.061020
$0.424287
$0.280074
$0.073185
$0.420195
$0.277373
$0.072479
$0.419360
$0.278073
$0.074003
$0.411204
$0.273122
$0.073175
$0.408747
$0.273310
$0.072477
$0.417324
$0.279568
$0.072058
$0.431590
$0.290998
$0.075187
$0.423250
$0.285375
$0.073734
$0.414905
$0.279749
$0.072281
The average worksharing discount in Table 10 is not simply equal to the difference
3
between the single-piece and workshared price indices. Rather, it is the difference
4
between the First-Class workshared letters price and a hypothetical single-piece letters
5
price index that would be constructed using First-Class workshared letters billing
6
determinants. The worksharing discount is intended to measure how much an average
7
worksharing mailer saves as a result of worksharing.
8
One reason why the price index for First-Class single-piece letters is greater than
9
the price index for First-Class workshared letters is that, according to the 2004 billing
10
determinants, the average First-Class single-piece letter contained 0.33 additional
USPS-T-7
69
1
ounces, while the average First-Class workshared letters only contained 0.06 additional
2
ounces. This does not, however, mean that a typical piece of First-Class Mail sheds
3
0.27 additional ounces by virtue of being workshared. Rather, this indicates that a
4
typical First-Class letter which is likely to be workshared is lighter than a typical First-
5
Class letter that is not likely to be workshared. In other words, from the last row in
6
Table 10 above, about 47 percent of the difference in the average price of single-piece
7
and First-Class workshared letters in 2005 is because of intrinsic differences in the
8
nature of the mailpieces themselves (i.e., the number of additional ounces) and 53
9
percent of the difference ($0.072 divided by ($0.415 – $0.280)) is because of the
10
11
worksharing discounts offered by the Postal Service.
Given the current level of real First-Class letters prices, as shown in the last row of
12
Table 10, and the price elasticities presented in Tables 7 and 9 above, then, a 10
13
percent increase in the price of First-Class single-piece letters, holding the price of First-
14
Class workshared letters constant, will lead to a 5.6 percent reduction in First-Class
15
single-piece letters volume, offset in part by a 4.4 percent increase in First-Class
16
workshared letters volume, leading to an overall 0.4 percent decline in total First-Class
17
letters volume.
18
A 10 percent increase in the price of First-Class workshared letters, holding the price
19
of First-Class single-piece letters constant, will lead to an 8.1 percent reduction in First-
20
Class workshared letters volume, offset in part by a 5.1 percent increase in First-Class
21
single-piece letters volume, leading to an overall 1.7 percent decline in total First-Class
22
letters volume.
23
A 10 percent increase in the prices of both single-piece and First-Class workshared
24
letters will also lead to a 10 percent increase in the average worksharing discount. In
25
this case, the positive impact of the change in the discount on First-Class workshared
USPS-T-7
70
1
letters volume will be offset by the negative impact of the decrease in the discount on
2
First-Class single-piece letters volume. Hence, in this case, the overall effect on total
3
First-Class letters volume will be equivalent to the average of the own-price elasticities
4
for single-piece and First-Class workshared letters, leading to a 2.3 percent change in
5
volume. Thus, the net price elasticity of First-Class letters is estimated to be
6
approximately equal to −0.23 in this case.
7
8
8. First-Class Cards
a. Volume History
9
First-Class cards volume growth has generally tracked that of First-Class letters,
10
with cards volume growing slightly more rapidly overall. Since 1976, for example, First-
11
Class cards annual growth rates have exceeded First-Class letters growth rates by
12
approximately 1.2 percent per year, although this difference has varied a good bit from
13
year to year.
14
One possible reason for the more rapid growth of First-Class cards is that a
15
somewhat greater proportion of First-Class cards than letters are advertising mail.
16
Advertising mail volume has historically grown more rapidly than other types of First-
17
Class Mail. According to the 2002 Household Diary Study, for example, the total
18
number of pieces of First-Class Mail from business and government received by
19
households per week rose from 6.4 to 8.7 between 1987 and 2002 (13.6 percent), while
20
the number of such pieces that were characterized as “Advertising Only” rose by more
21
than 120 percent, from 0.9 to 2.0 over the same time period.
22
The reason why a greater proportion of First-Class cards than letters are advertising
23
is that First-Class cards rates are lower than Standard Regular rates (for which there
24
are no card-specific rates) for much card-shaped mail, so it is actually cheaper for many
USPS-T-7
71
1
advertisers to send card-sized advertising First-Class instead of Standard. First-Class
2
letter rates, on the other hand, are universally higher than Standard rates.
3
Another possible reason why First-Class cards volume has grown somewhat more
4
rapidly than First-Class letters volume is that First-Class cards prices have increased
5
somewhat less than First-Class letters prices. In 1970, a First-Class single-piece card
6
cost 5 cents while a First-Class single-piece letter cost 6 cents. Since then, the price of
7
a single-piece letter has risen to 37 cents − an increase of 517 percent − while the price
8
of a First-Class single-piece card has risen to only 23 cents − a mere 360 percent
9
increase. More recently, since 1991, the First-Class single-piece card rate has
10
increased by 21.1 percent, from 19 cents to 23 cents, while the First-Class single-piece
11
letter rate has increased by 27.6 percent, from 29 cents to 37 cents.
12
b. Factors Affecting First-Class Cards Volume
13
A single equation was estimated for total First-Class cards. As in past rate cases,
14
separate equations for single-piece and workshared cards were not feasible, primarily
15
due to the somewhat erratic volume history of workshared cards.
16
17
i. Economy
The First-Class cards equation does not include any macroeconomic variables
18
because, historically, First-Class cards volume has been relatively insensitive to
19
changes in economic conditions. One reason for this may be that during economic
20
downturns some mailers may react to the more difficult economic conditions by shifting
21
from letters to cards, thereby saving on both postage and paper costs. Shifts of this
22
nature could offset more general reductions in correspondence and transactions mail
23
volumes at these times.
USPS-T-7
72
ii. Internet
1
2
First-Class cards volume has been somewhat negatively affected by the Internet.
3
This effect is modeled here by including the Internet Experience variable described
4
above in the First-Class cards equation. As with First-Class single-piece letters,
5
Internet Experience is entered into the First-Class cards equation using a Box-Cox
6
specification. For First-Class cards, the estimated Box-Cox coefficient is equal to 0.180
7
(t-statistic of 2.344). This is considerably less than the Box-Cox coefficient from the
8
single-piece letters equation (0.326). Because the Box-Cox coefficient here is closer to
9
zero, the impact of the Internet on First-Class cards volume is estimated to be closer to
10
a constant-elasticity model. That is, the effect of the Internet on First-Class cards
11
volume is closer to being a constant percentage change in volume for a given
12
percentage change in Internet Experience. The result of this is that, because the
13
growth rate of Internet Experience is declining in percentage terms over time, the
14
percentage change in First-Class cards volume being attributed to Internet Experience
15
is also declining over time.
iii. Prices
16
17
First-Class cards volume is modeled to be affected by the prices of both First-Class
18
cards and Standard Regular letters. The effect of the price of First-Class cards is
19
measured through a price index for First-Class cards. The effect of the price of
20
Standard Regular letters on First-Class cards volume, however, is somewhat more
21
complicated.
USPS-T-7
73
1
Prior to R87-1, which took effect in 1988Q3, Standard Regular letters prices were
2
uniformly less than First-Class cards rates. In R87-1, however, First-Class cards were
3
priced below Standard Regular letters. As a result, many direct-mail advertisers shifted
4
from Standard Regular to First-Class cards. In R90-1 (1991Q2), this rate relationship
5
was reversed for most mail. Since that time, however, the percentage of Standard
6
Regular letters for which First-Class cards rates are less expensive has tended to
7
change whenever rates have changed.
8
The effect of Standard Regular rates on First-Class cards volume is measured by a
9
variable which I refer to as the crossover dummy. The crossover dummy variable is
10
equal to the percentage of Standard Regular letters for which First-Class cards rates
11
are less than corresponding Standard Regular rates. This variable has a value of zero
12
until R87-1 (1988Q3), at which time it reached a value of 100 percent. The value of this
13
variable fell to a value of 20.0 percent with the implementation of R90-1 rates (1991Q2).
14
It has a current value of 38.2 percent.
15
16
17
18
19
20
iv. Summary of Demand Equation Specification
First-Class cards volume was found to be affected by the following variables:
Electronic Diversion
Prices of First-Class Cards and Standard Regular Mail
The effect of these variables on First-Class cards volume over the past ten years is
21
shown in Table 11 on the next page. Table 11 also shows the projected impacts of
22
these variables through GFY 2007.
USPS-T-7
74
1
The Test Year before-rates volume forecast for First-Class cards is 5,544.356 million
2
pieces, a 2.1 percent increase from GFY 2004. The Postal Service’s proposed rates in
3
this case are predicted to reduce the Test Year volume of First-Class cards by 1.5
4
percent, for a Test Year after-rates volume forecast for First-Class cards of 5,463.895
5
million.
USPS-T-7
75
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.21%
1.14%
1.24%
1.16%
1.17%
1.38%
1.23%
1.31%
1.28%
1.24%
1994 - 2004
Total
Avg per Year
13.08%
1.24%
-8.91%
-0.93%
2.48%
0.25%
1.15%
0.11%
7.29%
0.71%
0.44%
0.04%
1.69%
0.17%
17.02%
1.58%
2001 - 2004
Total
Avg per Year
3.88%
1.28%
-2.27%
-0.76%
-5.01%
-1.70%
-2.07%
-0.69%
2.01%
0.67%
0.39%
0.13%
2.10%
0.70%
-1.26%
-0.42%
Forecas t
1.22%
1.12%
1.11%
-0.65%
-0.61%
-0.57%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.78%
0.63%
0.71%
-0.21%
0.01%
-0.42%
-0.34%
0.14%
0.00%
0.79%
1.29%
0.83%
3.49%
1.15%
-1.81%
-0.61%
0.00%
0.00%
0.00%
0.00%
2.14%
0.71%
-0.62%
-0.21%
-0.20%
-0.07%
2.94%
0.97%
After-Rates Volum e Forecas t
2005
1.22%
2006
1.12%
2007
1.11%
-0.65%
-0.61%
-0.57%
0.00%
-1.45%
-0.38%
0.00%
0.00%
0.00%
0.78%
0.63%
0.71%
-0.21%
0.01%
-0.42%
-0.34%
0.14%
0.00%
0.79%
-0.18%
0.45%
-1.81%
-0.61%
-1.82%
-0.61%
0.00%
0.00%
2.14%
0.71%
-0.62%
-0.21%
-0.20%
-0.07%
1.06%
0.35%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 11
Estimated Impact of Factors Affecting First-Class Cards Volume, 1994 – 2007
Total Change
Pos tal Prices
Other Factors
Internet
Firs t-Clas s
Standard
Inflation Econom etric
Other
in Volum e
-1.08%
-0.98%
2.19%
0.88%
0.81%
0.73%
3.79%
-1.20%
0.37%
0.88%
0.74%
0.37%
-0.01%
2.28%
-1.14%
8.76%
-0.42%
0.76%
-0.73%
-0.45%
7.95%
-0.88%
0.11%
0.00%
0.43%
0.89%
1.48%
3.22%
-0.84%
-0.05%
0.00%
0.41%
-0.47%
-4.23%
-4.04%
-0.90%
-0.06%
0.00%
0.95%
0.02%
2.61%
4.04%
-0.95%
-0.18%
0.62%
0.88%
-0.81%
-0.41%
0.35%
-0.85%
-2.00%
-0.96%
0.58%
0.08%
1.30%
-0.60%
-0.72%
-3.07%
-1.12%
0.71%
-0.11%
-1.65%
-4.65%
-0.72%
0.00%
0.00%
0.71%
0.42%
2.48%
4.18%
2004 - 2007
Total
Avg per Year
3.49%
1.15%
USPS-T-7
76
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 11 above.
5
The Internet has had a relatively modest negative impact on First-Class cards
6
volume. The impact of the Internet is projected to decline modestly in the forecast
7
period to approximately 0.6 percent per year. This is about one-fourth of the level of the
8
econometric impact of the Internet on First-Class letters volume.
9
The own-price elasticity of First-Class cards was calculated to be equal to -0.376
10
(t−statistic of -3.411). The percentage of Standard Regular letter mail for which First-
11
Class cards prices are less expensive fell from 76.7 percent in June, 2001 to 38.2
12
percent 13 months later. This had the effect of reducing First-Class cards volume by
13
approximately 2.6 percent over this time period. This variable has remained unchanged
14
since then.
15
The Postal price impacts shown in Table 11 above are the result of changes in
16
nominal prices. Prices enter the demand equations developed here in real terms,
17
however. The impact of inflation reported in Table 11 measures the impact that a
18
change in real Postal prices, in the absence of nominal rate changes, has on the
19
volume of First-Class cards mail.
20
21
Other econometric variables include mainly seasonal variables. A more detailed
look at the econometric demand equation for First-Class cards follows.
c. Econometric Demand Equation
22
23
24
25
The demand equation for First-Class cards in this case models First-Class cards
volume per adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
USPS-T-7
77
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
·
Crossover Dummy Variable as described above
·
Crossover Trend, which is a simple time trend beginning in 1988Q4 and
plateauing in 1990Q4 to reflect lagged reaction by mailers to the initial R87-1
rate crossover
·
Dummy variables equal to one in 1988Q4 and 1991Q3, respectively, and zero
elsewhere, to adjust for apparent outliers in these quarters. These outliers
are most likely the result of the R87-1 rate crossover which priced First-Class
cards less than Standard Regular mail and the R90-1 rates which reversed
this relationship for most mail as noted above.
·
Internet Experience, constructed from data on consumption expenditures on
Internet Service Providers as described above
The Internet Experience variable is entered into the demand equation with a Box-Cox
coefficient as described above.
·
Dummy variable for MC95-1, which took effect in 1996Q4
·
Current and one lag of the price of First-Class cards
Details of the econometric demand equation are shown in Table 12 below. A
24
detailed description of the econometric methodologies used to obtain these results can
25
be found in Section III below.
USPS-T-7
78
1
2
TABLE 12
ECONOMETRIC DEMAND EQUATION FOR FIRST-CLASS CARDS
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.376
-3.411
-0.084
-0.395
Current
Lag 1
-0.292
-1.551
Pct of Cards Rates less than Std Regular
0.069
3.775
Time Trend Interacted with Std Regular
0.024
11.17
Rate Crossover (1988Q4 – 1990Q4)
Internet Experience
Box-Cox Coefficient
0.180
2.344
Coefficient
-0.168
-4.848
Dummy for 1988Q4
0.150
4.303
Dummy for 1991Q3
0.222
5.688
Dummy for MC95-1
0.080
4.303
Seasonal Coefficients
September 16 – 30
0.370
1.778
October
0.175
0.742
November 1 – December 17
-0.067
-0.465
December 18 – 31
-0.357
-0.708
January – March
0.199
2.041
April – May
-0.140
-2.865
Quarter 1 (October – December)
0.123
1.573
Quarter 2 (January – March)
-0.150
-1.530
Quarter 3 (April – June)
0.100
3.029
Quarter 4 (July – September)
-0.073
-2.225
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.060391
Quarter 2 (January – March)
1.013951
Quarter 3 (April – June)
0.972532
Quarter 4 (July – September)
0.955056
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
51
Mean-Squared Error
0.000712
Adjusted R-Squared
0.887
3
USPS-T-7
79
1
2
3
C. Standard Mail
1. Overview of Direct-Mail Advertising
More than 90 percent of Standard Mail could be characterized as direct-mail
4
advertising. Hence, understanding the demand for direct-mail advertising is the key to
5
understanding the demand for Standard Mail volume.
6
7
Table 13 below presents data on advertising expenditures by major media as
reported by Robert Coen of Universal McCann-Erickson.
8
Direct mail’s share of total advertising expenditures grew by more than 40 percent
9
between 1970 and 2004. Most of this growth was between 1980 and 1991, when the
10
share of total advertising expenditures that was spent on direct mail rose from 14.18
11
percent to 19.06 percent. While the growth in direct-mail advertising’s market share has
12
slowed considerably since that time, total direct-mail advertising expenditures still grew
13
by 13.6 percent from 1991 through 2004.
USPS-T-7
80
1
Table 13
Adve r tis ing Expe nditure s by M ajor M e dia
(millions of dollars, nominal)
New spapers
2
Magazines
Radio
Television
Direct-Mail
Other
Y ear
Total
Expenditures
Share
Expenditures
Share
Expenditures
Share
Expenditures
Share
Expenditures
Share
Expenditures
Share
1970
19,550
5,704
29.18%
1,292
6.61%
1,308
6.69%
3,596
18.39%
2,766
14.15%
4,884
24.98%
1971
20,700
6,167
29.79%
1,370
6.62%
1,445
6.98%
3,534
17.07%
3,067
14.82%
5,117
24.72%
1972
23,210
6,938
29.89%
1,440
6.20%
1,612
6.95%
4,091
17.63%
3,420
14.74%
5,709
24.60%
1973
24,980
7,481
29.95%
1,448
5.80%
1,723
6.90%
4,460
17.85%
3,698
14.80%
6,170
24.70%
1974
26,620
7,842
29.46%
1,504
5.65%
1,837
6.90%
4,854
18.23%
4,054
15.23%
6,529
24.53%
1975
27,900
8,234
29.51%
1,465
5.25%
1,980
7.10%
5,263
18.86%
4,124
14.78%
6,834
24.49%
1976
33,300
9,618
28.88%
1,789
5.37%
2,330
7.00%
6,721
20.18%
4,786
14.37%
8,056
24.19%
1977
37,440
10,751
28.72%
2,162
5.77%
2,634
7.04%
7,612
20.33%
5,164
13.79%
9,117
24.35%
1978
43,330
12,214
28.19%
2,597
5.99%
3,052
7.04%
8,955
20.67%
5,987
13.82%
10,525
24.29%
1979
48,780
13,863
28.42%
2,932
6.01%
3,310
6.79%
10,154
20.82%
6,653
13.64%
11,868
24.33%
1980
53,570
14,794
27.62%
3,149
5.88%
3,702
6.91%
11,488
21.44%
7,596
14.18%
12,841
23.97%
1981
60,460
16,528
27.34%
3,533
5.84%
4,230
7.00%
12,889
21.32%
8,944
14.79%
14,336
23.71%
1982
66,670
17,694
26.54%
3,710
5.56%
4,670
7.00%
14,713
22.07%
10,319
15.48%
15,564
23.34%
1983
76,000
20,582
27.08%
4,233
5.57%
5,210
6.86%
16,879
22.21%
11,795
15.52%
17,301
22.76%
1984
88,010
23,522
26.73%
4,932
5.60%
5,817
6.61%
20,043
22.77%
13,800
15.68%
19,896
22.61%
1985
94,900
25,170
26.52%
5,155
5.43%
6,490
6.84%
21,287
22.43%
15,500
16.33%
21,298
22.44%
1986
102,370
26,990
26.37%
5,317
5.19%
6,949
6.79%
23,199
22.66%
17,145
16.75%
22,770
22.24%
1987
110,270
29,412
26.67%
5,607
5.08%
7,206
6.53%
24,262
22.00%
19,111
17.33%
24,672
22.37%
1988
118,750
31,197
26.27%
6,072
5.11%
7,798
6.57%
26,131
22.01%
21,115
17.78%
26,437
22.26%
1989
124,770
32,368
25.94%
6,716
5.38%
8,323
6.67%
27,459
22.01%
21,945
17.59%
27,959
22.41%
1990
129,968
32,281
24.84%
6,803
5.23%
8,726
6.71%
29,247
22.50%
23,370
17.98%
29,541
22.73%
1991
128,352
30,409
23.69%
6,524
5.08%
8,476
6.60%
28,606
22.29%
24,460
19.06%
29,877
23.28%
1992
133,750
30,737
22.98%
7,000
5.23%
8,654
6.47%
31,079
23.24%
25,392
18.98%
30,888
23.09%
1993
140,956
32,025
22.72%
7,357
5.22%
9,457
6.71%
32,471
23.04%
27,266
19.34%
32,380
22.97%
1994
153,024
34,356
22.45%
7,916
5.17%
10,529
6.88%
36,342
23.75%
29,638
19.37%
34,243
22.38%
1995
165,147
36,317
21.99%
8,580
5.20%
11,338
6.87%
38,886
23.55%
32,866
19.90%
37,160
22.50%
1996
178,113
38,402
21.56%
9,010
5.06%
12,269
6.89%
43,824
24.60%
34,509
19.37%
40,099
22.51%
1997
191,307
41,670
21.78%
9,821
5.13%
13,491
7.05%
45,643
23.86%
36,890
19.28%
43,792
22.89%
1998
206,697
44,292
21.43%
10,518
5.09%
15,073
7.29%
49,513
23.95%
39,620
19.17%
47,681
23.07%
1999
222,308
46,648
20.98%
11,433
5.14%
17,215
7.74%
52,581
23.65%
41,403
18.62%
53,028
23.85%
2000
247,472
49,050
19.82%
12,370
5.00%
19,295
7.80%
60,257
24.35%
44,591
18.02%
61,909
25.02%
2001
231,287
44,255
19.13%
11,095
4.80%
17,861
7.72%
54,617
23.61%
44,725
19.34%
58,734
25.39%
2002
236,875
44,031
18.59%
10,995
4.64%
18,877
7.97%
58,365
24.64%
46,067
19.45%
58,540
24.71%
2003
245,477
44,843
18.27%
11,435
4.66%
19,100
7.78%
60,746
24.75%
48,370
19.70%
60,983
24.84%
2004
263,699
46,935
17.80%
12,121
4.60%
19,779
7.50%
67,089
25.44%
52,240
19.81%
65,535
24.85%
2005
280,617
49,618
17.68%
13,006
4.63%
20,981
7.48%
69,243
24.68%
57,203
20.38%
70,566
25.15%
USPS-T-7
81
1
The demand for Standard Mail volume is the result of a choice by advertisers
2
regarding how much to spend on direct-mail advertising expenditures. The decision
3
process made by direct-mail advertisers can be decomposed into three separate, but
4
interrelated, decisions:
5
(1) How much to invest in advertising?
6
(2) Which advertising media to use?
7
(3) Which mail category to use to send mail-based advertising?
8
9
These three decisions are integrated into the demand equations associated with
Standard Mail volume by including a set of explanatory variables in the demand
10
equations for Standard Mail that addresses each of these three decisions. Each of
11
these three decisions, and the implications for Standard Mail equations, are considered
12
separately below.
13
14
2. Advertising Decisions and Their Impact on Mail Volume
a. How Much to Invest in Advertising
15
The amount of advertising expenditures made by a business is a decision made as
16
part of a profit-maximizing optimization problem. Advertising expenditures are chosen
17
so that the expected profits from the additional sales generated by the last dollar of
18
advertising are equal to the cost of the advertising. Hence, advertising expenditures
19
can be expected to be a function of expected sales.
20
Several alternate measures of economic activity were investigated in the Standard
21
Mail equations, including personal consumption expenditures, personal disposable
22
income, and retail sales. Various lags of these variables were also investigated. Based
23
on these experiments, current retail sales are included in the Standard Mail equations
24
presented here.
USPS-T-7
82
1
While advertising expenditures track consumption, sales, and income over the long
2
run, in the shorter-run advertising expenditures have a tendency to Attachment A much
3
stronger cyclical pattern than the economy in general. During economic recessions,
4
advertising is more negatively affected than the economy as a whole. This is evident in
5
Table 13, for example, where total advertising expenditures declined by 6.5 percent
6
from 2000 to 2001. In contrast, total retail sales actually rose by 3.1 percent from 2000
7
to 2001. Similarly, advertising expenditures declined by 1.2 percent from 1990 to 1991,
8
while retail sales grew 0.6 percent during that same time period. Conversely, during
9
boom periods, advertising grows more rapidly than overall economic activity. On this
10
other side, advertising expenditures exploded in 2000, increasing by 11.3 percent from
11
1999. Retail sales, on the other hand, grew by only 6.6 percent from 1999 to 2000.
12
Advertising represents a form of business investment. Like advertising, most types
13
of business investment tend to Attachment A very strong cyclical pattern. Hence, in
14
addition to retail sales, the Standard Mail equations also include gross private domestic
15
investment as a measure of the overall demand for business investment.
16
In addition to these macroeconomic factors, the overall level of advertising is also
17
affected by certain other regular events. In particular, in the United States, the election
18
cycle is a key factor which drives advertising demand. In the case of Standard Mail, the
19
election cycle is particularly important with respect to preferred-rate mail, i.e., Standard
20
Nonprofit and Nonprofit Enhanced Carrier Route (ECR) mail. Variables which coincide
21
with the timing of Federal elections are included in the Standard bulk nonprofit demand
22
equation used to forecast Standard Nonprofit and Nonprofit ECR mail in this case.
23
These variables are described in more detail in the discussion of the Standard bulk
24
nonprofit demand equation below.
USPS-T-7
83
1
2
b. Which Advertising Media to Use
The choice of advertising media can be thought of as primarily a pricing decision, so
3
that the demand equation for Standard Mail ought to include the price of direct-mail
4
advertising as well as the prices of alternate advertising media.
5
i. Price of Direct-Mail Advertising
6
The price of direct-mail advertising is decomposed into three distinct prices in this
7
case: postage costs, paper and printing costs, and technological costs. These different
8
types of costs are considered below.
9
(a) Postage Costs
10
Postage costs are included in the Standard Mail equations through fixed-weight
11
price indices which measure the average postage paid by Standard Mailers. Details on
12
the construction and implementation of the price indices used here were discussed at
13
the beginning of this section.
14
15
(b) Paper and Printing Costs
Non-postage costs associated with direct-mail advertising are modeled through the
16
inclusion of the Bureau of Labor Statistics’ producer price index for direct-mail
17
advertising printing. This variable is included in the Standard Enhanced Carrier Route
18
(ECR) demand equation, as well as the Bound Printed Matter demand equation, which
19
is discussed in section E. below. Attempts to include this variable in the econometric
20
equations for Standard Regular and bulk nonprofit mail were less successful, however,
21
due to multicollinearity between this variable and several of the other economic
22
variables included in these equations.
23
24
25
(c) Technological Costs
One of the principal advantages of direct-mail advertising over other forms of
advertising is that direct-mail advertising allows an advertiser to address customers on a
USPS-T-7
84
1
one-on-one basis. By identifying specifically who will receive a particular piece of direct-
2
mail advertising, direct-mail advertising is able to provide an inherent level of targeting
3
that is not necessarily available through other advertising media.
4
The ability to target a direct mailing to specific individuals, based on specific
5
advertiser-chosen criteria, has increased dramatically as a result of technological
6
advances, particularly over the past fifteen to twenty years. The ease with which one is
7
able to identify specific consumers or businesses at whom to target direct-mail
8
advertising is a key component of the cost of direct-mail advertising.
9
This aspect of direct-mail advertising costs, called “technological costs” here, was
10
modeled by Dr. Tolley in past rate cases through the use of a logistic market penetration
11
variable, or “z-variable”. In R97-1, technological costs were modeled through the price
12
of computer equipment. The actual variable used was the implicit price deflator of
13
consumption expenditures on computers and related equipment, as tracked by the
14
Bureau of Economic Analysis. The price of computer equipment has fallen dramatically
15
over time, reflecting the increasing attractiveness of technology over time. In R2000-1,
16
the square of the price of computer equipment was also included in the Standard
17
Regular equation.
18
In R2001-1, the price of computer equipment was replaced by a simple linear time
19
trend. This is done again in this case. A linear time trend is included in the Standard
20
Regular equation. This time trend has a positive coefficient, reflecting the positive
21
influence of targeting described above.
22
Like most economic phenomena, the benefits of improving targeting are likely to be
23
subject to the law of diminishing returns. Hence, it may be reasonable to expect the
24
positive impact of this trend to attenuate over time. Historically, the positive trend in
25
direct-mail advertising has continued to persist as diminishing returns from new
USPS-T-7
85
1
technologies have largely been offset by newer technologies and additional advantages
2
to direct-mail advertising. For example, the recent creation of the National Do Not Call
3
registry has provided direct-mail advertising with a competitive advantage relative to
4
telemarketing.
5
Ultimately, the potential attenuation of this positive trend is a legitimate concern in
6
projecting Standard Mail volumes into the future. To reflect this concern in this case,
7
the time trend in the Standard Regular equation is projected to increase at a slightly
8
slower rate than it has historically through the forecast period presented in this case.
9
The rate of increase of the Standard Regular time trend is attenuated such that the
10
impact of the time trend will fall from its historical level of approximately 3.2 percent per
11
year to a level of 0.0 percent per year over a ten year period.
12
13
The Standard ECR equation also includes a time trend. In this case, the time trend
has a negative coefficient. This trend is discussed in section c. below.
ii. Competing Advertising Media
14
15
Direct-mail advertising competes with other advertising media for a fairly fixed level
16
of total advertising expenditures. Direct-mail advertising is relatively distinct from many
17
other advertising media because of its particular ability to be targeted. To some extent,
18
the degree to which an advertising medium competes with direct-mail advertising is a
19
direct function of the degree to which the advertising medium allows for precise
20
targeting.
USPS-T-7
86
1
In looking carefully at this issue, the two closest substitutes for Standard Mail appear
2
to be newspapers and the Internet. Measures of both of these media are included in
3
the Standard Enhanced Carrier Route equation used in this case. Substitution with
4
newspapers is modeled through a cross-price variable with respect to newspaper
5
advertising. The price of newspaper advertising is taken from the Bureau of Labor
6
Statistics, which reports a producer price index associated with newspaper advertising.
7
The relationship between direct-mail advertising and the Internet is discussed in the
8
next section.
iii. Relationship between the Internet and Direct-Mail Advertising
9
10
The relationship between direct-mail advertising and the Internet is somewhat more
11
complicated than the relationship between direct-mail advertising and, say, newspaper
12
advertising.
13
At one level, the Internet and the mail are competitors for limited advertising dollars.
14
The Interactive Advertising Bureau (IAB) reports total Internet advertising expenditures
15
on a quarterly basis, which are compiled by PricewaterhouseCoopers. This series is
16
divided by total advertising expenditures to calculate the share of total advertising
17
expenditures spent on the Internet. This variable is then included in the Standard
18
Enhanced Carrier Route demand equation used here to reflect this competition.
USPS-T-7
87
1
The measure of Internet advertising expenditures reported by the IAB is not a
2
perfect measure of Internet-based advertising for our purposes here. Internet
3
advertising expenditures here refers primarily to online ads, that is, advertising that
4
appears on the Internet. This excludes much of what is probably the type of Internet-
5
based advertising that is the closest substitute for direct-mail advertising: e-mail
6
advertising. That is, the emergence of electronic mail (e-mail) provides an alternate
7
delivery network through which advertisers can send advertisements to specific
8
customers and potential customers. The IAB’s Internet advertising variable may serve
9
as a useful proxy for the extent of e-mail advertising as these two types of advertising –
10
online advertising and e-mail advertising – may have exhibited similar historical growth
11
trends. Nevertheless, this does limit this variable’s direct usefulness.
12
For now, the Internet is not exclusively a competitive threat to direct-mail advertising.
13
In some ways, the Internet complements direct-mail advertising by providing a network
14
for making catalog purchases, substituting for telephone orders, for example. In the
15
long run, however, the Internet represents a strong potential threat to direct-mail
16
advertising. One could, for example, envision catalogs becoming available online
17
exclusively rather than being distributed through the mail.
18
For the current case, the projected impact of the Internet on Standard Mail volume is
19
measured through the inclusion of Internet advertising expenditures, as measured by
20
the IAB, in the Standard Enhanced Carrier Route equation. Internet advertising
21
expenditures are forecasted to grow through the forecast period in this case at a rate
22
similar to that exhibited over its most recent history. The share of total advertising
23
expenditures spent on Internet advertising is projected to increase from 3.6 percent in
24
2004 to 4.1 percent in 2006. Internet advertising expenditures are forecasted in Section
25
IV of my testimony below.
USPS-T-7
88
1
The potential long-run impact of the Internet on Standard Mail volumes, however, is
2
much more significant than this. Adjusting the Standard Mail forecasts to reflect this
3
potential would be important if the forecast were to span a longer time horizon than is
4
presented here.
5
6
c. How to Send Mail-Based Advertising
Direct-mail advertising can be sent via any of at least six subclasses of mail: First-
7
Class letters, First-Class cards, Standard Regular, Standard ECR, Standard Nonprofit,
8
and Standard Nonprofit ECR. The latter two of these are combined into a single
9
demand equation here, which I refer to as bulk nonprofit mail. Because the price
10
differences between bulk nonprofit and commercial rates are fairly substantial, there is
11
no real price-based substitution between these categories.
12
The Standard Regular demand equation developed here incorporates explicit
13
measures of possible price-based substitution between Standard Regular mail and two
14
types of First-Class Mail: First-Class workshared letters and First-Class cards. Price-
15
based substitution between First-Class workshared letters and Standard Regular letters
16
is modeled through the inclusion of a measure of the average price savings of sending a
17
workshared one-ounce letter as Standard Mail as compared to First-Class Mail.
18
This variable is constructed by calculating price indices which measure the average
19
price of a one-ounce First-Class workshared letter and a one-ounce Standard Regular
20
letter. Both of these price indices are constructed using 2004 billing determinants for
21
these respective categories of mail. The difference between these two price indices is
22
then included as an explanatory variable in the First-Class workshared letters and
23
Standard Regular demand equations presented in this testimony.
24
25
In R2001-1, for example, the nominal discount for Standard Regular letters relative
to First-Class workshared letters measured in this way increased from 9.36 cents to
USPS-T-7
89
1
10.51 cents, a 12.3 percent increase. This increase is estimated to have led to a shift of
2
approximately 100 million letters from First-Class to Standard Mail.
3
The discount elasticity with respect to First-Class workshared letters is freely
4
estimated in the Standard Regular demand equation. The elasticity is then
5
stochastically constrained in the First-Class workshared letters equation based on the
6
results from the Standard Regular equation using the Slutsky-Schultz symmetry
7
condition. The Slutsky-Schultz symmetry condition is described in detail in Section III
8
below.
9
Substitution between Standard Regular and First-Class cards is modeled through a
10
variable which measures the percentage of Standard Regular mail for which First-Class
11
card rates are less expensive as well as a time trend from 1988 through 1990 reflecting
12
the lagged initial reaction to this rate relationship. These variables are described in
13
more detail in the discussion of First-Class cards above. The coefficients on these
14
variables are freely estimated in the First-Class cards equation. These coefficients are
15
then constrained in the Standard Regular equation based upon the results from the
16
First-Class cards equation and the Slutsky-Schultz symmetry condition.
17
In addition, substitution between Standard Regular and ECR mail is modeled
18
through two aspects of the econometric demand equations presented here. First, in
19
R97-1 (1999Q2), some Standard Regular mail (Automation 5-digit letters) was priced
20
below some Standard ECR mail (basic letters). This caused some Standard ECR mail
21
to be sent as Standard Regular mail instead. This event is modeled by the inclusion of
22
a dummy variable equal to one starting with the implementation of R97-1 rates in
23
January, 1999 in the Standard Regular as well as the Standard ECR equation.
24
25
In addition, Standard Regular mail volume has generally grown more rapidly than
Standard ECR volume for at least the past decade. One possible reason for this is
USPS-T-7
90
1
because of the technological advances that have benefited the use of mail as an
2
advertising medium that were described above. Specifically, the ability to target
3
individual customers, based on individual characteristics of the specific customer has
4
increased tremendously over recent years. The general trend in database marketing
5
has been away from targeting based on broad demographic characteristics, and toward
6
targeting based on an individual’s past history of actual purchases. This has led to a
7
less geographically dense set of target customers. By targeting individuals rather than
8
neighborhoods, then, advertisers are less likely to have enough density within carrier
9
routes to qualify for Standard ECR rates. This is measured in the Standard ECR
10
11
equation by the inclusion of a negative time trend.
This loss of Standard ECR volume has not necessarily hurt the Postal Service, as
12
this loss has largely been to the benefit of Standard Regular mail volume. The gains to
13
Standard Regular mail volume because of shifts from ECR to Regular rate are
14
embedded within the positive time trend which is included in the Standard Regular
15
demand equation and was discussed briefly above.
16
17
18
19
The specific demand equations developed for Standard Mail volumes in this case
are outlined next.
3. Final Equation Specifications for Standard Mail
a. Overview of Standard Mail Subclasses
20
Standard Mail is divided into four subclasses: Regular, Enhanced Carrier Route,
21
Nonprofit, and Nonprofit ECR. The latter two of these – Nonprofit and Nonprofit ECR –
22
are estimated within a single demand equation here.
23
Table 14 presents volumes for the four subclasses of Standard Mail from 1970
24
through 2004. Table 15 presents the percentage change in Standard Mail volumes
25
since 1971.
USPS-T-7
91
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
Commercial
Regular
14,963.843
15,528.332
16,529.452
16,700.081
16,317.566
15,455.601
15,797.140
16,748.263
18,554.479
16,723.242
14,835.765
14,092.412
13,653.396
14,430.525
16,272.368
17,544.415
19,721.447
21,707.301
22,425.936
21,954.025
23,878.085
22,920.692
24,104.548
25,918.411
27,520.957
29,260.223
30,287.719
32,179.198
34,777.134
38,490.810
43,030.853
44,699.352
43,552.691
46,639.790
50,776.236
Ta ble 14
Sta nda rd Ma il Volume
(millions of pieces)
Subclasses
Preferred Subclasses
ECR
Nonprofit Nonprofit ECR
0.000
4,200.122
0.000
0.000
4,408.746
0.000
0.000
4,649.455
0.000
0.000
5,194.870
0.000
0.000
5,476.533
0.000
0.000
5,564.876
0.000
0.000
5,973.001
0.000
0.000
6,513.447
0.000
0.000
7,167.851
0.000
3,045.295
7,468.517
0.000
6,999.788
7,810.128
93.503
10,551.129
7,917.974
608.213
13,693.853
7,904.931
1,123.115
16,640.131
8,017.065
1,326.232
21,344.765
8,727.633
1,624.243
23,305.208
9,005.323
1,927.156
24,067.531
8,712.661
2,125.565
26,598.539
8,596.633
2,340.983
28,979.527
8,919.036
2,234.108
28,656.609
9,219.354
2,616.149
27,572.830
9,359.714
2,661.048
27,254.513
9,185.687
2,747.552
25,973.416
8,983.433
2,934.036
27,832.932
8,939.330
2,860.983
29,878.546
8,903.971
2,908.486
30,155.479
9,340.052
3,003.086
29,369.180
9,398.197
2,874.515
31,268.167
10,000.852
2,880.172
33,848.366
10,551.254
2,649.059
32,769.071
10,933.949
2,940.701
32,776.017
11,325.657
2,924.638
30,940.526
11,275.136
3,081.870
29,671.452
11,310.268
2,696.226
29,324.722
11,549.738
2,977.984
30,345.448
11,791.584
2,650.253
note: Data show n are f or Postal Fiscal Y ears through 1999, by Government Fiscal Y ears 2000 - 2004
Combined
4,200.122
4,408.746
4,649.455
5,194.870
5,476.533
5,564.876
5,973.001
6,513.447
7,167.851
7,468.517
7,903.631
8,526.187
9,028.046
9,343.297
10,351.876
10,932.479
10,838.226
10,937.616
11,153.144
11,835.503
12,020.762
11,933.239
11,917.469
11,800.313
11,812.457
12,343.138
12,272.712
12,881.024
13,200.313
13,874.650
14,250.295
14,357.006
14,006.494
14,527.723
14,441.837
USPS-T-7
92
Ta ble 15
Pe rce nta nge Cha nge in Sta nda rd Ma il Volume
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
2
Commercial Subclasses
Regular
ECR
3.77%
6.45%
1.03%
-2.29%
-5.28%
2.21%
6.02%
10.78%
-9.87%
-11.29%
129.86%
-5.01%
50.73%
-3.12%
29.79%
5.69%
21.52%
12.76%
28.27%
7.82%
9.18%
12.41%
3.27%
10.07%
10.52%
3.31%
8.95%
-2.10%
-1.11%
8.76%
-3.78%
-4.01%
-1.15%
5.17%
-4.70%
7.52%
7.16%
6.18%
7.35%
6.32%
0.93%
3.51%
-2.61%
6.25%
6.47%
8.07%
8.25%
10.68%
-3.19%
10.35%
-0.70%
3.88%
-5.60%
-2.57%
-4.10%
7.09%
-1.17%
8.87%
3.48%
Preferred Subclasses
Nonprofit Nonprofit ECR
4.97%
5.46%
11.73%
5.42%
1.61%
7.33%
9.05%
10.05%
4.19%
4.57%
1.38%
550.47%
-0.16%
84.66%
1.42%
18.09%
8.86%
22.47%
3.18%
18.65%
-3.25%
10.30%
-1.33%
10.13%
3.75%
-4.57%
3.37%
17.10%
1.52%
1.72%
-1.86%
3.25%
-2.20%
6.79%
-0.49%
-2.49%
-0.40%
1.66%
4.90%
3.25%
0.62%
-4.28%
6.41%
0.20%
5.50%
-8.02%
3.63%
11.01%
3.24%
-1.79%
-0.45%
5.38%
0.31%
-12.51%
2.12%
10.45%
2.09%
-11.01%
Combined
4.97%
5.46%
11.73%
5.42%
1.61%
7.33%
9.05%
10.05%
4.19%
5.83%
7.88%
5.89%
3.49%
10.79%
5.61%
-0.86%
0.92%
1.97%
6.12%
1.57%
-0.73%
-0.13%
-0.98%
0.10%
4.49%
-0.57%
4.96%
2.48%
5.11%
2.17%
0.75%
-2.44%
3.72%
-0.59%
note: Data show n are f or Postal Fiscal Y ears through 2000, by Government Fiscal Y ears 2001 - 2004
Standard Mail volume experienced considerable growth through the 1980s. From
3
1980 through 1988, total Standard Mail volume more than doubled, increasing at an
4
average annual rate of 9.7 percent. Since that time, the growth rate of Standard Mail
5
volume has slowed considerably. In fact, Standard Mail volume has not grown by more
USPS-T-7
93
1
than 7.2 percent in any year since 1988. The Standard Mail demand equations used in
2
this case are estimated over a sample period which begins in 1988Q1.
3
4
5
6
The three demand equations used to forecast Standard Mail volumes are described
below.
b. Standard Regular Mail
i. Factors Affecting Standard Regular Mail Volume
7
Standard Regular mail volume consists of Standard Mail volume that does not
8
qualify for preferred non-profit rates and is not geographically dense enough to qualify
9
for the Enhanced Carrier Route (ECR) subclass.
10
The factors underlying the demand for Standard Regular mail volume are basically
11
those outlined above. A common problem in empirical econometric work is the
12
tendency of many potential explanatory variables to be highly correlated with each
13
other, making it difficult to isolate the unique impact of each of these variables. In this
14
case, the prices of newspaper and direct-mail advertising are both highly correlated to
15
more general macroeconomic variables such as retail sales and investment.
16
As a result, neither the price of newspaper nor direct-mail advertising was found to
17
help to explain Standard Regular mail volume. Hence, the Standard Regular demand
18
equation used in this case does not include any explicit measures of substitution with
19
other advertising media. To some extent, this is reflective of the relative uniqueness of
20
Standard Regular mail as a means of directly targeting specific customers.
21
The effect of the Internet on Standard Regular mail volume is somewhat ambiguous.
22
As noted above, Internet advertising expenditures represent direct competition with
23
direct-mail advertising for limited advertising dollars. As such, one might expect
24
Standard Mail volumes to be negatively related to Internet advertising expenditures. In
USPS-T-7
94
1
fact, one finds this very relationship in the case of Standard Enhanced Carrier Route
2
mail.
3
At present, however, the Internet is at least somewhat complementary to direct-mail
4
advertising. Some evidence exists, for example, to suggest that direct-mail catalogs
5
increase traffic and sales on retailer Web sites. The econometric evidence suggests
6
that this complementarity between direct-mail and Internet advertising has offset the
7
substitution between these media, so that Internet advertising expenditures have not
8
been found to have affected Standard Regular mail volumes at least historically. This is
9
assumed to continue to be the case through the forecast period presented here. It will
10
be important moving forward, however, to carefully re-evaluate this assumption as new
11
data become available and as advertisers’ use of the Internet continues to evolve.
12
13
14
15
16
17
18
19
Overall, then, Standard Regular mail volume was found to be primarily affected by
the following variables:
Retail Sales
Investment
Time Trends
Prices of First-Class Letters and Cards
Price of Standard Regular Mail
The effect of these variables on Standard Regular Mail volume over the past ten
20
years is shown in Table 16 on the next page. Table 16 also shows the projected
21
impacts of these variables through GFY 2007.
22
The Test Year before-rates volume forecast for Standard Regular mail is 56,985.773
23
million pieces, a 12.2 percent increase from GFY 2004. The Postal Service’s proposed
24
rates in this case are predicted to reduce the Test Year volume of Standard Regular
25
mail by 0.9 percent, for a Test Year after-rates volume forecast for Standard Regular
26
mail of 56,478.638 million.
USPS-T-7
95
Other
1.69%
-0.05%
-0.02%
0.13%
-0.42%
-0.34%
0.90%
-1.98%
1.74%
0.61%
Total Change
in Volum e
6.32%
3.51%
6.25%
8.07%
10.68%
11.80%
3.88%
-2.57%
7.09%
8.87%
0.71%
0.07%
2.22%
0.22%
84.50%
6.32%
1.00%
0.33%
-0.23%
-0.08%
0.33%
0.11%
13.60%
4.34%
0.00%
0.00%
0.00%
0.38%
0.32%
0.36%
-0.21%
0.02%
0.50%
-0.83%
0.25%
0.00%
6.25%
5.62%
5.38%
0.00%
0.00%
0.00%
0.00%
1.07%
0.35%
0.30%
0.10%
-0.58%
-0.19%
18.26%
5.75%
0.00%
-1.29%
-0.12%
0.00%
0.00%
0.00%
0.00%
0.40%
0.00%
0.38%
0.32%
0.36%
-0.21%
0.02%
0.50%
-0.83%
0.25%
0.00%
6.25%
4.68%
5.25%
-1.41%
-0.47%
0.00%
0.00%
0.40%
0.13%
1.07%
0.35%
0.30%
0.10%
-0.58%
-0.19%
17.07%
5.39%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.24%
1.14%
1.22%
1.20%
1.22%
1.43%
1.26%
1.29%
1.34%
1.25%
Retail Sales
0.32%
0.19%
0.23%
0.22%
0.50%
0.49%
-0.06%
0.04%
0.08%
0.40%
1994 - 2004
Total
Avg per Year
13.33%
1.26%
2.45%
0.24%
9.90%
0.95%
37.53%
3.24%
-9.06%
-0.95%
6.94%
0.67%
1.36%
0.14%
3.61%
0.36%
2001 - 2004
Total
Avg per Year
3.93%
1.29%
0.52%
0.17%
-0.17%
-0.06%
9.98%
3.22%
-3.00%
-1.01%
0.00%
0.00%
0.98%
0.33%
Forecas t
1.25%
1.14%
1.13%
0.36%
0.09%
0.12%
1.97%
0.58%
0.34%
3.21%
3.14%
2.84%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
3.57%
1.17%
0.57%
0.19%
2.91%
0.96%
9.48%
3.06%
0.00%
0.00%
After-Rates Volum e Forecas t
2005
1.25%
2006
1.14%
2007
1.13%
0.36%
0.09%
0.12%
1.97%
0.58%
0.34%
3.21%
3.14%
2.84%
0.57%
0.19%
2.91%
0.96%
9.48%
3.06%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 16
Estimated Impact of Factors Affecting Standard Regular Mail Volume, 1994 – 2007
Pos tal Prices
Other Factors
Inves tm ent
Trends
Std Regular
Std ECR
Firs t-Clas s
Inflation Econom etric
2.04%
3.24%
-2.05%
0.00%
-0.05%
0.44%
-0.62%
-0.01%
3.21%
-1.71%
0.00%
0.02%
0.37%
0.35%
2.63%
3.19%
-0.74%
0.00%
0.21%
0.38%
-0.93%
2.58%
3.20%
0.11%
0.00%
0.00%
0.21%
0.21%
1.22%
3.24%
-0.28%
4.61%
0.32%
0.22%
-0.33%
1.65%
3.36%
-0.20%
2.22%
0.16%
0.48%
2.01%
-0.39%
3.28%
-1.54%
0.00%
-0.28%
0.44%
0.27%
-2.31%
3.18%
-1.42%
0.00%
0.21%
0.28%
-1.76%
0.00%
3.28%
-1.60%
0.00%
0.77%
0.36%
0.98%
2.19%
3.21%
0.00%
0.00%
0.00%
0.36%
0.57%
2004 - 2007
Total
Avg per Year
3.57%
1.17%
USPS-T-7
96
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 16 above.
5
Standard Regular mail volume has an elasticity with respect to retail sales of 0.104
6
(t-statistic of 1.243), meaning that a 10 percent increase in retail sales will lead to a 1.04
7
percent increase in the volume of Standard Regular Mail. Standard Regular mail
8
volume also has an elasticity with respect to investment of 0.228 (t-statistic of 5.893).
9
Taking these two variables together, the economy contributed about 2.0 percent per
10
year Standard Regular volume growth from 1994 through 2000. In 2001 and 2002,
11
Standard Regular volume declined by 2.7 percent because of the recession. The
12
economy is projected to add approximately 1.2 percent per year to Standard Regular
13
mail volume through the forecast period shown in Table 16.
14
The time trend in the Standard Regular demand equation has added approximately
15
3.2 percent per year to Standard Regular mail volume historically. This effect is
16
expected to diminish slightly through the forecast period shown in Table 16. As noted
17
above, some of this long-run gain in Standard Regular mail volume is a shift from
18
Enhanced Carrier Route mail due to increased targeting. As shown in Table 18 below,
19
the long-run time trend in the Standard ECR equation explains a decline in Standard
20
ECR mail volume of approximately 4.1 percent per year. Taken together, these trends
21
are expected to lead to an overall increase in Standard commercial volumes of 0.4
22
percent per year for the forecast period used in this case.
23
The own-price elasticity of Standard Regular mail was calculated to be equal to
24
-0.267 (t−statistic of -3.521). This is considerably lower than the own-price elasticity of
USPS-T-7
97
1
Standard ECR mail (-1.093), reflecting the fairly solid niche which Standard Regular
2
mail volume has carved out for itself within the advertising market.
3
The Postal price impacts shown in Table 16 above are the result of changes in
4
nominal prices. Prices enter the demand equations developed here in real terms,
5
however. The impact of inflation reported in Table 16 measures the impact that a
6
change in real Postal prices, in the absence of nominal rate changes, has on the
7
volume of Standard Regular mail.
8
Other econometric variables include seasonal variables and a dummy variable to
9
account for the temporary impact of the September 11, 2001, terrorist attacks. A more
10
detailed look at the econometric demand equation for Standard Regular mail follows.
ii. Econometric Demand Equation
11
12
The demand equation for Standard Regular Mail in this case models Standard
13
Regular Mail volume per adult per delivery day as a function of the following explanatory
14
variables:
15
16
17
18
19
20
21
22
23
24
25
26
·
Seasonal Variables
·
Retail Sales
·
Real Gross Private Domestic Investment (lagged one quarter)
·
Linear Time Trend
·
Dummy variable equal to one since the implementation of MC95-1 (1996Q4)
·
Dummy variable for September 11th, equal to one in 2002Q1, zero elsewhere
USPS-T-7
98
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
·
Difference in Price between a one-ounce First-Class workshared Letter and a
one-ounce Standard Regular letter
·
Percentage of Standard Regular mail for which First-Class cards rates are
lower than Standard Regular rates
·
Crossover Trend, which is a simple time trend beginning in 1988Q4 and
plateauing in 1990Q4 to reflect lagged reaction by mailers to the initial R87-1
rate crossover
The coefficients on the two preceding variables are constrained from the First-Class
cards equation.
·
Dummy variable equal to one since the implementation of R97-1 (1999Q2)
which set Standard Regular automation 5-digit letter rates below Standard
ECR basic letter rates
·
Current and one lag of the price of Standard Regular Mail
Details of the econometric demand equation are shown in Table 17 below. A
20
detailed description of the econometric methodologies used to obtain these results can
21
be found in Section III below.
USPS-T-7
99
1
2
3
TABLE 17
ECONOMETRIC DEMAND EQUATION FOR STANDARD REGULAR MAIL
Coefficient
T-Statistic
Own-Price Elasticity
Long-Run
-0.267
-3.521
Current
-0.176
-1.957
Lag 1
-0.091
-1.251
Avg. Standard Regular Letters Discount
0.075
0.942
(relative to First-Class)
Pct of First-Class Cards Rates less than
-0.0086
(N/A)
Standard Regular
Time Trend Interacted with Std Regular Rate
-0.0030
(N/A)
Crossover (1988Q4 – 1990Q4)
Retail Sales
0.104
1.243
Total Private Investment
0.228
5.893
Time Trend
0.0080
17.33
Dummy for MC95-1
-0.053
-3.868
Dummy for R97-1 (Rate Crossover with ECR
0.072
8.014
Mail)
Dummy for 2002Q1 (9/11 Effect)
-0.044
-2.303
Seasonal Coefficients
1.209
6.285
September
0.803
5.849
October 1 – December 10
1.017
3.835
December 11 – 15
-0.327
-0.526
December 16 – 17
9.859
3.848
December 18 – 24
-15.51
-5.089
December 25 – 31
1.373
7.193
January – March
-1.259
-3.661
April 1 – 15
1.115
6.702
April 16 – May
2.016
4.745
June
Quarter 1 (October – December)
0.421
6.593
Quarter 2 (January – March)
-0.556
-6.846
Quarter 3 (April – June)
-0.240
-2.726
Quarter 4 (July – September)
0.375
4.115
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.042293
Quarter 2 (January – March)
1.015593
Quarter 3 (April – June)
0.977137
Quarter 4 (July – September)
0.966481
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
AR-4: -0.443
Degrees of Freedom
41
Mean-Squared Error
0.000239
Adjusted R-Squared
0.995
4
5
USPS-T-7
100
1
2
3
c. Standard Enhanced Carrier Route (ECR)
i. Factors Affecting Standard ECR Mail Volume
Standard ECR mail volume consists of Standard Mail volume that does not qualify
4
for preferred non-profit rates and consists of at least ten pieces being sent to each
5
carrier route in a mailing.
6
The factors underlying the demand for Standard ECR mail volume are those outlined
7
above. To review, Standard ECR mail volume was found to be primarily affected by the
8
following variables:
9
10
11
12
13
14
15
16
Retail Sales
Investment
Price of Newspaper Advertising
Price of Direct-Mail Advertising
Internet Advertising Expenditures
Time Trend
Price of Standard ECR Mail
The effect of these variables on Standard ECR Mail volume over the past ten years
17
is shown in Table 18 on the next page. Table 18 also shows the projected impacts of
18
these variables through GFY 2007.
19
The Test Year before-rates volume forecast for Standard ECR mail is 33,328.906
20
million pieces, a 9.8 percent increase from GFY 2004. The Postal Service’s proposed
21
rates in this case are predicted to reduce the Test Year volume of Standard ECR mail
22
by 3.4 percent, for a Test Year after-rates volume forecast for Standard ECR mail of
23
32,187.100 million.
USPS-T-7
101
1
Other
-1.29%
-1.56%
-0.29%
1.67%
-1.58%
1.13%
-0.83%
-0.21%
-0.13%
-1.22%
Total Change
in Volum e
0.93%
-2.61%
6.47%
8.25%
-3.19%
0.02%
-5.60%
-4.10%
-1.17%
3.48%
0.36%
0.04%
-4.27%
-0.44%
1.56%
0.16%
6.06%
1.98%
0.75%
0.25%
-1.55%
-0.52%
-1.92%
-0.65%
0.00%
0.00%
0.00%
2.26%
1.94%
2.01%
-0.99%
-0.12%
-1.14%
1.92%
-0.31%
0.00%
6.43%
3.20%
1.94%
1.13%
0.38%
0.00%
0.00%
6.33%
2.07%
-2.23%
-0.75%
1.61%
0.53%
11.96%
3.84%
1.31%
3.70%
3.20%
0.83%
0.02%
0.29%
0.00%
-3.43%
-2.40%
2.26%
1.94%
2.01%
-0.99%
-0.12%
-1.14%
1.92%
-0.31%
0.00%
6.43%
-0.34%
-0.51%
8.42%
2.73%
1.13%
0.38%
-5.74%
-1.95%
6.33%
2.07%
-2.23%
-0.75%
1.61%
0.53%
5.53%
1.81%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.21%
1.11%
1.22%
1.20%
1.19%
1.34%
1.20%
1.28%
1.30%
1.22%
Retail Sales
1.35%
0.81%
1.00%
0.96%
2.15%
2.08%
-0.21%
0.19%
0.29%
1.69%
1994 - 2004
Total
Avg per Year
12.96%
1.23%
10.78%
1.03%
10.06%
0.96%
-33.89%
-4.05%
-2.44%
-0.25%
50.10%
4.14%
4.62%
0.45%
-38.26%
-4.71%
22.75%
2.07%
2001 - 2004
Total
Avg per Year
3.85%
1.27%
2.17%
0.72%
-1.48%
-0.49%
-11.71%
-4.07%
-0.19%
-0.06%
11.68%
3.75%
2.59%
0.86%
-11.67%
-4.05%
Forecas t
1.25%
1.13%
1.11%
1.62%
0.39%
0.51%
2.56%
0.82%
0.29%
-4.09%
-4.12%
-4.11%
-0.24%
-0.12%
-0.06%
1.31%
3.70%
3.20%
0.83%
0.02%
0.29%
3.53%
1.16%
2.54%
0.84%
3.69%
1.22%
-11.81%
-4.10%
-0.42%
-0.14%
8.42%
2.73%
After-Rates Volum e Forecas t
2005
1.25%
2006
1.13%
2007
1.11%
1.62%
0.39%
0.51%
2.56%
0.82%
0.29%
-4.09%
-4.12%
-4.11%
-0.24%
-0.12%
-0.06%
2.54%
0.84%
3.69%
1.22%
-11.81%
-4.10%
-0.42%
-0.14%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
2
Table 18
Estimated Impact of Factors Affecting Standard Enhanced Carrier Route Mail Volume, 1994 – 2007
Price of Price of Direct-Mail Advertis ing
Other Factors
Inves tm ent
Trends
Internet News paper Adv.
Printing
Pos tage
Inflation Econom etric
2.68%
-4.04%
0.00%
3.85%
0.24%
-5.05%
2.41%
-0.06%
-0.06%
-4.02%
-0.05%
7.90%
-0.24%
-7.68%
2.27%
-0.41%
2.19%
-4.09%
-0.18%
4.37%
-0.33%
-1.29%
2.24%
1.69%
2.78%
-4.10%
-0.29%
4.10%
0.85%
-0.09%
1.52%
-0.42%
1.54%
-4.02%
-0.51%
3.42%
0.38%
-7.38%
1.25%
0.81%
1.31%
-4.04%
-1.23%
2.71%
0.34%
-4.45%
2.36%
-1.22%
0.74%
-4.03%
-0.01%
3.93%
0.72%
-8.65%
2.73%
-0.75%
-2.52%
-4.08%
0.45%
4.52%
0.08%
-6.14%
1.95%
0.74%
-0.35%
-4.10%
-0.10%
3.38%
1.14%
-4.46%
1.93%
0.21%
1.43%
-4.02%
-0.53%
3.35%
1.35%
-1.49%
2.07%
-0.20%
2004 - 2007
Total
Avg per Year
3.53%
1.16%
USPS-T-7
102
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 18 above.
5
Standard ECR mail volume has an elasticity with respect to retail sales of 0.454 (t-
6
statistic of 2.305), meaning that a 10 percent increase in retail sales will lead to a 4.54
7
percent increase in the volume of Standard ECR Mail. Standard ECR mail volume also
8
has an elasticity with respect to investment of 0.233 (t-statistic of 2.820). Taking these
9
two variables together, the economy contributed about 3.2 percent per year to Standard
10
ECR volume growth from 1994 through 2000. From 2001 through 2003, Standard ECR
11
volume declined by 2.4 percent because of the recession. The economy is projected to
12
add approximately 2.1 percent per year to Standard ECR mail volume through the
13
forecast period shown in Table 18.
14
The prices of newspaper and direct-mail advertising have added an additional 4.6
15
percent per year to Standard ECR volume over the past ten years, and are projected to
16
add approximately 3.1 percent per year through the forecast period.
17
18
The time trend in the Standard ECR demand equation explains a decline in
Standard ECR mail volume of approximately 4.1 percent per year.
19
The own-price elasticity of Standard ECR mail is calculated to be equal to -1.093
20
(t−statistic of -4.973). This is among the highest own-price elasticities estimated in my
21
testimony, reflecting the competitiveness of the advertising market and the extent to
22
which relatively close substitutes exist for Standard ECR mail. In addition to the price of
23
Standard ECR mail, dummy variables are also included for the implementation of the
24
R97-1 and R2000-1 rate cases which caused some mail to migrate from the Standard
25
ECR subclass to the Regular subclass. The combined effects of these things have
USPS-T-7
103
1
combined to explain a 38.3 percent decline in Standard ECR mail volume over the past
2
ten years.
3
The Postal price impacts shown in Table 18 above are the result of changes in
4
nominal prices. Prices enter the demand equations developed here in real terms,
5
however. The impact of inflation reported in Table 18 measures the impact that a
6
change in real Postal prices, in the absence of nominal rate changes, has on the
7
volume of Standard ECR mail. The impact of inflation, in this case, has been to
8
increase Standard ECR mail volume by 22.8 percent over the past ten years.
9
10
Other econometric variables include seasonal variables. A more detailed look at the
econometric demand equation for Standard ECR mail follows.
ii. Econometric Demand Equation
11
12
The demand equation for Standard Enhanced Carrier Route Mail in this case models
13
Standard ECR Mail volume per adult per delivery day as a function of the following
14
explanatory variables:
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
·
Seasonal Variables
·
Retail Sales
·
Real Gross Private Domestic Investment (lagged two quarters)
·
Producer Price Index for Newspaper Advertising
·
Producer Price Index for Direct-Mail Advertising (lagged four quarters)
·
Linear Time Trend
·
Internet Advertising Expenditures as a Share of Total Advertising
Expenditures
·
Dummy variable equal to one since the implementation of R97-1 (1999Q2)
which set Standard Regular automation 5-digit letter rates below Standard
ECR basic letter rates
USPS-T-7
104
1
2
3
4
·
Dummy variable equal to one since the implementation of R2000-1 (2001Q2)
·
Current and four lags of the price of Standard ECR Mail
5
Details of the econometric demand equation are shown in Table 19 below. A detailed
6
description of the econometric methodologies used to obtain these results can be found
7
in Section III below.
USPS-T-7
105
1
2
3
4
TABLE 19
ECONOMETRIC DEMAND EQUATION FOR
STANDARD ENHANCED CARRIER ROUTE (ECR) MAIL
Coefficient
T-Statistic
Own-Price Elasticity
Long-Run
-1.093
-4.973
Current
-0.463
-2.548
Lag 1
-0.123
-0.694
Lag 2
-0.128
-0.691
Lag 3
-0.121
-0.647
Lag 4
-0.258
-1.898
Retail Sales
0.454
2.305
Total Private Investment
0.233
2.820
Price of Newspaper Advertising
1.353
2.300
Price of Direct-Mail Advertising
-0.483
-1.437
Time Trend
-0.0104
-2.492
Internet Advertising Expenditures
-0.689
-0.641
(as share of total advertising exp.)
Dummies for Rate Crossover w/ Std Regular
R97-1
-0.114
-4.382
R2000-1
-0.090
-5.542
Seasonal Coefficients
0.073
0.280
September 16 – 30
1.024
4.436
October
-0.483
-2.517
November 1 – December 15
2.239
2.758
December 16 – 17
-1.098
-0.514
December 18 – 24
6.350
1.317
December 25 – 31
-0.537
-1.181
January – February
0.021
0.100
March
0.103
0.192
April 1 – 15
-0.077
-0.405
April 16 – June
Quarter 1 (October – December)
-0.342
-1.198
Quarter 2 (January – March)
0.340
1.169
Quarter 3 (April – June)
0.002
0.045
Quarter 4 (July – September)
-0.000
-0.010
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.130268
Quarter 2 (January – March)
0.972741
Quarter 3 (April – June)
0.924201
Quarter 4 (July – September)
0.978039
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.187
Degrees of Freedom
40
Mean-Squared Error
0.000464
Adjusted R-Squared
0.953
5
USPS-T-7
106
1
d. Standard Bulk Nonprofit
i. Factors Affecting Standard Bulk Nonprofit Mail Volume
2
3
The Postal Service offers preferred rates to not-for-profit organizations sending
4
Standard Mail volumes. There are two subclasses of preferred-rate Standard Mail
5
which exactly parallel the two commercial subclasses discussed above: Nonprofit and
6
Nonprofit Enhanced Carrier Route (Nonprofit ECR). These two subclasses are
7
combined and a single demand equation is estimated for this type of mail. I refer to the
8
combination of the Standard Nonprofit and Nonprofit ECR subclasses as Standard bulk
9
nonprofit mail.
10
The demand for Standard bulk nonprofit mail volume will be affected by some of the
11
same factors that drive the demand for Standard Regular and ECR mail volumes, as
12
outlined above. Because these mailers are not-for-profit, however, the decision of
13
whether to send bulk nonprofit mail volume is not necessarily a business investment
14
decision per se. Hence, investment is not included in the Standard bulk nonprofit
15
equation used here. Also, because of the preferred rates offered by the Postal Service,
16
there is relatively little price-based competition between the mail and other media for
17
this type of advertising. Hence, no measures of competition with other media are
18
included in the Standard bulk nonprofit equation.
19
20
21
Thus, Standard bulk nonprofit mail volume was found to be primarily affected by
simply retail sales and the price of Standard bulk nonprofit mail.
The effect of these variables on Standard bulk nonprofit mail volume over the past
22
ten years is shown in Table 20 on the next page. Table 20 also shows the projected
23
impacts of these variables through GFY 2007.
24
25
The Test Year before-rates volume forecast for Standard bulk nonprofit mail is
15,502.729 million pieces, a 7.3 percent increase from GFY 2004. The Postal Service’s
USPS-T-7
107
1
proposed rates in this case are predicted to reduce the Test Year volume of Standard
2
bulk nonprofit mail by 0.5 percent, for a Test Year after-rates volume forecast for
3
Standard bulk nonprofit mail of 15,418.326 million.
USPS-T-7
108
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Table 20
Estimated Impact of Factors Affecting Standard Bulk Nonprofit Mail Volume, 1994 – 2007
Federal
Total Change
Other Factors
Inflation
Elections Econom etric
Other
in Volum e
Population Retail Sales Pos tage Price
1.22%
1.24%
-0.83%
0.67%
1.05%
0.49%
0.59%
4.49%
1.11%
0.78%
-1.69%
0.67%
-0.63%
0.20%
-0.99%
-0.57%
1.22%
0.96%
0.24%
0.66%
0.77%
-0.25%
1.27%
4.96%
1.16%
0.89%
2.84%
0.52%
-1.02%
-0.20%
-1.67%
2.48%
1.21%
2.09%
-1.14%
0.33%
1.78%
-0.55%
1.32%
5.11%
1.34%
1.96%
-2.40%
0.55%
0.46%
0.44%
0.38%
2.71%
1.24%
-0.21%
-0.96%
0.82%
-2.05%
0.69%
1.25%
0.75%
1.27%
0.17%
-1.70%
0.62%
-0.98%
-0.64%
-1.18%
-2.44%
1.33%
0.29%
-1.37%
0.51%
-0.84%
1.73%
2.06%
3.72%
1.19%
1.54%
-0.72%
0.59%
-0.15%
1.30%
-4.23%
-0.59%
1994 - 2004
Total
Avg per Year
13.00%
1.23%
10.11%
0.97%
-7.55%
-0.78%
6.11%
0.59%
-1.64%
-0.17%
3.23%
0.32%
-1.35%
-0.14%
22.26%
2.03%
2001 - 2004
Total
Avg per Year
3.84%
1.26%
2.00%
0.66%
-3.74%
-1.26%
1.73%
0.57%
-1.95%
-0.65%
2.40%
0.79%
-3.40%
-1.15%
0.59%
0.20%
Forecas t
1.26%
1.11%
1.13%
1.55%
0.37%
0.49%
0.00%
0.00%
0.00%
0.67%
0.59%
0.57%
-1.35%
0.54%
-0.71%
-0.23%
-0.19%
0.91%
3.56%
-0.68%
0.00%
5.51%
1.74%
2.40%
3.54%
1.17%
2.43%
0.80%
0.00%
0.00%
1.84%
0.61%
-1.53%
-0.51%
0.49%
0.16%
2.85%
0.94%
9.92%
3.20%
After-Rates Volum e Forecas t
2005
1.26%
2006
1.11%
2007
1.13%
1.55%
0.37%
0.49%
0.00%
-0.54%
-1.17%
0.67%
0.59%
0.57%
-1.35%
0.54%
-0.71%
-0.23%
-0.19%
0.91%
3.56%
-0.68%
0.00%
5.51%
1.19%
1.20%
2.43%
0.80%
-1.71%
-0.57%
1.84%
0.61%
-1.53%
-0.51%
0.49%
0.16%
2.85%
0.94%
8.04%
2.61%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
2004 - 2007
Total
Avg per Year
3.54%
1.17%
USPS-T-7
109
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 20 above.
5
Standard bulk nonprofit mail volume has an elasticity with respect to retail sales of
6
0.430 (t-statistic of 14.84), meaning that a 10 percent increase in retail sales will lead to
7
a 4.30 percent increase in the volume of Standard bulk nonprofit mail.
8
9
The own-price elasticity of Standard bulk nonprofit mail is calculated to be equal to
-0.319 (t−statistic of -7.801). The Postal price impacts shown in Table 20 are the result
10
of changes in nominal prices. Prices enter the demand equations developed here in
11
real terms, however. The impact of inflation reported in Table 20 measures the impact
12
that a change in real Postal prices, in the absence of nominal rate changes, has on the
13
volume of Standard bulk nonprofit mail.
14
One additional factor appears in Table 20 which affects Standard bulk nonprofit that
15
is not used in the demand equations for Standard Regular and Enhanced Carrier Route
16
mail. This is the impact of Federal elections on Standard bulk nonprofit mail volume.
17
Standard bulk nonprofit mail volume exhibits a fairly pronounced two-year seasonal
18
pattern, whereby this type of mail tends to be greater during Federal election years (i.e,
19
even-numbered years) than during non-election years.
20
Many groups with particular interests in political issues are eligible for preferred rates
21
from the Postal Service. The volume of advertising and solicitation from such groups
22
increases noticeably around elections. Hence, it is not surprising that Standard bulk
23
nonprofit mail volumes exhibit this election-cycle seasonal pattern.
24
I looked at the relationship of Federal elections and Standard bulk nonprofit mail
25
volume quite extensively. As a result of this investigation, I added four variables to the
USPS-T-7
110
1
Standard bulk nonprofit demand equation in order to model the impact of elections on
2
Standard bulk nonprofit mail volume. These four variables are described in the next
3
section.
4
One thing worth noting with respect to Table 20 is that the years shown in Table 20
5
refer to Postal Fiscal Years. Postal Fiscal Years begin in the preceding fall (October 1
6
since FY 2000 in Table 20, at various times in September prior to 2000). Hence, fall of
7
2004, for example, fell during the first quarter of Fiscal 2005. Because of this, general
8
Federal elections actually fall in the odd-numbered Fiscal Years shown in Table 20,
9
while the impact of general Federal elections actually spans multiple Fiscal Years – i.e.,
10
the 2004 general election impacted Standard bulk nonprofit mail volumes at the end of
11
FY 2004 (September, 2004) and at the beginning of FY 2005 (October, 2004). Hence,
12
the division of years shown in Table 20 is not especially useful in isolating the full
13
impacts of Federal elections on Standard bulk nonprofit mail volumes.
14
Other econometric variables include seasonal variables and two dummy variables
15
that are described below. A more detailed look at the econometric demand equation for
16
Standard bulk nonprofit mail follows.
ii. Econometric Demand Equation
17
18
The demand equation for Standard bulk nonprofit mail in this case models Standard
19
bulk nonprofit mail volume per adult per delivery day as a function of the following
20
explanatory variables:
21
22
23
24
25
26
27
·
Seasonal Variables
·
Retail Sales
·
Variable Measuring the Impact of non-Presidential General Elections on
Standard Bulk Nonprofit Mail Volume
USPS-T-7
111
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
This variable is set equal to the proportion of the quarter which falls between September
16th and October 31st of non-Presidential Federal election years (e.g., 1994, 1998,
2002).
·
Non-Presidential General Election Variable Interacted with a Time Trend
which increases by one for every non-Presidential election cycle
·
Variable Measuring the Impact of Presidential General Elections on Standard
Bulk Nonprofit Mail Volume
This variable is set equal to the proportion of the quarter which falls between September
1st and October 31st of Presidential election years (e.g., 1992, 1996, 2000).
·
Presidential General Election Variable Interacted with a Dummy Variable
equal to One During the Year 2000, zero elsewhere
That is, the impact of the 2000 Presidential general election on Standard bulk nonprofit
mail volume is uniquely estimated here.
·
Variable Equal to One in the Third Quarter of Federal Election Years (both
Presidential and Non-Presidential)
This variable measures the impact of primary campaigns on Standard bulk nonprofit
mail volumes. No significant difference was noted between Presidential and nonPresidential election years in this regard.
·
Dummy Variable equal to one starting in 1994Q1, reflecting a rule change
which restricted nonprofit eligibility
·
Dummy variable for September 11th, equal to one in 2002Q1, zero elsewhere
·
Current and four lags of the price of Standard bulk nonprofit mail
33
Details of the econometric demand equation are shown in Table 21 below. A detailed
34
description of the econometric methodologies used to obtain these results can be found
35
in Section III below.
36
USPS-T-7
112
1
2
3
TABLE 21
ECONOMETRIC DEMAND EQUATION FOR STANDARD BULK NONPROFIT MAIL
Own-Price Elasticity
Long-Run
Current
Lag 1
Lag 2
Lag 3
Lag 4
Retail Sales
Election Season Dummies
Off-Year: mid-Sep – Oct
Off-Year interacted w/ Trend
Presidential: Sep – Oct
Dummy for 2000 Pres. Election
Coefficient
T-Statistic
-0.319
-0.061
-0.018
-0.020
-0.071
-0.150
0.430
-7.801
-0.538
-0.095
-0.098
-0.373
-1.332
14.84
0.078
0.034
1.772
1.505
0.112
0.135
3.408
2.812
1.970
0.030
Off-Year and Pres: Spring
Dummy for 1994 Rule Change
-0.024
-3.568
Dummy for 2002Q1 (9/11 Effect)
-0.034
-1.575
Seasonal Coefficients
0.157
1.388
September 16 – 30
0.558
3.951
October
0.054
0.625
November 1 – December 24
2.129
0.696
December 25 – 31
-0.098
-0.307
January – February
0.241
2.763
March
-0.152
-1.008
April 1 – 15
Quarter 1 (October – December)
-0.119
-0.581
Quarter 2 (January – March)
0.089
0.438
Quarter 3 (April – June)
-0.007
-0.273
Quarter 4 (July – September)
0.037
2.093
Seasonal Multipliers (GFY 2005)
1.152301
Quarter 1 (October – December)
1.014124
Quarter 2 (January – March)
0.875281
Quarter 3 (April – June)
0.964245
Quarter 4 (July – September)
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
AR-4: -0.393
Degrees of Freedom
40
Mean-Squared Error
0.000360
Adjusted R-Squared
0.971
USPS-T-7
113
D. Expedited Delivery Services
1
1. General Overview
2
The defining characteristic of expedited delivery services is the speed of delivery.
3
4
The mail being sent could consist of correspondence or transactions, mail-order
5
purchases, business documents, or other items. To some extent, several categories
6
of mail could provide expedited delivery services, depending on what one means
7
precisely by the term “expedited,” including First-Class Mail and Priority Mail.
The Postal Service only offers one product with a guaranteed delivery window,
8
9
however: Express Mail, which competes in the overnight delivery market. The dominant
10
player in the overnight delivery market is Federal Express. Other companies which
11
offer comparable services include United Parcel Service (UPS), Airborne and DHL
12
(which have recently merged). Table 22 below compares volumes for Federal Express
13
and Express Mail since 1985.2
2
Because of differences in the timing of the Fiscal Years used by the Postal Service and Federal
Express, I have adjusted the numbers shown in Table 1 for Federal Express to express them based on
the Postal calendar.
USPS-T-7
114
Fiscal Year
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
Ta ble 22
Ex pe dite d Volum e s De live re d by Fe de ra l Ex pre ss a nd the Unite d Sta te s Posta l Se rvice
(m illions of pie ce s)
Federal Express
Express Mail
Volume
Growth
Pct. of Total
Volume
Growth
Pct. of Total
113.220
72.10%
43.813
27.90%
148.298
30.98%
78.77%
39.974
-8.76%
21.23%
186.141
25.52%
81.81%
41.381
3.52%
18.19%
227.709
22.33%
83.42%
45.243
9.33%
16.58%
266.583
17.07%
83.41%
53.023
17.20%
16.59%
304.948
14.39%
83.92%
58.449
10.23%
16.08%
315.794
3.56%
84.54%
57.732
-1.23%
15.46%
364.068
15.29%
87.32%
52.889
-8.39%
12.68%
419.274
15.16%
88.93%
52.199
-1.30%
11.07%
479.182
14.29%
89.56%
55.861
7.02%
10.44%
546.505
14.05%
90.59%
56.735
1.57%
9.41%
585.993
7.23%
91.12%
57.124
0.69%
8.88%
658.077
12.30%
91.27%
62.914
10.13%
8.73%
709.186
7.77%
91.47%
66.129
5.11%
8.53%
733.471
3.42%
91.47%
68.366
3.38%
8.53%
762.627
3.98%
91.55%
70.377
2.94%
8.45%
725.952
-4.81%
91.31%
69.121
-1.78%
8.69%
696.117
-4.11%
91.91%
61.280
-11.34%
8.09%
701.232
0.73%
92.63%
55.831
-8.89%
7.37%
706.759
0.79%
92.89%
54.123
-3.06%
7.11%
Fiscal Y ears are Postal Fiscal Y ears f rom 1985 - 2000, Government Fiscal Y ears f rom 2001 - 2004.
As a whole, the overnight delivery market experienced tremendous growth through
2
3
the mid-1990s. From 1985 through 1998, Express Mail volume grew at an average
4
annual rate of 3.2 percent. Yet, Federal Express grew nearly five times faster, with
5
annual growth of 15.2 percent over this time period.
Since 1998, the overnight delivery market appears to have stopped growing so
6
7
dramatically. Since 2001, however, Express Mail’s share of the total market, as
8
suggested in Table 22, which had been fairly stable from 1996 – 2001, has worsened
9
appreciably.3
The key drivers of the demand for Express Mail are described and quantified below.
10
3
Table 22 does not, of course, present the entire Expedited delivery market.
USPS-T-7
115
1
2
2. Factors Affecting Express Mail Volume
The demand for Express Mail can be thought of as the product of two demands: the
3
demand for overnight delivery services and the demand for Express Mail as the
4
overnight delivery service of choice. This distinction provides a useful way of
5
understanding the variables that are included in the Express Mail demand equation.
a.
6
7
Demand for Overnight Delivery Services
The demand for overnight delivery services will, of course, be largely driven by the
8
demand for the goods being delivered overnight. Hence, the demand for overnight
9
delivery services would be expected to be strongly affected by the overall level of the
10
11
economy.
Overnight deliveries are primarily business-related. Total private employment is one
12
good business measure of the overall level of the economy and is therefore included in
13
the Express Mail equation.
14
As noted above from Table 22, the combined volume of overnight mail delivered by
15
Federal Express and the Postal Service grew significantly over the time period of
16
interest here. This growth is accounted for in the demand equation through the
17
inclusion of a linear time trend over the full sample period.
18
Overnight delivery services were severely adversely affected by the September 11,
19
2001, terrorist attacks and their immediate aftermath. Immediately following these
20
attacks, air traffic was banned for several days. For obvious reasons, overnight delivery
21
was hit hard by this restriction. This is modeled in the demand equation for Express
22
Mail by including a dummy variable equal to one during the quarter of the September
23
11th attacks and zero elsewhere.
USPS-T-7
116
b.
Demand for Express Mail as Overnight Delivery Service of Choice
1
2
3
As noted above, the dominant delivery service provider in the overnight market is
4
Federal Express. The impact of Federal Express on Express Mail volume is modeled in
5
two ways. First, the price of Federal Express, which is measured through Federal
6
Express’s average revenue per piece, excluding Ground and Freight services, is
7
included in the Express Mail equation. Second, the loss of Express Mail market share
8
since 2001 that is evident in Table 22 is modeled through a linear time trend which
9
starts in the third quarter of 2001.
10
Besides Federal Express, there are several other companies which provide
11
overnight delivery services, including United Parcel Service (UPS). UPS workers
12
engaged in a general strike in the summer of 1997 (1997Q4). This strike had a positive
13
impact on Express Mail volume as mailers who might have normally used UPS to
14
deliver their overnight documents and packages were forced to find an alternate
15
delivery service. This impact is modeled in the demand equation for Express Mail by
16
including a dummy variable equal to one during the quarter of the UPS strike and zero
17
elsewhere.
18
19
20
21
22
23
24
25
3. Demand Equation for Express Mail
a.
Factors Affecting Express Mail Volume
In summary, then, Express Mail volume was found to be affected by two dummy
variables as well as the following variables:
Total Private Employment
Long- and Short-Run Trends
Prices of Federal Express and Express Mail
USPS-T-7
117
1
The effect of these variables on Express Mail volume over the past ten years is
2
shown in Table 23 on the next page. Table 23 also shows the projected impacts of
3
these variables through GFY 2007.
4
The Test Year before-rates volume forecast for Express Mail is 52.945 million
5
pieces, a 2.2 percent decline from GFY 2004. The Postal Service’s proposed rates in
6
this case are predicted to reduce the Test Year volume of Express Mail by 4.8 percent,
7
for a Test Year after-rates volume forecast for Express Mail of 50.388 million.
USPS-T-7
118
Other
-0.20%
0.84%
-0.68%
-0.26%
1.00%
1.76%
-2.02%
1.49%
-1.76%
0.83%
Total Change
in Volum e
1.57%
0.69%
10.13%
5.11%
3.38%
3.76%
-2.56%
-11.34%
-8.89%
-3.06%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.22%
1.13%
1.23%
1.22%
1.21%
1.40%
1.22%
1.25%
1.27%
1.20%
1994 - 2004
Total
Avg per Year
13.06%
1.24%
4.22%
0.41%
-8.02%
-0.83%
-28.72%
-3.33%
31.84%
2.80%
-4.34%
-0.44%
-1.46%
-0.15%
0.92%
0.09%
-3.11%
-0.32%
2001 - 2004
Total
Avg per Year
3.77%
1.24%
-6.77%
-2.31%
-18.12%
-6.45%
-9.33%
-3.21%
8.45%
2.74%
-1.34%
-0.45%
1.35%
0.45%
0.53%
0.18%
-21.70%
-7.83%
Forecas t
1.23%
1.11%
1.09%
0.86%
0.64%
0.04%
-6.51%
-6.47%
-6.44%
0.00%
0.00%
0.00%
3.16%
2.63%
2.79%
1.07%
-0.18%
0.00%
-0.23%
0.00%
-0.97%
0.68%
0.35%
0.00%
-0.02%
-2.15%
-3.69%
3.46%
1.14%
1.55%
0.51%
-18.20%
-6.48%
0.00%
0.00%
8.83%
2.86%
0.89%
0.30%
-1.19%
-0.40%
1.03%
0.34%
-5.79%
-1.97%
After-Rates Volum e Forecas t
2005
1.23%
2006
1.11%
2007
1.09%
0.86%
0.64%
0.04%
-6.51%
-6.47%
-6.44%
0.00%
-4.83%
-2.82%
3.16%
2.63%
2.79%
1.07%
-0.18%
0.00%
-0.23%
0.00%
-0.97%
0.68%
0.35%
0.00%
-0.02%
-6.88%
-6.41%
1.55%
0.51%
-18.20%
-6.48%
-7.51%
-2.57%
8.83%
2.86%
0.89%
0.30%
-1.19%
-0.40%
1.03%
0.34%
-12.87%
-4.49%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 23
Estimated Impact of Factors Affecting Express Mail Volume, 1994 – 2007
Expres s Mail Price
Other Factors
Em ploym ent Tim e Trends
Nom inal
Inflation
FedEx Price Econom etric
2.92%
2.13%
-3.55%
3.23%
-2.93%
-1.04%
1.40%
2.11%
-6.68%
3.05%
-1.15%
0.32%
2.05%
2.16%
-0.32%
2.98%
-0.48%
2.86%
2.33%
2.17%
0.00%
1.99%
0.77%
-3.11%
1.70%
2.15%
-3.43%
1.60%
0.26%
-1.02%
1.67%
2.21%
-6.99%
3.37%
-0.03%
0.68%
-0.81%
-1.15%
-2.43%
3.62%
0.52%
-1.40%
-3.90%
-6.43%
-4.72%
2.68%
-0.84%
-1.10%
-2.35%
-6.43%
-3.89%
2.63%
-0.33%
1.94%
-0.65%
-6.48%
-0.98%
2.91%
-0.18%
0.52%
2004 - 2007
Total
Avg per Year
3.46%
1.14%
USPS-T-7
119
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 23 above.
5
Express Mail has an employment elasticity of 1.286 (t-statistic of 4.406), meaning
6
that a 10 percent increase in employment will lead to a 12.86 percent increase in the
7
volume of Express Mail.
8
9
As outlined above, the Express Mail equation includes two trend variables. The first
is a full-sample trend which reflects historical growth in the overnight delivery market.
10
This trend explains annual growth of approximately 2.1 percent throughout the time
11
period shown in Table 23. The second time trend, which has only been in effect since
12
2001Q3, reflects Express Mail’s declining market share. This trend explains an annual
13
decline in volume of approximately 8.4 percent. The combined effect of these time
14
trends is to reduce Express Mail volume by approximately 6.5 percent per year through
15
the forecast period.
16
In real dollars, FedEx’s revenue per piece declined by more than 8 percent from
17
1994 to 2004. The cross-price elasticity of Express Mail with respect to FedEx was
18
calculated to be equal to 0.420 (t-statistic of 4.656), so that Express Mail volume
19
declined by more than 4.3 percent from 1994 through 2004 due to declining real FedEx
20
prices. For this case, FedEx prices are assumed to remain constant in real dollars (i.e.,
21
FedEx prices are expected to increase at the same rate as overall inflation) through the
22
forecast period.
23
The own-price elasticity of Express Mail was calculated to be equal to -1.470
24
(t−statistic of -13.45). This is the highest own-price elasticity of any mail product
25
discussed in my testimony.
USPS-T-7
120
1
The Postal price impacts shown in Table 23 above are the result of changes in
2
nominal prices. Prices enter the demand equations developed here in real terms,
3
however. The impact of inflation reported in Table 23 measures the impact that a
4
change in real Postal prices, in the absence of nominal rate changes, has on the
5
volume of Express Mail.
6
Other econometric variables include seasonal variables and dummy variables to
7
account for the temporary impacts of the UPS strike in the summer of 1997 and the
8
September 11, 2001, terrorist attacks. A more detailed look at the econometric demand
9
equation for Express Mail follows.
b. Econometric Demand Equation
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
The demand equation for Express Mail in this case models Express Mail volume per
adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
·
Total Private Employment
·
Linear Time Trend over the Full Sample Period
·
Linear Time Trend starting in 2001Q3
·
Dummy variable for the UPS strike in the summer of 1997, equal to one in
1997Q4, zero elsewhere
·
Dummy variable for September 11th, equal to one in 2002Q1, zero elsewhere
·
Current and three lags of average revenue per piece for Federal Express
(excluding Freight and Ground services)
·
Current and four lags of the price of Express Mail
Details of the econometric demand equation are shown in Table 24 below. A
31
detailed description of the econometric methodologies used to obtain these results can
32
be found in Section III below.
USPS-T-7
121
1
2
3
TABLE 24
ECONOMETRIC DEMAND EQUATION FOR EXPRESS MAIL
Coefficient
T-Statistic
Own-Price Elasticity
-13.45
Long-Run
-1.470
-0.556
-4.419
Current
-0.338
-2.241
Lag 1
-0.001
-0.007
Lag 2
-0.517
-3.353
Lag 3
-0.058
-0.500
Lag 4
Federal Express Cross-Price Elasticity
Long-Run
0.420
4.656
Current
0.007
0.042
Lag 1
0.165
0.903
Lag 2
0.189
1.113
Lag 3
0.059
0.428
Employment
1.286
4.406
Time Trends
Full-Sample
0.005
3.829
Since 2001Q3
-0.022
-7.040
Combined Effect of Trends
(annual percentage)
1990Q1 – 2001Q2
+2.2%
2001Q3 onward
-6.4%
Dummy for UPS Strike (1997Q4)
0.097
6.835
Dummy Starting in 2001Q3
-0.023
-1.409
Dummy for 2002Q1 (9/11 Effect)
-0.070
-4.979
Seasonal Coefficients
0.214
1.073
September 1 – 15
-0.043
-0.985
September 16 – October 31
0.067
0.938
November 1 – December 12
0.532
3.846
December 13 – 17
-0.180
-1.085
December 18 – 21
0.768
1.638
December 22 – 24
-5.943
-2.281
December 25 – 31
0.688
2.645
January 1 - February
-0.085
-1.996
March
0.449
3.123
April 1 – 15
Quarter 1 (October – December)
0.381
2.295
Quarter 2 (January – March)
-0.341
-2.058
Quarter 3 (April – June)
-0.023
-1.709
Quarter 4 (July – September)
-0.017
-2.029
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.012223
Quarter 2 (January – March)
1.017516
Quarter 3 (April – June)
1.005470
Quarter 4 (July – September)
0.961822
REGRESSION DIAGNOSTICS
Sample Period
1985Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.656
Degrees of Freedom
50
Mean-Squared Error
0.000197
Adjusted R-Squared
0.987
4
USPS-T-7
122
1
2
3
E. Package Delivery Services
1. Overview of Ground Package Delivery Market
Package delivery services refer broadly to the delivery of goods other than
4
Periodicals, advertisements, and correspondence. Examples of this type of mail include
5
mail-order deliveries (such as clothes) and the delivery of books, tapes, or CDs (such as
6
from book or CD clubs), as well as packages sent by households (e.g., Christmas
7
presents). This encompasses the Priority Mail subclass as well as the Package
8
Services mail class.
9
The demand for package delivery services is a derived demand, emanating from the
10
demand for the products being delivered. As such, the demand for package delivery
11
services would be expected to be a function of the usual factors affecting demand. The
12
demand for package delivery services offered by the Postal Service will be affected not
13
only by the price charged by the Postal Service for these services but also by the
14
availability and price of alternate delivery forms, including non-Postal alternatives.
15
My testimony is specifically focused on the ground package delivery market sub-
16
market of the package delivery market. The ground package delivery market refers to
17
packages delivered via ground. Such delivery may take anywhere from a few days to
18
several weeks. The Postal Service offers four products which compete in this market:
19
Priority Mail, Parcel Post, Bound Printed Matter, and Media Mail.
20
Technically, Priority Mail is not delivered via ground. In terms of delivery time,
21
Priority Mail falls somewhere between expedited delivery − e.g., one- or two-day
22
guaranteed delivery − and ground delivery. While the average delivery time for Priority
23
Mail of two to three days may be comparable to some more expedited services, Priority
24
Mail is hampered in its ability to compete effectively in this market by a lack of a
25
guaranteed delivery standard. The ground package market has evolved over time to
USPS-T-7
123
1
include full tracking and tracing and guaranteed delivery service. Therefore, I combine
2
Priority Mail with Package Services Mail as part of what I am calling the ground package
3
delivery market.
4
For many years, the dominant player in the ground package delivery market has
5
been United Parcel Service (UPS). Federal Express has entered the ground market
6
fairly aggressively within the past decade. Table 25 below compares ground package
7
volumes for UPS, Federal Express, and the Postal Service since 1990.4 The percent of
8
total figures in Table 25 indicate percentages of the combined volume of UPS Ground,
9
FedEx Ground, and the Postal Service volumes shown there. Of course, these are not
10
equal to market shares as data for other package delivery companies are not shown
11
here.
4
I have adjusted the UPS and FedEx numbers shown in Table 25 to express them based on the Postal
calendar.
USPS-T-7
124
Ta ble 25
Pa cka ge De live ry Volume s: 1990 - 2004
(m illions of pie ce s)
Fiscal Year
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
United Parcel Service
Volume
Growth Pct. of Total
2,499.608
67.92%
2,535.862
1.45%
67.49%
2,574.471
1.52%
65.82%
2,536.335
-1.48%
63.32%
2,517.332
-0.75%
59.94%
2,534.852
0.70%
58.23%
2,544.179
0.37%
56.84%
2,497.574
-1.83%
51.69%
2,420.364
-3.09%
48.66%
2,501.290
3.34%
49.18%
2,644.463
5.72%
49.43%
2,616.139
-1.07%
50.06%
2,558.141
-2.22%
50.00%
2,552.889
-0.21%
50.15%
2,689.261
5.34%
50.92%
Federal Express
Volume
Growth Pct. of Total
0.000
0.00%
0.000
0.00%
0.000
0.00%
0.000
0.00%
0.000
0.00%
0.000
0.00%
0.000
0.00%
231.605
4.79%
354.738
53.16%
7.13%
354.289
-0.13%
6.97%
376.725
6.33%
7.04%
395.988
5.11%
7.58%
484.618
22.38%
9.47%
549.254
13.34%
10.79%
611.507
11.33%
11.58%
Fiscal Y ears are Postal Fiscal Y ears f rom 1990 - 2000, Government Fiscal Y ears f rom 2001 - 2004.
Volume
1,180.728
1,221.477
1,336.987
1,469.098
1,682.391
1,818.356
1,932.079
2,102.770
2,198.793
2,230.713
2,328.260
2,214.014
2,073.238
1,988.102
1,980.562
Postal Service
Growth Pct. of Total
32.08%
3.45%
32.51%
9.46%
34.18%
9.88%
36.68%
14.52%
40.06%
8.08%
41.77%
6.25%
43.16%
8.83%
43.52%
4.57%
44.21%
1.45%
43.86%
4.37%
43.52%
-4.91%
42.36%
-6.36%
40.52%
-4.11%
39.06%
-0.38%
37.50%
USPS-T-7
125
1
Overall, the ground package delivery market has experienced modest growth since
2
1990, with average annual growth of 2.6 percent. This almost certainly overstates the
3
actual growth of ground package delivery over this time, as, for example, RPS, the
4
precursor to FedEx Ground, delivered some volume prior to its acquisition by Federal
5
Express. In terms of relative shares, the biggest story evident in Table 25 is a gain by
6
Federal Express since 1997. Federal Express’s initial gain in 1997 was largely at the
7
expense of UPS. Since 1998, however, UPS’s share of the total shown in Table 25 has
8
remained relatively stable. Subsequent gains in Federal Express’s share of the total
9
appear, therefore, to have been at the expense of the Postal Service.
10
As in the overnight delivery market above, the demand for a particular type of
11
ground package delivery can be thought of as the product of two demands: the demand
12
for ground package delivery services in general and the demand for the ground
13
package delivery service of interest as the delivery service of choice.
14
15
2. Final Demand Equations
The Postal Service offers several product offerings which compete within the ground
16
package delivery market. Table 26 presents volumes and shares of total across these
17
products. The shares of total shown in Table 26 are shares of the total volumes
18
presented in Table 25 above and so sum to the Postal Service share in that table rather
19
than 100 percent. The three products shown in Table 26 are discussed below.
USPS-T-7
126
1
Ta ble 26
Posta l Service Volum e s by Subcla ss
2
Fiscal Year
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Volume
517.140
529.070
578.655
674.084
779.475
852.036
936.211
1,065.555
1,167.999
1,187.813
1,215.581
1,117.088
998.151
859.587
848.633
Priority Mail
Growth
2.31%
9.37%
16.49%
15.63%
9.31%
9.88%
13.82%
9.61%
1.70%
2.34%
-8.10%
-10.65%
-13.88%
-1.27%
Pct. of Total
14.05%
14.08%
14.79%
16.83%
18.56%
19.57%
20.92%
22.05%
23.48%
23.35%
22.72%
21.38%
19.51%
16.89%
16.07%
Volume
128.700
138.457
164.203
232.845
258.972
258.845
262.495
291.650
319.991
326.021
323.073
353.146
372.591
386.944
375.618
Parcel Post
Growth
7.58%
18.60%
41.80%
11.22%
-0.05%
1.41%
11.11%
9.72%
1.88%
-0.90%
9.31%
5.51%
3.85%
-2.93%
Pct. of Total
3.50%
3.68%
4.20%
5.81%
6.17%
5.95%
5.86%
6.04%
6.43%
6.41%
6.04%
6.76%
7.28%
7.60%
7.11%
Other Package Services
Volume
Growth Pct. of Total
534.888
14.53%
553.951
3.56%
14.74%
594.129
7.25%
15.19%
562.169
-5.38%
14.04%
643.944
14.55%
15.33%
707.475
9.87%
16.25%
733.373
3.66%
16.38%
745.565
1.66%
15.43%
710.803
-4.66%
14.29%
716.879
0.85%
14.09%
789.606
10.14%
14.76%
743.780
-5.80%
14.23%
702.496
-5.55%
13.73%
741.571
5.56%
14.57%
756.311
1.99%
14.32%
USPS-T-7
127
a. Priority Mail
1
i. Factors Affecting Priority Mail Volume
2
3
The demand for ground package delivery services will be largely driven by the
4
demand for the goods being delivered. For the Priority Mail demand equation, this
5
relationship is modeled through the inclusion of total retail sales as an explanatory
6
variable.
7
The choice of delivery service will be made based on several factors, including price,
8
level of service, and access. Price is measured by including the price of Priority Mail in
9
its demand equation. Besides its own price, Priority Mail volume is also affected by the
10
prices charged by its competitors, measured by the combined average revenue per
11
piece of UPS and FedEx Ground services.
12
13
The level of service is modeled in the Priority Mail equation by including the average
number of days to deliver Priority Mail as an explanatory variable.
14
In addition, the Priority Mail equation includes a number of trend variables which
15
largely reflect changes in the market conditions under which Priority Mail has competed.
16
These trends are described in more detail below.
17
18
19
20
21
22
23
24
In summary, then, Priority Mail volume was found to be affected primarily by the
following variables:
Retail Sales
Time Trends
Average Delivery Time
Prices of UPS, Federal Express, and Priority Mail
The effect of these variables on Priority Mail volume over the past ten years is
25
shown in Table 27 on the next page. Table 27 also shows the projected impacts of
26
these variables through GFY 2007.
USPS-T-7
128
1
The Test Year before-rates volume forecast for Priority Mail is 842.705 million
2
pieces, a 0.7 percent decline from GFY 2004. The Postal Service’s proposed rates in
3
this case are predicted to reduce the Test Year volume of Priority Mail by 5.1 percent,
4
for a Test Year after-rates volume forecast for Priority Mail of 799.324 million.
USPS-T-7
129
Other
-0.40%
0.53%
-0.11%
-0.59%
-0.76%
2.28%
-1.11%
0.84%
-1.52%
0.12%
Total Change
in Volum e
9.31%
9.88%
13.82%
9.61%
1.70%
2.92%
-8.62%
-10.65%
-13.88%
-1.27%
-13.28%
-1.41%
-0.76%
-0.08%
8.87%
0.85%
3.66%
1.20%
-2.67%
-0.90%
-0.57%
-0.19%
-24.03%
-8.75%
2.01%
1.68%
1.97%
-2.20%
-1.84%
0.00%
-0.44%
-0.54%
-0.66%
-0.30%
-0.23%
0.00%
-0.28%
-0.42%
1.85%
0.00%
0.00%
5.77%
1.89%
-3.99%
-1.35%
-1.64%
-0.55%
-0.53%
-0.18%
1.13%
0.38%
0.01%
0.34%
0.30%
0.00%
-5.15%
0.00%
2.01%
1.68%
1.97%
-2.20%
-1.84%
0.00%
-0.44%
-0.54%
-0.66%
-0.30%
-0.23%
0.00%
-0.28%
-5.55%
1.85%
0.65%
0.22%
-5.15%
-1.75%
5.77%
1.89%
-3.99%
-1.35%
-1.64%
-0.55%
-0.53%
-0.18%
-4.07%
-1.38%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.25%
1.16%
1.25%
1.23%
1.20%
1.38%
1.19%
1.25%
1.22%
1.21%
Retail Sales
0.44%
0.28%
0.32%
0.32%
0.68%
0.67%
-0.06%
0.02%
0.08%
0.55%
1994 - 2004
Total
Avg per Year
13.04%
1.23%
3.34%
0.33%
13.80%
1.30%
1.62%
0.16%
-35.46%
-4.28%
17.29%
1.61%
23.71%
2.15%
2001 - 2004
Total
Avg per Year
3.72%
1.23%
0.64%
0.21%
-19.89%
-7.13%
1.95%
0.64%
-15.91%
-5.61%
5.64%
1.85%
Forecas t
1.22%
1.11%
1.11%
0.49%
0.12%
0.16%
-1.02%
-1.02%
-1.03%
0.01%
0.34%
0.30%
0.00%
0.00%
0.00%
3.49%
1.15%
0.78%
0.26%
-3.03%
-1.02%
0.65%
0.22%
After-Rates Volum e Forecas t
2005
1.22%
2006
1.11%
2007
1.11%
0.49%
0.12%
0.16%
-1.02%
-1.02%
-1.03%
0.78%
0.26%
-3.03%
-1.02%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 27
Estimated Impact of Factors Affecting Priority Mail Volume, 1994 – 2007
Avg Delivery
UPS/FedEx
Priority Mail Price
Other Factors
Tim e Trends
Tim e
Nom inal
Inflation
Prices Econom etric
5.57%
0.45%
-2.00%
1.49%
2.32%
0.00%
5.46%
-0.05%
-0.78%
1.28%
1.80%
-0.07%
5.58%
-0.79%
0.00%
1.59%
1.65%
3.70%
5.61%
0.15%
0.00%
0.86%
5.35%
-3.37%
5.40%
0.14%
-8.20%
1.04%
5.77%
-2.87%
5.57%
-0.02%
-4.75%
2.24%
0.20%
-4.27%
2.84%
-0.20%
-9.72%
2.06%
0.96%
-4.30%
-7.36%
0.46%
-7.49%
1.44%
0.90%
-0.72%
-8.26%
1.14%
-9.11%
2.03%
1.19%
-0.85%
-5.74%
0.34%
0.00%
2.07%
1.52%
-1.13%
2004 - 2007
Total
Avg per Year
3.49%
1.15%
USPS-T-7
130
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 27 above.
5
Priority Mail has a retail sales elasticity of 0.144 (t-statistic of 0.668), meaning that a
6
10 percent increase in retail sales will lead to a 1.44 percent increase in the volume of
7
Priority Mail.
8
The average time it takes to deliver a piece of Priority Mail, as measured by the
9
Postal Service, has improved from 2.2 days in GFY 2001 to less than 2.1 days in GFY
10
2004. This improvement in service has helped to improve Priority Mail volume by nearly
11
2 percent over these three years.
12
A key factor which affects price elasticities in general is the overall competitiveness
13
of a market, i.e., the number and closeness of substitutes for a product. The
14
emergence of FedEx Ground over the past few years as a significant player in the
15
ground package market has led to an increase in the level of competition in the ground
16
package delivery market. This increased competition is modeled through increasing
17
price elasticities of Priority Mail volume with respect to the prices of Priority Mail as well
18
as UPS and FedEx Ground. These price elasticities are modeled to have increased (in
19
absolute value) as FedEx Ground’s market reach expanded.
20
Prior to the existence of FedEx Ground, the own-price elasticity of Priority Mail was
21
calculated to be equal to -0.650 (t-statistic of -5.086). Upon the initial introduction of
22
FedEx Ground into the ground package market, FedEx ground was assumed to have a
23
market reach of approximately 50 percent. That is, it was assumed that FedEx Ground
24
delivery service was available to approximately 50 percent of United States addresses.
USPS-T-7
131
1
The addition of this new competitor increased the Priority Mail own-price elasticity to
2
approximately -0.827. FedEx aggressively expanded their market reach starting around
3
mid-1999, reaching 70 percent market reach in early 2001, 90 percent market reach in
4
early 2002, and 100 percent market reach by early 2003. The current own-price
5
elasticity of Priority Mail, given 100 percent FedEx Ground market reach, is -1.004 (t-
6
statistic of -6.824).
7
Prior to the existence of FedEx Ground, the cross-price elasticity of Priority Mail with
8
respect to UPS Ground prices was calculated to be equal to 1.204 (t-statistic of 5.112).
9
Since FedEx Ground achieved 100 percent market reach, this elasticity (now with
10
respect to the average of UPS and FedEx Ground prices) has risen to 1.446 (t-statistic
11
of 4.599).
12
The market position of Priority Mail within the ground package delivery market is
13
affected by factors beyond simply prices and average delivery time. Many of these
14
factors are difficult to quantify and do not necessarily lend themselves to direct inclusion
15
in an econometric demand equation. Tables 25 and 26 provide some helpful insight
16
into how the market position of Priority Mail has changed over time. For most of the
17
time period outlined in those tables, Priority Mail volume growth outpaced UPS Ground
18
volume growth, so that Priority Mail’s estimated share of total volume grew from 14
19
percent in the early 1990s to 23 percent by the end of the decade.
20
Beginning in 2000, and accelerating rapidly in 2001, 2002, and 2003, Priority Mail’s
21
market share tumbled from 23.4 percent in FY 1999 to 16.9 percent in FY 2003. Much
22
of this loss was apparently at the expense of FedEx Ground, which saw its market
23
share increase from 7.0 to 10.8 percent over this same time period. This is the time
24
period during which FedEx was aggressively expanding the market reach of its ground
25
delivery operations.
USPS-T-7
132
1
This rapid decline in Priority Mail market share appears to have slowed in 2004.
2
This slowdown is likely due to many factors. First, FedEx Ground achieved 100 percent
3
market reach. Hence, any subsequent growth of FedEx Ground’s market share will
4
have to involve increasing shares in existing markets. Second, UPS appears to be
5
voluntarily pricing itself out of certain markets. In 1999, UPS began to charge a
6
surcharge for residential deliveries within certain, mostly rural, ZIP Codes. In 2004, this
7
surcharge was expanded to include commercial deliveries within these same ZIP
8
Codes. This surcharge has made Priority Mail a more attractive option for many of
9
these customers, helping to ameliorate recent losses.
10
To reflect the changes described above, the Priority Mail equation includes three
11
trend variables. The first is a full-sample trend consistent with Priority Mail’s growing
12
market share over much of the early sample period here. This trend explains annual
13
growth of approximately 5.5 percent throughout the time period shown in Table 27. The
14
second time trend, which has only been in effect since 2001Q3, reflects Priority Mail’s
15
declining market share over this more recent time period. This trend, coupled with the
16
full-sample time trend, explains an annual decline in volume of more than 8 percent
17
from mid-2001 through 2003. The final trend in the Priority Mail equation begins in the
18
second quarter of 2004. This trend is timed to coincide with UPS’s expansion of the
19
rural residential surcharge to include commercial deliveries as well as residential ones.
20
Taken together, the three trends explain an annual decline in Priority Mail volume of
21
approximately 1.0 percent from 2004Q2 onward. The effect of these trends is projected
22
to continue at an annual level of approximately -1.0 percent throughout the forecast
23
period presented in this case.
USPS-T-7
133
1
Other econometric variables include seasonal variables and several dummy variables
2
which are described below. A more detailed look at the econometric demand equation
3
for Priority Mail follows.
ii. Econometric Demand Equation
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
The demand equation for Priority Mail in this case models Priority Mail volume per
adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
·
Retail Sales
·
Average Number of Days to Deliver Priority Mail
·
Linear Time Trend over the Full Sample Period
·
Linear Time Trend starting in 2001Q3
·
Linear Time Trend starting in 2004Q2
·
Dummy variable for the UPS strike in the summer of 1997, equal to one in
1997Q4, zero elsewhere
·
Dummy variable for the quarter immediately following the 1997 UPS strike,
equal to one in 1998Q1, zero elsewhere
·
Dummy variable equal to one since the introduction of FedEx Ground
The coefficient on this dummy variable does not reflect the impact of the introduction of
FedEx Ground on Priority Mail volume so much as it reflects a change to the constant
term as a result of interacting the price terms with the market reach of FedEx Ground.
·
Dummy variable equal to one starting in 2000Q3
This dummy variable reflects a level shift in Priority Mail volume at this time, which
served as a precursor to the more dramatic declines in Priority Mail which began the
next year and are measured through the linear time trend starting in 2001Q3.
USPS-T-7
134
1
2
3
4
5
6
·
Current and three lags of Average revenue per piece for UPS and FedEx
Ground Delivery Services
·
Current UPS/FedEx price times the estimated market reach of FedEx Ground
·
Current price of Priority Mail
7
8
9
·
Current Priority Mail price times the estimated market reach of FedEx Ground
Details of the econometric demand equation are shown in Table 28 below. A detailed
10
description of the econometric methodologies used to obtain these results can be found
11
in Section III below.
USPS-T-7
135
1
2
3
TABLE 28
ECONOMETRIC DEMAND EQUATION FOR PRIORITY MAIL
Own-Price Elasticity
Since Introduction of FedEx Ground
Current Long-Run Elasticity
Interaction with FedEx Market Reach
Full-Sample
Long-Run (current only)
UPS / FedEx Ground Market Price
Since Introduction of FedEx Ground
Current Long-Run Elasticity
Interaction with FedEx Market Reach
Full-Sample
Long-Run
Current
Lag 1
Lag 2
Lag 3
Lag 4
Retail Sales
Time Trends
Full-Sample
Since 2001Q3
Since 2004Q2
Combined Effect of Trends
(annual percentage)
1990Q1 – 2001Q2
2001Q3 – 2004Q1
2004Q2 onward
Average Delivery Days, Priority Mail
Dummy Variables for UPS Strike
1997Q4 (Quarter of UPS Strike)
1998Q1 (Quarter following Strike)
Dummy Variable for Start of FedEx Ground
Dummy Variable Starting in 2000Q3
Coefficient
T-Statistic
-1.004
-0.355
-6.824
-1.895
-0.650
-5.086
1.446
0.241
4.599
1.369
1.204
0.281
0.210
0.179
0.011
0.524
0.144
5.112
1.862
0.930
0.666
0.064
3.959
0.668
0.013
-0.036
0.020
7.440
-10.44
2.780
+5.5%
-8.6%
-1.0%
-0.162
-2.218
0.142
0.066
0.106
-0.083
3.661
1.420
2.055
-4.115
USPS-T-7
136
1
2
3
TABLE 28 (continued)
ECONOMETRIC DEMAND EQUATION FOR PRIORITY MAIL
Coefficient T-Statistic
Seasonal Coefficients
November – December
0.513
7.145
January – March
0.199
7.542
April – June
0.263
5.183
Quarter 1 (October – December)
0.012
0.429
Quarter 2 (January – March)
0.013
0.420
Quarter 3 (April – June)
-0.110
-3.614
Quarter 4 (July – September)
0.085
3.120
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.160419
Quarter 2 (January – March)
1.007362
Quarter 3 (April – June)
0.949624
Quarter 4 (July – September)
0.886953
REGRESSION DIAGNOSTICS
Sample Period
1990Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: -0.541
Degrees of Freedom
34
Mean-Squared Error
0.000565
Adjusted R-Squared
0.991
4
USPS-T-7
137
b. Parcel Post
1
i. General Overview
2
3
Parcel Post volume is divided into two categories for the purpose of estimating
4
demand equations here: non-destination-entry Parcel Post and destination-entry Parcel
5
Post.
6
Non-destination-entry Parcel Post is Parcel Post volume that does not receive
7
destination-entry discounts. For the most part, it represents packages that are sent one
8
or a few at a time and are mailed at the local Post Office. Therefore, non-destination
9
entry parcel post could be thought of as single-piece Parcel Post.
10
Destination-entry Parcel Post, on the other hand, is Parcel Post volume that does
11
receive destination-entry discounts. It is entered at either the destination Bulk Mail
12
Center (BMC), Sectional Center Facility (SCF), or Delivery Unit (DU) and must be
13
entered as part of a mailing of 50 or more pieces. This volume could be thought of,
14
then, as bulk Parcel Post.
15
Table 29 below presents volumes for non-destination-entry and destination-entry
16
Parcel Post mail from 1996 to the present. Table 30 below presents the annual
17
percentage change in volume over the same period last year for these mail categories
18
over this same time period.
19
One somewhat curious thing that is evident from Table 30 is that the growth rates for
20
non-destination-entry and destination-entry Parcel Post have the same sign in only 8 of
21
the last 37 quarters.
USPS-T-7
138
1
Quarter
1996PQ1
1996PQ2
1996PQ3
1996PQ4
1997PQ1
1997PQ2
1997PQ3
1997PQ4
1998PQ1
1998PQ2
1998PQ3
1998PQ4
1999PQ1
1999PQ2
1999PQ3
1999PQ4
2000GQ1
2000GQ2
2000GQ3
2000GQ4
2001GQ1
2001GQ2
2001GQ3
2001GQ4
2002GQ1
2002GQ2
2002GQ3
2002GQ4
2003GQ1
2003GQ2
2003GQ3
2003GQ4
2004GQ1
2004GQ2
2004GQ3
2004GQ4
2005GQ1
Ta ble 29
Pa rcel Post Volume
(millions of pieces)
Non-Destination Entry
Destination Entry
27.164
43.018
26.628
37.419
21.974
37.490
28.029
40.772
24.672
51.149
26.563
44.375
21.329
38.646
34.268
50.648
26.827
59.815
28.617
49.584
23.410
47.532
28.089
56.117
26.439
68.382
29.146
52.922
19.477
49.614
23.065
56.977
26.929
79.905
19.937
56.947
17.143
54.416
15.884
53.005
21.845
85.612
27.892
56.355
25.682
57.791
22.386
55.583
35.368
83.471
26.931
58.843
24.024
58.944
22.303
62.708
32.778
92.825
26.235
66.039
23.087
63.320
23.437
59.223
34.760
90.087
27.976
59.242
23.470
58.823
23.758
57.502
35.497
87.863
USPS-T-7
139
1
Quarter
1996PQ1
1996PQ2
1996PQ3
1996PQ4
1997PQ1
1997PQ2
1997PQ3
1997PQ4
1998PQ1
1998PQ2
1998PQ3
1998PQ4
1999PQ1
1999PQ2
1999PQ3
1999PQ4
2000PQ1
2000PQ2
2000PQ3
2000PQ4
2001GQ1
2001GQ2
2001GQ3
2001GQ4
2002GQ1
2002GQ2
2002GQ3
2002GQ4
2003GQ1
2003GQ2
2003GQ3
2003GQ4
2004GQ1
2004GQ2
2004GQ3
2004GQ4
2005GQ1
Ta ble 30
Pe rce nta ge Cha nge ove r Sa me Pe riod La st Ye a r
Pa rcel Post Volume
Non-Destination Entry
Destination Entry
-28.38%
22.86%
-20.65%
10.02%
-12.46%
23.03%
-1.34%
18.70%
-9.18%
18.90%
-0.24%
18.59%
-2.93%
3.08%
22.26%
24.22%
8.74%
16.94%
7.73%
11.74%
9.75%
22.99%
-18.03%
10.80%
-1.45%
14.32%
1.85%
6.73%
-16.80%
4.38%
-17.89%
1.53%
-16.03%
-4.69%
-23.74%
11.37%
-12.73%
14.01%
-16.87%
8.46%
-18.88%
7.14%
39.90%
-1.04%
49.81%
6.20%
40.94%
4.87%
61.90%
-2.50%
-3.45%
4.42%
-6.46%
1.99%
-0.37%
12.82%
-7.32%
11.21%
-2.58%
12.23%
-3.90%
7.43%
5.08%
-5.56%
6.05%
-2.95%
6.64%
-10.29%
1.66%
-7.10%
1.37%
-2.91%
2.12%
-2.47%
USPS-T-7
140
1
Destination-entry Parcel Post volume experienced strong, virtually uninterrupted,
2
growth from its introduction in 1991 through the third quarter of 2003. Much of this was
3
the typical sort of initial growth that would be expected with the introduction of a new
4
product. Mail shifted from non-destination-entry to destination-entry as more mailers
5
were able to take advantage of these newer, lower rates. Several parcel consolidators
6
came into business which enabled shippers of fewer parcels to be able to take
7
advantage of these discounts as well. Destination-entry discounts were also expanded
8
considerably in R97-1 (1999Q2) with the introduction of DSCF and DDU discounts.
9
The more recent quarters, however, tell a somewhat different story. Destination-
10
entry Parcel Post appears to have fully matured as a product and is beginning to feel
11
the pressure of increasing competition in the ground parcel market. UPS introduced a
12
new product, UPS Basic, in November, 2003 to compete directly with destination-entry
13
Parcel Post. Because of this and other competitive pressure, destination-entry Parcel
14
Post volume has declined over the same period last year for the last six consecutive
15
quarters.
16
In many ways, the story of non-destination-entry Parcel Post is a mirror-image of the
17
destination-entry story. From 1971 through the first quarter of 2001, non-destination-
18
entry Parcel Post volume had a negative SPLY in 85 of 121 quarters (70.2 percent).
19
This was most true through the 1970s and 1980s, when Parcel Post volume (all Parcel
20
Post was non-destination-entry at that time) declined in 62 of 76 quarters (81.6 percent),
21
with a total decline in volume of 78.5 percent. Parcel Post volume first showed signs of
22
recovery in 1990. With the introduction of destination BMC discounts in 1991, however,
23
most of this recovery became limited to destination-entry Parcel Post volume, while
24
non-destination-entry volume continued to decline. From 1991 to 2000, non-
25
destination-entry Parcel Post volume declined an additional 39.6 percent.
USPS-T-7
141
1
2
3
4
5
6
7
8
9
10
11
12
More recently, however, it appears that the Postal Service has found a niche in the
ground package delivery market with non-destination-entry Parcel Post.
The specific demand equations estimated here for non-destination-entry and
destination-entry Parcel Post mail are described below.
ii. Non-Destination Entry Parcel Post Mail
(a) Factors Affecting Non-Destination Entry Parcel Post Mail
Non-destination-entry Parcel Post volume was found to be principally affected by the
following variables:
Price of UPS Ground delivery
Price of non-destination-entry Parcel Post mail
The effect of these variables on non-destination-entry Parcel Post volume over the
13
past ten years is shown in Table 31 on the next page. Table 31 also shows the
14
projected impacts of these variables through GFY 2007.
15
The Test Year before-rates volume forecast for non-destination entry Parcel Post
16
is 116.209 million pieces, a 5.7 percent increase from GFY 2004. The Postal Service’s
17
proposed rates in this case are predicted to reduce the Test Year volume of non-
18
destination entry Parcel Post by 1.1 percent, for a Test Year after-rates volume forecast
19
for non-destination entry Parcel Post of 114.911 million.
USPS-T-7
142
1997
1998
1999
2000
2001
2002
2003
2004
Table 31
Estimated Impact of Factors Affecting Non-Destination-Entry Parcel Post Volume, 1996 – 2007
Parcel Pos t Price
Other Factors
Total Change
Population
Nom inal
Inflation
UPS Prices Econom etric
Other
in Volum e
1.16%
-0.03%
0.80%
0.85%
3.65%
-3.41%
2.93%
1.20%
0.00%
0.52%
1.68%
-4.48%
1.31%
0.10%
1.15%
-1.30%
0.38%
2.63%
-11.00%
0.24%
-8.24%
1.23%
-3.54%
0.78%
1.62%
-20.58%
2.53%
-18.58%
1.27%
-1.26%
0.95%
1.32%
18.62%
0.92%
22.42%
1.44%
-2.79%
0.70%
1.96%
7.37%
2.18%
11.06%
1.27%
-5.15%
0.66%
1.03%
0.26%
-0.81%
-2.84%
1.24%
-1.41%
0.70%
1.39%
1.82%
0.40%
4.19%
1996 - 2004
Total
Avg per Year
10.39%
1.24%
-14.56%
-1.95%
5.62%
0.69%
13.18%
1.56%
-9.01%
-1.17%
3.27%
0.40%
5.94%
0.72%
2001 - 2004
Total
Avg per Year
4.00%
1.31%
-9.09%
-3.13%
2.08%
0.69%
4.45%
1.46%
9.62%
3.11%
1.76%
0.58%
12.43%
3.98%
Forecas t
1.23%
1.13%
1.11%
0.00%
0.00%
0.00%
0.80%
0.65%
0.69%
0.78%
0.89%
-0.07%
0.05%
0.12%
-0.30%
0.09%
-0.19%
0.00%
2.98%
2.62%
1.44%
3.50%
1.15%
0.00%
0.00%
2.16%
0.72%
1.60%
0.53%
-0.13%
-0.04%
-0.09%
-0.03%
7.20%
2.34%
After-Rates Volum e Forecas t
2005
1.23%
2006
1.13%
2007
1.11%
0.00%
-1.12%
-0.88%
0.80%
0.65%
0.69%
0.78%
0.89%
-0.07%
0.05%
0.12%
-0.30%
0.09%
-0.19%
0.00%
2.98%
1.47%
0.55%
-1.99%
-0.67%
2.16%
0.72%
1.60%
0.53%
-0.13%
-0.04%
-0.09%
-0.03%
5.07%
1.66%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
2004 - 2007
Total
Avg per Year
3.50%
1.15%
USPS-T-7
143
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 31 above.
5
The non-destination-entry Parcel Post equation does not include any macro-
6
economic variables. Historically, non-destination-entry Parcel Post volume has been
7
driven primarily by competitive factors which tend to overwhelm more subtle impacts of
8
the economy as a whole.
9
The UPS cross-price variable used in this equation is a weighted average of
10
published residential rates for UPS Ground. The weights used to construct this price
11
index are non-destination-entry Parcel Post billing determinants, the same as are used
12
to construct the non-destination-entry Parcel Post price index. The estimated cross-
13
price elasticity of non-destination-entry Parcel Post mail with respect to UPS prices is
14
0.443 (t-statistic of 1.542).
15
UPS published rates, as measured in this way, have consistently increased by more
16
than the rate of inflation over the past decade. This has served to increase non-
17
destination-entry Parcel Post volume by approximately 1.6 percent per year over this
18
time period. For the forecast period, UPS rates are assumed to increase at the rate of
19
inflation in January of each year. The timing of these increases is consistent with when
20
UPS has historically raised their rates.
21
The own-price elasticity of non-destination-entry Parcel Post mail was calculated to
22
be equal to -0.382 (t−statistic of -1.811). This gives non-destination-entry Parcel Post
23
the lowest own-price elasticity of any of the package delivery products discussed in this
24
section of my testimony. This is consistent with my earlier observation that non-
USPS-T-7
144
1
destination-entry Parcel Post appears to have established something of a niche in the
2
package delivery market.
3
The Postal price impacts shown in Table 31 above are the result of changes in
4
nominal prices. Prices enter the demand equations developed here in real terms,
5
however. The impact of inflation reported in Table 31 measures the impact that a
6
change in real Postal prices, in the absence of nominal rate changes, has on the
7
volume of non-destination-entry Parcel Post mail.
8
9
10
Other econometric variables include seasonal variables and several dummy
variables. A more detailed look at the econometric demand equation for nondestination-entry Parcel Post follows.
(b) Econometric Demand Equation
11
12
The demand equation for non-destination-entry Parcel Post mail volume in this case
13
models non-destination-entry Parcel Post mail volume per adult per delivery day as a
14
function of the following explanatory variables:
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
·
Seasonal Variables
·
Current and three lags of a fixed-weight price index of published residential
UPS Ground rates, constructed using weights calculated using 2004 nondestination-entry Parcel Post billing determinants
·
Dummy variable for the UPS strike in the summer of 1997, equal to one in
1997Q4, zero elsewhere
·
Dummy variable for the introduction of delivery confirmation, equal to one
starting in the spring of 1999
When delivery confirmation was introduced, electronic delivery confirmation was made
available for Priority Mail at no additional charge. Delivery confirmation for Parcel Post
was only available manually at a cost of 40 cents. Because of this difference, Priority
Mail with electronic delivery confirmation was actually cheaper than Parcel Post with
delivery confirmation for much mail. This led to a modest shift of some mail from Parcel
Post, primarily non-destination-entry Parcel Post, to Priority Mail at this time.
USPS-T-7
145
1
2
3
4
5
6
7
8
9
10
11
12
·
Dummy variable equal to one since the implementation of R2000-1 in 2001Q2
In R2000-1, the Postal Service began to allow mail weighing less than one pound to be
mailed as Parcel Post. Prior to this, mail weighing less than one pound had to be
mailed at First-Class or Priority Mail rates. As indicated in Table 29 above, nondestination-entry Parcel Post volume increased by 33.4 percent in the first four quarters
after the implementation of R2000-1. This dummy variable measures the extent to
which this was the result of this rule change.
·
Current and three lags of the price of non-destination-entry Parcel Post
Details of the econometric demand equation are shown in Table 32 below. A
13
detailed description of the econometric methodologies used to obtain these results can
14
be found in Section III below.
USPS-T-7
146
1
2
3
TABLE 32
ECONOMETRIC DEMAND EQUATION FOR NON-DESTINATION ENTRY PARCEL POST
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.382
-1.811
Current
-0.081
-0.284
Lag 1
-0.034
-0.090
Lag 2
-0.237
-0.677
Lag 3
-0.030
-0.092
UPS Ground Price Elasticity
Long-Run
0.443
1.542
Current
0.021
0.016
Lag 1
0.117
0.066
Lag 2
0.177
0.138
Lag 3
0.128
0.168
Dummy for UPS Strike (1997Q4)
0.091
0.946
Introduction of Delivery Confirmation
-0.362
-10.96
Dummy for R2000-1
0.294
11.63
Seasonal Coefficients
October – December
0.093
0.613
January – March
0.159
1.241
April – June
-0.306
-1.970
Quarter 1 (October – December)
0.140
2.563
Quarter 2 (January – March)
-0.127
-3.895
Quarter 3 (April – June)
0.173
2.458
Quarter 4 (July – September)
-0.186
-1.820
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.264783
Quarter 2 (January – March)
1.033659
Quarter 3 (April – June)
0.877309
Quarter 4 (July – September)
0.832002
REGRESSION DIAGNOSTICS
Sample Period
1996Q1 – 2005Q1
Autocorrelation Coefficients
AR-4: -0.689
Degrees of Freedom
14
Mean-Squared Error
0.002162
Adjusted R-Squared
0.942
4
5
USPS-T-7
147
1
2
3
4
5
6
7
8
9
iii. Destination Entry Parcel Post Mail
(a) Factors Affecting Destination Entry Parcel Post Mail
Destination-entry Parcel Post volume was found to be principally affected by the
following variables:
Time Trend Starting in 2003Q4
Price of UPS and FedEx Ground
Price of destination-entry Parcel Post mail
The effect of these variables on destination-entry Parcel Post volume over the past
10
ten years is shown in Table 33 on the next page. Table 33 also shows the projected
11
impacts of these variables through GFY 2007.
12
The Test Year before-rates volume forecast for destination entry Parcel Post is
13
237.852 million pieces, a 10.5 percent decline from GFY 2004. The Postal Service’s
14
proposed rates in this case are predicted to reduce the Test Year volume of destination
15
entry Parcel Post by 6.9 percent, for a Test Year after-rates volume forecast for
16
destination entry Parcel Post of 221.536 million.
USPS-T-7
148
1999
2000
2001
2002
2003
2004
Table 33
Estimated Impact of Factors Affecting Destination-Entry Parcel Post Volume, 1998 – 2007
UPS & FedEx
Total Change
Parcel Pos t Price
Other Factors
Nom inal
Inflation
Prices Econom etric
Other
in Volum e
Population Tim e Trends
1.21%
0.00%
-0.18%
1.47%
3.75%
0.82%
-0.23%
6.97%
1.36%
0.00%
0.36%
3.32%
2.00%
0.39%
-0.40%
7.19%
1.24%
0.00%
-3.74%
3.29%
2.09%
-1.74%
3.53%
4.53%
1.30%
0.00%
-1.35%
1.96%
3.09%
-0.89%
-0.70%
3.38%
1.35%
-0.37%
2.14%
2.88%
1.28%
1.08%
-1.86%
6.61%
1.18%
-4.81%
1.34%
2.80%
1.04%
-5.67%
-1.28%
-5.60%
1998 – 2004
Total
Avg per Year
7.89%
1.27%
-5.17%
-0.88%
-1.53%
-0.26%
16.76%
2.62%
13.97%
2.20%
-6.03%
-1.03%
-1.03%
-0.17%
24.69%
3.75%
2001 – 2004
Total
Avg per Year
3.89%
1.28%
-5.17%
-1.75%
2.12%
0.70%
7.83%
2.54%
5.49%
1.80%
-5.51%
-1.87%
-3.79%
-1.28%
4.04%
1.33%
Forecas t
1.20%
1.08%
1.09%
-6.10%
-6.01%
-6.12%
0.10%
0.00%
0.00%
2.78%
2.31%
2.73%
-4.56%
-2.32%
0.00%
-0.05%
-0.45%
-0.48%
1.75%
-0.17%
0.00%
-5.11%
-5.64%
-2.97%
3.41%
1.12%
-17.15%
-6.08%
0.10%
0.03%
8.03%
2.61%
-6.78%
-2.31%
-0.98%
-0.33%
1.58%
0.52%
-13.12%
-4.58%
After-Rates Volum e Forecas t
2005
1.20%
2006
1.08%
2007
1.09%
-6.10%
-6.01%
-6.12%
0.10%
-6.86%
0.00%
2.78%
2.31%
2.73%
-4.56%
-2.32%
0.00%
-0.05%
-0.45%
-0.48%
1.75%
-0.17%
0.00%
-5.11%
-12.11%
-2.97%
-17.15%
-6.08%
-6.76%
-2.31%
8.03%
2.61%
-6.78%
-2.31%
-0.98%
-0.33%
1.58%
0.52%
-19.08%
-6.81%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
2004 - 2007
Total
Avg per Year
3.41%
1.12%
USPS-T-7
149
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 33 above.
5
The destination-entry Parcel Post demand equation includes a negative time trend
6
starting in 2003Q4. This trend is included to reflect new competitive pressures over this
7
time period, including the introduction of UPS Basic. This trend explains a decline in
8
destination-entry Parcel Post volume of approximately 6 percent per year through the
9
forecast period.
10
Like non-destination-entry Parcel Post, the destination-entry Parcel Post equation
11
does not include any macroeconomic variables, as destination-entry Parcel Post volume
12
has been driven historically primarily by continuous volume increases for its first decade
13
of existence and by competitive factors which tend to overwhelm more subtle impacts of
14
the economy as a whole since then.
15
The combined average revenue per piece for UPS and FedEx Ground is used to
16
measure the effect of competitor prices on destination-entry Parcel Post volume. The
17
estimated cross-price elasticity of destination-entry Parcel Post mail with respect to UPS
18
and FedEx prices is 1.821 (t-statistic of 3.335). The own-price elasticity of non-
19
destination-entry Parcel Post mail was calculated to be equal to -1.351 (t−statistic of
20
−2.044).
21
The Postal price impacts shown in Table 33 above are the result of changes in
22
nominal prices. Prices enter the demand equations developed here in real terms,
23
however. The impact of inflation reported in Table 33 measures the impact that a
24
change in real Postal prices, in the absence of nominal rate changes, has on the
25
volume of non-destination-entry Parcel Post mail.
USPS-T-7
150
1
Other econometric variables include seasonal variables and several dummy
2
variables. A more detailed look at the econometric demand equation for destination-
3
entry Parcel Post follows.
(b) Econometric Demand Equation
4
5
The demand equation for destination-entry Parcel Post mail volume in this case
6
models destination-entry Parcel Post mail volume per adult per delivery day as a
7
function of the following explanatory variables:
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
·
Seasonal Variables
·
Linear Time Trend starting in 2003Q4
·
Dummy variable equal to one starting in 2003Q4
·
Dummy variable equal to one since R97-1, to reflect the introduction of DSCF
and DDU discounts at this time
·
Dummy variable equal to one since R2001-1, to reflect the introduction of a
separate rate for Parcel Post weighing less than one pound. Prior to this, the
lowest price for Parcel Post applied to all mail weighing up to two pounds.
·
Dummy variable for September 11, 2001, and its immediate aftermath
·
Current and three lags of average revenue per piece for UPS and FedEx
Ground Delivery Services
·
Current price of destination-entry Parcel Post
Details of the econometric demand equation are shown in Table 34 below. A
28
detailed description of the econometric methodologies used to obtain these results can
29
be found in Section III below.
USPS-T-7
151
1
2
3
TABLE 34
ECONOMETRIC DEMAND EQUATION FOR DESTINATION-ENTRY PARCEL POST MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (current only)
-1.351
-2.044
UPS / FedEx Ground Market Prices
Long-Run
1.821
3.335
0.365
0.607
Current
Lag 1
0.530
0.923
0.587
2.444
Lag 2
Lag 3
0.338
1.409
Time Trend Starting in 2003Q4
-0.016
-1.956
Dummy Variable Starting in 2003Q4
-0.097
-3.828
Dummy for 2002Q1 (9/11 Effect)
-0.040
-1.423
Dummy for R97-1
-0.368
-1.911
Dummy for R2001-1
0.114
3.187
Seasonal Coefficients
0.742
8.540
October – December
0.162
2.877
January – March
0.209
1.555
April – June
Quarter 1 (October – December)
-0.144
-3.733
Quarter 2 (January – March)
0.050
1.032
Quarter 3 (April – June)
-0.051
-0.667
Quarter 4 (July – September)
0.145
2.129
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.353250
Quarter 2 (January – March)
0.920394
Quarter 3 (April – June)
0.872249
Quarter 4 (July – September)
0.861214
REGRESSION DIAGNOSTICS
Sample Period
1998Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
12
Mean-Squared Error
0.000495
Adjusted R-Squared
0.987
4
5
USPS-T-7
152
1
2
c. Other Package Services
In addition to Parcel Post, there are three additional subclasses in the Package
3
Services class: Bound Printed Matter, Media Mail, and Library Rate Mail. Each of these
4
subclasses offer reduced rates for mail which satisfies certain content restrictions.
5
Bound Printed Matter refers to any mail that is bound and printed, and weighs
6
between one and fifteen pounds. Generally, Bound Printed Matter falls into one of three
7
categories: catalogs, books (including telephone books in some areas), and direct-mail
8
advertising weighing sixteen ounces or more. The Media Mail subclass is reserved for
9
books, tapes, and CDs. The Library Rate subclass is a preferred subclass, generally
10
corresponding to the Media Mail subclass, available to libraries and certain other
11
institutions. In this testimony, a single demand equation is estimated for the combined
12
volume of Media Mail and Library Rate mail.
13
14
15
16
i. Overview of Bound Printed Matter and Media Mail
(a) History of Bound Printed Matter and Media Mail Subclasses
Prior to 1976, the Bound Printed Matter subclass was called the Catalog subclass,
17
and was composed entirely of catalogs. Beginning on or around the fourth quarter of
18
1976, an informal rule change occurred, whereby certain Post Offices began to allow
19
books, which had previously been sent as Media Mail (then called special rate mail), to
20
be sent as Bound Printed Matter with the inclusion of a single page of advertising. This
21
rule was gradually adopted by most Post Offices over the next several years.
22
In most cases, Bound Printed Matter rates were, and still are, less expensive than
23
Media Mail rates. However, Bound Printed Matter rates are zoned, whereas Media Mail
24
rates are unzoned. Thus, in order for mailers to shift from the Media Mail to the Bound
25
Printed Matter subclass, mailers had to switch from unzoned rates to zoned rates. This
26
structural adaptation, along with an apparent lag in realization by mailers of the
USPS-T-7
153
1
existence of this rule change, made it difficult for mailers to shift immediately from Media
2
Mail to Bound Printed Matter.
3
Shifts between these two subclasses were particularly erratic in the first two years
4
after this rule change was first implemented gradually. It was decided that it would be
5
best econometrically, therefore, to avoid this early period entirely. Consequently, in
6
R97-1, the demand equations for Bound Printed Matter and Media Mail volume were
7
modeled using data starting in 1979Q1, allowing two full years for Media Mailers to
8
begin to adapt to the enhanced opportunities available through Bound Printed Matter.
9
Even after this time period, however, gradual migration from Media Mail into Bound
10
Printed Matter continued. In R97-1 and R2000-1, this effect was modeled by including
11
logistic market penetration variables in the demand equations for Bound Printed Matter
12
and Media Mail volumes. The market penetration variable in the Bound Printed Matter
13
equation was positive to reflect market penetration into Bound Printed Matter, while the
14
market penetration variable in the Media Mail equation was negative to reflect market
15
penetration out of the Media Mail subclass.
16
For the present case, the Bound Printed Matter and Media Mail sample periods have
17
been further truncated to begin in 1988Q1. This provides an opportunity to remove the
18
market penetration variables from these equations.
19
20
The history of Bound Printed Matter, Media Mail, and Library Rate mail volumes are
summarized in Tables 35 and 36 below.
USPS-T-7
154
Ta ble 35
Othe r Pa cka ge Se rvice Volume s
(millions of pieces)
Bound Printed Matter
Media Mail
Library Rate
1976
75.398
299.748
36.490
1977
85.106
273.570
60.319
1978
87.273
280.202
72.005
1979
101.441
246.310
61.449
1980
114.905
247.025
59.654
1981
118.479
224.725
58.996
1982
164.728
206.550
55.610
1983
166.956
180.595
55.130
1984
193.936
187.531
54.388
1985
210.715
165.176
45.624
1986
247.811
166.932
42.742
1987
253.626
163.535
51.921
1988
303.319
173.007
47.191
1989
311.707
153.341
39.796
1990
344.468
149.462
40.959
1991
359.520
154.168
40.263
1992
386.450
164.635
43.044
1993
353.968
168.556
39.645
1994
413.166
194.084
36.694
1995
463.774
214.199
29.503
1996
511.069
191.838
30.467
1997
513.524
204.733
27.308
1998
492.665
190.534
27.604
1999
488.627
200.243
28.010
2000
560.218
215.934
28.112
2001
556.726
163.081
23.973
2002
507.702
174.639
20.154
2003
544.807
179.181
17.583
2004
553.666
186.229
16.415
1
no te: D ata s ho wn are fo r P o stal F is cal Years thro ugh 1999, by Go v ernm ent Fis cal Years 2000 - 2004
USPS-T-7
155
Ta ble 36
Pe rce nta ge Cha nge in Volum e s
Othe r Packa ge Se rvice s
Bound Printed Matter
Media Mail
Library Rate
1977
12.88%
-8.73%
65.30%
1978
2.55%
2.42%
19.37%
1979
16.23%
-12.10%
-14.66%
1980
13.27%
0.29%
-2.92%
1981
3.11%
-9.03%
-1.10%
1982
39.04%
-8.09%
-5.74%
1983
1.35%
-12.57%
-0.86%
1984
16.16%
3.84%
-1.35%
1985
8.65%
-11.92%
-16.11%
1986
17.60%
1.06%
-6.32%
1987
2.35%
-2.03%
21.48%
1988
19.59%
5.79%
-9.11%
1989
2.77%
-11.37%
-15.67%
1990
10.51%
-2.53%
2.92%
1991
4.37%
3.15%
-1.70%
1992
7.49%
6.79%
6.91%
1993
-8.41%
2.38%
-7.90%
1994
16.72%
15.15%
-7.44%
1995
12.25%
10.36%
-19.60%
1996
10.20%
-10.44%
3.27%
1997
0.48%
6.72%
-10.37%
1998
-4.06%
-6.94%
1.08%
1999
-0.82%
5.10%
1.47%
2000
11.59%
8.02%
0.08%
2001
-0.62%
-24.48%
-14.72%
2002
-8.81%
7.09%
-15.93%
2003
7.31%
2.60%
-12.76%
2004
1.63%
3.93%
-6.64%
1
no te: D ata s ho wn are fo r P o stal F is cal Years thro ugh 2000, by Go v ernm ent Fisc al Years 2001 - 2004
USPS-T-7
156
(b) Factors affecting Demand for Bound Printed Matter and Media
Mail
1
2
3
4
The demand for package delivery services will be largely driven by the demand for
5
the goods being delivered. In the cases of Bound Printed Matter and Media Mail, this
6
relationship is modeled through the inclusion of mail-order retail sales as an explanatory
7
variable.
8
9
Bound Printed Matter and Media Mail receive somewhat preferred rates from the
Postal Service based on their content. Because of this, these subclasses face less
10
price-based competition from other package delivery companies than Priority Mail and
11
Parcel Post. Because of this, competitor prices are not included in the Bound Printed
12
Matter and Media Mail equations. Bound Printed Matter and Media Mail do, however,
13
include cross-price variables with respect to each other in their demand equations.
14
The specific demand equations for Bound Printed Matter and Media Mail are
15
presented in more detail below.
ii. Bound Printed Matter Demand Equation
16
17
18
19
(a) Factors Affecting the Bound Printed Matter Demand Equation
The demand for Bound Printed Matter would, of course, be expected to be a function
20
of the price of Bound Printed Matter. In addition to postage, however, the cost of
21
producing Bound Printed Matter includes paper and printing costs. To account for these
22
non-postage costs, the Bound Printed Matter equation includes the producer price index
23
for direct-mail advertising. This variable was described in more detail earlier in the
24
discussion of Standard Enhanced Carrier Route Mail.
25
26
Overall, Bound Printed Matter volume was found to be affected by the following
variables:
USPS-T-7
157
1
2
3
4
5
Mail-Order Retail Sales
Price of Direct-Mail Advertising
Prices of Media Mail and Bound Printed Matter
The effect of these variables on Bound Printed Matter volume over the past ten
6
years is shown in Table 37 on the next page. Table 37 also shows the projected
7
impacts of these variables through GFY 2007.
8
The Test Year before-rates volume forecast for Bound Printed Matter is 598.339
9
million pieces, an 8.1 percent increase from GFY 2004. The Postal Service’s proposed
10
rates in this case are predicted to increase the Test Year volume of Bound Printed
11
Matter by 1.3 percent, for a Test Year after-rates volume forecast for Bound Printed
12
Matter of 605.996 million.
USPS-T-7
158
Other
-7.30%
5.42%
-3.94%
7.52%
-1.16%
2.94%
-5.29%
-3.05%
3.84%
-3.78%
Total Change
in Volum e
12.25%
10.20%
0.48%
-4.06%
-0.82%
14.65%
-0.62%
-8.81%
7.31%
1.63%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.26%
1.18%
1.21%
1.15%
1.17%
1.43%
1.22%
1.26%
1.34%
1.22%
1994 - 2004
Total
Avg per Year
13.16%
1.24%
37.14%
3.21%
7.54%
0.73%
-16.61%
-1.80%
11.96%
1.14%
1.64%
0.16%
-10.18%
-1.07%
-5.79%
-0.59%
34.01%
2.97%
2001 - 2004
Total
Avg per Year
3.87%
1.27%
2.90%
0.96%
4.76%
1.56%
-11.52%
-4.00%
3.27%
1.08%
0.53%
0.18%
-0.18%
-0.06%
-3.14%
-1.06%
-0.55%
-0.18%
Forecas t
1.22%
1.14%
1.12%
3.78%
2.64%
1.97%
0.06%
0.44%
0.72%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.20%
0.22%
0.08%
-0.29%
0.01%
-0.15%
-1.69%
0.17%
0.00%
3.24%
4.68%
3.78%
3.52%
1.16%
8.61%
2.79%
1.23%
0.41%
0.00%
0.00%
0.00%
0.00%
0.51%
0.17%
-0.42%
-0.14%
-1.53%
-0.51%
12.16%
3.90%
After-Rates Volum e Forecas t
2005
1.22%
2006
1.14%
2007
1.12%
3.78%
2.64%
1.97%
0.06%
0.44%
0.72%
0.00%
-1.52%
-1.79%
0.00%
2.84%
0.00%
0.20%
0.22%
0.08%
-0.29%
0.01%
-0.15%
-1.69%
0.17%
0.00%
3.24%
6.02%
1.93%
8.61%
2.79%
1.23%
0.41%
-3.28%
-1.10%
2.84%
0.94%
0.51%
0.17%
-0.42%
-0.14%
-1.53%
-0.51%
11.56%
3.71%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 37
Estimated Impact of Factors Affecting Bound Printed Matter Volume, 1994 – 2007
Mail-Order
Price of
Pos tal Prices
Other Factors
Retail Sales Paper/Printing Bound Printed
Media Mail
Inflation Econom etric
2.51%
-0.39%
-1.78%
6.55%
0.19%
11.69%
5.65%
-0.44%
-4.82%
2.64%
0.26%
0.28%
4.58%
1.29%
-0.73%
0.01%
0.17%
-1.89%
4.24%
0.57%
0.00%
0.03%
0.34%
-16.16%
4.44%
0.42%
0.65%
-2.69%
-0.02%
-3.41%
6.22%
1.09%
2.33%
-1.48%
-0.18%
1.61%
1.75%
0.10%
-1.39%
3.36%
0.34%
-0.50%
0.35%
1.51%
-7.50%
2.17%
0.37%
-3.86%
-0.48%
2.04%
-3.69%
1.08%
0.05%
3.10%
3.04%
1.14%
-0.68%
0.00%
0.11%
0.71%
2004 - 2007
Total
Avg per Year
3.52%
1.16%
USPS-T-7
159
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 37 above.
5
Bound Printed Matter has a retail sales elasticity of 0.424 (t-statistic of 5.067),
6
meaning that a 10 percent increase in mail-order retail sales will lead to a 4.24 percent
7
increase in the volume of Bound Printed Matter.
8
9
Bound Printed Matter has an elasticity with respect to the price of direct-mail
advertising of -0.716 (t-statistic of -1.400). Falling printing costs have added nearly 5
10
percent to Bound Printed Matter volume over the past three years. This has helped to
11
partially offset the negative impact of the R2000-1 and R2001-1 rate increases.
12
The cross-price elasticity of Bound Printed Matter with respect to Media Mail is
13
estimated to be equal to 0.510 (t-statistic of 2.707). This elasticity is stochastically
14
constrained from the Media Mail demand equation using the Slutsky-Schultz symmetry
15
condition. The Slutsky-Schultz symmetry condition is described in detail in Section III
16
below. The own-price elasticity of Bound Printed Matter is calculated to be equal to
17
-0.604 (t−statistic of -2.165).
18
The Postal price impacts shown in Table 37 above are the result of changes in
19
nominal prices. Prices enter the demand equations developed here in real terms,
20
however. The impact of inflation reported in Table 37 measures the impact that a
21
change in real Postal prices, in the absence of nominal rate changes, has on the
22
volume of Bound Printed Matter mail.
23
24
Other econometric variables include seasonal variables, a dummy variable to
account for the temporary impact of the September 11, 2001, terrorist attacks, and two
USPS-T-7
160
1
other dummy variables. A more detailed look at the econometric demand equation for
2
Bound Printed Matter follows.
(b) Econometric Demand Equation
3
4
The demand equation for Bound Printed Matter in this case models Bound Printed
5
Matter volume per adult per delivery day as a function of the following explanatory
6
variables:
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
·
Seasonal Variables
·
Mail-Order Retail Sales
·
Producer Price Index for Direct-Mail Advertising
·
Dummy variable equal to one starting in 1998Q1
This dummy variable is included in the Bound Printed Matter demand equation to
account for an otherwise unexplained decline in Bound Printed Matter of approximately
15 percent since 1998.
·
Dummy variable for the cancellation of the main Sears catalog, equal to one
from 1993Q2 – 1994Q1, zero elsewhere
·
Dummy variable for September 11, 2001, equal to one in 2002Q1, zero
elsewhere
·
Price of Media and Library Rate Mail
The cross-price elasticity of Bound Printed Matter mail with respect to the price of Media
Mail is stochastically constrained from the Media Mail equation using the SlutskySchultz symmetry condition. The Slutsky-Schultz symmetry condition is described in
detail in Section III below.
·
Current and four lags of the price of Bound Printed Matter
Details of the econometric demand equation are shown in Table 38 below. A
34
detailed description of the econometric methodologies used to obtain these results can
35
be found in Section III below.
USPS-T-7
161
1
2
3
TABLE 38
ECONOMETRIC DEMAND EQUATION FOR BOUND PRINTED MATTER
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.604
-2.165
Current
0.000
(N/A)
Lag 1
-0.274
-0.539
Lag 2
-0.117
-0.170
Lag 3
-0.012
-0.018
Lag 4
-0.200
-0.417
Media Mail Price
0.510
2.707
Mail-Order Retail Sales
0.424
5.067
Price of Direct-Mail Advertising
-0.716
-1.400
Dummy Since 1998Q1
-0.157
-3.326
Dummy for Cancellation of Sears
-0.128
-2.914
Catalog (1993Q2 – 1994Q1)
Dummy for 2002Q1 (9/11 Effect)
-0.152
-1.347
Seasonal Coefficients
0.745
0.549
September 16 – 30
-1.437
-1.353
October
November 1 – December 10
-1.095
-1.186
December 11 – 23
-0.987
-0.726
December 24 – February
-0.910
-1.465
March
-0.172
-0.153
April 1 – 15
-3.122
-1.366
April 16 – May
-1.099
-1.357
-2.849
-1.227
June
Quarter 1 (October – December)
0.280
3.464
Quarter 2 (January – March)
-0.262
-1.751
Quarter 3 (April – June)
0.932
1.880
Quarter 4 (July – September)
-0.949
-1.876
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.029035
Quarter 2 (January – March)
1.024889
Quarter 3 (April – June)
0.841635
Quarter 4 (July – September)
1.109419
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
46
Mean-Squared Error
0.006096
Adjusted R-Squared
0.886
USPS-T-7
162
iii. Media and Library Rate Mail Demand Equation
1
2
3
4
5
6
7
8
9
10
(a) Factors Affecting the Media and Library Rate Demand
Equation
Media and Library Rate Mail volume was found to be affected by the following
variables:
Mail-Order Retail Sales
Prices of Bound Printed Matter and Media Mail
The effect of these variables on Media and Library Rate Mail volume over the past
11
ten years is shown in Table 39 on the next page. Table 39 also shows the projected
12
impacts of these variables through GFY 2007.
13
The Test Year before-rates volume forecast for Media and Library Rate Mail is
14
209.679 million pieces, a 3.5 percent increase from GFY 2004. The Postal Service’s
15
proposed rates in this case are predicted to reduce the Test Year volume of Media and
16
Library Rate Mail by 0.6 percent, for a Test Year after-rates volume forecast for Media
17
and Library Rate Mail of 208.348 million.
USPS-T-7
163
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Table 39
Estimated Impact of Factors Affecting Media and Library Rate Mail Volume, 1994 – 2007
Mail-Order
Total Change
Pos tal Prices
Other Factors
Media Mail Bound Printed
Inflation Econom etric
Other
in Volum e
Population Retail Sales
1.22%
1.47%
-7.40%
4.97%
0.63%
-0.34%
5.47%
5.60%
1.10%
3.07%
-7.15%
2.03%
0.58%
-0.61%
-7.58%
-8.78%
1.20%
2.72%
-0.10%
0.00%
0.61%
1.01%
-1.10%
4.38%
1.14%
2.44%
-0.03%
0.00%
0.41%
-11.96%
2.67%
-5.99%
1.21%
2.67%
3.38%
-1.94%
0.25%
0.58%
-1.49%
4.64%
1.40%
3.73%
3.99%
-1.03%
0.56%
-0.59%
-1.19%
6.92%
1.08%
1.13%
-2.82%
4.41%
0.63%
-25.13%
-1.91%
-23.35%
1.35%
-0.31%
-4.60%
5.51%
0.54%
1.57%
0.27%
4.14%
1.30%
-0.32%
-2.05%
1.62%
0.46%
0.10%
-0.07%
1.01%
1.24%
1.80%
-0.28%
0.00%
0.50%
2.81%
-3.02%
2.99%
1994 - 2004
Total
Avg per Year
12.96%
1.23%
19.90%
1.83%
-16.40%
-1.78%
16.36%
1.53%
5.30%
0.52%
-31.07%
-3.65%
-8.17%
-0.85%
-12.19%
-1.29%
2001 - 2004
Total
Avg per Year
3.95%
1.30%
1.16%
0.38%
-6.82%
-2.33%
7.22%
2.35%
1.52%
0.50%
4.53%
1.49%
-2.82%
-0.95%
8.33%
2.70%
Forecas t
1.20%
1.14%
1.11%
2.27%
1.56%
1.15%
0.00%
0.00%
0.00%
0.00%
0.00%
0.00%
0.60%
0.49%
0.50%
-0.49%
0.10%
-1.20%
-4.42%
1.12%
0.00%
-0.97%
4.49%
1.55%
3.49%
1.15%
5.06%
1.66%
0.00%
0.00%
0.00%
0.00%
1.60%
0.53%
-1.59%
-0.53%
-3.35%
-1.13%
5.07%
1.66%
After-Rates Volum e Forecas t
2005
1.20%
2006
1.14%
2007
1.11%
2.27%
1.56%
1.15%
0.00%
-3.30%
-0.86%
0.00%
2.76%
0.00%
0.60%
0.49%
0.50%
-0.49%
0.10%
-1.20%
-4.42%
1.12%
0.00%
-0.97%
3.82%
0.67%
5.06%
1.66%
-4.14%
-1.40%
2.76%
0.91%
1.60%
0.53%
-1.59%
-0.53%
-3.35%
-1.13%
3.50%
1.15%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
2004 - 2007
Total
Avg per Year
3.49%
1.15%
USPS-T-7
164
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 39 above.
5
Media and Library Rate Mail has a retail sales elasticity of 0.249 (t-statistic of 3.903),
6
meaning that a 10 percent increase in mail-order retail sales will lead to a 2.49 percent
7
increase in the volume of Media Mail.
8
The cross-price elasticity of Media Mail with respect to Bound Printed Matter is
9
estimated to be equal to 0.509 (t-statistic of 1.090). The own-price elasticity of Media
10
and Library Rate Mail was calculated to be equal to -0.796 (t−statistic of -2.135).
11
The Postal price impacts shown in Table 39 above are the result of changes in
12
nominal prices. Prices enter the demand equations developed here in real terms,
13
however. The impact of inflation reported in Table 39 measures the impact that a
14
change in real Postal prices, in the absence of nominal rate changes, has on the
15
volume of Media Mail.
16
Other econometric variables include seasonal variables, and two dummy variables.
17
A more detailed look at the econometric demand equation for Bound Printed Matter
18
follows.
(b) Econometric Demand Equation
19
20
The demand equation for Media and Library Rate Mail in this case models Media
21
and Library Rate Mail volume per adult per delivery day as a function of the following
22
explanatory variables:
23
24
25
26
·
Seasonal Variables
·
Mail-Order Retail Sales
USPS-T-7
165
1
2
3
4
5
6
7
8
9
10
11
12
13
14
·
Dummy variable equal to one starting in 1998Q1
This dummy variable is included in the Media Mail equation to account for an otherwise
unexplained decline in Media Mail volume of approximately 15 percent since 1998.
·
Dummy variable equal to one starting in 2001Q1
This dummy variable is included in the Media Mail equation to account for an otherwise
unexplained decline in Media Mail volume of approximately 30 percent since 2001.
·
Price of Bound Printed Matter
·
Current and four lags of the price of Media and Library Rate Mail
Details of the econometric demand equation are shown in Table 40 below. A
15
detailed description of the econometric methodologies used to obtain these results can
16
be found in Section III below.
USPS-T-7
166
1
2
3
TABLE 40
ECONOMETRIC DEMAND EQUATION FOR MEDIA AND LIBRARY RATE MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.796
-2.135
Current
-0.514
-1.188
Lag 1
-0.002
-0.006
Lag 2
-0.202
-0.432
Lag 3
-0.077
-0.252
Bound Printed Matter Price
0.509
1.090
Mail-Order Retail Sales
0.249
3.903
Dummy Since 1998Q1
-0.145
-3.345
Dummy Since 2001Q1
-0.309
-4.845
Seasonal Coefficients
September 16 – 30
-0.640
-1.038
October
1.768
2.521
November 1 – December 10
-0.693
-1.245
December 11 – 17
0.526
0.684
December 18 – 21
-5.209
-1.127
December 22 – 24
-1.176
-0.183
December 25 – 31
1.636
0.085
January – February
0.451
0.261
March – May
0.095
2.599
Quarter 1 (October – December)
0.019
0.017
Quarter 2 (January – March)
-0.191
-0.173
Quarter 3 (April – June)
0.023
0.754
Quarter 4 (July – September)
0.149
1.576
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.087643
Quarter 2 (January – March)
1.014446
Quarter 3 (April – June)
0.971293
Quarter 4 (July – September)
0.927557
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
48
Mean-Squared Error
0.005122
Adjusted R-Squared
0.768
4
USPS-T-7
167
1
2
3
F. Periodicals Mail
1. General Overview
The Periodicals Mail class is eligible for mail that is sent at regular intervals and
4
contains at least a minimum level of editorial (i.e., non-advertising) content. This
5
type of mail may include magazines, newspapers, journals, and newsletters. The
6
Periodicals Mail class is divided into four subclasses, Periodicals Regular and three
7
subclasses which offer preferred rates for certain eligible mailers. Periodicals within-
8
county mail is open to Periodicals which are sent within the same county as they are
9
printed. Periodicals nonprofit mail is open to Periodicals sent by qualified not-for-
10
profit organizations. Periodicals classroom mail is open to Periodicals sent to
11
educational institutions for educational purposes.
12
13
14
Periodicals Mail volumes since 1970 are shown in Table 41 below. Annual
percentage changes in volume are shown in Table 42.
In looking at Tables 41 and 42, it is apparent that Periodicals Mail volume growth
15
has been fairly modest throughout most of the past 30 years. In fact, from 1970 to
16
2004, total Periodicals mail volume actually declined by approximately 8 percent.
17
Total Periodicals Mail volume peaked in 1990 at 10.66 billion pieces, while
18
Periodicals regular rate mail volume peaked in 2000 at 7.25 billion pieces. From
19
2000 to 2004, Periodicals regular rate mail volume declined by 10.9 percent, an
20
average annual rate of 2.8 percent. Total Periodicals Mail volume has declined at
21
an average annual rate of 3.1 percent over this same time period.
USPS-T-7
168
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1
Regular Rate
5,970.662
6,019.930
5,804.333
5,653.185
5,719.654
5,633.860
5,613.723
5,566.403
5,645.916
5,805.582
5,736.325
5,696.010
5,816.571
5,935.799
6,065.635
6,330.263
6,537.416
6,525.971
6,578.137
6,516.084
6,811.320
6,980.415
6,604.694
6,826.551
6,830.920
6,887.174
6,950.136
7,196.851
7,153.952
7,205.661
7,250.346
7,113.293
6,787.814
6,517.359
6,462.075
Ta ble 41
Pe riodica ls Ma il Volum e
(millions of pieces)
W ithin County
Nonprofit
Classroom
1,739.723
2,125.735
104.457
1,709.765
2,257.599
87.175
1,644.889
2,254.418
77.453
1,552.021
2,242.085
69.072
1,458.083
2,262.144
61.743
1,436.681
2,449.099
73.260
1,404.271
2,264.599
57.370
1,353.909
2,283.626
65.696
1,282.353
2,369.475
59.069
1,209.718
2,240.673
64.062
1,360.512
2,914.044
74.856
1,315.717
2,808.166
55.529
1,261.045
2,361.083
37.386
1,300.757
1,892.787
40.700
1,357.328
2,036.800
31.108
1,829.727
2,128.915
45.308
1,732.644
2,240.605
37.212
1,476.494
2,240.023
48.131
1,476.514
2,292.619
60.811
1,460.602
2,472.047
55.213
1,378.944
2,430.412
38.383
1,184.679
2,184.299
42.941
1,176.758
2,405.420
63.250
1,054.291
2,288.623
94.389
1,006.731
2,264.157
79.214
893.943
2,280.332
64.343
873.823
2,211.104
59.124
945.056
2,153.719
62.327
920.217
2,139.225
60.682
894.488
2,136.552
59.816
897.069
2,153.400
63.969
884.908
2,070.942
63.403
849.911
1,991.459
60.575
793.521
1,948.261
60.764
760.020
1,850.746
62.430
note: Data show n are f or Postal Fiscal Y ears through 1999, by Government Fiscal Y ears 2000 - 2004
Total
9,940.577
10,074.468
9,781.093
9,516.363
9,501.624
9,592.899
9,339.962
9,269.633
9,356.814
9,320.035
10,085.737
9,875.423
9,476.085
9,170.044
9,490.871
10,334.213
10,547.877
10,290.619
10,408.081
10,503.946
10,659.058
10,392.334
10,250.122
10,263.854
10,181.021
10,125.792
10,094.188
10,357.953
10,274.076
10,296.517
10,364.784
10,132.547
9,689.758
9,319.905
9,135.272
USPS-T-7
169
1
Ta ble 42
Pe rce nta nge Cha nge in Pe riodica ls Ma il Volume
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2
Regular Rate
0.83%
-3.58%
-2.60%
1.18%
-1.50%
-0.36%
-0.84%
1.43%
2.83%
-1.19%
-0.70%
2.12%
2.05%
2.19%
4.36%
3.27%
-0.18%
0.80%
-0.94%
4.53%
2.48%
-5.38%
3.36%
0.06%
0.82%
0.91%
3.55%
-0.60%
0.72%
-0.86%
-1.89%
-4.58%
-3.98%
-0.85%
W ithin County
-1.72%
-3.79%
-5.65%
-6.05%
-1.47%
-2.26%
-3.59%
-5.29%
-5.66%
12.47%
-3.29%
-4.16%
3.15%
4.35%
34.80%
-5.31%
-14.78%
0.00%
-1.08%
-5.59%
-14.09%
-0.67%
-10.41%
-4.51%
-11.20%
-2.25%
8.15%
-2.63%
-2.80%
-0.98%
-1.36%
-3.95%
-6.63%
-4.22%
Nonprofit
6.20%
-0.14%
-0.55%
0.89%
8.26%
-7.53%
0.84%
3.76%
-5.44%
30.05%
-3.63%
-15.92%
-19.83%
7.61%
4.52%
5.25%
-0.03%
2.35%
7.83%
-1.68%
-10.13%
10.12%
-4.86%
-1.07%
0.71%
-3.04%
-2.60%
-0.67%
-0.12%
-0.45%
-3.83%
-3.84%
-2.17%
-5.01%
Classroom
-16.54%
-11.15%
-10.82%
-10.61%
18.65%
-21.69%
14.51%
-10.09%
8.45%
16.85%
-25.82%
-32.67%
8.86%
-23.57%
45.65%
-17.87%
29.34%
26.34%
-9.21%
-30.48%
11.88%
47.30%
49.23%
-16.08%
-18.77%
-8.11%
5.42%
-2.64%
-1.43%
5.80%
-0.88%
-4.46%
0.31%
2.74%
note: Data show n are f or Postal Fiscal Y ears through 2000, by Government Fiscal Y ears 2001 - 2004
Total
1.35%
-2.91%
-2.71%
-0.15%
0.96%
-2.64%
-0.75%
0.94%
-0.39%
8.22%
-2.09%
-4.04%
-3.23%
3.50%
8.89%
2.07%
-2.44%
1.14%
0.92%
1.48%
-2.50%
-1.37%
0.13%
-0.81%
-0.54%
-0.31%
2.61%
-0.81%
0.22%
-0.75%
-2.24%
-4.37%
-3.82%
-1.98%
USPS-T-7
170
1
2. Factors Affecting Demand for Periodicals Mail
2
The demand for Periodicals mail is a derived demand, which is derived from the
3
demand of consumers for magazines and newspapers. Those factors which
4
influence the demand for newspapers and magazines would therefore be expected
5
to be the principal drivers of the demand for Periodicals mail.
6
The factors which would be expected to influence the demand for newspapers and
7
magazines are drawn from basic micro-economic theory. These factors include
8
consumer income, the price of newspapers and magazines, and the demand for goods
9
which may serve as substitutes for newspapers and magazines.
10
The Periodicals demand equations used here include total private employment. This
11
variable is a proxy for consumer income and tracks the business cycle in a fairly
12
obvious way. Employment worked better econometrically at explaining Periodicals mail
13
volumes than other variables tested, including personal disposable income,
14
consumption expenditures, and retail sales. The use of employment as a macro-
15
economic variable in my demand equations is also discussed above in connection with
16
First-Class Mail.
17
The price of newspapers and magazines is divided into two components for the
18
purposes of modeling demand equations for Periodicals mail. The first component is
19
the price of postage paid by publishers (and paid implicitly by consumers through
20
subscription rates). In addition to affecting the price of newspapers and magazines by
21
being incorporated into subscription rates, the price charged by the Postal Service will
22
also affect the demand for Periodicals mail directly by affecting publishers’ decisions
23
over how to deliver their Periodicals. For example, the delivery requirements of many
24
weekly newspapers can be satisfied by either mail or private delivery.
USPS-T-7
171
1
The second component of the price of newspapers and magazines considered in
2
this analysis is the price of paper, modeled by the Bureau of Labor Statistics’ producer
3
price index for pulp, paper, and allied products. This index, which I sometimes also
4
refer to as the price of paper and printing, is used in the Periodicals mail equations to
5
track the non-Postal price of Periodicals. This component of the price of Periodicals will
6
only affect the demand for Periodicals mail indirectly insofar as it is incorporated into
7
subscription prices.
8
9
The Periodicals demand equations used here also include long-run time trends. As
noted above, total Periodicals mail volume declined by approximately 8 percent from
10
1970 to 2004. Periodicals mail volume per adult has declined at an average annual rate
11
of 1.6 percent over the last thirty years.
12
This long-run trend is the result of long-run demographic shifts away from reading.
13
As Tables 41 and 42 suggest, this trend has been going on for decades. One of the
14
principal initial factors thought to be contributing to this long-run trend was television.
15
Yet, even as television viewership has reached maturity and leveled off, the negative
16
trends in Periodicals mail volume appear to be continuing. There are several
17
explanations for the persistence of this trend.
18
First, new substitutes have emerged just as older substitutes have reached maturity.
19
For example, as television reached market saturation, cable television came into
20
existence, creating new outlets for news and entertainment that competed with
21
Periodicals. As cable television growth began to slow, the Internet emerged as an
22
alternative to both cable television as well as Periodicals mail. Second, it appears that
23
substitution away from magazines and newspapers is as much the result of a
24
demographic shift as of substitution with specific media. That is, people today simply
25
read less than comparable people did a generation or two ago, even when one controls
USPS-T-7
172
1
for things such as age and alternate uses of time. This factor is discussed in some
2
more detail in the testimony of Peter Bernstein in this case (USPS-T-8).
3
All of these considerations led, then, to the decision to include simple linear time
4
trends in the Periodicals demand equations used in this case, rather than try to model
5
the impact of specific types of substitution such as television viewing or cable television
6
expenditures, as was attempted in previous rate cases.
7
One specific form of substitution faced by Periodicals is explicitly modeled here,
8
however. That is the Internet. The Internet is somewhat unique in its relationship with
9
magazines and newspapers, as compared with something like cable television. This is
10
because the Internet represents a direct substitute for a hardcopy magazine or
11
newspaper. For example, Chicagotribune.com is an almost-perfect substitute for a hard
12
copy of the Chicago Tribune newspaper. This is less true, however, of the degree of
13
substitution between, say, the Chicago Tribune and a cable news network, even a cable
14
news network that focuses primarily on Chicago news.
15
In this case, the impact of the Internet on Periodicals mail volume is measured by
16
including the number of broadband subscribers in the Periodicals mail equations. The
17
number of broadband subscribers was chosen to model this relationship for two
18
reasons.
19
First, there may be a direct relationship between consumers’ use of high-speed
20
Internet connections and the substitution from print Periodicals to Internet Periodicals.
21
Faster Internet connection speeds enable Internet users to do more things more quickly.
22
In particular, high-speed connections facilitate better Internet graphics. These improved
23
graphics, in turn, make Internet Periodicals more comparable to their print counterparts.
24
Second, the timing of the technological advances that have made broadband access
25
more affordable and, hence, more universal, coincides with the timing of the
USPS-T-7
173
1
technologies which have led to magazines and newspapers putting their content online.
2
Although the presence of magazines and newspapers on the Internet is ubiquitous by
3
now, the negative impact on print Periodicals and Periodicals mail volumes is likely to
4
continue for some time as consumers continue to shift their Periodicals reading habits
5
from hardcopy to the Internet. The continuing negative impact of the Internet on
6
Periodicals mail volumes is a prominent feature of the Periodicals mail volume forecasts
7
developed here.
8
9
Periodicals mail is divided into one regular subclass and three preferred subclasses:
within-county, nonprofit, and classroom mail. For estimation purposes, Periodicals
10
nonprofit and classroom mail were combined and estimated in a single demand
11
equation. Hence, three demand equations were modeled, one each for Periodicals
12
regular rate, within-county, and nonprofit and classroom mail. Periodicals regular mail
13
accounts for more than 70 percent of total Periodicals mail and is considered first below.
USPS-T-7
174
1
2
3
3. Regular Rate
a. Factors Affecting Periodicals Regular Rate Volume
Approximately 70 percent of Periodicals mail is sent at regular rates. More than 75
4
percent of Periodicals regular rate mail could be classified as magazines. Hence, the
5
demand for Periodicals regular rate mail volume is, to a large extent, simply the demand
6
for magazine subscriptions. The general factors considered here were described in the
7
previous section.
8
9
10
11
12
13
14
15
16
To summarize, the factors which affect Periodicals regular rate mail volume include
the following variables:
Total Employment
Price of Paper and Printing
Number of Broadband Subscribers
Linear Time Trend
Price of Periodicals Regular Mail
The effect of these variables on Periodicals regular rate volume over the past ten
17
years is shown in Table 43 on the next page. Table 43 also shows the projected
18
impacts of these variables through GFY 2007.
19
The Test Year before-rates volume forecast for Periodicals regular rate mail is
20
6,438.348 million pieces, a 0.4 percent decline from GFY 2004. The Postal Service’s
21
proposed rates in this case are predicted to reduce the Test Year volume of Periodicals
22
regular rate mail by 0.3 percent, for a Test Year after-rates volume forecast for
23
Periodicals regular rate mail of 6,416.651 million.
USPS-T-7
175
Other
-0.61%
-1.07%
1.91%
-1.32%
0.53%
2.43%
-0.37%
-1.99%
-0.75%
1.05%
Total Change
in Volum e
0.82%
0.91%
3.55%
-0.60%
0.72%
0.62%
-1.89%
-4.58%
-3.98%
-0.85%
1.87%
0.19%
-0.28%
-0.03%
-5.40%
-0.55%
1.05%
0.35%
0.52%
0.17%
-1.70%
-0.57%
-9.15%
-3.15%
0.00%
0.00%
0.00%
0.41%
0.35%
0.34%
-0.85%
0.02%
0.50%
1.16%
-0.48%
0.00%
-0.03%
-0.33%
0.74%
-0.13%
-0.04%
0.00%
0.00%
1.10%
0.36%
-0.34%
-0.11%
0.68%
0.22%
0.37%
0.12%
-1.66%
-1.39%
-1.17%
-0.09%
-0.04%
0.01%
0.00%
-0.34%
-0.67%
0.41%
0.35%
0.34%
-0.85%
0.02%
0.50%
1.16%
-0.48%
0.00%
-0.03%
-0.67%
0.06%
-4.16%
-1.41%
-0.13%
-0.04%
-1.01%
-0.34%
1.10%
0.36%
-0.34%
-0.11%
0.68%
0.22%
-0.64%
-0.21%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.21%
1.15%
1.21%
1.18%
1.19%
1.39%
1.23%
1.29%
1.29%
1.21%
Em ploym ent
0.83%
0.81%
0.46%
0.69%
0.70%
0.57%
0.41%
-0.60%
-1.25%
-0.70%
1994 - 2004
Total
Avg per Year
13.05%
1.23%
1.91%
0.19%
-1.82%
-0.18%
-15.75%
-1.70%
-0.30%
-0.03%
-5.45%
-0.56%
3.66%
0.36%
2001 - 2004
Total
Avg per Year
3.83%
1.26%
-2.53%
-0.85%
-0.55%
-0.18%
-6.14%
-2.09%
0.03%
0.01%
-3.71%
-1.25%
Forecas t
1.23%
1.11%
1.11%
-0.01%
0.30%
0.15%
-0.18%
-0.18%
-0.18%
-1.66%
-1.39%
-1.17%
-0.09%
-0.04%
0.01%
3.49%
1.15%
0.44%
0.15%
-0.55%
-0.18%
-4.16%
-1.41%
After-Rates Volum e Forecas t
2005
1.23%
2006
1.11%
2007
1.11%
-0.01%
0.30%
0.15%
-0.18%
-0.18%
-0.18%
0.44%
0.15%
-0.55%
-0.18%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 43
Estimated Impact of Factors Affecting Periodical Regular Rate Volume, 1994 – 2007
Periodical Prices
Other Factors
Trends
Internet Paper/Printing
Pos tage
Inflation Econom etric
-0.18%
0.00%
-0.34%
-0.30%
0.40%
-0.17%
-0.18%
0.00%
-0.03%
-1.93%
0.41%
1.80%
-0.18%
-0.73%
0.18%
0.60%
0.40%
-0.33%
-0.18%
-1.74%
-0.07%
0.93%
0.32%
-0.36%
-0.18%
-2.03%
0.04%
-0.08%
0.19%
0.39%
-0.19%
-3.05%
-0.13%
-0.62%
0.35%
-0.04%
-0.18%
-3.12%
0.02%
-0.39%
0.49%
0.07%
-0.18%
-2.32%
0.06%
-1.73%
0.38%
0.49%
-0.18%
-2.08%
-0.01%
-1.27%
0.30%
-0.07%
-0.18%
-1.87%
-0.02%
-0.76%
0.36%
0.10%
2004 - 2007
Total
Avg per Year
3.49%
1.15%
USPS-T-7
176
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 43 above.
5
Periodicals Regular Rate Mail has an elasticity with respect to total employment of
6
0.420 (t-statistic of 4.726), so that a 10 percent increase in employment is projected to
7
lead to a 4.20 percent increase in Periodicals regular rate mail volume.
8
9
The linear time trend in the Periodicals regular rate equation is very small, but is
statistically significant, with a coefficient of -0.0005 and a t-statistic of -2.588. This
10
explains annual declines in Periodicals mail volume of approximately 0.18 percent per
11
year.
12
The Internet, on the other hand, is estimated to have had a much more significant
13
negative effect on Periodicals regular rate mail volume, reducing Periodicals regular
14
rate volume by more than 15 percent over the past eight years. As with the Internet
15
Experience variable in the First-Class single-piece letters and First-Class cards
16
equations, the number of broadband subscribers is entered into the Periodicals regular
17
rate equation using a Box-Cox specification. The estimated Box-Cox coefficient is
18
equal to 0.365 (t-statistic of 5.974). Based on this specification, the negative effect of
19
the Internet on Periodicals regular rate volume is expected to decline moderately
20
through the forecast period. Even with this moderation, however, the Internet is still
21
expected to reduce Periodicals regular rate mail volume by more than 4 percent over
22
the next three years.
23
Periodicals regular rate mail volume is affected by two prices, the price of paper and
24
printing and the price of postage. The elasticity of Periodicals regular rate mail with
25
respect to the price of paper and printing is estimated here to be equal to -0.040
USPS-T-7
177
1
(t−statistic of −0.641). The own-price elasticity of Periodicals regular rate mail was
2
calculated to be equal to -0.193 (t−statistic of -10.33).
3
The Postal price impacts shown in Table 43 above are the result of changes in
4
nominal prices. Prices enter the demand equations developed here in real terms,
5
however. The impact of inflation reported in Table 43 measures the impact that a
6
change in real Postal prices, in the absence of nominal rate changes, has on the
7
volume of Periodicals regular rate mail.
8
9
Other econometric variables include seasonal variables. A more detailed look at the
econometric demand equation for Periodicals regular rate mail follows.
b. Econometric Demand Equation
10
11
The demand equation for Periodicals regular rate mail in this case models
12
Periodicals regular rate mail volume per adult per delivery day as a function of the
13
following explanatory variables:
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
·
Seasonal Variables
·
Total Private Employment (lagged three quarters)
·
Producer Price Index for Pulp, Paper, and Allied Products
·
Number of Broadband Subscribers
The broadband variable is entered into the Periodicals Regular rate equation with a
Box-Cox transformation as described above.
·
Linear Time Trend
·
Current and four lags of the price of Periodicals Regular Rate Mail
Details of the econometric demand equation are shown in Table 44 below. A
29
detailed description of the econometric methodologies used to obtain these results can
30
be found in Section III below.
USPS-T-7
178
1
2
3
TABLE 44
ECONOMETRIC DEMAND EQUATION FOR PERIODICALS REGULAR RATE MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.193
-10.33
Current
-0.022
-0.343
Lag 1
-0.000
-0.000
Lag 2
-0.051
-0.510
Lag 3
-0.069
-0.695
Lag 4
-0.050
-0.768
Employment (lagged three quarters)
0.420
4.726
Price of Paper
-0.040
-0.641
Time Trend
-0.0005
-2.588
Number of Broadband subscribers
Box-Cox Coefficient
0.365
5.974
Coefficient
-0.533
-15.50
Seasonal Coefficients
September 1 – 15
-0.538
-1.220
September 16 – 30
-0.247
-2.542
October 1 – December 10
-0.038
-0.486
December 11 – 12
0.883
1.841
December 13 – 17
-1.050
-4.350
December 18 – 21
0.323
1.187
December 22 – 31
-0.099
-0.756
January 1 – April 15
-0.049
-0.622
April 16 – May
-0.082
-1.072
June
-0.203
-2.042
Quarter 1 (October – December)
-0.008
-0.724
Quarter 2 (January – March)
0.001
0.049
Quarter 3 (April – June)
0.025
1.649
Quarter 4 (July – September)
-0.017
-1.112
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.000258
Quarter 2 (January – March)
1.046924
Quarter 3 (April – June)
1.001588
Quarter 4 (July – September)
0.952455
REGRESSION DIAGNOSTICS
Sample Period
1973Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
106
Mean-Squared Error
0.000577
Adjusted R-Squared
0.919
USPS-T-7
179
1
2
4. Preferred Periodicals Subclasses
a. Overview
3
The Postal Service offers preferred rates for certain types of Periodicals mailers.
4
Preferred Periodicals mail is divided into three subclasses on the basis of either the
5
mailer or the mail content: within-county mail, which is mail sent within a particular
6
county and is comprised primarily of small local publications (mostly newspapers);
7
nonprofit mail, which is mail sent by not-for-profit organizations; and classroom mail,
8
which is mail for students sent to classrooms and educational institutions. These latter
9
two subclasses are combined in my analysis.
10
11
12
The basic theory of demand for the preferred categories of Periodicals mail is
expected to be similar to the theory outlined at the introduction to this section.
The price of paper was investigated in these demand equations, consistent with the
13
theory outlined above. The price of paper was not found to affect the volume of
14
Periodicals within-county mail, however. This could have occurred for a variety of
15
reasons, including the possibility that within-county mailers are less sensitive to these
16
prices or that there are fewer substitutes for printed material within these contexts, so
17
that this type of mail would be less price-sensitive in general. In addition, the number of
18
broadband subscribers did not work in the within-county demand equation, either.
19
These omissions from the within-county equation are discussed below in the discussion
20
of the Periodicals within-county demand equation used in this case.
21
Linear time trends were included in the preferred Periodicals demand equations, just
22
as in the Periodicals regular equation. Both of the preferred Periodicals demand
23
equations had larger negative time trends than Periodicals regular mail. This is
24
discussed in more detail in the discussion of these specific demand equations below.
USPS-T-7
180
1
2
3
4
5
The specific demand equations for Periodicals within county and nonprofit (including
classroom) mail are described below.
b. Within-County
i. Factors Affecting Periodicals Within-County Mail Volume
Periodicals within-county mail is mail sent primarily within the county of publication.
6
In general, Periodicals within-county mail volume is affected by the same factors as
7
other types of Periodicals mail. There are, however, two significant omissions from the
8
Periodicals within-county demand equation: the price of paper and printing and the
9
number of broadband subscribers. Neither of these variables was found to influence
10
Periodicals within-county mail volume. It is not entirely clear why these variables
11
appeared to have no effect on within-county mail volume. My hypothesis is that the
12
producer price index for pulp, paper, and allied products may be a poor estimate of the
13
cost of preparing within-county mail and that the specific nature of within-county mail
14
makes it somewhat less vulnerable to Internet diversion.
15
16
17
18
19
20
21
In summary, then, the factors which affect Periodicals within-county mail volume
include the following variables:
Total Employment
Linear Time Trend
Price of Periodicals Within-County Mail
The effect of these variables on Periodicals within-county mail volume over the past
22
ten years is shown in Table 45 on the next page. Table 45 also shows the projected
23
impacts of these variables through GFY 2007.
24
The Test Year before-rates volume forecast for Periodicals within-county mail is
25
743.285 million pieces, a 2.2 percent decline from GFY 2004. The Postal Service’s
26
proposed rates in this case are predicted to increase Test Year volume of Periodicals
USPS-T-7
181
1
within-county mail by 1.4 percent, for a Test Year after-rates volume forecast for
2
Periodicals within-county mail of 753.578 million.
USPS-T-7
182
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.11%
1.15%
1.23%
1.17%
1.18%
1.38%
1.21%
1.30%
1.28%
1.19%
1994 - 2004
Total
Avg per Year
12.89%
1.22%
2.54%
0.25%
-34.10%
-4.08%
-4.27%
-0.44%
4.55%
0.45%
-3.09%
-0.31%
2.03%
0.20%
-24.51%
-2.77%
2001 - 2004
Total
Avg per Year
3.82%
1.26%
-4.89%
-1.66%
-11.81%
-4.10%
-1.12%
-0.37%
1.26%
0.42%
0.87%
0.29%
-2.35%
-0.79%
-14.11%
-4.94%
Forecas t
1.24%
1.10%
1.09%
0.60%
0.44%
0.03%
-4.16%
-4.06%
-4.07%
0.00%
0.00%
0.00%
0.48%
0.40%
0.46%
0.14%
-0.04%
-0.94%
2.69%
-0.83%
0.00%
0.86%
-3.04%
-3.47%
3.47%
1.14%
1.07%
0.36%
-11.79%
-4.10%
0.00%
0.00%
1.35%
0.45%
-0.84%
-0.28%
1.84%
0.61%
-5.59%
-1.90%
After-Rates Volum e Forecas t
2005
1.24%
2006
1.10%
2007
1.09%
0.60%
0.44%
0.03%
-4.16%
-4.06%
-4.07%
0.00%
1.38%
0.00%
0.48%
0.40%
0.46%
0.14%
-0.04%
-0.94%
2.69%
-0.83%
0.00%
0.86%
-1.69%
-3.47%
1.07%
0.36%
-11.79%
-4.10%
1.38%
0.46%
1.35%
0.45%
-0.84%
-0.28%
1.84%
0.61%
-4.28%
-1.45%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 45
Estimated Impact of Factors Affecting Periodical W ithin County Volume, 1994 – 2007
Total Change
Other Factors
Em ploym ent
Trends
Pos tage
Inflation Econom etric
Other
in Volum e
1.87%
-3.77%
-0.57%
0.50%
1.94%
-12.06%
-11.20%
0.97%
-4.14%
-0.46%
0.47%
-5.32%
5.43%
-2.25%
1.43%
-4.21%
-1.12%
0.50%
-0.95%
11.71%
8.15%
1.56%
-4.05%
-0.35%
0.25%
0.12%
-1.26%
-2.63%
1.16%
-4.05%
0.23%
0.30%
0.76%
-2.28%
-2.80%
1.15%
-4.22%
0.34%
0.64%
0.91%
0.21%
0.29%
-0.55%
-4.10%
-1.29%
0.53%
-1.29%
4.32%
-1.36%
-2.85%
-4.19%
-0.86%
0.34%
0.49%
1.89%
-3.95%
-1.65%
-4.07%
-0.26%
0.45%
0.03%
-2.51%
-6.63%
-0.46%
-4.05%
0.00%
0.46%
0.35%
-1.69%
-4.22%
2004 - 2007
Total
Avg per Year
3.47%
1.14%
USPS-T-7
183
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 45 above.
5
Periodicals within-county mail has an elasticity with respect to total employment of
6
0.893 (t-statistic of 1.621), so that a 10 percent increase in employment is projected to
7
lead to an 8.93 percent increase in Periodicals within-county mail volume.
8
The Periodicals within-county equation has the largest time trend of the three
9
Periodicals demand equations. The within-county trend coefficient is equal to -0.010
10
with a t-statistic of -8.597. This explains annual declines in Periodicals within-county
11
mail volume of approximately 4.1 percent per year.
12
The own-price elasticity of Periodicals within-county mail was calculated to be equal
13
to -0.235 (t−statistic of -1.740). The Postal price impacts shown in Table 45 above are
14
the result of changes in nominal prices. Prices enter the demand equations developed
15
here in real terms, however. The impact of inflation reported in Table 45 measures the
16
impact that a change in real Postal prices, in the absence of nominal rate changes, has
17
on the volume of Periodicals within-county mail.
18
Other econometric variables shown in Table 45 include seasonal variables. A more
19
detailed look at the econometric demand equation for Periodicals within-county mail
20
follows.
ii. Econometric Demand Equation
21
22
The demand equation for Periodicals within-county mail in this case models
23
Periodicals within-county mail volume per adult per delivery day as a function of the
24
following explanatory variables:
25
·
Seasonal Variables
USPS-T-7
184
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
·
Total Private Employment
·
Dummy variable equal to one starting in 1985Q1 to reflect a change to the
methodology used by the Postal Service for reporting within-county volume
·
Dummy variable equal to one starting in 1987Q1 to reflect a rule change
restricting within-county eligibility
·
Dummy variable equal to one starting in 1993Q2 to reflect a change in the
sampling methodology used by the Postal Service to calculate Periodical
within-county volumes
·
Linear Time Trend
·
Current price of Periodicals Within-County Mail
Details of the econometric demand equation are shown in Table 46 below. A
18
detailed description of the econometric methodologies used to obtain these results can
19
be found in Section III below.
USPS-T-7
185
1
2
3
TABLE 46
ECONOMETRIC DEMAND EQUATION FOR PERIODICALS WITHIN-COUNTY MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (current only)
-0.235
-1.740
Employment
0.893
1.621
Time Trend
-0.010
-8.597
Dummy for 1985 Reporting Change
0.304
4.827
Dummy for 1987 Rule Change
-0.103
-1.740
Dummy for 1993 Sampling Change
-0.181
-3.736
Seasonal Coefficients
September
-0.067
-0.190
0.387
0.858
October
November 1 – December 10
-0.102
-0.239
December 11 – 12
-2.713
-1.920
December 13 – 19
1.106
2.296
December 20 – 24
-0.714
-1.081
December 25 – 31
0.709
1.480
January – May
0.051
0.331
0.227
0.541
June
Quarter 1 (October – December)
-0.028
-0.792
Quarter 2 (January – March)
0.021
0.847
Quarter 3 (April – June)
-0.058
-0.678
Quarter 4 (July – September)
0.065
0.848
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.983372
Quarter 2 (January – March)
0.974676
Quarter 3 (April – June)
1.039572
Quarter 4 (July – September)
1.003498
REGRESSION DIAGNOSTICS
Sample Period
1983Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.496
Degrees of Freedom
68
Mean-Squared Error
0.004282
Adjusted R-Squared
0.964
4
USPS-T-7
186
1
2
3
4
5
c. Nonprofit and Classroom Mail
i. Factors Affecting Periodicals Nonprofit and Classroom Mail
Volume
Periodicals nonprofit mail is offered to qualified not-for-profit organizations. Certain
6
mail sent to educational institutions is eligible for classroom rates. The Periodicals
7
nonprofit demand equation exactly parallels the Periodicals regular rate demand
8
equation described above. Hence, Periodicals nonprofit and classroom mail volume are
9
modeled as a function of the following variables:
Total Employment
Price of Paper and Printing
Number of Broadband Subscribers
Linear Time Trend
Price of Periodicals Nonprofit and Classroom Mail
10
11
12
13
14
15
16
The effect of these variables on Periodicals nonprofit and classroom mail volume
17
over the past ten years is shown in Table 47 on the next page. Table 47 also shows the
18
projected impacts of these variables through GFY 2007.
19
The Test Year before-rates volume forecast for Periodicals nonprofit and
20
classroom mail is 1,896.987 million pieces, a 0.8 percent decline from GFY 2004. The
21
Postal Service’s proposed rates in this case are predicted to reduce the Test Year
22
volume of Periodicals nonprofit and classroom mail by 0.9 percent, for a Test Year after-
23
rates volume forecast for Periodicals nonprofit and classroom mail of 1,879.593 million.
USPS-T-7
187
Other
0.15%
0.66%
0.36%
-1.09%
-3.43%
4.01%
-0.36%
3.33%
1.75%
-5.25%
Total Change
in Volum e
0.06%
-3.18%
-2.39%
-0.73%
-0.16%
0.96%
-3.74%
-3.86%
-2.10%
-4.77%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.21%
1.12%
1.18%
1.16%
1.18%
1.36%
1.20%
1.31%
1.29%
1.19%
1994 - 2004
Total
Avg per Year
12.90%
1.22%
2.45%
0.24%
-10.36%
-1.09%
-10.21%
-1.07%
-7.44%
-0.77%
-9.73%
-1.02%
4.56%
0.45%
0.61%
0.06%
-0.23%
-0.02%
-18.36%
-2.01%
2001 - 2004
Total
Avg per Year
3.84%
1.26%
-4.56%
-1.54%
-3.25%
-1.09%
-3.93%
-1.33%
-0.49%
-0.16%
-3.52%
-1.19%
1.27%
0.42%
0.47%
0.15%
-0.38%
-0.13%
-10.36%
-3.58%
Forecas t
1.23%
1.10%
1.09%
0.58%
0.42%
0.03%
-1.10%
-1.08%
-1.09%
-1.05%
-0.87%
-0.72%
-0.70%
-1.01%
-2.13%
0.00%
0.00%
0.00%
0.51%
0.40%
0.45%
0.12%
0.05%
-0.56%
1.84%
-1.21%
0.00%
1.39%
-2.21%
-2.93%
3.47%
1.14%
1.02%
0.34%
-3.23%
-1.09%
-2.61%
-0.88%
-3.80%
-1.28%
0.00%
0.00%
1.36%
0.45%
-0.40%
-0.13%
0.61%
0.20%
-3.75%
-1.26%
After-Rates Volum e Forecas t
2005
1.23%
2006
1.10%
2007
1.09%
0.58%
0.42%
0.03%
-1.10%
-1.08%
-1.09%
-1.05%
-0.87%
-0.72%
-0.70%
-1.01%
-2.13%
0.00%
-0.92%
-0.33%
0.51%
0.40%
0.45%
0.12%
0.05%
-0.56%
1.84%
-1.21%
0.00%
1.39%
-3.10%
-3.25%
1.02%
0.34%
-3.23%
-1.09%
-2.61%
-0.88%
-3.80%
-1.28%
-1.25%
-0.42%
1.36%
0.45%
-0.40%
-0.13%
0.61%
0.20%
-4.95%
-1.68%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 47
Estimated Impact of Factors Affecting Periodical Nonprofit and Classroom Volume, 1994 – 2007
Periodical Prices
Other Factors
Em ploym ent
Trends
Internet Paper/Printing
Pos tage
Inflation Econom etric
1.87%
-1.09%
0.00%
-0.28%
-2.61%
0.54%
0.33%
0.89%
-1.08%
0.00%
-2.55%
-2.69%
0.49%
0.02%
1.28%
-1.07%
-0.46%
-5.63%
1.64%
0.50%
-0.02%
1.45%
-1.08%
-1.08%
-0.57%
0.39%
0.29%
-0.17%
1.10%
-1.09%
-1.28%
3.94%
-1.14%
0.25%
0.47%
1.06%
-1.12%
-1.94%
-1.88%
-1.43%
0.58%
0.46%
-0.52%
-1.08%
-1.95%
0.02%
-0.71%
0.56%
-0.94%
-2.63%
-1.11%
-1.51%
-2.10%
-1.46%
0.40%
-0.01%
-1.56%
-1.09%
-1.31%
0.47%
-1.87%
0.44%
-0.16%
-0.42%
-1.08%
-1.16%
1.17%
-0.23%
0.43%
0.63%
2004 - 2007
Total
Avg per Year
3.47%
1.14%
USPS-T-7
188
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 47 above.
5
Periodicals nonprofit mail has an elasticity with respect to total employment of 0.835
6
(t-statistic of 1.116), so that a 10 percent increase in employment is projected to lead to
7
an 8.35 percent increase in Periodicals nonprofit mail volume.
8
9
10
11
The linear time trend in the Periodicals nonprofit equation has a coefficient of
−0.0028 with a t-statistic of -1.395. This trend explains annual declines in Periodicals
nonprofit mail volume of approximately 1.1 percent per year.
The Internet is estimated to have had a somewhat more significant negative effect
12
on Periodicals nonprofit mail volume over the past three years, with an average annual
13
decline of 1.3 percent attributable to the Internet over this time period. The negative
14
effect of the Internet on Periodicals nonprofit mail volume is expected to decline
15
moderately through the forecast period, to an annual level of less than one percent for
16
GFY 2006 and GFY 2007.
17
The elasticity of Periodicals nonprofit mail with respect to the price of paper and
18
printing is estimated here to be equal to -1.2, while the own-price elasticity of
19
Periodicals nonprofit and classroom mail was calculated to be equal to -0.237
20
(t−statistic of -1.722).
21
The Postal price impacts shown in Table 47 above are the result of changes in
22
nominal prices. Prices enter the demand equations developed here in real terms,
23
however. The impact of inflation reported in Table 47 measures the impact that a
24
change in real Postal prices, in the absence of nominal rate changes, has on the
25
volume of Periodicals nonprofit and classroom mail.
USPS-T-7
189
1
2
Other econometric variables include seasonal variables. A more detailed look at the
econometric demand equation for Periodicals nonprofit and classroom mail follows.
ii. Econometric Demand Equation
3
4
The demand equation for Periodicals nonprofit mail in this case models Periodicals
5
nonprofit and classroom mail volume per adult per delivery day as a function of the
6
following explanatory variables:
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
·
Seasonal Variables
·
Total Private Employment
·
Producer Price Index for Pulp, Paper, and Allied Products (lagged two and
eight quarters)
·
Number of Broadband Subscribers
The broadband variable is entered into the Periodicals nonprofit equation with a BoxCox transformation. The Box-Cox coefficient in this case is taken from the Periodicals
regular rate equation.
·
Linear Time Trend
·
Current and two lags of the price of Periodicals nonprofit and classroom mail
Details of the econometric demand equation are shown in Table 48 below. A
24
detailed description of the econometric methodologies used to obtain these results can
25
be found in Section III below.
USPS-T-7
190
1
2
3
4
TABLE 48
ECONOMETRIC DEMAND EQUATION FOR
PERIODICALS NONPROFIT AND CLASSROOM MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.237
-1.722
Current
-0.087
-0.948
Lag 1
-0.056
-0.726
Lag 2
-0.093
-1.035
Employment
0.835
1.116
Price of Paper
Lag 2
-0.306
-0.525
Lag 8
-0.901
-1.666
Time Trend
-0.0028
-1.395
Number of Broadband subscribers
Box-Cox Coefficient
0.365
(N/A)
Coefficient
-0.332
-1.200
Seasonal Coefficients
September
0.694
3.132
0.199
1.784
October
0.614
3.317
November 1 – December 24
December 25 – February
0.411
5.193
March – April 15
0.527
3.075
April 16 – May
0.307
5.145
June
0.830
3.261
Quarter 1 (October – December)
-0.014
-0.854
Quarter 2 (January – March)
0.028
1.277
Quarter 3 (April – June)
-0.120
-3.326
Quarter 4 (July – September)
0.106
2.661
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.032503
Quarter 2 (January – March)
1.067765
Quarter 3 (April – June)
0.984084
Quarter 4 (July – September)
0.918837
REGRESSION DIAGNOSTICS
Sample Period
1978Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.538
AR-2: 0.451
AR-4: -0.313
Degrees of Freedom
83
Mean-Squared Error
0.003456
Adjusted R-Squared
0.908
USPS-T-7
191
1
G. Other Mail Categories
2
In addition to the mail volumes described above, demand equations are also
3
modeled for three additional categories of mail: Mailgrams, U.S. Postal Service mail
4
(also called Postal Penalty), and Free for the Blind and Handicapped Mail.
5
Mailgrams are telegrams delivered by the Postal Service under an agreement with
6
Western Union. Postal Penalty mail refers to mail sent by the Postal Service. Free for
7
the Blind and Handicapped Mail is mail that is delivered free of charge by the Postal
8
Service under certain circumstances.
9
Because there is no direct price charged for Mailgrams, Postal Penalty, and Free for
10
the Blind and Handicapped Mail, price was not included in the demand specifications for
11
these categories of mail. The primary factor in each of these equations is a simple
12
linear time trend. These equations are described briefly below.
13
1. Mailgrams
14
Mailgrams volume is modeled as a function of a linear time trend and seasonal
15
variables. The last ten years of Mailgrams volume are summarized in Table 49 on the
16
next page. Table 49 also shows Mailgrams projections through GFY 2007. The details
17
of the econometric demand equation for Mailgrams are shown in Table 50.
18
The Test Year before-rates volume forecast for Mailgrams is 1.359 million pieces, a
19
17.5 percent decline from GFY 2004. The Mailgrams forecast does not include price as
20
an explanatory variable. Hence, the Test Year after-rates volume forecast for
21
Mailgrams is equal to the Test Year before-rates forecast.
USPS-T-7
192
1
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2
Table 49
Estimated Impact of Factors Affecting Mailgrams Volume, 1994 – 2007
Total Change
Other Factors
Trends Econom etric
Other
in Volum e
Population
1.11%
-13.91%
-0.14%
-9.04%
-20.93%
1.09%
-14.24%
2.36%
16.29%
3.19%
1.41%
-17.92%
0.39%
54.05%
28.73%
1.02%
-13.20%
-2.28%
-7.58%
-20.82%
1.20%
-15.21%
-0.42%
17.76%
0.62%
1.27%
-14.17%
-0.17%
-2.14%
-15.09%
1.22%
-14.95%
7.61%
-0.83%
-8.13%
1.22%
-13.95%
9.94%
-14.30%
-17.93%
1.38%
-15.69%
-1.07%
19.72%
1.23%
0.95%
-11.67%
-8.88%
-27.33%
-40.95%
1994 - 2004
Total
Avg per Year
12.50%
1.19%
-79.13%
-14.50%
6.30%
0.61%
28.32%
2.53%
-67.98%
-10.76%
2001 - 2004
Total
Avg per Year
3.59%
1.18%
-35.91%
-13.78%
-0.89%
-0.30%
-25.44%
-9.32%
-50.94%
-21.13%
2005
2006
2007
1.27%
1.04%
1.05%
-15.38%
-13.98%
-14.27%
-0.59%
0.27%
0.14%
14.47%
-2.96%
0.00%
-2.48%
-15.42%
-13.25%
2004 - 2007
Total
Avg per Year
3.40%
1.12%
-37.59%
-14.54%
-0.18%
-0.06%
11.09%
3.57%
-28.45%
-10.56%
USPS-T-7
193
1
2
3
TABLE 50
ECONOMETRIC DEMAND EQUATION FOR MAILGRAMS
Coefficient T-Statistic
Time Trend
-0.039
-16.83
Seasonal Coefficients
September
1.194
1.619
2.096
1.619
October
November 1 – December 12
-1.571
-1.416
5.110
3.851
December 13 – 19
December 20 – 24
-1.265
-0.953
December 25 –June
0.437
2.253
Quarter 1 (October – December)
0.034
0.337
Quarter 2 (January – March)
0.052
0.676
Quarter 3 (April – June)
0.089
1.175
Quarter 4 (July – September)
-0.175
-1.711
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.966991
Quarter 2 (January – March)
1.085036
Quarter 3 (April – June)
1.126472
Quarter 4 (July – September)
0.822852
REGRESSION DIAGNOSTICS
Sample Period
1983Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.590
Degrees of Freedom
76
Mean-Squared Error
0.051194
Adjusted R-Squared
0.952
4
5
USPS-T-7
194
2. Postal Penalty Mail
1
Postal Penalty volume is modeled as a function of two linear time trends, one over
2
3
the full sample period and another starting in the first quarter of 1997, and seasonal
4
variables. The last ten years of Postal Penalty volume are summarized in Table 51 on
5
the next page. Table 51 also shows Postal Penalty projections through GFY 2007. The
6
details of the econometric demand equation for Postal Penalty mail are shown in Table
7
52.
8
9
The Test Year before-rates volume forecast for Postal Penalty mail is 666.538
million pieces, a 25.9 percent increase from GFY 2004 (although virtually all of this
10
increase is because actual volume in 2005Q1 was 59.8 percent greater than volume in
11
2004Q1). The Postal Penalty mail forecast does not include price as an explanatory
12
variable. Hence, the Test Year after-rates volume forecast for Postal Penalty mail is
13
equal to the Test Year before-rates forecast.
USPS-T-7
195
1
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2
Table 51
Estimated Impact of Factors Affecting Postal Penalty Mail Volume, 1994 – 2007
Total Change
Other Factors
Trends
Econom etric
Other
in Volum e
Population
1.20%
-12.73%
-5.75%
8.90%
-9.35%
1.03%
-11.62%
-1.10%
-1.08%
-12.64%
1.15%
-1.36%
0.81%
2.47%
3.06%
1.19%
5.11%
1.05%
-6.77%
0.21%
1.20%
5.13%
0.22%
-5.09%
1.21%
1.32%
4.96%
0.96%
-11.51%
-4.99%
1.27%
5.24%
0.43%
-2.02%
4.88%
1.40%
5.39%
3.40%
1.04%
11.65%
1.25%
4.83%
-4.73%
-8.91%
-7.90%
1.35%
5.60%
-0.20%
26.63%
35.25%
1994 - 2004
Total
Avg per Year
13.07%
1.24%
8.35%
0.81%
-5.13%
-0.53%
-1.31%
-0.13%
14.70%
1.38%
2001 - 2004
Total
Avg per Year
4.05%
1.33%
16.66%
5.27%
-1.69%
-0.57%
16.54%
5.23%
39.07%
11.62%
2005
2006
2007
1.39%
1.09%
1.13%
5.78%
4.96%
5.19%
-0.88%
-0.30%
-0.19%
18.38%
-5.41%
0.00%
25.84%
0.07%
6.18%
2004 - 2007
Total
Avg per Year
3.66%
1.20%
16.79%
5.31%
-1.37%
-0.46%
11.98%
3.84%
33.71%
10.17%
USPS-T-7
196
1
2
3
TABLE 52
ECONOMETRIC DEMAND EQUATION FOR POSTAL PENALTY MAIL
Coefficient T-Statistic
Time Trends
Full-Sample
-0.033
-4.169
Starting in 1997Q1
0.045
3.840
Dummy for 2002Q1 (9/11 Effect)
0.149
1.268
Seasonal Coefficients
September 1 – 15
-3.589
-2.041
-0.251
-0.826
September 16 – December 19
December 20 – 31
-1.114
-2.547
-0.415
-1.316
January – June
Quarter 1 (October – December)
-0.028
-0.510
Quarter 2 (January – March)
-0.019
-0.378
Quarter 3 (April – June)
-0.042
-0.845
Quarter 4 (July – September)
0.089
1.620
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.056628
Quarter 2 (January – March)
1.015066
Quarter 3 (April – June)
0.992232
Quarter 4 (July – September)
0.937936
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.298
AR-2: 0.454
Degrees of Freedom
54
Mean-Squared Error
0.015811
Adjusted R-Squared
0.768
4
5
USPS-T-7
197
1
3. Free for the Blind and Handicapped Mail
2
Free for the Blind and Handicapped Mail volume is modeled as a function of a linear
3
time trend, Internet Experience (adjusted by the Box-Cox coefficient estimated from the
4
First-Class single-piece letters equation as described above), dummy variables for the
5
time periods 2000Q1 through 2001Q4 and 2003Q1 through 2003Q3, and seasonal
6
variables. The last ten years of Free for the Blind and Handicapped Mail volume are
7
summarized in Table 53 on the next page. Table 53 also shows Free for the Blind and
8
Handicapped Mail volume projections through GFY 2007. The details of the
9
econometric demand equation for Free for the Blind and Handicapped Mail are shown
10
11
in Table 54.
The Test Year before-rates volume forecast for Free for the Blind and Handicapped
12
Mail is 75.317 million pieces, a 6.0 percent increase from GFY 2004. The Free for the
13
Blind and Handicapped Mail forecast does not include price as an explanatory variable.
14
Hence, the Test Year after-rates volume forecast for Free for the Blind and
15
Handicapped Mail is equal to the Test Year before-rates forecast.
USPS-T-7
198
1
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2
Table 53
Estimated Impact of Factors Affecting Free-for-the-Blind-and-Handicapped Mail Volume, 1994 – 2007
Total Change
Other Factors
Trends
Internet
Econom etric
Other
in Volum e
Population
1.15%
3.16%
-1.11%
3.77%
-7.65%
-1.11%
1.09%
3.17%
-1.37%
-8.98%
5.41%
-1.30%
1.21%
3.26%
-1.40%
-10.48%
14.82%
5.92%
1.21%
3.33%
-1.23%
-4.55%
1.03%
-0.40%
1.19%
3.26%
-1.24%
-2.35%
-0.94%
-0.19%
1.32%
3.22%
-1.29%
-20.06%
7.20%
-11.53%
1.17%
3.13%
-1.46%
-1.73%
-5.41%
-4.43%
1.46%
3.59%
-1.57%
24.01%
-0.63%
27.48%
1.44%
3.61%
-1.42%
15.53%
3.48%
23.87%
1.22%
3.26%
-1.27%
-11.37%
10.42%
0.99%
1994 - 2004
Total
Avg per Year
13.18%
1.25%
38.35%
3.30%
-12.58%
-1.34%
-21.39%
-2.38%
28.78%
2.56%
38.56%
3.32%
2001 - 2004
Total
Avg per Year
4.18%
1.37%
10.83%
3.49%
-4.20%
-1.42%
26.98%
8.29%
13.54%
4.32%
59.47%
16.83%
2005
2006
2007
1.23%
1.13%
1.11%
3.27%
3.30%
3.26%
-1.22%
-1.17%
-1.11%
0.78%
0.04%
-1.61%
-1.61%
0.19%
0.00%
2.38%
3.49%
1.59%
2004 - 2007
Total
Avg per Year
3.50%
1.15%
10.16%
3.28%
-3.45%
-1.17%
-0.81%
-0.27%
-1.42%
-0.48%
7.64%
2.49%
USPS-T-7
199
1
2
3
4
TABLE 54
ECONOMETRIC DEMAND EQUATION FOR
FREE FOR THE BLIND AND HANDICAPPED MAIL
Coefficient T-Statistic
Time Trend
0.008
6.652
Internet Experience
Box-Cox Coefficient
0.326
(N/A)
Coefficient
-0.163
-1.159
Dummy Variable for 2000 – 2001
-0.237
-2.116
Dummy Variable for 2003Q1 – 3
0.174
0.965
Seasonal Coefficients
December 11 – 21
-1.335
-1.808
2.272
0.936
December 22 – 24
Quarter 1 (October – December)
0.135
1.231
Quarter 2 (January – March)
-0.102
-0.943
Quarter 3 (April – June)
0.020
0.186
Quarter 4 (July – September)
-0.053
-0.484
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.084695
Quarter 2 (January – March)
0.916961
Quarter 3 (April – June)
1.035918
Quarter 4 (July – September)
0.962469
REGRESSION DIAGNOSTICS
Sample Period
1971Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
127
Mean-Squared Error
0.076556
Adjusted R-Squared
0.482
USPS-T-7
200
1
2
H. Special Services
1. General Overview
3
Except for money orders and Post Office Boxes, special services are not mail
4
volumes, but represent add-ons to mail volumes. That is, a certified letter would be
5
counted as both a piece of Certified Mail as well as a First-Class letter. Therefore, the
6
volumes of special services are not included in a calculation of total Postal Service
7
volume.
8
The special service forecasts presented here are for domestic special services only.
9
For some early years, the Postal Service’s volume data from the RPW system did not
10
distinguish between domestic and international special services. Hence, the volumes
11
for those quarters which are used in estimating the demand equations include both
12
domestic and international special service volumes. For most special services, this is
13
trivial, as the volume of international special services is insignificant relative to domestic
14
volumes. The one exception to this is Registered Mail. International registered volume
15
is approximately 50 percent as large as domestic Registered Mail volume. This is dealt
16
with by including a dummy variable equal to zero for the quarters for which domestic
17
and international special service data are combined and equal to one for the quarters
18
for which the data are only for domestic Registered Mail (1998Q1 through the present).
19
Because special services are add-ons to existing mail volumes, the demand for
20
special services may be affected directly by the demand for complementary categories
21
of mail. For example, the volumes of both Registered and Certified Mail are modeled in
22
part as functions of the volume of First-Class letters since most Registered and Certified
23
Mail is, in fact, First-Class Mail. In addition, insured mail volume is modeled in part as a
24
function of the volume of Parcel Post mail since a large portion of insured mail volume is
USPS-T-7
201
1
sent as Parcel Post mail. Similarly, the volume of Return Receipts is a function of the
2
volume of Certified Mail since most Return Receipts accompany Certified Mail.
3
In addition, the special service volumes modeled here have generally exhibited long-
4
run trends. For this reason, a time trend is included in the demand equation associated
5
with each of the special services (except for Return Receipts, Post Office Boxes, and
6
stamped cards).
7
Finally, of course, the demand for special service mail is also a function of the price
8
charged by the Postal Service for these services. In addition, most of the special
9
service equations also include some equation-specific variables, which are described
10
below. In particular, special service classification reform, MC96-3, which took effect in
11
June, 1997, involved significant changes to eligibility and pricing standards for several
12
special services. The techniques used to model these changes are described below in
13
the descriptions of the individual special services.
14
15
Specific demand equations for the nine special services forecasted here are
described in detail below.
2. Registered Mail
16
a. Factors Affecting Registered Mail Volume
17
18
19
20
21
22
23
24
Registered Mail volume was found to be principally affected by the following
variables:
First-Class Letters Volume
Linear Time Trend
Price of Registered Mail
The effect of these variables on Registered Mail volume over the past ten years is
25
shown in Table 55 on the next page. Table 55 also shows the projected impacts of
26
these variables through GFY 2007.
USPS-T-7
202
1
The Test Year before-rates volume forecast for Registered Mail is 3.990 million
2
pieces, a 20.3 percent decline from GFY 2004. The Postal Service’s proposed rates in
3
this case are predicted to reduce the Test Year volume of Registered Mail by 6.3
4
percent, for a Test Year after-rates volume forecast for Registered Mail of 3.738 million.
USPS-T-7
203
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.19%
1.06%
1.20%
1.02%
1.16%
1.35%
1.17%
1.18%
1.28%
1.15%
1994 - 2004
Total
Avg per Year
12.38%
1.17%
17.29%
1.61%
-65.14%
-10.00%
-6.26%
-0.64%
1.92%
0.19%
-47.80%
-6.29%
-3.13%
-0.32%
-77.80%
-13.97%
2001 - 2004
Total
Avg per Year
3.64%
1.20%
-9.48%
-3.26%
-26.68%
-9.83%
-1.38%
-0.46%
0.54%
0.18%
-0.87%
-0.29%
-4.83%
-1.64%
-35.66%
-13.67%
Forecas t
1.18%
1.08%
1.06%
3.91%
-2.64%
-1.82%
-10.11%
-10.19%
-10.03%
0.00%
0.00%
0.00%
0.20%
0.18%
0.20%
-2.36%
-0.99%
-0.88%
-4.71%
3.14%
0.00%
-11.91%
-9.57%
-11.34%
3.36%
1.11%
-0.67%
-0.22%
-27.37%
-10.11%
0.00%
0.00%
0.58%
0.19%
-4.17%
-1.41%
-1.72%
-0.58%
-29.37%
-10.94%
After-Rates Volum e Forecas t
2005
1.18%
2006
1.08%
2007
1.06%
3.91%
-3.86%
-2.45%
-10.11%
-10.19%
-10.03%
0.00%
-5.11%
0.00%
0.20%
0.18%
0.20%
-2.36%
-0.99%
-0.88%
-4.71%
3.14%
0.00%
-11.91%
-15.27%
-11.91%
-2.54%
-0.86%
-27.37%
-10.11%
-5.11%
-1.73%
0.58%
0.19%
-4.17%
-1.41%
-1.72%
-0.58%
-34.25%
-13.04%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 55
Estimated Impact of Factors Affecting Registered Mail Volume, 1994 – 2007
Firs t-Clas s
Pos tage
Total Change
Other Factors
Letters Volum e
Trends
Price
Inflation Econom etric
Other
in Volum e
6.15%
-10.33%
-0.58%
0.23%
-5.12%
0.69%
-8.30%
3.90%
-9.72%
-0.27%
0.19%
-1.87%
-4.62%
-11.35%
5.53%
-10.45%
-0.08%
0.21%
-5.68%
-1.10%
-10.68%
4.16%
-8.89%
-0.36%
0.10%
-35.31%
-5.31%
-41.43%
3.97%
-10.38%
-1.41%
0.13%
-2.41%
2.49%
-6.93%
3.40%
-10.74%
-0.78%
0.28%
-4.22%
12.22%
0.04%
-0.58%
-10.01%
-1.56%
0.23%
-0.82%
-1.59%
-12.84%
-3.20%
-9.39%
-0.81%
0.14%
-2.51%
-6.17%
-19.36%
-3.96%
-10.21%
-0.57%
0.20%
3.21%
-2.66%
-12.58%
-2.63%
-9.89%
0.00%
0.19%
-1.49%
4.20%
-8.72%
2004 - 2007
Total
Avg per Year
3.36%
1.11%
USPS-T-7
204
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 55 above.
5
Registered Mail has an elasticity with respect to First-Class letters volume of 1.173
6
(t-statistic of 3.238), so that Registered Mail volume is very close to proportional to First-
7
Class letters volume.
8
The time trend has a coefficient of -0.027 (t-statistic of -32.33) in the Registered Mail
9
equation. This explains annual declines in Registered Mail volume of approximately 10
10
percent per year.
11
The own-price elasticity of Registered Mail was calculated to be equal to -0.099
12
(t−statistic of -0.971). The Postal price impacts shown in Table 55 above are the result
13
of changes in nominal prices. Prices enter the demand equations developed here in
14
real terms, however. The impact of inflation reported in Table 55 measures the impact
15
that a change in real Postal prices, in the absence of nominal rate changes, has on the
16
volume of Registered Mail.
17
Other econometric variables include seasonal variables, a dummy variable for the
18
fact that the volume data prior to 1998 include both domestic and international
19
Registered Mail volumes, and a dummy variable to account for the temporary impact of
20
the September 11, 2001, terrorist attacks. A more detailed look at the econometric
21
demand equation for Registered Mail follows.
b. Econometric Demand Equation
22
23
The demand equation for Registered Mail in this case models Registered Mail
24
volume per adult per delivery day as a function of the following explanatory variables:
25
·
Seasonal Variables
USPS-T-7
205
1
2
3
4
5
6
7
8
9
10
11
12
13
·
First-Class Letters Volume (per adult per delivery day)
·
Linear Time Trend
·
Dummy variable for the mixing of domestic and international Registered Mail
volume, equal to zero through 1997Q4, and equal to one from 1998Q1
forward
·
Dummy variable for September 11th, equal to one in 2002Q1, zero elsewhere
·
Current price of Registered Mail
Details of the econometric demand equation are shown in Table 56 below. A
14
detailed description of the econometric methodologies used to obtain these results can
15
be found in Section III below.
USPS-T-7
206
1
2
3
TABLE 56
ECONOMETRIC DEMAND EQUATION FOR REGISTERED MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (current only)
-0.099
-0.971
First-Class Letters Volume
1.173
3.238
Time Trend
-0.027
-32.33
Dummy for Separation of Domestic
-0.403
-10.70
and International Registered Volumes
Dummy for 2002Q1 (9/11 Effect)
-0.145
-1.951
Seasonal Coefficients
September 1 – 15
0.672
0.487
September 16 – December 10
-1.087
-1.656
December 11 – 17
-0.069
-0.071
December 18 – 24
-4.895
-1.200
December 25 – 31
8.739
1.612
January – March
-1.718
-2.121
April – May
-0.640
-1.055
June
-3.735
-1.954
Quarter 1 (October – December)
-0.398
-2.095
Quarter 2 (January – March)
0.667
2.457
Quarter 3 (April – June)
0.657
1.698
Quarter 4 (July – September)
-0.926
-2.215
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.980981
Quarter 2 (January – March)
0.983513
Quarter 3 (April – June)
1.005482
Quarter 4 (July – September)
1.029102
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
52
Mean-Squared Error
0.004528
Adjusted R-Squared
0.992
4
5
USPS-T-7
207
1
2
3
4
5
6
7
8
9
10
11
3. Postal Insurance
a. Factors Affecting Insured Mail Volume
Insured Mail volume was found to be principally affected by the following variables:
Parcel Post Volume
Time Trends
Price of Postal Insurance
The effect of these variables on Insured Mail volume over the past ten years is
shown in Table 57 on the next page. Table 57 also shows the projected impacts of
these variables through GFY 2007.
The Test Year before-rates volume forecast for Insured mail is 35.903 million pieces,
12
a 30.3 percent decline from GFY 2004. The Postal Service’s proposed rates in this
13
case are predicted to reduce the Test Year volume of Insured mail by 1.5 percent, for a
14
Test Year after-rates volume forecast for Insured mail of 35.366 million.
USPS-T-7
208
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.14%
1.09%
1.22%
1.22%
1.28%
1.47%
1.25%
1.29%
1.29%
1.17%
1994 - 2004
Total
Avg per Year
13.15%
1.24%
8.72%
0.84%
53.99%
4.41%
-12.04%
-1.27%
4.34%
0.43%
-8.76%
-0.91%
1.92%
0.19%
61.67%
4.92%
2001 - 2004
Total
Avg per Year
3.80%
1.25%
-7.21%
-2.46%
-12.06%
-4.19%
-5.05%
-1.71%
1.34%
0.44%
7.98%
2.59%
-0.32%
-0.11%
-12.26%
-4.27%
Forecas t
1.13%
1.05%
1.03%
-3.30%
-3.88%
-3.66%
-17.44%
-18.23%
-18.38%
0.00%
0.00%
0.00%
0.49%
0.52%
0.43%
3.85%
3.16%
3.64%
-1.98%
2.46%
0.00%
-17.41%
-15.62%
-17.32%
3.25%
1.07%
-10.45%
-3.61%
-44.90%
-18.02%
0.00%
0.00%
1.44%
0.48%
11.03%
3.55%
0.43%
0.14%
-42.37%
-16.78%
After-Rates Volum e Forecas t
2005
1.13%
2006
1.05%
2007
1.03%
-3.30%
-5.32%
-3.73%
-17.44%
-18.23%
-18.38%
0.00%
0.00%
-1.00%
0.49%
0.52%
0.43%
3.85%
3.16%
3.64%
-1.98%
2.46%
0.00%
-17.41%
-16.88%
-18.21%
-11.86%
-4.12%
-44.90%
-18.02%
-1.00%
-0.33%
1.44%
0.48%
11.03%
3.55%
0.43%
0.14%
-43.85%
-17.50%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 57
Estimated Impact of Factors Affecting Postal Insurance Volume, 1994 – 2007
Parcel Pos t
Pos tage
Total Change
Other Factors
Volum e
Trends
Price
Inflation Econom etric
Other
in Volum e
2.84%
-4.99%
0.00%
0.44%
-2.99%
-9.09%
-12.46%
2.20%
-5.06%
0.04%
0.50%
-1.54%
3.62%
0.60%
5.17%
-5.39%
-1.12%
0.48%
8.73%
7.73%
17.20%
5.09%
44.97%
-2.98%
0.37%
-15.58%
-6.07%
19.08%
2.00%
22.82%
-1.00%
0.22%
-2.79%
-2.06%
19.85%
-0.95%
10.01%
-1.73%
0.32%
-3.40%
14.99%
21.08%
-0.17%
4.76%
-0.78%
0.60%
2.62%
-4.76%
3.31%
-1.63%
2.31%
-2.25%
0.52%
4.38%
-4.68%
-0.34%
-1.85%
-0.15%
-1.28%
0.32%
1.42%
-1.37%
-1.66%
-3.89%
-13.91%
-1.60%
0.49%
2.00%
6.03%
-10.48%
2004 - 2007
Total
Avg per Year
3.25%
1.07%
USPS-T-7
209
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 57 above.
5
Postal insurance has an elasticity with respect to Parcel Post volume of 0.296
6
(t−statistic of 5.378). Declining Parcel Post volume reduced the level of insured mail by
7
7.2 percent over the past three years. Parcel Post volume is projected to continue to
8
decline over the next three years, leading to an additional 10.4 percent decline in the
9
volume of Postal insurance from GFY 2004 through GFY 2007 (before-rates).
10
The insurance equation includes three time trends. The first of these is a linear time
11
trend over the full sample period. This time trend has an estimated coefficient of -0.013
12
(t-statistic of -11.95). This trend explains long-run annual declines in the volume of
13
insured mail of approximately 5 percent per year.
14
In June, 1997, special service classification reform (MC96-3) was enacted. One
15
aspect of this classification reform was an increase in the maximum amount of
16
insurance available from the Postal Service from $600 to $5,000. This change led to an
17
increase in the volume of mail which was insured. This increase is modeled here
18
through a logistic time trend starting in 1997Q4. A logistic time trend has the feature
19
that it increases at a decreasing rate. Hence, for example, the logistic time trend
20
explains a 51.1 percent increase in insurance volume in the first year after MC96-3 (FY
21
1998), and a 28.8 percent increase in volume the next year (1999). By 2004, the impact
22
of the logistic time trend has fallen to 5.0 percent, and by 2007, it is projected to fall still
23
further, to a mere 3.3 percent positive impact.
USPS-T-7
210
1
The third trend that is included in the insurance equation is a linear time trend
2
starting in 2003Q4. This time trend has a coefficient of -0.047, a t-statistic of -3.993,
3
and translates into 17 percent annual declines in the volume of postal insurance.
4
While the Postal Service charges a fee to insure a package which it delivers (via
5
Priority Mail or Parcel Post, for example), the Postal Service’s main competitors,
6
including UPS and FedEx, do not. Rather, UPS and FedEx provide free insurance as
7
part of their basic price.
8
In 2003, UPS introduced UPS Basic service to compete directly with the Postal
9
Service’s destination-entry Parcel Post mail category. As a result of this and other
10
competitive factors occurring around this time, destination-entry Parcel Post volume has
11
declined considerably since 2003Q4. The demand equation for destination-entry Parcel
12
Post volume includes a time trend starting in 2003Q4 to account for this. This was
13
described earlier in my testimony, in the section on destination-entry Parcel Post.
14
Because UPS and other competitors provide insurance at no extra charge, insured
15
Parcel Post and Priority Mail may be more vulnerable to competition than uninsured
16
Parcel Post and Priority Mail. This hypothesis is supported by volume data. From
17
2003Q4 through 2005Q1, for example, the percentage change in destination-entry
18
Parcel Post volume over the same period a year earlier averaged −5.2 percent, while
19
the percentage change in insured mail volume averaged -10.5 percent over this same
20
time period.
21
The own-price elasticity of insured mail was calculated to be equal to -0.230
22
(t−statistic of -3.486). The Postal price impacts shown in Table 57 above are the result
23
of changes in nominal prices. Prices enter the demand equations developed here in
24
real terms, however. The impact of inflation reported in Table 57 measures the impact
USPS-T-7
211
1
that a change in real Postal prices, in the absence of nominal rate changes, has on the
2
volume of insured mail.
3
Other econometric variables include seasonal variables and several dummy
4
variables, which are described below. A more detailed look at the econometric demand
5
equation for Postal insurance follows.
b. Econometric Demand Equation
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
The demand equation for insured mail in this case models insured mail volume per
adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
·
Parcel Post Volume (per adult per delivery day)
·
Linear Time Trend
·
Logistic Time Trend starting in1997Q4
·
Linear Time Trend starting in 2003Q4
·
Dummy variable equal to one starting in 1993Q1 for a change in the
methodology used to report Parcel Post volumes
·
Dummy variable for the general UPS strike in the summer of 1997, equal to
one in 1997Q4, zero elsewhere
·
Dummy variable for the implementation of MC96-3, equal to one starting in
1997Q4
·
Price of Postal Insurance lagged four quarters
Details of the econometric demand equation are shown in Table 58 below. A
30
detailed description of the econometric methodologies used to obtain these results can
31
be found in Section III below.
USPS-T-7
212
1
2
3
TABLE 58
ECONOMETRIC DEMAND EQUATION FOR INSURANCE
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (Lag 4 Only)
-0.230
-3.486
Parcel Post Mail Volume
0.296
5.378
Time Trends
Full-Sample Linear Trend
-0.013
-11.95
Logistic Trend since 1997Q4
0.330
10.31
Linear Trend since 2003Q4
-0.047
-3.993
Dummy for 1993 RPW Change
-0.147
-2.514
Dummy for UPS Strike (1997Q4)
0.339
3.646
Dummy for MC96-3
-0.058
-0.819
Seasonal Coefficients
September 1 – 15
-1.957
-1.603
September 16 – October
-0.827
-2.505
November 1 – December 10
0.422
1.364
December 11 – 17
1.209
1.861
December 18 – 21
0.433
0.558
December 22 – 24
-2.778
-3.194
December 25 – 31
-0.805
-1.279
January – February
-0.223
-0.986
March
-0.838
-2.402
April 1 – 15
0.751
0.980
April 16 –June
-0.641
-1.863
Quarter 1 (October – December)
-0.080
-1.990
Quarter 2 (January – March)
0.098
2.969
Quarter 3 (April – June)
0.009
0.260
Quarter 4 (July – September)
-0.027
-0.530
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.145063
Quarter 2 (January – March)
1.005732
Quarter 3 (April – June)
0.959860
Quarter 4 (July – September)
0.890328
REGRESSION DIAGNOSTICS
Sample Period
1971Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.218
Degrees of Freedom
112
Mean-Squared Error
0.005216
Adjusted R-Squared
0.983
4
5
USPS-T-7
213
1
2
3
4. Certified Mail
a. Factors Affecting Certified Mail Volume
Certified Mail volume was found to be principally affected by the following variables:
First-Class Letters Volume
Linear Time Trend
Price of Certified Mail
4
5
6
7
8
The effect of these variables on Certified Mail volume over the past ten years is
9
shown in Table 59 on the next page. Table 59 also shows the projected impacts of
10
these variables through GFY 2007.
11
The Test Year before-rates volume forecast for Certified Mail is 282.145 million
12
pieces, a 3.1 percent increase from GFY 2004. The Postal Service’s proposed rates in
13
this case are predicted to reduce the Test Year volume of Certified Mail by 1.2 percent,
14
for a Test Year after-rates volume forecast for Certified Mail of 278.811 million.
USPS-T-7
214
Other
-1.22%
-0.63%
3.24%
-3.68%
-4.53%
4.86%
-2.61%
0.59%
3.48%
-1.94%
Total Change
in Volum e
5.65%
3.28%
6.93%
-2.96%
-3.78%
1.30%
-0.58%
5.40%
-4.25%
0.84%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.19%
1.17%
1.25%
1.17%
1.18%
1.36%
1.25%
1.34%
1.28%
1.23%
1994 - 2004
Total
Avg per Year
13.13%
1.24%
10.00%
0.96%
35.24%
3.06%
-13.92%
-1.49%
3.37%
0.33%
-23.24%
-2.61%
-2.87%
-0.29%
11.66%
1.11%
2001 - 2004
Total
Avg per Year
3.90%
1.28%
-9.01%
-3.10%
9.42%
3.05%
-7.39%
-2.53%
0.99%
0.33%
3.05%
1.01%
2.07%
0.69%
1.76%
0.58%
Forecas t
1.21%
1.13%
1.11%
0.14%
-2.16%
-1.69%
3.02%
3.10%
3.08%
0.00%
0.00%
0.00%
0.37%
0.32%
0.32%
-1.11%
-0.17%
-1.02%
-3.89%
1.30%
0.00%
-0.40%
3.50%
1.76%
3.50%
1.15%
-3.68%
-1.24%
9.48%
3.07%
0.00%
0.00%
1.02%
0.34%
-2.28%
-0.77%
-2.64%
-0.89%
4.89%
1.61%
After-Rates Volum e Forecas t
2005
1.21%
2006
1.13%
2007
1.11%
0.14%
-3.16%
-2.27%
3.02%
3.10%
3.08%
0.00%
-0.16%
-0.24%
0.37%
0.32%
0.32%
-1.11%
-0.17%
-1.02%
-3.89%
1.30%
0.00%
-0.40%
2.27%
0.91%
-5.23%
-1.77%
9.48%
3.07%
-0.40%
-0.13%
1.02%
0.34%
-2.28%
-0.77%
-2.64%
-0.89%
2.79%
0.92%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 59
Estimated Impact of Factors Affecting Certified Mail Volume, 1994 – 2007
Firs t-Clas s
Pos tage
Other Factors
Letters Volum e
Trends
Price
Inflation Econom etric
2.57%
3.01%
-0.32%
0.38%
-0.02%
3.70%
3.10%
-1.15%
0.37%
-3.15%
3.73%
3.17%
-0.12%
0.37%
-4.65%
3.31%
3.02%
-2.55%
0.27%
-4.24%
2.84%
3.01%
-1.24%
0.17%
-4.95%
2.23%
3.11%
-0.45%
0.34%
-9.48%
0.86%
3.10%
-1.42%
0.45%
-2.08%
-3.04%
3.10%
-5.04%
0.35%
8.53%
-3.98%
2.97%
-2.00%
0.30%
-5.99%
-2.26%
3.07%
-0.49%
0.33%
1.00%
2004 - 2007
Total
Avg per Year
3.50%
1.15%
USPS-T-7
215
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 59 above.
5
Certified Mail has an elasticity with respect to First-Class letters volume of 0.767
6
(t−statistic of 3.894), so that Certified Mail volume is fairly proportional to First-Class
7
letters volume.
8
The time trend in the Certified Mail equation has a coefficient of 0.008 (t-statistic of
9
12.56). This trend explains annual increases in Certified Mail volume of approximately
10
11
3 percent per year.
The own-price elasticity of Certified Mail was calculated to be equal to -0.183
12
(t−statistic of -3.420). The Postal price impacts shown in Table 59 above are the result
13
of changes in nominal prices. Prices enter the demand equations developed here in
14
real terms, however. The impact of inflation reported in Table 59 measures the impact
15
that a change in real Postal prices, in the absence of nominal rate changes, has on the
16
volume of Certified Mail.
17
Other econometric variables include seasonal variables, a dummy variable for the
18
introduction of delivery confirmation, which provides similar services as Certified Mail at
19
a much smaller price, and dummy variables to account for the temporary impact of the
20
2001 terrorist and bioterrorist attacks. A more detailed look at the econometric demand
21
equation for Certified Mail follows.
b. Econometric Demand Equation
22
23
24
25
The demand equation for Certified Mail in this case models Certified Mail volume
per adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
USPS-T-7
216
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
·
First-Class Letters Volume (per adult per delivery day)
·
Linear Time Trend
·
Dummy variable equal to one starting in 1988Q1, to reflect a change in the
reporting of volume. Prior to 1988, mail sent by the Federal government was
classified separately from non-government mail; since 1988, government mail
has been distributed to the appropriate mail categories and special services.
·
Dummy variables equal to one in 2002Q1, 2002Q2, and 2002Q3,
respectively, zero elsewhere, to reflect the short-term impact of September 11
and bioterrorism on Certified Mail volume
·
Dummy variable equal to one since the introduction of delivery confirmation
by the Postal Service
· Current and four lags of the price of Certified Mail
Details of the econometric demand equation are shown in Table 60 below. A
20
detailed description of the econometric methodologies used to obtain these results can
21
be found in Section III below.
USPS-T-7
217
1
2
3
TABLE 60
ECONOMETRIC DEMAND EQUATION FOR CERTIFIED MAIL
Coefficient
T-Statistic
Own-Price Elasticity
Long-Run
-0.183
-3.420
Current
-0.001
-0.008
Lag 1
-0.073
-0.295
Lag 2
-0.039
-0.145
Lag 3
-0.026
-0.096
Lag 4
-0.043
-0.264
First-Class Letters Volume
0.767
3.894
Time Trend
0.008
12.56
Dummy for Distribution of Government Mail
0.070
2.049
(1988Q1)
Introduction of Delivery Confirmation
-0.126
-4.059
Dummies for Quarters Immediately Following
9/11, etc.
0.050
0.526
2002Q1
0.125
1.442
2002Q2
2002Q3
0.075
0.856
Seasonal Coefficients
-0.994
-0.677
September 1 – 15
1.026
2.705
September 16 – 30
0.584
1.492
October
-0.453
-1.211
November 1 – December 10
0.388
0.969
December 11 – 31
-0.110
-0.426
January – February
0.795
2.505
March – April 15
-0.362
-1.259
April 16 – May
0.969
2.528
June
Quarter 1 (October – December)
-0.005
-0.120
Quarter 2 (January – March)
-0.038
-0.959
Quarter 3 (April – June)
-0.076
-1.527
Quarter 4 (July – September)
0.119
1.854
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.929163
Quarter 2 (January – March)
1.030982
Quarter 3 (April – June)
1.057155
Quarter 4 (July – September)
0.986077
REGRESSION DIAGNOSTICS
Sample Period
1971Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
112
Mean-Squared Error
0.005856
Adjusted R-Squared
0.961
4
USPS-T-7
218
1
2
3
4
5
6
7
5. Collect on Delivery (COD) Mail
a. Factors Affecting COD Mail Volume
COD Mail volume was found to be principally affected by the following variables:
Linear Time Trend
Price of COD Mail
The effect of these variables on COD volume over the past ten years is shown in
8
Table 61 on the next page. Table 61 also shows the projected impacts of these
9
variables through GFY 2007.
10
The Test Year before-rates volume forecast for COD mail is 1.693 million pieces, an
11
11.1 percent decline from GFY 2004. The Postal Service’s proposed rates in this case
12
are predicted to reduce the Test Year volume of COD mail by 1.1 percent, for a Test
13
Year after-rates volume forecast for COD mail of 1.673 million.
USPS-T-7
219
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.18%
1.10%
1.16%
1.16%
1.20%
1.29%
1.13%
1.20%
1.24%
1.16%
1994 - 2004
Total
Avg per Year
12.47%
1.18%
-50.03%
-6.70%
-26.30%
-3.01%
11.97%
1.14%
-35.67%
-4.32%
15.25%
1.43%
-65.62%
-10.13%
2001 - 2004
Total
Avg per Year
3.65%
1.20%
-18.36%
-6.54%
-3.94%
-1.33%
3.21%
1.06%
-29.77%
-11.11%
15.32%
4.87%
-32.05%
-12.08%
-6.68%
-7.04%
-6.83%
0.00%
0.00%
0.00%
1.18%
1.15%
1.08%
-1.43%
-0.16%
-0.28%
-4.69%
4.27%
0.00%
-10.25%
-1.00%
-5.07%
3.43%
1.13%
-19.18%
-6.85%
0.00%
0.00%
3.45%
1.14%
-1.86%
-0.62%
-0.62%
-0.21%
-15.65%
-5.52%
After-Rates Volum e Forecas t
2005
1.18%
2006
1.13%
2007
1.08%
-6.68%
-7.04%
-6.83%
0.00%
-1.14%
-1.95%
1.18%
1.15%
1.08%
-1.43%
-0.16%
-0.28%
-4.69%
4.27%
0.00%
-10.25%
-2.13%
-6.93%
-19.18%
-6.85%
-3.08%
-1.04%
3.45%
1.14%
-1.86%
-0.62%
-0.62%
-0.21%
-18.25%
-6.50%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 61
Estimated Impact of Factors Affecting COD Mail Volume, 1994 – 2007
Pos tage
Total Change
Other Factors
Trends
Price
Inflation Econom etric
Other
in Volum e
-6.84%
-3.63%
1.31%
-1.01%
5.75%
-3.66%
-6.74%
-12.43%
1.43%
-3.12%
12.07%
-9.07%
-6.71%
-2.68%
1.23%
8.10%
-4.31%
-3.83%
-6.83%
0.00%
1.04%
-8.85%
-3.26%
-16.02%
-7.11%
-1.27%
0.67%
-1.09%
11.14%
2.71%
-6.75%
-3.60%
1.07%
0.25%
11.24%
2.63%
-6.42%
-1.86%
1.45%
-2.25%
-26.32%
-32.14%
-6.39%
-3.29%
1.13%
-28.08%
22.16%
-18.60%
-6.60%
-0.67%
0.92%
-4.14%
-10.42%
-18.60%
-6.62%
0.00%
1.12%
1.87%
5.39%
2.56%
2004 - 2007
Total
Avg per Year
Forecas t
1.18%
1.13%
1.08%
3.43%
1.13%
USPS-T-7
220
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 61 above.
5
6
The time trend in the COD equation has a coefficient of -0.018 (t-statistic of -19.56).
This coefficient translates into an annual trend of approximately -7 percent.
7
The own-price elasticity of COD mail was calculated to be equal to -0.592 (t−statistic
8
of -2.800). The Postal price impacts shown in Table 61 above are the result of changes
9
in nominal prices. Prices enter the demand equations developed here in real terms,
10
however. The impact of inflation reported in Table 61 measures the impact that a
11
change in real Postal prices, in the absence of nominal rate changes, has on the
12
volume of COD mail.
13
Other econometric variables include seasonal variables and dummy variables which
14
measure the impact of UPS’s 1997 strike as well as the 2001 terrorist and bioterrorist
15
attacks. A more detailed look at the econometric demand equation for COD Mail
16
follows.
b. Econometric Demand Equation
17
18
19
20
21
22
23
24
25
26
27
28
The demand equation for COD mail in this case models COD mail volume per
adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
·
Linear Time Trend
·
Dummy variable equal to one in 1997Q4, zero elsewhere, reflecting the
impact of the UPS strike in the summer of 1997 on mail volumes
·
Dummy variable equal to one in 2002Q1, zero elsewhere, and a second
dummy variable equal to one from 2002Q2 onward, reflecting the possible
USPS-T-7
221
1
2
3
4
5
long-run impact on COD volumes of bioterrorism and security concerns
arising as a result of the September 11, 2001, and anthrax attacks
·
Current and four lags of the price of COD Mail
Details of the econometric demand equation are shown in Table 62 below. A
6
detailed description of the econometric methodologies used to obtain these results can
7
be found in Section III below.
USPS-T-7
222
1
2
3
TABLE 62
ECONOMETRIC DEMAND EQUATION FOR COLLECT-ON-DELIVERY MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.592
-2.800
Current
-0.174
-0.817
Lag 1
-0.000
-0.000
Lag 2
-0.005
-0.018
Lag 3
-0.165
-0.650
Lag 4
-0.248
-1.138
Time Trend
-0.018
-19.56
Dummy for UPS Strike (1997Q4)
0.299
3.205
Dummy for 2002Q1 (9/11 Effect)
-0.159
-1.474
Dummy Since 2002Q2
-0.387
-4.200
Seasonal Coefficients
September 1 – 15
4.101
2.918
September 16 – 30
0.625
2.746
October – June
0.895
3.607
Quarter 1 (October – December)
-0.066
-1.992
Quarter 2 (January – March)
-0.015
-0.469
Quarter 3 (April – June)
-0.041
-1.292
Quarter 4 (July – September)
0.122
3.334
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.971806
Quarter 2 (January – March)
1.022711
Quarter 3 (April – June)
0.996876
Quarter 4 (July – September)
1.008465
REGRESSION DIAGNOSTICS
Sample Period
1971Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.654
Degrees of Freedom
119
Mean-Squared Error
0.010895
Adjusted R-Squared
0.984
4
5
USPS-T-7
223
6. Return Receipts
1
a. Factors Affecting Return Receipts Volume
2
3
4
5
6
7
8
9
10
11
Return Receipts volume was found to be principally affected by the following
variables:
Certified Mail Volume
Price of Return Receipts
The effect of these variables on Return Receipts volume over the past ten years is
shown in Table 63 on the next page. Table 63 also shows the projected impacts of
these variables through GFY 2007.
The Test Year before-rates volume forecast for Return Receipts is 250.973 million
12
pieces, a 5.2 percent increase from GFY 2004. The Postal Service’s proposed rates in
13
this case are predicted to reduce the Test Year volume of Return Receipts by 2.0
14
percent, for a Test Year after-rates volume forecast for Return Receipts of 245.970
15
million.
USPS-T-7
224
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Table 63
Estimated Impact of Factors Affecting Return Receipts Volume, 1994 – 2007
Certified Mail
Pos tage
Total Change
Other Factors
Population
Volum e
Price
Inflation Econom etric
Other
in Volum e
1.30%
8.48%
-1.15%
0.41%
14.92%
-1.73%
23.18%
1.12%
3.51%
-0.50%
0.34%
-3.60%
-1.49%
-0.76%
1.30%
7.86%
0.02%
0.37%
-1.22%
6.41%
15.30%
1.12%
-0.42%
0.07%
0.18%
-4.96%
-6.71%
-10.49%
1.19%
-3.85%
-1.10%
0.23%
-2.43%
2.96%
-3.11%
1.38%
-0.09%
-0.60%
0.49%
-0.60%
1.76%
2.34%
1.24%
-1.23%
0.35%
0.40%
-0.54%
-0.95%
-0.75%
1.36%
2.48%
-0.39%
0.26%
2.76%
1.18%
7.85%
1.27%
-5.20%
-1.52%
0.35%
0.57%
-2.86%
-7.32%
1.24%
-0.82%
0.00%
0.36%
0.51%
1.89%
3.19%
1994 – 2004
Total
Avg per Year
13.22%
1.25%
10.27%
0.98%
-4.73%
-0.48%
3.45%
0.34%
4.20%
0.41%
-0.12%
-0.01%
28.06%
2.50%
2001 – 2004
Total
Avg per Year
3.91%
1.29%
-3.64%
-1.23%
-1.91%
-0.64%
0.97%
0.32%
3.87%
1.27%
0.14%
0.05%
3.14%
1.04%
Forecas t
1.23%
1.13%
1.11%
-1.36%
1.77%
0.50%
0.00%
0.00%
0.00%
0.36%
0.30%
0.35%
-0.01%
0.23%
-0.05%
1.86%
-0.37%
0.00%
2.07%
3.08%
1.92%
3.52%
1.16%
0.88%
0.29%
0.00%
0.00%
1.01%
0.34%
0.17%
0.06%
1.48%
0.49%
7.23%
2.35%
After-Rates Volum e Forecas t
2005
1.23%
2006
1.13%
2007
1.11%
-1.36%
0.75%
-0.21%
0.00%
-1.00%
0.00%
0.36%
0.30%
0.35%
-0.01%
0.23%
-0.05%
1.86%
-0.37%
0.00%
2.07%
1.02%
1.20%
-0.83%
-0.28%
-1.00%
-0.34%
1.01%
0.34%
0.17%
0.06%
1.48%
0.49%
4.35%
1.43%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
2004 - 2007
Total
Avg per Year
3.52%
1.16%
USPS-T-7
225
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 63 above.
5
Return Receipts have an elasticity with respect to Certified Mail volume of 0.845
6
(t−statistic of 11.75), so that return receipt volume is fairly proportional to Certified Mail
7
volume. The own-price elasticity of Return Receipts was calculated to be equal to
8
-0.180 (t−statistic of -1.036).
9
The Postal price impacts shown in Table 63 above are the result of changes in
10
nominal prices. Prices enter the demand equations developed here in real terms,
11
however. The impact of inflation reported in Table 63 measures the impact that a
12
change in real Postal prices, in the absence of nominal rate changes, has on the
13
volume of return receipt mail.
14
Other econometric variables include seasonal variables and several dummy
15
variables which are described below. A more detailed look at the econometric demand
16
equation for Return Receipts follows.
b. Econometric Demand Equation
17
18
19
20
21
22
23
24
25
26
27
28
29
The demand equation for Return Receipts in this case models return receipt volume
per adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
·
Certified Mail Volume (per adult per delivery day)
·
Dummy variable equal to one starting in 1995Q2 which measures the
apparent impact of a 1995 RPW sampling change on the measurement of
return receipt volumes
·
Two dummy variables equal to one in 1995Q2 and 1997Q2, respectively, and
zero elsewhere
USPS-T-7
226
1
2
3
4
5
6
7
8
9
10
·
Dummy variable equal to one in 2002Q1, zero elsewhere, and a second
dummy variable equal to one from 2002Q2 onward, reflecting the possible
long-run impact on Return Receipt volumes of security concerns arising from
the terrorism and bioterrorism attacks in the fall of 2001
·
Current price of Return Receipts
Details of the econometric demand equation are shown in Table 64 below. A
detailed description of the econometric methodologies used to obtain these results can
be found in Section III below.
USPS-T-7
227
1
2
3
TABLE 64
ECONOMETRIC DEMAND EQUATION FOR RETURN RECEIPTS
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (current only)
-0.180
-1.036
Certified Mail Volume
0.845
11.75
Dummy for RPW Sampling Change (1995Q2)
0.121
8.476
Dummy Variables for Individual Quarters
1995Q2
0.245
6.975
1997Q2
0.115
2.710
Dummy Variables for 9/11 and Aftermath
Initial Impact (2002Q1)
0.016
0.545
Long-Run Impact (2002Q2 onward)
0.020
1.830
Seasonal Coefficients
November 1 – December 10
0.056
2.102
December 11 – 24
-6.114
-3.109
December 25 – 31
13.00
2.853
January – March
0.105
2.904
Quarter 1 (October – December)
0.088
3.511
Quarter 2 (January – March)
-0.093
-3.184
Quarter 3 (April – June)
0.013
1.318
Quarter 4 (July – September)
-0.008
-0.806
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.996477
Quarter 2 (January – March)
1.007423
Quarter 3 (April – June)
1.008795
Quarter 4 (July – September)
0.987407
REGRESSION DIAGNOSTICS
Sample Period
1993Q1 – 2005Q1
Autocorrelation Coefficients
AR-1: 0.212
AR-4: -0.575
Degrees of Freedom
28
Mean-Squared Error
0.000903
Adjusted R-Squared
0.913
4
5
USPS-T-7
228
1
2
3
7. Money Orders
a. History of Money Orders Volume
Table 65 below shows the history of money orders volume since 1988. From 1988
4
through 2000, money orders volume showed consistent annual growth which averaged
5
4 percent per year. Starting in the second quarter of 2001, however, money orders
6
volume has declined over the same period the year before for 16 consecutive quarters.
7
Over the past few years, money orders have faced increasing competition, which
8
has taken two basic forms. First, the number of places where money orders are
9
available has increased. For example, Wal-Mart began selling money orders within the
10
past five to six years. Second, a number of alternatives to money orders are now
11
available. One example of this is pre-paid debit cards, whereby people without bank
12
accounts can be paid with a debit card. Therefore, the un-banked do not need to
13
purchase money orders to pay bills. Also, in some cities, major utility bills can be paid
14
directly at currency exchanges, again eliminating the need for money orders in many
15
circumstances.
16
The recent downturn in money order volumes observed in Table 65 is modeled
17
econometrically through a linear time trend starting in the fourth quarter of 2000. Money
18
orders volume is also affected by economic conditions, modeled by including total
19
private employment in the money orders equation. Finally, of course, money orders
20
volume is modeled as being a function in part of the price of money orders.
USPS-T-7
229
Table 65
Money Orders Volume
(millions of units)
FY
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000*
2001
2002
2003
2004
Quarter
2000.1
2000.2
2000.3
2000.4
2001.1
2001.2
2001.3
2001.4
2002.1
2002.2
2002.3
2002.4
2003.1
2003.2
2003.3
2003.4
2004.1
2004.2
2004.3
2004.4
2005.1
Volume
142.614
150.075
154.848
161.673
176.233
185.513
195.816
200.271
211.510
206.947
207.389
219.059
231.213
227.004
216.867
198.454
187.211
Annual Change
0.62%
5.23%
3.18%
4.41%
9.01%
5.27%
5.55%
2.28%
5.61%
-2.16%
0.21%
5.63%
4.76%
-1.82%
-4.47%
-8.49%
-5.67%
Change from Same Quarter
Previous Year
55.723
60.346
58.563
56.581
57.271
2.78%
58.779
-2.60%
57.522
-1.78%
53.432
-5.57%
53.527
-6.54%
55.715
-5.21%
55.379
-3.73%
52.247
-2.22%
50.525
-5.61%
50.778
-8.86%
50.981
-7.94%
46.169
-11.63%
47.747
-5.50%
48.632
-4.23%
45.897
-9.97%
44.935
-2.67%
45.646
-4.40%
* Volume show n for 2000 is GFY; percentage change is
1
show n for PFY 2000
USPS-T-7
230
b. Factors Affecting Money Orders Volume
1
2
3
The volume of money orders was found to be principally affected by the following
variables:
Total Private Employment
Time Trend Starting in 2000Q4
Price of Money Orders
4
5
6
7
8
The effect of these variables on money orders volume over the past ten years is
9
shown in Table 66 on the next page. Table 66 also shows the projected impacts of
10
these variables through GFY 2007.
11
The Test Year before-rates volume forecast for money orders is 181.567 million
12
pieces, a 3.0 percent decline from GFY 2004. The Postal Service’s proposed rates in
13
this case are predicted to reduce the Test Year volume of money orders by 0.9 percent,
14
for a Test Year after-rates volume forecast for money orders of 179.939 million.
USPS-T-7
231
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.22%
1.13%
1.22%
1.18%
1.21%
1.42%
1.23%
1.29%
1.28%
1.19%
1994 – 2004
Total
Avg per Year
13.09%
1.24%
2.69%
0.27%
-14.61%
-1.57%
-12.58%
-1.34%
12.06%
1.14%
3.94%
0.39%
-5.30%
-0.54%
-4.39%
-0.45%
2001 – 2004
Total
Avg per Year
3.80%
1.25%
-4.68%
-1.58%
-11.48%
-3.98%
-9.86%
-3.40%
3.44%
1.13%
0.95%
0.32%
0.04%
0.01%
-17.53%
-6.22%
Forecas t
1.21%
1.11%
1.09%
0.58%
0.43%
0.03%
-3.97%
-4.01%
-4.00%
0.00%
0.00%
0.00%
1.22%
1.15%
1.07%
-0.37%
0.00%
-0.57%
-0.38%
0.16%
0.00%
-1.79%
-1.25%
-2.44%
3.45%
1.14%
1.04%
0.35%
-11.50%
-3.99%
0.00%
0.00%
3.48%
1.15%
-0.94%
-0.31%
-0.22%
-0.07%
-5.38%
-1.83%
After-Rates Volum e Forecas t
2005
1.21%
2006
1.11%
2007
1.09%
0.58%
0.43%
0.03%
-3.97%
-4.01%
-4.00%
0.00%
-0.90%
-2.20%
1.22%
1.15%
1.07%
-0.37%
0.00%
-0.57%
-0.38%
0.16%
0.00%
-1.79%
-2.13%
-4.59%
1.04%
0.35%
-11.50%
-3.99%
-3.08%
-1.04%
3.48%
1.15%
-0.94%
-0.31%
-0.22%
-0.07%
-8.29%
-2.84%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 66
Estimated Impact of Factors Affecting Money Orders Volume, 1994 – 2007
Total Change
Other Factors
Em ploym ent
Trends
Price
Inflation Econom etric
Other
in Volum e
1.96%
0.00%
-1.15%
1.27%
-0.67%
-0.33%
2.28%
0.93%
0.00%
-4.55%
1.33%
6.79%
0.17%
5.61%
1.35%
0.00%
-1.17%
1.21%
-3.87%
-0.83%
-2.16%
1.52%
0.00%
0.00%
1.03%
0.07%
-3.50%
0.21%
1.15%
0.00%
0.51%
0.64%
1.09%
0.91%
5.63%
1.14%
-0.27%
2.33%
1.05%
0.61%
-0.82%
5.55%
-0.55%
-3.28%
1.12%
1.52%
-0.79%
-1.00%
-1.82%
-2.70%
-4.03%
-2.58%
1.25%
0.93%
1.46%
-4.47%
-1.59%
-3.97%
-6.17%
0.99%
-0.33%
1.24%
-8.49%
-0.44%
-3.96%
-1.39%
1.16%
0.35%
-2.61%
-5.67%
2004 - 2007
Total
Avg per Year
3.45%
1.14%
USPS-T-7
232
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 66 above.
5
Money orders have an elasticity with respect to employment of 0.867 (t−statistic of
6
10.27), so that a 10 percent increase in employment would be expected to lead to an
7
8.67 percent increase in money orders volume.
8
9
10
11
The time trend since 2000Q4 has an estimated coefficient of -0.010 with a t-statistic
of -18.72. This coefficient translates into an annual trend of approximately -4 percent.
This trend is expected to continue throughout the forecast period presented in this case.
The own-price elasticity of money orders was calculated to be equal to -0.604
12
(t−statistic of -26.89). The Postal price impacts shown in Table 66 above are the result
13
of changes in nominal prices. Prices enter the demand equations developed here in
14
real terms, however. The impact of inflation reported in Table 66 measures the impact
15
that a change in real Postal prices, in the absence of nominal rate changes, has on the
16
volume of money order mail.
17
Other econometric variables include seasonal variables and a dummy variable for
18
one unusual quarter of data. A more detailed look at the econometric demand equation
19
for money orders follows.
c. Econometric Demand Equation
20
21
22
23
24
25
26
The demand equation for money orders in this case models money orders
volume per adult per delivery day as a function of the following explanatory variables:
·
Seasonal Variables
·
Total Private Employment
USPS-T-7
233
1
2
3
4
5
6
·
Linear Time Trend starting in 2000Q4
·
Dummy variable equal to one in 1996Q4, zero elsewhere
·
Current and four lags of the price of Money Orders
Details of the econometric demand equation are shown in Table 67 below. A
7
detailed description of the econometric methodologies used to obtain these results can
8
be found in Section III below.
USPS-T-7
234
1
2
3
TABLE 67
ECONOMETRIC DEMAND EQUATION FOR MONEY ORDERS
Coefficient T-Statistic
Own-Price Elasticity
Long-Run
-0.604
-26.89
Current
-0.092
-0.944
Lag 1
-0.083
-0.556
Lag 2
-0.002
-0.015
Lag 3
-0.078
-0.503
Lag 4
-0.349
-3.565
Employment
0.867
10.27
Time Trend Since 2000Q4
-0.010
-18.72
Dummy Variable for 1996Q4
0.181
8.120
Seasonal Coefficients
September 1 – 15
-1.086
-2.486
September 16 – 30
0.008
0.050
October
0.727
3.651
November 1 – December 12
-0.849
-4.577
December 13 – 17
0.517
1.818
December 18 – 24
2.389
1.833
December 25 – 31
-1.881
-0.445
January – February
-0.233
-0.562
March – May
-0.103
-1.904
Quarter 1 (October – December)
-0.069
-0.262
Quarter 2 (January – March)
0.112
0.428
Quarter 3 (April – June)
-0.056
-2.890
Quarter 4 (July – September)
0.013
0.517
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.005089
Quarter 2 (January – March)
1.043144
Quarter 3 (April – June)
0.991791
Quarter 4 (July – September)
0.961875
REGRESSION DIAGNOSTICS
Sample Period
1988Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
48
Mean-Squared Error
0.000432
Adjusted R-Squared
0.962
4
5
USPS-T-7
235
1
2
3
8. Delivery and Signature Confirmation
a. Factors Affecting Delivery and Signature Confirmation
Delivery confirmation was introduced by the Postal Service in April, 1999 for Priority
4
Mail and Parcel Post. Signature confirmation was introduced for these same mail
5
categories in January, 2001. Delivery and signature confirmation were both expanded
6
to include First-Class and Standard packages in June, 2002.
7
8
Delivery and signature confirmation volumes are shown in Table 68 below. Units in
Table 68 are millions of pieces.
Table 68
Delivery & Signa ture Confirma tion
9
Quarter
2000.1
2000.2
2000.3
2000.4
2001.1
2001.2
2001.3
2001.4
2002.1
2002.2
2002.3
2002.4
2003.1
2003.2
2003.3
2003.4
2004.1
2004.2
2004.3
2004.4
2005.1
Volume
28.247
27.248
33.184
34.363
45.447
49.239
45.935
48.754
60.130
66.494
70.577
85.752
141.336
122.978
128.467
121.868
170.268
141.945
145.438
141.691
194.095
Change from
Previous Year
60.89%
80.70%
38.43%
41.88%
32.31%
35.04%
53.64%
75.89%
135.05%
84.95%
82.02%
42.12%
20.47%
15.42%
13.21%
16.27%
13.99%
10
Given the relative newness and gradual expansion of this special service, it is not
11
too surprising that year-to-year growth in delivery and signature confirmation volume
12
growth has been double-digit throughout its history.
USPS-T-7
236
1
Not surprisingly, therefore, the dominant feature of the delivery and signature
2
confirmation equation used in this case is a time trend over the full sample period. A
3
logistic time trend is used here to reflect the fact that while delivery and signature
4
confirmation volume has continued to grow throughout its history, it is doing so at an
5
ever-decreasing rate.
6
7
8
9
10
11
In summary, then, delivery and signature confirmation volume was modeled as a
function of the following variables:
Time Trend
Price of Delivery and Signature Confirmation
The effect of these variables on delivery and signature confirmation volume over the
12
past three years is shown in Table 69 on the next page. Table 69 also shows the
13
projected impacts of these variables through GFY 2007.
14
The Test Year before-rates volume forecast for delivery and signature confirmation
15
is 724.011 million pieces, a 20.8 percent increase from GFY 2004. The Postal Service’s
16
proposed rates in this case are predicted to reduce the Test Year volume of delivery
17
and signature confirmation by 3.9 percent, for a Test Year after-rates volume forecast
18
for delivery and signature confirmation of 695.440 million.
USPS-T-7
237
2002
2003
2004
Table 69
Estimated Impact of Factors Affecting Delivery & Signature Confirmation Volume, 2001 – 2007
Pos tage
Total Change
Other Factors
Trends
Price
Inflation Econom etric
Other
in Volum e
Population
1.50%
21.71%
26.10%
0.57%
0.73%
-5.32%
49.41%
1.75%
17.01%
49.94%
0.80%
-3.33%
4.55%
81.89%
1.30%
10.31%
0.00%
0.94%
1.00%
2.23%
16.46%
2001 - 2004
Total
Avg per Year
57.10%
16.25%
89.07%
23.65%
2.33%
0.77%
-1.65%
-0.55%
1.20%
0.40%
216.48%
46.82%
8.16%
6.66%
5.72%
0.00%
0.00%
0.00%
0.91%
0.76%
0.89%
0.85%
-0.11%
-0.33%
0.44%
-0.65%
0.00%
11.98%
7.88%
7.52%
3.61%
1.19%
21.96%
6.84%
0.00%
0.00%
2.58%
0.85%
0.41%
0.14%
-0.21%
-0.07%
29.89%
9.11%
After-Rates Volum e Forecas t
2005
1.29%
2006
1.15%
2007
1.14%
8.16%
6.66%
5.72%
0.00%
-3.95%
0.00%
0.91%
0.76%
0.89%
0.85%
-0.11%
-0.33%
0.44%
-0.65%
0.00%
11.98%
3.62%
7.52%
21.96%
6.84%
-3.95%
-1.33%
2.58%
0.85%
0.41%
0.14%
-0.21%
-0.07%
24.76%
7.65%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
2004 - 2007
Total
Avg per Year
4.62%
1.52%
Forecas t
1.29%
1.15%
1.14%
3.61%
1.19%
USPS-T-7
238
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 69 above.
5
The logistic time trend included in the delivery confirmation equation increases at a
6
decreasing rate. This trend explains a 21.7 percent increase in the volume of delivery
7
and signature confirmation mail in 2002. The magnitude of this trend has decreased
8
over time, explaining a 17.0 percent increase in volume for 2003, and a 10.3 percent
9
increase in delivery and signature confirmation volume in 2004. The impact of this trend
10
is projected to continue to decline to 6.7 percent for GFY 2006 and only 5.7 percent by
11
GFY 2007.
12
The own-price elasticity of delivery and signature confirmation mail was calculated to
13
be equal to -0.466 (t−statistic of -0.925). The Postal price impacts shown in Table 69
14
above are the result of changes in nominal prices. Prices enter the demand equations
15
developed here in real terms, however. The impact of inflation reported in Table 69
16
measures the impact that a change in real Postal prices, in the absence of nominal rate
17
changes, has on the volume of delivery and signature confirmation mail.
18
Other econometric variables include seasonal variables and a dummy variable for
19
R2001-1, which expanded delivery and signature confirmation to First-Class and
20
Standard Mail. A more detailed look at the econometric demand equation for delivery
21
and signature confirmation follows.
USPS-T-7
239
b. Econometric Demand Equation
1
2
The demand equation for delivery and signature confirmation mail in this case
3
models delivery and signature confirmation mail volume per adult per delivery day as a
4
function of the following explanatory variables:
5
6
7
8
9
10
11
12
13
·
Simple quarterly dummy variables
·
Logistic Time Trend
·
Dummy variable for R2001-1, which significantly expanded the availability of
both delivery and signature confirmation
·
Current price of Delivery and Signature Confirmation Mail
Details of the econometric demand equation are shown in Table 70 below. A
14
detailed description of the econometric methodologies used to obtain these results can
15
be found in Section III below.
USPS-T-7
240
1
2
3
4
TABLE 70
ECONOMETRIC DEMAND EQUATION FOR
DELIVERY AND SIGNATURE CONFIRMATION MAIL
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (current only)
-0.466
-0.925
Logistic Time Trend
0.395
5.346
Dummy for R2001-1
0.380
1.826
Seasonal Coefficients
Quarter 1 (October – December)
0.329
5.802
Quarter 2 (January – March)
0.173
2.881
Quarter 3 (April – June)
0.123
2.074
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.181625
Quarter 2 (January – March)
1.010937
Quarter 3 (April – June)
0.962051
Quarter 4 (July – September)
0.850389
REGRESSION DIAGNOSTICS
Sample Period
2000Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
14
Mean-Squared Error
0.008566
Adjusted R-Squared
0.978
USPS-T-7
241
1
9. Post Office Boxes
a. Estimating the Number of Post Office Boxes
2
3
Post Office Boxes are different from other special services in several ways. First,
4
unlike all other special services modeled here except for money orders, Post Office
5
Boxes do not represent an add-on to another type of mail volume but instead represent
6
a separate service. Also, Post Office Boxes are unique in that they are not physical
7
items purchased one at a time. Instead, a Post Office Box is rented on an ongoing
8
basis. Finally, the number of Post Office Boxes being rented at any given time is not
9
reported by the Postal Service’s RPW (Revenue, Pieces, and Weight) system. Instead,
10
the RPW system only reports the revenue received by the Postal Service for Post Office
11
Box rents.
12
Because Post Office Box volumes are not reported by the Postal Service, it was
13
necessary for me to first estimate a measure of the number of occupied Post Office
14
Boxes before I could attempt to model a demand equation for Post Office Boxes. To do
15
this, a price index was constructed for Post Office Box rents using GFY 2004 billing
16
determinants. Post Office Box revenue, as reported by the RPW system, was then
17
divided by this price index to obtain a volume index for Post Office Boxes.
18
19
The estimated number of Post Office Boxes constructed in this way is shown in
Table 71 below. Revenues are expressed in millions of dollars in Table 71.
USPS-T-7
242
Table 71
Estimated Number of Post Office Boxes
1
Quarter
1993PQ1
1993PQ2
1993PQ3
1993PQ4
1994PQ1
1994PQ2
1994PQ3
1994PQ4
1995PQ1
1995PQ2
1995PQ3
1995PQ4
1996PQ1
1996PQ2
1996PQ3
1996PQ4
1997PQ1
1997PQ2
1997PQ3
1997PQ4
1998PQ1
1998PQ2
1998PQ3
1998PQ4
1999PQ1
1999PQ2
1999PQ3
1999PQ4
2000GQ1
2000GQ2
2000GQ3
2000GQ4
2001GQ1
2001GQ2
2001GQ3
2001GQ4
2002GQ1
2002GQ2
2002GQ3
2002GQ4
2003GQ1
2003GQ2
2003GQ3
2003GQ4
2004GQ1
2004GQ2
2004GQ3
2004GQ4
2005GQ1
Revenue
$110.407
$110.407
$110.407
$147.209
$111.876
$111.876
$111.758
$145.668
$114.164
$114.164
$124.820
$169.241
$126.036
$126.037
$133.095
$178.165
$132.974
$132.974
$132.974
$177.299
$136.798
$136.798
$141.599
$195.399
$142.213
$152.114
$152.114
$216.819
$168.877
$171.373
$173.952
$170.006
$173.039
$171.438
$177.725
$176.151
$162.902
$183.642
$198.133
$205.965
$190.012
$200.066
$215.027
$182.975
$186.069
$196.933
$209.122
$187.753
$197.325
Price Index
$32.771170
$32.771170
$32.771170
$32.771170
$32.771170
$32.771170
$32.771170
$32.771170
$32.771170
$34.677394
$35.338737
$35.338737
$35.338737
$35.338737
$35.338737
$35.338737
$35.338737
$35.338737
$35.338737
$37.356781
$37.660148
$37.660148
$37.660148
$37.660148
$37.660148
$40.402312
$42.422854
$42.422854
$42.422854
$42.422854
$42.422854
$42.422854
$42.422854
$46.008435
$46.264548
$46.264548
$46.264548
$46.264548
$46.315353
$50.887873
$50.887873
$50.887873
$50.887873
$50.887873
$50.887873
$50.887873
$50.887873
$50.887873
$50.887873
Volume Index
3.369
3.369
3.369
4.492
3.414
3.414
3.410
4.445
3.484
3.292
3.532
4.789
3.567
3.567
3.766
5.042
3.763
3.763
3.763
4.746
3.632
3.632
3.760
5.188
3.776
3.765
3.586
5.111
3.981
4.040
4.100
4.007
4.079
3.726
3.841
3.807
3.521
3.969
4.278
4.047
3.734
3.932
4.225
3.596
3.656
3.870
4.109
3.690
3.878
Change from
Previous Year
1.33%
1.33%
1.22%
-1.05%
2.05%
-3.56%
3.57%
7.74%
2.38%
8.33%
6.63%
5.27%
5.50%
5.50%
-0.09%
-5.86%
-3.47%
-3.47%
-0.08%
9.32%
3.96%
3.65%
-4.63%
-1.50%
5.42%
7.29%
14.36%
-21.59%
2.46%
-7.76%
-6.31%
-4.99%
-13.68%
6.53%
11.36%
6.30%
6.04%
-0.95%
-1.23%
-11.16%
-2.07%
-1.57%
-2.75%
2.61%
6.05%
USPS-T-7
243
b. Factors Affecting Post Office Boxes
1
2
The demand for Post Office Boxes was modeled using traditional demand theory.
3
The number of Post Office Box rentals is affected by the overall economy and by the
4
price charged by the Postal Service.
5
6
7
8
9
10
Specifically, Post Office Box rentals were found to be principally affected by the
following variables:
Total Private Employment
Price of Post Office Boxes
The effect of these variables on Post Office Box volume over the past ten years is
11
shown in Table 72 on the next page. Table 72 also shows the projected impacts of
12
these variables through GFY 2007.
13
The Test Year before-rates volume forecast for Post Office Boxes is 16.100 million
14
units, a 5.1 percent increase from GFY 2004. The Postal Service’s proposed rates in
15
this case are predicted to reduce the Test Year volume of Post Office Boxes by 3.3
16
percent, for a Test Year after-rates volume forecast for Post Office Boxes of 15.573
17
million.
USPS-T-7
244
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Population
1.20%
1.16%
1.22%
1.16%
1.20%
1.37%
1.21%
1.30%
1.32%
1.21%
1994 - 2004
Total
Avg per Year
13.08%
1.24%
7.24%
0.70%
-22.75%
-2.55%
12.10%
1.15%
-0.52%
-0.05%
-0.97%
-0.10%
3.44%
0.34%
2001 - 2004
Total
Avg per Year
3.88%
1.28%
-7.98%
-2.73%
0.85%
0.28%
3.24%
1.07%
-0.32%
-0.11%
-0.04%
-0.01%
-0.83%
-0.28%
Forecas t
1.24%
1.12%
1.12%
-0.71%
0.94%
0.72%
0.00%
0.00%
0.00%
1.21%
1.01%
1.18%
-0.41%
0.06%
0.17%
1.51%
-0.99%
0.19%
2.85%
2.14%
3.43%
3.52%
1.16%
0.94%
0.31%
0.00%
0.00%
3.44%
1.13%
-0.18%
-0.06%
0.69%
0.23%
8.65%
2.81%
After-Rates Volum e Forecas t
2005
1.24%
2006
1.12%
2007
1.12%
-0.71%
0.94%
0.72%
0.00%
-3.27%
0.00%
1.21%
1.01%
1.18%
-0.41%
0.06%
0.17%
1.51%
-0.99%
0.19%
2.85%
-1.20%
3.43%
0.94%
0.31%
-3.27%
-1.10%
3.44%
1.13%
-0.18%
-0.06%
0.69%
0.23%
5.10%
1.67%
Before-Rates Volum e
2005
2006
2007
2004 - 2007
Total
Avg per Year
1
Table 72
Estimated Impact of Factors Affecting Post Office Boxes, 1994 – 2007
Other Factors
Total Change
Em ploym ent
Price
Inflation Econom etric
Other
in Volum e
2.40%
-2.84%
1.38%
0.11%
0.56%
2.76%
3.22%
-1.24%
1.19%
1.11%
0.05%
5.57%
1.56%
-1.27%
1.25%
0.32%
-2.43%
0.58%
2.18%
-4.72%
0.65%
0.81%
1.16%
1.10%
2.52%
-4.16%
0.79%
-1.29%
1.26%
0.18%
1.89%
-2.27%
1.63%
-0.86%
-3.14%
-1.49%
1.72%
-9.40%
1.39%
-0.38%
1.71%
-4.18%
-0.91%
-2.39%
0.86%
1.47%
2.06%
2.34%
-4.60%
3.33%
1.16%
-0.99%
-2.12%
-2.08%
-2.65%
0.00%
1.18%
-0.79%
0.06%
-1.04%
2004 - 2007
Total
Avg per Year
3.52%
1.16%
USPS-T-7
245
1
All of the demand equations presented here model mail volume per adult per Postal
2
delivery day as a function of various explanatory variables. Hence, total mail volume is
3
projected to grow proportionally to adult population. This is reflected in the first column
4
of Table 72 above.
5
Post Office Boxes have an elasticity with respect to employment of 1.417 (t−statistic
6
of 4.160), so that a 10 percent increase in employment would be expected to lead to a
7
14.17 percent increase in the volume of Post Office Boxes being rented.
8
9
The own-price elasticity of Post Office Boxes was calculated to be equal to -0.608
(t−statistic of -2.795). The Postal price impacts shown in Table 72 above are the result
10
of changes in nominal prices. Prices enter the demand equations developed here in
11
real terms, however. The impact of inflation reported in Table 72 measures the impact
12
that a change in real Postal prices, in the absence of nominal rate changes, has on the
13
volume of Post Office Box mail.
14
Other econometric variables include seasonal variables and dummy variables
15
associated with three recent rate changes, which involved changes to the structure of
16
Post Office Box rents, the impacts of which are inadequately modeled by a simple price
17
index. The impact of these three dummy variables is included in the estimated impact
18
of Post Office Box prices shown in Table 72.
19
20
A more detailed look at the econometric demand equation for Post Office Boxes
follows.
c. Econometric Demand Equation
21
22
The demand equation for Post Office Boxes in this case models the estimated
23
number of Post Office Boxes per adult per delivery day as a function of the following
24
explanatory variables:
25
·
Seasonal Variables
USPS-T-7
246
1
2
3
4
5
6
7
8
9
10
·
Total Private Employment (lagged four quarters)
·
Dummy variables equal to one starting with the implementation of MC96-3
(1997Q4), R2000-1 (2001Q2), and R2001-1 (the last day of 2002Q3)
·
Current price of Post Office Boxes
Details of the econometric demand equation are shown in Table 73 below. A
detailed description of the econometric methodologies used to obtain these results can
be found in Section III below.
USPS-T-7
247
1
2
3
TABLE 73
ECONOMETRIC DEMAND EQUATION FOR POST OFFICE BOXES
Coefficient T-Statistic
Own-Price Elasticity
Long-Run (current only)
-0.608
-2.795
Employment (lagged four quarters)
1.417
4.160
Dummies for Specific Rate Cases
MC96-3
-0.026
-1.160
R2000-1
-0.084
-3.745
R2001-1
0.097
3.195
Seasonal Coefficients
September
-1.876
-2.468
October
-1.283
-1.524
November 1 – December 10
-1.521
-2.072
December 11 – 31
-0.996
-1.026
January – March
-1.652
-2.349
April – May
-1.474
-2.071
June
-4.390
-2.073
Quarter 1 (October – December)
-0.209
-1.444
Quarter 2 (January – March)
0.171
0.853
Quarter 3 (April – June)
0.979
2.016
Quarter 4 (July – September)
-0.942
-1.901
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
0.978775
Quarter 2 (January – March)
1.029910
Quarter 3 (April – June)
1.031181
Quarter 4 (July – September)
0.960359
REGRESSION DIAGNOSTICS
Sample Period
1993Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
33
Mean-Squared Error
0.000911
Adjusted R-Squared
0.766
4
USPS-T-7
248
1
2
10. Stamped Cards
a. Estimating Stamped Cards Volume
3
Like Post Office Boxes, the volume of Stamped Cards is not reported within the
4
Postal Service’s RPW system. Instead, the RPW system only reports the revenue
5
generated by Stamped Cards fees. Therefore, as I did with Post Office Boxes above, I
6
estimated Stamped Cards volumes by dividing RPW revenue by fees. The calculation
7
is somewhat simpler here than with Post Office Boxes, however, because all Stamped
8
Cards have the same fee, two cents. Hence, Stamped Cards volume is simply equal to
9
Stamped Card fee revenue divided by $0.02.
10
11
Stamped Cards volumes calculated in this way are presented in Table 74 below.
Revenues are expressed in millions of dollars in Table 74.
Table 74
Estimated Volume of Stamped Cards
12
Quarter
2000GQ1
2000GQ2
2000GQ3
2000GQ4
2001GQ1
2001GQ2
2001GQ3
2001GQ4
2002GQ1
2002GQ2
2002GQ3
2002GQ4
2003GQ1
2003GQ2
2003GQ3
2003GQ4
2004GQ1
2004GQ2
2004GQ3
2004GQ4
2005GQ1
Revenue
$0.674
$0.602
$0.422
$0.262
$0.361
$0.890
$1.078
$0.966
$1.481
$0.807
$0.965
$1.437
$1.092
$0.604
$0.876
$0.703
$0.554
$0.376
$0.740
$0.266
$0.417
Price Index
$0.010
$0.010
$0.010
$0.010
$0.010
$0.019
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
$0.020
Volume
67.414
60.240
42.249
26.217
36.052
46.057
53.884
48.301
74.031
40.333
48.241
71.826
54.580
30.221
43.785
35.137
27.714
18.814
36.975
13.304
20.836
Change from
Previous Year
-46.52%
-23.54%
27.54%
84.24%
105.35%
-12.43%
-10.47%
48.70%
-26.27%
-25.07%
-9.24%
-51.08%
-49.22%
-37.74%
-15.55%
-62.14%
-24.82%
USPS-T-7
249
1
The last column of Table 74 shows that Stamped Cards volumes are highly erratic,
2
with year-to-year changes in volume that range from -62 percent to +105 percent.
3
Because of this, it is very difficult to make very much econometric sense of Stamped
4
Cards volume.
5
Compounding the problem of the variability of the Stamped Cards volume data is the
6
fact that the Stamped Cards fee has only changed once in its history (twice if you count
7
the initial change from zero to one cent in 1999). Because of the combination of
8
excessive variation in Stamped Cards volume and relative lack of variation in the price
9
of Stamped Cards, it was not possible to estimate an own-price elasticity associated
10
11
12
with Stamped Cards.
b. Factors Affecting Stamped Cards Volume
Stamped Cards are a subset of First-Class cards and, more specifically, a subset of
13
First-Class single-piece cards. Hence, if the volume of First-Class single-piece cards
14
declines, one would expect the volume of Stamped Cards to similarly decline.
15
Therefore, the volume of First-Class single-piece cards is the primary explanatory
16
variable included in the econometric demand equation for Stamped Cards.
17
Ultimately, Stamped Cards volume was simply modeled as a function of First-Class
18
single-piece cards volume and several dummy variables which were included to help to
19
control for some of the more severe volume swings evident in Table 74.
20
The effect of these variables on Stamped Cards volume over the past three years is
21
shown in Table 75. Table 75 also shows the projected impacts of these variables
22
through GFY 2007. A more detailed look at the econometric demand equation for
23
Stamped Cards follows.
24
25
The Test Year before-rates volume forecast for Stamped Cards is 90.352 million
pieces, a 6.7 percent decline from GFY 2004. The Postal Service’s proposed rates in
USPS-T-7
250
1
this case are predicted to reduce the Test Year volume of Stamped Cards by 1.0
2
percent, for a Test Year after-rates volume forecast for Stamped Cards of 89.429
3
million.
USPS-T-7
251
2002
2003
2004
2001 - 2004
Total
Avg per Year
3.73%
1.23%
-4.80%
-1.63%
-57.80%
-24.99%
26.04%
8.02%
-47.47%
-19.31%
Before-Rates Volum e Forecas t
2005
1.19%
2006
1.17%
2007
1.10%
0.02%
-2.28%
-2.33%
-17.94%
2.35%
1.18%
5.42%
5.34%
0.00%
-12.44%
6.59%
-0.09%
3.51%
1.16%
-4.54%
-1.54%
-15.02%
-5.28%
11.05%
3.56%
-6.75%
-2.30%
After-Rates Volum e Forecas t
2005
1.19%
2006
1.17%
2007
1.10%
0.02%
-3.28%
-2.62%
-17.94%
2.35%
1.18%
5.42%
5.34%
0.00%
-12.44%
5.50%
-0.38%
-5.79%
-1.97%
-15.02%
-5.28%
11.05%
3.56%
-7.97%
-2.73%
2004 - 2007
Total
Avg per Year
1
Table 75
Estimated Im pact of Factors Affecting Stamped Cards Volume, 2001 – 2007
Firs t-Clas s
Other Factors
Total Change
Cards Volum e
Econom etric
Other
in Volum e
Population
1.48%
-1.27%
30.40%
-2.64%
27.20%
1.19%
-3.88%
-31.59%
4.97%
-30.16%
1.01%
0.32%
-52.69%
23.33%
-40.87%
2004 - 2007
Total
Avg per Year
3.51%
1.16%
USPS-T-7
252
c. Econometric Demand Equation
1
2
The demand equation for Stamped Cards in this case models estimated Stamped
3
Cards volume per adult per delivery day as a function of the following explanatory
4
variables:
5
6
7
8
9
10
11
12
13
14
15
·
Seasonal Variables
·
First-Class single-piece Cards Volume (per adult per delivery day)
·
Dummy variable equal to one starting in 2000Q4
·
Dummy variables equal to one in 2001Q4, 2002Q1, 2002Q4, and 2003Q4,
respectively, zero elsewhere
·
Dummy variable equal to one starting in 2004Q1
Details of the econometric demand equation are shown in Table 76 below. A
16
detailed description of the econometric methodologies used to obtain these results can
17
be found in Section III below.
USPS-T-7
253
1
2
3
TABLE 76
ECONOMETRIC DEMAND EQUATION FOR STAMPED CARDS
Coefficient T-Statistic
First-Class single-piece Cards Volume
0.782
0.473
Dummy Variables
2000Q4 onward
-0.224
-1.284
2001Q4
0.725
2.259
2002Q1
0.497
1.833
2002Q4
1.066
3.505
2003Q4
0.375
1.148
2004Q1 onward
-0.544
-2.741
Seasonal Coefficients
Quarter 1 (October – December)
0.602
2.926
Quarter 2 (January – March)
0.470
2.303
Quarter 3 (April – June)
0.670
3.145
Seasonal Multipliers (GFY 2005)
Quarter 1 (October – December)
1.146155
Quarter 2 (January – March)
1.003894
Quarter 3 (April – June)
1.226180
Quarter 4 (July – September)
0.627669
REGRESSION DIAGNOSTICS
Sample Period
2000Q1 – 2005Q1
Autocorrelation Coefficients
None
Degrees of Freedom
10
Mean-Squared Error
0.056589
Adjusted R-Squared
0.737
USPS-T-7
254
1
III. Econometric Demand Equation Methodology
2
A. Functional Form of the Equation
3
4
5
6
7
The demand equations modeled here take the form of Equation 1, presented earlier:
Vt = a·x1te1·x2te2·…·xnten·εt
(Equation 1)
This demand function is used because it has been found to model mail volume quite
8
well historically, and because it possesses two desirable properties. First, by taking
9
logarithmic transformations of both sides of Equation 1, the natural logarithm of Vt can
10
11
12
13
14
be expressed as a linear function of the natural logarithms of the Xi variables as follows:
ln(Vt) = ln(a) + e1•ln(x1t) + e2•ln(x2t) + e3•ln(x3t) +...+ en•ln(xnt) + ln(εt)
(Equation 1L)
Equation 1L satisfies traditional least squares assumptions and is amenable to
15
solution by Ordinary Least Squares. To acknowledge this property, this demand
16
function is sometimes referred to as a log-log demand function, to reflect the fact that
17
the natural logarithm of volume is a linear function of the natural logarithm of the
18
explanatory variables.
19
The second desirable property of Equation 1 is that the ei parameters in Equation 1L
20
are exactly equal to the elasticities with respect to the various explanatory variables.
21
Hence, the estimated elasticities do not vary over time, nor do they vary with changes to
22
either the volume or any of the explanatory variables. Because of these properties, this
23
demand function is sometimes also referred to as a constant-elasticity demand
24
specification.
25
B. Data Used in Modeling Demand Equations
26
Quarterly mail volumes for the various mail categories are used in each regression
27
as the dependent variable in the demand equations presented here. Data are reported
28
by Postal quarter through 1999. Data from 2000 to the present are reported by
USPS-T-7
255
1
Gregorian quarter. These quarterly volume figures are taken from the Postal Service’s
2
RPW (Revenue, Pieces, and Weight) system.
3
4
5
Quarterly volumes are divided by the number of delivery days in the quarter to obtain
volume per delivery day.
One factor affecting mail volume historically is population. As the population of the
6
United States grows, mail volume would be expected to grow in proportion. It is
7
extremely difficult to estimate the impact of population growth on mail volume growth
8
econometrically, however, due to the relatively smooth changes to population
9
historically. An assumption that a one percent change in the adult population of the
10
United States would lead to a comparable one percent change in mail volume for all
11
categories of mail seemed to provide a reasonable way around this unfortunate
12
shortcoming. For this reason, mail volumes are further divided by the number of people
13
22 years of age and older prior to being used in the demand equations.
14
The resulting series of quarterly volume per delivery day per adult is then used as
15
the dependent variable in the demand equations underlying the Postal Service’s
16
forecasting system.
17
For consistency, quantity-based macroeconomic data (retail sales, employment,
18
number of broadband subscribers, etc.) are also divided by adult population prior to
19
their inclusion in the demand equations presented here. That is, mail volume per adult
20
is modeled as a function of retail sales per adult, private investment per adult, Internet
21
Experience per adult, etc.
22
The natural logarithm of mail volume per adult per delivery day is modeled as a
23
function of a set of explanatory variables of the form of Equation 1L. In general, the
24
explanatory variables are entered into the demand equation in logarithmic form.
USPS-T-7
256
1
An exception, however, is made for those variables which take on a value equal to
2
zero over some portion of their relevant history. The natural logarithm of zero does not
3
exist. Consequently, variables which take on a value of zero at some point in the
4
regression period must be entered into the demand equations in some other form. In
5
the case of dummy variables and seasonal variables, these variables are simply
6
entered in their natural state.
7
8
9
10
11
12
13
In some other cases, however, these variables are adjusted by a Box-Cox
transformation, so that
Ln(Volume) = a + b•(Variable)λ + ...
In these cases, the values of b are technically not elasticities.
An example of a variable for which a Box-Cox transformation is made would be the
14
Internet Experience variable used to measure electronic diversion in the First-Class
15
single-piece letters and First-Class cards equations. In these equations, a value of λ
16
equal to one would be equivalent to entering Internet Experience directly in the demand
17
equation, and would mean that a given increase in the level of Internet Experience
18
would lead to the same percentage decrease in mail volume.
19
As the value of λ approaches zero, this equation approaches the equivalent of
20
entering the natural logarithm of Internet Experience in the demand equation, and would
21
mean that a given percentage increase in the level of Internet Experience would lead to
22
the same percentage decrease in mail volume. The values of λ used here are
23
estimated econometrically using nonlinear least squares in a preliminary step prior to
24
the full estimation of the other elasticity estimates. The specific applications of the Box-
25
Cox transformation are described in the relevant portions of Section II above.
USPS-T-7
257
1
C. Basic Ordinary Least Squares Model
2
Equation 1L can be re-written in matrix form as follows:
3
4
5
6
where y is equal to ln(Vt), expressed as a vector, X is a matrix with columns equal to
7
explanatory variables, ln(X1), ln(X2), ln(X3), etc., expressed as vectors, β is a vector of
8
e1, e2, e3, etc., and ε is equal to εt, expressed as a vector.
9
y = Xβ + ε
(Equation III.1)
If E(εt) = 0, and var(εt) is equal to σ2 for all t, so that var(ε) = σ2IT (where IT is a T-by-
10
T identity matrix), then the best linear unbiased estimate of the coefficient vector, β, is
11
equal to
12
13
b = (X’X)-1X’y
(Equation III.2)
This is the Ordinary Least Squares (OLS) estimate and is among the oldest and
14
most traditional results in all of econometrics. If the error term is not identically
15
distributed (i.e., var(εt) is not equal to σ2 for all t), or if the error term is not uncorrelated
16
through time (i.e., cov(εt, εt-j)≠0 for some j≠0), then the variance-covariance matrix of ε
17
can be expressed as, var(ε) = σ2Σ, and the restriction on the variance of εt can be eased
18
by introducing Σ into equation III.2 as follows:
19
20
21
22
Equation III.3 is called the Generalized Least Squares (GLS) estimate of β.
23
D. Adjustments to the Basic Ordinary Least Squares Model
24
25
b = (X’Σ-1X)-1X’Σ-1y
(Equation III.3)
1. Introduction of Outside Restrictions into OLS Estimation
To introduce restrictions into the OLS estimator, define a vector of restrictions, d,
26
and a restriction matrix, C, such that C•β = d. If the restrictions are known with
27
certainty, as for example, the restrictions imposed upon the seasonal variables that
USPS-T-7
258
1
concurrent seasons with comparable coefficients are constrained to have equal
2
coefficients, then the OLS estimator is modified as follows to yield a Restricted Least
3
Squares (RLS) estimate of the regression coefficients:
4
5
6
7
8
To introduce restrictions which are not known with certainty (i.e., stochastic
9
restrictions), define a restriction matrix, R and a vector of restrictions, r, such that
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
(OLS Estimator)
(RLS Estimator)
b = (X’X)-1X’y
b^ = b + (X’X)-1C’[C (X’X)-1C’]-1•(d - Cb)
(Equation III.4)
r = Rβ + v
where v is a random variable, such that E(v) = 0 and var(v) = σ2Ω.
In all cases where stochastic restrictions are introduced here, the matrix Ω is a
diagonal matrix with the variances associated with r along the diagonal.
The OLS estimator is modified as follows to yield a Least Squares estimate with
stochastic restrictions:
(Stochastic Restrictions Estimator)
b* =
(X’X + R’Ω-1R)-1(X’y + R’Ω-1r)
(Equation III.5)
Finally, exact and stochastic restrictions can be combined within a single estimator,
which satisfies the following formula:
(OLS Estimator incorporating outside information)
b* = (X’X + R’Ω-1R)-1(X’y + R’Ω-1r)
b** = b* + (X’X + R’Ω-1R)-1C’[C (X’X + R’Ω-1R)-1C’]-1•(d-Cb*)
(Equation III.6)
USPS-T-7
259
1
2
3
4
5
6
If E(Rβ) = r, then the most efficient, unbiased GLS estimator incorporating outside
information is similarly modified from Equation III.5 as follows:
b* = (X’Σ-1X + R’Ω-1R)-1(X’ Σ-1y + R’Ω-1r)
b** = b* + (X’ Σ-1X + R’Ω-1R)-1C’[C (X’ Σ-1X + R’Ω-1R)-1C’]-1•(d-Cb*)
(Equation III.7)
For a full treatment of the introduction of outside restrictions into the OLS model,
7
see, for example, The Theory and Practice of Econometrics, 2nd ed., by Judge, et al.,
8
pp. 51 - 62.
9
10
11
12
Equation III.7 forms the basis for estimating the demand equations developed in my
testimony.
2. Multicollinearity
In order for the OLS estimator, b, to be defined, the value of (X’X)-1 must be defined.
13
This requires that the matrix (X’X) must be of rank k if (X’X) is a k-by-k matrix. This will
14
be strictly true as long as there is no independent variable in X which can be expressed
15
as a linear combination of the other variables that make up X. So long as this is the
16
case, perfect multicollinearity will not exist, and Equation III.7 above will be uniquely
17
solvable.
18
As a practical matter, if there are variables within X which are near-perfect linear
19
combinations of one another, however, some degree of multicollinearity will exist. In
20
such a case, the OLS estimators will be unbiased, but may have extremely large
21
variances about the estimates.
22
23
24
25
26
27
Suppose, for example, that the X-matrix of explanatory variables in equation III.1
were to be divided into two separate matrices, X1 and X2, so that
y = X1β1 + X2β2 + ε
Suppose further that the explanatory variables that make up X1 (e.g., x1, x2, x3) are
highly correlated, so that, for example, x1 ≈ a1•x2 + a2•x3, for some constants, a1 and a2.
USPS-T-7
260
1
The aggregate impact of these variables on the dependent variable (X1β1) will be
2
accurately estimated. The estimated standard errors associated with the coefficients on
3
x1, x2, and x3 will be quite large, however, so that the values of b1, b2, and b3, associated
4
with x1, x2, and x3, respectively, will be poorly estimated.
5
If one’s goal is simply to fit y as well as possible (i.e., to minimize ε), then Ordinary
6
Least Squares should be sufficient. If, however, one’s goal is to obtain the best
7
possible estimate for each individual coefficient, βi, it may be necessary to develop
8
independent estimates of some of the elasticities, in cases where high multicollinearity
9
is known to exist.
10
The need for additional information in such cases is expounded on quite clearly in
11
The Theory and Practice of Econometrics, 2nd edition, by George G. Judge, et al.
12
(1985):
13
14
15
16
17
18
19
“Once detected, the best and obvious solution to [this] problem is to ... incorporate
more information. This additional information may be reflected in the form of new data,
a priori restrictions based on theoretical relations, prior statistical information in the form
of previous statistical estimates of some of the coefficients and/or subjective
information.” (p. 897)
20
econometric work. In my work, multicollinearity is particularly acute with regard to a
21
high degree of correlation between current and lagged prices of Postal products and a
22
high degree of correlation between the prices of competing Postal products. The
23
techniques by which the demand equation estimation procedure is refined to account for
24
these types of multicollinearity are described below.
25
a. Shiller Smoothness Priors
Multicollinearity will be a problem to at least some degree in any empirical
26
Experience suggests that there may be a lagged reaction by mailers to changes in
27
prices, so that mail volumes are affected not only by the current price of mail but also by
USPS-T-7
261
1
lagged prices. Because Postal prices change relatively infrequently, however, the
2
current Postal price is highly correlated with lagged Postal prices. This represents a
3
classic case of the multicollinearity problem outlined above. The aggregate effect of
4
price on mail volume can be very accurately modeled, while the coefficients on the
5
individual lags of price may be highly erratic and unstable.
6
Because the lags of price play an important role in forecasting mail volumes,
7
especially immediately after proposed future price changes, however, it is important not
8
only that the long-run (i.e., cumulative) impact of price on mail volume be accurately
9
modeled, but also that the impacts of the individual lags be accurately modeled.
10
Dr. Robert Shiller proposed a solution to this problem in a 1973 article in
11
Econometrica (Robert J. Shiller, "A Distributed Lag Estimator Derived from Smoothness
12
Priors," Econometrica, July 1973, pp. 775-788). Dr. Shiller’s technique allows a
13
polynomial equation to be used to adjust a set of coefficients so that the coefficients will
14
follow a reasonable pattern. Following Dr. Shiller’s technique, a quadratic pattern is
15
stochastically imposed on the price coefficients.
16
Dr. Shiller’s proposed technique represents a special case of a stochastic restriction.
17
In particular, the GLS estimator is modified as follows to generate Shiller distributed
18
lags:
19
20
21
22
A unique matrix, Si, is developed for each price distribution for which Shiller
23
restrictions are applied. P refers to the number of such distributions. If there are N
24
explanatory variables in the equation and variables j through j+4 are the current and first
25
through fourth lag of price i, the Si matrix will assume the following form:
bS = (X’X + Σi=1P ki2•Si’Si)-1(X’y + Σi=1P ki2•Si’Si)
(Equation III.8)
USPS-T-7
262
1
2
3
4
5
6
7
Si =
x1
0
0
0
x2
0
0
0
...
...
...
...
xj-1
0
0
0
xj
1
0
0
xj+1
-2
1
0
xj+2
1
-2
1
xj+3
0
1
-2
xj+4
0
0
1
xj+5
0
0
0
...
...
...
...
xN
0
0
0
The variable ki2 is equal to the variance of the full model (σ2) divided by the variance
8
of the smoothness restriction (ρi2). As ρi2 approaches zero, ki2 will approach infinity, and
9
bS will approach a strict quadratic distributed lag (also called an Almon lag). As ρi2
10
approaches infinity, ki2 will approach zero, and bS will approach the GLS estimator, b. A
11
unique value of ki2 is estimated for each price to which the Shiller restriction is being
12
applied.
13
The values of ki2 are chosen prior to estimation. The goal of the estimation
14
procedure is to minimize the value of ki2, subject to a prior expectation about the general
15
shape of the price distribution. The values of ki2 are minimized through a search
16
technique that evaluates the price distribution for each value of ki2. An acceptable
17
pattern for price coefficients is defined as one for which all price coefficients have the
18
same sign.
19
The smallest value of ki2 for each price distribution which yields price coefficients
20
which are all of the same sign is chosen and used in making the final coefficient
21
estimates used to make volume forecasts.
22
The quadratic Shiller restriction, at the limit, restricts each price lag coefficient to be
23
equal to the average of the coefficients of the lags before and after. Given this
24
restriction, it is technically possible for a price distribution to yield price coefficients
25
which are not all the same sign for any value of ki2. If, however, one of the end-point
26
price coefficients (i.e., either the current price or the price lagged four quarters) is
27
constrained to be equal to zero, then there will definitely exist some value of ki2 for
USPS-T-7
263
1
which all of the price coefficients will be of the same sign. In cases where such a
2
constraint is necessary, the preferred solution is to constrain the price lagged four
3
quarters to be equal to zero. In a few cases, however, constraining the fourth lag is
4
somewhat problematic. In these cases, the current price is constrained to have a
5
coefficient equal to zero, while the coefficient on the price lagged four quarters is left
6
unconstrained.
7
If, given the optimal value of ki2, the coefficient on the fourth price lag is negligible,
8
then the coefficient on the fourth lag of price is constrained to be equal to zero, and the
9
value of ki2 is re-optimized. If, given this new optimal value of ki2, the coefficient on the
10
third price lag is negligible, then the coefficient on the third lag of price is constrained to
11
be equal to zero, and the value of ki2 is re-optimized. If, given this new optimal value of
12
ki2, the coefficient on the second price lag is negligible, then the coefficient on the
13
second lag of price is constrained to be equal to zero, and the value of ki2 is re-
14
optimized. Finally, if, given this new optimal value of ki2, the coefficient on the first price
15
lag is negligible, then the coefficient on the first lag of price is constrained to be equal to
16
zero. In this last case, only the current price appears in the demand equation, so that
17
no Shiller restriction is necessary.
18
19
b. Special Note on Price Lags
As mentioned earlier, for certain mail volumes the full impact of a change in Postal
20
prices will not be felt until one year after the date of a rate change. In the tables in
21
Section II showing the estimated impact of factors affecting the demand for mail
22
volumes, volume changes from 2003 to 2004 were affected by nominal Postal prices for
23
some categories of mail despite the fact that Postal prices have remained unchanged
24
since June, 2002 (the last day of 2002Q3).
USPS-T-7
264
1
For mail categories which are affected by prices with a four-quarter lag, the effect of
2
the R2001-1 rate change on the last day of 2002Q3 was not felt in full until four quarters
3
later, i.e., the last day of 2003Q3. Hence, for the year 2003, the percentage change in
4
volume attributable to the R2001-1 rate change was less than the percentage change in
5
volume implied by the long-run price elasticity of the relevant mail category. By 2004,
6
however, the percentage change in volume attributable to the R2001-1 rate change
7
would be equal to the percentage change in volume implied by the long-run own-price
8
elasticity.
9
On a quarter-by-quarter basis, then, changes in volumes could have been affected
10
by R2001-1 as late as 2003Q4. Thus, on a year-by-year basis, the effect in 2003
11
(which includes only a partial impact of R2001-1 on the first three quarters of volume)
12
may have been less than the effect in 2004 (which would have the full impact of
13
R2001−1 for all four quarters).
14
15
c. Slutsky-Schultz Symmetry Condition
In addition to being highly correlated with their own lags, Postal prices are also
16
highly correlated with one another. Postal prices tend to be increased at the same time
17
as part of omnibus rate cases. Between rate cases, all real Postal prices fall together at
18
the rate of inflation.
19
In order to alleviate some of the multicollinearity across Postal prices, the
20
econometric estimation of cross-price relationships modeled here is helped by a
21
relationship known as the Slutsky-Schultz symmetry condition.
22
The Slutsky-Schultz cross-price relationship is premised on an assumption that, for
23
two goods i and j, the change in the volume of good i attributable to a change in the
24
price of good j is equal to the change in the volume of good j attributable to a change in
25
the price of good i, or, mathematically,
USPS-T-7
265
1
2
3
4
5
6
7
8
∂Vi / ∂pj = ∂Vj / ∂pi
The elasticity of Vi with respect to pj is equal to
eij = [∂Vi / ∂pj]•(pj / Vi),
11
∂Vi / ∂pj = eij•(Vi / pj),
14
15
(Equation III.10b)
Combining equations (III.9) and (III.10b), then, yields the Slutsky-Schultz symmetry
condition:
eij / eji = Vj•pj / Vi•pi
12
13
(Equation III.10)
so that, rearranging terms,
9
10
(Equation III.9)
(Equation III.11)
In words, the Slutsky-Schultz symmetry condition states that the ratio of cross-price
elasticities is equivalent to the ratio of expenditures on goods i and j.
The Slutsky-Schultz symmetry condition can be used to gauge the reasonableness
16
of the cross-price elasticities between Postal categories estimated from the quarterly
17
time series data, and, if necessary, to adjust the cross-price elasticities to more
18
reasonable values.
19
When necessary, the Slutsky-Schultz condition is implemented by freely estimating
20
the relevant cross-price elasticity in one of the two equations of interest. The equation
21
in which the elasticity is freely estimated is chosen based upon the reasonableness of
22
the freely-estimated elasticities in the two equations of interest.
23
The Slutsky-Schultz condition is then applied to the freely-estimated cross-price
24
elasticity and used as the basis for calculating a restriction which is applied to the other
25
equation(s) in which this cross-price relationship appears. In general, the Slutsky-
26
Schultz restriction is entered as a stochastic restriction, with the variance of the
27
restriction being derived from the variance of the freely-estimated elasticity estimate.
USPS-T-7
266
1
For example, the cross-price elasticity of Media Mail with respect to the price of
2
Bound Printed Matter is freely estimated in the Media Mail demand equation. This
3
cross-price elasticity is then used to calculate a stochastic restriction which is applied to
4
the estimated cross-price elasticity with respect to Media Mail in the Bound Printed
5
Matter equation.
6
7
3. Autocorrelation
The restriction on the OLS estimator that var(εt) = σ2 requires an assumption that the
8
error term is independently distributed, so that cov(εt, εt-k) = 0 for all t, k≠0. If this is not
9
the case, the residuals are said to be autocorrelated. In this case, the Least Squares
10
estimator will be unbiased. It will not, however, be efficient. That is, the estimated
11
variance of b will be very high, and the traditional least squares test statistics may not
12
be valid.
13
Autocorrelation is tested for and corrected in the residuals using a traditional
14
econometric method called the Cochrane-Orcutt procedure (D. Cochrane and G. H.
15
Orcutt, "Application of Least Squares Regressions to Relationships Containing
16
Autocorrelated Error Terms," Journal of the American Statistical Association, vol. 44,
17
1949, pp. 32-61).
18
An OLS regression (with outside restrictions as outlined above) is initially run. The
19
residuals from this regression are then inspected to assess the presence of
20
autocorrelation.
21
Three degrees of autocorrelation are tested for: first-order autocorrelation, whereby
22
residuals are affected by residuals one quarter earlier; second-order autocorrelation,
23
whereby residuals are affected by residuals two quarters earlier; and fourth-order
24
autocorrelation, whereby residuals are affected by residuals four quarters, i.e., one year,
25
earlier.
USPS-T-7
267
1
The exact nature of the autoregressive process is identified by testing the
2
significance of the partial autocorrelation of the residuals at one, two, and four lags. In
3
general, a 95 percent confidence level is used to test for the presence of
4
autocorrelation. The following relationship is then fit to the residuals:
5
6
7
8
where ut is assumed to satisfy the OLS assumptions. The values of ρ1, ρ2, and ρ4 are
9
estimated using traditional OLS. If significant fourth-order autocorrelation is not
et = ρ1•et-1 + ρ2•et-2 + ρ4•et-4 + ut
(Equation III.12)
10
identified, ρ4 is set equal to zero, if second-order autocorrelation is not identified as
11
significant, then ρ2 = 0, and if significant first-order autocorrelation is not identified, then
12
ρ1 = 0.
13
The values of ρ1, ρ2, and ρ4 are used to adjust the variance-covariance matrix of the
14
residuals, Σ, and the β-vector is re-estimated using the Generalized Least Squares
15
equation:
16
17
18
19
β^ = (X’Σ-1X)-1X’Σ-1y
The variance-covariance matrix of the residuals, Σ, is set equal to (P’P)-1, where P is
20
a (T-i)-by-T matrix (where T is the total number of observations in the sample period
21
and i is the largest lag for which significant autocorrelation was detected) that takes on
22
the following form:
USPS-T-7
268
1
2
3
4
5
6
7
8
9
10
P0 =
1
-ρ1
-ρ2
0
-ρ4
0
0
...
0
0
1
-ρ1
-ρ2
0
-ρ4
0
0
0
1
-ρ1
-ρ2
0
-ρ4
0
0
0
1
-ρ1
-ρ2
0
0
0
0
0
1
-ρ1
-ρ2
0
0
0
0
0
1
-ρ1
0
0
0
0
0
0
1
0
0
0
0
0
0
0
...
...
...
...
...
...
...
0
0
0
0
0
0
0
0
0
...
0
-ρ4
0
-ρ2
-ρ1
1
11
where P0 is a T-by-T matrix, and P is equal to the last T-i rows of P0. In other words, if
12
i=0, then ρ1=ρ2=ρ4=0, P is simply equivalent to P0 (which, in such a case, would simply
13
be a T-dimensional identity matrix), and the GLS equation above is exactly equivalent to
14
Ordinary Least Squares. If i=1, then ρ2=ρ4=0, and the first row of P is equal to [-ρ1 1 0 0
15
... 0]. If i=2, then ρ4=0, and the first row of P is equal to [-ρ2 -ρ1 1 0 0 ... 0]. Finally, if
16
i=4, the first row of P is equal to [-ρ4 0 −ρ2 -ρ1 1 0 0 ... 0] .
17
Modifying Σ in this way and estimating β using Generalized Least Squares is
18
equivalent to using the rho-coefficients (ρ1, ρ2, and ρ4) to transform the dependent
19
variable as well as all of the independent variables as follows:
20
21
22
23
removing the first i observations of the regression period, re-defining y and X using the
24
transformed data, and re-estimating β using the OLS estimator on the transformed
25
variables.
x’t = xt - ρ1•xt-1 - ρ2•xt-2 - ρ4•xt-4
(Equation III.13)
26
The values of ρ1, ρ2, and ρ4 are optimized through a simple iteration process. First,
27
the β-vector is solved for as described above, assuming that ρ1, ρ2, and ρ4 are equal to
28
zero. Given the value of β, ρ1, ρ2, and ρ4 are then estimated using Equation 8. Given
29
these values for ρ1, ρ2, and ρ4, β is re-estimated. Given β, ρ1, ρ2, and ρ4 are then re-
30
estimated. This iteration process continues until the estimated values of ρ1, ρ2, and ρ4
USPS-T-7
269
1
do not vary between iterations. This is mathematically equivalent to estimating the β-
2
vector simultaneously with ρ1, ρ2, and ρ4.
3
E. Summary of Demand Equation to be Estimated
4
To summarize, then, demand equations are estimated of the form:
5
6
7
8
9
10
11
12
13
ln(Vt) = ln(a) + e1•ln(x1t) + e2•ln(x2t) + e3•ln(x3t) + ... + en•ln(xnt) + εt
(Equation 1L)
using a Generalized Least Squares Technique with fixed and stochastic restrictions:
b* = (X’Σ-1X + R’Ω-1R)-1(X’Σ-1y + R’Ω-1r)
b = b + (X’Σ X+R’Ω R+Σi=1Pki2•Si’Si)-1C’[C(X’Σ-1X+R’Ω-1R+Σi=1Pki2•Si’Si)-1C’]-1•(d-Cb*)
**
*
-1
-1
where Σ is adjusted for the possibility of autocorrelation, so that
14
15
16
17
where P is a (T-i)-by-T matrix (where T is the total number of observations in the
18
sample period and i is the largest lag for which significant autocorrelation was detected)
19
that takes on the following form:
20
21
22
23
24
25
26
27
28
29
30
31
32
Σ = (P’P)-1
P0 =
1
-ρ1
-ρ2
0
-ρ4
0
0
...
0
0
1
-ρ1
-ρ2
0
-ρ4
0
0
0
1
-ρ1
-ρ2
0
-ρ4
0
0
0
1
-ρ1
-ρ2
0
0
0
0
0
1
-ρ1
-ρ2
0
0
0
0
0
1
-ρ1
0
0
0
0
0
0
1
0
0
0
0
0
0
0
...
...
...
...
...
...
...
0
0
0
0
0
0
0
0
0
...
0
-ρ4
0
-ρ2
-ρ1
1
where P0 is a T-by-T matrix, and P is equal to the last T-i rows of P0.
All of these various terms are defined and described above.
USPS-T-7
270
1
F. Step-by-Step Examples
2
Two examples of the econometric processes used in this case are shown below:
3
First-Class workshared letters and Standard Regular mail.
4
1. First-Class Workshared Letters
5
a. Basic Demand Equation
6
As described in Section II.B.6 above, the demand for First-Class workshared letters
7
is modeled as a function of seasonal variables, retail sales, the number of broadband
8
subscribers lagged one year, a time trend starting in 2002Q4, dummy variables starting
9
in 1993Q1 and 1996Q4, the average discounts for Standard Regular letters (relative to
10
First-Class workshared letters) and First-Class workshared letters (relative to First-
11
Class single-piece letters), and current and four lags of the price of First-Class
12
workshared letters. The details about these variables are described in Section II above.
13
14
15
16
17
18
19
Mathematically, the demand equation for First-Class workshared letters can be
specified as follows:
Ln(VolWS / Population / Delivery Days)t =
a+b1•Ln(Retail Sales)t+b2•(Broadband)t-4+b3•T02Q4t+b4•D93Q1t+b5•D96Q4t+ b6•Ln(DSTD L)t+
b7•Ln(DWS)/(VWS/VSP)’t+b8•(Pws)t+b9•(Pws)t-1+b10•(Pws)t-2+b11•(Pws)t-3+b12•(Pws)t-4+Σi=1S(si)t+et
Putting this into matrix format, then, the y-vector contains the natural logarithm of
20
First-Class workshared letters volume per adult per delivery day, Ln(VolWS / Population /
21
Delivery Days)t, for t = 1991Q1 through 2005Q1. The X-matrix contains 21 columns of
22
data for the 21 explanatory variables: constant, retail sales, broadband, T02Q4 (trend
23
since 2002Q4), D93Q1 (dummy since 1993Q1), D96Q4 (dummy since 1996Q4), DSTD L
24
(average discount for Standard Regular letters), average First-Class workshared letters
25
discount, current and four lags of the price of First-Class workshared letters, and eight
26
seasonal variables.
27
The β-vector to be solved for contains the following elements:
USPS-T-7
271
1
2
3
βWS = [a b1 b2 b3 b4 b5 b6 b7 b8 b9 b10 b11 b12 s1 s2 … s8]
b. Seasonal Variables
As described in Section II.A.c.vi above, a total of 22 seasonal variables are included
4
in the demand equations presented in my testimony. In an effort to maximize the
5
explanatory power of the seasonal variables, the coefficients on adjoining seasons that
6
are similar in sign and magnitude are constrained to be equal. In most cases, this
7
involves entering 21 seasonal variables into the demand equation of interest, and
8
imposing fixed restrictions across some of these variables.
9
For equations estimated over sufficiently short sample periods (typically, sample
10
periods starting in 1990 or later), however, some of the adjoining seasonal variables
11
had to be constrained to avoid perfect multicollinearity. In these cases, the seasonal
12
variables outlined earlier in my testimony were combined prior to their inclusion in the
13
demand equation.
14
These two techniques – constraining the variable coefficients within the equation
15
and combining the variables prior to inclusion in the equation – are mathematically
16
equivalent. Because the First-Class workshared letters equation is estimated over a
17
sample period which begins in 1991Q1, the latter option – combining seasonal variables
18
prior to inclusion in the demand equation – was chosen.
19
The seasonal variables that are included in the First-Class workshared letters
20
equation span the following time periods: September 1 – December 10, December 11 –
21
31, January – May, June, Gregorian Quarter 1 (October – December, since FY 2000),
22
Gregorian Quarter 2 (January – March, since FY 2000), Gregorian Quarter 3 (April –
23
June, since FY 2000), and Gregorian Quarter 4 (July – September, since FY 2000).
USPS-T-7
272
c. Restriction Matrices
1
The First-Class workshared letters equation includes two restrictions: one fixed
2
3
restriction and one stochastic restriction. The fixed restriction is that the sum of the
4
coefficients on the four Gregorian quarterly dummy variables must be equal to zero.
5
This restriction is applied to all of the demand equations estimated in my testimony.
6
The rationale for this restriction was discussed in Section II.A.c.vi. above. The
7
restriction matrix, C, is a (1-by-21 matrix) of the following form:
C = [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1]
8
The vector, d, associated with this restriction is equal to [0].
9
The stochastic restriction applied to the First-Class workshared letters equation
10
11
restricts the coefficient on the Standard Regular letters discount (DSTD L) from the
12
Standard Regular equation using the Slutsky-Schultz symmetry condition. The Slutsky-
13
Schultz symmetry condition was described in Section III.D.2 above. Applying the
14
Slutsky-Schultz equation III.11 to the discount elasticity from the Standard Regular
15
equation (0.075), using GFY 2004 revenues for First-Class workshared letters and
16
Standard Regular mail yields a stochastic constraint of -0.0509035 with a variance of
17
0.002919. The stochastic restriction matrices, R, r, and Ω, therefore will take the
18
following forms:
19
R = [0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
20
r = [-0.050903]
21
Ω = [0.002919]
5
Because the price here is expressed as a discount, rather than a price, the elasticities in the Standard
Regular and First-Class workshared letters equations will have opposite signs.
USPS-T-7
273
1
2
3
d. Final Econometric Estimates
The demand equation for First-Class workshared letters contains a single price to
which a Shiller restriction is imposed, PWS. The S-matrix is equal to the following:
4
5
6
7
0 0 0 0 0 0 0 0 1 -2 1 0 0 0 0 0 0 0 0 0 0
S = 0 0 0 0 0 0 0 0 0 1 -2 1 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 1 -2 1 0 0 0 0 0 0 0 0
8
The minimum value of k2 which yielded a reasonable price distribution was chosen
9
based on a search of alternate values for k2. The chosen k2 value was 0.006250.
10
Based on the procedure outlined in Section III.D.3 above, autocorrelation
11
coefficients ρ1, ρ2, and ρ4 were estimated to be equal to 0, 0, and -0.366801,
12
respectively. The variance-covariance matrix of the residuals, Σ, was adjusted using
13
these values as described in section D.3 above.
14
15
16
17
18
19
20
21
Taken together, then, the resulting β-vector associated with First-Class workshared
letters was estimated to be equal to the following:
βWS = [-0.744 0.459 -1.261 -0.004 -0.069 -0.068 -0.097 0.108 -0.128 -0.040 -0.002 -0.043 -0.116 0.157
0.462 0.122 0.229 -0.063 0.070 -0.054 0.047]
2. Standard Regular Mail
a. Basic Demand Equation
As described in Section II.C.3.b above, the demand for Standard Regular mail is
22
modeled as a function of seasonal variables, retail sales, private domestic investment
23
lagged one quarter, a linear time trend, a dummy variable starting in 1996Q4, a dummy
24
variable for the implementation of R97-1 rates (1999Q2), a dummy variable equal to
25
one in 2002Q1 (zero elsewhere), the average discount for Standard Regular letters
26
relative to First-Class workshared letters, the percentage of Standard Regular mail for
27
which First-Class cards rates are lower, a time trend beginning in 1988Q4 and
USPS-T-7
274
1
plateauing in 1990Q4, and current and four lags of the price of Standard Regular mail.
2
The details about these variables are described in Section II above.
3
Mathematically, the demand equation for Standard Regular mail can be specified as
4
follows:
5
6
7
8
9
Ln(VolReg / Population / Delivery Days)t =
a+b1•Ln(Retail Sales)t+b2•(Investment)t-1+b3•Trendt+b4•D96Q4t++b5•(DR97)t + b6•D02Q1t+ b7•Ln(DSTD L)t+
b8•(Cards Crossover)t+b9•(Trend88-90)t+b10•(PReg)t+b11•(PReg)t-1 +b12•(PReg)t-2 +b13•(PReg)t-3 +b14•(PReg)t-4 + Σi=121(si)t+et
Putting this into matrix format, then, the y-vector contains the natural logarithm of
10
Standard Regular mail volume per adult per delivery day, Ln(VolReg / Population /
11
Delivery Days)t, for t = 1988Q1 through 2005Q1. The X-matrix contains 36 columns of
12
data for the 36 explanatory variables: constant, retail sales, investment, Trend, D96Q4,
13
DR97, D02Q1, DSTD L (average discount for Standard Regular letters), Cards Crossover,
14
Trend88-90, current and four lags of the price of Standard Regular mail, and twenty-one
15
seasonal variables (as described in Section II.A.2.c.vi above).
16
17
18
19
20
The β-vector to be solved for contains the following elements:
βReg = [a b1 b2 b3 b4 b5 b6 b7 b8 b9 b10 b11 b12 b13 b14 s1 s2 … s21]
b. Restriction Matrices
The Standard Regular equation includes thirteen restrictions, all fixed. These
restrictions fall into three categories.
21
First, there are eight restrictions associated with the seasonal coefficients, whereby
22
adjoining seasonal variables with similar estimated coefficients are constrained to have
23
equal coefficients. The rationale for doing this was discussed in Section II.A.c.vi. above.
24
The following seasonal coefficients were constrained to be equal in this way: September
25
1 – 15 is constrained to equal September 16 – 30, October is constrained to equal
26
November 1 – December 10, December 11 – 12 is constrained to equal December 13 –
27
15, the seasonal variables encompassing the time period from December 18 – 24 are
USPS-T-7
275
1
constrained to be equal, and January – February is constrained to equal March. In
2
addition, the sum of the coefficients on the four Gregorian quarterly dummy variables is
3
constrained to be equal to zero. This restriction is applied to all of the demand
4
equations estimated in my testimony, and is also discussed in Section II.A.c.vi. above.
5
Second, the coefficients on the price of Standard Regular mail lagged two, three,
6
and four quarters are constrained to be equal to zero. The rationale for these
7
restrictions is discussed in Section III.D.2.a. above.
8
Finally, the coefficients on the First-Class cards cross-price variables, (Cards
9
Crossover) and Trend88-90, are both constrained from the First-Class cards equation.
10
This is described in Section II.C.3.b. above. The restrictions used here, which are
11
calculated from the First-Class cards equation, are -0.008628 for the Cards Crossover
12
variable, and -0.002954 for Trend88-90.
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
The restriction matrix, C, is a (13-by-36 matrix) of the following form:
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
1 -1
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
1 -1
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
1 -1
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 0 0 0
0 0 0 0
0 0 0 0
1 -1 0 0
0 1 -1 0
0 0 1 -1
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0 0 0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0 0
0 0
0 0
0 0
0 0
0 0
1 -1
0 0
0 0
0 0
0 0
0 0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
The vector, d, associated with this restriction is equal to the following:
d = [0 0 0 0 0 0 0 0 0 0 0 -0.008628 -0.002954]
c. Final Econometric Estimates
The demand equation for Standard Regular mail contains a single price to which a
Shiller restriction is imposed, PReg. The S-matrix is equal to the following:
USPS-T-7
276
1
2
3
4
5
0 0 0 0 0 0 0 0 0 0 1 -2 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 -2 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 1 -2 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
The minimum value of k2 which yielded a reasonable price distribution was chosen
6
based on a search of alternate values for k2. In this case, a k2 value of zero worked, so
7
that no Shiller restriction was necessary.
8
9
Based on the procedure outlined in Section III.D.3. above, autocorrelation
coefficients ρ1, ρ2, and ρ4 were estimated to be equal to 0, 0, and -0.442743,
10
respectively. The variance-covariance matrix of the residuals, Σ, was adjusted using
11
these values as described in section D.3. above.
12
Taken together, then, the resulting β-vector associated with Standard Regular mail
13
was estimated to be equal to the following:
14
15
βReg = [-2.858 0.104 0.228 0.008 -0.053 0.072 -0.044 0.075 -0.009 -0.003 -0.176 -0.091 0 0 0 1.209 1.209 0.803 0.803 1.017 1.017 0.327 9.859 9.859 9.859 9.859 -15.511 1.373 1.373 -1.259 1.115 2.016 0.421 -0.556 -0.240 0.375]
USPS-T-7
277
1
IV. Volume Forecasting Methodology
2
A. Volume Forecasting Equation
3
As discussed earlier, the basic forecasting equation employed here, Equation 1,
4
relates volume at time t to a series of explanatory variables according to the following
5
formula:
Vt = a·x1te1·x2te2·…·xnten·εt
6
7
8
9
10
Equation 1 is assumed to hold both historically as well as into the forecast period.
Of particular interest, Equation 1 is assumed to hold over the most recent time period,
called the Base Period. That is,
VB = a·x1Be1·x2Be2·…·xnBen·εB
11
12
13
14
15
16
17
18
(Equation 1)
(Equation 1B)
Dividing Equation 1 by Equation 1B, for forecast time period t and multiplying both
sides by VB yields the following equation:
Vt = VB · [x1t/x1B]e1 · [x2t/x2B]e2 · … · [xnt/xnB]en · [εt/εn]
(Equation IV.1)
Equation IV.1 forms the basis for the volume forecasts used in this case. The
19
volume forecasting methodology used here is sometimes referred to as a base-volume
20
forecasting methodology. The logic of this name can be seen quite readily in Equation
21
IV.1. Using this forecasting methodology, volume at time t (Vt) is projected to be equal
22
to volume in the base period (VB) times a series of multipliers of the form [xit/xiB]ei which
23
reflect the extent to which the explanatory variables have changed from the base period
24
to time t.
USPS-T-7
278
1
B. Base Period Used in Forecasting
2
The base period used in forecasting is typically the most recent four Postal quarters.
3
In this case, this is 2004GQ2 through 2005GQ1. This corresponds to calendar 2004
4
(January – December).
5
Base volume (VB) in Equation IV.1 is equal to the sum of the volume in these four
6
quarters. Base values for explanatory variables, xB, on the other hand, are equal to the
7
average value from the base period.
8
C. Projection Factors
9
The [xt/xB]e terms in Equation 2 can be called Projection Factors, as they project the
10
impact of a particular factor on mail volume. Projection factors fall into three general
11
categories – rate effect multipliers, non-rate multipliers, and other multipliers.
12
13
By aggregating the multipliers in this way, the volume forecasting equation for mail
category i can be simplified as follows:
14
15
16
17
where RMti is the rate effect multiplier, NRMti is the non-rate effect multiplier, and CMti is
18
the composite multiplier. These three multipliers are described next.
19
Vti = VBi•RMt i•NRM t i•CM t i
1. Rate Effect Multiplier
20
The rate effect multiplier includes projection factors based upon prices. This
21
includes both Postal prices and, if appropriate, competitor (UPS, FedEx) prices.
22
Generally, prices are included in Equation 1 for both the current time period, t, as well
23
as lagged one through four quarters. Hence, each price within the rate effect multiplier
24
generates five projection factors. That is, the total projection factor for price j on mail
25
category i at time t is the following:
26
27
RMtij = (ptj / pBj)e0j•(pt-1j / pB-1j)e1j•(pt-2j / pB-2j)e2j•(pt-3j / pB-3j)e3j•(pt-4j / pB-4j)e4j
(Equation IV.2)
USPS-T-7
279
1
2
Prices enter the rate effect multiplier in real terms. The total rate multiplier, RMti, for
3
each category of mail is then equal to the product of RMtij for j equals all of the prices
4
included in the forecasting equation, i.e.,
5
RMti = πj RMtij
6
Before-rates nominal Postal prices are, of course, taken to remain constant
(Equation IV.3)
7
throughout the forecast period. After-rates nominal Postal prices are set equal to the
8
rates requested by the Postal Service in this case. These prices are deflated by the
9
implicit price deflator for personal consumption expenditures. The forecast for the
10
11
consumption deflator comes from Global Insight.
UPS and FedEx prices are assumed to remain constant through the forecast period.
12
In the case of UPS and FedEx average revenue, these are assumed to literally remain
13
constant in real terms for each quarter of the forecast period. In the case of published
14
UPS Ground rates, which are used in the forecast for non-destination-entry Parcel Post
15
mail, rates are assumed to increase in January of each year at the rate of inflation since
16
the previous January. This is done to reflect the fact that UPS has historically raised
17
their published rates only once each year.
18
19
2. Non-Rate Effect Multiplier
The non-rate multiplier includes projection factors for all variables which are included
20
in the econometric demand equations, except for prices, which are included in the rate
21
effect multiplier described above, and seasonal variables, which are included in the
22
composite multiplier described below.
23
24
25
For each non-rate variable, j, the non-rate multiplier for mail category i at time t is
calculated as follows:
USPS-T-7
280
1
2
3
4
NRMtij = (xtj / xBj)eij
The total non-rate multiplier, NRMti, for mail category i is then equal to the product of
NRMtij across all j’s,
NRMti = πj NRMtij
5
6
(Equation IV.4)
(Equation IV.5)
3. Forecasts of Non-Rate Variables
7
Non-rate variables can be divided into three categories: mechanical variables,
8
economic variables forecasted by Global Insight, and economic variables that are not
9
forecasted by Global Insight.
10
11
a. Mechanical Non-Rate Variables
Mechanical variables are simply dummy variables and time trends. These variables
12
are projected forward mechanically. That is, a time trend that increases by one each
13
quarter historically is assumed to continue to increase by one each quarter of the
14
forecast period. This need not be true, of course. The magnitude of some trends may
15
well change in the future. It is important to consider this possibility in developing
16
forecasts. In this case, all of the econometric trends are assumed to continue to grow at
17
their historical rate through the test year except for the trend in the Standard Regular
18
demand equation.
19
The Standard Regular time trend is projected to grow at a somewhat slower rate
20
than it has grown historically. Specifically, starting in 2006Q1, the Standard Regular
21
time trend is increased at a progressively decreasing rate. For each quarter, the rate at
22
which this trend is projected to increase declines by 0.025, so that, whereas through
23
2005Q4, the Standard Regular time trend increased by one each quarter, in 2006Q1,
24
the Standard Regular time trend increases by 0.975. In 2006Q2, the Standard Regular
25
time trend increases by 0.950, etc.
USPS-T-7
281
1
2
b. Macroeconomic Variables Forecasted by Global Insight
Many of the macroeconomic variables used here for forecasting are forecasted by
3
Global Insight. These include retail sales, employment, the implicit price deflator for
4
personal consumption expenditures, adult population, investment, and the producer
5
price index of pulp, paper, and allied products. Global Insight makes forecasts of these
6
and other macroeconomic variables monthly. Each month, Global Insight provides a
7
baseline forecast, which they consider the most likely outcome, and two alternative
8
forecasts, typically an optimistic and a pessimistic scenario. For this case, Global
9
Insight’s baseline forecasts for January, 2005 were used.
10
11
c. Non-Rate Variables Forecasted Here
Finally, several of the non-price variables used here are not forecasted by Global
12
Insight. In these cases, forecasts of these variables are made here. These variables
13
include mail-order retail sales, advertising expenditures, the producer price indices for
14
newspaper and direct-mail advertising printing, average delivery days for Priority Mail,
15
and the three Internet measures used here: consumption expenditures on Internet
16
Service Providers, the number of broadband subscribers, and total Internet advertising
17
expenditures. The forecasts for each of these variables are considered next.
18
19
i.
Mail-Order Retail Sales
Retail sales data are compiled by the Census Bureau. In addition to total retail
20
sales, retail sales data are calculated for a number of breakdowns, including one
21
entitled “Electronic Shopping and Mail-Order Houses.” This data series is identified
22
here as mail-order retail sales. Data on mail-order retail sales are available from
23
January, 1978 through November, 2004. Table IV-1 below compares total and mail-
24
order retail sales by year from 1978 through 2004 (first 11 months only).
USPS-T-7
282
1
2
Global Insight forecasts total retail sales. They do not, however, forecast mail-order
retail sales. Hence, a forecast of mail-order retail sales is developed here.
USPS-T-7
283
1
Year
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004*
* Thru November, 2004
2
3
Ta ble IV-1
Ma il-Orde r Re ta il Sa le s
(historical)
Mail-Order as
Retail Sales
Percentage of
Total
Mail-Order Total Retail Sales
848.368
8.481
1.00%
946.565
9.437
1.00%
1,009.496
10.706
1.06%
1,097.762
10.989
1.00%
1,128.585
11.228
0.99%
1,234.261
12.645
1.02%
1,357.260
14.951
1.10%
1,452.493
15.790
1.09%
1,532.190
17.079
1.11%
1,626.887
20.776
1.28%
1,742.957
23.736
1.36%
1,859.134
26.222
1.41%
1,951.873
26.457
1.36%
1,962.892
29.817
1.52%
2,055.747
35.071
1.71%
2,200.391
40.219
1.83%
2,379.971
46.976
1.97%
2,504.895
52.505
2.10%
2,648.701
60.785
2.29%
2,781.008
70.188
2.52%
2,917.924
79.418
2.72%
3,162.430
92.072
2.91%
3,372.438
109.969
3.26%
3,475.898
109.453
3.15%
3,564.420
114.439
3.21%
3,754.731
120.990
3.22%
3,690.082
122.939
3.33%
Table IV-1 shows that the share of retail sales that are classified as mail-order has
4
trended upward over time. Looking more closely, it also appears that the share of retail
5
sales that are mail-order has tended to decline during economic recessions (1982,
6
1990, 2001).
USPS-T-7
284
1
A regression equation is fitted for mail-order retail sales as a percentage of total
2
retail sales. The percentage of retail sales that are mail order is modeled as a function
3
of total private employment and a time trend. This equation is fitted using monthly data
4
from January, 1992 through November, 2004 and is summarized below. The
5
regression starts in 1992 because of a break in the data prior to that. Data prior to 1992
6
were classified under the Standard Industry Classification (SIC) system. Data from
7
1992 forward are classified under the North American Industry Classification System
8
(NAICS). The pre-1992 data are adjusted here to conform to the newer NAICS data.
9
Nevertheless, for the purpose of fitting a regression on mail-order retail sales, it was
10
decided to exclude these older, pre-1992 data.
USPS-T-7
285
Ta ble IV-2
Re gre ssion Equa tion for
Ma il-Orde r Re ta il Sa le s a s a Pe rce nta ge of Tota l Re ta il Sa le s
Constant
Total Private Employment
Time Trend
1
2
Adjusted R2
Degrees of Freedom
Coefficient
-0.022241
0.000423
0.000062
T-Statistic
-10.016
17.491
15.999
0.980
152
This equation is then used to project the share of retail sales that is expected to be
3
mail-order sales through the forecast period, given Global Insight’s forecast of total
4
employment and a straight mechanical projection of the time trend. Total mail-order
5
retail sales are then calculated by multiplying these projected shares times Global
6
Insight’s forecasts for total retail sales.
7
Forecasted values of mail-order retail sales are shown in Table IV-3 below.
USPS-T-7
286
GQ
1
2
3
ii.
Ta ble IV-3
Ma il-Orde r Re ta il Sa le s
(forecast)
Mail-Order as
Retail Sales
Percentage of
Total
Mail-Order Total Retail Sales
Actual
2004.1
2004.2
2004.3
2004.4
2005.1
320.190
328.244
333.531
338.155
346.576
10.467
10.802
11.145
11.344
11.681
3.27%
3.29%
3.34%
3.35%
3.37%
Forecast
2005.2
2005.3
2005.4
2006.1
2006.2
2006.3
2006.4
2007.1
2007.2
2007.3
2007.4
352.018
355.210
357.063
360.076
363.902
367.255
371.421
375.428
379.383
383.166
387.319
12.177
12.441
12.651
12.889
13.147
13.403
13.685
13.947
14.206
14.469
14.745
3.46%
3.50%
3.54%
3.58%
3.61%
3.65%
3.68%
3.71%
3.74%
3.78%
3.81%
Total Advertising Expenditures
Total advertising expenditures are not used directly in any of the demand equations
4
presented in my testimony. Internet advertising expenditures are included in the
5
Standard Enhanced Carrier Route equation as a share of total advertising expenditures,
6
however. Hence, historical and forecasted data on total advertising expenditures are
7
needed to express Internet advertising expenditures as a share of total advertising
8
expenditures. In addition, advertising expenditures are an explanatory variable in the
9
forecasts of newspaper and direct-mail advertising developed below.
10
Historical advertising expenditures data are reported annually by Robert Coen of
11
Universal McCann-Erickson. Mr. Coen also makes one-year-ahead forecasts. The
USPS-T-7
287
1
advertising expenditures data used here through 2005 are taken from this source.
2
These data are presented in Table IV-4 below.
Ta ble IV-4
Annua l Adve rtising Expe nditure s
3
4
Year
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Nominal
(billions of $)
$130.0
$128.4
$133.8
$141.0
$153.0
$165.1
$178.1
$191.3
$206.7
$222.3
$247.5
$231.3
$236.9
$245.5
$263.7
$280.6
Real (2000 dollars)
per Adult
$946.78
$888.71
$885.94
$899.04
$943.26
$984.90
$1,028.34
$1,073.25
$1,136.20
$1,187.66
$1,274.53
$1,152.24
$1,148.18
$1,152.98
$1,197.51
$1,236.96
For our purposes, these data need to be expanded in two ways. First, the annual
5
data need to be converted to monthly data. Second, advertising expenditures need to
6
be forecasted through the forecasting period used in this case.
7
To accomplish this, two regression equations are estimated. First, a regression
8
equation is fitted which explains the natural logarithm of real annual advertising
9
expenditures per adult as a function of real consumption expenditures and advertising
10
employment (both per adult). This equation, which is estimated using data from 1990
11
through 2004, is summarized in Table IV-5 below.
USPS-T-7
288
Ta ble IV-5
Re gre ssion Equa tion for Annua l Adve rtising Ex pe nditure s
Constant
Consumption Expenditures
Advertising Employment
1
2
Adjusted R2
Degrees of Freedom
Coefficient
-3.589386
0.645661
0.903065
T-Statistic
-16.922
6.336
13.592
0.960
12
A second regression is run using monthly data which attempts to estimate monthly
3
seasonal coefficients from advertising employment data. This equation models
4
advertising employment as a function of consumption expenditures and seasonal
5
dummy variables for each of the twelve months. This equation is estimated using data
6
from January, 1990 through November, 2004 and is summarized in Table IV-6 below.
Ta ble IV-6
Re gre ssion Equa tion for Adve rtising Employm e nt
Consumption Expenditures
January
February
March
April
May
June
July
August
September
October
November
December
7
Adjusted R2
Degrees of Freedom
Coefficient
0.262231
-0.100274
-0.101774
-0.100773
-0.102011
-0.102288
-0.097449
-0.102050
-0.103542
-0.105289
-0.100368
-0.098812
-0.092979
T-Statistic
5.508
-0.608
-0.617
-0.610
-0.618
-0.619
-0.589
-0.617
-0.625
-0.636
-0.606
-0.596
-0.562
0.988
166
8
Based upon this latter equation, a seasonally-adjusted series is developed for
9
advertising employment by subtracting the sum of the seasonal coefficients times the
10
seasonal dummies from the natural logarithm of advertising employment and taking the
USPS-T-7
289
1
resulting anti-log. A set of seasonal multipliers is then developed by dividing advertising
2
employment by seasonally-adjusted advertising employment.
3
Monthly advertising expenditures data are constructed in three stages. First,
4
advertising expenditures are fit into the first equation above using monthly data for
5
advertising employment and consumption expenditures. Next, the seasonal pattern
6
associated with advertising employment is then applied to advertising expenditures by
7
multiplying the fitted monthly advertising expenditures data from the first step times the
8
seasonal multipliers developed for advertising employment. Finally, a year-specific
9
adjustment-factor is added to the monthly advertising expenditures data to tie the
10
11
monthly advertising expenditures data to McCann-Erickson’s annual totals.
Annual advertising expenditures are forecasted to grow with consumption
12
expenditures. These annual forecasts are then converted to monthly forecasts using
13
the three-step process outlined above. The final monthly forecasts of advertising
14
expenditures for the forecast period used in this case are shown in Table IV-7 below.
USPS-T-7
290
Ta ble IV-7
Fina l Monthly Adve rtising Ex pe nditure s
1
Month
2004.01
2004.02
2004.03
2004.04
2004.05
2004.06
2004.07
2004.08
2004.09
2004.10
2004.11
2004.12
2005.01
2005.02
2005.03
2005.04
2005.05
2005.06
2005.07
2005.08
2005.09
2005.10
2005.11
2005.12
2006.01
2006.02
2006.03
2006.04
2006.05
2006.06
2006.07
2006.08
2006.09
2006.10
2006.11
2006.12
2007.01
2007.02
2007.03
2007.04
2007.05
2007.06
2007.07
2007.08
2007.09
2007.10
2007.11
2007.12
Nominal
(billions of $)
$21.777
$21.838
$21.959
$21.854
$21.918
$21.992
$21.875
$21.744
$21.950
$22.194
$22.234
$22.364
$23.269
$23.234
$23.257
$23.318
$23.312
$23.425
$23.394
$23.359
$23.318
$23.507
$23.544
$23.681
$24.328
$24.292
$24.316
$24.407
$24.400
$24.519
$24.530
$24.493
$24.450
$24.672
$24.710
$24.855
$25.571
$25.532
$25.558
$25.624
$25.617
$25.741
$25.723
$25.685
$25.640
$25.883
$25.924
$26.075
Real (2000 dollars)
per Adult
$100.62
$100.54
$100.69
$99.97
$99.79
$99.79
$99.01
$98.37
$99.20
$99.62
$99.67
$100.25
$103.63
$103.48
$103.58
$103.12
$103.09
$103.59
$102.80
$102.65
$102.47
$102.55
$102.71
$103.31
$105.34
$105.18
$105.29
$105.02
$104.99
$105.50
$104.76
$104.60
$104.42
$104.53
$104.69
$105.30
$107.46
$107.30
$107.40
$106.86
$106.83
$107.35
$106.42
$106.26
$106.08
$106.22
$106.38
$107.01
USPS-T-7
291
iii.
1
2
Prices of Newspaper and Direct-Mail Advertising
The prices of newspaper and direct-mail advertising would be expected to be
3
functions of the expected demand for these types of advertising as well as the costs of
4
newspaper and direct-mail production.
5
The demand for newspaper and direct-mail advertising is modeled here as a
6
function of two factors: total advertising expenditures and simple time trends. Increases
7
in total advertising expenditures reflect increases in the demand for all types of
8
advertising, including, of course, newspaper and direct-mail advertising. Time trends
9
are included to measure changes in the demands for these particular types of
10
advertising over time. In the case of newspaper advertising, the trend term is
11
augmented by also including the time trend squared to reflect the fact that the positive
12
trend in the price of newspaper advertising appears to have been lessening over time.
13
The cost of both newspaper and direct-mail production is largely a function of paper
14
and printing costs. These are modeled by the producer price index for pulp, paper, and
15
allied products.
16
Taken together, then, the price of newspaper advertising is modeled as a function of
17
advertising expenditures, the producer price index for pulp, paper, and allied products
18
(lagged twelve months), a linear time trend, and the time trend squared. The dependent
19
variable in this equation is the natural logarithm of the real producer price index for
20
newspaper advertising, where this index has been deflated by the implicit price deflator
21
for consumption expenditures. The regression equation is fitted using monthly data
22
from January, 1984 through December, 2004. This equation is summarized in Table IV-
23
8 below.
USPS-T-7
292
Ta ble IV-8
Re gre ssion Equa tion for Produce r Price Inde x for Ne w spa pe r Adve rtising
Constant
Advertising Expenditures
Paper Price (lag 12 months)
Time Trend
Time Trend Squared
1
2
3
Adjusted R2
Degrees of Freedom
Coefficient
0.775712
0.049805
0.211200
0.002611
-0.000002
T-Statistic
14.728
3.311
8.451
37.984
-7.746
0.995
174
The price of direct-mail advertising is modeled as a function of advertising
4
expenditures, the producer price index for pulp, paper, and allied products (current and
5
lagged twelve months), and a linear time trend. The dependent variable in this equation
6
is the natural logarithm of the real producer price index for direct-mail advertising, where
7
this index has been deflated by the implicit price deflator for consumption expenditures.
8
The regression equation is fitted using monthly data from January, 1984 through
9
December, 2004. This equation is summarized in Table IV-9 below.
Ta ble IV-9
Re gre ssion Equa tion for Produce r Price Inde x for Dire ct-Ma il Advertising
Constant
Advertising Expenditures
Paper Price (current)
Paper Price (lag 12 months)
Time Trend
10
Adjusted R2
Degrees of Freedom
Coefficient
0.625237
0.122664
0.203293
0.154950
-0.001137
T-Statistic
13.771
9.472
8.068
7.110
-44.357
0.964
175
11
The prices of newspaper and direct-mail advertising are then forecasted based on
12
these equations using the forecast for advertising expenditures presented above, Global
13
Insight’s forecast of the producer price index of pulp, paper, and allied products, and a
14
straight mechanical projection of the linear time trend. The forecasted values of the
USPS-T-7
293
1
newspaper and direct-mail advertising prices are adjusted by a constant multiplier which
2
sets the forecasted values in the last historical period (December, 2004) equal to the
3
actual value in that period. This avoids possible problems in transitioning from actual to
4
forecasted data.
5
iv.
6
Average Delivery Days, Priority Mail
The average number of days to deliver Priority Mail is included as an explanatory
7
variable in the Priority Mail demand equation presented above. Average delivery days
8
are reported quarterly by the Postal Service by Postal quarter through 2003 and by
9
Gregorian quarter from 2003 through the first Gregorian quarter of 2005. Predicted
10
values of average delivery days are needed for two time periods, by Gregorian quarter
11
from 2000 through 2002, and throughout the forecast period (also by Gregorian quarter,
12
of course).
13
This is done by fitting a regression equation which models the natural logarithm of
14
average delivery days as a function of average delivery days one year earlier (also
15
logged), a simple linear time trend, and seasonal variables.
16
Five Gregorian seasonal variables are included. These measure the percentage of
17
the quarter which falls within the particular Gregorian seasonal. The five time periods
18
considered are January – March, April – June, July – September, October – November,
19
and December. For Gregorian quarters, of course, these variables are constant year to
20
year. In the case of the first three of these, in fact, these become simple dummy
21
variables, equal to one in the quarter of interest, zero otherwise. For Postal quarters,
22
however, these variables are equal to the number of days within the quarter that fall
23
within the Gregorian quarter of interest divided by the total number of days in the
24
quarter. In addition to these seasonal dummies, a dummy variable equal to one in
USPS-T-7
294
1
2002PQ1 is included to reflect the unique impact on Priority Mail delivery of the
2
September 11, 2001, terrorist attacks.6
This regression equation is estimated using data by Postal quarter from 1980
3
4
through 2003 and by Gregorian quarter from 2003 through the first quarter of 2005. The
5
results are summarized in Table IV-10 below.
Ta ble IV-10
Re gre ssion Equa tion for Ave ra ge De live ry Da ys for Priority Ma il
Average Delivery Days One Year Ago
Time Trend
January – March
April – June
July – September
October – November
December
2002PQ1 (September 11th Effect)
Coefficient
0.341598
-0.000427
0.545852
0.503031
0.504940
0.486587
0.735782
0.237187
Adjusted R2
Degrees of Freedom
6
T-Statistic
3.801
-2.779
7.356
7.325
7.349
7.266
6.639
4.885
0.986
97
This equation is then used to estimate fitted values of average delivery days by
7
8
Gregorian quarter for 2000, 2001, and 2002. For these purposes, the impact of the
9
September 11th dummy is assumed to have affected 2001GQ4 and 2002GQ1 equally
10
(i.e., the 2002PQ1 dummy variable is given a value of 0.5 in 2001GQ4 and 0.5 in
11
2002GQ1). These results are shown in Table IV-11 below.
6
The value of average delivery lagged one year earlier used to predict average delivery in 2003PQ1 is
adjusted to remove the unique impact of 9/11, i.e., average delivery in 2003PQ1 is a function of what
average delivery would have been in 2002PQ1 had there been no 9/11 terrorist attacks.
USPS-T-7
295
Ta ble IV-11
Ave ra ge De livery Days, 2000 - 2003
1
2
Postal
2.09
2.43
2.16
2.09
2.17
2.62
2.22
2.15
2.60
2.69
2.18
2.08
2.04
2.26
2.05
2.02
Quarter
2000.1
2000.2
2000.3
2000.4
2001.1
2001.2
2001.3
2001.4
2002.1
2002.2
2002.3
2002.4
2003.1
2003.2
2003.3
2003.4
Gregorian
2.24
2.19
2.04
2.04
2.25
2.18
2.03
2.29
2.53
2.17
2.03
2.03
2.19
2.04
2.02
2.06
Average delivery days for Priority Mail are forecasted using this same equation.
3
Forecasted values of the average delivery days variables used here are shown in Table
4
IV-12 below.
Ta ble IV-12
Avera ge De livery Days, 2004 - 2007
5
Quarter
2004.1
2004.2
2004.3
2004.4
2005.1
2005.2
2005.3
2005.4
2006.1
2006.2
2006.3
2006.4
2007.1
2007.2
2007.3
2007.4
Avg. Delivery Days
2.19
2.05
2.00
2.02
2.28
2.11
2.01
2.02
2.24
2.13
2.00
2.01
2.22
2.13
2.00
2.00
USPS-T-7
296
v.
1
Measures of Electronic Diversion Used Here
2
Several of the demand equations presented here include measures of Internet
3
activity as a means of measuring electronic diversion of various mailstreams. These
4
variables are discussed in some detail in Section II above. All of the Internet variables
5
used in my testimony are forecasted by me.
6
As measures of a relatively new technology, most measures of Internet activity are
7
experiencing significant market growth. There is a great deal of literature on how best
8
to model market penetration. Some examples of mechanical methods for dealing with
9
such things include Bass Curves and the “z-variables” used by Dr. Tolley and myself in
10
past rate cases. The dominant feature of any type of market penetration-fueled growth
11
will be that the rate of increase of such a variable has a tendency to decrease over time.
12
This is true of the Internet variables used in this case and is reflected in the forecasting
13
models which I develop below.
14
The Internet forecasts developed here have a dual nature. Obviously, every effort is
15
made to make forecasts that are reasonable and accurate measures of the expected
16
values of these specific Internet variables. Yet, it is important to keep in mind that the
17
ultimate goal here is not to be able to accurately assert how many households are
18
expected to have broadband access in 2006, for example, but to accurately assess how
19
much mail volume is likely to be reduced due to electronic alternatives over that time
20
period.
21
As discussed in Section II above, mail volume is affected by two dimensions of the
22
Internet: the breadth of the Internet (how many people use the Internet) and the depth of
23
the Internet (how many different things people use the Internet to do). In order to
24
accurately assess the risk of electronic diversion going forward, it is important to make
25
forecasts which reflect both of these dimensions. The effect of this dichotomy of
USPS-T-7
297
1
breadth versus depth on the choice of Internet variables used here is discussed in detail
2
in Section II above. It was important that I keep this issue in mind as I developed the
3
Internet forecasts used in this case.
4
I believe that the forecasts presented here do a very good job of satisfying both the
5
need for accurate forecasts of the specific variables being forecasted as well as
6
providing accurate projections of the expected impact of the Internet and electronic
7
diversion on mail volumes.
8
9
10
There are three raw Internet variables which are used in my testimony: consumption
expenditures on Internet Service Providers, the number of broadband subscribers, and
Internet advertising expenditures. These three variables are forecasted below.
11
(a) Consumption Expenditures on Internet Service Providers
12
Consumption expenditures on Internet Service Providers are used to construct the
13
Internet Experience variable that is included in the First-Class single-piece letters, First-
14
Class cards, and Free for the Blind and Handicapped Mail equations presented earlier
15
in my testimony. The construction of this variable is described in Section II above. In
16
order to construct forecasted values of Internet Experience, it is necessary to make
17
forecasts for total consumption expenditures on Internet Service Providers (ISP
18
consumption) as well as for the price index for ISP consumption. The latter number is
19
simply assumed to remain constant throughout the forecast period. The forecast of ISP
20
consumption is described next.
21
Consumption expenditures on Internet Service Providers have increased almost
22
without exception since 1988. The rate at which these expenditures have increased
23
has decreased considerably, however, from over 100 percent annual growth for its first
24
three years to less than 20 percent per year for the last three years.
USPS-T-7
298
1
ISP consumption data from 1988 through 2004 are shown in Table IV-13 below.
2
Table IV-13 also presents data on ISP consumption as a share of total consumption
3
expenditures. As the last column of Table IV-13 indicates, the rate of growth of ISP
4
consumption as a share of total consumption expenditures has fallen considerably. For
5
the first eleven months of 2004, ISP consumption as a share of total consumption grew
6
by less than 5 percent.
Ta ble IV-13
Consumption Ex penditure s, Interne t Se rvice Provide rs
(historical)
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004*
7
8
9
10
11
Total
(billions of dollars)
$0.025
$0.050
$0.100
$0.200
$0.305
$0.412
$0.805
$1.611
$2.675
$3.575
$4.709
$7.004
$9.481
$12.156
$13.228
$15.435
$17.084
Percentage Change
from Previous Year
100.67%
100.00%
100.08%
52.39%
35.12%
95.43%
100.03%
66.08%
33.65%
31.71%
48.75%
35.37%
28.21%
8.82%
16.69%
10.68%
Percentage of
Total Consumption
0.001%
0.001%
0.003%
0.005%
0.007%
0.009%
0.017%
0.032%
0.051%
0.064%
0.080%
0.111%
0.141%
0.172%
0.179%
0.199%
0.208%
Percentage Change
from Previous Year
87.01%
87.42%
92.75%
43.43%
27.80%
84.49%
90.68%
57.20%
26.65%
24.27%
39.20%
26.19%
22.48%
4.08%
10.90%
4.74%
* Thru November, 2004
In the long run, consumption expenditures on Internet Service Providers are likely to
trend toward being simply a constant percentage of total consumption expenditures.
This forms the basis for developing the forecast of ISP consumption used here.
A regression equation is fitted which has as its dependent variable ISP consumption
12
as a percentage of total consumption expenditures. The explanatory variables in this
13
equation are a simple linear time trend and the time trend squared. This equation is
USPS-T-7
299
1
fitted using monthly data from January, 1997 through November, 2004. The results are
2
summarized in Table IV-14 below.
Ta ble IV-14
Re gre ssion Equa tion for ISP Consumption a s a Pe rce nta ge of Tota l Consumption
Constant
Time Trend
Time Trend Squared
3
4
Adjusted R2
Degrees of Freedom
Coefficient
0.000411
0.000027369
-0.000000094
T-Statistic
19.358
26.821
-9.141
0.983
92
The positive coefficient on the Time Trend indicates that ISP consumption as a
5
share of total consumption expenditures is expected to increase over time. The
6
negative coefficient on the Time Trend Squared term, however, indicates that this rate
7
of increase is expected to decline over time.
8
9
Consumption expenditures on Internet Service Providers are then forecasted using
this equation. The forecasted values of ISP consumption are adjusted by a constant
10
multiplier which sets the forecasted value in the last historical period (November, 2004)
11
equal to the actual value in that period. This avoids possible problems in transitioning
12
from actual to forecasted data. The forecasted values of consumption expenditures on
13
Internet Service Providers used in this case are presented in Table IV-15 below.
USPS-T-7
300
1
Ta ble IV-15
Consumption Ex pe nditures, Inte rne t Service Provide rs
(forecast)
Total
Percentage of
(billions of dollars)
Total Consumption
2004.01
$16.606
0.207%
2004.02
$16.740
0.208%
2004.03
$16.945
0.209%
2004.04
$16.623
0.205%
2004.05
$16.648
0.203%
2004.06
$17.054
0.209%
2004.07
$17.201
0.208%
2004.08
$17.372
0.210%
2004.09
$17.466
0.211%
2004.10
$17.588
0.210%
2004.11
$17.683
0.211%
2004.12
$17.762
0.211%
2005.01
$18.022
0.212%
2005.02
$18.098
0.213%
2005.03
$18.173
0.214%
2005.04
$18.455
0.215%
2005.05
$18.528
0.216%
2005.06
$18.599
0.217%
2005.07
$18.854
0.217%
2005.08
$18.922
0.218%
2005.09
$18.989
0.219%
2005.10
$19.260
0.220%
2005.11
$19.324
0.221%
2005.12
$19.387
0.221%
2006.01
$19.684
0.222%
2006.02
$19.744
0.223%
2006.03
$19.802
0.223%
2006.04
$20.095
0.224%
2006.05
$20.150
0.225%
2006.06
$20.204
0.225%
2006.07
$20.522
0.226%
2006.08
$20.573
0.226%
2006.09
$20.623
0.227%
2006.10
$20.930
0.227%
2006.11
$20.977
0.228%
2006.12
$21.022
0.228%
2007.01
$21.304
0.229%
2007.02
$21.346
0.229%
2007.03
$21.386
0.230%
2007.04
$21.680
0.230%
2007.05
$21.717
0.230%
2007.06
$21.752
0.231%
2007.07
$22.056
0.231%
2007.08
$22.088
0.232%
2007.09
$22.118
0.232%
2007.10
$22.439
0.232%
2007.11
$22.466
0.232%
2007.12
$22.491
0.233%
USPS-T-7
301
1
2
(b) Number of Broadband Subscribers
Table IV-16 below shows the number of broadband subscribers by quarter. As with
3
ISP consumption, the number of broadband subscribers has continued to increase
4
throughout its history, but at a decreasing rate.
5
Ta ble IV-16
Numbe r of Broa dband Subscribe rs
(historical)
Percentage Change
Subscribers from Previous Quarter
Quarter
1997Q1
0.007
1997Q2
0.014
100.00%
1997Q3
0.029
100.00%
1997Q4
0.058
100.00%
1998Q1
0.115
100.00%
1998Q2
0.230
100.00%
1998Q3
0.345
50.00%
1998Q4
0.460
33.33%
1999Q1
0.695
51.09%
1999Q2
0.930
33.81%
1999Q3
1.165
25.27%
1999Q4
1.400
20.17%
2000Q1
2.500
78.57%
2000Q2
3.600
44.00%
2000Q3
4.700
30.56%
2000Q4
5.800
23.40%
2001Q1
6.692
15.39%
2001Q2
8.260
23.42%
2001Q3
9.530
15.38%
2001Q4
11.000
15.42%
2002Q1
12.179
10.72%
2002Q2
13.793
13.25%
2002Q3
15.654
13.49%
2002Q4
17.369
10.96%
2003Q1
19.068
9.78%
2003Q2
20.667
8.39%
2003Q3
22.686
9.77%
2003Q4
24.624
8.54%
2004Q1
26.899
9.24%
2004Q2
28.625
6.42%
2004Q3
30.953
8.13%
6
The number of broadband subscribers is forecasted by fitting an equation which
7
models the natural logarithm of the percentage change in the number of broadband
USPS-T-7
302
1
subscribers from the previous quarter as a function of a simple linear time trend. This
2
equation is fitted over a sample period from 2001Q3 through 2004Q3 and is
3
summarized in Table IV-17 below.
Ta ble IV-17
Re gre ssion Equa tion for Qua rte rly Pe rce nta ge Cha nge in Num be r of Broa dba nd Subscribe rs
Constant
Time Trend
Adjusted R 2
Degrees of Freedom
4
5
Coefficient
-1.840034
-0.060620
T-Statistic
-25.167
-6.581
0.779
11
By taking the natural logarithm of the change in the number of broadband
6
subscribers as the explanatory variable in the above equation, the forecasted growth
7
rate for broadband subscribers is guaranteed to remain positive throughout the forecast
8
period. The negative coefficient on the time trend indicates that the average growth rate
9
has declined over time. The result is that the number of broadband subscribers is
10
11
projected to increase throughout the forecast period but at a decreasing rate.
The forecasted growth rate for the number of broadband subscribers is projected by
12
fitting this equation. The total number of broadband subscribers is then forecasted by
13
applying forecasted growth rates to the projected number of broadband subscribers
14
from the previous quarter.
15
16
Forecasted values for the number of broadband subscribers are shown in Table IV18 below.
USPS-T-7
303
Ta ble IV-18
Numbe r of Broa dband Subscribe rs
(forecast)
Percentage Change
Quarter
Subscribers from Previous Quarter
1
2
3
Actual
2003Q1
2003Q2
2003Q3
2003Q4
2004Q1
2004Q2
2004Q3
19.068
20.667
22.686
24.624
26.899
28.625
30.953
9.78%
8.39%
9.77%
8.54%
9.24%
6.42%
8.13%
Forecast
2004Q4
2005Q1
2005Q2
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
2007Q3
2007Q4
33.056
35.171
37.289
39.402
41.503
43.586
45.646
47.675
49.670
51.627
53.541
55.409
57.228
6.80%
6.40%
6.02%
5.67%
5.33%
5.02%
4.72%
4.45%
4.18%
3.94%
3.71%
3.49%
3.28%
(c) Internet Advertising Expenditures
Internet advertising expenditures have been reported quarterly by the Interactive
4
Advertising Bureau (IAB) since 1996, based on data compiled by
5
PricewaterhouseCoopers. These data are shown in Table IV-19 below. Table IV-19
6
also presents these numbers as percentages of total advertising expenditures, as
7
measured by Robert Coen of McCann-Erickson.
USPS-T-7
304
Ta ble IV-19
Interne t Adve rtising Ex pe nditure s
(historical)
Internet Advertising
Percent of
Quarter
(millions of dollars)
Total Advertising
1996Q1
30
0.07%
1996Q2
52
0.12%
1996Q3
76
0.17%
1996Q4
110
0.24%
1997Q1
130
0.28%
1997Q2
214
0.45%
1997Q3
227
0.47%
1997Q4
336
0.68%
1998Q1
351
0.69%
1998Q2
423
0.82%
1998Q3
491
0.95%
1998Q4
656
1.24%
1999Q1
693
1.28%
1999Q2
934
1.69%
1999Q3
1,217
2.18%
1999Q4
1,777
3.10%
2000Q1
1,922
3.15%
2000Q2
2,091
3.40%
2000Q3
1,951
3.14%
2000Q4
2,123
3.37%
2001Q1
1,872
3.17%
2001Q2
1,848
3.17%
2001Q3
1,773
3.11%
2001Q4
1,641
2.88%
2002Q1
1,520
2.54%
2002Q2
1,458
2.46%
2002Q3
1,451
2.47%
2002Q4
1,580
2.67%
2003Q1
1,632
2.67%
2003Q2
1,660
2.69%
2003Q3
1,793
2.94%
2003Q4
2,182
3.53%
2004Q1
2,230
3.40%
2004Q2
2,369
3.60%
2004Q3
2,430
3.71%
1
2
Internet advertising expenditures accounted for 3.6 percent of total advertising
3
expenditures for the first three quarters of 2004. This share has increased by 50
4
percent in two years.
USPS-T-7
305
1
Total Internet advertising expenditures are forecasted by fitting an equation which
2
models Internet advertising expenditures as a function of the change in total
3
employment and a simple linear time trend.
4
A second set of time trends is also included in the equation to reflect an unusual
5
boom/bust phenomenon from 1999 – 2001. Basically, a model with just the change in
6
employment and a single time trend over the entire sample does an excellent job of
7
fitting total Internet advertising expenditures through early 1999 and also since 2002.
8
This is shown graphically in Figure IV-1 below.
Figure IV-1
Internet Advertising Expenditures
3,000
2,500
2,000
Actual
1,500
Fitted
1,000
500
0
1997
1998
1999
2000
2001
2002
2003
2004
9
10
As can be seen in Figure IV-1, however, Internet advertising expenditures grew at
11
an unprecedented rate in 1999 and 2000. Then, in 2001 and 2002, Internet advertising
12
expenditures fell at an unprecedented rate. As Figure IV-1 indicates, though, Internet
13
advertising expenditures at the end of this period were very close to what would have
14
been expected as of the beginning of 1999. To fit the 1999 – 2002 period, two
15
additional variables are included in the Internet advertising expenditures regression, a
USPS-T-7
306
1
time trend equal to zero through 1999Q1, increasing by one from 1999Q2 through
2
2002Q4, and remaining constant thereafter, and this time trend squared.
3
4
This equation is fitted over a sample period from 1997Q1 through 2004Q3 and is
summarized in Table IV-20 below.
Ta ble IV-20
Re gre ssion Equa tion for Inte rne t Adve rtising Ex pe nditure s
Constant
Change in Employment
Time Trend
Time Trend, 1999Q2 – 2002Q4
1999Q2 Time Trend Squared
5
6
Adjusted R 2
Degrees of Freedom
Coefficient
-799.979
185.193
80.507
281.295
-15.428
T-Statistic
-10.118
8.038
7.640
10.353
-17.183
0.979
26
Internet advertising expenditures are then forecasted based on this equation, using
7
Global Insight’s forecast of total employment. The forecasted levels of Internet
8
advertising are adjusted by a constant multiplier which sets the forecasted value in the
9
last historical period (2004Q3) equal to the actual value in that period. This procedure
10
11
12
avoids possible problems in transitioning from actual to forecasted data.
Forecasted values for Internet advertising expenditures, both in millions of dollars
and as a percent of total advertising expenditures, are shown in Table IV-21 below.
USPS-T-7
307
Ta ble IV-21
Interne t Adve rtising Ex pe nditure s
(forecast)
Internet Advertising
Percent of
Quarter
(millions of dollars)
Total Advertising
1
2
3
4
Actual
2003Q1
2003Q2
2003Q3
2003Q4
2004Q1
2004Q2
2004Q3
1,632.0
1,660.0
1,793.0
2,182.0
2,230.0
2,369.0
2,430.0
2.67%
2.69%
2.94%
3.53%
3.40%
3.60%
3.71%
Forecast
2004Q4
2005Q1
2005Q2
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
2007Q3
2007Q4
2,572.2
2,691.9
2,740.4
2,858.0
2,919.5
2,955.3
3,007.0
3,062.6
3,116.7
3,180.6
3,240.7
3,302.5
3,383.5
3.85%
3.86%
3.91%
4.08%
4.13%
4.05%
4.10%
4.17%
4.20%
4.15%
4.21%
4.29%
4.34%
4. Composite Multiplier
The composite multiplier is made up of all of the projection factors which are not
included in the rate-effect or non-rate effect multipliers.
5
In general, the components of the composite multiplier fall into four categories:
6
seasonal multipliers, share forecasts, non-econometric multipliers, and a quarterly
7
adjustment multiplier.
8
9
10
The first of these, seasonal multipliers, are constructed based upon seasonal
variables included in the econometric demand equations. The classification of
seasonality as used in the demand equations was described in the introduction to
USPS-T-7
308
1
Section II above. As described there, the seasonal variables and their coefficients can
2
be combined into a single seasonal index, SIt = Σ(seasonals) bi•Sti, so that volume can be
3
related to the seasonal index as follows:
Ln(Volume at time t) = a + Σ bti•xti + SIt + et
4
5
A seasonal multiplier can then be constructed from this seasonal index in a way
6
similar to that used to construct projection factors from other variables, by taking the
7
anti-log of the seasonal index. That is,
SMti = exp(SIti) / exp(SIBi)
8
9
10
where exp(.) indicates taking the anti-log (the anti-log of x is equal to ex).
The second component of the composite multiplier, the share forecast, is applied to
11
some categories of mail for which forecasts are made at a finer level of detail than the
12
level at which I estimate demand equations. Share equations are described in detail in
13
Section V below.
14
If any non-econometric information is introduced in the forecasting equations, it
15
would be entered as part of the composite multiplier. The volume forecast used in this
16
case contains no explicit adjustments of this type.
17
The final component of the composite multiplier, the quarterly-adjustment multiplier,
18
is identical across all mail categories. It is equal to the number of delivery days within
19
quarter t divided by the number of Postal delivery days within the base period. This
20
multiplier adjusts the forecast to reflect the fact that VB in Equation IV.1 is an annual
21
volume, while Vt is a quarterly volume.
USPS-T-7
309
1
V. Forecasts of Component Mail Categories
2
A. Overview
3
In many cases, the level of detail at which the Postal Service estimates demand
4
equations is not sufficiently fine as to be satisfactory for accurately projecting Postal
5
Service revenues and costs. In these cases, it is therefore necessary to make more
6
detailed forecasts of individual mail categories within certain subclasses of mail.
7
In this specific case, two types of further breakdowns of mail categories are
8
considered here: mailers’ use of presort and automation discounts offered by the Postal
9
Service and differences in the shape of the mail. Share equations which focus on the
10
former of these – presort and automation share equations – are developed in this case
11
for First-Class workshared letters, First-Class cards, Standard Regular, and Standard
12
Nonprofit mail. Share equations of the latter type – letters versus non-letters – are
13
developed here for Standard Regular mail. In all other cases, the relative shares of mail
14
categories within a particular subclass of mail are treated as if they were assumed to
15
remain constant throughout the forecast period.
16
The equations used to project the shares of First-Class and Standard Mail that will
17
take advantage of specific presort and automation discounts are presented in section B
18
below. Letter versus non-letter shares of Standard Regular mail are discussed in
19
section C below.
B. Presort and Automation Shares
20
1. Basic Theory of Consumer Worksharing
21
22
Traditionally, economists have modeled consumer demand as an effort by
23
consumers to maximize utility given income. On the other side of the same consumer
24
demand coin, however, is the basic problem of minimizing costs for a given level of
25
utility.
USPS-T-7
310
1
Mathematically, consumers’ cost-minimization problem can be expressed as:
2
3
4
5
where x is the quantity of the good of interest, U is the consumer’s utility function, C is
6
the consumer’s cost function, and uR is the consumer’s reservation utility.
7
8
9
10
11
12
min C(x) s.t. U(x)>uR
(Equation V.1)
In general, C(x) is equivalent to the price of good x, including any transactions costs,
so that
C(x) = p•x + transactions costs
(Equation V.2)
where p is the price of good x.
Assuming that transactions costs are exogenous to the consumer and the consumer
13
takes price as given in Equation V.2, the minimand of the cost-minimization equation,
14
Equation V.1, will simply be x.
15
For some categories of mail, however, the Postal Service offers discounts to mailers
16
who presort or barcode their mail, thereby making the Postal Service’s job easier. In
17
such a case, the cost-minimization equation can be re-written as follows:
18
19
20
21
where d is the discount obtained by the consumer for doing additional work, and u is the
22
unit cost to the consumer of doing the additional work, which may vary with x. In this
23
case, in addition to choosing x in Equation V.3, the consumer will also choose the level
24
of worksharing.
25
C(x) = (p-d+u(x))•x + transactions costs
(Equation V.3)
For any given value of x, minimizing C(x) is equivalent to minimizing the price paid
26
for good x, or minimizing [p - d + u(x)]. Taking p as fixed for the consumer, this can be
27
further simplified to a simple choice of minimizing [-d + u(x)] or, rearranging terms,
28
maximizing [d - u(x)]. That is, a consumer will choose the worksharing option that
USPS-T-7
311
1
maximizes his or her benefit of worksharing, where the consumer’s benefit to
2
worksharing is equal to d - u.
3
In general, the level of worksharing will not be a continuous function, but will instead
4
involve a choice from among discrete levels of worksharing. Mathematically, this
5
becomes:
6
7
8
9
maxi (di - ui(x))
(Equation V.4)
for i equals the set of all possible worksharing options, where di is the discount
10
associated with worksharing option i, ui is the cost to the consumer of qualifying for
11
worksharing option i, and x is the quantity of the good consumed.
12
13
2. Derivation of Basic Share Equation
Solving Equation V.4 requires information about the user costs associated with all
14
possible worksharing categories. If there are N worksharing options, this becomes an
15
N-dimensional problem. If N is very large at all, this can quickly become an intractable
16
problem.
17
One possible way of making Equation V.4 a more tractable problem is to introduce
18
the concept of opportunity costs into u(x). Economists generally think of the opportunity
19
cost associated with a product as the forgone benefit of not doing anything different with
20
the product. In the context of Equation V.4, then, the opportunity cost of using
21
worksharing option i is the maximum benefit, where benefit is defined as d - u, which
22
could be achieved by using a different worksharing category. Explicitly incorporating
23
opportunity costs into Equation V.4 yields the following consumer maximization
24
problem:
25
26
27
maxi [di - (wi(x) + maxj≠i(dj-uj))]
(Equation V.5)
USPS-T-7
312
1
where wi equals the cost of qualifying for worksharing option i, excluding opportunity
2
costs, and ui = (wi(x) + maxj≠i(dj-uj)).
3
If maxj≠i(dj-uj) > di - wi, for some worksharing option j, then di - (wi(x) + maxj≠i(dj-uj))
4
will be strictly less than zero. If worksharing discounts are defined as discounts from a
5
base price for which consumers are eligible at no additional cost (i.e., d=0 and w=0 for
6
the base worksharing option), then maxi≠i(di - ui )≥0, since, if any given worksharing
7
option were more costly to the consumer than the discount earned as a result of
8
qualifying for the option, the consumer could still choose to do no worksharing at no
9
cost.
10
11
12
13
Combining these two facts yields the following result:
di - ui ≥ 0 if, and only if, di - wi ≥ dj - wj for all worksharing options j.
Stated in words, then, a consumer will utilize a worksharing option if, and only if, the
14
costs to the consumer of doing so are less than the discount offered by the seller for
15
doing so.
16
3. Modeling Consumers’ Use of Worksharing Options
17
This reduces Equation V.4 from an N-dimensional problem to a system of N one-
18
dimensional problems. A consumer will use worksharing option i if, and only if, di - ui ≥
19
0. Mathematically, this can be represented by Equation V.6 below:
20
21
22
23
24
(Percentage of mail within a category) = ∫0d p.d.f. (u) du
(Equation V.6)
Thus, the share of a good that will be sent as part of a particular worksharing option
can be solved for by estimating Equation V.6.
USPS-T-7
313
1
a. Modeling User-Cost Distributions
2
The first step in solving Equation V.6 is to define what type of distribution best
3
describes the user-cost distribution. The most likely candidate would seem to be the
4
normal distribution.
5
Probably the most common empirical distribution is the normal distribution. A
6
number of social and economic variables have been shown to be generally normally
7
distributed, including income. In addition, user costs that decline at a constant rate
8
would lead to logistic growth in the use of worksharing options. This is generally
9
consistent with historical growth patterns in the use of presortation and automation
10
discounts offered by the Postal Service.
11
Finally, the Central Limit Theorem states that:
12
13
14
15
16
17
If an arbitrary population distribution has a mean µ and finite variance σ2, then
the distribution of the sample mean approaches the normal distribution with mean µ
and variance σ2/n as the sample size n increases. (Anderson and Bancroft,
Statistical Theory in Research, McGraw-Hill, 1952, p. 71)
This means that any sample distribution with finite mean and variance is
18
approximately normal. A consumer user-cost distribution would certainly be expected to
19
have both a finite mean and variance. Thus, it is reasonable to assume that user costs
20
are normally distributed for consumer worksharing options. Despite the appeal of the
21
normal distribution, it is not without its limitations, however.
22
23
i. Issue 1: Population of Potential Worksharers
One issue to be resolved in modeling the share of consumers that will use a
24
particular worksharing option is to properly identify the consumer population of potential
25
work sharers. For example, not everybody who mails a letter has a realistic option of
26
presorting or automating their mail due to limitations imposed by the Postal Service that
27
presorted mailings must include at least 500 pieces or practical limitations against
USPS-T-7
314
1
purchasing barcoding equipment that can cost more than $100,000. On the other hand,
2
consider a mailer who sends a letter to every address in a particular city (e.g., utility bills
3
and saturation advertising). Such a mailer will likely either presort as finely as possible
4
(carrier-route presorting or saturation presorting) or not presort at all, but would have
5
little reason to consider intermediate presort options (e.g., 3- or 5-digit presorting).
6
In reality, therefore, user-cost distributions may have several clusters of consumers.
7
For example, the user-cost distribution associated with 3-digit Automated mail may have
8
multiple peaks, rather than being a purely symmetric distribution. At one extreme may
9
be mailers who mail letters one or two at a time. The “costs” to these mailers of
10
qualifying for the Postal Service’s 3-digit presort requirement would basically involve
11
preparing an additional 400-500 letters to meet the minimum mailing requirement for the
12
3-digit presort requirement. In addition, such mailers may have to purchase barcoding
13
equipment, which would be prohibitively expensive. A middle “hump” of the distribution
14
may be associated with mailers who would never consider only 3-digit presorting their
15
mail as long as more attractive discounts existed for 5-digit or carrier-route presorting.
16
In such a case, the user-cost distribution could reasonably be thought of as being
17
normally distributed only over the small subset of mailers who have sufficient density
18
and low opportunity costs associated with 3-digit Automation (although these
19
opportunity costs will likely still be prohibitive for some mailers). As long as the discount
20
for the worksharing category falls within this area of the user-cost distribution, however,
21
then a normal distribution over that subset of consumers will be a valid approximation to
22
the true user-cost distribution.
23
24
25
ii. Issue 2: Negative User Costs
Technically, a normal user-cost distribution would assume that user costs could take
on any value from -∞ to +∞. If user costs are defined as the costs associated with
USPS-T-7
315
1
qualifying for a worksharing category, above and beyond the cost of qualifying for the
2
corresponding non-workshared category, then this means that the true distribution of
3
user costs associated with any worksharing option must be non-negative. Thus, the
4
true user-cost distribution associated with any worksharing category for which a non-
5
worksharing option exists will have a lower bound of zero user costs.
6
7
iii. Issue 3: Non-Integrability of Normal p.d.f.
Finally, an empirical problem with a normal user-cost distribution is that the normal
8
probability density function (p.d.f.) is not integrable, so that Equation V.6 would be non-
9
solvable. Solving Equation V.6 for a normal user-cost distribution would require either a
10
discrete approximation to the normal c.d.f., or an approximation to the normal p.d.f.
11
which is integrable. The latter of these two options is chosen here.
12
13
iv. Resolution of Issues
A distribution that is often used to approximate the normal distribution, due to its
14
similarity to the normal distribution and numerical simplicity, is the logistic distribution.
15
(See, for example, Judge, et al., The Theory and Practice of Econometrics, 2nd edition,
16
John Wiley and Sons, 1985, p. 762).
17
18
19
20
The logistic p.d.f. takes the following form:
Logistic p.d.f. = e-((x-µ)/σ) / {σ[1+e-((x-µ)/σ)]2}
(Equation V.7)
The main advantage of the logistic distribution over the normal distribution is that the
21
logistic p.d.f. is integrable. Inserting the logistic p.d.f. into Equation V.6 allows the
22
equation to be solved as follows:
23
24
25
26
27
(Pct. of good x within worksharing category i) = ∫-∞dj e-((ui-µi)/σi) / { σi [1+e-((ui-µi)/σi)]2 dui
or, integrating the logistic p.d.f.
USPS-T-7
316
(Pct. of good x within worksharing category i) = 1 / [1+e-((di-µi)/σi)]
(Equation V.8)
1
2
3
As discussed above, user costs may be normally (or logistically) distributed only
4
over a subset of the total consumers of good x. Equation V.8 actually measures the
5
percentage of good x for which the user-cost distribution is logistically distributed which
6
will be sent within category i. The percentage of all of good x within worksharing
7
category i is the product of Equation V.8 and the percentage of good x over which the
8
user-cost distribution associated with worksharing category i is logistically distributed, or
9
10
11
12
where αi is the percentage of good x for which user costs associated with worksharing
13
category i are logistically distributed. The parameter αi represents the maximum
14
percentage of good x which would ever take advantage of worksharing category i, for
15
any likely discount associated with category i. Thus, αi may be called the “ceiling”
16
share associated with worksharing category i.
17
(Pct. of good x within worksharing category i) = αi / [1+ e-((di-µi)/σi)]
(Equation V.9)
The logistic distribution has the same drawback as the normal distribution in that the
18
logistic distribution assumes that user costs can take on any value from -∞ to +∞. In
19
reality, however, user costs have a lower bound of zero, by definition, for reasons
20
discussed above.
21
The simplest way of constraining user costs to be greater than or equal to zero is to
22
assume that user costs falling below zero are actually exactly equal to zero. This
23
procedure leads to a censored logistic distribution associated with user costs. As long
24
as di>0, Equation V.9 will be unchanged due to this type of censoring.
25
26
27
b. Changes in the User-Cost Distribution over Time
If Equation V.9 is to be used in evaluating the use of worksharing options over time
or in forecasting the future use of worksharing options, then the user-cost distribution
USPS-T-7
317
1
outlined in Equation V.9 must be allowed to vary over time. There is no reason to
2
believe that user costs are constant for any or all consumers over time. In fact, if the
3
shares of worksharing categories change independently of changes in discounts, as has
4
happened with Postal worksharing categories, then the user-cost distributions
5
associated with these categories must be changing over time.
6
The crucial need, then, in modeling the use of worksharing categories is to
7
adequately model the changes in user-cost distributions over time. There are four types
8
of changes in user-cost distributions which may occur over time: changes in the type of
9
distribution, changes in the standard deviation of the distribution (σ), changes in the
10
percentage of the good over which user costs are normally distributed (α), and changes
11
in the mean of the user-cost distribution (µ). These four issues are considered
12
separately below.
13
14
i. Changes in the Type of Distribution
Arbitrary changes in the general shape of user-cost distributions over time would be
15
extremely problematic empirically. At the extreme, if the type of user-cost distribution
16
changed over time, then it would not be valid to base forecasts of future use of
17
worksharing categories on historical patterns, as there would be no guarantee that the
18
distribution might not change shape in the future.
19
Fortunately, there is no reason to believe that user-cost distributions would change
20
type over time. The Central Limit Theorem suggests that, if anything, user-cost
21
distributions ought to appear more normal over time. Thus, as an empirical matter, it is
22
likely to be a valid assumption that all user-cost distributions are logistically distributed
23
over their entire histories.
USPS-T-7
318
1
ii. Changes in the Standard Deviation of the Distribution
2
There is no a priori reason to assume that the standard deviation of the user-cost
3
distribution, σ, would remain constant over time. A potential difficulty in modeling
4
changes in σ, however, arises in interpreting changes in σ over time.
5
The effects of changes in σ are dependent on where the discount lies along the
6
user-cost distribution. A decline in the standard deviation of the distribution will lead to
7
an increase in the use of the worksharing option if the discount is greater than the mean
8
of the user-cost distribution, but will lead to a decrease in the use of the worksharing
9
option if the discount is less than the mean.
10
Another empirical difficulty in permitting σ to change over time is a computational
11
difficulty in modeling unique shifts in d, µ, and σ over time in Equation V.9. A
12
convergent solution to Equation V.9 is facilitated if one takes either the numerator (i.e.,
13
−(d-µ)) or the denominator (i.e., σ) of the exponential expression as constant over time.
14
Since d is exogenous, and can be expected to change over time, it is convenient to hold
15
σ constant.
iii. Changes in the Ceiling of the Distribution
16
17
If a new category is introduced, the opportunity costs associated with older lower-
18
discount categories may rise dramatically for many consumers as they shift into the
19
newer more-discounted worksharing category. Alternately, there may be long-run shifts
20
in the concentration of mail. Such shifts may shift some mail either into or out of the
21
portion of the user-cost distribution over which user costs are normally (logistically)
22
distributed.
23
Shifts of this nature over time could be modeled in Equation V.9 through a change in
24
the value of α over time. Empirically, it should be noted, however, that it might be
25
difficult to isolate gradual changes in α (modeled, for example, through a simple time
USPS-T-7
319
1
trend) from changes in µ that will be discussed below. Thus, it may be desirable as a
2
practical matter to be cautious in modeling changes in α over time.
iv. Changes in the Mean of the User-Cost Distribution
3
4
In estimating the share of a good which would take advantage of a particular work-
5
sharing option over time, the variable which would generally be expected to change the
6
most over time (except, perhaps, for the discount) would be the mean of the user-cost
7
distribution. Changing the mean of the user-cost distribution suggests that user costs
8
shift proportionally across all consumers. This would generally be true of such things as
9
fixed capital costs associated with worksharing (e.g., barcoding machines to prebarcode
10
mail), shocks to costs from changes in worksharing requirements, and falling user costs
11
in the initial periods following the introduction of worksharing options as consumers
12
optimize their costs of worksharing.
13
Estimating the share of a good, x, that takes advantage of a particular worksharing
14
option, i, historically then becomes a matter of incorporating historical changes in the
15
discount associated with worksharing option i, the mean user-cost associated with
16
worksharing option i, and the percentage of good x for which user costs associated with
17
worksharing option i are logistically distributed into Equation V.9. Forecasting the share
18
of good x that would be expected to use worksharing option i would require forecasts of
19
di, µi, and αi.
20
c. Opportunity Costs
21
For consumer goods with multiple worksharing options (e.g., separate discounts for
22
various levels of presortation offered by the Postal Service), a critical component of the
23
user costs of worksharing will be opportunity costs as outlined above. Opportunity costs
24
as derived in Equation V.5 can be decomposed into the opportunity costs associated
25
with not using all other categories. That is,
USPS-T-7
320
1
2
3
4
oci = Σocnot using j for all j≠i
(Equation V.10)
For any individual mailer, the opportunity costs associated with not using category j
5
will be equal to zero for all categories except for the one category that the mailer
6
actually chooses. For the distribution of all mailers, however, the above equation
7
makes the calculation of opportunity costs rather straightforward.
8
A logistic user-cost distribution is uniquely defined by three parameters – α, µ, and
9
σ. In general, opportunity costs do not directly affect σ. For computational simplicity, it
10
is best to treat α as remaining constant over time. Thus, opportunity costs would only
11
affect α implicitly.
12
13
The mean of the user-cost distribution, µ, can be decomposed into the following
equation, based on the theoretical implications of Equation V.5 above.
14
15
16
17
where µnon-oc is equal to the mean user cost, excluding opportunity costs, and ocij is the
18
forgone benefit of using category i instead of category j.
µi = µnon-oc + Σj≠i E(ocij)
(Equation V.11)
19
For those consumers for whom category j is the most attractive worksharing option
20
(and would, thus, use worksharing category j), ocij will equal dj - uj, the benefit of using
21
category j. For those consumers for whom category j is not the most attractive
22
worksharing option, ocij is equal to zero. This leads to the following formula for the
23
expected value of ocij:
24
25
26
27
where ūj is equal to the average cost of using worksharing category j by consumers who
28
actually use category j, and ŝij is equal to the percentage of good x for which user costs
E(ocij) = (dj - ūj)•(ŝij)
(Equation V.12)
USPS-T-7
321
1
associated with worksharing category i are logistically distributed that take advantage of
2
worksharing category j.
3
Opportunity cost, as expressed in Equation V.12, is an important component of the
4
theoretical model underpinning the presort and automation share equations presented
5
here. For simplicity reasons, however, opportunity costs are not explicitly modeled in
6
the actual share equations used in this case. Empirical aspects of Equation V.12 were
7
discussed in my testimony in previous rate cases. Because opportunity costs are not
8
explicitly modeled in this case, this discussion is not repeated here.
9
d. Empirical Problem to be Solved to Model Use of Worksharing
10
For a good x, whose seller offers consumers discounts from the basic price of good
11
x associated with N distinct mutually exclusive worksharing options to consumers,
12
identified as option 1, option 2, ..., option N, where option 1 reflects no worksharing and
13
is offered for the base-line price of good x, the share of good x that will take advantage
14
of each of the N various worksharing categories can be determined by a system of N
15
equations, (N-1) of which are variations of Equation V.9 as follows:
16
sit = αit / [1 + e-(dit-{µit+Σj≠i ocijt})/σi], for i, j = 2, …, N
17
The share of good x that will take advantage of the base worksharing category,
18
19
20
(Equation V.13)
category 1, is then simply equal to
s1 = 1 - Σi=2,...,N si
(Equation V.14)
The dependent variables of this equation system are sit, i = 1 to N. Values of dit
21
must be taken as given. The values for αit, µit, and σi, for i = 2 to N, are then the
22
parameters to be estimated in this system of equations.
23
24
25
4. Econometric Share Equations
Equation V.13 is fit historically for each worksharing category of First-Class letters,
First-Class cards, Standard Regular, and Standard Nonprofit mail. The resulting
USPS-T-7
322
1
econometric values of αt, µt, and σ are then used to forecast the shares of the various
2
worksharing categories.
3
First-Class letters are divided into two categories for forecasting purposes: single-
4
piece and workshared letters. Share equations are used to separate First-Class
5
workshared letters into seven categories: presort nonautomated letter, flats, and IPPs;
6
automation basic letters; automation 3-digit letters; automation 5-digit letters;
7
automation basic flats; automation 3/5-digit flats; and carrier-route presort letters.
8
9
10
11
First-Class cards are treated as a single group of mail. Share equations are
estimated for five categories of workshared cards: presort nonautomation, automation
basic, automation 3-digit, automation 5-digit, and carrier-route presort.
Standard Regular and Standard Nonprofit mail are divided into four categories each
12
for forecasting purposes: basic letters, basic nonletters, presort letters, and presort
13
nonletters. Three of these four categories (basic letter, basic nonletters, and presort
14
nonletters) are divided into nonautomation and automation through share equations.
15
Share equations are used to divide presort letters into three categories: nonautomation,
16
automation 3-digit, and automation 5-digit.
17
For this rate case, share equations are estimated over a sample period of 2000Q1
18
through 2005Q1. The year 2000 is the first year for which volume data are available by
19
Gregorian (or calendar) quarter. Hence, starting in 2000Q1 eliminates any potential
20
problems with mixing Postal and Gregorian quarters. Of course, this is done at a cost of
21
limiting each of these equations to only 21 observations and 2 nominal rate changes.
22
Econometric values are estimated for αt, µt, and σ for each of the share equations
23
estimated for this case. First-Class nonautomation presort, First-Class single-piece
24
cards, and Standard Regular and Nonprofit nonautomation share equations were not
USPS-T-7
323
1
estimated; the share of these categories is instead equal to one minus the share of
2
more highly workshared mail within the relevant mail category.
3
4
5
Because of data limitations, αt was assumed to be constant over the entire sample
period for each of the share equations presented here.
The specification used to model the mean of the user-cost distribution, µt, for the
6
First-Class and Standard share equations presented below modeled µt as function of
7
quarterly dummy variables and a simple linear time trend:
µt = µ0 + µT·Trend + µ1·Q1 + µ2·Q2 + µ3·Q3
8
9
10
(V.15)
As mentioned earlier, no opportunity cost relationships between mail categories
were explicitly modeled for this case.
11
Because of the interrelationships between αt, dt, µt, and σ, it can sometimes be
12
difficult to freely estimate all of these parameters simultaneously. Because of this, in
13
some cases, the share equations were actually estimated using a two-step iterative
14
procedure, whereby αt and/or σ were estimated holding µt constant prior to estimating
15
the other parameters. This procedure was then repeated to ensure convergence. This
16
procedure will lead to unbiased and efficient coefficient estimates, just as if all of the
17
parameters were estimated simultaneously. Because of the nature of the estimation,
18
however, the coefficient estimates do not have true standard error estimates.
19
Nevertheless, t-statistics are presented here, although these numbers should be viewed
20
with caution.
21
The goodness-of-fit measure used to evaluate these equations is mean absolute
22
percentage error. Given a set of fitted shares, ft, and actual shares, st, the mean
23
absolute percentage error is calculated as follows:
24
25
(mean abs. pct. error) = Σ i=1N [(ft / st) - 1]2
where N is the number of observations in the equation.
(V.16)
USPS-T-7
324
1
2
The specific econometric share equations are described in section 6 below.
5. Share Forecast Equations
3
Presort and automation shares are forecasted using a base-share forecasting
4
methodology that parallels the base-volume forecasting methodology used to make
5
volume forecasts.
6
7
The share equation for category i expresses the share of category i at time t as
follows:
sit = αi / [1 + e-(dit-µit)/σi]
8
9
This is also true for the base period, B, i.e.,
siB = αi / [1 + e-(diB-µiB)/σi]
10
11
12
13
(Equation V.17)
(Equation V.18)
Dividing equation V.17 by equation V.18 then produces the forecasting equation,
(V.19):
sit = siB • [1 + e-(diB-µiB)/σi] / [1 + e-(dit-µit)/σi]
(Equation V.19)
14
Hence, the share of category i can be projected given base values for si, di, and µi,
15
and forecasts for di and µi. In this way, equation V.19 is used to make the presort and
16
automation shares described below.
17
18
19
6. Presort and Automation Share Equations by Mail Subclass
a. First-Class Workshared Letters
Share equations are estimated for six categories of First-Class workshared letters:
20
automation basic letters (which includes both mixed-ADC and AADC presort),
21
automation basic flats, automation 3-digit letters, automation 5-digit letters, automation
22
3- and 5-digit flats (combined), and automation carrier-route letters. Shares are of total
23
First-Class workshared letters. Discounts are calculated relative to basic (presort
24
nonautomation) worksharing rates. Results are summarized below.
USPS-T-7
325
i. Automation Basic Letters
1
2
Automation basic letters are letters that are automated but not presorted to the 3-
3
digit level or finer. Both AADC and mixed-ADC letters are treated as “basic” letters
4
here.
5
6
The coefficients for the automation basic letters share equation (t-statistics are in
parentheses) are:
α =
µ0 =
µT =
µ1 =
µ2 =
µ3 =
σ =
7
8
9
10
11
12
13
14
15
16
17
0.120171
0.028706
0.000000
0.000000
0.000000
0.000000
0.010039
(N/A)
(N/A)
(N/A)
(N/A)
(N/A)
(N/A)
(1.134)
Mean Absolute Percentage Error
2.087%
ii. Automation Basic Flats
Automation basic flats are flats that are automated but not presorted to the 3-digit
18
level or finer. The coefficients for the automation basic flats share equation (t-statistics
19
in parentheses) are:
α = 0.008962
µ0 = 0.063163
µT = -0.001168
µ1 = 0.001167
µ2 = 0.000824
µ3 = -0.001168
σ = 0.019514
20
21
22
23
24
25
26
27
28
29
30
(0.596)
(0.895)
(-1.463)
(0.826)
(0.621)
(-0.981)
(0.886)
Mean Absolute Percentage Error
4.449%
iii. Automation 3-Digit Letters
Automation 3-digit letters are letters that are automated and presorted to the 3-digit
31
level. The coefficients for the automation 3-digit letters share equation (t-statistics in
32
parentheses) are:
USPS-T-7
326
α = 0.484246
µ0 = -0.137291
µT = 0.000000
µ1 = 0.016956
µ2 = -0.010541
µ3 = 0.003484
σ = 0.049984
1
2
3
4
5
6
7
8
9
10
11
(0.002)
(-0.007)
(N/A)
(0.337)
(-0.205)
(0.153)
(10.86)
Mean Absolute Percentage Error
0.668%
iv. Automation 5-Digit Letters
Automation 5-digit letters are letters that are automated and presorted to the 5-digit
12
level. The coefficients for the automation 5-digit letters share equation (t-statistics in
13
parentheses) are:
14
15
16
17
18
19
20
21
22
α = 0.372626
µ0 = 0.037024
µT = -0.003339
µ1 = -0.000996
µ2 = 0.002309
µ3 = 0.000136
σ = 0.049610
(0.260)
(0.152)
(-0.901)
(-0.211)
(0.411)
(0.029)
(1192.6)
Mean Absolute Percentage Error
1.728%
USPS-T-7
327
1
2
v. Automation 3- and 5-Digit Flats
Automation 3/5-digit flats are flats that are automated and presorted to the 3-digit or
3
5-digit level. The coefficients for the automation 3/5-digit flats share equation (t-
4
statistics in parentheses) are:
α =
µ0 =
µT =
µ1 =
µ2 =
µ3 =
σ =
0.010413
0.055149
-0.005670
-0.005792
-0.003580
-0.010330
0.013884
(41.19)
(4.633)
(-2.331)
(-0.769)
(-0.544)
(-0.952)
(1.163)
5
6
7
8
9
10
11
12
13
Mean Absolute Percentage Error
14
vi. Automation Carrier-Route Letters
4.276%
15
Automation carrier-route letters are letters that are automated and presorted to the
16
carrier-route level. The coefficients for the automation carrier-route share equation (t-
17
statistics in parentheses) are:
USPS-T-7
328
1
2
3
4
5
6
7
8
9
10
11
α = 0.052645
µ0 = 0.110678
µT = 0.004709
µ1 = -0.016336
µ2 = -0.004601
µ3 = 0.001166
σ = 0.113080
(42.78)
(15.74)
(11.07)
(-2.296)
(-0.616)
(0.157)
(15.90)
Mean Absolute Percentage Error
4.292%
b. First-Class Cards
i. Nonautomation Presort
12
Nonautomation presort First-Class cards are those cards which are presorted but
13
are not automated. The coefficients for the nonautomation presort share equation (t-
14
statistics in parentheses) are:
15
16
17
18
19
20
21
22
23
24
25
α = 0.997374
µ0 = 0.154872
µT = 0.002542
µ1 = 0.000216
µ2 = -0.005320
µ3 = -0.00227
σ = 0.067878
(518636)
(0.198)
(0.164)
(0.030)
(-0.169)
(-0.145)
(0.165)
Mean Absolute Percentage Error
9.136%
ii. Automation Basic
Automation basic cards are cards that are automated but not presorted to the 3-digit
26
level or finer. This includes AADC and mixed-ADC cards. The coefficients for the
27
automation basic share equation (t-statistics in parentheses) are:
USPS-T-7
329
α = 0.200362
µ0 = 0.068613
µT = -0.000394
µ1 = 0.000324
µ2 = -0.005015
µ3 = -0.008403
σ = 0.099951
1
2
3
4
5
6
7
8
9
10
11
(0.095)
(0.033)
(-0.453)
(0.070)
(-0.732)
(-0.920)
(0.119)
Mean Absolute Percentage Error
2.693%
iii. Automation 3-Digit
Automation 3-digit cards are cards that are automated and presorted to the 3-digit
12
level. The coefficients for the automation 3-digit share equation (t-statistics in
13
parentheses) are:
α = 0.393507
µ0 = 0.057481
µT = -0.000803
µ1 = -0.002838
µ2 = 0.001438
µ3 = -0.001033
σ = 0.049998
14
15
16
17
18
19
20
21
22
23
24
(1.028)
(0.182)
(-0.915)
(-0.781)
(0.448)
(-0.359)
(7521.7)
Mean Absolute Percentage Error
3.368%
iv. Automation 5-Digit
Automation 5-digit cards are cards that are automated and presorted to the 5-digit
25
level. The coefficients for the automation 5-digit share equation (t-statistics in
26
parentheses) are:
27
28
29
30
31
32
33
34
35
α =
µ0 =
µT =
µ1 =
µ2 =
µ3 =
σ =
0.419567
0.102548
-0.001331
-0.008332
-0.004976
-0.002303
0.049983
(0.188)
(0.209)
(-1.289)
(-0.962)
(-0.819)
(-0.454)
(0.359)
Mean Absolute Percentage Error
6.930%
USPS-T-7
330
1
2
v. Automation Carrier-Route
Automation carrier-route cards are cards that are automated and presorted to the
3
carrier route. The coefficients for the automation carrier-route share equation (t-
4
statistics in parentheses) are:
5
6
7
8
9
10
11
12
13
14
15
16
α = 0.999539
µ0 = 0.111207
µT = 0.000000
µ1 = -0.003287
µ2 = -0.001628
µ3 = -0.000838
σ = 0.012315
(0.008)
(0.067)
(N/A)
(-2.647)
(-1.448)
(-0.800)
(0.516)
Mean Absolute Percentage Error
7.924%
c. Standard Regular Mail
i. Automation Basic Letters
Automation basic letters are letters that are automated but not presorted to the 3-
17
digit level or finer. The share of automation basic letters is taken as a share of total
18
basic letters. The coefficients for the automation basic letters share equation (t-
19
statistics in parentheses) are:
20
21
22
23
24
25
26
27
28
29
30
31
α = 0.864775
µ0 = -0.135902
µT = -0.012732
µ1 = 0.009034
µ2 = -0.002224
µ3 = 0.004585
σ = 0.113982
(61.66)
(-0.181)
(-0.253)
(2.639)
(-1.854)
(0.234)
(0.252)
Mean Absolute Percentage Error
0.598%
ii. Automation Basic Flats
Automation basic flats are flats that are automated but not presorted to the 3-digit
level or finer. The share of automation basic flats is taken as a share of total basic
USPS-T-7
331
1
nonletters. The coefficients for the automation basic flats share equation (t-statistics in
2
parentheses) are:
α = 0.638494
µ0 = 0.114144
µT = -0.011977
µ1 = 0.014362
µ2 = -0.006498
µ3 = -0.006886
σ = 0.061689
3
4
5
6
7
8
9
10
11
12
13
(1.679)
(0.359)
(-0.619)
(0.487)
(-0.351)
(-0.356)
(1.106)
Mean Absolute Percentage Error
3.637%
iii. Automation 3-Digit Letters
Automation 3-digit letters are letters that are automated and presorted to the 3-digit
14
level. The share of automation 3-digit letters is taken as a share of total presorted
15
letters. The coefficients for the automation 3-digit letters share equation (t-statistics in
16
parentheses) are:
α = 0.477746
µ0 = -0.033834
µT = 0.000000
µ1 = 0.000000
µ2 = 0.000000
µ3 = 0.000000
σ = 0.000159
17
18
19
20
21
22
23
24
25
26
27
(0.000)
(-0.000)
(N/A)
(N/A)
(N/A)
(N/A)
(0.000)
Mean Absolute Percentage Error
2.944%
iv. Automation 5-Digit Letters
Automation 5-digit letters are letters that are automated and presorted to the 5-digit
28
level. The share of automation 5-digit letters is taken as a share of total presorted
29
letters. The coefficients for the automation 5-digit letters share equation (t-statistics in
30
parentheses) are:
USPS-T-7
332
α = 0.519989
µ0 = 0.004850
µT = -0.002450
µ1 = -0.008692
µ2 = 0.002833
µ3 = -0.001333
σ = 0.033946
1
2
3
4
5
6
7
8
9
10
11
(9.130)
(0.070)
(-0.406)
(-1.417)
(0.463)
(-0.225)
(0.722)
Mean Absolute Percentage Error
1.964%
v. Automation 3/5-Digit Flats
Automation 3/5-digit flats are flats that are automated and presorted to at least the 3-
12
digit level. The share of automation 3/5-digit flats is taken as a share of total presorted
13
nonletters. The coefficients for the automation 3/5-digit flats share equation (t-statistics
14
in parentheses) are:
15
16
17
18
19
20
21
22
23
α = 0.934449
µ0 = -0.009810
µT = -0.001578
µ1 = 0.001062
µ2 = -0.000478
µ3 = 0.002599
σ = 0.061689
(43.32)
(-0.573)
(-10.41)
(0.786)
(-0.346)
(1.833)
(1.106)
Mean Absolute Percentage Error
0.403%
USPS-T-7
333
1
2
3
d. Standard Nonprofit Mail
i. Automation Basic Letters
Automation basic letters are letters that are automated but not presorted to the 3-
4
digit level or finer. The share of automation basic letters is taken as a share of total
5
basic letters. The coefficients for the automation basic letters share equation (t-
6
statistics in parentheses) are:
α = 0.815268
µ0 = -0.453994
µT = -0.028615
µ1 = -0.003240
µ2 = 0.016381
µ3 = 0.001621
σ = 0.412779
7
8
9
10
11
12
13
14
15
16
17
(29.72)
(-1.043)
(-1.357)
(-0.253)
(0.797)
(0.123)
(1.062)
Mean Absolute Percentage Error
0.467%
ii. Automation Basic Flats
Automation basic flats are flats that are automated but not presorted to the 3-digit
18
level or finer. The share of automation basic flats is taken as a share of total basic
19
nonletters. The coefficients for the automation basic flats share equation (t-statistics in
20
parentheses) are:
21
22
23
24
25
26
27
28
29
30
31
32
α =
µ0 =
µT =
µ1 =
µ2 =
µ3 =
σ =
0.498374
0.039122
-0.027707
-0.007144
0.028195
-0.009053
0.061689
(0.997)
(0.121)
(-1.673)
(-0.359)
(0.931)
(-0.433)
(1.106)
Mean Absolute Percentage Error
2.249%
iii. Automation 3-Digit Letters
Automation 3-digit letters are letters that are automated and presorted to the 3-digit
level. The share of automation 3-digit letters is taken as a share of total presorted
USPS-T-7
334
1
letters. The coefficients for the automation 3-digit letters share equation (t-statistics in
2
parentheses) are:
α =
µ0 =
µT =
µ1 =
µ2 =
µ3 =
σ =
3
4
5
6
7
8
9
10
11
12
13
0.550000
-0.013195
-0.000507
0.013293
-0.001932
-0.002564
0.017097
(128458)
(-0.489)
(-2.999)
(1.790)
(-0.604)
(-0.740)
(0.837)
Mean Absolute Percentage Error
1.419%
iv. Automation 5-Digit Letters
Automation 5-digit letters are letters that are automated and presorted to the 5-digit
14
level. The share of automation 5-digit letters is taken as a share of total presorted
15
letters. The coefficients for the automation 5-digit letters share equation (t-statistics in
16
parentheses) are:
α = 0.450000
µ0 = 0.032263
µT = -0.002847
µ1 = -0.024050
µ2 = 0.001690
µ3 = 0.005692
σ = 0.045512
17
18
19
20
21
22
23
24
25
26
27
(2391.2)
(1.585)
(-2.859)
(-1.939)
(0.382)
(1.106)
(1.847)
Mean Absolute Percentage Error
3.137%
v. Automation 3/5-Digit Flats
Automation 3/5-digit flats are flats that are automated and presorted to at least the 3-
28
digit level. The share of automation 3/5-digit flats is taken as a share of total presorted
29
nonletters. The coefficients for the automation 3/5-digit flats share equation (t-statistics
30
in parentheses) are:
USPS-T-7
335
1
2
3
4
5
6
7
8
9
10
α = 0.898185
µ0 = -0.092011
µT = -0.004513
µ1 = 0.015820
µ2 = 0.020413
µ3 = 0.010953
σ = 0.049987
(0.123)
(-0.215)
(-0.984)
(0.852)
(0.835)
(0.823)
(12.97)
Mean Absolute Percentage Error
0.617%
7. Final Presort and Automation Share Forecasts
11
Presort and automation shares are forecasted using equation V.19, which was
12
derived above, given base values for si, di, and µi, and forecasts for di and µi. The
13
values used to make these forecasts are presented in Tables V-1 through V-4 below.
14
Forecasted values in Tables V-2 through V-4 are summarized by Government Fiscal
15
Year. The actual forecasting equation is implemented on a quarter-by-quarter basis.
16
The resulting quarterly share forecasts are presented in Tables V-5 and V-6 below.
USPS-T-7
336
Ta ble V-1
Base Va lues used in Fore ca sting
Share
Discount (real) Mean User Cost
First-Class W orkshare d Le tte rs
(Nonautomated Presort)
(Automated)
(Basic Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3- & 5-Digit Flats)
(Carrier-Route Letters)
First-Class Cards
-- Single-Piece
(Nonautomated Presort)
(Automated)
(Basic)
(3-Digit)
(5-Digit)
(Carrier-Route)
Standa rd Regular Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
1
Standa rd Nonprofit Mail
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
Sigma
4.366%
11.120%
0.288%
47.497%
34.255%
1.010%
1.462%
$0.043434
$0.015382
$0.055663
$0.068651
$0.044363
$0.071434
$0.028706
$0.041579
-$0.134826
-$0.024397
-$0.054666
$0.179228
$0.010039
$0.019514
$0.049984
$0.049610
$0.013884
$0.034548
46.192%
7.039%
$0.016699
$0.200055
$0.067878
9.014%
20.594%
15.849%
1.313%
$0.036163
$0.043602
$0.050096
$0.055663
$0.058042
$0.042018
$0.074062
$0.109782
$0.099951
$0.049998
$0.049983
$0.012315
15.810%
84.190%
$0.049100
-$0.368607
$0.113982
57.047%
42.953%
$0.040819
-$0.107232
$0.199986
3.007%
47.406%
49.587%
$0.041747
$0.053807
-$0.033834
-$0.042260
$0.000159
$0.033946
8.845%
91.155%
$0.025048
-$0.038205
$0.017484
25.153%
74.847%
$0.023013
-$0.979727
$0.412779
54.368%
45.632%
$0.038036
-$0.470557
$0.199972
13.892%
50.422%
35.686%
$0.022265
$0.036181
-$0.020428
-$0.024487
$0.017097
$0.045512
12.436%
87.564%
$0.015771
-$0.163797
$0.049987
USPS-T-7
337
Ta ble V-2
Va lue s use d in Fore ca sting: GFY 2005
Real Discount
Forecasted Share
Before-Rates
After-Rates Mean User Cost
Before-Rates
After-Rates
First-Cla ss W orksha re d Le tte rs
(Nonautomated Presort)
(Automated)
(Basic Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3- & 5-Digit Flats)
(Carrier-Route Letters)
First-Cla ss Ca rds
-- Single-Piece
(Nonautomated Presort)
(Automated)
(Basic)
(3-Digit)
(5-Digit)
(Carrier-Route)
Sta nda rd Re gula r Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
1
Sta nda rd Nonprofit Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
3.828%
3.828%
$0.042839
$0.015171
$0.054900
$0.067710
$0.043755
$0.070455
$0.042839
$0.015171
$0.054900
$0.067710
$0.043755
$0.070455
$0.028706
$0.038087
-$0.134816
-$0.034411
-$0.071671
$0.181528
10.974%
0.327%
47.411%
35.086%
1.025%
1.350%
10.974%
0.327%
47.411%
35.086%
1.025%
1.350%
$0.016470
$0.016470
$0.207671
45.993%
6.366%
45.993%
6.366%
$0.035667
$0.043005
$0.049410
$0.054900
$0.035667
$0.043005
$0.049410
$0.054900
$0.056873
$0.039607
$0.070034
$0.109768
9.097%
21.012%
16.317%
1.215%
9.097%
21.012%
16.317%
1.215%
$0.048427
$0.048427
-$0.406796
15.270%
84.730%
15.270%
84.730%
$0.040260
$0.040260
-$0.143113
54.743%
45.257%
54.743%
45.257%
$0.041175
$0.053070
$0.041175
$0.053070
-$0.033834
-$0.049625
2.681%
47.158%
50.161%
2.681%
47.158%
50.161%
$0.024705
$0.024705
-$0.042946
8.347%
91.653%
8.347%
91.653%
$0.022697
$0.022697
-$1.065536
24.081%
75.919%
24.081%
75.919%
$0.037515
$0.037515
-$0.553579
53.059%
46.941%
53.059%
46.941%
$0.021960
$0.035685
$0.021960
$0.035685
-$0.021905
-$0.033121
12.659%
49.943%
37.398%
12.659%
49.943%
37.398%
$0.015555
$0.015555
-$0.177254
12.123%
87.877%
12.123%
87.877%
USPS-T-7
338
Ta ble V-3
Va lue s use d in Fore ca sting: GFY 2006
Real Discount
Forecasted Share
Before-Rates
After-Rates Mean User Cost
Before-Rates
After-Rates
First-Cla ss W orksha re d Le tte rs
(Nonautomated Presort)
(Automated)
(Basic Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3- & 5-Digit Flats)
(Carrier-Route Letters)
First-Cla ss Ca rds
-- Single-Piece
(Nonautomated Presort)
(Automated)
(Basic)
(3-Digit)
(5-Digit)
(Carrier-Route)
Sta nda rd Re gula r Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
1
Sta nda rd Nonprofit Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
3.246%
2.346%
$0.042126
$0.014919
$0.053986
$0.066583
$0.043027
$0.069283
$0.044355
$0.015819
$0.056686
$0.070182
$0.045628
$0.072882
$0.028706
$0.033414
-$0.134816
-$0.047769
-$0.094350
$0.184614
10.851%
0.390%
47.443%
35.871%
1.011%
1.189%
11.320%
0.403%
47.494%
36.111%
1.012%
1.314%
$0.016196
$0.015296
$0.217837
45.386%
5.469%
45.435%
5.398%
$0.035074
$0.042289
$0.048588
$0.053986
$0.035074
$0.042289
$0.048588
$0.054886
$0.055297
$0.036395
$0.064711
$0.109768
9.095%
21.469%
17.421%
1.160%
9.088%
21.442%
17.390%
1.247%
$0.047622
$0.049926
-$0.457725
14.678%
85.322%
14.661%
85.339%
$0.039590
$0.042289
-$0.191021
51.839%
48.161%
51.675%
48.325%
$0.040490
$0.052187
$0.042289
$0.054886
-$0.033834
-$0.059425
1.978%
47.406%
50.616%
1.839%
47.401%
50.761%
$0.024294
$0.026093
-$0.049259
7.779%
92.221%
7.621%
92.379%
$0.022320
$0.023648
-$1.179998
22.778%
77.222%
22.764%
77.236%
$0.036891
$0.038690
-$0.664407
52.243%
47.757%
52.233%
47.767%
$0.021595
$0.035091
$0.022494
$0.036891
-$0.023934
-$0.044511
10.943%
50.648%
38.409%
10.539%
50.830%
38.631%
$0.015296
$0.016196
-$0.195308
11.380%
88.620%
11.356%
88.644%
USPS-T-7
339
Ta ble V-4
Va lue s use d in Fore ca sting: GFY 2007
Real Discount
Forecasted Share
Before-Rates
After-Rates Mean User Cost
Before-Rates
After-Rates
First-Cla ss W orksha re d Le tte rs
(Nonautomated Presort)
(Automated)
(Basic Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3- & 5-Digit Flats)
(Carrier-Route Letters)
First-Cla ss Ca rds
-- Single-Piece
(Nonautomated Presort)
(Automated)
(Basic)
(3-Digit)
(5-Digit)
(Carrier-Route)
Sta nda rd Re gula r Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
1
Sta nda rd Nonprofit Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
2.815%
1.958%
$0.041312
$0.014631
$0.052944
$0.065297
$0.042196
$0.067944
$0.043498
$0.015513
$0.055591
$0.068827
$0.044747
$0.071474
$0.028706
$0.028741
-$0.134816
-$0.061127
-$0.117028
$0.187700
10.670%
0.457%
47.425%
36.573%
1.011%
1.050%
11.155%
0.471%
47.474%
36.771%
1.012%
1.159%
$0.015883
$0.015001
$0.228003
44.807%
4.722%
44.877%
4.660%
$0.034396
$0.041473
$0.047649
$0.052944
$0.034396
$0.041473
$0.047649
$0.053826
$0.053722
$0.033183
$0.059388
$0.109768
9.141%
21.950%
18.313%
1.067%
9.134%
21.915%
18.269%
1.145%
$0.046702
$0.048962
-$0.508654
14.317%
85.683%
14.307%
85.693%
$0.038825
$0.041473
-$0.238928
49.271%
50.729%
49.126%
50.874%
$0.039708
$0.051179
$0.041473
$0.053826
-$0.033834
-$0.069225
1.559%
47.406%
51.035%
1.451%
47.400%
51.149%
$0.023825
$0.025589
-$0.055572
7.391%
92.609%
7.271%
92.729%
$0.021889
$0.023191
-$1.294460
21.810%
78.190%
21.795%
78.205%
$0.036178
$0.037943
-$0.775236
51.641%
48.359%
51.639%
48.361%
$0.021177
$0.034413
$0.022060
$0.036178
-$0.025964
-$0.055900
9.418%
50.916%
39.666%
9.070%
51.082%
39.848%
$0.015001
$0.015883
-$0.213362
10.988%
89.012%
10.968%
100.000%
USPS-T-7
340
First-Cla ss W orksha re d Le tte rs
(Nonautomated Presort)
(Automated)
(Basic Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3- & 5-Digit Flats)
(Carrier-Route Letters)
1
First-Cla ss Ca rds
-- Single-Piece
(Nonautomated Presort)
(Automated)
(Basic)
(3-Digit)
(5-Digit)
(Carrier-Route)
2005Q1
2005Q2
Ta ble V-5
Qua rte rly Be fore -Ra te s Pre sort a nd Autom a tion Sha re Fore ca sts
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
4.115%
3.858%
3.759%
3.507%
3.581%
3.186%
3.142%
2.952%
3.110%
2.692%
2.729%
2.583%
10.949%
0.285%
46.954%
35.093%
1.064%
1.538%
11.013%
0.312%
47.723%
34.747%
1.011%
1.336%
10.973%
0.361%
47.457%
35.171%
1.011%
1.268%
10.941%
0.351%
47.525%
35.416%
1.011%
1.249%
10.899%
0.350%
47.118%
35.705%
1.011%
1.336%
10.854%
0.370%
47.708%
35.685%
1.011%
1.185%
10.820%
0.425%
47.438%
36.037%
1.011%
1.127%
10.776%
0.414%
47.506%
36.233%
1.011%
1.107%
10.726%
0.413%
47.090%
36.467%
1.011%
1.183%
10.674%
0.434%
47.691%
36.449%
1.011%
1.049%
10.626%
0.494%
47.415%
36.731%
1.011%
0.994%
10.575%
0.482%
47.484%
36.890%
1.011%
0.976%
45.288%
6.631%
45.768%
6.747%
45.877%
6.237%
47.089%
5.828%
44.432%
5.602%
45.317%
5.835%
45.340%
5.391%
46.526%
5.034%
43.853%
4.836%
44.771%
5.038%
44.766%
4.652%
45.913%
4.340%
9.047%
21.408%
16.283%
1.344%
9.128%
20.482%
16.609%
1.266%
9.311%
21.108%
16.301%
1.166%
8.903%
21.031%
16.075%
1.073%
8.899%
21.729%
17.969%
1.370%
9.177%
20.986%
17.510%
1.175%
9.361%
21.617%
17.205%
1.086%
8.951%
21.531%
16.964%
0.995%
8.946%
22.221%
18.878%
1.267%
9.223%
21.475%
18.409%
1.083%
9.405%
22.094%
18.087%
0.996%
8.994%
22.004%
17.838%
0.910%
2007Q3
2007Q4
USPS-T-7
341
Sta nda rd Re gula r Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
1
Sta nda rd Nonprofit Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
Ta ble V-5 (continue d)
Qua rte rly Be fore -Ra te s Pre sort a nd Autom a tion Sha re Fore ca sts
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
2005Q1
2005Q2
2007Q3
2007Q4
15.782%
84.218%
15.232%
84.768%
15.156%
84.844%
14.950%
85.050%
14.911%
85.089%
14.677%
85.323%
14.627%
85.373%
14.493%
85.507%
14.468%
85.532%
14.316%
85.684%
14.284%
85.716%
14.197%
85.803%
57.033%
42.967%
54.623%
45.377%
53.846%
46.154%
53.536%
46.464%
53.696%
46.304%
51.732%
48.268%
51.038%
48.962%
50.766%
49.234%
50.909%
49.091%
49.178%
50.822%
48.574%
51.426%
48.338%
51.662%
2.924%
46.406%
50.670%
2.937%
47.406%
49.657%
2.469%
47.406%
50.125%
2.407%
47.406%
50.187%
1.789%
47.406%
50.805%
2.306%
47.406%
50.288%
1.935%
47.406%
50.659%
1.890%
47.406%
50.704%
1.407%
47.406%
51.187%
1.813%
47.406%
50.781%
1.525%
47.406%
51.069%
1.491%
47.406%
51.103%
8.686%
91.314%
8.255%
91.745%
8.430%
91.570%
8.013%
91.987%
7.977%
92.023%
7.731%
92.269%
7.856%
92.144%
7.557%
92.443%
7.532%
92.468%
7.355%
92.645%
7.447%
92.553%
7.232%
92.768%
24.650%
75.350%
24.343%
75.657%
23.804%
76.196%
23.457%
76.543%
23.114%
76.886%
23.023%
76.977%
22.600%
77.400%
22.329%
77.671%
22.062%
77.938%
21.990%
78.010%
21.662%
78.338%
21.451%
78.549%
52.856%
47.144%
53.708%
46.292%
52.934%
47.066%
52.751%
47.249%
52.448%
47.552%
52.512%
47.488%
52.045%
47.955%
51.936%
48.064%
51.756%
48.244%
51.794%
48.206%
51.518%
48.482%
51.454%
48.546%
13.209%
46.336%
40.455%
12.733%
51.430%
35.837%
12.770%
51.606%
35.624%
11.830%
51.236%
36.934%
11.897%
47.940%
40.163%
10.835%
51.705%
37.460%
10.851%
51.869%
37.280%
10.075%
51.526%
38.399%
10.412%
48.478%
41.110%
9.206%
51.953%
38.842%
9.217%
52.100%
38.683%
8.589%
51.782%
39.629%
12.909%
87.091%
12.326%
87.674%
11.778%
88.222%
11.320%
88.680%
11.645%
88.355%
11.649%
88.351%
11.260%
88.740%
10.936%
89.064%
11.166%
88.834%
11.170%
88.830%
10.895%
89.105%
10.666%
89.334%
USPS-T-7
342
First-Cla ss W orksha re d Le tte rs
(Nonautomated Presort)
(Automated)
(Basic Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3- & 5-Digit Flats)
(Carrier-Route Letters)
1
First-Cla ss Ca rds
-- Single-Piece
(Nonautomated Presort)
(Automated)
(Basic)
(3-Digit)
(5-Digit)
(Carrier-Route)
2005Q1
2005Q2
Ta ble V-6
Qua rte rly Afte r-Ra te s Pre sort and Automa tion Sha re Fore ca sts
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
4.115%
3.858%
3.759%
3.507%
2.647%
2.294%
2.260%
2.086%
2.227%
1.849%
1.892%
1.759%
10.949%
0.285%
46.954%
35.093%
1.064%
1.538%
11.013%
0.312%
47.723%
34.747%
1.011%
1.336%
10.973%
0.361%
47.457%
35.171%
1.011%
1.268%
10.941%
0.351%
47.525%
35.416%
1.011%
1.249%
11.362%
0.363%
47.191%
35.949%
1.011%
1.477%
11.321%
0.383%
47.752%
35.929%
1.011%
1.311%
11.290%
0.438%
47.495%
36.259%
1.011%
1.246%
11.249%
0.427%
47.560%
36.443%
1.011%
1.224%
11.203%
0.426%
47.164%
36.662%
1.011%
1.307%
11.155%
0.448%
47.735%
36.644%
1.011%
1.158%
11.111%
0.508%
47.472%
36.909%
1.011%
1.097%
11.063%
0.496%
47.537%
37.057%
1.011%
1.077%
45.288%
6.631%
45.768%
6.747%
45.877%
6.237%
47.089%
5.828%
44.400%
5.531%
45.302%
5.762%
45.326%
5.324%
46.515%
4.971%
43.820%
4.776%
44.753%
4.976%
44.751%
4.594%
45.900%
4.287%
9.047%
21.408%
16.283%
1.344%
9.128%
20.482%
16.609%
1.266%
9.311%
21.108%
16.301%
1.166%
8.903%
21.031%
16.075%
1.073%
8.899%
21.729%
17.969%
1.474%
9.177%
20.986%
17.510%
1.263%
9.361%
21.617%
17.205%
1.167%
8.951%
21.531%
16.964%
1.069%
8.946%
22.221%
18.878%
1.361%
9.223%
21.475%
18.409%
1.163%
9.405%
22.094%
18.087%
1.069%
8.994%
22.004%
17.838%
0.977%
2007Q3
2007Q4
USPS-T-7
343
Sta nda rd Re gula r Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
1
Sta nda rd Nonprofit Ma il
Basic Letters
(Nonautomated)
(Automated)
Basic Nonletters
(Nonautomated)
(Automated)
Presort Letters
(Nonautomated)
(3-Digit Automation)
(5-Digit Automation)
Presort Nonletters
(Nonautomated)
(Automated)
Table V-6 (continue d)
Qua rte rly Afte r-Ra te s Pre sort and Automa tion Sha re Fore ca sts
2005Q3
2005Q4
2006Q1
2006Q2
2006Q3
2006Q4
2007Q1
2007Q2
2005Q1
2005Q2
2007Q3
2007Q4
15.782%
84.218%
15.232%
84.768%
15.156%
84.844%
14.950%
85.050%
14.886%
85.114%
14.657%
85.343%
14.608%
85.392%
14.477%
85.523%
14.452%
85.548%
14.303%
85.697%
14.272%
85.728%
14.187%
85.813%
57.033%
42.967%
54.623%
45.377%
53.846%
46.154%
53.536%
46.464%
53.527%
46.473%
51.577%
48.423%
50.888%
49.112%
50.618%
49.382%
50.761%
49.239%
49.045%
50.955%
48.446%
51.554%
48.213%
51.787%
2.924%
46.406%
50.670%
2.937%
47.406%
49.657%
2.469%
47.406%
50.125%
2.407%
47.406%
50.187%
1.661%
47.406%
50.933%
2.142%
47.406%
50.452%
1.799%
47.406%
50.795%
1.757%
47.406%
50.837%
1.309%
47.406%
51.285%
1.687%
47.406%
50.907%
1.420%
47.406%
51.174%
1.388%
47.406%
51.206%
8.686%
91.314%
8.255%
91.745%
8.430%
91.570%
8.013%
91.987%
7.824%
92.176%
7.602%
92.398%
7.715%
92.285%
7.446%
92.554%
7.423%
92.577%
7.264%
92.736%
7.347%
92.653%
7.153%
92.847%
24.650%
75.350%
24.343%
75.657%
23.804%
76.196%
23.457%
76.543%
23.100%
76.900%
23.009%
76.991%
22.588%
77.412%
22.317%
77.683%
22.051%
77.949%
21.980%
78.020%
21.652%
78.348%
21.442%
78.558%
52.856%
47.144%
53.708%
46.292%
52.934%
47.066%
52.751%
47.249%
52.433%
47.567%
52.497%
47.503%
52.034%
47.966%
51.926%
48.074%
51.747%
48.253%
51.785%
48.215%
51.512%
48.488%
51.448%
48.552%
13.209%
46.336%
40.455%
12.733%
51.430%
35.837%
12.770%
51.606%
35.624%
11.830%
51.236%
36.934%
11.422%
48.242%
40.336%
10.447%
51.845%
37.708%
10.467%
52.001%
37.532%
9.707%
51.673%
38.620%
9.997%
48.754%
41.249%
8.873%
52.079%
39.048%
8.888%
52.218%
38.893%
8.274%
51.915%
39.811%
12.909%
87.091%
12.326%
87.674%
11.778%
88.222%
11.320%
88.680%
11.616%
88.384%
11.621%
88.379%
11.238%
88.762%
10.920%
89.080%
11.146%
88.854%
11.150%
88.850%
10.879%
89.121%
10.655%
89.345%
USPS-T-7
344
1
C. Standard Regular Letters versus Non-Letters
2
In most cases, the individual categories of mail within a particular subclass are
3
assumed to grow in proportion, being equally affected by factors through the forecast
4
period. One exception to this assumption is the shares of First-Class and Standard Mail
5
that are automated. This was dealt with in the previous section. Another exception in
6
this case is Standard Regular letters versus Standard Regular non-letters.
7
Table V-7 below shows the annual growth rate for Standard Regular letters versus
8
Standard Regular non-letters by quarter since 1998. Since 1998, Standard Regular
9
letter volume has grown more rapidly than Standard Regular non-letter volume for 29
10
consecutive quarters. On average, the Standard Regular letter growth rate over this
11
time period has been more than 10 percent greater than the Standard Regular non-
12
letter growth rate.
13
Such a dramatic difference in growth rates, if it continues through the forecast
14
period, could have a significant impact on the revenues and costs associated with
15
Standard Regular mail volume. Therefore, it was decided that some recognition of this
16
historical difference in growth rates should be reflected in the volume forecasts used in
17
this case.
USPS-T-7
345
1
2
1998PQ1
1998PQ2
1998PQ3
1998PQ4
1999PQ1
1999PQ2
1999PQ3
1999PQ4
2000PQ1
2000PQ2
2000PQ3
2000PQ4
2001GQ1
2001GQ2
2001GQ3
2001GQ4
2002GQ1
2002GQ2
2002GQ3
2002GQ4
2003GQ1
2003GQ2
2003GQ3
2003GQ4
2004GQ1
2004GQ2
2004GQ3
2004GQ4
2005GQ1
Ta ble V-7
Pe rce nta ge Grow th Ra te
over Sa me Pe riod Last Ye a r (SPLY)
Standard Regular
Letters
Non-Letters
9.42%
5.50%
6.32%
4.40%
11.02%
6.83%
11.38%
5.93%
7.83%
6.70%
13.97%
2.34%
19.91%
3.34%
19.22%
0.98%
22.99%
0.96%
19.57%
-0.54%
13.64%
3.69%
10.15%
0.26%
12.04%
0.05%
11.28%
-3.26%
9.45%
-6.18%
0.81%
-10.41%
0.71%
-14.03%
-4.50%
-14.80%
0.76%
-8.79%
9.38%
-7.36%
12.96%
-0.18%
10.78%
2.30%
6.40%
-1.49%
8.11%
2.95%
5.49%
-1.11%
13.48%
1.25%
13.66%
2.49%
13.91%
2.95%
9.56%
3.98%
USPS-T-7
346
1
To recognize this difference in trends, a simple share equation was modeled, which
2
expressed the share of Standard Regular mail that was letter-sized as a function of
3
quarterly dummy variables (to reflect differences in the seasonal pattern of Standard
4
Regular letters and non-letters), a time trend, and a time trend squared. This equation
5
was estimated using data from 2000GQ1 through 2005GQ1, so as to avoid having to
6
deal with the problem of combining data for Postal and Gregorian quarters.
7
The results of this equation were as follows (t-statistics in parentheses):
8
9
10
11
Letters Share = 0.6406 + 0.002•Q1 + 0.005•Q2 + 0.027•Q3 + 0.008431•(Trend) – 0.000145•(Trend)2
(212.4) (0.967)
(1.892)
(10.83)
(14.15)
(-4.791)
(Equation V.20)
This equation has an adjusted-R2 of 0.988.
12
Equation V.20 is then used to project the share of Standard Regular mail that will be
13
letter-sized through the forecast period. The share of Standard Regular mail that will be
14
non-letter-sized will, of course, be equal to one minus the share that is letter-sized.
15
The combined impact of the trend and trend-squared terms in Equation V.20 is that
16
the share of Standard Regular mail that is letter-sized is increasing but at a decreasing
17
rate. In fact, mathematically, at some point the negative impact of the trend-squared
18
term will become larger in absolute value than the positive impact of the trend term. At
19
this point, the combined effect of these variables will be to reduce the share of Standard
20
Regular letters. It seems quite reasonable to think that the actual rate at which the
21
letters share of Standard Regular mail volume is increasing will decline over time. It
22
seems much more speculative, however, to posit that the current positive trend will
23
reverse itself and become a net negative impact at some point in the forecast period.
24
Hence, the combined impact of the the trend and trend-squared variables is restricted to
25
be non-negative throughout the forecast period. In this case, this restriction becomes
USPS-T-7
347
1
binding starting in 2007Q3. That is, from 2007Q3 onward, the combined impact of the
2
trend and trend-squared terms is constrained to be exactly equal to zero.
3
The share of Standard Regular mail volume that is expected to be letter-sized
4
resulting from applying equation V.20, subject to the trend restriction described in the
5
preceding paragraph, is summarized in Table V-8 below.
Ta ble V-8
Fore casted Share s of Sta ndard Re gula r Ma il Volum e
Letters
Non-Letters
6
Actual
2004GQ1
2004GQ2
2004GQ3
2004GQ4
2005GQ1
74.06%
74.81%
77.14%
75.15%
75.05%
25.94%
25.19%
22.86%
24.85%
24.95%
Forecast
2005GQ2
2005GQ3
2005GQ4
2006GQ1
2006GQ2
2006GQ3
2006GQ4
2007GQ1
2007GQ2
2007GQ3
2007GQ4
75.86%
78.35%
75.79%
76.20%
76.57%
78.94%
76.27%
76.56%
76.81%
78.94%
76.27%
24.14%
21.65%
24.21%
23.80%
23.43%
21.06%
23.73%
23.44%
23.19%
21.06%
23.73%
USPS-T-7
348
2005GQ1
FIRST-CLASS MAIL
First-Class Letters & Flats
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC Letters)
(Mixed-ADC Flats)
(AADC Letters)
(AADC Flats)
(3-Digit Letters)
(5-Digit Letters)
(3-Digit Flats)
(5-Digit Flats)
(Carrier-Route Letters)
First-Class Cards
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC)
(AADC)
(3-Digit)
(5-Digit)
(Carrier-Route)
TOTAL FIRST-CLASS MAIL
Priority Mail
Express Mail
Mailgrams
PERIODICAL MAIL
W ithin County
Nonprofit
Classroom
Regular Rate
TOTAL PERIODICAL MAIL
STANDARD MAIL
Regular Rate Bulk
Regular
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Enhanced Carrier-Route
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
Atta chm e nt A
R2005-1 Volume Fore ca st
(m illions of pie ce s)
2005GQ2
2005GQ3
2005GQ4
pa ge 1 of 10
2005GFY
25,003.717
12,366.264
12,637.453
520.084
12,117.369
728.126
10.475
655.528
25.595
5,933.832
4,434.909
61.444
73.079
194.380
1,424.490
645.118
779.372
94.451
684.921
71.486
57.383
304.952
231.953
19.148
26,428.207
23,305.434
10,803.560
12,501.874
483.953
12,017.922
723.276
10.925
654.085
28.093
5,966.155
4,342.138
61.290
65.069
166.890
1,355.859
620.549
735.310
91.475
643.835
70.455
53.312
277.709
225.195
17.163
24,661.293
22,685.907
10,767.342
11,918.566
450.055
11,468.510
687.182
12.058
621.444
31.006
5,656.041
4,189.292
58.436
62.040
151.010
1,357.016
622.558
734.458
84.640
649.818
71.923
54.423
286.443
221.206
15.823
24,042.924
22,137.272
10,327.722
11,809.550
416.514
11,393.035
679.003
11.631
614.047
29.908
5,612.390
4,179.372
57.895
61.465
147.323
1,336.344
629.275
707.069
77.878
629.191
67.728
51.249
281.053
214.817
14.345
23,473.616
93,132.331
44,264.888
48,867.443
1,870.606
46,996.836
2,817.588
45.090
2,545.105
114.602
23,168.419
17,145.712
239.065
261.652
659.603
5,473.710
2,517.501
2,956.209
348.444
2,607.765
281.592
216.367
1,150.156
893.171
66.479
98,606.041
239.616
13.517
0.434
209.179
13.569
0.422
205.693
13.852
0.440
191.769
13.173
0.310
846.257
54.111
1.607
203.448
498.549
15.972
1,622.199
2,340.169
186.711
483.534
17.206
1,645.334
2,332.784
189.522
462.188
16.447
1,636.136
2,304.293
186.873
430.545
15.321
1,556.153
2,188.891
766.554
1,874.816
64.945
6,459.821
9,166.137
22,446.154
13,502.675
842.680
191.874
114.893
260.762
275.150
12,659.995
480.691
543.236
86.556
4,138.413
4,518.593
2,892.508
8,943.479
8,422.388
565.555
3,719.424
110.781
488.034
796.881
2,741.712
521.091
20,959.109
13,337.138
819.556
195.705
116.121
259.347
248.383
12,517.582
512.296
576.803
96.465
4,186.381
4,385.184
2,760.452
7,621.971
7,146.833
521.233
2,892.624
115.488
428.800
716.029
2,472.660
475.139
21,169.130
13,538.883
767.964
204.153
104.243
228.590
230.978
12,770.918
537.568
605.257
89.353
4,388.963
4,640.693
2,509.084
7,630.247
7,154.593
521.798
2,895.765
115.614
429.265
716.806
2,475.344
475.655
21,674.589
13,573.273
773.689
195.290
116.167
216.140
246.092
12,799.584
522.614
588.420
100.821
4,256.500
4,506.184
2,825.045
8,101.316
7,596.296
554.013
3,074.540
122.751
455.767
761.060
2,628.165
505.020
86,248.982
53,951.968
3,203.889
787.022
451.424
964.840
1,000.603
50,748.080
2,053.169
2,313.716
373.196
16,970.257
18,050.654
10,987.089
32,297.014
30,320.109
2,162.599
12,582.353
464.634
1,801.867
2,990.776
10,317.881
1,976.905
USPS-T-7
349
2005GQ1
STANDARD MAIL
Nonprofit Rate Bulk
Nonprofit
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Nonprofit ECR
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
TOTAL STANDARD MAIL
Atta chm e nt A
R2005-1 Volume Fore ca st
(m illions of pie ce s)
2005GQ2
2005GQ3
2005GQ4
pa ge 2 of 10
2005GFY
4,508.217
3,422.092
536.509
139.471
28.436
300.080
68.521
2,885.583
219.349
206.979
25.363
1,052.618
919.011
462.263
1,086.125
1,023.771
148.404
551.416
17.067
30.319
148.293
128.272
62.354
26,954.372
3,697.884
2,947.185
454.013
127.920
25.701
243.813
56.580
2,493.172
202.171
195.405
22.152
984.781
686.214
402.450
750.699
700.599
83.662
310.460
17.467
13.656
159.256
116.098
50.100
24,656.993
3,339.290
2,661.388
405.461
112.959
22.874
220.807
48.822
2,255.927
183.866
177.712
20.338
892.332
615.985
365.694
677.902
632.660
75.549
280.354
15.773
12.332
143.812
104.840
45.242
24,508.420
3,691.851
2,942.377
426.297
123.062
25.201
226.157
51.876
2,516.079
204.205
197.371
22.573
979.470
706.056
406.405
749.474
699.456
83.526
309.953
17.438
13.634
158.996
115.908
50.018
25,366.440
15,237.242
11,973.042
1,822.280
503.412
102.212
990.857
225.799
10,150.761
809.590
777.467
90.427
3,909.201
2,927.266
1,636.811
3,264.201
3,056.487
391.141
1,452.183
67.746
69.942
610.357
465.118
207.714
101,486.224
PACKAGE SERVICES
Parcel Post
Non-Destination Entry
(Inter-BMC)
(Intra-BMC)
Destination Entry
(DBMC)
(DSCF)
(DDU)
Bound Printed Matter
Media Mail
Library Rate
TOTAL PACKAGE SERVICES MAIL
123.359
35.497
24.926
10.571
87.863
26.029
0.998
60.835
140.177
46.947
4.202
314.685
85.085
28.307
19.893
8.415
56.778
18.878
0.627
37.273
143.128
46.264
4.091
278.568
79.759
25.227
17.728
7.499
54.532
18.131
0.602
35.798
123.699
46.454
4.108
254.020
77.108
24.211
17.014
7.197
52.897
17.587
0.584
34.725
164.582
44.660
3.950
290.300
365.311
113.242
79.561
33.682
252.069
80.626
2.812
168.631
571.586
184.324
16.352
1,137.573
Postal Penalty
Free-for-the-Blind
199.738
19.177
154.178
16.269
159.277
19.278
152.911
18.054
666.104
72.778
56,509.915
52,323.257
51,508.196
51,695.464
212,036.831
1.129
13.262
63.250
0.367
57.403
45.646
191.851
2.244
375.152
1.138
10.744
70.532
0.447
62.632
46.979
160.765
2.208
355.445
1.090
9.883
72.167
0.448
64.417
46.398
162.801
2.236
359.440
1.055
8.659
66.664
0.447
59.027
44.837
146.995
2.019
329.704
4.412
42.547
272.612
1.710
243.480
183.861
662.412
8.707
1,419.741
3.878
20.836
3.916
22.242
4.105
27.489
3.864
14.199
15.762
84.765
TOTAL DOMESTIC MAIL
DOMESTIC SPECIAL SERVICES
Registry
Insurance
Certified
Collect-on-Delivery
Return Receipts
Money Orders
Delivery Confirmation
Signature Confirmation
TOTAL SPECIAL SERVICES
Post Office Boxes
Stamped Cards
USPS-T-7
350
2006GQ1
FIRST-CLASS MAIL
First-Class Letters & Flats
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC Letters)
(Mixed-ADC Flats)
(AADC Letters)
(AADC Flats)
(3-Digit Letters)
(5-Digit Letters)
(3-Digit Flats)
(5-Digit Flats)
(Carrier-Route Letters)
First-Class Cards
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC)
(AADC)
(3-Digit)
(5-Digit)
(Carrier-Route)
TOTAL FIRST-CLASS MAIL
Priority Mail
Express Mail
Mailgrams
PERIODICAL MAIL
W ithin County
Nonprofit
Classroom
Regular Rate
TOTAL PERIODICAL MAIL
STANDARD MAIL
Regular Rate Bulk
Regular
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Enhanced Carrier-Route
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
Atta chm e nt A
R2005-1 Volum e Fore ca st: Be fore -Ra te s
(m illions of pie ce s)
2006GQ2
2006GQ3
2006GQ4
pa ge 3 of 10
2006GFY
24,096.627
11,834.189
12,262.437
442.247
11,820.190
702.445
12.051
635.247
30.987
5,777.582
4,374.340
60.117
63.824
163.598
1,449.024
643.833
805.191
81.168
724.023
73.402
55.542
314.853
260.368
19.858
25,545.650
23,002.108
10,554.668
12,447.439
399.772
12,047.667
710.238
12.925
642.295
33.235
5,938.410
4,437.377
61.022
64.785
147.379
1,371.169
621.377
749.792
80.003
669.789
71.631
54.202
287.749
240.095
16.111
24,373.277
22,367.229
10,514.322
11,852.907
375.821
11,477.085
674.323
14.135
609.815
36.346
5,622.795
4,266.499
58.107
61.689
133.377
1,372.554
622.313
750.241
74.001
676.241
73.140
55.344
296.710
236.142
14.905
23,739.783
21,858.194
10,084.562
11,773.631
351.178
11,422.453
667.160
13.682
603.338
35.182
5,593.318
4,260.630
57.715
61.274
130.155
1,351.609
628.849
722.759
68.034
654.725
68.870
52.113
291.016
229.281
13.445
23,209.802
91,324.157
42,987.742
48,336.414
1,569.019
46,767.396
2,754.166
52.793
2,490.694
135.750
22,932.106
17,338.845
236.961
251.572
574.508
5,544.356
2,516.372
3,027.984
303.206
2,724.778
287.043
217.201
1,190.328
965.886
64.320
96,868.513
237.853
13.416
0.343
206.542
13.215
0.366
205.459
13.493
0.382
192.850
12.821
0.269
842.705
52.945
1.359
193.024
470.431
16.740
1,592.819
2,273.014
182.684
480.062
17.083
1,646.026
2,325.854
185.160
457.784
16.290
1,639.564
2,298.798
182.418
423.527
15.071
1,559.938
2,180.953
743.285
1,831.804
65.183
6,438.348
9,078.621
23,638.366
14,451.282
758.782
208.515
121.943
171.937
256.387
13,692.500
559.695
630.170
105.156
4,556.452
4,883.165
2,957.861
9,187.084
8,614.379
628.264
3,486.601
139.203
516.851
863.060
2,980.401
572.705
21,958.882
14,059.300
767.855
200.641
112.533
216.658
238.024
13,291.445
548.644
617.728
104.997
4,454.245
4,725.052
2,840.780
7,899.582
7,407.138
540.217
2,997.980
119.695
444.418
742.108
2,562.720
492.444
22,095.815
14,229.675
719.461
208.647
101.012
189.753
220.049
13,510.214
572.814
644.941
96.904
4,647.761
4,966.650
2,581.143
7,866.140
7,375.781
537.930
2,985.289
119.188
442.536
738.967
2,551.871
490.360
22,621.616
14,245.516
731.304
199.953
113.339
179.220
238.793
13,514.212
554.905
624.777
109.921
4,495.376
4,808.122
2,921.111
8,376.100
7,853.950
572.804
3,178.824
126.915
471.226
786.873
2,717.308
522.150
90,314.679
56,985.773
2,977.402
817.754
448.827
757.568
953.252
54,008.371
2,236.058
2,517.617
416.978
18,153.833
19,382.990
11,300.895
33,328.906
31,251.248
2,279.215
12,648.694
505.000
1,875.031
3,131.008
10,812.300
2,077.658
USPS-T-7
351
2006GQ1
STANDARD MAIL
Nonprofit Rate Bulk
Nonprofit
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Nonprofit ECR
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
TOTAL STANDARD MAIL
Atta chm e nt A
R2005-1 Volum e Fore ca st: Be fore -Ra te s
(m illions of pie ce s)
2006GQ2
2006GQ3
2006GQ4
pa ge 4 of 10
2006GFY
4,318.043
3,441.446
499.562
141.835
29.307
266.005
62.416
2,941.884
239.909
231.880
26.571
1,071.900
898.028
473.596
876.596
818.094
97.693
362.526
20.396
15.947
185.964
135.568
58.502
27,956.408
3,769.963
3,004.631
414.985
123.340
25.618
211.510
54.517
2,589.646
209.708
202.690
23.167
1,009.351
731.270
413.460
765.332
714.255
85.293
316.511
17.807
13.923
162.360
118.361
51.077
25,728.846
3,506.346
2,794.531
382.243
112.611
23.615
197.008
49.010
2,412.287
196.114
189.551
21.759
941.752
676.867
386.244
711.816
664.310
79.329
294.379
16.562
12.949
151.007
110.084
47.505
25,602.161
3,908.377
3,114.946
407.242
124.015
26.267
203.903
53.057
2,707.704
219.367
212.025
24.309
1,042.785
777.116
432.102
793.431
740.479
88.425
328.132
18.461
14.434
168.321
122.706
52.952
26,529.993
15,502.729
12,355.554
1,704.034
501.801
104.807
878.426
218.999
10,651.521
865.098
836.146
95.806
4,065.787
3,083.281
1,705.402
3,147.175
2,937.138
350.740
1,301.549
73.227
57.252
667.652
486.720
210.036
105,817.408
PACKAGE SERVICES
Parcel Post
Non-Destination Entry
(Inter-BMC)
(Intra-BMC)
Destination Entry
(DBMC)
(DSCF)
(DDU)
Bound Printed Matter
Media Mail
Library Rate
TOTAL PACKAGE SERVICES MAIL
116.599
36.197
25.437
10.760
80.402
26.733
0.888
52.781
150.168
51.357
4.542
322.665
82.878
29.275
20.573
8.702
53.603
17.822
0.592
35.188
148.627
47.589
4.209
283.303
78.397
25.970
18.250
7.720
52.427
17.431
0.579
34.416
128.404
47.753
4.223
258.777
76.188
24.767
17.405
7.362
51.421
17.097
0.568
33.756
171.139
45.944
4.063
297.334
354.061
116.209
81.665
34.545
237.852
79.083
2.627
156.142
598.339
192.642
17.037
1,162.079
Postal Penalty
Free-for-the-Blind
170.402
19.977
164.032
16.796
169.440
19.903
162.664
18.642
666.538
75.317
56,539.729
53,112.231
52,308.196
52,605.328
214,565.484
1.087
11.789
67.238
0.415
59.833
45.472
203.168
2.790
391.791
1.007
8.842
72.421
0.426
64.355
46.349
175.051
2.404
370.854
0.963
8.139
74.028
0.426
66.125
45.664
176.723
2.427
374.495
0.934
7.134
68.457
0.426
60.661
44.083
159.260
2.187
343.141
3.990
35.903
282.145
1.693
250.973
181.567
714.201
9.809
1,480.281
3.851
26.260
4.032
22.319
4.236
27.547
3.980
14.225
16.100
90.352
TOTAL DOMESTIC MAIL
DOMESTIC SPECIAL SERVICES
Registry
Insurance
Certified
Collect-on-Delivery
Return Receipts
Money Orders
Delivery Confirmation
Signature Confirmation
TOTAL SPECIAL SERVICES
Post Office Boxes
Stamped Cards
USPS-T-7
352
2007GQ1
FIRST-CLASS MAIL
First-Class Letters & Flats
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC Letters)
(Mixed-ADC Flats)
(AADC Letters)
(AADC Flats)
(3-Digit Letters)
(5-Digit Letters)
(3-Digit Flats)
(5-Digit Flats)
(Carrier-Route Letters)
First-Class Cards
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC)
(AADC)
(3-Digit)
(5-Digit)
(Carrier-Route)
TOTAL FIRST-CLASS MAIL
Priority Mail
Express Mail
Mailgrams
PERIODICAL MAIL
W ithin County
Nonprofit
Classroom
Regular Rate
TOTAL PERIODICAL MAIL
STANDARD MAIL
Regular Rate Bulk
Regular
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Enhanced Carrier-Route
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
Atta chm e nt A
R2005-1 Volum e Fore ca st: Be fore -Ra te s
(m illions of pie ce s)
2007GQ2
2007GQ3
2007GQ4
pa ge 5 of 10
2007GFY
23,561.883
11,481.854
12,080.029
379.913
11,700.116
681.441
14.004
616.253
36.009
5,688.552
4,399.088
59.214
62.865
142.690
1,447.499
634.774
812.725
70.003
742.721
73.711
55.776
321.643
273.254
18.338
25,009.382
23,039.984
10,460.385
12,579.599
342.824
12,236.775
706.289
15.352
638.723
39.476
5,999.732
4,578.409
61.663
65.465
131.667
1,406.771
629.825
776.946
70.877
706.070
73.861
55.889
302.109
258.970
15.241
24,446.755
21,992.999
10,176.345
11,816.654
326.762
11,489.892
660.579
16.398
597.386
42.165
5,603.151
4,333.562
57.920
61.492
117.238
1,389.962
622.233
767.729
64.655
703.074
74.416
56.309
307.105
251.403
13.841
23,382.961
21,307.901
9,741.953
11,565.948
303.034
11,262.914
643.525
15.665
581.963
40.280
5,492.394
4,259.586
56.690
60.185
112.625
1,346.191
618.075
728.116
58.429
669.687
68.923
52.153
296.222
240.135
12.254
22,654.092
89,902.767
41,860.536
48,042.231
1,352.534
46,689.697
2,691.834
61.419
2,434.325
157.930
22,783.829
17,570.645
235.487
250.007
504.221
5,590.422
2,504.907
3,085.515
263.964
2,821.552
290.910
220.127
1,227.079
1,023.762
59.674
95,493.189
239.359
12.566
0.299
213.816
13.067
0.321
210.229
13.042
0.331
194.860
12.314
0.228
858.264
50.990
1.179
181.358
455.327
16.202
1,624.922
2,277.809
180.453
472.560
16.816
1,672.070
2,341.899
180.356
443.804
15.792
1,642.904
2,282.855
175.360
406.533
14.466
1,546.121
2,142.480
717.527
1,778.224
63.276
6,486.016
9,045.044
24,297.958
15,137.767
724.257
212.935
119.274
142.298
249.750
14,413.510
592.119
666.677
115.016
4,795.460
5,177.969
3,066.269
9,160.191
8,589.163
626.425
3,476.395
138.795
515.338
860.533
2,971.677
571.028
23,148.526
14,925.846
740.193
208.431
112.388
181.465
237.909
14,185.653
586.790
660.677
116.144
4,743.868
5,081.564
2,996.612
8,222.680
7,710.094
562.312
3,120.599
124.590
462.595
772.461
2,667.537
512.586
23,467.000
15,390.929
711.733
220.379
103.981
161.747
225.626
14,679.196
622.051
700.378
110.086
5,027.055
5,415.461
2,804.166
8,076.071
7,572.624
552.286
3,064.959
122.369
454.347
758.688
2,619.975
503.446
23,112.108
14,595.342
690.197
200.678
110.570
144.820
234.129
13,905.145
570.500
642.336
118.173
4,605.768
4,964.996
3,003.372
8,516.766
7,985.848
582.424
3,232.208
129.046
479.140
800.088
2,762.942
530.918
94,025.591
60,049.884
2,866.379
842.423
446.213
630.329
947.414
57,183.504
2,371.460
2,670.068
459.418
19,172.150
20,639.989
11,870.419
33,975.708
31,857.729
2,323.447
12,894.163
514.801
1,911.419
3,191.770
11,022.130
2,117.979
USPS-T-7
353
2007GQ1
STANDARD MAIL
Nonprofit Rate Bulk
Nonprofit
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Nonprofit ECR
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
TOTAL STANDARD MAIL
Atta chm e nt A
R2005-1 Volum e Fore ca st: Be fore -Ra te s
(m illions of pie ce s)
2007GQ2
2007GQ3
2007GQ4
pa ge 6 of 10
2007GFY
4,697.139
3,743.583
497.082
147.261
31.459
253.256
65.106
3,246.501
264.544
255.691
29.324
1,179.096
999.882
517.964
953.556
889.917
106.270
394.353
22.187
17.347
202.290
147.470
63.638
28,995.097
3,897.985
3,106.664
387.791
121.811
26.126
185.807
54.047
2,718.873
219.737
212.383
24.316
1,048.624
783.990
429.822
791.321
738.510
88.189
327.260
18.412
14.395
167.873
122.380
52.811
27,046.511
3,484.267
2,776.934
343.896
107.256
23.229
166.291
47.120
2,433.038
197.242
190.641
21.860
939.979
697.925
385.393
707.333
660.127
78.829
292.525
16.458
12.868
150.056
109.391
47.206
26,951.267
3,795.438
3,024.935
360.013
115.698
25.272
168.792
50.251
2,664.922
215.435
208.225
23.843
1,017.693
778.838
420.889
770.504
719.082
85.869
318.650
17.928
14.017
163.457
119.161
51.422
26,907.546
15,874.830
12,652.116
1,588.781
492.026
106.085
774.147
216.523
11,063.335
896.958
866.940
99.344
4,185.391
3,260.634
1,754.068
3,222.714
3,007.636
359.158
1,332.789
74.984
58.626
683.677
498.402
215.077
109,900.421
PACKAGE SERVICES
Parcel Post
Non-Destination Entry
(Inter-BMC)
(Intra-BMC)
Destination Entry
(DBMC)
(DSCF)
(DDU)
Bound Printed Matter
Media Mail
Library Rate
TOTAL PACKAGE SERVICES MAIL
113.540
36.351
25.545
10.806
77.189
25.665
0.853
50.672
153.698
49.940
4.416
321.594
83.108
30.200
21.223
8.977
52.908
17.591
0.584
34.732
156.122
49.678
4.393
293.302
77.588
26.438
18.579
7.859
51.149
17.007
0.565
33.578
137.030
49.085
4.341
268.044
74.444
24.891
17.492
7.399
49.553
16.476
0.547
32.530
174.118
46.917
4.149
299.629
348.680
117.881
82.839
35.042
230.799
76.738
2.549
151.512
620.968
195.621
17.300
1,182.569
Postal Penalty
Free-for-the-Blind
180.670
19.368
176.850
17.581
180.243
20.558
169.994
19.007
707.756
76.514
57,056.145
54,550.102
53,309.531
52,400.149
217,315.926
0.958
9.916
68.817
0.390
61.077
43.720
216.762
2.977
404.618
0.904
7.392
75.105
0.411
66.843
46.076
191.501
2.630
390.862
0.848
6.567
74.662
0.406
66.960
44.780
190.496
2.616
387.335
0.827
5.811
68.512
0.401
60.910
42.563
169.172
2.324
350.519
3.537
29.687
287.097
1.607
255.790
177.139
767.931
10.547
1,533.334
3.943
25.958
4.225
22.677
4.370
27.610
4.114
14.028
16.652
90.272
TOTAL DOMESTIC MAIL
DOMESTIC SPECIAL SERVICES
Registry
Insurance
Certified
Collect-on-Delivery
Return Receipts
Money Orders
Delivery Confirmation
Signature Confirmation
TOTAL SPECIAL SERVICES
Post Office Boxes
Stamped Cards
USPS-T-7
354
2006GQ1
FIRST-CLASS MAIL
First-Class Letters & Flats
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC Letters)
(Mixed-ADC Flats)
(AADC Letters)
(AADC Flats)
(3-Digit Letters)
(5-Digit Letters)
(3-Digit Flats)
(5-Digit Flats)
(Carrier-Route Letters)
First-Class Cards
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC)
(AADC)
(3-Digit)
(5-Digit)
(Carrier-Route)
TOTAL FIRST-CLASS MAIL
Priority Mail
Express Mail
Mailgrams
PERIODICAL MAIL
W ithin County
Nonprofit
Classroom
Regular Rate
TOTAL PERIODICAL MAIL
STANDARD MAIL
Regular Rate Bulk
Regular
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Enhanced Carrier-Route
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
Atta chm e nt A
R2005-1 Volume Fore ca st: Afte r-Ra te s
(m illions of pie ce s)
2006GQ2
2006GQ3
2006GQ4
pa ge 7 of 10
2006GFY
23,920.285
11,744.407
12,175.878
324.720
11,851.158
727.134
12.396
657.575
31.873
5,745.086
4,374.215
59.776
63.462
179.641
1,443.095
641.014
802.081
79.807
722.274
73.089
55.305
313.453
259.166
21.261
25,363.380
22,738.854
10,406.051
12,332.803
285.267
12,047.535
733.989
13.256
663.774
34.085
5,888.365
4,427.757
60.547
64.281
161.481
1,346.199
611.057
735.142
77.513
657.630
70.269
53.172
282.047
235.154
16.987
24,085.053
22,108.115
10,366.273
11,741.842
267.892
11,473.949
697.025
14.466
630.346
37.197
5,576.157
4,253.825
57.644
61.199
146.091
1,347.558
611.996
735.562
71.697
663.866
71.749
54.291
290.830
231.283
15.712
23,455.673
21,578.760
9,942.564
11,636.196
245.309
11,390.887
688.369
13.973
622.518
35.931
5,533.549
4,236.601
57.125
60.648
142.173
1,327.043
618.463
708.581
65.917
642.664
67.561
51.122
285.248
224.563
14.170
22,905.804
90,346.014
42,459.296
47,886.718
1,123.189
46,763.529
2,846.518
54.091
2,574.212
139.086
22,743.156
17,292.398
235.093
249.589
629.387
5,463.895
2,482.529
2,981.366
294.933
2,686.433
282.668
213.891
1,171.577
950.167
68.130
95,809.909
225.609
13.025
0.343
195.910
12.602
0.366
194.882
12.867
0.382
182.923
11.894
0.269
799.324
50.388
1.359
195.697
468.256
16.664
1,590.963
2,271.581
185.213
476.423
16.956
1,644.108
2,322.700
187.724
452.073
16.091
1,633.284
2,289.171
184.944
418.243
14.887
1,548.297
2,166.370
753.578
1,814.995
64.598
6,416.651
9,049.822
23,334.207
14,374.458
736.434
207.128
120.900
158.879
249.527
13,638.024
556.974
627.107
105.001
4,531.862
4,869.698
2,947.383
8,959.749
8,400.884
613.410
3,402.122
135.625
504.491
841.655
2,903.582
558.865
21,570.205
13,917.056
739.624
198.436
111.042
199.313
230.834
13,177.432
543.356
611.774
104.301
4,408.620
4,692.909
2,816.472
7,653.149
7,175.708
524.113
2,906.397
115.816
431.019
718.906
2,479.458
477.441
21,653.868
14,085.732
693.803
206.368
99.680
174.610
213.145
13,391.928
567.286
638.717
96.246
4,600.154
4,930.147
2,559.379
7,568.137
7,095.922
518.453
2,874.531
114.498
426.331
710.909
2,451.199
472.215
22,107.457
14,101.393
706.664
197.803
111.849
164.977
232.035
13,394.729
549.532
618.728
109.167
4,449.330
4,772.412
2,895.561
8,006.065
7,506.448
548.615
3,041.285
121.093
451.102
752.034
2,592.319
499.617
88,665.738
56,478.638
2,876.525
809.734
443.472
697.779
925.540
53,602.113
2,217.148
2,496.325
414.714
17,989.965
19,265.167
11,218.794
32,187.100
30,178.961
2,204.590
12,224.335
487.032
1,812.943
3,023.503
10,426.558
2,008.138
USPS-T-7
355
2006GQ1
STANDARD MAIL
Nonprofit Rate Bulk
Nonprofit
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Nonprofit ECR
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
TOTAL STANDARD MAIL
Atta chm e nt A
R2005-1 Volume Fore ca st: Afte r-Ra te s
(m illions of pie ce s)
2006GQ2
2006GQ3
2006GQ4
pa ge 8 of 10
2006GFY
4,303.939
3,430.404
487.149
141.292
29.206
254.590
62.061
2,943.256
239.194
231.189
26.494
1,075.167
898.970
472.242
873.534
815.227
97.354
361.283
20.315
15.892
185.320
135.064
58.307
27,638.146
3,753.909
2,992.062
405.512
122.749
25.505
203.104
54.153
2,586.550
208.881
201.890
23.078
1,007.805
733.017
411.880
761.848
710.993
84.907
315.096
17.715
13.860
161.627
117.787
50.855
25,324.115
3,487.689
2,779.923
373.175
111.958
23.489
189.072
48.657
2,406.748
195.136
188.605
21.651
939.159
677.860
384.337
707.766
660.519
78.881
292.735
16.455
12.876
150.155
109.418
47.247
25,141.557
3,872.789
3,087.080
396.106
122.837
26.031
194.737
52.501
2,690.974
217.466
210.188
24.098
1,036.308
774.565
428.349
785.709
733.249
87.570
324.996
18.258
14.295
166.696
121.435
52.460
25,980.246
15,418.326
12,289.469
1,661.942
498.835
104.232
841.502
217.372
10,627.527
860.676
831.871
95.321
4,058.439
3,084.412
1,696.808
3,128.857
2,919.989
348.712
1,294.110
72.744
56.923
663.797
483.704
208.869
104,084.064
PACKAGE SERVICES
Parcel Post
Non-Destination Entry
(Inter-BMC)
(Intra-BMC)
Destination Entry
(DBMC)
(DSCF)
(DDU)
Bound Printed Matter
Media Mail
Library Rate
TOTAL PACKAGE SERVICES MAIL
110.930
36.043
25.329
10.714
74.886
24.888
0.827
49.171
154.274
51.356
4.537
321.096
79.025
29.099
20.449
8.650
49.926
16.593
0.552
32.781
150.468
47.582
4.203
281.278
74.325
25.494
17.915
7.579
48.831
16.229
0.540
32.062
129.185
47.239
4.171
254.920
72.169
24.275
17.059
7.216
47.893
15.917
0.529
31.447
172.069
45.263
3.996
293.497
336.448
114.911
80.752
34.160
221.536
73.627
2.448
145.461
605.996
191.440
16.908
1,150.792
Postal Penalty
Free-for-the-Blind
170.402
19.977
164.032
16.796
169.440
19.903
162.664
18.642
666.538
75.317
56,023.557
52,402.851
51,538.796
51,722.309
211,687.513
1.022
11.616
66.857
0.411
58.948
45.256
195.150
2.680
381.941
0.942
8.718
71.557
0.422
63.066
45.931
168.143
2.309
361.089
0.901
8.011
73.015
0.422
64.703
45.247
169.749
2.331
364.380
0.873
7.020
67.382
0.418
59.253
43.505
152.975
2.101
333.528
3.738
35.366
278.811
1.673
245.970
179.939
686.017
9.422
1,440.938
3.725
26.171
3.900
22.029
4.097
27.189
3.850
14.041
15.573
89.429
TOTAL DOMESTIC MAIL
DOMESTIC SPECIAL SERVICES
Registry
Insurance
Certified
Collect-on-Delivery
Return Receipts
Money Orders
Delivery Confirmation
Signature Confirmation
TOTAL SPECIAL SERVICES
Post Office Boxes
Stamped Cards
USPS-T-7
356
2007GQ1
FIRST-CLASS MAIL
First-Class Letters & Flats
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC Letters)
(Mixed-ADC Flats)
(AADC Letters)
(AADC Flats)
(3-Digit Letters)
(5-Digit Letters)
(3-Digit Flats)
(5-Digit Flats)
(Carrier-Route Letters)
First-Class Cards
-- Single-Piece
-- W orkshared
(Nonautomated Presort)
(Automated)
(Mixed-ADC)
(AADC)
(3-Digit)
(5-Digit)
(Carrier-Route)
TOTAL FIRST-CLASS MAIL
Priority Mail
Express Mail
Mailgrams
PERIODICAL MAIL
W ithin County
Nonprofit
Classroom
Regular Rate
TOTAL PERIODICAL MAIL
STANDARD MAIL
Regular Rate Bulk
Regular
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Enhanced Carrier-Route
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
Atta chm e nt A
R2005-1 Volume Fore ca st: Afte r-Ra te s
(m illions of pie ce s)
2007GQ2
2007GQ3
2007GQ4
pa ge 9 of 10
2007GFY
23,185.498
11,320.181
11,865.317
267.344
11,597.974
699.170
14.213
632.285
36.547
5,595.288
4,345.585
58.254
61.846
154.786
1,421.031
623.964
797.067
67.828
729.239
72.309
54.715
315.269
267.631
19.315
24,606.530
22,668.573
10,313.095
12,355.478
231.388
12,124.090
725.048
15.567
655.688
40.029
5,897.322
4,522.570
60.660
64.401
142.806
1,381.104
619.329
761.775
68.680
693.095
72.456
54.827
296.121
253.641
16.049
24,049.678
21,638.718
10,033.055
11,605.663
222.584
11,383.079
678.449
16.592
613.546
42.664
5,508.912
4,278.338
56.976
60.489
127.112
1,364.601
611.890
752.711
62.652
690.058
73.001
55.239
301.018
246.229
14.571
23,003.319
20,963.764
9,604.779
11,358.985
202.750
11,156.235
661.274
15.854
598.015
40.766
5,399.413
4,203.898
55.763
59.202
122.050
1,321.677
607.841
713.835
56.621
657.215
67.612
51.161
290.352
235.194
12.896
22,285.441
88,456.554
41,271.110
47,185.444
924.066
46,261.378
2,763.941
62.226
2,499.534
160.006
22,400.936
17,350.391
231.654
245.938
546.754
5,488.413
2,463.025
3,025.388
255.781
2,769.607
285.378
215.942
1,202.760
1,002.696
62.831
93,944.968
227.038
11.622
0.299
202.809
12.085
0.321
199.407
12.062
0.331
184.829
11.389
0.228
814.082
47.158
1.179
183.869
449.646
16.005
1,608.530
2,258.050
182.952
466.665
16.610
1,655.202
2,321.429
182.853
438.267
15.600
1,626.330
2,263.050
177.788
401.461
14.289
1,530.524
2,124.063
727.463
1,756.039
62.504
6,420.586
8,966.592
23,618.657
14,984.613
702.231
210.654
117.707
131.105
242.766
14,282.381
586.382
660.218
114.229
4,746.340
5,135.862
3,039.351
8,634.044
8,095.060
592.020
3,280.802
130.519
486.716
810.995
2,794.008
538.985
22,525.253
14,774.871
716.072
206.239
110.932
167.178
231.723
14,058.799
581.083
654.252
115.305
4,695.276
5,043.134
2,969.749
7,750.382
7,266.561
531.429
2,945.024
117.161
436.903
727.993
2,508.051
483.821
22,847.467
15,235.273
689.368
218.071
102.642
149.105
219.549
14,545.905
615.997
693.562
109.276
4,975.563
5,372.231
2,779.276
7,612.194
7,136.999
521.953
2,892.515
115.072
429.113
715.013
2,463.333
475.195
22,475.284
14,447.708
669.699
198.600
109.150
133.558
228.391
13,778.009
564.937
636.072
117.296
4,558.591
4,925.068
2,976.045
8,027.576
7,526.451
550.435
3,050.353
121.351
452.529
754.030
2,597.752
501.125
91,466.661
59,442.464
2,777.370
833.564
440.430
580.947
922.429
56,665.094
2,348.399
2,644.103
456.106
18,975.770
20,476.295
11,764.421
32,024.197
30,025.070
2,195.837
12,168.694
484.103
1,805.260
3,008.031
10,363.144
1,999.126
USPS-T-7
357
2007GQ1
STANDARD MAIL
Nonprofit Rate Bulk
Nonprofit
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(Presort Letters)
(Presort Nonletters)
-- Automated
(Mixed-ADC Letters)
(AADC Letters)
(Basic Flats)
(3-Digit Letters)
(5-Digit Letters)
(3/5-Digit Flats)
Nonprofit ECR
-- Nonautomated
(Basic Letters)
(Basic Nonletters)
(High-Density Letters)
(High-Density Nonletters)
(Saturation Letters)
(Saturation Nonletters)
-- Automated
TOTAL STANDARD MAIL
Atta chm e nt A
R2005-1 Volume Fore ca st: Afte r-Ra te s
(m illions of pie ce s)
2007GQ2
2007GQ3
2007GQ4
pa ge 10 of 10
2007GFY
4,616.808
3,680.680
478.670
144.703
30.934
239.146
63.887
3,202.010
260.200
251.492
28.839
1,165.678
986.347
509.453
936.128
873.601
104.340
387.275
21.728
17.034
198.622
144.602
62.527
28,235.465
3,831.305
3,054.447
374.572
119.696
25.690
176.150
53.035
2,679.875
216.128
208.895
23.914
1,033.324
774.855
422.759
776.859
724.970
86.588
321.386
18.031
14.136
164.829
120.000
51.889
26,356.559
3,424.665
2,730.260
332.222
105.398
22.842
157.726
46.257
2,398.038
194.000
187.507
21.497
926.128
689.867
379.039
694.406
648.024
77.398
287.275
16.117
12.635
147.335
107.263
46.382
26,272.132
3,730.510
2,974.088
347.829
113.697
24.851
159.933
49.347
2,626.260
211.892
204.800
23.448
1,002.976
769.212
413.931
756.421
705.897
84.310
312.931
17.557
13.764
160.493
116.843
50.524
26,205.794
15,603.289
12,439.475
1,533.292
483.494
104.318
732.955
212.526
10,906.183
882.219
852.694
97.698
4,128.107
3,220.281
1,725.184
3,163.814
2,952.492
352.637
1,308.866
73.432
57.568
671.280
488.708
211.322
107,069.950
PACKAGE SERVICES
Parcel Post
Non-Destination Entry
(Inter-BMC)
(Intra-BMC)
Destination Entry
(DBMC)
(DSCF)
(DDU)
Bound Printed Matter
Media Mail
Library Rate
TOTAL PACKAGE SERVICES MAIL
107.523
35.629
25.037
10.591
71.895
23.894
0.794
47.206
152.886
49.200
4.344
313.953
78.879
29.600
20.801
8.799
49.279
16.378
0.545
32.356
155.298
48.943
4.321
287.440
73.554
25.913
18.210
7.703
47.641
15.833
0.526
31.281
136.306
48.358
4.270
262.488
70.550
24.397
17.144
7.252
46.154
15.339
0.510
30.305
173.198
46.223
4.081
294.052
330.506
115.538
81.192
34.346
214.968
71.444
2.375
141.149
617.688
192.724
17.016
1,157.934
Postal Penalty
Free-for-the-Blind
180.670
19.368
176.850
17.581
180.243
20.558
169.994
19.007
707.756
76.514
55,852.995
53,424.752
52,213.589
51,294.796
212,786.132
0.892
9.660
67.443
0.378
59.442
42.374
208.209
2.860
391.257
0.841
7.206
73.597
0.398
65.047
44.657
183.944
2.526
378.217
0.790
6.399
73.163
0.393
65.162
43.401
182.978
2.513
374.800
0.770
5.662
67.135
0.389
59.272
41.252
162.496
2.232
339.208
3.293
28.928
281.338
1.557
248.923
171.685
737.626
10.131
1,483.482
3.814
25.611
4.087
22.380
4.227
27.250
3.980
13.846
16.107
89.088
TOTAL DOMESTIC MAIL
DOMESTIC SPECIAL SERVICES
Registry
Insurance
Certified
Collect-on-Delivery
Return Receipts
Money Orders
Delivery Confirmation
Signature Confirmation
TOTAL SPECIAL SERVICES
Post Office Boxes
Stamped Cards