Postal Rate Commission
Submitted 12/21/2006 4:05 pm
Filing ID: 55481
Accepted 12/21/2006
UNITED STATES OF AMERICA
Before the
POSTAL REGULATORY COMMISSION
WASHINGTON, D.C. 20268-0001
Postal Rate and Fee Changes, 2006
)
Docket No. R2006-1
Initial Brief
of
The Office of the Consumer Advocate
SHELLEY S. DREIFUSS
DIRECTOR
EMMETT RAND COSTICH
KENNETH E. RICHARDSON
ATTORNEYS
December 21, 2006
-i-
TABLE OF CONTENTS
I.
Page
EXECUTIVE SUMMARY...................................................................................... 1
II.
ECONOMETRIC ANALYSIS OF MAIL PROCESSING COSTS......................... 12
A.
B.
C.
II.
A Comparison of Alternative Models of Mail Processing Costs ............... 13
1.
Witness Robert’s model of mail processing .................................. 14
2.
Witness Bozzo’s model of mail processing ................................... 26
3.
Implications of the model of mail processing for
measurement and allocation of volume variable cost in
mail processing. ............................................................................ 29
Response to Rebuttal Testimony............................................................. 33
1.
Definition of processing categories ............................................... 34
2.
Quarterly dummy variables and shifts in sorting
technology..................................................................................... 36
3.
Sample period analyzed................................................................ 39
Conclusion ............................................................................................... 40
ECONOMETRIC ANALYSIS OF CITY CARRIER COSTS................................. 42
A.
Witness Bradley’s Volume Variabilities Presented in Docket No.
R2005-1 Are Flawed................................................................................ 42
B.
The Methodology in the Original Carrier Cost Study Needs
Improvement............................................................................................ 44
1.
The density variable should not have been used in the
econometric equations. ................................................................. 44
2.
In rebuttal testimony and during oral cross-examination,
witness Bradley was unsuccessful in demonstrating the
relevance of the density variable................................................... 49
3.
Witness Bradley did not correctly compute the density
variable. ........................................................................................ 51
4.
Witness Bradley presented carrier cost variabilities based
on a restricted, rather than full, quadratic, a major
deficiency apparently due to computational problems. ................. 52
- ii -
C.
Witness Smith Provided Alternative Variabilities. .................................... 54
D.
Witness Smith Has Concluded That Parcels and Accountables
Should Be 100 Percent Attributable in Cost. ........................................... 56
E.
The Use of the DOIS Database Would Solve Much of the Data
Problem. .................................................................................................. 57
1.
F.
Witness Bradley’s Conclusions Are Now Outdated. ................................ 61
1.
III.
Witness Smith performed an analysis of volume
variability using the DOIS database. ............................................. 60
Delivery technology is changing.................................................... 61
G.
Witness Crowder’s Criticisms Are Largely Irrelevant ............................... 63
H.
Concluding Comments ............................................................................ 64
WINDOW SERVICE COSTS ............................................................................. 66
A.
In This Proceeding, the Postal Service Updated the Previous
Window Service Transaction Supply Side Variabilities............................ 66
1.
Postal Service witness Nieto’s updated window service
transaction time study ................................................................... 68
2.
The field study data collection effort appears to be
seriously flawed because witness Nieto could not provide
adequate justification of the sample design for data
collection. ...................................................................................... 69
3.
The transaction time study was not representative of
small and large sites. .................................................................... 71
4.
The Postal Service offers no substantiation that the
collection of data over a two week period in April/May
resulted in a database that was representative of postal
transaction times or representative of the mix of postal
products. ....................................................................................... 73
5.
The Postal Service’s elimination of significant amounts of
data raises the issue of whether the window service
transaction study included sufficient quality control. ..................... 75
6.
The erratic walk-time data from facility to facility and day
to day are evidence that there may have been errors in
the field study data collection process. ......................................... 77
- iii -
7.
B.
C.
IV.
The Postal Service did not analyze whether enough data
was gathered for each type of window transaction in
order to have a statistically adequate sample. .............................. 78
Using the Data Presented by Witness Nieto, Postal Service
Witness Bradley Updated the Window Service Supply Side
Variabilities with a Methodology Inconsistent with His Previous
Testimony Before This Commission and Problematic in Several
Other Respects........................................................................................ 84
1.
Witness Bradley’s model in the current estimation
procedure is linear, but he has used a quadratic
approach in other testimony.......................................................... 84
2.
Witness Bradley’s model is sensitive to changes in the
data and thus is not robust............................................................ 86
3.
Witness Bradley should have included walk-time in the
modeling effort. ............................................................................. 87
4.
Witness Bradley’s estimations of variability are flawed................. 89
In Conclusion, There Are Significant Problems with the Updated
Window Service Transaction Side Variabilities........................................ 89
FIRST-CLASS RATE PROPOSAL..................................................................... 92
A.
B.
OCA Witness Thompson’s First-Class Rate Proposal Simplifies
the Postal Service’s Increasingly Complex Schedule .............................. 92
1.
OCA’s proposed four-ounce incremental rates, for letter-,
flat- and parcel-shaped mail pieces, significantly reduce
the complexity of the First-Class single-piece rate
schedule proposed by the Postal Service ..................................... 92
2.
Elimination of the additional ounce rate is feasible and
justifiable ....................................................................................... 96
The Commission Should Reject the Postal Service’s Proposal
for a Presort “Base Rate” in Favor of Continuing to Use the
Traditional Bulk Metered Mail Benchmark when Establishing
Presort Rates........................................................................................... 98
1.
USPS Witness Taufique erroneously argues that the
average First-Class single piece is the appropriate
benchmark. ................................................................................. 100
2.
OCA opposes giving intervenors greater discounts .................... 103
- iv -
C.
V.
Declining First Class Mail Volumes May Force a Review to
Determine Which Mail Class Bears an Increasing Portion of the
Cost of Universal Service....................................................................... 107
CONFIRM RATE PROPOSAL ......................................................................... 109
A.
B.
C.
The Postal Service’s Proposal to Restructure the Existing
Subscription-Based Fee Schedule for Confirm Service Is
Unnecessary to Recover Costs and Can Be Achieved by
Adjusting Current Fees, as Proposed by the OCA ................................ 109
1.
The Postal Service fails to meet its burden of proof that
the classification changes it proposes for confirm service
are superior to the existing subscription-based fee
structure that was held fair and equitable in Docket No.
MC2002-1 ................................................................................... 110
2.
The Postal Service’s proposal will increase costs for, and
reduce usage by, Confirm subscribers........................................ 111
3.
The OCA’s proposal minimizes fee increases for most
Confirm subscribers in order to expand usage while
recovering costs .......................................................................... 114
The OCA’s Subscription-Based Fees for Confirm Service Will
Recover Institutional Costs Unlike the Transaction-Based Fees
Proposed by the Postal Service............................................................. 115
1.
OCA’s Estimate of Confirm subscribers and estimated
revenues is realistic..................................................................... 115
2.
Witness Callow’s revenue estimate is likely to be
achieved...................................................................................... 119
3.
The Postal Service’s estimate of Confirm transactions
lacks support, resulting in a revenue estimate that will not
materialize................................................................................... 123
The OCA’s Fee Structure for Confirm Service Is Presumptively
Fair and Equitable and the Postal Service’s Fees Are Arbitrary
and Not Fair and Equitable. ................................................................... 127
1.
OCA proposes to retain the existing, presumptively fair
and equitable subscription-based fee structure previously
recommended by the Commission.............................................. 127
2.
The Postal Service’s differential unit-based fees for
blocks of one-million scans is arbitrary, and unfair and
inequitable................................................................................... 130
-v-
VI.
D.
The OCA’s Fee Proposal for Confirm Service Preserves the
Simplicity Inherent in the Existing Subscription-Based Fee
Structure ................................................................................................ 132
E.
The Potential for Future Enhancements to Confirm Service Is
Not a Valid Reason to Accept the Postal Service’s Unit Pricing
Proposal In Lieu of the OCA’s Subscription-Based Fees Which
Can Also Accommodate Future Enhancements.................................... 134
F.
The Existing Requirement to Provide Electronic Notifications for
Entry of Confirm Mailings Should Be Retained to Permit
Development of System-Wide Measurement of Service
Performance .......................................................................................... 138
FOREVER STAMP........................................................................................... 142
A.
B.
VII.
The Postal Service’s Proposed “Forever Stamp” Mail
Classification Should Be Recommended by the Commission to
Enhance Convenience for Individuals and Smaller Mailers................... 142
1.
The Postal Service’s Forever Stamp proposal is a
desirable postage option for individuals, smaller mailers
and the Postal Service, and has minimal financial risk. .............. 143
2.
The Postal Service’s cooperative development of the
Forever Stamp proposal is commendable. ................................. 150
The Commission Should Recommend Witness Carlson’s
Proposed DMCS Language to Resolve Ambiguities Created by
the Postal Service’s Proposed DMCS Language About the Use
of the Forever Stamp ............................................................................. 151
1.
The Postal Service’s proposed DMCS language is
ambiguous about the weight and classes of mail on
which the Forever Stamp may be used....................................... 151
2.
The Postal Service’s proposed DMCS language should
be modified as proposed by intervenor witness Carlson............. 156
THE POSTAL SERVICE’S INSURANCE AND COD PRODUCTS HAVE A
LOW VALUE OF SERVICE THAT HAS NOT BEEN TAKEN INTO
ACCOUNT IN THE COST COVERAGES PROPOSED FOR THESE
SERVICES ....................................................................................................... 157
A.
The Cost Coverage For Insurance Should Be Set Close to Zero .......... 157
B.
The Cost Coverage of Collect on Delivery (COD) Should Be Set
Near Zero............................................................................................... 162
-6-
Docket No. R2006-1
I.
-1-
Initial Brief of the OCA
EXECUTIVE SUMMARY
In the current rate case, OCA has focused its resources on three major cost
areas: volume variability of mail processing costs, volume variability of city carrier
costs, and supply-side variabilities for window service costs. OCA has also made two
major rate proposals: a user-friendly First-Class rate design that sets rate levels using
the well established bulk metered mail benchmark and a Confirm pricing proposal that
has broad support within the mailing community. OCA also gives its endorsement to
the Postal Service’s Forever Stamp proposal. Finally, OCA recommends sharp
reductions in cost coverage for Insurance and COD as a consequence of the poor
quality of service provided to purchasers of these special services. OCA’s treatment of
each of these issues is summarized as follows.
Mail Processing Costs
The Postal Service’s mail processing cost model incorporates so many restrictive
assumptions that it virtually guarantees volume variabilities of less than 100 percent.
As professor Mark Roberts (OCA-T-1) shows, the Postal Service eliminates by
assumption one of the paths by which changes in mail volume can cause changes in
cost. Specifically, the Postal Service model prohibits a change in volume from affecting
the mix of processing operations (e.g., manual, OCR, DBCS) used in a plant. The
Postal Service assumes that there can never be an increase in volumes so great that it
would cause a plant to use high-cost sorting operations more intensively. As witness
Oronzio tells it, there is always enough DBCS capacity in every plant to process all
machinable letters. This may be the view from headquarters, but MODS data tell a
different story. Professor Roberts’s model allows the MODS data to speak freely.
Docket No. R2006-1
-2-
Initial Brief of the OCA
The Postal Service’s estimation procedure also serves to reduce volume
variability. The labor demand equation that witness Bozzo estimates removes the
effects on cost of seasonal variations in volume. However, as professor Roberts
shows, it is important to measure how a plant responds to seasonal volume changes.
In particular, if the only seasonal effect known is an increase in nonmachinable letters
during the peak holiday mailing season, this change in volume should be explicitly
modeled. Professor Roberts’s model can do so (by incorporating a “nonmachinable”
processing category), unlike the Postal Service’s model.
City Carrier Costs
OCA recommends that the Commission reject the volume variability results
presented by Postal Service witness Bradley in favor of those developed by OCA
witness Jed Smith (OCA-T-3). The Postal Service’s city carrier analysis is plagued by a
number of fatal defects: (1) the restrictions imposed by witness Bradley on the
quadratic equations he employs are excessive; (2) the City Carrier Street Time Study
(CCSTS) database on which he relies suffers from multicollinearity, resulting in
meaningless results for the Full Quadratic equation, high Variance Inflation Factors,
and potentially incorrect regressors; (3) the 2002 CCSTS database is no longer
representative of the major technology used to deliver mail, i.e., a “three bundle
approach,” consisting of the carrier routinely carrying one Delivery Point Sequenced
(DPS) bundle, one Cased Mail bundle, and one Sequenced Mail bundle to the street;
the currency of the three-bundle model is advocated by Dr. Bradley himself in a paper
he recently presented in Bern; (4) witness Bradley incorporates a density variable in his
model that witness Smith has discredited; and (5) the Postal Service model is not
Docket No. R2006-1
-3-
Initial Brief of the OCA
founded on a fully articulated and substantiated economic theory that relates the
econometric estimation approach to the effect on city carrier costs from changes in
volume, by mail type.
Witness Smith provides alternative variabilities, summarized in Table 2 of this
brief. At the request of the Presiding Officer (in POIR 25) to modify the Postal Service
Full Quadratic model, witness Smith willingly did so. Witness Smith re-ran witness
Bradley’s model, but removed the density variable and the cross-products terms with
“spr.” Witness Smith refers to this case as “CC3: Witness Bradley Recommended
Case without the Density Variable, Rerun for full quadratic, spr cross products
eliminated.”
Witness Smith performed a second run, that was also provided in response to
POIR 25. In this case, he re-ran the model with density defined correctly and the “spr”
cross-products terms again removed. This equation is denoted as “CC2: Rerun for Full
Quadratic, spr cross products eliminated” in the table.
OCA advocates serious consideration by the Commission of two other cases that
were presented in witness Smith’s testimony. The first is equation CC5A, a full
quadratic model based on the “three bundle” technology, using the CCSTS database.
A second equation favored by OCA is equation ND6, which uses the DOIS database.
OCA believes that all four of these approaches are superior to the variability
results presented by the Postal Service. If the Commission decides to use the CCSTS
database, over DOIS, and eschews the three-bundle model, equation CC3 would be
the best alternative.
Window Service Costs
Docket No. R2006-1
-4-
Initial Brief of the OCA
In the current rate case, the Postal Service presents new supply-side variabilities
for window service transactions. These variabilities are produced from a new window
service survey sponsored by witness Nieto that furnishes inputs for an econometric
analysis by witness Bradley. Before the current case, variabilities were based on a
1996 study and earlier methodologies. In the last ten years, there have been changes
in window service procedures, and new products and services have been added.
Moreover, the Postal Service has identified errors in the 1996 study. For all of these
reasons, the Postal Service decided that an update of window service variabilities was
necessary.
OCA witness Jed Smith (OCA-T-2) is in agreement that an update of the 1996
study is desirable. However, he finds that mistakes were made both in the data
collection process and the modeling process. Justification for the sample design of the
new survey has never been furnished. The decisions concerning the number, and
types, of sites to be surveyed, and the number of observations needed for particular
postal products were never determined using formal statistical methods. Rather, highly
subjective determinations underlie the data collection exercise. A decision to collect
data only at one time of year – April and May – has not been defended. It is well
established in the record of this case that the types and numbers of products sold at
retail windows vary throughout the year, the winter holiday season being the most
notable example. Poorly planned data collection techniques led to the deletion of a
significant amount of data and inadequate samples. Although not intending to do so,
rebuttal witness Kelley confirms these conclusions.
Docket No. R2006-1
-5-
Initial Brief of the OCA
The volume variabilities estimated by witness Bradley are necessarily dependent
on the inadequate database resulting from flawed data collection. For window service
analysis, witness Bradley decided to use a linear model, although he has often favored
quadratic models, both in this and earlier dockets. At the very least, justification for his
decision not to use a flexible functional form should be provided. This is particularly
important in the case of window service, since the use of a linear form is essentially
equivalent to assuming 100 percent volume variability for a number of postal products
whose sale is associated with other products.
On behalf of OCA, Witness Smith testifies to the importance of including walktime as part of the transaction when analyzing the supply-side variabilities. Although
the impact of including walk-time appears to be minimal, this impact is somewhat
dependent on the questionable accuracy of the underlying walk-time data.
The major issues that the Commission must decide are whether sample
composition is acceptable, whether walk-time has been treated appropriately, and
whether the underlying equation has been properly specified. Changes in these areas
could produce substantially different variabilities. OCA concludes with a
recommendation for the design and collection of a significantly improved database as
well as upgrades to the modeling effort.
First-Class Rate Proposal
OCA witness Pam Thompson (OCA-T-4) presents a major restructuring of FirstClass Mail that greatly simplifies the First-Class rate schedule. OCA supports the
USPS proposal of shape-based rates for First-Class Mail because it aligns costs with
the mail characteristics that cause them. However, one drawback of the USPS shape-
Docket No. R2006-1
-6-
Initial Brief of the OCA
based rates is increased complexity of the First-Class rate schedule. Consumers face
an array of new rates with complex new requirements for First-Class letter-, flat- and
parcel-shaped pieces.
Witness Thompson’s rate design compensates for complexities created by the
new shape-based rates by collapsing single-ounce rate cells into a much smaller
number of four-ounce cells. As a result, the current First-Class rate schedule is
dramatically simplified –144 rate cells are reduced to 28. Except for the newly
proposed First-Class Mail Business Parcels rate category, rates proposed for singlepiece First-Class Mail are multiples of 42-cents.
One of the distinct achievements of witness Thompson’s proposal is the
elimination of the additional ounce rate in First Class. USPS witness Taufique testifies
that, formerly, the additional ounce rate bore not only the cost of additional weight but
the costs caused by differences in shape as well. With shape-based rates, mailers
should be freed from the burden of weighing each piece of First-Class Mail so as to
account for every ounce increment that bumps up the postage tab. Unfortunately,
witness Taufique suffered from excessive timidity and would only go so far as to reduce
the additional ounce rate from 24 cents to 20 cents.
The Postal Service has again ignored the Commission’s oft-repeated instructions
to submit data justifying the additional ounce rate. Witness Taufique readily admits that
the additional ounce charge has been a significant source of revenue for First-Class
and the Postal Service as a whole. However, OCA witness Thompson’s First-Class
proposal proves that the additional ounce rate is neither justified nor necessary to meet
the First-Class or overall revenue requirement.
Docket No. R2006-1
-7-
Initial Brief of the OCA
OCA advocates retention of the BMM benchmark when establishing presort
rates. This is in contrast with the Postal Service’s position in the current rate case. In
his rebuttal testimony, witness Taufique claims that the Postal Service is unable to
develop a single benchmark that adequately reflects all of the varieties of mail pieces
that can convert to, or from, presort mail. He also testifies that the BMM benchmark the
USPS formerly used has been a source of controversy among presort mailers, who
believe they are entitled to greater discounts based on the worksharing they perform for
the USPS. In keeping with bulk mailer preferences, witness Taufique proposes an
alternative “base rate” benchmark of $0.346. The USPS benchmark is significantly
lower ($0.074 or 18 percent) than the Commission’s traditional BMM benchmark of
$0.42 (which would be the figure used in this case). Consequently, the USPS proposed
Presort rates are much lower than would have occurred if the traditional BMM
benchmark had been used. The effect of unwarranted reductions in Presort automation
rates is to shift more of the cost burden of universal service to First-Class single
- piece
mailers with few alternatives, thereby ignoring Congress’ justification for granting the
Postal Service monopolies over letter mail and access to the mailbox.
Confirm Service Rate Proposal
In this proceeding, the Postal Service proposes a fundamental restructuring of
the schedule of fees for Confirm—a special service that provides mailers with near realtime tracking information on outgoing and incoming automation-compatible mailpieces
entered as First-Class Mail, Standard Mail, or Periodicals. The Postal Service
maintains that restructuring the existing fee schedule is necessary to generate sufficient
revenues to cover costs so that the Postal Service can continue offering Confirm
Docket No. R2006-1
-8-
Initial Brief of the OCA
Service. To generate sufficient revenues to cover Confirm service costs, the Postal
Service abandons its current subscription-based fee schedule, establishing instead a
single annual user fee, and a “declining block” schedule of fees for the purchase of
additional blocks of 1 million scans by Confirm mailers. As a consequence, the Postal
Service’s greater reliance on transaction-based fees will discourage usage by imposing
significantly higher fees on most Confirm subscribers, particularly resellers, who provide
important value-added services to mailers, and further diminish Confirm service as a
tool for measuring service performance by business mailers and the Postal Service.
The OCA disagrees that restructuring the existing fee schedule is necessary to
cover the costs of Confirm service. Rather, the Postal Service’s goal of covering
Confirm service costs can be achieved better by simply adjusting current fees, as
proposed by the OCA. Toward that end, the OCA’s proposed fees produce a cost
coverage of 127.3 percent, which is consistent with the Postal Service’s proposal, while
preserving the benefits of the existing “subscription-based” fee schedule—that of
encouraging the expanded use of Confirm service for use in promoting service
performance measurement. The testimony of OCA witness James F. Callow (OCA-T5), presents the development of the proposed fees for Confirm service.
The OCA also proposes, jointly with the Postal Service, retention of the
requirement for an electronic preshipment notification contained in DMCS §991.31,
which serves to “start the clock” for the Confirm mailings, and is essential to developing
transparent, system-wide service performance measurement for business mail.
Endorsement of the Forever Stamp Proposal
Docket No. R2006-1
-9-
Initial Brief of the OCA
The Postal Service proposes a new mail product that offers a beneficial postage
option to individuals and smaller mailers who predominantly use First-Class Mail. The
Postal Service’s proposed “Forever Stamp” classification will be a non-denominated
stamp with a postage value equivalent to the first-ounce, single-piece First-Class Mail
rate prevailing at the time of its use—even if usage occurs subsequent to changes in
postage rates years after purchase of the stamp. The Postal Service proposes that the
Forever Stamp, if recommended by the Commission, be sold initially at the approved
single-piece First-Class Mail rate (presumably $0.42). As a result, the Forever Stamp
will provide greater convenience for individuals and smaller mailers as they adjust to
First-Class Mail rate changes that occur at the next change in rates—after the $0.42
rate is implemented. Postal Service witness Altaf H. Taufique presents the Forever
Stamp proposal in two separate pieces of testimony. USPS-T-32 and USPS-T-48.
The Postal Service developed the Forever Stamp proposal presented in this
proceeding in cooperation with the OCA and the Greeting Card Association (GCA).
The OCA strongly supports Commission recommendation of the Forever Stamp
proposal. In recommending the Forever Stamp proposal, the OCA also urges the
Commission to resolve ambiguities in the DMCS language proposed by the Postal
Service regarding the use of the Forever Stamp in favor of convenience and simplicity
to individuals and smaller mailers. Toward that end, the OCA supports the changes to
DMCS §241 proposed by intervenor witness Douglas F. Carlson (DFC-T-1) that
prescribes more clearly the permissible use of the Forever Stamp.
Recommendation of Sharply Reduced Cost Coverage for Insurance and COD
Docket No. R2006-1
- 10 -
Initial Brief of the OCA
OCA urges the Commission to recommend the lowest possible cost coverage
(one percent) for Insurance service. This radical request stems from a years-long
investigation by OCA of the quality of service provided by Insurance. The dismal
experience of many Insurance claimants was first brought to the attention of OCA by
contacts from numerous individuals who had spent months, sometimes years, filing and
re-filing and re-filing again the documentation needed to receive indemnification for an
Insurance claim. From the stories told by individuals who contacted OCA, a common
pattern could be discerned.
First, clerks launched the claimants on their less-than-excellent adventures by
not filling out the Insurance claim forms correctly; not submitting the form; and/or
sending the form to the wrong place. This set off a chain of misery for claimants that
involved repeated visits to the post office to plead with clerks to perform the
fundamental activities needed to get the claim to the St. Louis Accounting Service
Center. Visiting a post office four, six, or eight times was not unusual.
Second, once the claim arrived at the Accounting Service Center,
pronouncements that documentation was insufficient seemed to be the rule, rather than
the exception. It sometimes happened that files would appear, disappear, and
reappear. Telephone calls by concerned claimants went unanswered. When letters of
rejection arrived months later, there would be inscrutable statements given for the
reasons claims were being denied – mere boilerplate was used for any and all rejected
claims with no trouble taken to explain to the particular claimant why his/her claim was
being denied.
Docket No. R2006-1
- 11 -
Initial Brief of the OCA
The length of time spent processing claims may give Insurance the distinction of
having the poorest performance of all postal products out of a well populated field of
candidates. The Postal Service has established 10 days as a respectable period for
the processing of an Insurance claim (once received and entered into the electronic
system in St. Louis); but the average length of time to process claims is five times the
established standard – nearly 50 days.
According to the Postal Service, it is not necessary to provide potential Insurance
purchasers with information about the claims process because damage or loss to
Insured items is rare. If this is so, then a public institution like the Postal Service has a
duty to inform potential purchasers that they may be wasting their money – processing,
transportation, and delivery of packages is so highly proficient that Insurance is almost
certainly unnecessary. Furthermore, in order to make an informed decision whether to
purchase Insurance, customers should automatically be given detailed printed
information letting them know how complicated the claims process is, how long a wait
they will have before the Postal Service acts on their claim, and the percentage of
claims denied. For purposes of setting rates, the Commission should bear in mind this
important fact about Insurance – customers are paying primarily for sale of the service
by a clerk and the obtaining of a recipient’s signature by a carrier. Indemnification is a
mere 15.5 percent of revenues, which may be the lowest level of indemnification among
the various types of Insurance sold in this country.
Purchasers in good faith of COD service have found themselves at the mercy of
the managers of the Accounting Service Center and the Consumer Advocate in the last
year and a half. The COD payment policies were dramatically altered so that
Docket No. R2006-1
- 12 -
Initial Brief of the OCA
reimbursement for non-delivered COD packages is random by package type and
random with respect to the amounts to be paid. The new policy was applied
immediately before an official decision to do so was ever made. COD service has now
been so seriously devalued as to warrant a cost coverage approaching zero.
II.
ECONOMETRIC ANALYSIS OF MAIL PROCESSING COSTS
The mail processing model which the Postal Service has advocated in rate
hearings since 1997 is flawed and should not be used as the basis for allocating mail
processing costs across rate classes. The framework makes strong, untested
assumptions about the relationship between the volume of mail processed in the
system and the use of individual sorting operations in each mail processing plant.
These assumptions eliminate one potentially important channel through which changes
in the volume of mail in a plant can affect labor hours and cost. Specifically, changes in
the plant’s volume of mail are assumed to have no effect on the mix of sorting
operations used in the plant. As a result, the framework does not even recognize the
possibility of the increased use of manual sorting in high volume periods or of capacity
constraints that might alter the mix of sorting operations used. Both are channels
through which an increase in mail volume can lead to rising marginal cost and both are
eliminated from consideration. The practical implication is that volume variabilities
estimated by the USPS using the MODS data are too narrowly defined and only
account for part of the relationship between mail volume and cost. In fact, the term
“volume variability” for these estimates is a misnomer. When these variabilities are
used to scale the labor expenditures in each sorting operation to create pools of volume
Docket No. R2006-1
- 13 -
Initial Brief of the OCA
variable costs, they will tend to underestimate the size of the VVC pools, or
equivalently, underestimate the marginal cost of processing mail.
This limitation is not the result of using a particular econometric method, sample
period, functional form for the labor demand equation, disaggregation of sorting
operations, or definition of categories of processing output. It is a more fundamental
problem that results from the way that information in the plant-level MODS data is
combined with aggregate data on the mix of mail by rate class (the distribution key).
The combination of steps requires that all of the relationship between mail volume and
the mix of sorting operations utilized in the plant be captured by the distribution key
step, but the assumptions underlying this step are particularly restrictive and constrain
how changes in mail volume are allowed to impact labor hours and costs. The
limitations of the USPS framework can be resolved by allowing a more general or
flexible relationship between plant-level mail volume and plant-level labor hours. This
more general relationship is a key component of the mail processing models developed
in Roberts (2002, 2006) and used in OCA-T-1.
A.
A Comparison of Alternative Models of Mail Processing Costs
The differences between the mail processing models underlying the most recent
USPS testimony (R2006-1, USPS-T-12) and the testimony in OCA-T-1 have been
discussed in detail in Roberts (2002, Sections II and III), Roberts (2006, Sections II, III,
and IV), OCA-T1, Section 4, pp. 8-11, and USPS/OCA-T1-24, Tr. 23/8337-8340. Some
of the key differences have been examined empirically, including the separability
assumption (USPS/OCA-T1-3c, Tr. 23/8290), the proportionality assumption (OCA-T-1,
Section IV, pp. 12-16, particularly Table 1), and the assumption that TPF is an
Docket No. R2006-1
- 14 -
Initial Brief of the OCA
exogenous regressor (USPS/OCA-T1-9, Tr. 23/8303 and USPS/OCA-T1-43, Tr.
23/8377-8388). In each case, the empirical evidence does not support the USPS
framework. In this section we do not review all of the issues raised but will highlight just
some of the more important distinctions. We will focus on the processing of lettershaped mail because all of the important distinctions can be made there and carried
over to processing of other shapes.
1.
Witness Robert’s model of mail processing
The goal of the mail processing model used in OCA-T-1 is to measure how a
change in the volume of mail arriving in a plant affects the use of man hours. This is
what is needed to measure the marginal cost of processing an additional piece of mail
in a processing plant. Figure 1 provides a summary of the important components of the
model in OCA-T-1.
1
The figure illustrates three stages to the processing and cost
allocation process. Beginning on the left, each box represents the system-wide volume
of mail in each of the 17 CRA rate classes (VRC1, VRC2, … VRC17). This mail then
moves to processing plants, and one plant is represented in the remaining panels of
Figure 1. The dashed vertical line separating the first and second stages distinguishes
the movement of mail that is external to the plant, and therefore is not measured in the
plant-level MODS data, from the internal movement of mail in the processing plant,
which is quantified in the MODS data.
1
The development of this model over time is reviewed in USPS/OCA-T1-35, Tr. 23/8364-8366.
Docket No. R2006-1
- 15 -
Initial Brief of the OCA
Once inside the processing plant, mail that will receive piece sorting in the facility
is recombined into distinct processing categories based on its shape, machinability,
level of mailer processing, and level of final dispatch. The number of processing
Docket No. R2006-1
- 16 -
Initial Brief of the OCA
Figure 1
Letter-Shaped Mail: OCA Model
In ternal to P lant
Exte rnal to P lant
Volum e of m ail by rate class
Vo lum e of m ail b y
processing category
So rting operations
VR C 1
V IN
Ho urs,
TP F
OC R
VOUT
Hours,
TP F
BC S
VR C 2
!
VR C 1 7
Hours,
TP H
Cost Pool Allocation Stage
M anual
Cost Pool Estim ation Stage
Docket No. R2006-1
- 17 -
Initial Brief of the OCA
categories used in the empirical modeling can be varied by the researcher but, in
practice, will be limited by the amount and quality of the data on plant mail volume.
Figure 1 distinguishes two processing categories, incoming processing and outgoing
processing. VIN and VOUT represent the volume of mail in each processing category
arriving in the plant.2 The mail in each processing category moves through the
separate sorting operations, represented as OCR, BCS, and manual, generating both
labor hours and piece feedings/handlings in each operation. Sorting the mail in each
processing category will generally require a different mix of sorting operations and thus
have a different quantity of labor hours and cost. For example, processing VIN will
require more BCS labor hours and fewer OCR hours than processing VOUT
(USPS/OCA-T1-45(d), Tr. 23/8386-87), and this can result in different marginal costs
for the two processing categories.
The arrows running from each processing category to each sorting operation
represents one of the hours-volume relationships that are estimated in OCA-T-1 using
the MODS data. In the case of two processing categories and three sorting operations
there are six output elasticities of labor demand that are estimated.3 The relationship
that is estimated is between the total volume of mail in each processing category
incoming or outgoing and the total hours of labor used in each sorting operation. It is
not necessary to distinguish the exact pathways that letters follow as they pass through
2
Two processing categories are used to simplify this explanation, but this can be extended, and is
extended in the empirical analysis in OCA-T-1, to a larger number of processing categories.
3
For this case, the six elasticities are reported in the first two lines of OCA-T-1, Table 3, p. 34.
Additional estimates for alternative specifications and data samples are reported in Table #4, p.37 and for
additional output processing categories in Table 5, p. 42.
Docket No. R2006-1
- 18 -
Initial Brief of the OCA
the sorting operations. All that is necessary is to estimate the relationship between the
volume in each processing category and total hours in each operation. For example, it
is not necessary to know how much of the mail volume moves from the OCR to the
BCS processing stage or how much is diverted to manual. It is also possible that some
of the pathways may not be used. For example, if the processing categories distinguish
mail that is barcoded, from mail that is not barcoded and the barcoded mail enters
processing in the BCS operation, then this category will not generate any labor hours in
OCR. This is fine. The estimated elasticities linking the volume of mail in each
processing category with the hours in each sorting operation will reflect whatever
relationships exist in the plant-level data.
4
A second point that is illustrated in Figure 1 is that there is no special role for
information on TPF/TPH. As the volume of mail processed in an operation increases,
amount of both labor the hours and machine time used in processing increase. We will
observe an increase in both TPF/TPH and man hours in the operation. In the
automated OCR and BCS operations, TPF is a count, not of how many unique pieces
of mail are processed, but how many passes are made through the sorting machinery.
This is a measure of the amount of work that the machinery does in contributing to the
4
Some efficiency can be gained in estimation by imposing restrictions on the elasticities if the
restrictions are true. For example, if barcoded mail never runs through an OCR operation then the
elasticity of OCR labor with respect to the barcoded mail category could be set equal to zero (i.e. the
volume of mail in this processing category would be dropped from the OCR labor demand equation). The
problem with this procedure is, if the restriction is not true, then the output elasticities for the other
processing categories can be biased. This reflects the standard tradeoff between bias and efficiency
present when specifying variables to omit from an econometric model. Some caution must be used when
imposing restrictions on the empirical model based on stylized models of the processing plant such as
USPS/OCA-T1-XE-1 at Tr. 23/8395 since this may ignore relationships that are present in the data (Tr.
23/8434-8435 lines 1-11). In the processing plants even processing pathways that seem obviously
impossible to exist apparently do. For example, even the movement of mail from manual back to
automated operations occurs (Tr. 36/12290 at line 15 and 36/12291 at lines 1-13).
Docket No. R2006-1
- 19 -
Initial Brief of the OCA
overall sorting of the mail volume in the plant. Equivalently, it is a measure of the flow
of capital services or capital input that is used to sort the pieces of mail processed in
that operation (see Roberts 2006, Section IV.C, pp. 35-38 for discussion). It is exactly
analogous to the flow of labor services, measured as man hours, that is used for the
labor demand estimation. Machine services and labor services are the two inputs used
to sort mail in the automated operations. The first can be measured by TPF and the
5
second by man hours in the operation. In manual operations there is only a single
input, labor hours, and, of course, TPH is not a measure of capital input in this
operation. Manual TPH is a measure of the number of pieces of mail that are handled
in the manual sorting operation and reflects both pieces entered directly into manual
operations and pieces rejected from automated operations. It is not useful information
for measuring volume elasticities because there is no way, using the MODS data, to
relate this variable to the volume of mail in the plant or the volume of mail in any
relevant processing category.
6
Finally, it is important to recognize that estimated labor-output elasticities will
reflect at least two mechanisms through which plant-level mail volume and hours are
related. One is the relationship that exists within each sorting operation. If the number
of pieces of mail arriving for processing in the BCS operation increases, then the
5
Of course, as witness Bozzo points out (USPS-T-12 at 27-29), TPF and man hours will be related.
Capital and labor inputs in any production process will be related, but measuring the relationship between
them does not provide an estimate of the relationship between output and labor use. Witness Bozzo’s
attempt to explain away this interpretation in USPS-T-12, p. 29, lines 1-8, by saying that it implies capital
input and labor input are identical makes no sense. As the use of the automated operation increases,
both labor hours and machine counts will rise. They do not have to increase by the same amount, and the
USPS estimates of elasticities of labor hours with respect to piece feedings show they do not. The capital
and labor inputs can scale up by different amounts as the use of the operation increases.
Docket No. R2006-1
- 20 -
Initial Brief of the OCA
increase in BCS labor hours will reflect any economies or diseconomies that exist within
the operation. This will capture the role of setup and take-down time, runtime, waiting
time, and overhead activities within the operation that are always emphasized in the
USPS testimony (USPS-T-12, Section II.F, p.26-32). But this is not the only
mechanism affecting the relationship between the volume of mail processed in the plant
and cost. A second mechanism is the scaling up or down of the individual sorting
operations in response to changes in a plant’s mail volume. During high-volume
periods, all operations are scaled up to meet the increase plant volume but, as the
operations staff has testified, the mix of sorting operations may change. Capacity
constraints and timing issues on DBCS equipment can result in mail being processed
on OCR equipment (Tr. 36/12282, line 19 to 12284, line 14). The use of manual sorting
increases at the holiday time to handle the increase in nonmachinable mail volume
(Docket No. R97-1, USPS-T-4 at 20). If the mix of sorting operations changes as the
volume of mail changes, this will be a second pathway through which volume changes
affect cost. This pathway will be reflected, as it should be, in the estimated output
elasticities in the OCA framework.
Once the relationships between the plant’s mail volume in each processing
category and hours in each sorting operation have been estimated using the MODS
data, the marginal cost of mail in each processing category can be calculated using the
formula in Roberts (2006, equation 5, page 10). Notice that this is, to use the USPS
names, a “constructed marginal cost” for the volume of mail in the processing category.
6
In particular, it is not a measure of the volume of non-machinable mail entered directly into the
manual sorting operations.
Docket No. R2006-1
- 21 -
Initial Brief of the OCA
It is estimated for the processing category using data on volume and hours for the
7
category of mail. Multiplying the marginal cost by the volume of mail in the processing
category gives the volume variable cost pool for each processing category. The terms
marginal cost and volume variable cost are used interchangeably in this discussion. It
is important to note that volume variable cost (VVC) pools are defined for each
processing category, not each sorting operation. This is appropriate. It is the sorting of
a volume of mail with a given set of processing characteristics that generates labor
hours and costs, and it is the marginal cost of processing an increased volume of mail
with a given set of processing characteristics that we are trying to measure. It is also
important to note that the VVC pool for each processing category will depend on the
mix of sorting operations used for that category of mail. For example, the VVC cost
pool for VIN will reflect a different mix of OCR, BCS, and manual labor inputs than the
VVC pool for VOUT because each uses a different mix of the three sorting operations.
The outgoing sorting will use a higher proportion of the OCR operation and a lower
proportion of the BCS operation than the incoming operation (USPS/OCA-T1-45(d), Tr.
23/8382). Since the construction of the VVC pools relies on the estimated volumehours relationship, the bottom of Figure 1 labels this step the “Cost-Pool Estimation
Stage.”
7
One additional strength of this modeling framework is that, since it provides estimates of the
marginal cost of mail in each processing category, it can be used as a basis for calculating price discounts
for the mail in different processing categories. For example, even if only the incoming and outgoing
categories are used, the differences in marginal costs provide a basis for calculating discounts for
presorted mail that skips the outgoing processing stage entirely. Since the USPS framework never
estimates marginal costs for processing categories, it can’t be used in this way. This is one area which
should be explored in the future if the Commission adopts the use of the OCA framework.
Docket No. R2006-1
- 22 -
Initial Brief of the OCA
Once the VVC pools have been constructed for each processing category, it is
necessary to allocate each cost pool across the 17 CRA rate classes. (Greater detail
on how this allocation should be done is described in USPS/OCA-T1-24, Tr. 23/83378340). This is represented on the left side of Figure 1 as the “Cost-Pool Allocation
Stage.” What is needed to do this allocation is information on how the system-wide
volume of mail in each rate class is linked to the volume of mail in each processing
category in the processing plants. The needed linkages are represented by the arrows
on the left side of the figure. If we make the assumption that a proportional expansion
in the volume of mail in each rate class, system-wide, results in an equal proportional
expansion in the plant-level volume of mail in each processing category, then each VVC
pool will be allocated across the rate classes using the share of each rate class in the
cost pool volume. Specifically, the VVC pool for the processing category of VIN will be
distributed across the rate classes using the shares of the volume of mail in each rate
class in VIN. These shares cannot be measured using the MODS data because this
data set does not contain any information on the division of mail by rate class. To
implement this fully will require a sampling method that samples the plant mail stream in
each processing category and constructs shares of each rate class by processing
category. Notice that these are not shares of piece handlings, as is used in the USPS
methodology, but shares of the number of pieces of mail. These shares are not
reported in the current USPS testimony and thus the appropriate distribution step
cannot be constructed at this time.
Until the required distribution key is constructed, a simpler, but more restrictive,
alternative, may be available to the Commission. Rather than allocate cost pools for
Docket No. R2006-1
- 23 -
Initial Brief of the OCA
VIN and VOUT separately, a single VVC pool is constructed for letters where the scaling
factor is the aggregated volume elasticity for letters (OCA-T-1, Table 4, p. 37, elasticity
of total letters with respect to a change in FHPIN and FHPOUT). Then a single
distribution key expressing the mail volume in each rate class as a share of the total
mail volume sorted in all piece handling operations in the plant could be used as an
alternative. This process does recognize in the estimation stage that VIN and VOUT use
different input mixes and therefore have different marginal costs but does not recognize
in the allocation stage that the mail contained in the VIN category may contain a
different mix of rate classes than the mail contained in VOUT. Ultimately, to take full
advantage of the flexibility present in the estimation stage, the distribution key will have
to be brought up to date to reflect the mix of rate classes in each processing category in
the plant.
The final point to recognize from Figure 1 is that the cost pool estimation and
allocation stages are separated “at the plant door.” The allocation stage only deals with
the relationship between system volume by rate class and plant volume by processing
category. This will likely be done using aggregate data based on sampling the mail
stream, so that there is a single distribution key which is used for all plants. The
allocation stage reflects what occurs externally to the plant and makes no assumptions
and places no restrictions on how the mail is processed within the plant. The
estimation stage only deals with the relationship that occurs within the plant and it
captures the entire relationship between the mail volume in each processing category
that arrives at the plant and the labor hours used in sorting. Failing to distinguish the
dividing line between processing relationships that occur within the plant and
Docket No. R2006-1
- 24 -
Initial Brief of the OCA
relationships between mail volumes that are external to the plant is a major problem
with the USPS framework and results in it missing one piece of the volume-hours
relationship that is needed to estimate the marginal cost of processing mail.
Docket No. R2006-1
- 25 -
Initial Brief of the OCA
Figure 2
Letter-Shaped Mail: USPS Model
Internal to Plant
External to Plant
Volume of mail by rate class
Volume of mail by
processing category
Sorting Operations
VRC1
TPF
OCR
VIN
Hours
VRC2
!
VOUT
TPF
BCS
VRC1 7
Hours
TPH
Manual
Hours
Cost-Pool Allocation Stage
Cost-Pool Estimation Stage
Docket No. R2006-1
2.
- 26 -
Initial Brief of the OCA
Witness Bozzo’s model of mail processing
The model of mail processing used in the USPS testimony since 1997 can be
represented in Figure 2. Roberts (2006), Section III, pp. 12-23 provides detailed
discussion of this model and its underlying assumptions about the sorting technology.
There are several differences from Figure 1. First, there is no role for the plant-level
volume of mail, either in total or disaggregated by processing category. This is
recognized by the fact that there are no arrows joining VIN and VOUT to any other part of
the diagram. From this fact alone, it is clear that, despite their name, the volume
variabilities estimated with the USPS model do not contain any information on the
relationship between the volume of mail received in the plant and labor use.
Second, the cost pool estimation stage, which is represented on the right side of
the diagram, relies on the estimated relationship between TPF/TPH and hours in each
of the three sorting operations. Causality is assumed to run from TPF/TPH to labor
hours in each operation. The vertical arrow from the TPF/TPH box to the labor-hours
box on the right side of Figure 2 represents the elasticity that is estimated by the USPS.
If there are three sorting operations, there are three elasticities estimated. Unlike the
OCA model in Figure 1, which had two processing categories and six output elasticities,
the number of elasticities in the USPS model is determined solely by the number of
sorting operations, three in this case, and is not affected by the number of processing
categories for the mail.
This also illustrates why the USPS framework is more restrictive than the
framework implemented in OCA-T-1. The OCA framework estimates a different
relationship between the volume of mail in each processing category and each sorting
Docket No. R2006-1
- 27 -
Initial Brief of the OCA
operation, while the separability assumption imposed on the sorting plants in the USPS
framework requires the whole relationship to operate through the intermediate step of a
cost driver in each operation. This issue is discussed in greater detail in Roberts (2006,
Sections IV.A and IV.B, pp. 27-35).
It is also the case that the USPS framework cannot provide an estimate of the
marginal cost of handling a piece of mail in a processing plant. Given the MODS data
on TPF/TPH and hours, which the USPS model uses, there is no way to calculate how
the three sorting operations are combined to handle the processing of a given category
of mail. Thus, there is no way to tell that an increase in VIN, for example, will have a
different impact on labor use of OCR relative to BCS than an equivalent increase in
VOUT. In other words, it is impossible to use the estimates to measure the relative costs
of processing two different categories of mail in the plant.
Once the elasticity between piece feedings and labor hours is estimated for each
sorting operation, a VVC pool for each sorting operation is constructed by scaling the
labor expenditure in the cost pool by the estimated elasticity. It is important to
recognize that the size of a VVC pool will depend on an estimated elasticity that
captures only the relationship between TPF/TPH and hours within that sorting
operation. Unlike the output elasticities in the OCA model in Figure 1, the elasticity will
contain no information on the relationship between the volume of mail in the plant and
the level of use of the sorting operation.
The final step of the USPS methodology allocates the cost pools across rate
classes using a distribution key. If the USPS methodology is to be capable of
measuring the relationship between mail volume and the usage of each sorting
Docket No. R2006-1
- 28 -
Initial Brief of the OCA
operation, it will have to occur in the cost-pool allocation stage. Unfortunately, the
USPS methodology is particularly restrictive at this stage. For each sorting operation,
the distribution key measures the share of piece feedings that are accounted for by
each rate class. This cost allocation is represented in Figure 2 by the long arrows from
the separate boxes identifying the volume of mail for each rate class (VRC1, VRC2, …
VRC17) to the piece feedings in each sorting operation. These shares sum to one
across rate classes for each sorting operation. The relationship measured by the
distribution key is labeled the “Cost-Pool Allocation Stage” in Figure 2. Note, in
particular, that it is not limited to the relationship between system-wide mail volume and
plant volume, as the OCA framework is. It does not end at the plant door, but rather
crosses into the plant and involves an assumption about the response of the individual
sorting operations. It is at this point that any chance of allowing flexibility in how the mix
of sorting operations responds to changes in mail volume is lost.
The assumption underlying this distribution method is that a proportional
increase in the volume of mail in all rate classes in the system will result in an equal
proportional increase in the piece feedings in every sorting operation. This assumption,
which is labeled the proportionality assumption, is described as a first-order
approximation to a more general relationship (R2005-1, USPS-T-12, p. 20) or as
measuring the effect of small, local changes in volume on the cost driver in each
operation (R2005-1, USPS-RT-3, p.11). However described, it is not an innocuous
assumption. Figure 2 shows that the proportionality assumption is an assumption
about how the plant responds to changes in system volume that, presumably, work their
way through the volume of mail arriving in the plant. All operations in the plant are
Docket No. R2006-1
- 29 -
Initial Brief of the OCA
scaled up proportionately in response to a system-wide increase in mail volume.
Another way of stating this assumption is that the plant uses the sorting operations,
OCR, BCS, and manual in fixed proportions. Thus, if a change in the volume of mail
processed in the plant results in a different mix of sorting operations being used, the
USPS framework will miss the cost implications of this change. In particular, if the shift
is to higher cost operations such as manual, the USPS framework misses a potentially
important source of increased cost that is caused by the plant’s response to increased
mail volume.
Looking over the whole USPS framework as illustrated in Figure 2, it can be
seen that the limitation of the methodology is that it skips over the relationship between
plant volume and the relative use of the sorting operations. Thus it can measure, at
best, only one of the two mechanisms by which mail volume can affect cost in a plant.
It does not recognize any mechanism through which an increase in plant volume results
in a change in the mix of sorting operations. In particular, if the increase in plant
volume results in the more intensive use of higher-cost sorting operations, such as
manual, the USPS framework will miss this source of rising marginal processing cost.
At a minimum this response of the mix of sorting operations to changes in plant mail
volume is something that should be examined using the plant-level MODS data, rather
than being assumed fixed.
3.
Implications of the model of mail processing for measurement and
allocation of volume variable cost in mail processing.
One issue which is clearly illustrated in Figures 1 and 2 is that the USPS
framework and the OCA framework are not estimating the same economic relationships
from the MODS data. By construction, the USPS estimates of elasticities of labor hours
Docket No. R2006-1
- 30 -
Initial Brief of the OCA
with respect to TPF in an operation can only estimate economies or diseconomies that
arise within a single operation. If production in the operation is driven by a combination
of a fixed setup time followed by a long run time, then it is not surprising to see
elasticities of labor hours with respect to TPF that are less than one, as the hours used
in setup get spread over a larger number of piece feedings. In fact, as illustrated in
Table 1, the preferred USPS estimates of these elasticities from the last four rate cases
are virtually always less than one. In contrast, the OCA model encompasses a much
broader range of activities that can affect the relationship between mail volume and
labor hours in the plant. This includes the economies or diseconomies that arise at the
operation level, but also the effect of a changing mix of sorting operations as plant mail
volume changes or as capacity constraints on operations bind.
Docket No. R2006-1
- 31 -
Initial Brief of the OCA
Table 1
USPS Preferred Estimates
(standard errors in parentheses)
Sources
R 2000-1
USPS-T-15
Table 6,
p. 119
Table 7,
p. 120
p. 121, line 12
R 2001- 1
USPS-T-14
Table 9, p. 61
Table 10, p. 62
Table 11, p. 63
Table A-1, Col. 1,
p. 71
R 2005-1
USPS-T-12
Table 5, p.
48
Table 6, p.
49
Table 7, p.
50
Table 1, p. 3
R 2006- 1
USPS-T-12
Table 26,
p. 99
Table 1, p. 3
Letters
OCR
BCS
BCS/DBCS
BCS outgoing
BCS incoming
Manual
LSM
.751 (.038)
.895 (.030)
---.735 (.024)
.954 (.022)
.77 (.06)
.94 (.05)
.87 (.05)
--.58 (.04)
.90 (.06)
.78 (.06)
.90 (.05)
.85 (.07)
--.87 (.07)
--
.78 (.05)
-.88
1.06 (.06)
.82 (.07)
.89 (.09)
--
Flats
FSM
FSM 1000
AFSM 100
Manual
.817 (.026)
--.772 (.027)
.74 (.05)
.74 (.05)
-.71 (.05)
1.01 (.04)
.73 (.03)
1.03 (.09)
.90 (.13)
-.72 (.03)
.99 (.08)
.94 (.07)
Parcels
SPBS
Manual
.641 (.045)
.522 (.028)
.66 (.05)
.44 (.04)
.77 (.06)
.78 (.18)
.87 (.05)
.80 (.18)
Priority
Manual
.522 (.025)
.55 (.05)
.76 (.08)
.75 (.09)
Cancellation/Prep
Composite
.549 (.037)
.760*
-.71
.46 (.07)
.83
.50 (.07)
.85
Docket No. R2006-1
- 32-
Initial Brief of the OCA
A significant limitation of the USPS framework for mail processing costs is that it
assumes that the plant manager cannot utilize the sorting operations in different
proportions as plant volume expands or contracts. At a minimum, the assumption that
sorting operations scale up and down in fixed proportions in response to volume
changes should be tested empirically. Using the USPS framework, and testing it in an
econometrically appropriate way that recognizes the endogeneity of the key TPF
variables, has only been done by witness Roberts. OCA-T-1, Table 1, reports evidence
that the proportionality assumption does not hold.8 In rebuttal testimony (USPS-RT-5,
Table 10) Dr. Bozzo claims to provide evidence that the proportionality assumption is
valid, but his evidence does not show this. The proportionality assumption is the
requirement that, in every operation, the elasticity of TPF with respect to plant volume
equal one. Dr. Bozzo reports a specification in which the sum of TPF across all sorting
operations has an elasticity with respect to plant volume equal to one. This does not
imply that all the individual operations have elasticities equal to one, which is what the
proportionality assumption implies. His evidence does not speak to the issue of how
the mix of sorting operations changes in response to changes in plant mail volume and
Dr. Bozzo agrees (Tr. 36/12481 at 6-13). While the purpose of Table 10 is unclear, it
certainly cannot be interpreted as evidence supporting the proportionality assumption.
8
Dr. Bozzo also criticizes the specification of the regression model underlying Table 1 in OCA-T-1
(USPS RT-5, Section VI.B.2). The purpose of those regressions was not to estimate a complete model of
the mix of operations utilized in the plant (although the equations would be the starting point for that
model) but rather to show, in the clearest, simplest way possible, that the proportionality assumption which
underlies the USPS analysis is not true in the plant-level data. The results in Table 1 demonstrate that
clearly. The USPS has not provided any empirical evidence justifying the use of the proportionality
assumption, and raising criticisms of this evidence still does not establish that the proportionality
assumption is correct.
Docket No. R2006-1
- 33-
Initial Brief of the OCA
If the use of higher cost sorting methods, particularly manual, rises with plant
volume, then an important source of rising marginal cost is missed. The practical way
that this error appears in the estimates of VVC is that the labor demand elasticities with
respect to piece feedings that are used to scale the cost pools to construct VVC pools
underestimate the scaling factor for operations that are used more heavily when volume
increases. Since this is likely to be manual, the VVC pool for the manual operation is
too small. Since manual hours are a substantial fraction of total hours, 35.8 percent in
letter processing and 28.6 percent in flats processing in 2005, this is likely to lead to
substantial underestimates of marginal cost.
Since the within-plant relationship between the volume of mail and the use of the
different sorting operations is a potentially important component of the marginal cost of
processing mail, it should not be restricted, or eliminated, unnecessarily. Restricting
the potential relationship between mail volume and labor hours as the USPS framework
does, makes it inappropriate as a basis for constructing and allocating volume variable
processing costs across rate classes. Overall, the USPS methodology needs to be
generalized so that it does not place a priori restrictions on the way that the plant
responds to changes in mail volume. The framework utilized in OCA-T-1 does not
restrict the technology in this way.
B.
Response to Rebuttal Testimony
The rebuttal testimony offered by Dr. Bozzo includes a number of criticisms of
the testimony presented by Dr. Roberts in OCA-T-1 and a great deal of speculation
about the motivation behind that testimony. There appear to be four themes that run
through the relevant parts of USPS-RT-5. One of them, the need for a distribution key,
Docket No. R2006-1
- 34-
Initial Brief of the OCA
has already been discussed above. It can be added that the need for a distribution key
to allocate cost pools to rate classes has always been obvious and Dr. Roberts has
never claimed that one is not necessary. The remaining three themes involve the
definition of processing categories, the use of quarterly dummy variables, and the
sample period analyzed. In each case, the criticisms reflect errors in describing the
OCA model or an incomplete reading of the papers written by Dr. Roberts.
1.
Definition of processing categories
In OCA-T-1, Dr. Roberts divides the volume of letter and flat mail processed in
the plant into volume in two processing categories, the outgoing operation and the
incoming operation, and estimates separate output elasticities for each category.
These are reported in OCA-T-1, Table 4, p. 37. He also further disaggregates letter
mail into four categories: incoming with a barcode, incoming without a barcode,
outgoing with a barcode, and outgoing without a barcode, and estimates separate
output elasticities for each. These are reported in OCA-T-1, Table 5, p. 42. When all
four categories are used (Table 5, Panel D), the number of output elasticities estimated
increases to 12, compared with 6 when only two processing categories are used. In
the discussion of these results in Section VIII.C, pp. 41-44, Dr. Roberts points out that
as the level of disaggregation of the processing categories increases, the precision of
the individual estimates decreases. However, the aggregated elasticities for the
incoming and outgoing categories are not highly affected by whether two or four
processing categories are defined. In addition, the aggregated elasticity for the whole
plant is basically the same whether one, two, or four processing categories are defined.
What this demonstrates is that the MODS data is useful for estimating volume-hours
Docket No. R2006-1
- 35-
Initial Brief of the OCA
relationships for some fairly broad processing categories, but will be more problematic
as the level of disaggregation increases.
Dr. Bozzo argues that this specification renders the results useless because the
nonbarcoded processing category includes both nonmachinable pieces that have to be
entered directly into manual sorting and machinable pieces that will be entered into one
of the address recognition processing steps (USPS-RT-5, pp. 59-61). At the same
time he argues that the FHP entered directly into the manual processing stage is a
minority of the overall category and so it will be difficult to distinguish the “costs for ‘true’
nonautomation mail from automation mail.” This criticism is misdirected. Dr. Roberts
advocated the use of two-output models because, as summarized above, further
disaggregations resulted in imprecise parameter estimates. In addition, it will be difficult
to estimate output elasticities for processing categories that contain very little mail
volume. The concern about distinguishing the costs of nonautomation from automation
mail is puzzling, since the USPS model cannot distinguish these differences, and it has
never been raised as a limitation in their previous testimony.
Dr. Bozzo also provides evidence that the disaggregation into what he defines as
manual and automated output categories produce “better-behaved” results when the
manual output is eliminated from the automated operations labor demand curves.
When doing this he ignores the distinction between incoming and outgoing operations
and specifies the model differently than the OCA model. He then finds that the old
model is more “inefficient” than the USPS models at uncovering a relationship that is
basically imposed in the USPS model. A more useful way to proceed would have been
to maintain the distinction between incoming and outgoing processing categories, since
Docket No. R2006-1
- 36-
Initial Brief of the OCA
those have been shown to have different volume elasticities, and then demonstrate that
separating each of those further into (i) nonmachinable, (ii) machinable without a
barcode, and (iii) machinable with a barcode processing categories could be estimated
and produce precise parameter estimates. This amounts to a further disaggregation of
the four-output model used in OCA-T-1 into a six-output model. If that model could be
estimated, that would be useful to know, but Dr. Bozzo’s discussion does not shed any
light on the issue.
2.
Quarterly dummy variables and shifts in sorting technology
The model estimated in Roberts (2006) and in OCA-T-1 does not include
quarterly dummies in the labor demand function. Quarterly dummy variables are
justified in the labor demand models if the technology for mail sorting changes
systematically from quarter to quarter. They are not justified simply because the
volume of mail, or the composition of mail between, for example, barcoded and non
barcoded processing categories, changes. Changes in the volume and composition of
mail should be controlled for with the output measures used in the model and not by
including quarterly dummies that represent changes in the technology.
There is no evidence that the technology available in the plant shifts on a
quarter-by-quarter basis. Dr. Bozzo agrees (Tr. 36/12473 at 7-24). The only
operational testimony which the USPS offers that is relevant to this point is presented in
Docket No. R97-1 by witness Moden (USPS-T-4). He describes the seasonal effect of
the holidays as a change in “the volume and characteristics of the mail.” (R97-1, USPST-4 at 20). There is more mail volume to be processed, and the characteristics of the
mail change so that there is a higher proportion of nonmachinable mail. The response
Docket No. R2006-1
- 37-
Initial Brief of the OCA
of the plants to this exogenous shift in the mail stream is to use more manual labor,
particularly more temporary clerks. In other words, as the volume of mail rises and the
composition of the mail changes, the USPS uses a different mix of inputs, in particular
higher cost inputs that will result in higher processing costs. This is not a change in
technology that should be modeled with quarterly dummies. This is a change in the mix
and quantity of mail processed, not an exogenous shift in the processing technology,
and its impact should be recognized as a volume-driven change in cost.
The model implemented in OCA-T-1 does exactly this. The volume of mail in
each of several processing categories is included as the output controls in the labor
demand equations. The increase in volume in the holiday season is measured directly
using plant FHP, and the change in the mix of mail characteristics is measured directly
by disaggregating FHP into multiple processing categories. In OCA-T-1, this
disaggregation includes distinguishing incoming and outgoing mail as well as barcoded
and nonbarcoded. The holiday surge in nonmachinable mail would be measured
directly as an increase in the volume of the outgoing, nonbarcoded category. How the
plant responds to this change in the mail stream is exactly the effect that should be
measured by the volume elasticities in the labor demand curves.
Dr. Bozzo argues that Dr. Roberts is inconsistent in his treatment of quarterly
dummies over time. This is true. Roberts (2002) included quarterly dummies in the
labor demand models based on the USPS justification. As he explains in Tr. 23/8414 at
17 through 8418 at 9, that paper focused on a number of other specification problems
with the USPS model (the failure to incorporate volume measures as output, the use of
the manual ratio as an explanatory variable, and the failure to recognize the
Docket No. R2006-1
- 38-
Initial Brief of the OCA
endogeneity of TPF in the demand equations among other issues), and he did not
question the USPS justification for the use of quarterly dummies. In preparing his 2006
paper he recognized that the specification of quarterly effects was an important issue
and systematically explored their justification and implications for the results. This is
explained in Roberts (2006), Section V.E, and his conclusion was that quarterly
dummies were inappropriate in a model of labor demand. He also summarized the
effect that they had on the results in that paper on page 59 and in Table 8, p. 74.
Finally, in the current testimony (OCA-T-1, Section VII, pp. 27-29) he shows that
quarterly dummies could be also be used as instrumental variables. Based on
statistical tests for relevance and overidentification, they are useful and appropriate in
six out of 10 sorting operations using two-output processing categories (USPS/OCA-T112, Tr. 23/8309-8313) and in all but one sorting operation when more disaggregated
processing categories are used for letters (USPS/OCA-T1-34(a), Tr. 23/8361-8362).
Finally, Dr. Roberts also shows that, when quarterly dummy variables are
included in the labor demand equations, their magnitudes are unrealistically large
(USPS/OCA-T1-44, Tr. 23/8379-8380). For example, between different quarters of the
year, they show a 16.7 percent swing in manual labor use, a 15.0 percent swing in
DBCS labor hours, a 13.9 percent shift in manual flat hours and a 12.0 shift in AFSM
hours. Given that there is no testimony indicating that the technology changes from
quarter to quarter, it is hard to square these estimates with operational reality. More
realistically, they reflect quarter-to-quarter fluctuations in average labor hours that are
due to differences in the average mail volume. Overall, the discussion and treatment of
the quarterly data variation in Dr. Roberts’ papers demonstrates substantial
Docket No. R2006-1
- 39-
Initial Brief of the OCA
development over time and has better illuminated a set of issues that have not been
addressed in the USPS testimony.
3.
Sample period analyzed
Dr. Bozzo criticizes the use of the sample period 2002-2005 rather than the
period 1999-2004 used in Roberts (2006). As originally noted in Roberts (2006) pp. 5253, the volume elasticities for the FSM1000 and manual flat sorting operations were
sensitive to the time period, suggesting that the technology, specifically the relationship
between plant volume and hours in each operation, had changed as a result of the
introduction of the AFSM. In preparing the testimony in OCA-T-1, Dr. Roberts explored
this issue in more detail and found that introduction of the AFSM was responsible for a
change in coefficients in other operations. As a result, he limited the sample for flat
sorting to the plants and time periods in which the AFSM was in full operation. This is
appropriate if the change in technology results in different output elasticities for the
operation. It is also important to recognize that the OCA model is a model of a plant,
not a sorting operation, and introduction of a new technology as significant as the
AFSM is very likely to affect the output elasticities for other operations. Dr. Roberts
shows in Table 7, Panels A and D that the output elasticities do differ between
observations in which the AFSM was utilized and observations where it was not. Given
this evidence, it would be inappropriate to pool the observations from all plants and time
periods.
The sample period for letter sorting operations was also limited to the period
2002-2005. This was done to try to remove one source of potential criticism of the
analysis, that the samples were being selectively chosen for some reason. By using an
Docket No. R2006-1
- 40-
Initial Brief of the OCA
identical time period for all the analysis this would seem to remove the issue from
consideration, but apparently it did not. Finally, in his original testimony, (Table 4 Panel
D p. 37), Dr. Roberts reports a complete set of output elasticities for letters using the
extended time period 1999-2005 and finds no difference from those estimated on the
shorter time period.
C.
Conclusion
The framework underlying the USPS model of mail processing costs USPS-T-15 is
flawed and should not be used as the basis for measuring marginal or volume variable
costs or allocating costs for across classes of mail. The framework eliminates a
potentially important channel through which changes in the volume of mail in a plant
can affect labor hours and result in rising marginal processing costs. As shown in Table
1, the USPS estimates of volume variabilities are virtually always less than one and this
is consistent with the limitations inherent in the framework.
The alternative framework estimated in OCA-T-1 directly measures the
relationship between mail volume and labor hours at the plant level, and this is a major
source of the difference in volume variability estimates. In the case of letter processing,
the preferred estimate for the elasticity of total labor use with respect to an increase in
total plant FHP is 1.276 (Table 4, Panel B, p. 37), and this estimate is robust across a
number of alternative specifications. For priority mail the estimates are 1.184 and
1.033 depending on whether the plant uses the SPBS technology (Table 9, p. 54). For
cancellation operations the elasticity is .918 (Table 11, p. 56). In the case of flat sorting
in plants that use the AFSM technology the elasticity of total labor use is 1.098 (Table
7, Panel B, p.47). However, the underlying estimates for flat sorting operations,
Docket No. R2006-1
- 41-
Initial Brief of the OCA
particularly manual, are not estimated precisely and it is advisable to wait until the
AFSM operation has been in place longer and all adjustments in staffing of other
operations has occurred before estimating the elasticity for flat sorting operations.
Overall, it is clear that the estimates using the OCA framework are consistently larger
than those produced by the USPS model and do not support the Postal Service
arguments that there are significant economies of scale in mail processing.
We recommend that the Commission reject the modeling approach advocated
by the Postal Service and adopt the alternative presented in OCA-T-1.
Docket No. R2006-1
II.
- 42-
Initial Brief of the OCA
ECONOMETRIC ANALYSIS OF CITY CARRIER COSTS
A.
Witness Bradley’s Volume Variabilities Presented in Docket No. R2005-1
Are Flawed.
In Docket No. R2005-1 witness Bradley presented volume variabilities by shape
9
for City Carrier mail delivery. Witness Bradley testified in the current case that there
was insufficient time to revise the city carrier street time costing methodology which he
had presented in R2005-1 for presentation in the current case.10 Accordingly, the
Postal Service has adopted the conclusions and volume variabilities from that study for
use in the current case.
OCA witness Smith has reviewed witness Bradley’s City Carrier study and has
testified on the deficiencies in that study in the estimation of volume variabilities for
letters, flats, collection volume, small packages, sequenced mail, large packages, and
accountables. He found a number of deficiencies:
(1) The density variable is incorrectly used. The variable is not needed from a
theoretical viewpoint.
(2) Witness Bradley also computed the density variable incorrectly.
(3) The underlying database for the original study appears to suffer from
multicollinearity, as evidenced by high values for the Variance Inflation Factor
(VIF). This appears to be the cause of the generation of negative volume
variabilities in some cases—a result that is clearly incorrect.
(4) Furthermore, multicollinearity problems in the database can result in the
computation of regressors which are unreliable; accordingly, the subsequent
9
Testimony of Michael D. Bradley, Docket No. R2005-1, USPS-T-14.
10
Testimony of Michael D. Bradley, Docket No. R2006-1, USPS-T-14.
Docket No. R2006-1
- 43-
Initial Brief of the OCA
use of the regressors in the computation of volume variabilities also renders
the estimates of volume variability unreliable.
(5) Witness Smith recognized that the Commission presented a thorough and
comprehensive analysis of the deficiencies of the City Carrier Street Time
Survey (CCSTS) database in Docket No. R2005-1.
(6) Witness Smith demonstrated in his response to POIR 25 that problems
caused by multicollinearity could be reduced by eliminating selected cross
product terms associated with the variable “spr”.
(7) Witness Smith also demonstrated that the volume variability estimation
problem could be recast in terms of what may be called the “three bundle
approach,’ wherein carrier costs are modeled primarily in terms of Delivery
Point Sequenced (DPS), Cased, and Sequenced mail in addition to mail
collected at the customer’s site. Witness Bradley subsequently
acknowledged the relevance of using DPS and cased mail in modeling city
11
carrier delivery.
(8) Witness Smith has concluded that changing delivery technologies, as
acknowledged by witness Bradley, lead to the conclusion that witness
Bradley’s volume variabilities based on the 2002 database may be obsolete.
(9) Witness Smith also discusses the potential use of the Delivery Operations
Information System (DOIS) database. Although the database was not
received until late in the case, preliminary results indicated that it holds great
promise. He further indicated that the use of the Delivery Operations
Information System (DOIS) data could reduce the major data problem in this
case—multicollinearity.
(10) Witness Smith did not need to address the estimation methodology and
results related to the volume variabilities of parcels and accountables in
detail. As he discussed, parcels and accountables should be 100 percent
attributable in cost, for the Postal Service has been able to segregate the
costs (delineated by time) and has shown that the costs would not be
incurred without the associated mail volumes.
11
Michael D. Bradley, Jeff Colvin, and Mary K. Perkins, “Measuring Scale and Scope Economies
with a Structural Model of Postal Delivery,” presented at Conference on Postal and Delivery Economics
(2006, Bern, Switzerland) in Liberalization of the Postal and Delivery Sector, Michael A. Crew and Paul R.
Kleindorfer, editors, Elgar, 2006.
Docket No. R2006-1
B.
- 44-
Initial Brief of the OCA
The Methodology in the Original Carrier Cost Study Needs Improvement.
1.
The density variable should not have been used in the econometric
equations.
Witness Smith concluded that the use of the density variable was incorrect in the
current estimation process and that it should be eliminated from the analysis.12 The
City Carrier equations presented by witness Bradley model delivery time as a function
of the quantities of the various types of mail being delivered, delivery points, and
density. Inclusion of quantities of mail as a driver is obvious; mail is an output of the
delivery process. In addition, delivery points clearly should also be included in a
modeling effort, for carriers need to pass each delivery point in order to complete the
route: one of the outputs of the delivery process is the passage by a carrier past a
delivery point whether or not any mail is delivered.
One could propound the density variable as accounting for differences in the
physical layouts of ZIP codes—e.g., congestion, urban/suburban/rural, and other
factors. Inclusion of the density variable, however, is theoretically incorrect. In
modeling mail delivery there are a number of types of functions which could be
13
estimated :
12
13
(1)
Production function (Equation 8.2.1 in Intriligator): output as a
function of inputs.
(2)
Factor Demand function (Equation 8.2.28 in Intriligator): Derived
demand for inputs as a function of the price of inputs and price of the
product.
Direct Testimony of J. Edward Smith, Docket No. R2006-1, OCA-T-3 at 3.
POIR No. 25 at 13. The equations discussed are from Michael D. Intriligator, Ronald G. Bodkin,
Cheng Hsiao, Econometric Models, Techniques, and Applications, Second Edition, Prentice Hall, 1996.
Docket No. R2006-1
- 45-
Initial Brief of the OCA
(3)
Cost function (Equation 8.2.57 in Intriligator): Cost as a function of
input costs and output.
(4)
Cost curve (Equation 8.2.14 in Intriligator): Cost as a function of
output.
Witness Smith testified that in the modeling of an economic process one
generally expects to see the maximization or minimization of a process subject to some
type of constraint. This results in the development of various equilibrium conditions.
Witness Smith concluded that no theoretical analysis of the underlying economic
processes of mail delivery has been explicitly hypothesized in conjunction with the
modeling effort (a conclusion disputed by witness Bradley14). Witness Smith concluded
that witness Bradley’s equations model a cost function, with cost (measured in terms of
time) as a function of output (pieces of mail delivered or collected plus coverage of the
delivery points).
Witness Smith, therefore, is disputing the suitability of drivers which should enter
the estimating equation. There are several variables that impact the cost of mail
delivery: e.g., square miles of the subject ZIP code, route miles for the delivery of mail
in a ZIP code, environmental factors, and route structures as well as other potential
variables. As witness Smith indicated,
It is not unusual to develop equations based on the ad-hoc selection of
variables; there are numerous examples of ad-hoc estimation efforts in
the Operations Research literature, and the equations have in many
cases met the needs for which they were developed. Ad-hoc
specification is not necessarily bad, even though the equations are not
directly consistent with economic theory.15
14
Tr. 34/11563. Although not presented in this case, one can find some theoretical discussion on
this issue by witness Bradley et. al., “Measuring Scale and Scope Economies with a Structural Model of
Postal Delivery.”
15
POIR No. 25, at 12.
Docket No. R2006-1
- 46-
Initial Brief of the OCA
Accordingly, one would be free to include a variety of variables in an ad-hoc estimating
equation. However, in estimating a cost curve, one would use variables which are
drivers. Density is a variable which is subsumed in the production process. Witness
Smith has indicated that a cost curve for a firm models cost as a function of output.
The costs are a result of production decisions and tradeoffs made by the firm’s
management in an attempt to minimize overall costs. Accordingly, adjustments made
to account for a variety of factors are taken into account in arriving at a cost function
based on a production function. Witness Crowder has summarized the adjustment
process:
There are three principal workload variables affecting city carrier
costs: volumes, possible delivery points, and square mileage
describing a delivery unit’s service territory. The latter by definition is
ZIP-code square miles. Postal managers reconfigure and add routes
as necessary to minimize costs, subject to any operational constraints
(daily carrier hours for example). In this route restructuring, they are
responding to changes in these three primary variables.16
Witness Crowder is correct in stating that square miles are an important consideration;
so probably for that matter do other factors impact the cost of mail delivery. However,
as indicated by witness Crowder, postal managers perform a minimization process that
takes the multitude of factors into account; accordingly, square mileage, density, etc.
are variables addressed by the least cost solution of the production process and are not
necessary in the estimation of the cost curve.
Witness Bradley has described this concept:
16
Tr. 34/11647-8.
Docket No. R2006-1
- 47-
Initial Brief of the OCA
...duality theory establishes that there is a duality between
production cost, in the sense that for every technology there is an
associated cost function, and for every cost function there is an
associated technology. TR 11563 ….Consequently, researchers
no longer need to explicitly (or mathematically) derive the cost
function when they want to investigate characteristics of the
technology…
The level of costs is a result of decisions made by the firm’s managers to attain a
level of output for various levels of inputs.17 There are a wide number of decisions to
be made in the production process:
The equilibrium of the firm in the long run, when both inputs can be freely
varied, is at the tangency of an isocost to an isoquant. Only at such a
point is output maximized for a given cost, or, equivalently, is cost
minimized for a given output. The former follows by moving along any
one isocost: if at any one point it crosses an isoquant it is possible to
increase output with no additional cost—by moving toward the tangency
point. Similarly, moving along any one isoquant, if at any one point it
crosses an isocost, it is possible to decrease cost while holding output
constant—by moving toward the tangency point. The locus of tangency
points is the set of possible equilibrium points for the firm: it is called the
expansion path and is characterized by the equality of slopes of isocost
and isoquant. From the above results on these slopes, the geometric
tangency is in fact equivalent to the algebraic condition (89.2.7), stating
that, for profit maximization, the marginal rate of technical substitution
18
must equal the ratio of wages.
Based on the achievement of economic efficiency and the use of duality, one
can obtain a cost function. Accordingly, it appears that witness Bradley’s model is that
of a cost curve with C = f(y1, y2, …yn). Types of mail--e.g., letters, flats, sequenced
mail, collection volume, and small packages—as well as delivery points are the outputs
whose cost is being measured. From a theoretical viewpoint the use of the density
17
OCA-T-3 at 4.
18
Intriligator et. al., op. cit. at 278.
Docket No. R2006-1
- 48-
Initial Brief of the OCA
variable is wrong; it should be eliminated from the equations. 19 Insofar as a variety of
cost related factors are balanced and adjusted in arriving at a production point (termed
the tangency of an isoquant and isocost) it would appear that density is subsumed in
reaching a solution and is not needed even as a control variable.
Intriligator,20 has noted that “The modern approach to the theory of the firm is
based on the concept of duality….” He has outlined the variables used in a production
function (Equation 8.2.1), cost function (Equation 8.2.57), cost curve (Equation 8.2.14),
and factor demand function (Equation 8.2.28). The density variable does not have the
characteristics of any of the variables referenced by Intriligator in any of the functions
cited. The density variable does not represent output, factor prices, or product prices.
Rather, the density variable appears to measure delivery characteristics that are
subsumed in some type of maximization or adjustment process for efficient City Carrier
delivery; the process is then modeled by an equation with the economically relevant
variables.
21
Finally, it should be noted that many of the operational or environmental
characteristics allegedly explained by the density variable are also explained by the
delivery points variable when the variable is disaggregated. The disaggregated points
yield statistically meaningful results. Delivery points is a variable which is clearly
relevant and which could be disaggregated in an analysis: there are substantial
19
Equations with the density value are included in Table 1of witness Smith’s testimony, OCA-T-3 at
10, for purposes of comparison. Witness Smith, however, does not advocate their use.
20
21
Intriligator et al., op.cit., 283.
Although witness Smith testifies against the use of the density variable, his tables of equation
results contain equations with and without density in the interest of full disclosure, recognizing that the
Commission requires that all relevant work and alternatives be presented.
Docket No. R2006-1
- 49-
Initial Brief of the OCA
differences between curb, central, NDCBU, and “other” delivery points at the residential
and business level. For example, witness Bradley acknowledges that a neighborhood
cluster box could have 12 to 20 delivery points at a stop, and central deliveries might
have 50 to several hundred delivery points at a stop.22 The focus on density is
misplaced. Differentiation among the various types of delivery points is a much better
match for the way that deliveries are made across distances, or that may be co-located,
and drive carrier costs. Density is a very crude tool. Delivery point technology
information gives more precise data of the type needed.23
2.
In rebuttal testimony and during oral cross-examination, witness
Bradley was unsuccessful in demonstrating the relevance of the
density variable.
Witness Bradley presents a discussion of two hypothetical ZIP codes, one which
is 10 miles square, and a second that is 100 miles square.24 By his own computations,
both ZIP codes have routes that violate the 8 hour rule, and his argument is based on
the design of delivery routes that would never exist. Although it is easy to hypothesize
abstract situations to support a position, there is little relevance between the imagined
example and actual postal deliveries.
During cross-examination on the two squares examples, it was shown that
conclusions at variance with witness Bradley’s example could be obtained simply by
varying the placement or numbers of delivery points within the imaginary ZIP codes.
What was learned from this example was best summarized by witness Bradley himself:
22
Tr. 34/11615.
23
This concept was acknowledged in the Roy article cited by witness Bradley. Tr. 34/11610.
24
Tr. 34/11561.
Docket No. R2006-1
- 50-
Initial Brief of the OCA
Well, again I ‘m not doing any analysis of this type of zip code.
The only thing I’m learning here is perhaps it’s not a good idea to
25
give simple examples.
Witness Bradley also cited Roy’s article26 in support of the use of the density
variable. Witness Bradley implicitly confirmed that the article included the following
statement:
“Geographic density is often highlighted as the main factor
influencing delivery costs.”27
Subsequently, the Roy article indicated that
The initial results therefore tell us that population density is not
enough of a cost factor—far from it—to estimate delivery cost. The
inclusion of the single variable leads to extremely wide error
28
interval in estimating the cost of outdoor delivery work
In response to additional questions about the article and whether the magnitude
of the volume variability related to the density variable was small, witness Bradley
reiterated his contention about the importance of the density variable, but did not show
that the density variable was significant on the basis of the Roy article.29 Accordingly,
use of the density variable has not been shown to be appropriate, either from a
theoretical or computational basis.
25
Tr. 34/11630.
26
“Technico-Economic Analysis of The Costs of Outside Work in Postal Delivery,” Roy, B., in
Emerging Competition in Postal and Delivery Services, Crew, M. and Kleindorfer, P., (eds.) Kluwer 1999.
27
Tr. 34/11604.
28
Tr. 34/11605
29
Tr. 34/11606-09.
Docket No. R2006-1
3.
- 51-
Initial Brief of the OCA
Witness Bradley did not correctly compute the density variable.
Witness Smith concluded that the density variable is not correctly computed in
witness Bradley’s model.
30
This problem was further evident from the response given
by witness Bradley to interrogatory OCA/USPS-T14-2.31 It appears that density for a
ZIP code as computed by witness Bradley is a function of the number of routes
reporting deliveries in a given ZIP code and can vary from day to day. The total number
of delivery points, presumably indicative of area congestion and/or other physical
layout, does not appear to be correctly delineated in the density computations. This
was subsequently confirmed by witness Bradley:
As was discussed in Docket No. R2005-1, the CCSTS database
sometimes does not include reporting for all routes in ZIP code on a
given day. This means that the delivery points associated with
recorded delivery time and volumes can vary from day to day within a
ZIP code. Ideally, one would know the square mileage for the area
served by the included routes and would calculate density by dividing
that square mileage into the delivery points on the same routes.
Unfortunately, that data is not available and an approximation must be
made. In Docket No. R2005-1, I propose an approximation that would
account for the variation in routes. Dr. Smith disagrees, and would
prefer an approximation that does not vary from day to day, even
though the actual density associated with the recorded delivery time is
likely varying. 32
In computing density for a ZIP code, the correct approach would be to divide the land
area in a ZIP code by the maximum number of delivery points in the ZIP code. Density
should be a constant for a ZIP code, rather than being dependent on the number of
routes reporting.
30
OCA-T-3 at 8.
31
Tr. 13/3788-89.
Docket No. R2006-1
4.
- 52-
Initial Brief of the OCA
Witness Bradley presented carrier cost variabilities based on a
restricted, rather than full, quadratic, a major deficiency apparently
due to computational problems.
Witness Bradley attempted to measure volume variability using both a full and
restricted quadratic equation.33 It appears that a full quadratic model, which takes into
account the interactions between various types of mail during the delivery process,
would be more appropriate.
34
Casual observation of the delivery process shows that
small packages and letters may be bundled together with flats for insertion at the
mailbox. The mix of mail appears to have an impact on carrier actions. It appears that
there is interaction in handling procedures by city carriers in delivering the various types
of mail--letters, flats, sequenced mail, etc. The carrier’s actions in delivering DPS letters
and cased flats and letters also appear to be related to the handling of sequenced mail.
Accordingly, it appears that interaction terms are drivers of carrier time and should be
retained if one is modeling delivery time as a function of the shapes. Furthermore, the
full quadratic function is the more general of the two. For the modeling effort using the
2002 CCSTS database, estimation using the full quadratic was the appropriate
approach.
However, witness Bradley discarded the full quadratic form because he obtained
an incorrect sign in developing his model: a negative volume variability for small
packages. Obviously such an outcome is meaningless. The negative variability is
reflective of the database. Witness Bradley recognizes the problem is “an extreme
32
Tr. 34/11568-9.
33
The restricted form lacks cross products.
34
OCA-T-3 at 23.
Docket No. R2006-1
- 53-
Initial Brief of the OCA
symptom of multicollinearity.”35 Multicollinearity is caused by the nature of the available
sample data, involving cases in which variables have approximate linear relationships
with each other. Although true multicollinearity is rare, approximate relationships
involving a degree of multicollinearity are common. A significant level of
multicollinearity can occur when variables vary together; one might find such a situation
occurring in the case of mail delivery when one used a two-week time frame. There are
three solutions to multicollinearity: accept it (an acceptable solution if one is estimating
the dependent variable; a poor solution if one is going to use the regressors in
computing volume variability), eliminate a variable (which could be done by combining
the values of the variables, recognizing that one will be able to compute fewer volume
variabilities), or obtain additional data, possibly through the use of a longer time series
(for example, the DOIS database). Witness Bradley indicates that the existence of
substantial multicollinearity renders the estimated coefficients unreliable even if they
are not negative; the implications of the statement for witness Bradley’s propounded
volume variabilities appear to be highly negative.
36
That is, the volume variability
estimates subsequently computed from unreliable regressors may themselves be
unreliable.
Witness Bradley’s analysis effort has illustrated the problems of collinearity
associated with the appearance of unexpected signs and high VIF values. Negative
volume variability is meaningless. Given the problems of the underlying database as
evidenced by the types of results obtained, it appears that the Carrier Cost analysis
35
Tr. 34/11566
36
Tr. 34/11566-67.
Docket No. R2006-1
- 54-
Initial Brief of the OCA
presented in Docket No. R2005-1 is flawed and that additional analysis is needed.
Witness Smith in his analysis also obtained cases of negative volume variability, further
confirming the deficiencies of the CCSTS database.
Witness Smith Provided Alternative Variabilities.
C.
Even though there appear to be significant problems associated with the
underlying CCSTS database, OCA witness Smith performed an analysis investigating
possible alternatives to witness Bradley’s model. The analysis was presented in Tables
1 and 4 of his testimony. He evaluated the models on the basis of their economic
meaning.
37
Table 2 summarizes some of the variabilities in witness Smith’s testimony and
his response to POIR 25. Table 2 presents witness Bradley’s model, the CC5 model
recommended by witness Smith on the basis of his analysis of the CCSTS data, a
recommended model ND6 based on the DOIS data, CC2 with spr cross products
removed, and CC3 with spr cross products removed.
Witness Smith recommended adoption of the CC5 variabilities in his testimony
as the preferable model, based on the elimination of the density variable, positive signs
for volume variabilities, and a reasonable relationship between letters and flats in terms
of cost. He noted that the reason for the rejection of the full quadratic case for CC3
was a negative sign for volume variability in the case of small packages.
37
OCA-T-3 at 10-11 and 17.
Docket No. R2006-1
- 55-
Initial Brief of the OCA
TABLE 2
Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Mail
Volume
CC1: Witness Bradley Recommended Case
Full Quadratic
Volume Variability
0.24069 0.11606 0.00700031
0.04707 -0.019978 0.67068 -0.0777
Marginal Cost
1.53
2.26
0.46
2.22
-12.29
CC1: Witness Bradley Recommended Case
Restricted Quadratic Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.22268 0.0712
0.012912
0.088135 0.015836 0.66888 -0.0825
Marginal Cost
1.39
1.36
0.82
4
9.56
CC2: Recommended Case, density modified
Density redefined as dens = units/sqm;
Full Quadratic
Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.25654 0.13852
0.015301
0.039501 -0.022269 0.68459 -0.0551
Marginal Cost
1.62
2.68
0.99
1.81
-13.6
CC2: Recommended Case, density modified
Density redefined as dens = units/sqm;
Restricted Quadratic Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.24101 0.06943
0.0067851
0.080097 0.0047302 0.70117 -0.0556
Marginal Cost
1.52
1.34
0.43
3.68
2.89
CC2: Rerun for Full Quadratic, spr cross products eliminated
Density redefined as dens = units/sqm;
Full Quadratic
Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.24263 0.13802
0.016292
0.043027 0.0019312 0.66942 -0.0566
Marginal Cost
1.53
2.65
1.05
1.96
1.17
CC3: Witness Bradley Recommended Case without the Density Varible
Full Quadratic
Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.23729 0.09846 0.00961802
0.055261 -0.005995 0.67683
Marginal Cost
1.53
1.94
0.63376
2.58541 -3.73382
CC3: Witness Bradley Recommended Case without the Density Varible
Restricted Quadratic Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.1905 0.06018
0.015632
0.097381 0.034415 0.65003
Marginal Cost
1.21
1.17
1.02
4.49
21.12
CC3: Witness Bradley Recommended Case without the Density Varible
Rerun for full quadratic, spr cross products eliminated
Full Quadratic
Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.21346 0.09219
0.01125
0.063835 0.028562 0.65517
Marginal Cost
1.37
1.81
0.74
2.98
17.7
The following model recommended by witness Smith based on CCSTS database
CC5A: DPS Case
Letters contain only DPS mail; flats include all other letter mail plus flats.
Full Quadratic
Letters
Flats
Sequenced Collection Sm Prcls Deliv Pts Density
Volume Variability
0.16827 0.16695
0.014871
0.065867 0.0052223 0.66961
Marginal Cost
1.62
1.6
0.96645
3.03943
3.20827
The following model recommended by witness Smith based on DOIS database
Letters
Flats
Seq
Sm Prcls
Curb
Central NDCBU
ND6 Technologydelivpts Full Quadratic
Volume Variability 0.19372 0.12727
0.016056
0.035732
0.11285 0.083708 0.03286
Marginal Cost
1.92
3.2
1.75
66.03
Other
0.40691
Docket No. R2006-1
- 56-
Initial Brief of the OCA
Subsequently, the Commission issued POIR 25, requesting further analysis of
equations CC2 and CC3. An analysis of the matrix of correlations between regressors
and interaction terms showed that there was substantial correlation, most noticeably in
the case of small packages. When multicollinearity is present, one option is to drop the
highly correlated terms. Witness Smith analyzed the effects of dropping the crossproduct terms involving the “spr” variable. In the case of the full quadratic for CC3A, a
negative volume variability was originally reported for small packages. Once the
variables with cross products for “spr” were dropped, the computed negative variability
vanished, and the Variance Inflation Factors were substantially decreased.
Accordingly, CC3A with “spr” cross products removed is a full quadratic (minus
cross products for “spr”) with marginal costs that appear to comport with what one
would expect. This is consistent with witness Bradley’s approach but offers a number
of advantages: there is more consideration of interaction among types of mail because
fewer restrictions are imposed, there are no negative variabilities, and the model lacks
an incorrectly specified density variable.
In the case of the revised CC2 with “spr” cross products eliminated, again there
are no negative volume variabilities. This is an equation that includes the density
variable, which is, however, computed differently from the approach used by witness
Bradley. Density in this case is computed as housing units per square mile.
D.
Witness Smith Has Concluded That Parcels and Accountables Should Be
100 Percent Attributable in Cost.
All time for the delivery of large parcels and accountables should be attributable;
this has been demonstrated by the Postal Service’s ability to specifically and separately
record time and quantities for these delivery activities and time, and to model them
Docket No. R2006-1
- 57-
Initial Brief of the OCA
separately. If the accountables and large parcels were not delivered, then there would
be no time incurred as a result of a deviation in the route. If accountables and large
parcels are delivered, then the Postal Service knows exactly how much time is incurred
and that time arises solely from the route deviation needed to make a non-routine
delivery. Accordingly, it is clear that the time is 100 percent attributable. Witness
Bradley’s estimation procedures for Parcels and Accountables are irrelevant.38
E.
The Use of the DOIS Database Would Solve Much of the Data Problem.
There appears to be a substantial collinearity problem in the estimation process,
and this process appears to be exacerbated by the use of cross-product variables as
well as squared variables. Incorrect signs for volume variabilities and high VIF values
for regressors demonstrate the existence of the problem. In addition, the resulting
regressors may give unreliable values, which makes the computation of volume
variability potentially inaccurate.
The use of time series data over a time period substantially longer than that
available in the CCSTS should help to reduce multicollinearity. For example, in using
the Delivery Operations Information System (DOIS) database the Full Quadratic
equation for ND6, based on the DOIS data, had in general lower Variance Inflation
Factors (VIFs) that was the case for the Full Quadratic version of witness Bradley’s
CCSTS based model. The use of a longer time series of data from the DOIS improved
the VIF values as multicollinearity was reduced. Similarly, elimination of some cross
product terms associated with the variable “spr” also reduced multicollinearity, resulting
38
POIR 25 at 15
Docket No. R2006-1
- 58-
Initial Brief of the OCA
in lower VIF values and improved estimates of volume variability; the improved volume
variability estimates are presented as revised equation 3 of Table 2 in this Brief.
The Delivery Operations Information system (DOIS) is a potential solution to the
data problem. DOIS provides delivery unit supervisors with operations data for
management and decision making purposes. DOIS covers approximately 96 percent of
city carrier routes.39 A major advantage to the use of the DOIS data is the availability of
the data on an ongoing, long-term basis. The data available in the DOIS are similar to
that found in witness Bradley’s CCSTS data base with the following exceptions: there
is no differentiation between small and large packages, piece data are unavailable for
accountables, and there is no piece count of mail collected. Each of these deficiencies
will be addressed.
First, DOIS parcel information does not differentiate between large and small
parcels. The CCSTS database does, however, provide this differentiation. There are
relatively few large parcels, and the total amount of delivery time spent on both large
parcels and accountables is in the neighborhood of 7 percent. Using the CCSTS
database it should be possible to identify the mean amount of time involved in a large
parcel delivery as well as the mean number of large parcels which would typically be
delivered. We also have available updated CCSTS information, so there is an
adequate sample from a statistical viewpoint. A statistical adjustment could be
developed to remove large parcels from the DOIS database. It should be noted that the
use of statistical adjustments is common to much Postal Service data analysis, so
estimation is not a novel or radically new concept.
39
Tr. 8D/4782-83.
Docket No. R2006-1
- 59-
Initial Brief of the OCA
Second, DOIS information does not include information on accountable pieces.
Again, as in the case of large parcels, it should be possible to back out the
accountables delivery time, based on an estimate of accountables per ZIP code and
time to deliver accountables.
Finally, DOIS does not provide information on collection volume. Again, the
CCSTS database would permit an adjustment which could be performed on the DOIS
database.
Using the DOIS data with adjustments would permit the estimation of variabilities
for the major types of mail, i.e., letters, flats, and sequenced mail. The variabilities for
accountables and large parcels should be 100 percent, so the only type of mail for
which one would not have variabilities on an ongoing basis would be pieces collected.
The striking benefits of using a DOIS database are that data would be available from all
city carrier routes over a multi-year timeframe, as contrasted with the eleven-day
CCSTS timeframe. One would expect increased precision and accuracy in the
estimated results. The use of a limited number of statistical adjustments is not an
insurmountable problem: statistical adjustments are common to most database
development efforts.
Docket No. R2006-1
1.
- 60-
Initial Brief of the OCA
Witness Smith performed an analysis of volume variability using
the DOIS database.
The data were received several months into the current case,
40
so an extensive
quality control and data checking effort was not possible. However, data were available
over a 16-quarter, rather than 2-week, timeframe, and the number of ZIP code days
was 21,700. One would expect that the amount of usable data would inevitably be
reduced as data cleaning and verification efforts are mounted, but one would ultimately
expect to have several times the 1,545 ZIP code days worth of data from the original
CCSTS database.
In analyzing the DOIS database, witness Smith made the decision to develop
both full quadratic and limited quadratic cases generally consistent with those
presented in Table 1 of his testimony. He identified equation ND6 as presenting
volume variabilities based on the DOIS analysis.
The DOIS analysis also showed that delivery points could be analyzed by type:
curb, central, NDCBU, and “other”. The disaggregation by delivery point type accounts
for the substantial differences in technology used in delivering mail and the proximity of
delivery points to one another: for example, delivery operations associated with a
cluster box (multiple delivery points at one stop) are clearly different from operations on
a foot route (where delivery points are more widely dispersed).
40
USPS-LR-L-160—DOIS Data Provided in Response to OCA/USPS-T14-8.
received on July 21, 2006.
The data were
Docket No. R2006-1
F.
- 61-
Initial Brief of the OCA
Witness Bradley’s Conclusions Are Now Outdated.
1.
Delivery technology is changing.
Increasingly carriers are performing their work in terms of three bundles of mail:
(1) Delivery Point Sequenced letters, wherein the processing plant provides letters in a
sorted form requiring no additional carrier sorting (DPS); (2) Cased Mail, which is letters
and flats sorted in-office by the carrier before being taken to the street; and (3)
Sequenced Mail, which arrives as one or more sets.41 Some of witness Smith’s
equations, such as his CC5 equation, are in this form. Witness Bradley has testified
against modeling the CCSTS data for 2002 based on three bundles:
First, as I explained in my Docket No. R2005-1 testimony, Postal
Service delivery methods were mixed in FY2002 when the carrier data
were collected. At that time, DPS was not widespread as it is now and
there was still some separate casing of letters and flats. Thus the
“three bundle” approach was not universal as it is now. It is
questionable to specify a model based upon more recent technology for
estimation with a data set in which costs were generated more in
42
accord with an older technology.
Witness Bradley’s comments call into question the relevancy of his analysis,
which is based on the 2002 database. Granted, the 2002 CCSTS study is an
improvement over the previous City Carrier volume variability estimation approach,
which was decades old. However, the years since 2002 involve the implementation of
new delivery technologies. His conclusions are essentially moot. Future work will
consider seriously the three bundle approach. In addition, it is clear that having data
available on an ongoing basis from an existing, operative managerial data collection
41
As previously mentioned, this was discussed in Bradley et. al., “Measuring Scale and Scope
Economies with a Structural Model of Postal Delivery.”
42
Tr. 34/11572.
Docket No. R2006-1
- 62-
Initial Brief of the OCA
system would avoid the need for the commissioning of special studies and surveys; this
is a major advantage of the DOIS database.
In his testimony, witness Smith discussed the changing nature of collection volume,
which is significantly different today as contrasted with the 2002 data.43 Although one
tends to think of collection volume in terms of the carrier removing outgoing letters from
a receptacle, the Postal Service now has an online program allowing customers to
request that letter carriers pick up packages, including Express Mail, Priority Mail, and
Parcel Post (when combined with Express Mail or Priority Mail), on their regular delivery
routes. This type of pickup may require a major route deviation, akin to that involved in
delivering large parcels and accountables. This new parcel pickup service was not
offered to the public in 2002. Hence the mail pieces collected were almost exclusively
letters. Accordingly, the collection volume variability developed by witness Bradley is
now irrelevant. Witness Bradley strongly disagrees, arguing that “…Dr. Smith provides
44
no support for this strong claim.”
Actually, witness Smith was only quoting witness
Bradley’s client, the Postal Service, on this one. Witness Bradley indicates that the
proportion of letters and flats of total collected mail continues to be over 99 percent; he
apparently believes that the program has had little impact. However, the program is still
young and the service is attractive, so the likelihood of witness Smith’s comments
becoming increasingly relevant seems high.
43
OCA-T-3 at 21.
44
Tr. 34/11579 at 26.
Docket No. R2006-1
- 63-
Initial Brief of the OCA
Witness Crowder’s Criticisms Are Largely Irrelevant
G.
OCA disagrees with witness Crowder’s criticisms of witness Smith. Her
comments are best summarized in one of her earlier statements:
Dr. Smith’s analysis of the CCSTS data and analyses is superficial,
with very little in the way of new evaluation.”45
The introduction to this section of the brief summarizes 10 issues addressed by witness
Smith. Of particular interest in this case was the revised CC3 unrestricted quadratic
presented by witness Smith in POIR 25 and also presented in Table 2 of this Brief. The
equation solves the key problems with witness Bradley’s proposed variabilities: lack of
an analysis based on a full quadratic, the generation of positive volume variabilities, and
the elimination of the spurious density variable. Problems remain, most noticeably the
inherent deficiencies of the CCSTS database; in fact, many of witness Crowder’s
criticisms are basically criticisms of the underlying database. However, the alternative
equation presented by witness Smith is a significant advance.
Witness Crowder is correct that the DOIS database lacks a number of
variables.
46
She asserts
At a minimum, future use of the DOIS data for city carrier modeling
requires the addition of collection and accountable volumes, and
disaggregation of parcel volumes into large and small parcel
components. Without these changes, all DOIS city carrier models will
remain tainted.47
45
Tr.34/11640 at 22.
46
Tr. 34/11653
47
Tr. 34/11657 at 3.
Docket No. R2006-1
- 64-
Initial Brief of the OCA
However, as discussed, it is clear that one could make statistical adjustments to the
data in addressing accountables, packages, and collection volume. Accordingly,
witness Crowder’s comments do not identify any substantive problems beyond those
already addressed by witness Smith.
H.
Concluding Comments
OCA notes that the revised equation CC3 as presented in POIR 25 solves some
of the problems associated with witness Bradley’s approach. First, the equation is full
quadratic, allowing for the interaction of delivery products in the delivery process.
Volume variabilities are positive. There is a reasonable relationship among marginal
costs. The equation doses not contain a density variable.
Recognizing that the Commission needs to make a recommendation on the
appropriate choice of volume variabilities, the output of the revised CC3 is preferable to
that presented by witness Bradley.
However, OCA does not believe that CC3 as revised should be the final word in
volume variability. It is clear that packages and accountables should be treated as 100
percent attributable. In the case of other types of mail, however, additional concepts
need to be considered.
First, delivery technologies have changed, as recognized by both witnesses
Smith and Bradley. Accordingly, volume variabilities based on the three-bundle
approach and more current data need to be considered fully.
Second, further analysis of the drivers and technologies associated with the
delivery process is necessary. In just a short time the nature of the delivery process
has changed, as indicated by the three-bundle approach. Furthermore, additional
Docket No. R2006-1
- 65-
Initial Brief of the OCA
consideration of how cost minimization is achieved during the delivery process and how
cost functions are designed is appropriate.
Third, the DOIS data that are available, overall, are most suitable to the
estimation process. Although the database has limitations, it is also the case that the
CCSTS database is inferior, plagued with multicollinearity issues; suffering from limited
observations; and spanning too short a timeframe.
Docket No. R2006-1
III.
- 66-
Initial Brief of the OCA
WINDOW SERVICE COSTS
A.
In This Proceeding, the Postal Service Updated the Previous Window
Service Transaction Supply Side Variabilities.
Until the filing of the current case, transaction supply-side variabilities for window
service had been based on a 1996 study conducted by the Postal Service as well as
earlier methodologies.48 In this case, the Postal Service has presented a new survey of
window service transaction time through the testimony of witness Nieto (USPS-T-24).
Using the results of that survey, a new methodology and estimate of volume variabilities
was presented in the testimony of witness Bradley (USPS-T-17). OCA witness Smith
(OCA
-T- 2) filed testimony rebutting portions of their testimony. In turn, testimony
rebutting witness Smith was filed by Postal Service witnesses Bradley (USPS-RT-4 (Tr.
34/11548, 11581-11597)) and Kelley (USPS-RT-6 (Tr. 36/12244-12265)).
Witness Bradley testified that the prior Window Service volume variability was
deficient due to errors in methodology, and that the operating procedures currently
used by the Postal Service are different from those in use during the initial data
collection effort in 1996. The Postal Service’s implementation of the Point-of-Sale
(POS-ONE) system, the addition of new products and services, and the implementation
of operational changes triggered the need for updated variabilities. (USPS-T-24 at 1).
The OCA concurs with witness Bradley’s conclusions regarding the deficiencies of the
earlier study. (Smith, OCA-T-2 at 2).
48
Postal Service witness Bradley (USPS-T-17 at 1, note) references testimony of Michelle LaMorte
in Docket No. R90-1, USPS-T-6 and testimony by Christopher Brehm in Docket No. R97-1, USPS-T-21.
Docket No. R2006-1
- 67-
Initial Brief of the OCA
In conjunction with witness Bradley’s analysis, Postal Service witness Nieto
presented the results of a field study to collect window service data on transaction time,
by postal product, for input into the cost analysis which witness Bradley used to
calculate the variabilities of window service costs for various window service products.
OCA witness Smith identified a number of deficiencies associated with the
database presented by witness Nieto. The OCA is concerned that the statistical validity
of the database has not been demonstrated, resulting in witness Bradley’s variability
study’s conclusions being based on inadequate information. In addition, the OCA
believes that walk-time and data outliers should have been addressed differently than
they were handled in witness Bradley’s approach, and that the underlying methodology
for witness Bradley’s study may need improvement.
Although the OCA concurs with witness Bradley that the earlier studies are
deficient, the burden is on the Postal Service to demonstrate that the replacement study
is reasonable and that it should be accepted by the Commission. Given the
inadequacies of the outdated earlier study, the Commission should recommend an
update. However, OCA’s witness Smith has enumerated several deficiencies
embedded within the Postal Service’s updated variability computations and offers
alternative, though similar, variabilities. (Smith, T-2 at 17; library reference OCA-LR-3,
Window Service Analysis).49
Witness Smith also presented several caveats about the updated Postal Service
study of which the Commission should be aware and which the Commission should
49
Two additional library references relating to window services were filed by OCA: OCA-LR-7 filed
in response to POIR No. 17 and OCA-LR-9, Tables 1-4 WindowService.xls files, responsive to
USPS/OCA-T2-2 and T2-3.
Docket No. R2006-1
- 68-
Initial Brief of the OCA
advise the Postal Service to consider. By heeding these caveats, the Postal Service
should be able to improve the window service variability computations in future rate
cases. (Smith, T-2 at 18-26).
1.
Postal Service witness Nieto’s updated window service transaction
time study
Witness Nieto has provided information on the development of the new database
for window service transactions. She describes the data collection effort as consisting of
the field collection of transaction data through collector observations and the
subsequent merging of that data with POS-ONE records. The study design was based
on a data collection effort over a two-week period in April and May of 2005 at 27
randomly selected post offices. Collectors used Palm Pilots to gather transaction time
data. The collectors stood unobtrusively near postal window clerks recording
information on the time associated with each transaction. (USPS-T-24 at 4-6).
Witness Nieto stated that data were reviewed and cleaned to correct errors, and
were subsequently combined with the POS-ONE transactional data from the days and
sites observed. The transaction time data was matched with POS-ONE records to
determine the quantities and types of products involved with each recorded transaction
time. (Id. at 7). Then, product variables were created that consolidated the product
identification information into products that are included in the established window
service costing model. (Ibid.).
Docket No. R2006-1
2.
- 69-
Initial Brief of the OCA
The field study data collection effort appears to be seriously flawed
because witness Nieto could not provide adequate justification of
the sample design for data collection.
Witness Nieto has not provided an adequate justification for the selection of
sample size in terms of the total number of transactions, the number of sites sampled,
the breakout between small and large sites, nor the number of transactions by postal
product. For example, Witness Nieto states:
The proportion of 2 large sites to 1 small site was chosen to balance the
considerations of maximizing the number of transactions observed with
including small offices. Including more large offices than small is likely to
increase the number of transactions observed, but small offices were also
included to account for the possibility that they might have differences in
50
transaction times despite having fewer transactions per day.
OCA witness Smith determined the above comments relating to sample size are
not sufficient support for an adequate statistical analysis. (OCA-T-2 at 5). In reviewing
a data collection effort based on the sampling of transactions at sites, one would expect
to find an analysis of the population of sites, types of transactions and data to be
collected, justification for the selection of sites, computation of sample sizes in terms of
their statistical properties, and quality control procedures. (Ibid.). The information
provided in the testimony and the associated library reference, USPS-LR-L-78, was
inadequate to justify the sampling plan. (Ibid.).
Witness Nieto’s analysis appeared to be related to the concept of a stratified
sample; however, she has not provided adequate analysis of the statistical issues
associated with the development of a representative, stratified sample of data. In fact,
she admitted this omission when she indicated:
50
Tr. 5/700 (OCA/USPS-T24-1(h)).
Docket No. R2006-1
- 70-
Initial Brief of the OCA
The goal of the study was not [to] (sic) produce an estimate of total
annual transactions by type but rather to produce a dataset that
permitted an update of the established transaction time econometric
model.51
Witness Smith explained the significant impact that can result if there is a deficient
theoretical basis for the sampling. A database that is not based on statistical sampling
theory, and that cannot be verified to be representative of the set of transactions
studied, does not provide a foundation for the development of a transaction-time
econometric model. (Smith, OCA-T-2 at 6). In fact, witness Smith testified, a model
developed on the basis of incomplete and irrelevant data can give incorrect
conclusions. (Ibid.).
Witness Kelley (USPS-RT-6 (Tr. 36/12244)), rebutting witness Smith’s criticisms
of the study, maintained:
Dr. Smith wants more justification for the selection of sites included in the
study, but he is not specific as to what type of additional justification he is
seeking….” (USPS-RT-6 at 13 (Tr. 36/12259)).
To the contrary, witness Smith explained that one would wish to see a statistical
justification and analysis of the basis for any stratification (not provided by witness
Nieto), as well as a study of the required number of sample observations (also not
provided by witness Nieto, but provided by witness Smith and discussed by witness
Kelley). (OCA-T-2 at 14-16 and particularly Table 4 at 15; Kelley, USPS-RT-6 at 12-15
(Tr. 36/12258-12261)).
In rebutting witness Smith’s concerns about the sample size, witness Kelley
states:
51
Tr. 5/706 (OCA/USPS-T24-5).
Docket No. R2006-1
- 71-
Initial Brief of the OCA
However, since the purpose of this study was to provide the data needed to
update the econometric model and not estimate national level point estimates,
constructing such a table would not have been useful. (USPS-RT-6 at 4 (Tr.
36/12250)).
The statement is wrong. It would be very useful, even important, to know if the
underlying database is statistically adequate, because the data are subsequently used
in the econometric model. Witness Kelly’s contentions notwithstanding, the data are
used to build a national model of window service transactions supply-side variabilities.
We are left by the Postal Service without an understanding of the required sample size
for the study of postal products provided through window service; we are left only with
an inadequate assurance about the appropriate number of sites:
The sample size for the 2005 study (27 sites) was largely based on the size of
the previous 1996 Transaction Study (20 sites), which provided sufficient data for
the estimation of an econometric equation for transaction time, the results of
which were accepted by the Commission in Docket R97-1. (USPS-RT-6 at 5,
footnote omitted (Tr. 36/12251)).
In short, the previous study, based on a flawed methodology and now obsolete
operating practices, as carefully explained by witness Bradley, is, nevertheless,
supposed to serve as the basis for the current study without further statistical analysis.
Witness Smith has explicitly expressed his view that there was inadequate
consideration of the underlying statistical issues associated with sample design. (Smith,
OCA-T-2 at 4-9).
3.
The transaction time study was not representative of small and
large sites.
OCA witness Smith demonstrated in Table 1 of his testimony that the
transactions reported were not representative of the mix of small and large sites: there
is approximately a 50/50 distribution between small and large sites in Postal Service
Docket No. R2006-1
- 72-
Initial Brief of the OCA
offices, but small sites make up only 33 percent of the sample in terms of number of
sites. (Smith, OCA-T-2 at 7). Small sites appear to account for 23 percent of the
observations in the sample, but only 15 percent of the revenue. (Smith, OCA-T-2 at 67). Accordingly, small sites appear to be under-represented in terms of the number of
sites, but overly represented in terms of transaction volume
Witness Kelley appears to find a circular argument in witness Smith’s
conclusions. (Kelley, USPS- at 7-9 (Tr. 36/12252-12255)). Witness Kelley is wrong.
Witness Smith has simply called for (1)increasing the number of small sites sampled in
order to be proportional to the overall mix of small versus large sites, and (2) changing
the mix of observations from small and large sites. The application of standard
sampling theory to this approach should be relatively straightforward, recognizing that
overall sample size will also need to increase in order to obtain statistically accurate sub
samples for various products.
Witness Kelley’s discussion of probability sampling that invokes a quote from the
noted statistician W. Edwards Deming is misdirected.
52
Witness Smith was applying
recognized concepts of statistics with respect to sample size. Witness Smith was not
addressing the procedures with which the Postal Service selected its inadequate
sample of 27 locations. Witness Kelley appears to be addressing the issue of sample
selection, as related to specific sites; this appears to be the issue addressed by
Deming. However, this is not an issue on which witness Smith has offered criticism.
Witness Smith has addressed sample size in terms of numbers of observations and
52
Kelley (USPS-RT-6 at 7 (Tr. 36/12253)) citing W. Edwards Deming, Sample Design in Business
Research, John Wiley & Sons, 1990.
Docket No. R2006-1
- 73-
Initial Brief of the OCA
numbers of sites, not selection of sites. Witness Kelley’s comments on this point are
irrelevant.
4.
The Postal Service offers no substantiation that the collection of
data over a two week period in April/May resulted in a database
that was representative of postal transaction times or
representative of the mix of postal products.
Postal Service witness Hintenach, a manager with experience in Retail
Operations and 40 years of service with the Postal Service, stated in this record that
during heavy mail time periods, such as holidays or tax-filing deadlines, the Postal
Service manages its lobby operations differently than is the case at other times.
53
The
Postal Service now has a lobby program to conduct business with customers while they
wait in line. These factors could affect the transaction times at the window, particularly
at a time when lines are longer during the busiest holiday period.54 The explanations by
witness Nieto as to the reasons for selecting the April/May time period for the field study
do not support her position that the time periods selected are representative and,
therefore, valid, for a statistical study of year-round window service transaction costs.
Witness Nieto testified that the goal was not to produce “an estimate of total
annual transactions by type” but only to update “the established transaction time
econometric model.”55 Obviously, the task was not to merely estimate total annual
transactions, and in that OCA agrees. However, the task was to conduct a field study
that would yield a representative estimation of the window service transactions
53
Tr. 3/399-400.
54
Ibid.
55
Tr. 5/706 (OCA/USPS-24-5).
Docket No. R2006-1
- 74-
Initial Brief of the OCA
occurring during the entire year in order to accurately estimate window service costs in
the test year.
Witness Nieto explained she consulted the “USPS Retail Operations experts” as
to any operational issues associated with the selected period and strove to minimize
potential clerk disruption during tax time. These explanations do not constitute
evidence that the period selected for the study provided data representative of window
service operations for an entire year. Moreover, the established econometric model
she claims to be updating was thoroughly discredited by witness Bradley.
Although the implication of witness Nieto’s explanation is that her study relied
upon an already established model, that methodology has been discredited. The
Postal Service is not updating that model. It is actually generating a new model
supported by a new field study. The reason for a new study is to incorporate new
operational realities into the model. There is no evidence in this record to support her
assumption that, despite the change in operations resulting from the introduction of the
POS-ONE system which now requires an updated study, window service operations
have otherwise remained so similar in terms of window service activity during the year
and the number of representative locations, that the old methodology of the established
study is still valid. In fact, by the Postal Service’s own admission, operating procedures
are different from those in use in 1996. (Bradley, USPS-T-17 at 1, 3-6). The
introduction of the POS-ONE system may have impacted customer requirements for
window service. The lobby program alluded to by witness Hintenach may also have
impacted the window service transaction time at different times of the year.56 Also,
56
Tr. 3/399-400.
Docket No. R2006-1
- 75-
Initial Brief of the OCA
witness Kelley’s comparison of the 1996 study with the 2005 study clearly indicates the
earlier study was conducted at a different time of year than the new study. (Kelley,
USPS-RT-6 at 16 (Tr. 36/12262)). Either additional information is needed to determine
if a two week data collection time period in April and May would be representative of an
entire year or additional information is needed to determine whether different or
additional time periods must be selected to constitute an adequate statistical study.
5.
The Postal Service’s elimination of significant amounts of data
raises the issue of whether the window service transaction study
included sufficient quality control.
Approximately 30 percent of the nested transactions were dropped from the
database due to collection problems: only 133 of 190 nested transactions were retained
in the study.
57
Witness Nieto concluded that the 57 deleted nested transactions were
a small number. She stated:
Therefore, dropping those 57 out of the 190 that were included does
not create any unnecessary issues. It’s just a transaction that happens
to be separated into two parts.58
In addition, approximately 16 percent of all data collected was deemed unusable:
Data were gathered for 9,459 transactions, but 1,535 were dropped from the database
because they could not be matched.59 Witness Nieto testified data were dropped from
the database when the POS-ONE observations could not be matched with the records
57
Nested transactions occur when a transaction is interrupted by another transaction and is then
completed; for example, a customer initiates a Registered Mail transaction, steps away briefly from the
window to complete some forms while the clerk continues with another transaction, and then returns to the
window to complete the transaction. (Smith, OCA-T-2 at 10, note 8).
58
Tr. 5/720.
59
Tr. 5/707 (OCA/USPS-T24-6).
Docket No. R2006-1
- 76-
Initial Brief of the OCA
maintained by the data collectors. She testified that is a little on the high side but within
reason.
60
Yet, the Postal Service provided no analysis of the impact on the database
from the elimination of the unusable observations; that is, whether the 1,535 dropped
observations biased the resulting sample and significantly affected computed volume
variabilities. Witness Nieto concluded:
The data just could not be matched to the product specific
information, so I don’t have specific concerns about the number of
transactions that were dropped because we ended up with a number of
transactions that were equal to or greater than the transactions in the
previous study.61
A database with a number of transactions that are “equal to or greater than the
transactions in the previous study” is not a statistical criterion. (Smith, OCA-T-2 at 11).
Furthermore, witness Nieto did not refer to the statistical literature to determine whether
a significant amount of data had been dropped. She was unable to cite any statistical
literature covering the subject of the dropping of data.62 Witness Kelley has
acknowledged the importance of the matching process and that the collectors
eliminated a significant amount of data they had just collected but were unable to
match. Witness Kelley testified:
This matching process was crucial to validate the information recorded
by the data collectors as part of the study…Transactions that were
unable to be matched with the census of transactions were not
included in the final database. (USPS-RT-6 at 5-6 (Tr. 36/1225112252)).
60
Tr. 5/721.
61
Tr. 5/722.
62
Tr. 5/723.
Docket No. R2006-1
- 77-
Initial Brief of the OCA
While it may be that data that survived the matching process are correct, the Postal
Service should have been more concerned about the elimination of the data that did not
survive the process. Questions arise: what is different, important, or unique about the
data that was eliminated due to lack of matching, and why was data just recently
collected quickly eliminated? Witness Kelley’s own testimony highlighted the problem
by referring to the matching process as “crucial.” In commenting on the dropping of
observations, witness Bradley stated:
Dropping all observations with a studentized residual greater than 2.0
(in absolute value) would mean elimination of 250 data points. This is
over 3% of the collected data.63
Apparently, witness Bradley was concerned about dropping 250 data points (only a little
over 3 percent of the data). This pales in comparison to the 1,535 dropped data points
out of 9,459 data points (or more than 16 percent of the data) discussed by witness
Nieto.64 Thus, without support for the limited April/May data collection timeframe, and
because of the large percentage of dropped nested and non-nested transactions, the
quality of the database is questionable. (Smith, OCA-T-2 at 11).
6.
The erratic walk-time data from facility to facility and day to day are
evidence that there may have been errors in the field study data
collection process.
In Library Reference USPS-LR-L-159, witness Nieto provided walk-time data by
facility. Walk-time denotes the total time period (in seconds) during which the customer
approaches the window. Data collectors were instructed to gather walk-time data for
each transaction in cases where the walk was long enough to be measured. OCA
63
POIR No. 7, Question 7, note 1.
64
Tr. 5/776 (POIR No. 7, Question 8).
Docket No. R2006-1
- 78-
Initial Brief of the OCA
witness Smith provided his Table 3, summarizing the total walk-time data (in seconds)
on a facility-by-facility, daily basis. He has shown that data collection was apparently
erratic from facility to facility and day to day. (Smith, OCA-T-2 at 13).
7.
The Postal Service did not analyze whether enough data was
gathered for each type of window transaction in order to have a
statistically adequate sample.
Normally, in performing survey work, the researcher computes the required
sample size; i.e., how much data is necessary in order to have statistically reliable
conclusions from the sample at an acceptable level of confidence? The Postal Service
apparently did not consider the issue in this case. Witness Nieto testified:
No analysis of the number of transaction observations needed for each
product was performed, because the study was not designed to provide
national estimates of product-specific transaction times or product
volumes. The notion of “statistically representative” product observations
is not well-defined in the context of this update because many
transactions contain multiple products. Rather, the objective of the
transaction time study was to create a database that contained sufficient
transactions to allow an update of the established transaction time
econometric model.65
OCA witness Smith presented sample size calculations for each type of
transaction. The calculations for the computation of sample size shown in his Table 4
were based upon accepted methodology. (Smith, OCA-T-2 at 15).66 Witness Kelley, in
rebuttal to witness Smith, criticized witness Smith’s method of computation underlying
that table (Kelley, USPS-RT-6 at 12 (Tr. 36/12258)), but he did not find any error in
65
66
Tr. 5/701 (OCA/USPS-T24-2).
Gilbert A. Churchill, Jr., Market Research Methodological Foundations, Sixth Edition, Dryden
Press, 1995, at 631.
Docket No. R2006-1
- 79-
Initial Brief of the OCA
witness Smith’s calculations. Witness Kelley’s criticisms are irrelevant. Even if witness
Kelley’s objections are accepted for purposes of argument, such as the passport
example in his rebuttal testimony, the point of witness Smith’s testimony is not refuted;
rather, it is reinforced. That is, for the same products shown in witness Smith’s Table 4
as having insufficient samples, even if the number of required samples are computed
as proposed by witness Kelley, the actual sample sizes obtained by the field study were
far less than the number necessary for a proper statistical analysis.
Witness Kelley testifies:
Theoretically, in deriving the required sample sizes provided in Table 4,
witness Smith incorrectly uses H=4.5 seconds, which is five percent of the
mean transaction time over all single item transactions of 93 seconds.
The proper use of this formula is to calculate a different H for each
transaction type. If it is desired to have the same level of precision for all
transactions, then H is derived by taking five percent for example,
multiplied by the mean transaction time for that transaction type. (Kelley,
USPS-RT-6 at 13-14 (Tr. 36/12259-12260)).
Witness Kelley computes the required sample for Passport transactions, and
finds that the required sample size is 614, whereas witness Smith calculated a required
sample size of 49,343 using an accepted methodology. The actual sample taken by
the Postal Service consists of only 35 transactions, rather than the 614 which witness
Kelley calculates to be valid. This supports witness Smith’s conclusion that there is not
enough data. Given that witness Kelley has provided an alternative computational
method, OCA has recomputed the required samples using witness Kelley’s proposed
computational approach. The computations are presented as alternatives to witness
Smith’s Table 4 within the following Table 3.
Docket No. R2006-1
- 80-
Initial Brief of the OCA
Table 3, Required Sample
Sizes
A
Variable and Product Category
CERT (Certified Mail)
FC (First Class)
STMPSCN (Samps Scanned)
STMPNO (Stamps Not Scanned)
PM (Priority Mail)
MO (Money Order)
PP (Parcel Post)
OWR
Bounded Printed Matter
Library Mail
Media Mail
EM (Express Mail)
PVI (PVI Strips)
INS (Insurance)
RP (Ready Post Items)
INTERNATL
STMPEN (Stamped Envelopes)
REGINS (Registered with Insurance)
PASS (Passport)
RETAIL
BOX (PO Box Items)
DOMCOD (Domestic COD)
FCENCL (First Class with Enclosure)
OSS (Other Special Services)
Return Receipt
Delivery Confirmation
Signature Confirmation
Certificate of Mailing
Postage Due
SERVICES
Hold Mail
Pickup
Mailing Payments
E
F
G
H
I
J
K
L
Cases with one item in a
Total Time
Required
Actual
Required
Cases Where
transaction
for One Item
Sample Size
Sample
Sample Size
Computations
-------------------------------------------------- Transactions
Based on
Size
Based on
Identify
Transactions Time
Std.Dev. Col(E)*Col(F)
H = 4.5
USPS-RT-6
Inadequate
Seconds
(Smith Testimony)
(Kelley Methodology)
Data
674
1223
627
602
455
85
57
5
na
52
235
43
na
57
152
77
na
13
3
68
na
na
na
na
na
na
na
8
74
345
8
74.8
52.6
61.8
46.1
64.1
58.1
105
75.3
118.4
111.9
101.3
57.2
112.6
94.6
82
64
na
na
115.6
97
155.2
78.9
100.6
112.5
na
na
88.9
74.6
188.8
151.6
53.8
52.6
na
na
807
510
337.3
101.8
191.6
195.6
na
na
na
na
na
na
na
na
na
na
na
na
na
na
129.5
47.8
101.994
82
95.5
62.4
102
83
161.8
157.2
Total Transactions:
4863
na: insufficient data to permit computation.
Time per Transaction:
Total Time:
0
50415
75581
40191
63210
53872
8611
6418
410
0
6011
36472
4326
0
5067
28698
4143
0
10491
1012
13029
0
0
0
0
0
0
0
1036
7067
35190
1294
na
525
403
640
1076
2375
621
1698
777
na
1785
1181
2401
na
1056
4360
525
na
49343
1966
7258
na
na
na
na
na
na
na
433
1276
739
1307
4688
391
1777
2019
1266
1550
851
291
159
1
323
148
322
101
314
334
332
163
15
35
8
82
1
1
839
319
480
18
22
9
490
78
402
13
760
855
1262
790
1373
490
1085
936
x
x
x
x
1082
397
1922
x
x
x
1082
991
1469
x
x
x
614
140
1601
x
x
x
209
993
656
1017
1451
x
x
x
x
x
452544
93.058524
In Table 3, columns B, C, and D of witness Smith’s original table have not been
reproduced since they are not relevant to the current analysis. Columns E, F, and G
are reproduced from the original Table 4. Column I presents the required sample size
based on witness Smith’s original calculations. Column K presents the required sample
size based on witness Kelley’s approach. As shown in column L, in all instances where
witness Smith previously indicated a lack of sufficient data, there continues to be a
significant shortage of data.
Docket No. R2006-1
- 81-
Initial Brief of the OCA
Witness Kelley’s comments, therefore, do not refute witness Smith’s
conclusions, rather, his comments confirm the conclusions. As witness Smith explained
in his testimony, (Smith, OCA-T-2-16), there is insufficient data for a number of
products—Money Order, Parcel Post, Bounded Printed Matter, Express Mail, PVI
Strips, Ready Post Items, International, Stamped Envelopes, PO Boxes, Postage Due,
Hold Mail, Pickup, and Mailing Payments. Furthermore, in the case of certain types of
services—such as Return Receipt, Delivery Confirmation, Signature Confirmation, and
Certificate of Mailing—there is no basis for computing required sample size.
Witness Kelley is, however, correct on one issue:
…Dr. Smith, by constructing Table 4, assumes that the means and
standard deviations by transaction type are known before the survey is
conducted. In practice, usually crude estimates are used when applying
67
the sample size formula.
This is exactly the approach used by witness Smith in applying H = 4.5 in the estimation
problem across the board. Further refinement, based on witness Kelley’s comments
has led OCA to generate the new Table 3. In both cases, however, the conclusions are
unchanged.
Witness Kelley continues:
Now, on a practical level everyone wants larger sample sizes.
However, data collection is costly, so limits need to be set. Therefore,
constructing Table 4 might be a nice academic exercise, but as a practical
matter it is not useful.68
67
Kelley (USPS-RT-6 at 14 (Tr. 36/12260)).
68
Ibid.
Docket No. R2006-1
- 82-
Initial Brief of the OCA
Witness Kelley is wrong. He does not take into account the impact on the matter being
studied. Given the millions of dollars involved in window service costs, and the need to
collect only a few hundred more observations in most cases, the achievement of
statistical accuracy is more than “a nice academic exercise.” Furthermore,
achievement of the appropriate sample size does not appear to be beyond the
capabilities of survey research, nor is it unreasonable to expect the Postal Service to
conduct a competent, statistically reliable study. Witness Kelley stated, “The purpose
of the study was to update the established transaction time econometric model.”
(Kelley, USPS-RT-6 at 15 (Tr. 36/12261)). Obtaining an adequate database would be a
good start.
OCA witness Smith concluded that the database presented by witness Nieto
suffers from a number of statistical problems that render it unsuitable for a final
estimation of volume variabilities. On rebuttal, witness Kelley cited the model of the
previous study as justification for the current study (Kelley, USPS-RT-6 at 10-11, 16 (Tr.
36/12256-12257, 12262)), claiming “the current study has a superior dataset to the
previous one….” (Kelley, USPS-RT-6 at 12 (Tr. 36/12258)). Whether or not the
database is better than the previous database, witness Nieto has not established that
the database is adequate to support witness Bradley’s analysis. The Postal Service
has discredited the work on the variability model in the previous study. Given the
apparent errors uncovered in that methodology, that study was apparently not
subjected to intensive analysis and close scrutiny in the rate proceeding where the
Docket No. R2006-1
- 83-
Initial Brief of the OCA
variabilities have been applied.69 There is little justification to rely upon the remainder
of that same study relating to the data gathering process as a model to be followed
when witness Smith has pointed out critical flaws.
In rebuttal testimony, Witness Bradley stated:
I find that his [Dr. Smith’s] concerns about the updated data set are
speculative and not substantive, and they do not have the effects on the
variabilities that he asserts. (Bradley, USPS-RT-4 at 28 (Tr. 34/11581)).
He subsequently continues:
He does not point to even a single instance, out of nearly 8,000
transactions, in which there was an error in recording the type of
transaction. (Bradley, USPS-RT-4 at 29, emphasis omitted (Tr.
34/11582)).
Witness Kelley also claims the validation procedure led to an accurate data set. (Kelley,
USPS-RT-6 at 11 (Tr. 36/12257)). A review of witness Smith’s testimony shows that
both witnesses Bradley and Kelley are on the wrong track, insofar as approximately 16
percent of the transactions could not be matched, even though gathered as part of the
survey process. (Smith, OCA-T-2-10; Tr. 5/721). With such a large portion of the
transactions being unmatched, the question arises as to whether even those instances
that did match with the POS-ONE data are otherwise accurate as to the data on the
length of the transaction time separately recorded only by the data collectors. It seems
that if errors in recording a large proportion of transactions leads to their elimination
from the data base, it calls into question the statistical reliability of the remaining
transactions even though no errors were individually identified. The reliability of the
study and the elimination of bias are necessarily dependent upon a careful
69
3(b)).
In recent rate dockets, no party has objected to or criticized the study. Tr. 5/740 (OCA/USPS-T17-
Docket No. R2006-1
- 84-
Initial Brief of the OCA
measurement of a very large and statistically defined proportion of the events at the
locations randomly selected.
B.
Using the Data Presented by Witness Nieto, Postal Service Witness
Bradley Updated the Window Service Supply Side Variabilities with a
Methodology Inconsistent with His Previous Testimony Before This
Commission and Problematic in Several Other Respects.
Witness Bradley reviewed the established methodology for the computation of
window service supply side variabilities, concluding that there was no analytical support
for the existing calculation of window service volume variabilities (Bradley, USPS-T-17
at 9). He then mathematically derived a revised transaction time variability (Bradley,
USPS-T-17 at 10-18).
1.
Witness Bradley’s model in the current estimation procedure is
linear, but he has used a quadratic approach in other testimony.
Witness Bradley used a linear model as the basis for his analysis. He noted that
the model had been used in Docket Nos. R2001-1 and R2005-1, with no party criticizing
or objecting to the analysis, concluding that it seemed appropriate once again to adopt
a linear specification.
70
During the hearings, he reiterated his position and noted further
that the linear model provides variabilities which are consistent with the entire structure
of the window service costing model.71
It should be noted, however, that the use of a model other than a linear model,
such as a quadratic model for this phase of the window operation, would not be
theoretically inappropriate. In other cases, witness Bradley has strongly advocated the
70
Tr. 5/740 (OCA/USPS-T17-3(b)).
71
Tr. 5/848.
Docket No. R2006-1
- 85-
Initial Brief of the OCA
use of flexible functional forms. No advantage of a linear model has been established
for this purpose, nor would use of a different type of model by witness Bradley in any
way conflict with the analysis. The potential use of a quadratic model or other type of
suitable flexible functional form should have been explored. The choice of a linear
model should have been defended.
A deficiency of the linear model is that it effectively sets volume variabilities for a
number of products sold with other products, such as insurance and return receipt, at
100 percent. The products are always sold in conjunction with another product.
Witness Bradley explained the reason for the assumption of 100 percent variability in
those cases:
The assumption really comes from the fact that for those special
services they’re never sold by themselves. They’re always sold with
another product and so they have no transaction core related time. All
they have is the time associated with themselves in the transaction,
72
and that’s why you get the 100 percent.
Witness Bradley also stated:
In a linear model, like the established model, a product never sold by
itself would have no “fixed” or “common” time. It is easy to show that
[in] (sic) a linear model, a product with no “fixed” or “common” time has
a variability that is 100 percent.73
Witness Smith has recommended the consideration of a flexible
functional form in estimating the volume variabilities of window services. (Smith,
OCA-T-2 at 25-26).
72
Tr. 5/853.
73
Bradley (USPS-T-17 at 10).
Docket No. R2006-1
2.
- 86-
Initial Brief of the OCA
Witness Bradley’s model is sensitive to changes in the data and
thus is not robust.
Witness Bradley indicates that “his [witness Smith’s] theoretical attempt to
criticize the transaction time study database actually shows that the variability results
are robust, not fragile.” (Bradley, USPS-RT-4 at 29 (Tr. 34/11581)). Witness Bradley
devotes the first part of his rebuttal testimony to a discussion of the robustness of the
data. (Bradley, USPS-RT-4 at 29-34 (Tr. 34/11582-11587)). However, his computed
variabilities result from changes to a database that resulted from excluding over 16
percent of the original database. Given such a significant percentage of exclusions, the
degree of confidence expressed by witness Bradley is unwarranted.
Witness Bradley removed a number of outlying observations, as mentioned on
pages 22-24 of his testimony. (Bradley, USPS-T-17). A POIR requested that witness
Bradley supply a test for outliers. In response, witness Bradley noted that he had
removed a number of outliers, but that he had not used a formal statistical test.
74
Witness Bradley discussed the use of studentized residuals:
In general, a studentized residual is a residual divided by its
standard error, but in outlier analysis it is important to have an
“externally’ studentized residual. Externally studentized residuals have
the standard error calculated with the observation removed, thus
eliminating the possibility that a large outlier could contribute to a large
standard error and thus make it harder to identify the outlier itself.75
Witness Bradley’s response to Question 7 of POIR No. 7 provided a list of
studentized residuals with an absolute value above 2. He concluded that observations
with a studentized residual above absolute value 3 are likely outliers, and that those
74
POIR No. 7, Question 7. (The response to this POIR question was designated and commences at
Tr. 5/779, but most of the response, including the portion pertinent here, was not transcribed.)
Docket No. R2006-1
- 87-
Initial Brief of the OCA
with a studentized residual above 2 in absolute value bear investigation. He furnished a
list of 250 observations with a studentized residual above 2 in absolute value and 117
observations with a studentized residual above 3.
OCA witness Smith used the list of 250 observations with a studentized residual
above 2 to calculate the variabilities which appeared in column 5 of his Table 5.
Witness Bradley’s rebuttal testimony criticized the dropping of the observations (USPSRT-4 at 42); however, he does not appear to have investigated the observations on a
case-by-case basis. Accordingly, we are left with 250 observations which, statistically,
are suspect.
Witness Bradley also acknowledged that the SISQ (Single Item, Single Quantity)
76
value should be 9 for the product, “Other SS1.”
In the case of “Other SS1,” by
switching eight transactions, the corrected volume variability fell to 95.2 percent instead
of witness Bradley’s original calculation of 99.4 percent variability. This situation
demonstrates that the impact of failing to account for a few transactions can be
important: From this, it is fair to conclude witness Bradley’s model is sensitive to
changes in data.
3.
Witness Bradley should have included walk-time in the modeling
effort.
Witness Bradley did not include walk-time in his model, but, as witness Smith
testified, walk-time should have been included. Walk-time is defined as the time
75
Id. at 2.
76
Tr. 18A/5311 (POIR No. 11, Question 1).
Docket No. R2006-1
- 88-
Initial Brief of the OCA
required by a customer to reach the service window. In some cases—e.g., small post
offices with minimal walk-times—it is reported that the data collectors did not record
walk-time. However, walk-time was recorded for larger offices or where the time was
considered to be significant. Witness Bradley summarized the definition of a window
transaction, indicating:
The “transaction time” recorded in POS is simply the time from when
the window clerk presses the button to enter the first item to be sold
until the time when the receipt is printed. This amount of time is short
of the total transaction time and the gap between the actual time and
77
the POS ONE recorded time will vary with the type of transactions.
Witness Nieto explains that on-site data collectors recorded information on the time
associated with the customer approaching the window (if applicable), the time the
transaction began, and the time the transaction ended.78 She stated in written crossexamination that customer walk-time is not included in the transaction time:
…The purpose of the transaction time study was to construct a
database permitting an update of the econometric model of transaction
time, thus any non-transactional time was not relevant.
…The “walk” part of the transaction was recorded for those
transactions in which the walk was long enough to allow a data
collector to record a separate measurement. However, the time
associated with the ”walk” part of the transaction was not included in
the calculation of transaction time for any transactions….79
The approach of the customer requires the clerk to be available and prepared to
serve; whether the clerk is actually doing any other meaningful work is irrelevant.
77
USPS-T-17 at 5.
78
USPS-T-24 at 6, lines 13-17.
79
Tr. 5/703 (OCA/USPS-T24-3(e)).
Docket No. R2006-1
- 89-
Initial Brief of the OCA
Accordingly, customer walk-time should have been included as part of the window
service transactions.
Witness Bradley indicated:
Thus, including walk time in the regression is adding a nonvolume related amount of time to the dependent variable and such an
addition cannot, by definition, improve the estimation of the transaction
time coefficients.” (Bradley, USPS-RT-4 at 40 (Tr.34/11593)).
Walk time is, however, associated with each transaction; in some cases it may
be fairly minimal, and in other cases may be somewhat more substantial.
Nevertheless, walk-time is present and should be included. (Smith, T-2 at 20).
Witness Bradley disputes witness Smith’s claim that the reporting of walk time is
erratic. (Bradley, USPS-RT-4 at 37 (Tr.34/11590)). Although walk time is not recorded
as a large percentage of overall time, an examination of the reported walk time at sites
127869, 116806, and 85098—among others—(Smith, OCA-T-2, Table 3 at 13)
suggests that there are substantial variations, calling into question the overall recording
process.
4.
Witness Bradley’s estimations of variability are flawed.
Given the deficiencies in methodology and data, it is clear that the variabilities
estimated by witness Bradley are deficient. Witness Smith has provided alternative
variabilities that are not appreciably different from witness Bradley’s, but these
alternatives are themselves of limited value due to the underlying limitations of the
existing study database.
C.
In Conclusion, There Are Significant Problems with the Updated Window
Service Transaction Side Variabilities.
Docket No. R2006-1
- 90-
Initial Brief of the OCA
A database that is not representative of the population of data cannot yield
meaningful results. Witness Nieto has not provided affirmation that the database meets
the quality standards that are needed. Consequently, because witness Bradley’s model
seems to be very susceptible to potential errors in the database, his results are open to
question for the reasons enumerated by witness Smith.
Witness Smith’s analysis in Table 5 of his testimony presents relatively minor
modifications to the variabilities recommended by witness Bradley. (Smith, OCA-T-2 at
17) However, it should be noted that major changes in variabilities might result from the
development of an adequate database. This is, however, a matter about which we
have insufficient information.
Finally, witness Bradley has endorsed the use of a flexible functional form for
estimation purposes in other cases before this Commission. It is speculative as to
whether the use of a flexible functional form in this analysis would have resulted in
substantially different conclusions. It is clear, however, that for some postal products
included in witness Bradley’s testimony, such as insurance and return receipt, the use
of a linear form is essentially equivalent to assuming 100 percent volume variability. If a
flexible functional form were utilized for those products, their volume variability could be
affected.
The OCA, therefore, agrees with witness Bradley that the current window service
variabilities are outdated and should be revised but offers, in lieu of witness Bradley’s
variabilities, the volume variabilities as shown on Table 5 of witness Smith’s testimony.
OCA also requests that the Commission recommend improvement in the window
service transaction study. This would entail a study that includes:
Docket No. R2006-1
- 91-
Initial Brief of the OCA
a.
a statistically appropriate number of sites, types of sites, and time period;
b.
a sufficient number of transactions;
c.
a sufficient sample design providing for an appropriate number of
transaction samples for each product;
d.
data without an indication of errors in the data collection process signaling
potential quality control problems; and
e.
a database that does not exclude a substantial percentage of
transactions,
or, otherwise, an analysis of the impact of their exclusion.
Finally, the OCA suggests that the Commission should also recommend that a future
Postal Service supply-side window service volume variability model include walk-time
and better explain data outliers in the modeling effort together with a full consideration
of a flexible functional form.
Docket No. R2006-1
IV.
- 92-
Initial Brief of the OCA
FIRST-CLASS RATE PROPOSAL
A.
OCA Witness Thompson’s First-Class Rate Proposal Simplifies the Postal
Service’s Increasingly Complex Schedule
In this docket, the USPS is proposing shape-based rates for First-Class letters
and sealed parcels. OCA supports the USPS’s proposal for shaped-based rates,
because rates may be more nearly aligned with mail characteristic costs. However, the
USPS proposals for First-Class single
- piece mail dramatically increase the rate
schedule complexity. Now customers must understand the differences between letters,
flats and parcels, as defined by the USPS, as well as pay particular attention to the
actual weight of a mail piece—especially a letter whose weight (greater than 3.5
ounces) bumps it up to a flat rate category. (USPS-T32, Revised August 25, 2006, at
19.)
The OCA proposal is less complex. Except for cards, the rates proposed for
single-piece First-Class Mail would be, with one exception, multiples of $0.42. For FirstClass single-piece mailers, only one type of stamp needs to be stocked to satisfy most
mailing needs.
In addition, the OCA proposes to eliminate the additional-ounce rate as
unnecessary. The USPS failed to justify the additional-ounce rate because: (1) there
are no known additional costs of processing mail pieces weighing in excess of one
ounce; and (2) new USPS mail processing equipment is capable of processing lettershaped mail pieces up to six ounces.
1.
OCA’s proposed four-ounce incremental rates, for letter-, flat- and
parcel-shaped mail pieces, significantly reduce the complexity of
Docket No. R2006-1
- 93-
Initial Brief of the OCA
the First-Class single-piece rate schedule proposed by the Postal
Service
The USPS is proposing shape-based rates for First-Class Mail. Consumers will
need extensive education regarding the mailing requirements for First-Class letter -, flatand parcel-shaped pieces. In addition, consumers must be more aware of the mail
piece’s weight and flexibility because those two criteria impact a mail piece’s
classification as either a letter, flat or parcel and the postage required.
OCA supports the USPS’s proposal for shaped-based rates. However, under
the USPS proposal, the consumer must maintain several unique types of stamps for:
letters (especially since the USPS proposes to restrict letters to 0 to 3.5 ounces),
additional ounces, flats, and parcels. This means the general public will: (1) make more
trips to the post office; (2) generate more window transactions; and, (3) wait in longer
lines for window service, simply because consumers may not want to maintain more
than one type of stamp.
As reported by the Washington Post, on October 18, 2006,
80
stamp vending
machines will be a thing of the past and not a consumer option for purchasing stamps
for the First-Class additional-ounce, letter-, flat- and parcel-shaped mail pieces because
the USPS plans to eliminate the machines by 2010. As vending machines break, they
will not be repaired or replaced. Postal Service spokeswoman, Yvonne Yoerger,
indicated that the Postal Service is increasing its use of the automated postal centers
(APC). Yet, APCs are designed to print postage, for a specific mail piece, on demand.
At present, only a few APCs sell stamps. (Id.) The Postal Service has made no
80
http://www.washingtonpost.com/wp-dyn/content/article/2006/10/18/AR2006101801047.html
Docket No. R2006-1
- 94-
Initial Brief of the OCA
commitment to increase the number of APCs that will sell stamps such as those
currently sold through a vending machine. The USPS might argue that the consumer
can easily go to a convenience or grocery store to purchase stamps. However,
convenience and grocery stores currently carry First-Class single
- piece letter stamps for
the convenience of their customers. Stamps are not a significant source of revenue for
the stores. Given the amount of management involved in tracking inventory, distributing
several types of products and recording stamp sales, there is no guarantee that local
stores will maintain a diverse inventory of stamps unless they are given financial
incentives.
A review of the rate schedule proposed by the Postal Service reveals the variety
of stamp inventory a consumer or small business will have to maintain in order to have
the appropriate postage for varying types of First-Class Mail:
•
42-cent stamps for machinable letters up to one-ounce
•
20-cent stamps for each additional ounce of First Class above the first ounce
•
62-cent stamps to pay the postage on one-ounce First-Class flats
•
$1 stamps to pay the postage on one-ounce First-Class parcels
Admittedly, a 20-cent stamp could be paired with a 42-cent stamp to pay for a
one-ounce flat; and five 20-cent stamps could be applied to pay for a one-ounce FirstClass parcel. It should be noted, however, that, by itself, the workhorse 42-cent stamp
is of little use in paying for any type of mailpiece other than a machinable one-ounce
letter. For example, two 42-cent stamps would involve an overpayment of 22 cents on
a one-ounce flat, and a 2-cent overpayment on a two-ounce flat. Three of them would
involve a 26-cent overpayment on a one-ounce parcel, and a 6-cent overpayment on a
Docket No. R2006-1
- 95-
Initial Brief of the OCA
2-ounce parcel. Nearly every rate cell proposed by the Postal Service would produce
overpayment by the stamp-using public. In stark contrast, OCA’s proposed rate
schedule gives stamp users the ability to maintain only a stock of 42-cent stamps to pay
for almost any type of First-Class mailpiece.
OCA witness Thompson’s proposal resolves the problems associated with: (1)
maintaining a diverse inventory of denominated stamps for letter-, flat- and parcelshaped pieces; (2) more frequent post office trips to obtain various types of postage
stamps; (3) increases in the number of window transactions due to the Postal Service’s
proposal for more diverse postage denominations; and (4) increased lobby wait times
for patrons.
OCA adopts the USPS’s proposed rate of $0.42 for a First-Class single
- piece
letter-shaped mail piece. However, OCA witness Thompson recommends that the
$0.42 rate be applied to single-piece letter-shaped mailpieces weighing between 0 and
4 ounces. This eliminates the additional ounce rate which has not been justified by the
USPS. A 0-to-4-ounce weight increment effectively reduces the number of First-Class
single-piece letter-shaped rates from four (as proposed by the USPS) to one. Also,
witness Thompson proposes rate simplification for First-Class flat- and parcel-shaped
mail pieces. Her proposal recommends rates in four ounce increments—0 to 4, 4 to 8
and 8 to 13 ounces—and (except for one rate cell) all are priced in increments of $0.42.
Her proposal reduces the number of First-Class rate cells from 144 to 28 (excluding the
new USPS category for Business Parcels). Consumers will need to maintain only one
denomination of stamp to meet almost all of their First-Class single
- piece postage
requirements.
Docket No. R2006-1
2.
- 96-
Initial Brief of the OCA
Elimination of the additional ounce rate is feasible and justifiable
As noted above, witness Thompson’s proposal eliminates the additional-ounce
rate. In the current rate case, the Postal Service again fails to justify the additionalounce rate.
(a)
The Postal Service continues to ignore Commission
requests for data regarding the cost of processing FirstClass Mail in excess of one ounce
As noted in the PRC’s Op. R97-1, para. 5035,
Notwithstanding the extensive supporting material the Service has filed, a
glaring omission is information addressing the cost support for the FirstClass Mail additional-ounce rate. The Service’s failure to devote attention
to this long-requested review has hindered the Commission’s ability to
review the additional-ounce issue.
The Postal Service provided no information regarding the cost of
processing mail pieces weighing more than one ounce. On the contrary, USPS
witness Taufique essentially concedes that there is no cost justification. He
states candidly that additional-ounce charges are nothing more than revenue
generators. The additional-ounce fee is “substantial and an important source in
meeting the revenue requirements for the subclass and the Postal Service as a
whole.” (USPS-T32, Revised August 25, 2006, at 5.)
Further, witness Taufique states:
Historically, the additional ounce rate bore, not only the cost of additional
weight, but also recovered the cost caused by differences in shapes. As
the Postal Service explicitly recognizes the shape differences in FirstClass Mail rates, the additional ounce rate may be reduced ….” (USPST32 , Revised August 25, 2005, at 4-5.)
According to the USPS, all that remains as justification for having an additional-ounce
rate are weight-related costs. The USPS implies that an automation-compatible mail
Docket No. R2006-1
- 97-
Initial Brief of the OCA
piece costs more to process on a given piece of machinery simply because it weighs
more than one ounce. An automation-compatible mail piece is either machinable or it is
not. Other than rumors,
81
the USPS reports provided no evidence showing that an
automation compatible mail piece is processed differently on any given piece of
machinery due to weight. (Tr. 16/4938-40; 20/7362.) In fact, the Postal Service has
provided information to the contrary. This is discussed in the next section.
(b)
New USPS mail processing machinery processes lettershaped mail weighing up to six ounces
In the test year, USPS witness McCrery states that the USPS plans on having
617 DIOSS-EC machines operational. The machines are capable of handling lettershaped mail pieces weighing one
- to six-ounces. (Tr. 11/2754.) Therefore, there is no
weight-based cost justification for charging additional ounce fees for single-piece lettershaped mail weighing up to six ounces. Since the Postal Service proposes to keep the
weight cut-off for letter-shaped mail pieces well below the six-ounce limitation, no
additional ounce fee is justified.
(c)
The Postal Service’s additional ounce rate arbitrarily exacts
more money from the general public
In his testimony, witness Taufique readily admits that the
[r]evenue generated from additional ounces is substantial and an
important source in meeting the revenue requirements for the subclass
and the Postal Service as a whole. (USPS-T32, Revised August 25,
2006, at 5.)
81
The Postal Service cannot rely on the ancient “3.5 ounce Heavy Letter Field Evaluation Report.”
Neither witness McCrery nor witness Laws was involved with test design or report preparation. Neither
witness could answer questions about the meaning or interpretation of the report. (Tr. 34/11474-75;
39/13446.) Accordingly, the report’s conclusions have no evidentiary weight.
Docket No. R2006-1
- 98-
Initial Brief of the OCA
OCA witness Thompson’s rate proposal eliminates the additional ounce rate, more
closely reflects shape- and weight-related cost characteristics, and meets the USPS
revenue requirement. Thus, USPS witness Taufique’s argument, in support of an
additional ounce rate as a necessary fund raiser, is fatally flawed.
B.
The Commission Should Reject the Postal Service’s Proposal for a
Presort “Base Rate” in Favor of Continuing to Use the Traditional Bulk
Metered Mail Benchmark when Establishing Presort Rates
In this docket, the Postal Service proposes a new approach to pricing First-Class
Mail which has the overarching objective of equal unit institutional cost contribution for
single-piece and workshared First-Class Mail. The Postal Service’s reasoning, for the
new benchmark methodology, appears to be based on the fact that there has been
considerable controversy by presort mailers regarding discount levels. In addition, if the
USPS benchmark is not changed, First-Class presort rates will experience a double
digit increase. (Tr. 38/13349-50.)
Intervenors who send presort letters have contended that using what they
regard as a narrowly defined range of cost characteristics to establish rate
differentials between Single-Piece Letters and Presort Letters may ignore
cost-causing characteristics that, while not expressly associated with the
worksharing activity for which the cost avoidance and discount are being
measured and developed, nevertheless are associated with their mail.
(USPS-T32, Revised August 25, 2006, at 13.)
The presence of varying cost causing characteristics does not automatically
entitle each type of mailpiece to be recognized with a discount. The Commission has
long held that the appropriate cost savings to be recognized in establishing presort
automation discounts is based on the mail processing and delivery cost differential
between the Bulk Metered Mail piece and the mail piece’s depth of sort. By ignoring the
Commission’s practice of using the Bulk Metered Mail (BMM) benchmark, the USPS
Docket No. R2006-1
- 99-
Initial Brief of the OCA
proposal provides Presort automation mailers with significantly lower rates than
warranted. In his rebuttal testimony (Tr. 38/13348), USPS witness Taufique argues that
USPS is unable to develop a single benchmark that will adequately reflect all the
varieties of mail pieces that can convert to or from presort mail. He contends that the
BMM benchmark is no longer an adequate solution. USPS witness Taufique believes
that his average First-Class single-piece benchmark is a viable alternative. Id. OCA
disagrees. Dr. Panzar in his paper, “Efficient Work-Sharing Discounts with Mail
Heterogeneity,” at page 3, states:
Nevertheless, a PRC policy based on ECPR may have substantial merit.
This is because, in standard work-sharing models, discounts set on the
basis of avoided cost assure that mailers will perform work if and only if
they can do it more cost effectively than USPS and that USPS will perform
the work if and only if it can do it more cost effectively than mailers. Thus,
discounts comporting with ECPR lead to the lowest combined cost for the
postal sector. (Tr. 26/9163.)
It appears that a new and lower USPS presort benchmark was sought by the
USPS because mailers threatened to switch their First-Class presort mailings to the
internet if there is a double digit increase in presort rates. In his testimony, MMA
witness Gorham says:
CSG’s clients and mailers like them would be forced to re-think their
opinion as to the value of using direct mail as their “touch point” for
reaching out to their customers. A sharp reduction in discounts in this
case and the prospect of further reductions in the next case would
enhance the relative value of CSG’s EBPP service and accelerate
conversion of our clients’ customers to less costly electronic bill
presentment and payment [EBPP] options. (Tr. 38/13193.)
The First-Class monopoly exists so that the USPS can fund the cost of universal
service for those mailers who are not able to provide discounted mail. If the monopoly
did not exist, independent providers would select the most profitable clients to serve,
Docket No. R2006-1
- 100-
Initial Brief of the OCA
while the USPS would be left with the least profitable. The end result would be that
those clients, for whom the mail provider was the USPS, wouldpay prohibitively high
prices to send mail, and hence, universal service would beat risk .
1.
USPS Witness Taufique erroneously argues that the average FirstClass single piece is the appropriate benchmark.
The USPS argues that a Bulk Metered Mail piece is no longer the sole type of
mail to be a candidate to move from the First-Class single
- piece letter-shaped mail
category to the First-Class presorted letter-shaped mail category and visa versa. (Tr.
16/5027.) Instead, the Postal Service now believes that an average First-Class single
piece mail piece is the more appropriate benchmark to use when establishing presort
rates. Id.
The Commission recognizes that mail processing and delivery worksharing
related unit cost savings are based upon the mail piece’s depth of sort and, to the
extent possible, has set discounts equal to the costs avoided by worksharing. (Docket
No. R2006-1, Notice of Inquiry No. 3, dated July 26, 2006, at 3.) While the
Commission’s approach to calculating discounts has been previously challenged by
large mailers, the Commission has continued to maintain that the BMM benchmark
method is appropriate for setting First-Class automation rates.
The assumption the USPS uses to justify their new presort benchmark is fatally
flawed. The Commission has used the BMM benchmark because it represents the mail
most likely to either convert to or from presort mail. Historically, the Commission has
justified reduced rates for workshared mail because there is a “firm link between the
rate differential and the costs avoided by the Postal Service when worksharing is
performed by mailers.” (Id.) The Postal Service intends to destroy that link; and, OCA
Docket No. R2006-1
- 101-
Initial Brief of the OCA
opposes the USPS efforts to do so, because the USPS proposed discounts are
excessive. (OCA-T-4 at 13.)
USPS witness Taufique postulates that presort bureaus are currently collecting
office mail to presort. (Tr. 16/4938.) His argument is baseless, because there is no
supporting data. (Tr. 16/4940.)
In a class with diverse costs, like First-Class, the benchmark must
be the cost of the mail more likely to convert to worksharing. The cost of
this mail is lower than the average cost of the non-workshared mail.
Using the cost of the average piece will result in price signals that will lead
82
to inefficient production.
Thus, the OCA recommends that the Commission reject the proposed USPS
benchmark methodology that has the potential for shifting the burden of universal
service to those clients who are least able to provide discounted mail and who have few
if any mailing alternatives.
A presort automation compatible mail piece is not likely to convert to a nonautomation mail piece, because, “mailers find it advantageous to self-sort because the
self-sorting cost [ ] is less than the discount offered [ ].” (Dr. J. Panzar, “Efficient WorkSharing Discounts with Mail Heterogeneity” at 8. See also, Tr. 26/9163.) The FirstClass presort automation piece has been barcoded, trayed, presorted and entered
further into the mail stream than a non-presorted mail piece. The depth of sort and the
entry points are such that the presort mailer maximizes the goal of getting mail to its
destination more quickly. In addition, presort rates are lower than the First-Class single
-
82
Progress toward liberalization of the Postal and Delivery sector, edited by Michael A. Crew, Paul
R. Kleindorfer, New York, NY, Springer, 2006, “Worksharing: How Much Productive Efficiency, At What
Cost and at What Price?”, Robert Cohen, Matthew Robinson, John Waller, and Spyros Xenakis, at 151.
Docket No. R2006-1
- 102-
Initial Brief of the OCA
piece letter-shaped mail piece rate. Therefore, it is illogical to believe that a presorted
mail piece would deliberately shift to, and pay, a higher First-Class single
- piece rate.
USPS witness Taufique argues that a link between First-Class single
- piece and
presort will continue to be strong simply because the USPS currently has the goals of
equal unit contributions to institutional costs from single-piece and presort; and the
continuation of separate categories (single-piece and presort) within one subclass. (Tr.
38/13350.) The current USPS goal is highly susceptible to change by either the Postal
Service and/or the Commission, as USPS witness Taufique states in his testimony:
The goal of similar unit contributions from these two mail categories
[workshare and single-piece mail] is not an absolute one; other rate
design and rate impact considerations may require the Postal Service and
the Commission to deviate from this goal. However, to the extent
practicable, the Postal Service’s intention going forward is to equalize the
unit contribution from the Single-Piece Letter category and from the
Presort Letter category. (USPS-T-32, Revised: August 25, 2006, at 16.)
Since the goal of similar unit contributions from the two mail categories is fluid, the
Commission should not accept the Postal Service’s argument that the link will be
strong.
The Postal Service’s new presort benchmark is $0.346 (USPS-T32, Revised
August 25, 2006, at 29-30), i.e., $0.074 (18 percent) lower than the BMM rate of $0.42.
Consequently, all USPS proposed presort rates are lower than they should be. The
only rationale for using the USPS’s new benchmark is to foster lower presort rates for
mailers and justify those rates on the basis of similar unit contributions to that of FirstClass single-piece mail. As proposed, the USPS notes that the rate increases are
reasonable because the rate increase for a First-Class single
- piece letter-shaped mail
Docket No. R2006-1
- 103-
Initial Brief of the OCA
piece (7.7% rounded) is close to the increase proposed for First-Class presorted singlepiece letter-shaped automation MAADC mail piece (6.1% rounded).
USPS witness O’Hara states, during oral cross-examination by counsel for
VALPAK:
A.
… ECP is really aimed at getting the right person to do a particular
activity, right entity. I don’t think we are getting anybody to change
parcels into letters by keeping the contributions the same, having the
difference in rates the same. That’s something that involves what the
ECP is really aimed at. So I agree, you don’t want to apply equal percent
per piece contributions across the whole range of pieces within the subclass.
Q.
So that would relate to shape-related costs in your comments, and
also really weight-related costs. Correct?
A.
Yes. (Tr. 17/5261-62.)
The USPS’s proposal may encourage worksharing, but it does so at the expense
of First-Class single-piece mailers. Therefore, the OCA proposes that the Commission
continue to use the BMM benchmark when developing First-Class presort rates and
establish discounts based on the costs avoided by the activities of mailers that justified
the creation of the discount. (Tr. 20/7378.) When testifying in rebuttal, USPS witness
Abdirahman, USPS-RT-7, states: “[I]f the Commission decides not to accept the delinking proposal, the only alternative that I see is the BMM, and that’s as a costing
witness.” (Tr. 35/11968-69.)
2.
OCA opposes giving intervenors greater discounts
Some mailer-sponsored witnesses have argued that presort discounts should be
increased in the interest of economic efficiency. Others suggest that fairness requires
that mailers be compensated for work they perform to comply with Postal Service entry
requirements. Neither argument holds water. Indeed, the second argument refutes the
Docket No. R2006-1
- 104-
Initial Brief of the OCA
first: if mailers are already creating efficient mail for the Postal Service at current
discounts, there is no need to increase discounts in the name of efficiency.
Before considering the presort mailers’ policy arguments, it will be enlightening to
consider the rate-making requirements of the Postal Reorganization Act. First, there is
no requirement that rates be economically efficient. In fact, economic efficiency is
impossible given the break-even requirement. And numerous policies of the Act are
inconsistent with economic efficiency. Second, the Act imposes a universal service
obligation (USO) at reasonable rates. It makes no sense to increase the burden of
financing the USO on single-piece mailers—those who should benefit from the USO.
ABA witness Kent argues that there has been “an erosion of bulk First-Class
volumes and, if bulk First-Class discount rates were to disappear, the current migration
to use of the internet and email would become a stampede.” (Tr. 38/13226.) In fact, it
is single-piece volume that is declining, and bulk that is growing. Witness Kent
articulates that the de-linking of First-Class Presort rates with those of single-piece
rates is needed as an incentive for mailers to continue using First-Class mail. (Id.)
OCA is not proposing the elimination of First-Class presort discounts. In fact, OCA has
proposed presort rates that pass through 100 percent of the costs avoided. The
Commission should not be swayed by those who erroneously argue that the average
First-Class single-piece mail best represents the benchmark. As discussed earlier, the
argument is without merit; its sole purpose is to justify lower presort rates than would
otherwise have occurred.
ABA rebuttal witness Kent states in reference to First-Class Mail:
Docket No. R2006-1
- 105-
Initial Brief of the OCA
[S]ound business judgment[] suggest[s] that the more mail services cost,
the more a mailer should pay, and conversely the less mail services cost,
the less a mailer should pay. (Tr. 38/13227.)
Witness Kent ignores an important fact. First-Class Mail is covered by the Private
Express statutes. First-Class Mail is a monopoly mail service that exists to provide a
reasonably priced mail stream for everyone in support of universal service. To ignore
the rationale for the monopoly is tantamount to shifting the cost of universal service to
those First-Class single
- piece mailers with few alternatives.
An NAA witness testifies that imposing increasingly higher prices on single-piece
First-Class mailers repudiates the underlying reasons for a monopoly on First-Mail and
actually results in an unfair and inequitable rate schedule, in contravention of 39 U.S.C.
§3622(b) (Tr. 32/10835):
When consumers defect from First-class Mail to electronic forms of
communication, the customers who remain are the inframarginal users of
First-Class Mail, who have fewer competitive alternatives. It seems
contrary to the intent of the postal monopoly for the Postal Service to keep
raising the price charged to these inframarginal users of First-class Mail,
for the purpose of having a postal monopoly in the first place is to keep
the delivery of letters affordable so as to produce positive network
externalities of communication. Moreover, one can reasonably argue that
a downward spiral by which the remaining consumers of First-Class Mail
would be forced to pay continually increasing cost coverage would be
inconsistent with “the establishment and maintenance of a fair and
equitable schedule” that section 3622(b) prescribes.
Arguments suggesting that Standard Mail is a good substitute for First-Class Mail
ring hollow. First-Class Mail offers a level of service that far exceeds that of other mail
classes: (1) it affords the user a level of privacy unmatched by other mail categories; (2)
other than expedited mail, First Class offers the most expeditious and consistent mail
delivery service at a reasonable rate; and (3) recipients have a higher regard for First-
Docket No. R2006-1
- 106-
Initial Brief of the OCA
Class Mail than Standard Mail. MMA witness Gorham (Tr. 38/13187-88) states in his
testimony:
CSG utilizes First-Class Mail on a large scale to send tens of millions of
monthly statements to end users served by our clients. Often, the
monthly billing statement, which includes a Courtesy Reply Mail envelope,
is the only regular “touch-point” our clients have with their customer.
Consistency, quality and timely delivery of CSG’s mailings are very
important to our clients because these factors directly impact the
relationship our clients have with their customers.
Thus, First-Class Presorted Mail is unlikely to convert to Standard Mail. In addition,
monthly billing statements, if mailed, must be sent via First-Class Mail.
In his testimony, witness Gorham states that presort bureaus invest in expensive
equipment that is maintained and updated in order to meet USPS presort requirements.
Given the extensive amount of investment and ongoing employee training, presort
mailers want their rates to be low enough to justify their investment. (Tr. 38/13188-90.)
Otherwise, presort mailers will look to electronic services to meet their mailing needs,
even though their clients know that “monthly paper statements are still the preferred
method of customer communication.” (Tr. 38/13191.)
Presort mailers would not make the investments in equipment and personnel that
they have unless doing so generated profits. As NAPM witness Bell testified, her
presort bureau encourages its customers to provide bulk metered mail. (Tr. 38/13000,
ll. 3-6.) The current discounts provide sufficient incentives for even small mailers—e.g.,
200 pieces per day (Tr. 38/12986, l. 3)—to prepare bulk metered mail. The benefit to
the Postal Service of paying ever smaller mailers ever higher discounts would be
overwhelmed by the loss of revenue that would occur if current presorters also received
Docket No. R2006-1
- 107-
Initial Brief of the OCA
the larger discounts. Witness Panzar was presented with such an example (revenue
losses from higher discounts exceeding cost savings) and was asked,
Would you agree that application of ECPR will lead to no improvement in
cost efficiency if, relative to the status quo, it leads to no change in mailer
behavior?
His response was, “Yes.” (Tr. 26/9157.) Granting the increased discounts requested
by the Postal Service simply transfers a revenue burden from presort mailers to singlepiece mailers, without generating any meaningful change in mailer behavior.
Workshare mailers say they do not want special treatment. They just want rates
that are fair. (Tr. 38/13195.) However, what they claim is fair is nothing more than
trying to avoid their share of responsibility to support universal service. They want to
shift the entire burden of universal service to single-piece mailers.
C.
Declining First Class Mail Volumes May Force a Review to Determine
Which Mail Class Bears an Increasing Portion of the Cost of Universal
Service
NAA rebuttal witness Sidak, (Tr. 32/10836) notes in his testimony that the trend
is toward the electronic communication substitution for First-Class Mail. Further, he
notes that there may be differences regarding the extent:
… to which First-Class Mail has been diverted toward electronic
communication, [however,] there is no debate over the fact that electronic
diversion exists and has had, and will continue to have, an effect on FirstClass Mail that is significant in both an economic and statistical sense.
In contrast to First-Class Mail, there is no evidence in this rate case
that Standard Regular or ECR mail have experienced statistically
significant diversion at the hands of electronic communication. (Tr.
32/10838.)
NAA witness Sidak notes that Standard Regular volumes have increased even
though the cost coverage of Standard Regular Mail has also increased. ECR volumes
Docket No. R2006-1
- 108-
Initial Brief of the OCA
have remained constant even though the cost coverage of ECR has decreased. (Tr.
32/10841.) Thus, NAA witness Sidak suggests that there may come a time when ECR
and Standard Regular Mail will have to bear an increasing share of the contribution to
the institutional costs in support of universal service. This suggestion has merit.
However, the expectation in this case is that First-Class Mail will provide a significant
contribution to institutional costs. The Commission should not be swayed by arguments
to shift more of the USPS institutional costs to the First-Class single
- piece rate payer in
favor of lower rates for First-Class Presort mailers.
Docket No. R2006-1
V.
- 109-
Initial Brief of the OCA
CONFIRM RATE PROPOSAL
The Postal Service’s Proposal to Restructure the Existing SubscriptionBased Fee Schedule for Confirm Service Is Unnecessary to Recover
Costs and Can Be Achieved by Adjusting Current Fees, as Proposed by
the OCA
A.
Among the changes proposed by the omnibus rate filing is a proposal to change
the fees for Confirm service, Fee Schedule 991. Confirm service is integral to the
Postal Service’s effort to increase value to mailers, because, “Confirm service can help
mailers better manage their businesses and enhance their relationships with their
83
customers.”
The importance of Confirm service was echoed by witness Bentley
testifying for the Major Mailers Association.
84
Confirm service also provides the Postal
Service an opportunity to develop a better means of measuring and improving mail
processing and delivery service performance.
As discussed below, the Commission should reject the Postal Service’s proposal
as it is unnecessary and detrimental to users of Confirm service and the Postal Service.
Instead, the Commission should recommend the subscription fees for Confirm service
proposed by the OCA in order to encourage expanded usage of Confirm service. In
addition, the Commission should retain the requirement in the Domestic Mail
Classification Schedule (DMCS) for a “start the clock” notification for the entry of
Confirm mail to facilitate development of a service performance measurement system
for business mail.
83
USPS-T-40 (Mitchum), at 14-15.
84
MMA-T-1 (Bentley), at 32.
Docket No. R2006-1
1.
- 110-
Initial Brief of the OCA
The Postal Service fails to meet its burden of proof that the
classification changes it proposes for confirm service are superior
to the existing subscription-based fee structure that was held fair
and equitable in Docket No. MC2002-1
In this proceeding, the Postal Service is initiating a request to change the price
structure for Confirm service. The Postal Service has submitted a request under
§§3623 and 3624 of Title 39 for a recommended decision on the proposed
classification change. Section 3624 provides that “an opportunity for a hearing on the
record under sections 556 and 557 of title 5” (the Administrative Procedure Act) must
be accorded to mailers and an officer of the Commission. As the moving party, the
Postal Service has the burden of proving that the existing classification for Confirm, i.e.,
three-tiered pricing with an unlimited subscription option, is inferior to the new
classification structure proposed by witness Mitchum.
The Administrative Procedure Act makes clear that “the proponent of a rule or
order shall have the burden of proof.” The burden of proof establishes specific
requirements for the Postal Service’s evidence and the criteria against which that
evidence is to be measured. As the Supreme Court explained in the context of the
Natural Gas Act, a regulated entity “initiating an increase in rates . . . assumes the
hazards involved in that procedure. It bears the burden of establishing its rate schedule
as being ‘just and reasonable.’” FPC v. Tennessee Gas Pipeline Company, 371 U.S.
145, 152 (1962). Similarly, the Postal Service bears the burden of establishing that its
proposed change to the Confirm pricing framework is “fair and equitable.” 39 U.S.C.
§3623.
As explained in 4 Stein, Mitchell, Mezines, Administrative Law §24.01 at 24-5
through 24-9 (1991):
Docket No. R2006-1
- 111-
Initial Brief of the OCA
The term “burden of proof” often contemplates what the litigating
proponent must establish in order to persuade the trier of the facts of the
validity of his claim or affirmative defense and, at times, is referred to as
the “burden of persuasion” . . . . [I]t is a burden derived from substantive
law which becomes an active factor in the context of litigation when all of
the evidence has been submitted.
It is imperative to note that the burden of proof remains fixed throughout
the litigation on the proponent, who must sustain his burden of persuasion
in a qualitative manner, specifically with credible evidence.
With respect to the proposed change to the price structure of Confirm, the Postal
Service has failed to sustain its burden of providing qualitative and credible evidence
that (1) its proposed price structure and rates will cover costs better than witness
Callow’s proposed fees under the current classification, and (2) that its proposed price
structure and fees are more fair and equitable than witness’s Callow’s.
2.
The Postal Service’s proposal will increase costs for, and reduce
usage by, Confirm subscribers
The Postal Service’s proposal would eliminate the existing subscription-based
fee structure which currently offers three different levels of Confirm service: Silver,
Gold, or Platinum. Under this fee structure, subscribers receive either 3 months or 12
months of service and different quantities of scans, in the millions (blocks), depending
on the service level. Subscribers may also purchase additional blocks of scans and
Additional Identification (ID) Codes.
In place of the existing three service levels, the Postal Service proposes to
establish a single transaction-based annual user fee of $5,000, which includes one
annual ID Code and one million “units” to be applied towards scans of mail. Additional
scans, purchased in blocks of one million “units,” would be available according to a
schedule of fees that decline as the number of blocks purchased increases. This
Docket No. R2006-1
- 112-
Initial Brief of the OCA
“declining block” fee schedule allows users to purchase additional blocks one through
nine for a fee of $70 each. Purchases of the next 90 blocks (i.e., 10 to 99 blocks) are
available for a fee of $35 each. Users purchasing in excess of 99 blocks during the
annual subscription period can do so for a fee of $17.50 each.85 Additional ID Codes
would be available for use either for a quarter ($750) or for a year ($2,000), or until the
end of the subscription term, if sooner. Most problematic is the proposal to charge five
times as many “units” per scan for Standard Mail and Periodical Mail scans as for FirstClass Mail scans.
The Postal Service assumes for its proposal that the number of Confirm
subscriptions will remain at the current level through the test year.
86
By contrast, the
Postal Service estimates that the demand for scans will decrease 10 percent from the
current usage. Based upon these assumptions, the new fees would provide a cost
coverage of 126.3 percent on revenue of $1,517,297 and Test Year (2008) costs of
$1,200,890.87
Witness Drew Mitchum (USPS-T-40), testifying in support of the Postal Service’s
proposal, states that restructuring of the existing fee schedule is “intended to generate
revenue adequate to cover Confirm costs . . . [so that] the Postal Service can continue
offering Confirm Service.”88 He observes that revenues and demand for Confirm
service have not met the original forecast in Docket No. MC2002-1, the docket
85
USPS-T-40 (Mitchum), at 17.
86
Tr. 14/4156.
87
USPS-T-40 (Mitchum), at 19.
88
Tr. 14/3937 (OCA/USPS-T40-16).
Docket No. R2006-1
- 113-
Initial Brief of the OCA
establishing Confirm service as a permanent mail classification.89 Moreover, witness
Mitchum states his belief that raising the current subscription-based fees cannot cover
the costs of the Confirm service or produce a reasonable cost coverage.
90
Consequently, the Postal Service’s transaction-based fee proposal was devised.
Significantly, the Postal Service did not express any fundamental objections to the
current three-tier fee structure, nor did witness Mitchum discuss with current
subscribers how they would react to the change in the current pricing structure.91
The Postal Service’s proposal represents a fundamental restructuring of the
schedule of fees for Confirm service. The OCA believes such a restructuring of the
existing fee schedule is unnecessary to cover costs. The Postal Service’s goal of
covering Confirm service costs can be achieved better by simply adjusting current fees,
as proposed by the OCA.
Moreover, restructuring fees as proposed represents a significant departure from
the original subscription-based pricing approach presented by the Postal Service and
recommended by the Commission in Docket No. MC2002-1. As a result, the Postal
Service’s plans to restructure the existing fee schedule will discourage usage by
imposing significantly higher fees on most Confirm subscribers, and further diminish
Confirm service as a tool for measuring service performance by business mailers and
the Postal Service.
89
USPS-T-40 (Mitchum), at 19.
90
Tr. 14/4153-5, 4174.
91
Tr. 14/4174-75.
Docket No. R2006-1
3.
- 114-
Initial Brief of the OCA
The OCA’s proposal minimizes fee increases for most Confirm
subscribers in order to expand usage while recovering costs
Witness James F. Callow (OCA-T-5) presents the OCA proposal for Confirm
service. Under the OCA proposal, fees are $2,000 per quarter for a Silver subscription,
$5,200 for a Gold subscription, and $19,500 for a Platinum subscription, representing
increases of 0 percent, 16 percent, and 95 present for Silver, Gold, and Platinum
subscriptions, respectively. Witness Callow proposes no change in the fees for
additional scans for Silver and Gold subscribers, and a 50 percent increase in the fee,
to $750, for a quarterly Additional ID code. These fees produce test year revenues of
$1,529,500 and a cost coverage of 127.3 percent.
Witness Callow further testifies that development of OCA’s proposal for Confirm
service was directed toward addressing the following five concerns:
92
1) Retain the existing Silver, Gold, and Platinum service subscription-based levels;
2) Cover the costs of Confirm service with a cost coverage slightly above the cost
coverage proposed by the Postal Service;
3) Encourage the expanded use of Confirm service by preserving the subscriptionbased “internet” pricing model, including the “service expansion” features.
4) Eliminate the declining block user fees based upon “units,” which permits the
differential pricing of scan data as between First-Class Mail and the “other” (i.e.,
Standard Mail and Periodicals) mail classes.
5) Minimize the potential loss of subscribers by holding fees constant for Silver
subscribers and increasing them modestly for Gold subscribers, while assigning
92
OCA-T-5 (Callow), at 14-15.
Docket No. R2006-1
- 115-
Initial Brief of the OCA
the largest fee increases to large-volume Platinum subscribers in exchange for
preserving the option of unlimited scans.
The OCA’s proposed fees minimize the risk of revenue deficiency from reduced
demand caused by the Postal Service’s transaction-based schedule and resulting
higher combined total fees, and produce a cost coverage slightly larger than the cost
coverage proposed by the Postal Service. Moreover, by retaining the existing
subscription-based fee schedule, OCA’s fee proposal preserves the benefits of the
existing schedule—encouraging expanded use of Confirm service.
B.
The OCA’s Subscription-Based Fees for Confirm Service Will Recover
Institutional Costs Unlike the Transaction-Based Fees Proposed by the
Postal Service
1.
OCA’s Estimate of Confirm subscribers and estimated revenues is
realistic
(a)
Witness Callow’s assumptions about the number of Confirm
subscribers are credible
According to the Postal Service, Base Year (FY 2005) subscriptions for the
Silver, Gold, and Platinum service levels were 16, 119, and 45, respectively.93 In order
to estimate test year revenues, witness Callow assumes the same number of Confirm
subscriptions in the test year as in the Base Year. These assumptions are credible,
given the change in fees proposed by the OCA and the Postal Service’s lack of
knowledge as to the behavior of Confirm subscribers in response to changes in price.
93
USPS-LR-L-124 (REV 7-3-06), Excel file “REV-USPS-LR-124 7-3-06.xls,” worksheet tab “WP4
Confirm.”
Docket No. R2006-1
- 116-
Initial Brief of the OCA
Witness Callow assumes no change in the number of Silver subscriptions, since
no changes are proposed in the current Silver subscription fee.
94
In addition, witness
Callow assumes no change in the number of Gold subscriptions, based upon a
proposed fee increase of 15.6 percent.95 This modest fee increase, the first since
Confirm service was established more than four years ago, approximates the 5.4
percent system-wide average fee increase resulting from Docket No. R2005-1 and the
proposed 8.5 percent average increase proposed in this proceeding.
For Platinum subscribers, witness Callow proposes a fee increase of $9,500,
representing an increase of 95 percent. Witness Callow assumes such an increase will
have little or no effect on the number of Platinum subscribers, based upon the
demonstrated behavior of current subscribers.
96
The Postal Service questions the credibility of witness Callow’s assumptions with
respect to the number of Platinum subscribers. Witness Mitchum maintains that
witness Callow’s assumption is “unreasonable.”97 According to the Postal Service,
current Platinum subscribers will reduce their usage by “seeding” their mailings with
Confirm barcodes or move to the Gold subscription service “if the price differential
between the two [subscription] tiers were to be widened from $5,500 to $14,300, as the
94
OCA-T-5 (Callow), at 16-17.
95
OCA-T-5 (Callow), at 17.
96
Id.
97
Tr. 33/11296 (USPS-RT-13 (Mitchum), at 12).
Docket No. R2006-1
- 117-
Initial Brief of the OCA
OCA proposes.”98 Thus, witness Mitchum assumes “that Confirm subscribers choose
99
to pay as little as possible, and choose the Gold tier when it is less expensive.”
The Postal Service has no knowledge of the demand behavior of current
Platinum subscribers whose level of usage could be satisfied by a Gold subscription.
Current Platinum subscribers using less than 92 million scans have an incentive to
become Gold subscribers. At 92 million scans, a Platinum subscriber pays $109
($10,000 / 92) per million scans, while a Gold subscriber pays $106 (($4,500 + (7
additional blocks of 6 million scans * $750)) / 92) per million scans. Thus, a Gold
subscription is the least expensive option until a subscriber uses more than 92 million
scans.
According to the Postal Service, however, there are 29 current Platinum
subscribers that use 62 million scans or less.
100
At 62 million scans or less, a current
Platinum subscriber could save at least $4,000 ($10,000 – ($4,500 + (2 additional
blocks of 6 million scans * $750))) by becoming a Gold subscriber. Nevertheless, such
Platinum subscribers have not converted to Gold subscriptions.
The Postal Service cannot explain this known “unreasonable” behavior of current
Platinum subscribers. First, the Postal Service is unable to claim any knowledge of
subscriber behavior with respect to price changes for Confirm service, because there
have been no changes in Confirm service fees since its inception. In fact, the Postal
98
Tr. 33/11295 (USPS-RT-13 (Mitchum), at 11).
99
Tr. 33/11296 (USPS-RT-13 (Mitchum), at 12).
100
Tr. 14/3976 (OCA/USPS-T40-57).
Docket No. R2006-1
- 118-
Initial Brief of the OCA
Service “does not really know what motivates customers to overpay.”101 The Postal
Service never spoke to current Platinum subscribers that currently overpay as to why
they do so.
102
This behavior by the 29 identified Platinum subscribers suggests such
subscribers are not as price sensitive as thought by the Postal Service.103 As the
current behavior shows, Platinum subscribers do not always choose the least expensive
service level. They often respond to factors other than price. It also suggests witness
Callow’s assumption, that OCA’s Platinum fee increase “will have little or no effect on
the number of Platinum subscribers,”104 is not “unreasonable.”
Nevertheless, the Postal Service questions whether Platinum subscribers “who
could save as much as $5,500 by moving to the Gold tier under the existing fees will
decide to pay as much as $14,300 more than is necessary” under the OCA proposal.105
However, the Postal Service made no effort to learn of the response of Confirm
subscribers to such a change in prices. According to witness Mitchum, the Postal
Service never asked Confirm subscribers how they would change their behavior if
OCA’s proposed fees were recommended by the Commission.106 Moreover, the Postal
Service never “discussed with any mailer” whether OCA’s proposed $19,500 Platinum
101
Tr. 33/11340.
102
Tr. 33/11339.
103
Tr. 21/7792 (USPS/OCA-T5-2(e)-(g)).
104
OCA-T-5 (Callow), at 17.
105
Tr. 33/11295 (USPS-RT-13 (Mitchum), at 11).
106
Tr. 33/11337.
Docket No. R2006-1
- 119-
Initial Brief of the OCA
subscription fee would cause such subscribers to terminate their subscriptions.107 On
the other hand, as discussed below, witness Callow has gauged the response of
Confirm subscribers to the OCA’s proposed fees.
Second, the Postal Service wrongly attributes the mailer activity of “seeding”—
the application of PLANET Code barcodes to a sample of, rather than all, mailpieces—
to higher subscription-based fees. Seeding is a mailer response to transaction-based
fees of the type proposed by the Postal Service in this proceeding. The likelihood of
seeding was one of the reasons Postal Service witness James M. Kiefer (USPS-T-1) in
Docket No. MC2002-1, specifically rejected a transaction-based pricing approach.108
By contrast, witness Kiefer determined that a subscription-based fee approach would
®
“encourage Confirm subscribers to place barcodes on all their mail rather than limiting
usage to occasional mailings, or seeding barcodes within mailings.”109
2.
Witness Callow’s revenue estimate is likely to be achieved
OCA’s estimate of test year revenues is likely to be achieved for several
reasons. The subscription-based fees proposed by witness Callow have the support of
Confirm subscribers, or their representatives. According to the testimony of witness
Cameron Bellamy, President of Grayhair Software (GHS), Confirm subscriber who
submitted testimony in this proceeding, “Many of our customers and other large mailers
have told us that the OCA proposal is preferred by them over the USPS proposal with
107
Tr. 33/11353.
108
Tr. 33/11432a, Exhibit No. NPPC-X-2 (Docket No. MC2002-1, USPS-T-5 (Kiefer), at 4).
109
Id.
Docket No. R2006-1
- 120-
Initial Brief of the OCA
its per-scan charges that arbitrarily discriminate by mail class.”110 This view is echoed
by a coalition of mailers, many of whose members are Confirm subscribers.
111
We believe that there is a consensus among the users of Confirm, both
large and small, that covering the costs of Confirm by adjusting the annual
permit fees (e.g., as proposed by OCA) would better serve the interests of
mailers and the Postal Service than would the imposition of per-scan
charges in all tiers of service, including Platinum.
Moreover, in contrast to the Postal Service’s proposal, only adjusting the current
subscription-based fees maintains a more stable source of revenue that is likely to
produce the estimated test year revenues projected by witness Callow. Under the
existing subscription-based fee schedule, mailers can choose a subscription service
level—Silver, Gold or Platinum—based upon their projected need for Confirm scans.
Once selected, mailers pay their Confirm subscription fee up-front, permitting
subscribers to use scans up to the maximum allowed during the specified time period
for each subscription service level. In so doing, the Postal Service receives a known
amount of revenue whether or not mailers use scans up to the maximum allowed for
each service level. In effect, during the subscription period, mailers cannot treat
112
Confirm service as a “discretionary good.”
110
GHS-ST-1 (Bellamy), at 6-7.
111
Answer of Grayhair Software, Inc., American Bankers Association, Association for Postal
Commerce, Mail Order Association of America, Mailing & fulfillment Services Association, National
Association of Presort Mailers, National Postal Policy Council, and Office of Consumer Advocate to
Objection of Untied States Postal Service to Filing of Supplemental Testimony of Cameron Bellamy (GHSST-1), November 29, 2006, at 2; see also Del Polito, Condemning Confirm to Perdition, PostCom Bulletin
50-06 (December 8, 2006), at 6-7.
112
Tr. 33/11291 (USPS-RT-13 (Mitchum), at 7).
Docket No. R2006-1
- 121-
Initial Brief of the OCA
Witness Callow assumes, like the Postal Service, that no revenue will be
generated from the purchase of Additional ID codes in the test year.
113
This
assumption is reflected in the OCA’s estimate of test year revenue.114 However,
Confirm subscribers are expected to purchase Additional IDs in the test year, which will
generate additional revenues and increase the cost coverage of Confirm service
beyond the 127.3 percent proposed. The testimony of witness Bellamy states that
“GHS is certain it will continue to purchase additional ID codes.”115 The purchase of
Additional IDs provides a measure of insurance that witness Callow’s estimated test
year revenue will be achieved.
Moreover, the departure of some Platinum subscribers would not significantly
affect the OCA’s revenue estimate. Witness Mitchum projects, with respect to scans, a
10 percent reduction in demand for scans in response to the transaction-based fees for
Confirm service.
116
Assuming the Postal Service’s 10 percent reduction in demand is
“reasonable,” if applied to Platinum subscribers, this suggests a migration of five
(rounded) subscribers from the Platinum to Gold subscription service level. A shift of 5
subscribers would reduce the OCA's proposed cost coverage from 127.1 percent to
121.4 percent. In fact, it would take a shift of more than 22 Platinum subscribers to the
113
USPS-LR-L-124 (REV 7-3-06), Excel file “REV-USPS-LR-124 7-3-06.xls,” worksheet tab “WP-4
Confirm;” see also Tr. 14/3921 (“mailers would be unlikely to buy additional IDs when they were no longer
necessary, resulting in a reduction in the number of additional IDs.” MMA/USPS-T40-2(e)).
114
OCA-T-5, Attachment 1 (Revised 10-20-06).
115
GHS-T-1 (Bellamy), at 7.
116
Tr. 14/4132 (Response of Postal Service Witness Mitchum to POIR No. 4, Question 3)
Docket No. R2006-1
- 122-
Initial Brief of the OCA
Gold subscription service level to reduce the cost coverage below 100 percent.117 If,
however, OCA’s proposed fees caused five Platinum subscribers to drop Confirm
service entirely, the cost coverage would decrease to 119.2 percent. More than 16
Platinum subscribers would have to drop Confirm service to reduce the cost coverage
below 100 percent. Such changes in the number of Platinum subscribers are highly
unlikely to occur, and is another reason OCA’s revenue estimate is more likely to be
achieved, and more likely than the Postal Service’s estimate.
To the extent OCA’s proposed fee causes some Platinum subscribers to become
Gold subscribers or drop Confirm service altogether, the resulting lower cost coverage
would properly reflect the reduced value of service provided to Confirm mailers. As
originally proposed, Confirm service was designed to benefit not only mailers, but also
the Postal Service.
118
Benefits to the Postal Service would arise from widespread
usage of Confirm mailings, permitting the collected scan data to be used to monitor mail
processing operations and improve service performance measurement.119 The Postal
Service no longer relies on mailer-generated Confirm scans for performance
measurement.120 As a result, the Postal Service has eliminated one of the benefits of
Confirm service—greater transparency of system-wide service performance—expected
117
Tr. 21/7794 (USPS/OCA-T5-3(a)).
118
PRC Op. MC2002-1, at 1.
119
Tr. 33/11432 and 11438, Exhibit No. NPPC-X-2 (Docket No. MC2002-1, USPS-T-5 (Kiefer), at 3
and 15).
120
Tr. 33/11334.
Docket No. R2006-1
- 123-
Initial Brief of the OCA
by mailers.121 Even if revenues are slightly less than estimated, a cost coverage lower
than targeted would not be unreasonable.
3.
The Postal Service’s estimate of Confirm transactions lacks
support, resulting in a revenue estimate that will not materialize
(a)
Witness Mitchum’s estimate of a 10 percent reduction in
Confirm scans is unsupported and not developed on this
record
The Postal Service acknowledges that its transaction-based fees will adversely
affect demand for Confirm service. The testimony of witness Mitchum assumes a 10
percent reduction in demand for Confirm scans from current usage.
122
This reduction in
demand is based on the Postal Service’s further assumption that “some mailers will
begin seeding to reduce their costs.”123
Witness Mitchum’s estimate of a 10 percent reduction in demand is without
support. Witness Mitchum admits that his 10 percent estimate is arbitrary.124 The
Postal Service’s estimated reduction in demand “is not based on any market
research.”125 Demand for Confirm service is highly priced elastic, according to the
findings of witness Kiefer in Docket No. MC2002-1.126 Witness Mitchum provides no
evidence to suggest that there has been a change in the high price elasticity of demand
121
Docket No. MC2002-1, USPS-T-2 (Lubenow), at 1-2.
122
Tr. 14/4132 (Response of Postal Service Witness Mitchum to POIR No. 4, Question 3)
123
Tr. 33/11298 (USPS-RT-13 (Mitchum), at 14).
124
Tr. 33/11341; see also Tr. 14/3974 (OCA/USPS-T40-55(a)).
125
Tr. 33/11342.
126
Tr. 33/11435, Exhibit No. NPPC-X-2 (Docket No. MC2002-1, USPS-T-5 (Kiefer), at 9).
Docket No. R2006-1
- 124-
Initial Brief of the OCA
for Confirm service since Docket No. MC2002-1. Moreover, witness Mitchum’s
assumption that the transaction-based fees will reduce demand by 10 percent can not
be replicated or evaluated because of concern about the use of proprietary data.
(b)
127
Witness Mitchum’s transaction-based fees will discourage
usage and reduce revenue to the Postal Service from
Confirm
The Postal Service’s proposed fees will discourage expanded usage of Confirm
service by producing higher costs for most Confirm subscribers. Given the documented
high price elasticity of demand for Confirm service, witness Mitchum’s assumed 10
percent estimate likely understates the expected reduction in demand, resulting in lower
estimated revenues in the test year. This result of the Postal Service’s higher
transaction-based fees should not be surprising. Witness Bellamy observes that “we
should expect that some mailers will respond to the attempt of the Postal Service to
128
turn individual Confirm scans into a profit center by rationing their use of the service.”
For another reason, the Postal Service’s proposed transaction-based fees will
likely generate less revenue than estimated. The Postal Service observes:129
Many end user subscribers may well view Confirm service as a
discretionary good, which could be cut from the budget during tough times
or in the face of large fee increases that are applied recklessly.
To the extent this observation is accurate, the Postal Service’s transaction-based user
fees permit mailers to accord Confirm scans “discretionary” treatment. Transactionbased fees permit mailers to reduce usage immediately in response to tougher
127
Tr. 14/3973 (OCA/USPS-T40-54(a)).
128
GHS-T-1 (Bellamy), at 11.
129
Tr. 33/11291 (USPS-RT-13 (Mitchum), at 7).
Docket No. R2006-1
- 125-
Initial Brief of the OCA
economic times, mid-year cost-cutting initiatives, or other factors. In contrast, a
subscription-based fee schedule provides that subscribers pay up-front for a fixed
number of scans for a fixed period of time.
An indication of the likely effect on demand can be gleaned from the higher
combined total fees (i.e., the annual user fee plus declining block fees) applicable to
most current Confirm subscribers under the Postal Service’s proposal. Considering
only the less expensive scans for First-Class Mail, current Silver subscribers will
experience fee changes ranging from -14 to 190 percent, depending upon the number
of quarterly subscriptions purchased during the year and scans used.130 Confirm
mailers that purchased one, two, or three quarterly Silver subscriptions will pay between
6 and 190 percent more in combined total fees than currently.131 For the 12 month
period February 1, 2005 to January 31, 2006, there were seven quarterly Silver
subscriptions purchased by five Confirm mailers.132 Only Confirm mailers that
purchased four sequential quarterly Silver subscriptions will experience a decrease in
combined total fees of between -8 and -14 percent under the Postal Service’s
proposal.133 For the same 12 month period, there were 12 sequential quarterly Silver
subscriptions purchased by three Confirm mailers.134 In addition, assuming Gold
130
OCA-T-5, Attachment 1 (Revised 10-20-06).
131
Tr. 14/3976 (OCA/USPS-T40-57).
132
Tr. 19/6791 (Response of Postal Service Witness Mitchum to POIR No. 12, Question 6). The
Postal Service is unable to determine the number of Silver subscriptions in the Base Year. Id. Moreover,
information on the number of mailers that purchased one, two, three or four quarterly Silver subscriptions
during Fiscal Years 2003, 2004, and 2005 is not available from the Postal Service. Tr. 14/3975
(OCA/USPS-T40-56(a)).
133
OCA-T-5 (Callow), at 17 (Revised 10-20-06).
134
Tr. 19/6791 (Response of Postal Service Witness Mitchum to POIR No. 12, Question 6).
Docket No. R2006-1
- 126-
Initial Brief of the OCA
subscribers use only less expensive First-Class Mail scans, most Gold subscribers
would pay higher combined total fees that rise 11 to 56 percent with usage under the
Postal Service’s proposal.
135
Current Platinum subscribers using only First-Class Mail
scans will experience combined total fee changes ranging from -50 percent for 1 million
scans, rising to 102 percent for 750 million scans, up to 145 percent for 1 billion
scans.136
For current Confirm subscribers purchasing the average number of First-Class
Mail and Standard Mail scans per block of 1 million units, varying the number of scans
acquired per “unit” will increase combined total fees even more, further depressing
demand. The Postal Service estimates the number of scans in a typical block of one
million “units” at 357,143, based upon each Standard Mail scan equal to 5 units and
each First-Class Mail scan equal to one unit.137 Most of the 119 current Gold
subscribers purchasing the average number of scans will experience higher combined
total fees than under the OCA proposal. Combined total fees rise from 7 percent for 1
million scans, to 64 percent for 20 million scans, up to 158 percent for 50 million scans.
For Platinum subscribers purchasing the average number of scans under the
Postal Service’s proposal, combined total fee changes range from -52 percent for 1
million scans, to 0 percent for 60 million scans, exploding to an unacceptable 338
percent for 750 million scans, up to 460 percent for 1 billion scans.138 The only way for
135
OCA-T-5, Attachment 1 (Revised 10-20-06).
136
OCA-T-5, Attachment 1 (Revised 10-20-06). The Postal Service claims that the maximum
number of scans used by any current subscriber is near 750 million. Tr. 14/4144.
137
Tr. 14/3957 (OCA/USPS-T40-29).
138
OCA-T-5, Attachment 1 (Revised 10-20-06).
Docket No. R2006-1
- 127-
Initial Brief of the OCA
such Platinum subscribers to avoid such rising fees is to limit usage. According to the
Postal Service, “[a]ny [Platinum] subscriber[s] who chose to use fewer than
169,000,000 units would pay less under the proposed fee schedule.”139 (Emphasis
added). Based upon current usage, sixteen of the current 45 Platinum subscribers will
pay combined total fees higher than the $10,000 subscription fee.
140
However, all
Platinum subscribers lose the option of unlimited scans—a current feature of the
existing Platinum subscription service. OCA submits that it is the reckless application of
larger fee increases by the Postal Service that will reduce Confirm usage and also
reduce revenues.
C.
The OCA’s Fee Structure for Confirm Service Is Presumptively Fair and
Equitable and the Postal Service’s Fees Are Arbitrary and Not Fair and
Equitable.
1.
OCA proposes to retain the existing, presumptivelyf air and
equitable subscription-based fee structure previously
recommended by the Commission.
When considering new rates and fees, a primary reference point for the
Commission is the existing rate relationships. The Commission’s current fees for
Confirm service are presumptively fair and equitable. Although the Postal Service
wants to change the current fee structure, it does not claim the existing subscriptionbased fee structure for Confirm service is “unfair or inequitable.”141 Rather, with respect
139
Tr. 14/3976 (OCA/USPS-T40-57). The Postal Service estimates that a user could purchase 168
additional blocks and still spend less than the current $10,000 Platinum subscription fee, calculated as
follows: the $5,000 annual user fee plus $4,987.50 ((9 * $70) + (90 * $35) + (69 * $17.50)) in declining
block user fees. See also OCA-T-5, Attachment 1 (Revised 10-20-06).
140
Id.
141
Tr. 33/11356.
Docket No. R2006-1
- 128-
Initial Brief of the OCA
to the OCA fee proposal, witness Mitchum maintains that it is not “superlative in either
142
fairness or equity and as such has room for improvement.”
According to witness
Mitchum, the fairness and equity of the OCA proposal is “worse than what the existing
fee schedule is.”143 This view is not supported by the record. The OCA proposes to
retain the existing subscription-based fee structure previously recommended by the
Commission in Docket No. MC2002-1, and simply adjust current fees. The existing rate
relationships will not change.
Just as with the existing subscription-based fee structure, OCA’s proposal offers
mailers the same options they currently have: mailers will choose either of the three
service levels based upon their projected need for Confirm scans. Like now, two
different mailers will be charged the same amount for the same number of scans. A
Platinum subscriber purchasing 51 million scans would pay the same amount, and have
the same cost per scan, as any other Platinum subscriber purchasing 51 million scans.
Similarly, a Platinum subscriber purchasing 1 billion scans would pay the same amount,
and have the same cost per scan, as any other Platinum subscriber purchasing 1 billion
scans. On the other hand, the Postal Service’s differential unit pricing proposal would
charge two different mailers different prices for the same number of scans depending
upon the class of mail being scanned without any cost justification and so is unfair.
Under both the current fee structure and the proposed OCA fee structure, the
average price per scan decreases as the number of scans purchased increases. With
the OCA proposal, a Platinum subscriber purchasing 51 million scans would pay an
142
Tr. 14/ 3972 (OCA/USPS-T40-51(a)).
143
Tr. 33/11356.
Docket No. R2006-1
- 129-
Initial Brief of the OCA
average price of $382.35 per million scans whereas the same subscriber purchasing 1
billion scans would see a decrease in the average price to $19.50 per million scans.
144
The Postal Service does not object to deeper discounts for increased usage but
is concerned about the extent of the “volume taper” and believes the relatively higher
average costs for smaller mailers may be unfair and should be a consideration of the
Commission.145 A measure of the degree of perceived unfairness may be gained by
giving consideration to the judgment of Confirm subscribers. OCA is unaware in this
proceeding, or elsewhere, of any intervenor or mailer that has objected to the volume
taper feature in the existing subscription-based fee structure. When asked about
complaints regarding the current fee structure, Witness Mitchum did not point to any
customer complaints about the supposed unfairness of the unlimited scan option.
146
In
fact, he did not even inquire as to customers’ views about the fairness of OCA’s
subscription fee structure.147 On the other hand, this decrease in average price per
million scans provides a strong incentive for mailers to use PLANET barcodes on all of
their mailings. This volume taper price incentive will encourage mailers to make greater
use of Confirm service, but it is not a feature of the Postal Service’s proposal.
Witness Mitchum states his proposal improves fairness and equity, particularly
for smaller users.148 Yet his proposal is not necessarily cheaper for the smallest users.
144
Tr. 33/11302 (USPS-RT-13 (Mitchum) , Table 2, at 18); see also Tr. 33/11406-407.
145
Tr. 33/11357.
146
Tr. 33/11378.
147
Tr. 33/11337-38.
148
Tr. 33/11394.
Docket No. R2006-1
- 130-
Initial Brief of the OCA
The smallest Confirm subscribers under the OCA proposal, purchasers of Silver
subscriptions, would pay only $2,000 for a quarterly subscription or $4,000 for two
quarters, whereas each would be required to pay the $5,000 annual user fee under the
Postal Service proposal.149
In other respects, the existing subscription-based fees proposed by the OCA
much better serves the interests of the smallest mailers. Two Confirm re-sellers:
Grayhair Software and TrackMyMail offer access to low volume users for only $50 per
10,000 scans.
150
It is not at all clear that the Postal Service’s proposed elimination of
the current option of unlimited scans will continue to make possible such small-user
access. The “small” users that witness Mitchum wishes to protect are not mailers with
$50 accounts, but mailers whowouldfind it worthwhile to pay the $5,000 annual user
fee (100 times the size of Grayhair’s and TrackMyMail’s smallest account holders).
Thus, on balance, the unfairness perceived by the Postal Service in the OCA’s proposal
is not supported by the record, while the Postal Service’s proposed rate structure might
very well lead to much higher fees for the smallest mailers.
2.
The Postal Service’s differential unit-based fees for blocks of onemillion scans is arbitrary, and unfair and inequitable
The Postal Service’s proposal is arbitrary, unfair and inequitable and has not
been justified. For the first time with respect to Confirm service, the Postal Service
proposes to price additional blocks based on “units” rather than scans—the basis for
149
150
Tr. 22/11395-96.
Tr. 14/4147 (OCA-XE-Mitchum-#1). OCA’s cross-examination exhibit reveals TrackMyMail's
pricing for low volume mailers: $50 for the first 1,000 pieces and $2 for each additional 1,000 pieces.
OCA has been informed by witness Bellamy (GHS-T-1), President of GrayHair Software, that it offers a
similar product, at comparable prices, for low volume mailers.
Docket No. R2006-1
- 131-
Initial Brief of the OCA
pricing additional blocks under the existing fee schedule. The Postal Service claims
that its proposed “pricing structure is more fair and equitable than the three-tier
151
system.”
To the contrary, the Postal Service proposal to vary the number of scans
acquired per “unit” is unfair and inequitable on its face. Further inquiry has also
determined that it is arbitrary.
The Postal Service does not attempt to justify such differential treatment
between blocks of First-Class Mail and Standard Mail scans as a form of “value
pricing.”152 In fact, the Postal Service provides no basis for its plan to effectively charge
different prices for scan data based upon the class of mail. The Postal Service makes
no effort to justify why the ratio of 5-to-1 was selected in the first place, or why it is
preferable to a ratio of, say, 2-to-1. Without any rational support, such differential
pricing of scans is arbitrary.
Moreover, the differential pricing of scans cannot be justified on the basis of cost.
The cost per passive scan to the Postal Service of a barcoded mailpiece of First-Class
Mail or Standard Mail is the same.
153
Rather, the true purpose of such differential
treatment is to develop a pricing structure that generates revenues to cover costs.
154
The Commission should reject such arbitrary, unfair and inequitable pricing that cannot
be justified on the basis of costs.
151
Tr. 14/3937 (OCA/USPS-T40-16).
152
Tr. 14/3936 (OCA/USPS-T40-15(e)).
153
Tr. 14/3958 (OCA/USPS-T40-30(b)).
154
Tr. 14/3936 (OCA/USPS-T40-15(e)).
Docket No. R2006-1
D.
- 132-
Initial Brief of the OCA
The OCA’s Fee Proposal for Confirm Service Preserves the Simplicity
Inherent in the Existing Subscription-Based Fee Structure
The existing subscription-based fee schedule for Confirm service is simple in
structure. The OCA proposal retains the simplicity inherent in the existing subscriptionbased fee schedule by simply adjusting current fees.
The Postal Service states that the “OCA retains the existing structure and the
complexity of 3 different sets of subscription prices,” and that retention of that structure
will continue depressing the number of subscribers.
155
The Postal Service
acknowledges that the OCA proposal is no more “complex” than the existing
subscription-based fee schedule.156 Moreover, no existing Confirm subscribers have
expressed to the Postal Service concern about the complexity of the OCA proposal.157
Given that the Postal Service finds the OCA proposal no more complex than the
existing fee schedule, and can point to no complaints with respect to complexity, the
Postal Service view that the structure will depress the number of subscribers is pure
speculation.
By contrast, the Postal Service maintains that it’s proposal “is less complicated”
than the existing subscription-based fee schedule and, by extension, less complicated
than the OCA proposal. The Postal Service is mistaken. The transaction-based fees
proposed by the Postal Service introduce a number of complicating elements to the
Confirm fee schedule.
155
Tr. 33/11304 (USPS-RT-13 (Mitchum), at 20).
156
Tr. 33/11362.
157
Tr. 33/11363.
Docket No. R2006-1
- 133-
Initial Brief of the OCA
The Postal Service’s bow toward simplicity consists of replacing the Silver, Gold,
and Platinum subscription service levels with a single annual user fee of $5,000, which
158
includes 1 million “units.”
From there on, however, the Postal Service’s proposal
increases in complexity. First, the Postal Service proposes to replace the existing block
purchase of additional scans with a transaction-based user fee based upon “units.”
Users may purchase additional “units” in blocks of 1 million, and the number of
additional units purchased varies by the type (i.e., First-Class Mail, or “other”) of scans
desired.
The Postal Service then introduces another complicating element. As discussed
previously, the purchase of additional blocks of 1 million “units” follows a schedule of
fees that decrease as the number of blocks purchased increase beyond certain a
specified minimum, or threshold, quantity of scans. This “declining block fee” allows
users to purchase additional blocks one through nine for a fee of $70 each. Purchases
of the next 90 blocks (i.e., 10 to 99 blocks) are available for a fee of $35 each. Users
intending to purchase in excess of 99 blocks during the annual subscription period can
do so for a fee of $17.50 each.
159
Third, as discussed above, with the pricing of additional blocks based upon
“units,” the Postal Service proposes to vary the number of scans acquired per “unit.”
Thus, the purchase of each First-Class Mail scan will require one “unit,” while each
Standard Mail and Periodicals scan will require 5 “units.” Each of these elements
158
USPS-T-40 (Mitchum), at 17.
159
Id.
Docket No. R2006-1
- 134-
Initial Brief of the OCA
represents a departure from simplicity, and exposes the inaccuracy of the Postal
Service’s claim that its proposal is less complicated.
Moreover, Postal Service witness Kiefer, in Docket No. MC2002-1, cited
increases in administrative costs, since each transaction would have to be tracked and
billed, as a reason for rejecting a transaction-based fee structure. While technology
permits the Postal Service to keep track of how many scans every customer uses, a
transaction-based fee schedule would complicate billing for current Platinum
subscribers. Such subscribers would likely want to begin tracking how many scans they
use in order to make sure they were billed correctly.160 In turn, this may lead to a higher
frequency of billing inquiries and complaints by Confirm subscribers that can cause an
increase in the cost to the Postal Service of administering Confirm service.
E.
The Potential for Future Enhancements to Confirm Service Is Not a Valid
Reason to Accept the Postal Service’s Unit Pricing Proposal In Lieu of the
OCA’s Subscription-Based Fees Which Can Also Accommodate Future
Enhancements
The Postal Service asserts another reason supporting its proposed transactionbased fee schedule—that pricing based upon “units” provides flexibility for future
enhancements to offer other types of scan information for Confirm service mailers.
Witness Mitchum cites scanning containers as a potential future enhancement for
Confirm.161 However, without programming for unit pricing, fees cannot be
differentiated for scans on different types of materials.162 In the future, witness Mitchum
160
Tr. 33/11363.
161
Tr. 33/11307 (USPS-RT-13 (Mitchum), at 23).
162
Tr. 33/11367.
Docket No. R2006-1
- 135-
Initial Brief of the OCA
says, changes in the number of units per scan, which may or may not have a fee
attached, could be incurred without substantial programming costs. Such future
enhancements would be a means of obtaining additional revenue for information about
the mail.163
For several reasons, the “flexibility for future enhancements” argument is not a
basis for distinguishing between witness Mitchum’s proposal and witness Callow’s
proposal, and does not provide a justification for a recommendation in favor of the
Postal Service’s Confirm fee proposal. Although witness Mitchum states that he
discussed future enhancements of Confirm service with the Postal Service’s product
development group, he is aware of only one planned Confirm enhancement—
introduction of the four-state barcode for which unit pricing is not needed.
164
He admits
that other than the four-state barcode, there are currently no plans for the
implementation of scanning of containers; in fact, there are no plans for future
enhancements to Confirm during the test year.165
Moreover, the scanning of containers may not require a different fee, in which
case unit pricing may not be required to scan containers. Witness Mitchum does not
assert that container scans would necessarily be priced differently from other scans,
even if they are implemented. As witness Mitchum recognizes, the cost of scans would
be an important consideration in determining whether differential pricing is required:
“Units would allow [the] (sic) Postal Service to charge a different fee for container scans
163
USPS-T-40 (Mitchum), at 15.
164
Tr. 33/11324-25.
165
Tr. 33/11366.
Docket No. R2006-1
- 136-
Initial Brief of the OCA
than for mailpiece scans to the extent the generation and provision of such data is more
costly.” (Emphasis added).
166
Until the costing of container scans is measured, subject
to Commission review, it will not be known whether container scans have different costs
than scans of single pieces of mail and consequently whether they should be priced
differently.167
The Postal Service adds further support for its proposal that does not provide for
unlimited scans. Scans for future enhancements may not have the same very low
marginal cost as scans for mailpieces, so it is useful to have a fee schedule in place
which has a modest price in place for all scans. As noted above, the cost of scans for
future enhancements is not currently known. It is premature to eliminate the current
price structure providing for unlimited scans for a future eventuality that may not occur.
Witness Mitchum’s rebuttal testimony indicates some uncertainty with respect to future
enhancements: “Regardless of the fate of future enhancements, a modest charge for
168
incremental units helps cover the cost of the program.”
In other words, he
recognizes future enhancements may not necessarily occur.
According to witness Mitchum, programming for future unit pricing can be
relatively inexpensive if combined with programming for the new rate structure
proposed by the Postal Service.169 Programming OCA’s proposal for unit pricing would
incur additional costs in excess of those required to program for unit pricing under
166
Tr. 33/11307-08 (USPS-RT-13 (Mitchum), at 23-24).
167
Regardless of whether the Postal Service’s or the OCA’s proposal is implemented, the Postal
Service would be required to return to the Commission to set a price for scans of containers. Tr. 33/11369.
168
USPS-T-40 (Mitchum), at 18.
169
Tr. 33/11366, 11369-370.
Docket No. R2006-1
- 137-
Initial Brief of the OCA
witness Mitchum’s proposal. However, he is unable to quantify the number of hours or
the cost of reprogramming.
170
It is obvious that a few hours of a programmer’s time is
so trivial an expense that the Postal Service has probably wasted more of its, and
other’s, resources just arguing this point before the Commission. In any event,
incurring programming costs for enhancements now, solely because there may be a
need for unit scans in the future, ignores the time value of money prematurely
expended to prepare for what is only a potential pricing modification. The mere
possibility of saving some programming costs for potential future enhancements does
not justify elimination of the current pricing structure with an unlimited scan option and
eliminating a fee structure favored by many customers.
The Postal Service also neglects to point out that additional and substantially
more programming would be required to implement witness Mitchum’s new fee
structure than witness Callow’s proposed fees. The OCA proposal maintains the
existing subscription-based fee schedule and simply adjusts current fees, virtually
eliminating all the programming costs to be incurred under the Postal Service’s
proposal.
Although witness Callow did not address the question of the introduction of
manual scans of containers into Confirm service, that does not demonstrate
enhancements for Confirm service cannot be accommodated by the OCA proposal. If
witness Callow’s proposal is accepted, the Postal Service could program for scans
based upon units of various values when it is ready to implement future enhancements.
A fee structure with differential unit pricing, as proposed by the Postal Service, is not
170
Tr. 33/11370-371.
Docket No. R2006-1
- 138-
Initial Brief of the OCA
necessary to undertake programming to implement unit pricing of scans. In fact, the
Postal Service’s concerns about the alleged incompatibility of the OCA proposal with
unit pricing could apparently be eliminated simply by programming prices to be
expressed in units rather than dollars and cents.171
F.
The Existing Requirement to Provide Electronic Notifications for Entry of
Confirm Mailings Should Be Retained to Permit Development of SystemWide Measurement of Service Performance
Confirm service subscribers entering Destination Confirm mailpieces are
required to provide an electronic preshipment notification for every outgoing Confirm
mailing. This requirement, found in section 991.31 of the Domestic Mail Classification
Schedule (herein “DMCS”), specifies that authorized subscribers provide, for
Destination Confirm, “electronic notice of entering Confirm mail prior to or
172
contemporaneous with mail entry.”
Initially, the Postal Service proposed to eliminate this requirement from DMCS
§991.31. Subsequent to the filing of its request, however, the OCA reached an informal
agreement with the Postal Service concerning Confirm that was forged during the
period of discovery on the Postal Service’s case. Under this agreement, the Postal
Service no longer asks that the start-the-clock language currently in the DMCS be
removed. Specifically, the language that OCA and the Postal Service agree should be
retained is the following: “and, for Destination Confirm, providing electronic notice of
171
Id.
172
DMCS §991.31.
Docket No. R2006-1
- 139-
Initial Brief of the OCA
entering Confirm mail prior to or contemporaneous with mail entry, all. . . .” In its
entirety, DMCS §991.31 will read as follows:
991.31 Mailers [must] become Confirm subscribers by applying to, and
being authorized by the Postal Service. Authorization requires that a
customer demonstrate the capabilities of producing mailpieces with
Confirm-compatible barcodes, and, for Destination Confirm, providing
electronic notice of entering Confirm mail prior to or contemporaneous
with mail entry, all as specified by the Postal Service.
This is essentially the same language that is currently found in the DMCS. The
Postal Service’s Chief Counsel, Rates and Classifications, Daniel Foucheaux, has
informed the Director, OCA, that the Postal Service authorizes OCA to provide the
above statement to the Commission. The OCA wishes to thank the Postal Service for
its willingness to discuss this issue, and its work in concluding this agreement.
In reaching this agreement, DMCS §991.31 seeks to preserve the original intent
of the preshipment notification requirement. As originally proposed, Confirm service
was designed to benefit not only mailers, but also the Postal Service.
173
Benefits to the
Postal Service would arise from widespread usage of Confirm mailings, permitting the
Postal Service to use the collected scan data to monitor mail processing operations and
improve service performance measurement.174 Of critical importance to achieving
these benefits is the requirement of a “preshipment” notification:175
Preshipment notification enables the Postal Service to use Confirm
information to measure, diagnose, monitor, and improve mail processing
and delivery service performance.
173
PRC Op. MC2002-1, at 1.
174
Docket Tr. 33/11432 and 11438, Exhibit No. NPPC-X-2 (Docket No. MC2002-1, USPS-T-5
(Kiefer), at 3 and 15).
175
Postal Service Publication 197, Confirm Users Guide (September 2004), at 29.
Docket No. R2006-1
- 140-
Initial Brief of the OCA
To achieve this result, the preshipment notification, an electronic manifest that
provides a profile of the Confirm mailing,
176
serves to link entry scan data with PLANET Code mailpiece processing in
order to “start the clock” on the mailing and help measure processing and
delivery performance.
Since establishment of Confirm service, the Postal Service has had a change-ofheart as to the benefits of using collected scan data for service performance
measurement and improvement, as expressed in the testimony and discovery
responses of witness Mitchum. At some point, the Postal Service “concluded [ ] that
177
Confirm service is ill-suited to evaluation of system performance.”
This conclusion is
apparently based on operational issues, including178
i)
inconsistent mail preparation and barcoding methods by Confirm
mailers;
ii)
inconsistent induction procedures by mailers and the Postal
Service;
iii)
system infrastructure limitations that limit the ability to uniquely
identify mailpieces with a pre-shipment notification;
iv)
lack of integration with postal mail acceptance and verification
procedures, and;
v)
incorrect entry of information on mail processing equipment that
becomes associated with a Confirm scan.
As a result of these operational issues, the Postal Service simply seeds PLANET Code
barcodes on mailpieces that are reentered into the mailstream for its own limited
purposes.
176
Id., at 2.
177
Tr. 14/3943 (OCA/USPS-T40-20(a)-(b)).
178
Tr. 14/3948 (OCA/USPS-T40-23(c)).
Docket No. R2006-1
- 141-
Initial Brief of the OCA
The electronic preshipment notification requirement contained in DMCS §991.31
agreed to by the Postal Service and the OCA provides an opportunity to facilitate
development of a better means for entering Confirm mail and to measure and improve
mail processing and delivery service performance. In this regard, the OCA urges the
Postal Service to work with Confirm subscribers, interested mailers, and trade
associations to develop an accurate, reliable, and low-cost electronic notification
system for the entry of Confirm mailings that is simple-to-use by subscribers and the
Postal Service. Moreover, the purpose of such a system should be to ensure that the
electronic notification is recognized by the Postal Service and Confirm subscribers as
the official “start-the-clock” entry time for purposes of developing transparent, systemwide service performance measurement for business mail.
Docket No. R2006-1
VI.
- 142-
Initial Brief of the OCA
FOREVER STAMP
A.
The Postal Service’s Proposed “Forever Stamp” Mail Classification Should
Be Recommended by the Commission to Enhance Convenience for
Individuals and Smaller Mailers.
In this proceeding, the Postal Service proposes a new mail product styled as a
“Forever Stamp” that offers a new beneficial postage option to individuals and smaller
mailers who predominantly use First-Class Mail. The Forever Stamp will be a nondenominated postage stamp available for purchase at the current First-Class firstounce rate. Once purchased, according to the Postal Service, the “Forever Stamp
would constitute valid postage for the first-ounce of single-piece First-Class Mail letters,
179
even if the rate were to change.”
If recommended by the Commission, the Forever
Stamp would be sold at the approved single-piece First Class rate (presumably $0.42)
upon its introduction some weeks prior to the implementation date for the new rates
approved in this docket. The Board will set the specific effective date, if the Forever
Stamp is approved.180 After the subsequent rate increase, the stamp would be sold at
the rate applicable to such letters at the time of purchase.
Postal Service witness Altaf H. Taufique presents the Forever Stamp proposal in
two separate pieces of testimony: USPS-T-32 introduces the concept and expected
benefits for mailers,181 and USPS-T-48 offers proposed changes to the Domestic Mail
179
Notice of the United States Postal Service for Establishment of Settlement Procedures for Forever
Stamp Proposal, July 14, 2006, at 2.
180
USPS-T-48 at 19 (Taufique).
181
USPS-T-32 (Taufique) at 4, 26 and 27.
Docket No. R2006-1
- 143-
Initial Brief of the OCA
Classification Schedule (DMCS) and explains the Postal Service’s rationale for the
forever stamp, its availability, use, value, the market research that supports the
proposed Forever Stamp and the financial impact on the Postal Service.
182
The Postal Service has expended considerable time and effort to develop the
Forever Stamp mail classification and present it to the Commission in this proceeding.
The Postal Service is to be commended for its extensive cooperation with the OCA and
the Greeting Card Association (GCA) in developing the Forever Stamp proposal. The
Forever Stamp can provide important consumer benefits, and should be recommended
by the Commission in a form that achieves the goal of an “unprecedented level of
convenience” for individuals and smaller mailers.
1.
The Postal Service’s Forever Stamp proposal is a desirable
postage option for individuals, smaller mailers and the Postal
Service, and has minimal financial risk.
(a)
The Forever Stamp proposal offers greater convenience to
individuals and smaller mailers, and simplicity for the Postal
Service.
The Forever Stamp proposal offers greater convenience to individuals and
smaller mailers, and would simplify the rate change process for the Postal Service.
The Postal Service proposes the Forever Stamp to provide “an unprecedented level of
convenience for domestic postal rate payers as they adjust to the transition during rate
182
See generally USPS-T-48 (Taufique).
Docket No. R2006-1
- 144-
Initial Brief of the OCA
changes from one basic First-Class Mail rate to the next.”183 The Forever Stamp is
184
“intended to be . . . a convenience for single-piece mailers.”
Rate changes can be a source of inconvenience for some individuals and
smaller business mailers. Increases in postal rates are considered an inconvenience
by individuals and smaller mailers.185 In addition to an undesirable rise in postage costs
for mailers, special trips are sometimes needed to comply with the new postage rates.
At the time of a rate change, individuals and smaller mailers may be holding an
inventory of single-piece First-Class Mail postage stamps. Mailers are then required to
purchase make-up stamps in order to use-up their existing inventory.
Even when postage increases are announced sufficiently in advance for the
purchase of stamps denominated at the higher rate, often-times, stamps with the new
denominations are not available far enough in advance to permit customers to
conveniently purchase the new stamps. Customers are then faced with two choices.
They may purchase fewer stamps at the old rate than they would normally purchase
and make a return trip to the post office sooner than usual to purchase stamps at the
new rate when they become available. Alternatively, they may purchase the usual
amount of stamps at the old rate knowing they will last beyond the implementation date
of the rate increase and, at the same time, purchase enough make-up stamps to use
up the old-rate stamps that will be remaining on the date of implementation.
Sometimes, however, even make up stamps are not available due to an increase in
183
USPS-T-32 (Taufique) at 4.
184
USPS-T-48 (Taufique) at 5.
185
See generally USPS-LR-L-152 at 5, Insights #7, 8, and 9 (Forever Stamp Market Research).
Docket No. R2006-1
- 145-
Initial Brief of the OCA
demand for low value stamps around the time of a planned rate increase and a special
trip to purchase stamps may be necessary. Thus, rate changes often have the effect of
causing unscheduled trips to retail post offices to purchase new or make up postage.
Moreover, the inconvenience of unscheduled trips to retail post offices is
sometimes compounded when the Postal Service has difficulty distributing new rate
stamps or make-up rate postage to Post Offices or retail outlets when they are needed
by customers, despite the Postal Service’s best efforts.
186
By eliminating the Postal
Service’s rush to distribute stamps prior to a rate increase, the proposed Forever Stamp
holds the potential to simplify Postal Service management and administration of rate
changes applicable to individuals and smaller mailers.
The Forever Stamp would offer mailers greater convenience in dealing with rate
changes. In the future, mailers that obtain Forever Stamps can use such stamps
before and after a rate change, avoiding the need for unscheduled trips to retail post
offices for new-rate or make-up rate postage stamps.
187
For mailers that are concerned
about higher rates, the Forever Stamp would permit such mailers to defer or possibly
avoid higher rates through the use of Forever Stamps purchased at the lower rate in
advance of rate changes. To the extent the Postal Service is moving toward annual
postal rate increases,188 the availability of the Forever Stamp will minimize the further
inconvenience associated with annual rate changes on individuals and smaller mailers.
186
USPS-T-31 (O’Hara) at 6.
187
This benefit will not occur during this rate cycle unless mailers are willing to use the Forever
Stamp, to be initially offered at a price of $0.42, as postage for a $0.39 letter prior to the rate increase in
this docket.
188
“USPS Strategic Transformation Plan 2006-2010,” September 2005 at 6.
Docket No. R2006-1
- 146-
Initial Brief of the OCA
Another potential advantage of the Forever Stamp is that individuals and smaller
mailers may be more disposed toward the purchase and use of First-Class Mail
postage. Responses to the market survey undertaken by the Postal Service indicated
that by increasing convenience, “the Forever Stamp could encourage more use of the
mails than might otherwise occur.”189
However, achieving these beneficial results will depend in some measure upon
the rules in the DMCS and the DMM applicable to the Forever Stamp. Its introduction
should include clear statements as to the availability, applicability, value, and terms of
use. Unless the introduction of the Forever Stamp is explained fully to the public, many
of the benefits and goodwill that such a product would garner may be reduced or even
lost. These issues are discussed below in another section.
(b)
The Forever Stamp proposal could improve Postal Service
administration by simplifying the rate change process and
thereby reducing postal costs.
The Forever Stamp also has the potential to ease the Postal Service’s transition
to new postal rates. Once introduced, the continuous availability of the Forever Stamp
would eliminate the need to rush-order printing of new First-Class Mail postage stamps
and their distribution to retail post offices in response to rate changes. Moreover, the
Forever Stamp could reduce or eliminate the need to print and maintain an inventory of
make-up rate postage stamps. As a result, distribution difficulties, in which new-rate
189
USPS-T-32 (Taufique) at 27.
Docket No. R2006-1
- 147-
Initial Brief of the OCA
and make-up rate postage stamps are distributed in limited quantities to some retail
post offices, but not to all others, would be a thing of the past.
190
In addition, introduction of the Forever Stamp would, over time, make more
postage sales possible through a larger network of private retailers. Private retailers
offering postage stamps would no longer have to worry about having outdated
denominated postage because the Forever Stamp would never have to be recalled as a
result of a rate change. This in turn would simplify Postal Service management of the
Forever Stamp inventory thereby reducing administrative costs of the Postal Service.
Continuous sale of Forever Stamps through a larger number of private retailers could
also reduce unscheduled trips by mailers to retail post offices at the time of rate
changes and so contribute to a reduction in window service transactions and costs after
such a change and even throughout the year.
(c)
The financial impact of the Forever Stamp will be minimal.
Witness Taufique estimated the financial exposure to the Postal Service to result
from introduction of the Forever Stamp.
191
There is no exposure to financial risk during this rate cycle because the Postal
Service is proposing to introduce the sale of the Forever Stamp at the 42 cent rate
rather than the currently effective rate, even though it will be introduced before the new
rates become effective.192 Although there will be no financial cost to the Postal Service
190
See, for example, “Postal Customers Are Angry As Cheap Stamps Sell Out,” Sacramento Bee,
January 13, 2006, at D1.
191
USPS-T-48 at 22, note 14 (Taufique).
192
Id. at 20.
Docket No. R2006-1
- 148-
Initial Brief of the OCA
during this rate change, consumers would, nevertheless, derive some benefit if the nondenominated Forever Stamp is more widely available well before this rate increase than
the amount of time stamps were available prior to previous rate increases. Another
advantage of approval of the Forever Stamp DMCS language in this proceeding is that
it will permit the Postal Service to advertise the Forever Stamp to potential purchasers
and to assure them that the Forever Stamp they purchase during this rate cycle will be
valid forever as first-ounce letter mail postage.
Witness Taufique also notes that “experiences of foreign postal
administrations…suggest that [Forever Stamps] have had no effect, or a positive effect,
193
on revenues and costs.”
He provides a very rough estimate of the potential financial
exposure in future rate cases, stating that it “is not difficult” to estimate a $100 million
exposure in the first month after a rate increase.
While witness Taufique recognizes this is a conservative amount, in OCA’s view
the $100 million estimate in the first month after a rate increase is so conservative as to
be very improbable. That figure represents the absolute maximum exposure possible.
If rate cases occur every year and are limited to the increase in the cost of living, the
rate increase is not likely to be three cents but would be, instead, either one cent or two
cents. That fact alone would reduce the financial exposure using witness Taufique’s
method to $33 million or $66 million, respectively.
Witness Taufique’s estimate also assumes that every single piece of First-Class
Mail for one month after implementation of a three cent rate increase (about 330 million
pieces) will carry a Forever Stamp purchased prior to the rate change at the previous
193
USPS-T-48 at 20 (Taufique).
Docket No. R2006-1
- 149-
Initial Brief of the OCA
lower rate. That is unrealistic. A significant portion of mailers would not have the
opportunity or have the inclination to make a special trip to purchase Forever Stamps in
order to save a few pennies on their postage after a rate change. Nor is it reasonable
to assume that every mailer having Forever Stamps available will use them on every
one-once letter in the first month after a rate change.
Additionally, witness Taufique’s conservative analysis does not take into account
the dampening effect on demand of only selling the stamps in booklet form. Small
businesses and even individuals normally handling rolls of stamps may not bother with
Forever Stamps booklets, and some purchasers of only a few stamps would not want to
purchase a full booklet of Forever Stamps. Thus, in neither case would they take
advantage of the Forever Stamp discount and the financial impact would be reduced
ever further.
Moreover, there are offsets to the costs that are of value to the Postal Service
and others. There will be some savings in window service costs because there will be
fewer stamp purchases around the time of a rate increase, particularly costly stamp
purchase transactions. There is also the time value of the money received by the
Postal Service in advance of the actual use of Forever Stamps.
Of course, witness Taufiques’s estimate does not measure the savings in other
costs, such as the savings to mailers in time and money of eliminating additional trips to
purchase stamps at the new rate.
Thus, although witness Taufique’s estimate is not so large in OCA’s view as to
outweigh the very positive advantages of the Forever Stamp’s convenience and
Docket No. R2006-1
- 150-
Initial Brief of the OCA
goodwill, the actual financial risk to the Postal Service from the Forever Stamp is
considerably lower than the costs estimated by witness Taufique.
2.
The Postal Service’s cooperative development of the Forever
Stamp proposal is commendable.
The Postal Service has expended considerable time and effort to develop the
Forever Stamp mail classification and present it to the Commission in this proceeding.
Since conclusion of the omnibus rate proceeding in Docket No. R2005-1, the Postal
Service has cooperated fully with the OCA and the Greeting Card Association (GCA)
and proceeded with expedition to develop this proposal. Its efforts and cooperation are
much appreciated.
The idea of a non-denominated postage stamp to ease the transition to new
rates for individual mailers was originally introduced by the OCA and others at the May
2002 Ratemaking Summit.
R2005-1,
195
194
Pursuant to an agreement with the OCA in Docket No.
the Postal Service established a working group that met regularly with the
OCA and the GCA between November 2005 and May 2006. Those meetings identified
the need for specific research, and discussed issues surrounding introduction of a
Forever Stamp. As part of the identified research, the OCA conducted a multi-country
survey concerning the availability of non-denominated postage, and the terms and
features of its use, which is presented in OCA-LR-1.196 With input from the participants
194
See “Ratemaking Summit, Day One,” March 28, 2002, Tr. 1/20, 21, 32, and 53-54.
195
“Office of the Consumer Advocate Notice of Receipt of Letter from Postmaster General Potter
Detailing the Agreement Reached Between the Postal Service and OCA,” July 25, 2005.
196
OCA-LR-1: “Materials Collected by the Office of the Consumer Advocate Concerning the Use of
a Non-Value-Indicated Stamp in the United Kingdom and France.”
Docket No. R2006-1
- 151-
Initial Brief of the OCA
of the working group, the Postal Service committed the personnel and financial
resources necessary to conduct the Forever Stamp market research contained in
USPS-LR-L-152,
197
and prepared the paper entitled “Forever Stamp Issues for
Consideration.”198
Throughout the development of the Forever Stamp proposal, the Postal Service
was receptive to ideas and suggestions presented by the OCA and GCA. Many of
those ideas and suggestions were incorporated by the Postal Service into its marketing
research, including reorganizing and expanding the market research survey to more
deeply explore respondents’ views of the Forever Stamp concept as well as potential
purchasing patterns.
B.
The Commission Should Recommend Witness Carlson’s Proposed
DMCS Language to Resolve Ambiguities Created by the Postal Service’s
Proposed DMCS Language About the Use of the Forever Stamp 199
1.
The Postal Service’s proposed DMCS language is ambiguous
about the weight and classes of mail on which the Forever Stamp
may be used.
The testimony of witness Taufique (USPS-T-48) proposes changes to the
Domestic Mail Classification Schedule (DMCS) to establish the Forever Stamp as a
197
USPS-LR-L-152: “Forever Stamp, Report of Quantitative Market Research Conducted by Opinion
Research Corporation, April, 2006.”
198
Tr. 19/6843-48 (DFC/USPS-T48-22).
199
See generally USPS-T-48 (Taufique).
Docket No. R2006-1
- 152-
Initial Brief of the OCA
“means for applying postage to First-Class Mail letters that would not expire with future
200
rate changes.”
The proposal would not establish a subclassification of mail.
Witness Taufique proposes changes to the DMCS in two parts. First, DMCS
section 241, “Forever Stamp:”
Postage for the first ounce of a First-Class Mail Single-Piece letter may be
paid through the application of a Forever Stamp. The Forever Stamp is
sold at the prevailing rate for Single Piece Letters, First Ounce, in Rate
Schedule 221. Once purchased, the Stamp may be used for first-ounce
letter postage at any time in the future, regardless of the prevailing rate at
the time of use.201
A second part of the proposed change is to add a sentence to DMCS section 3030:
The use of the Forever Stamp, as described in section 241, is considered
full prepayment of postage for the first ounce of First-Class Mail, Single
Piece Letters. 202
Intervenor Douglas F. Carlson (DFC-T-1) maintains, and the OCA agrees,
that the Postal Service’s DMCS language “is not a model of clarity.”203 In
response to discovery requests, the Postal Service states that the “correct”
interpretation of its DMCS language is that “the Forever Stamp is intended for
use on single-piece First-Class Mail one-ounce letters [and . . . ] excludes [use
on] the first-ounce rate component of letters weighing more than one ounce.”204
(Emphasis added). However, with some degree of understatement, the Postal
200
DFC/USPS-80(c), Revised December 6, 2006.
201
USPS-T-48 (Taufique) at 17.
202
Id.
203
DFC-T-1 (Carlson) at 21.
204
Tr. 18C/6123 (DBP/USPS-510(b)).
Docket No. R2006-1
- 153-
Initial Brief of the OCA
Service also allows that the term “first ounce” as used in its DMCS language “is
possibly ambiguous.”
A plain reading of proposed DMCS section 3030 and Postal Service statements
interpreting such section appear straightforward: the Forever Stamp is only intended to
be valid postage for the first ounce of postage for single-piece First-Class Mail letters
and not for heavier mail or other classes of mail. Absent from the definition, but easily
presumed, is that the Forever Stamp will not be allowed for postage for ounces greater
than the first ounce or for classes of mail other than First-Class. Nevertheless, the
Postal Service states that “the proposed DMCS changes do not explicitly prohibit the
use of Forever Stamps for postage applications other than single-piece First-Class Mail
205
one-ounce letters.”
The reality facing the Postal Service is that even if the Postal Service attempts to
restrict usage to single piece one-ounce letters, “some mailers will at times use the
Forever Stamp for an unintended purpose, whether a First-Class Mail letter weighing
more than one ounce, a First-Class Mail flat or parcel, or another mail class
altogether.”206 Where mailers use the Forever Stamp for “an unintended purpose,” the
Postal Service has clearly stated in cross-examination that it “intends to give credit for
such uses at the currently applicable First-Class Mail single-piece 1-ounce letter
rate.”207
205
Id. (DBP/USPS-510(c)).
206
Id. (DBP/USPS-510(b)).
207
DPB/USPS-700. The Postal Service reached this decision belatedly, modifying its earlier
response to interrogatories. Previously, the Postal Service said it was “considering” giving credit for
postage purposes at the original purchase price. (Emphasis supplied). Tr. 18C/6123 (DBP/USPS-510(b))
Docket No. R2006-1
- 154-
Initial Brief of the OCA
In further explanation of its proposed DMCS language, the Postal Service states
that “the actual use of the Forever Stamp would seem to consist of its intended use to
pay postage for one-ounce First-Class Mali letters."208 (Emphasis in original). Yet the
Postal Service has explicitly stated, “alternative uses” on mail matter other than oneounce First-Class Mail letters will be “tolerated” through proposed changes to the
Domestic Mail Manual (DMM) §604.1.10.209
The lack of clarity within the Postal Service’s proposed DMCS and DMM
language would leave open to interpretation the postage value to be assigned to the
Forever Stamp when used by mailers on First-Class Mail heavier than one ounce or on
cards, or parcels and other subclasses or classes of mail in the future. Consequently,
the proposed DMCS language, “as currently written, runs a serious risk of making rate
changes more complicated and less convenient for customers than they are now.”210
Proposed DMM §604.1.10. would read, in part: “The postage value for each
211
forever stamp is the current First-Class Mail single-piece 1-ounce letter rate.”
With
this proposed DMM section, the Postal Service has stated it will allow the Forever
Stamp to be valued at the then current single-piece first-ounce rate when used on
mailpieces other than First-Class Mail pieces weighing one ounce, even though the
208
DFC/USPS-81(a).
209
Ibid.; see DFC/USPS-80(c); see also DFC/USPS-78(c)-(d), Revised December 6, 2006. In the
latter response, the Postal Service affirmed that customers may use Forever Stamps to pay the postage
on all mail for which customers can use stamps to pay the postage (such as, but not limited to, First-Class
flats and Priority Mail parcels) and provide postage value equal to the current rate for single-piece, oneounce First-Class letters for all mail for which customers can use stamps to pay the postage (such as, but
not limited to, First-Class flats and Priority Mail parcels).
210
DFC-T-1 (Carlson) at 20.
211
71 Fed. Reg. 56,587, September 27, 2006.
Docket No. R2006-1
- 155-
Initial Brief of the OCA
Postal Service’s proposed DMCS language does not explicitly permit what the Postal
Service apparently intends to allow through the DMM. The Postal Service maintains
that it is reserving the possibility that it may propose more restrictive amendments to the
DMCS language if it determines that alternative uses of the Forever Stamp can no
longer be “tolerated.”212 Inasmuch as the Postal Service can change the DMM without
public or Commission involvement, it is at least arguable that there will be no need for
the Postal Service to propose additional changes to the DMCS, or to request any
Commission review if the Postal Service determines such alternative uses can be
permitted, or restricted as necessary through the DMM.
In the absence of Commission (or Postal Service) clarification of the proposed
DMCS language, the DMCS would be ambiguous and could lead, in all probability, to
inconsistent enforcement in conflict with the stated intent of the proposed DMCS
language. The consequence of such a conflict will be unfair to individuals and smaller
mailers. Unfortunately, in all probability, many mailers will attempt to use the Forever
Stamp on single-piece First-Class Mail additional ounce letters, flats and parcels, as
well as Parcel Post parcels, and for some special services in the same manner as they
do now with denominated postage stamps. If clerks and mail carriers are not
sufficiently aware of the “tolerated” alternative use anticipated but ambiguously
permitted by the DMCS and DMM language, mailpieces presented to retail clerks or left
for postal carriers may be rejected or not picked up because its is believed they bear
improper Forever Stamp postage.
212
DFC/USPS-80(c) Revised December 6, 2006.
Docket No. R2006-1
- 156-
Initial Brief of the OCA
At best, the apparent, but unspoken, restrictions in the DMCS and DMM
language and the potential for the uneven application of the Forever Stamp, are likely to
create the perception of discriminatory treatment and may lead to confusion for clerks,
mailhandlers, and mailers. This confusion may also discourage potential customers
with Forever Stamps from using the mails because they do not believe Forever Stamps
are acceptable for the item they want to send. Mailers who interpret the postal
regulations as imposing restrictions on the use of the Forever Stamp may believe there
is a “need to purchase two different types of stamps for their mail—‘Forever Stamps’
plus regular [denominated] stamps that they could use on their flats and parcels.”213
Alternatively, mailers may simply forgo the purchase of Forever Stamps and rely solely
on denominated postage for all of their mailing needs to avoid perceived problems. In
either event, the benefit of convenience of the Forever Stamp would be negated.
2.
The Postal Service’s proposed DMCS language should be modified
as proposed by intervenor witness Carlson.
Rather than relying upon ambiguous DMCS language and interpretations as
proposed by the Postal Service, the Commission should adopt the proposed DMCS
language proposed by witness Carlson. Witness Carlson’s proposed alternative
language for DMCS, section 241, prescribes more clearly the permissible use of the
Forever Stamp and eliminates ambiguity:
The Forever Stamp is sold at the prevailing rate for Single Piece
Letters, First Ounce, in Rate Schedule 221. The Forever Stamp is an
adhesive stamp within the meaning of section 3040. Once purchased, the
Forever Stamp may be used for postage equal to the prevailing rate, at
213
DFC-T-1 (Carlson) at 22.
Docket No. R2006-1
- 157-
Initial Brief of the OCA
the time of use, for Single Piece Letters, First Ounce, in Rate Schedule
221.
In OCA’s view, this proposal would eliminate the confusing ambiguity of the Postal
Services language. Unlike the Postal Service’s proposal which expressly states in
section 241 that the Forever Stamp is for “first ounce letter” postage, witness Carlson’s
proposal removes the ambiguity by eliminating those three words (first ounce letter) and
thereby being silent about the types or classes of mail on which the Forever Stamp may
be applied as postage.
It is, therefore, critical that the Commission now recommend a clarification of the
various weight and classes of mail on which the Forever Stamp may be used.
Clarification is needed to assure the Forever Stamp fulfils the goal of providing
convenience to mailers. If the Commission recommends the Forever Stamp proposal,
it should resolve any ambiguities in the proposed DMCS language in favor of
convenience to individuals and smaller mailers.
VII.
THE POSTAL SERVICE’S INSURANCE AND COD PRODUCTS HAVE A LOW
VALUE OF SERVICE THAT HAS NOT BEEN TAKEN INTO ACCOUNT IN THE
COST COVERAGES PROPOSED FOR THESE SERVICES
A.
The Cost Coverage For Insurance Should Be Set Close to Zero
The Postal Service proposes a cost coverage of 131 percent for Insurance,
which is several percent points lower than the average Special Services cost coverage
of 145 percent. 214 At first glance, this relatively low coverage would appear to be
appropriate. Closer scrutiny, however, reveals that the cost coverage of Insurance
214
Witness O’Hara’s Exhibit 31B.
Docket No. R2006-1
- 158-
Initial Brief of the OCA
should be set at the lowest possible level owing to the very poor service accorded to
Insurance purchasers.
Witness Mitchum is responsible for defending the proposed cost coverage. He
claims that “[t]he value of service to insurance customers is very high.”215 This rosecolored view of the service gives a false picture of what it is like for many insurance
purchasers when it comes time to file a claim for indemnification.
With indemnification constituting a mere 15.5 percent of Insurance revenues,216
this service stands out as offering little value for the money expended by purchasers.
Witness Mitchum explains that the insurance product is very labor intensive, involving
window clerk and carrier costs that cause nearly all of the attributable costs of the
service. When an Insurance purchaser chooses to pay for postal insurance (s)he is
largely paying for the sale at the window and the extra time expended by the carrier to
obtain a signature. Very little of the revenue taken in for this service ever applies to
customer indemnification which, after all, is the essence of what the customer hopes to
obtain.
As witness Mitchum states during oral cross-examination, he is actually
proposing a reduction in most Insurance rates because the indemnification costs are
going down.
217
He speculates that the Postal Service is doing a better job of handling
insured items, thereby causing less damage and loss than in the past;218 but he offers
215
USPS-T-40 at 28.
216
Tr. 14/4113.
217
Tr. 14/4185.
218
Id. at 4186.
Docket No. R2006-1
- 159-
Initial Brief of the OCA
no proof to support his supposition. Another possibility is that customer indemnity
claims are so mishandled by postal clerks (who prepare the paperwork and mail the
claims forms) and by the Accounting Service Center in St. Louis that claimants become
too discouraged to see the process to its conclusion.219 OCA has been contacted by a
number of such individuals and postulates that being worn down by an excruciating
claims process causes many claimants to give up their efforts to be indemnified.
Evidence provided by the Postal Service in response to discovery requests and
questions on oral cross-examination suggest that the experience of Insurance
purchasers and claimants is far less informative than it ought to be. To begin with,
owing to the expense of providing information to all Insurance purchasers and to the
perception that the incidence of loss and damage is low, purchasers are not
automatically given information on important matters ranging from how to package an
item to withstand damage, to the documents they must retain in order to establish the
value of a lost or damaged item when submitting a claim, to the rigors of submitting a
claim.
Postal Service witness Hintenach, Manager, Customer Service Operations,
explained the policies of the Postal Service this way:
220
You know, on that I guess we have not had tremendous loss or damage.
Is there a need to go to that extreme, you know, to hand everyone who
comes up something on their parcel? Now, if someone would say hey,
listen, I'm not sure whether I know this is packaged properly or not, I think
we have -- what is it -- Notice 2122. We can provide them with packaging
tips, and we can provide them on our insurance process and our claims
219
Although witness Mitchum does not hold this view, he knows that others believe that poor
handling of claims may be responsible for a reduction in indemnification costs. Id.
220
Tr. 3/378.
Docket No. R2006-1
- 160-
Initial Brief of the OCA
process and times you need to take the claim and how much you have
depending on the type of service you use.
Personally, I think it would be a tremendous cost to do something like that
for a relatively minor occurrence. It's not like we're not answering the
customer's questions if they ask it. We're trying to get them by our
customers to lead them to find out if there is something in there that could
be damaged so that we do ask and try to make sure it's packaged
properly.
Postal clerks, with only the best intentions, give customers a false sense of
security that if an insured item is lost or damaged, indemnification will be quick and
easy to obtain. The clerks have next to no training on processing Insurance claim
paperwork. Module 22, the only training given to postal clerks on Insurance, lasts a
mere 60 minutes.
There is no training at all on what a claimant is likely to encounter
once embarking on the claims process.
221
The complaints system maintained by the Postal Service keeps postal
management in the dark about the number and reasons for complaints about
Insurance. When OCA asked witness Mitchum to provide copies of the 500 most
recent Insurance complaints, he responded: “This information is not available, because
records of complaints are not identified by special service.”222
The processing of claims, once they are received by the St. Louis Accounting
Center, is agonizingly slow – an average of 48 days, when the processing is supposed
to be completed in 10 days.223 In effect, the Postal Service has established a
221
A review of all training materials provided to the clerk shows that there is no description of the
process, the percentage of claims paid/denied, the length of time for processing claims, etc. Id. at 3996 –
4002.
222
Id. at 4112.
223
Id. at 3992-93.
Docket No. R2006-1
- 161-
Initial Brief of the OCA
reasonable service standard of 10 days, but its actual performance is nearly 5 times
slower than the standard. Although the Postal Service performs poorly in many areas,
such as retail parcel post and non-dropshipped, low volume Standard Mail,
224
this
surely must be one of the worst performances for any type of mail product. It must also
be noted that the clock starts running on the 10-day standard only after the claim has
been received in St. Louis and entered into its electronic claims system.
OCA has learned from Insurance claimants who registered complaints with us
that clerks sometimes allow the Insurance Claim Form 1000 to sit in a desk drawer for
weeks or months; do not know how to fill it out correctly so that it is rejected in St. Louis
and sent back for correct processing; address the envelope incorrectly so that the Form
#1000 ends up in a Mail Recovery Center; or a correctly addressed envelope is not
processed and delivered properly so that the envelope ends up in a Mail Recovery
Center. The record indicates that Postal Service retail offices keep no logs of claims
that have been submitted,
225
so mishandling can go undetected long enough that the
claim submission deadlines are not able to be met. OCA has learned that many
claimants simply give up. This, of course, may result in a downward trend in the
amount of money paid out for indemnification.
In the view of OCA, this litany of mistakes and mishaps warrant a strong pricing
signal by the Commission – a cost coverage of set close to zero percent so that postal
management will take seriously the need to develop a fully trackable system in which
224
Only 54 percent of retail Parcel Post packages are delivered on time.
http://www.usps.com/serviceperformance/welcome.htm
See also, Testimony of Janyce Pritchard (Flute-T-1).
225
Tr. 3/385.
Docket No. R2006-1
- 162-
Initial Brief of the OCA
clerks are required to log in every complaint; log out each complaint at the time it is
mailed to St. Louis; have all processing steps logged in electronically in St. Louis and
made available to claimants on the Internet or by telephone. In addition, all individuals
who inquire about Insurance should be given printed information that explains the
documentation they will need if it is necessary to file a claim; the steps that will be
followed in processing a claim; the low incidence of damage/loss which may make it a
“poor bet” to buy Insurance at all; the length of time it takes to process a claim – 48
days for the decision to be issued plus 30 days for the first appeal. Potential
purchasers should also be informed about the percentage of claims that are denied –
approximately one-fifth so that they will understand that indemnification is far from
being automatic.
226
Finally, until the Postal Service meets its 10-day standard in 95+
percent of claims, cost coverage should be set much lower than the 131 percent
proposed in this case. Given the Postal Service’s long list of failings in the current rate
case, OCA respectfully asks that cost coverage be set at one percent above costs.
B.
The Cost Coverage of Collect on Delivery (COD) Should Be Set Near
Zero
The Postal Service’s recent revisions to its payment policies for COD are
arbitrary and unfair, as demonstrated in the testimony of Growing Family witness
Robert Paul (GF-T-1). Beginning in May 2005, the Postal Service began to reduce the
payments to Growing Family for COD packages that were delivered, but not paid for, on
a “random” basis, with no discernible pattern for the types of packages selected for
reduced reimbursement nor for the amounts paid. (GF-T-1 at 3).
226
28,500/129,249 = 22%. Tr. 14/4109.
Docket No. R2006-1
- 163-
Initial Brief of the OCA
It is especially disturbing that the Postal Service applied its policy of unexplained,
reduced payments months before it ever issued a decision that set forth the rationale
for the policy (“Sentence first, judgment later”). Id. at 4.
Owing to the dramatic reduction in the value of service made unilaterally by the
Postal Service in 2005, the Commission should reduce the cost coverage of COD to a
level close to zero. When Congress granted the Postal Service a monopoly over
access to the mailbox, a concomitant duty arose for the Postal Service to be
reasonable, open, and fair in setting policies for the services that are made available
under the Domestic Mail Classification Schedule. The Postal Service has failed utterly
to discharge this duty.