Download File

Postal Rate Commission
Submitted 12/21/2006 3:12 pm
Filing ID: 55470
Accepted 12/21/2006
BEFORE THE
POSTAL RATE COMMISSION
WASHINGTON, D.C. 20268-0001
____________________________________________
:
Postal Rate and Fee Changes, 2006
:
:
Docket No. R2006-1
INITIAL POST-HEARING BRIEF OF
THE GREETING CARD ASSOCIATION
James N. Horwood
Peter J. Hopkins
Spiegel & McDiarmid
1333 New Hampshire Avenue, N.W.
Second Floor
Washington, D.C. 20036
Telephone: (202) 879-4000
Facsimile: (202) 393-2866
E-mail: [email protected]
[email protected]
David F. Stover
2970 South Columbus Street
No. 1B
Arlington, Virginia 22206-1450
Telephone: (703) 998-2568
Facsimile: (703) 998-2987
E-mail: [email protected]
Attorneys for Greeting Card Association
December 21, 2006
TABLE OF CONTENTS
Page
I.
STATEMENT OF THE CASE ............................................................................... 1
II.
GCA’S POSITION AND SUMMARY OF ARGUMENT.......................................... 2
III.
THE COMMISSION SHOULD SUPPORT THE INTRODUCTION OF THE
FOREVER STAMP ............................................................................................... 8
IV.
THE COMMISSION SHOULD REJECT THE PROPOSAL TO “DELINK”
WORKSHARED AND SINGLE-PIECE FIRST-CLASS RATES.......................... 10
A.
Introduction.............................................................................................. 10
B.
Rates Based On Delinking Would Be Inefficient. .................................... 11
1.
The Postal Service position. ......................................................... 12
2.
Efficient component pricing as explained by witnesses Sidak
and Panzar.................................................................................... 14
3.
Delinking violates ECP. ................................................................ 15
C.
Delinking-Based Rates Would Be Unfair And Inequitable....................... 18
D.
The Delinking Proposal Is Indistinguishable In Practice From
Bifurcation Of First-Class Letters Into Workshared And SinglePiece Subclasses. ................................................................................... 20
V.
THE COMMISSION SHOULD LIMIT THE PROPOSED CHANGE TO
FIRST CLASS LETTER MAIL FIRST OUNCE RATES TO NO MORE
THAN 41 CENTS................................................................................................ 23
A.
The Relevant Market Realities Have Changed Fundamentally. .............. 23
B.
In The Face Of Electronic Competition First Class Mail Volumes
Have Significantly Eroded While Standard Mail Volumes Have
Increased Dramatically. ........................................................................... 25
C.
There Is Good Reason To Believe That Electronic Competition Will
Continue To Erode First Class Mail Volumes Given The Rapid
Growth In Broadband Deployment And Users......................................... 29
D.
The Commission Should Ameliorate Rather Than Continue To
Increase The Disparity In Institutional Cost Coverage Between
First Class And Standard Mail. ................................................................ 32
E.
Price Matters: Market Research Shows That Future Postal Rate
Increases Play A Significant Trigger Role In Further Consumer
Diversion To Electronic Substitutes......................................................... 34
F.
The Commission Should Not Look To An Infra-Marginal Class Of
Less Educated, Older, And Less Wealthy Consumers To Bear
Further Revenue Increases. .................................................................... 38
i
G.
H.
The Postal Service’s First Class Elasticity Estimates Are Not
Credible. .................................................................................................. 40
1.
Daubert analysis. ......................................................................... 41
2.
Witness Thress’s test for model selection has no basis in
the literature or the econometric community. ................................ 43
3.
There is no theoretical or real world basis supporting
Witness Thress’ choice of variables for purposes of
estimating electronic diversion of First Class single piece or
workshared mail. ........................................................................... 48
4.
Other problems in witness Thress’s estimating of First Class
elasticities. .................................................................................... 50
5.
Witness Clifton’s elasticity estimates are a better measure of
market realities.............................................................................. 54
6.
The Commission does not have to resolve the battle of
elasticities in order to adopt GCA’s position. ................................ 61
The Commission Cannot Properly Look At The Instant Proposed
Rate Increase In A Vacuum..................................................................... 62
1.
Recently enacted postal reform legislation reduces
uncertainty and alleviates the need for a 1% contingency............ 62
2.
The Commission should not look at the instant proposed
rate increase in a vacuum............................................................. 63
VI.
THE SHAPE-BASED RATES PROPOSAL NEEDS IMPROVEMENT ................ 65
A.
Introduction.............................................................................................. 65
B.
Charging Low-Aspect-Ratio Letters As Though They Were Flats Is
Bad Mail Classification Practice. ............................................................. 68
1.
Introduction. .................................................................................. 68
2.
Cost justification impossible.......................................................... 69
3.
An appropriate incentive? ............................................................. 71
C.
Why Proper Treatment Of Low-Aspect-Ratio Letters Is Important. ......... 73
1.
Low-aspect-ratio letters are important to mailers.......................... 73
2.
Price matters. ................................................................................ 76
D.
GCA’s Case For Its Proposal Has Not Been Rebutted............................ 77
E.
Conclusion............................................................................................... 80
VII.
CONCLUSION.................................................................................................... 82
ii
TABLE OF AUTHORITIES
FEDERAL CASES
Amorgianos v. Nat’l R.R. Passenger Corp., 303 F.3d 256 (2d Cir. 2002)..................... 47
Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993)................. 40-42, 46
Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137 (1999)............................................ 41
Lobsters, Inc. v. Evans, 346 F. Supp.2d 340 (D. Mass. 2004)...................................... 42
Pasha v. Gonzalez, 433 F.3d 530 (7th Cir. 2005) ......................................................... 42
Village of Bensenville v. FAA, 376 F.3d 1114 (D.C. Cir. 2004) ..................................... 62
FEDERAL STATUTES AND REGULATIONS
5 U.S.C. § 556(d) ..................................................................................................... 73-75
39 U.S.C. § 3622(b) ................................................................................................ 12, 40
39 U.S.C. § 3622(b)(1)........................................................................................ 8, 18, 80
39 U.S.C. § 3622(b)(4).................................................................................. 4, 21, 67, 80
39 U.S.C. § 3622(b)(8)............................................................................................ 67, 80
39 U.S.C. § 3624(a) ...................................................................................................... 73
OTHER
39 C.F.R. § 3001.31(a) ................................................................................................. 42
39 C.F.R. § 3001.31(k)(iv)(e) ....................................................................... 42, 46-47, 52
39 C.F.R. § 3001.54(i) .................................................................................................. 40
iii
I.
STATEMENT OF THE CASE
This case is the first fully litigated comprehensive postal rate increase
proceeding since 2000 (R2000-1). In addition to rate increases for all classes of mail
(totaling almost $4 billion), the Postal Service proposes a classification change to
introduce the Forever Stamp to be available for First Class first ounce single piece
Letter mail regardless of the prevailing first ounce rate. Not surprisingly, intervenors
challenged virtually every rate increase proposal as being excessive for the class or
subclass affecting each intervenor.
Unique among the intervenors, the Greeting Card Association (“GCA”) is
representing solely the interests of the citizen mailers, those who mail First Class single
piece mail.1 GCA is the association of greeting card publishers – ranging from wellknown large corporations to small family-run “Mom and Pop” small businesses. U.S.
consumers spend approximately $7.5 billion a year on greeting cards; most of those
purchased travel through the mail, predominantly as First-Class letters. Greeting cards
account for most of the household-household letter mail the Postal Service carries.
Because Americans’ ability to communicate their thoughts and wishes to one
another through greeting cards depends on affordable, reliable, universal postal
services, the interests of GCA cause it to be an advocate for the citizen mail user. The
single-piece First-Class mail senders and recipients whose interests GCA here
1
The Commission’s Office of Consumer Advocate (“OCA”) and the American Postal
Workers Union (“APWU”) also are advocating for the interests of First-Class singlepiece mailers, along with the interests of mailers using other than First-Class singlepiece mail.
1
advocates will be substantially affected by the outcome of a rate increase for that class
of mail that results from this proceeding.
II.
GCA’S POSITION AND SUMMARY OF ARGUMENT
GCA presented the testimony of five witnesses dealing with what can be
grouped into two issues – (1) the Postal Service’s proposed increase in the surcharge
from 13 cents to 20 cents per piece for low-aspect ratio Letters (the “square card”
issue) and (2) the proposed increase in the first-ounce rate for single-piece First-Class
Letter mail. This brief will show that (1) the increase in the surcharge has not been
justified as it would apply to square greeting cards, and (2) the increase in single-piece
First-Class Letter mail rates should be limited to a first-ounce rate of no more than 41
cents, so that the unfair and economically irrational trend toward ever-greater
institutional cost burdens on first class is at least moderated. In addition, we discuss
two other matters – (3) the Postal Service’s proposal to introduce a Forever Stamp,
which GCA supports, and (4) the proposal to delink the rates from workshared and
single-piece First-Class mail, which GCA opposes.
The Forever Stamp. The Postal Service’s proposal to introduce a stamp that will
be valid for single letter-shaped pieces regardless of the prevailing first-ounce rate has
not been opposed by the testimony of any party, although one intervenor has raised
two objections regarding details of the proposed implementation. GCA welcomes the
Postal Service’s proposal as providing an important benefit for citizen mailers,
particularly mailers of greeting cards. GCA supports acceptance of the Forever Stamp
as proposed (although we would not object to the change in implementation as
proposed).
2
Square Card Surcharge. The Postal Service has proposed separate first-ounce
rates for Letters, Flats, and Parcels based on its position that today shape rather than
weight is the main influence on First-Class Letter costs. In general, this proposal
makes sense and deserves favorable consideration, but it is not problem-free. If
accepted, the Commission should recommend an important – though simple, and not
costly – adjustment. This adjustment would create a sub-category of “low-aspect ratio
Letters” – single-piece Letters with an aspect ratio less than 1.3:1 – and attach to it the
existing 13 cent non-machinable surcharge, rather than charging such pieces the 62
cent rate for a single-piece Flat as proposed by the Postal Service. The Postal
Service’s 62 cent proposal would levy an unjustified (and unjustifiable) 54-percent
increase in the extra cost of sending a square greeting card. Given the importance of
square designs to greeting card users, and the likelihood that increasing the additional
cost of mailing such cards would discourage their use, the Commission should make
this simple adjustment to the shaped-based rate proposal.
The Postal Service’s design proposal combines the existing definition of
machinability – which controls liability for the non-machinable surcharge – with a new
rate distinction by basic shape (Flat, Letter, or Parcel). This misalliance produces a
structure in which a Letter-shaped piece that fails any one of the current machinability
tests is treated, for rate purposes, as it were a Flat. Such a piece would still be
processed and delivered as a Letter. But, as regards processing and delivery, nothing
has changed that warrants the proposed increase in square card rates.
The issue is important for consumers because the square or near-square
greeting card is a distinctive, highly valued medium of personal communication.
3
Penalizing it through an unjustified 54 percent increase in the extra cost of sending
what is still a Letter, though a non-standard one, can be expected to discourage its use,
and in the process deprive the Postal Service of high-contribution First-Class volume.
This effect is not merely unfair to consumers; it also demands the Commission’s
attention under 39 U.S.C. § 3622(b)(4) (“effect of rate increases upon the general
public”).
Subjecting mailers of square greeting cards to the proposed $0.62 cent flat rate
can in no way contribute to the Postal Service’s goal of converting flat volume into letter
volume. Moreover, adoption of the sub-category proposed by GCA would not defeat
the objective of the Postal Service to increase rates for lighter weight flats and parcels
that may not otherwise be recovering their costs – a problem that has not been
demonstrated for low aspect ratio letter sized pieces, i.e. square envelopes.
First Class Letter Mail Rates. Since the last fully litigated rate proceeding in
2000, market realities central to postal ratemaking have changed. First Class mail
volumes have eroded substantially and continuously. Standard mail volumes have
increased substantially and continuously. There has been explosive growth in
broadband deployment and broadband usage. These facts are not in dispute.
Despite these facts, the Postal Service has proposed FCLM first ounce rates
that would increase the institutional cost coverage for FCLM and also would increase
the disparity in institutional cost coverage between FCLM and Standard mail. This
disparity has grown considerably since 1995. In light of the present and expected
market realities, the Commission should recommend rates that ameliorate this disparity
which is inequitable and ultimately likely to be self-defeating.
4
Significantly, the Postal Service projects recent market trends to continue: First
Class mail volumes will continue to erode and Standard mail volumes will continue to
increase. The Postal Service’s estimate in this proceeding is that in two years
Standard mail will become the dominant mail class and the volume of Standard mail will
exceed the volume of First Class mail. And this relationship is expected to continue
into the future. The primary cause of the decline in First Class mail volumes is (and will
continue to be) electronic diversion. First Class mail volumes lost to electronic
diversion will not subsequently return to the mail system. In contrast, the dynamic is
very different with respect to Standard mail. That mail has grown in volume and is
expected to continue to grow, notwithstanding the growth in Internet and, in particular,
broadband usage.
In addition, unrebutted forensic-quality market survey research shows that the
frequency and magnitude of FCLM rate increases are likely to be important behavioral
triggers as regards consumer decisions to shift FCLM volumes to electronic substitutes.
Stated simply, price matters.
The record evidence also shows that electronic diversion is positively correlated
with education and income. Whether described as “infra-marginal” customers, or
“hard-core” mailers, the Postal Service should not be looking to saddle the poor and
less educated with inequitable rate increases.
The Postal Service’s attempt to justify its proposed rate increase to 42 cents
based upon its proffered elasticity evidence is without merit. The Commission cannot
properly rely upon the Postal Service’s FCLM elasticity estimates. The elasticity study
presented by GCA’s economic witness shows that the elasticity for FCLM is higher than
5
Standard mail; the Postal Service’s witness study concludes that Standard mail, while
highly inelastic, is less inelastic than FCLM. The FCLM elasticity calculated in that
study is not credible and is entitled to no weight. The Postal Service’s witness did not
rely upon any accepted form of model selection, which is an unacceptable departure
from accepted econometric practice. Moreover, the witness did not look out into the
real world and decide what variable made the most real world sense as an explanatory
factor to measure and predict electronic diversion and simply see what results such an
empirically grounded model would produce. First Class mail volumes are eroding
because of emerging electronic, and in particular broadband, competition. An elasticity
model that attempts to estimate future First Class mail volumes but is not properly
designed to account for the effect of electronic diversion, is fundamentally flawed and
unreliable.
At the end of the day, however, the elasticity battle is subsumed by
considerations of the broader market realities. Regardless of whether the Commission
credits GCA’s or the Postal Service’s measures of First Class elasticity, or instead,
relies upon its own discretion and ratemaking expertise, principles of ratemaking policy
informed by the new market realities warrant (at most) an increase in FCLM rates to no
more than 41 cents.
Delinking. The Postal Service’s proposal to “delink” single-piece and
worksharing rates in First Class would scrap the Commission’s longstanding approach
to developing worksharing rates by discounts from a single-piece benchmark cost
based on costs that the worksharing avoids. The new approach would substitute
“bottom-up” development of separate workshared and single-piece revenue
6
requirements on the basis of Cost and Revenue Analysis (CRA) costs, obliterating the
distinction between cost savings from substitution of mailer activity for a Postal Service
operation and cost reductions due to “intrinsic” features of particular First-Class
mailpieces that have been excluded from past Commission definitions of costs avoided.
The Commission should reject the delinking proposal and adhere to the
established system of constructing worksharing discounts. Rates based on a delinking
policy would be inherently inefficient, unfairly allocate the benefits of lower-cost mail
among mailer groups, and result in a thinly-disguised but effective bifurcation of FirstClass Letters into the same separate subclasses the Commission rejected in Docket
MC95-1.
The Commission has repeatedly pointed out that, to be efficient, worksharing
rates must conform to the efficient component pricing (ECP) principle. Delinking-based
rates would violate a proper application of that principle.
The CRA-based delinking proposal cannot meet the ECP test, because it fails –
indeed, appears to fail deliberately – to distinguish between (1) costs saved (avoided)
by the mailer’s performing a function the Postal Service would otherwise have to
perform for it, and (2) reductions in the cost of a function the Postal Service must still
perform, which result from certain intrinsic characteristics of the mailpieces in question.
This distinction is of central importance. If the Postal Service does not
“compete” in the production of a mail characteristic, because the related function is not
one the Postal Service performs, then no efficiency-related purpose is served by
offering a price reduction for the presence of that characteristic. The Postal Service
does not formulate addresses or apply them to mail. Address quality therefore is not a
7
case where mailer activity can – and should, if lower in cost – be substituted for Postal
Service activity. It is a mailer activity pure and simple.
Section 3622(b)(1) of the Act requires the Commission to recommend a “fair and
equitable schedule” of rates. First-Class rates based on delinking would not meet this
standard. Delinking treats pieces having cost-reducing intrinsic characteristics as
eligible for reduced rates if they meet certain other tests but fails to recognize these
same characteristics in single-piece mail. This is unfair on its face; and, as we have
shown above, the unfairness is not excused by any gain in efficiency.
The delinking proposal is indistinguishable in practice from bifurcation of FirstClass Letters into workshared and single-piece subclasses. The cost bases of the two
proposed categories are precisely what they would be in a separate-subclasses
proposal, and the Postal Service’s suggestion – there is nothing that could reasonably
be called a commitment – that unit contributions would be equal is wholly inadequate to
prevent delinking from producing a de facto two-subclass structure in First-Class
Letters.
III.
THE COMMISSION SHOULD SUPPORT THE INTRODUCTION OF THE
FOREVER STAMP
As described in the testimony of Postal Service witness Taufique, the Postal
Service proposes a classification change “introducing a stamp that will be valid for First
Class Mail first ounce postage for single letter-shaped pieces, regardless of the
prevailing first-ounce rate.” (USPS-T-48 at 3). The proposed new stamp is colloquially
referred to as the Forever Stamp because consumers will be able to use it for first
ounce postage for single FCLM pieces even following changes in FCLM rates it. GCA
favors the implementation of the Forever Stamp.
8
The primary purpose of the Forever Stamp is to smooth transitions between rate
cases. In the face of an impending rate change, consumers could purchase Forever
Stamps and know that the Forever Stamps could be used for FCLM postage regardless
of the date of the rate change, and without needing to purchase “make up” stamps to
bridge the difference between the new and old rates. (USPS-T-48 at 4-6). From the
Postal Service’s perspective the Forever Stamp should simplify retail transactions
surrounding rate changes and may ease some of the administrative burden of
implementing new stamps or ensuring adequate supplies of otherwise necessary make
up stamps. (Id. at 7, 19 and 21).
Market research shows that consumers and businesses are likely to be receptive
to the Forever Stamp provided that it is not sold at a premium. (Id. at 13-15). In
addition, Great Britain and France have successfully implemented non-value indicated
stamps akin to the Forever Stamps. “According to the OCA, the experience in both
countries is that such stamps have been used successfully for approximately 14 years,
and they are viewed as a permanent option for customers.” (Id. at 12).
The Postal Service intends to make the Forever Stamp available at the FCLM
first ounce rate recommended and approved in this proceeding. Under this proposal the
Forever Stamp “would have no financial effect in the test year on revenues … different
from that estimated in the current filing for First-Class Mail.” (Id. at 20). The Postal
Service recommends monitoring the implementation and usage of the Forever Stamp to
guard against potential adverse consequences down the road, and Postal Service
witness Taufique notes that even a potential modest reduction in revenues could be
9
offset by an increase in customer good-will or possible retardation of electronic
diversion due to increased convenience.
No party opposes the introduction of the Forever Stamp. Intervenor Douglas
Carlson raises two objections to the Postal Service’s proposed implementation of the
stamp. Mr. Carlson contends that the proposed classification language can be read
such that the Forever Stamp could not be used except as postage on the first ounce of
FCLM. The Postal Service states that it is considering treating the Forever Stamp as
postage equivalent to its purchase price if used for other than the first ounce of FCLM.
Mr. Carlson proposes classification language that “Once purchased, the Forever Stamp
may be used for postage equal to the prevailing rate, at the time of use, for Single
Piece Letters, First Ounce, in Rate Schedule 221.” (DFC-T-1 at 28). Mr. Carlson also
contends that if the Forever Stamp is sold prior to the date of the rate change it must be
sold at 39 cents because that is the prevailing rate. The Postal Service’s apparent
position is that for implementation purposes only, prevailing rate can be read to mean
the approved and recommend rate. GCA disagrees with witness-intervenor Carlson that
the Postal Service’s positions are unlawful. From a policy perspective, GCA supports
the recommendation of either the Postal Service or Mr. Carlson regarding their
proposed treatments of these two implementation issues.
IV.
THE COMMISSION SHOULD REJECT THE PROPOSAL TO “DELINK”
WORKSHARED AND SINGLE-PIECE FIRST-CLASS RATES
A.
Introduction.
The Postal Service proposes that the Commission “delink” single-piece and
worksharing rates in First Class. Delinking would scrap the Commission’s longstanding
approach to developing worksharing rates by discounts from a single-piece benchmark
10
cost based on costs that the worksharing avoids. This approach has been refined by
the Commission over several rate cases to include a well-defined set of mail
processing and in-office delivery costs, and exclude other cost differences for reasons
clearly enumerated in various Commission Opinions and Recommended Decisions.
The new approach would substitute “bottom-up” development of separate workshared
and single-piece revenue requirements on the basis of Cost and Revenue Analysis
(CRA) costs, obliterating the distinction between cost savings from substitution of
mailer activity for a Postal Service operation and cost reductions due to “intrinsic”
features of particular First-Class mailpieces that have been excluded from past
Commission definitions of costs avoided.
GCA urges the Commission to reject the delinking proposal and adhere to the
established system of constructing worksharing discounts. Rates based on a delinking
policy would
·
be inherently inefficient,
·
unfairly allocate the benefits of lower-cost mail among mailer groups, and
·
result in a thinly-disguised but effective bifurcation of First-Class Letters
into the same separate subclasses the Commission rejected in Docket
MC95-1.
B.
Rates Based On Delinking Would Be Inefficient.
The Commission has repeatedly pointed out that, to be efficient,
worksharing rates must conform to the efficient component pricing (ECP) principle.2
2
See, e.g., PRC Op. MC95-1, ¶¶ 3074 ff., 4261 ff. (especially ¶¶ 4268-4272); PRC Op.
R2000-1, ¶¶ 5535 ff., 5805.
11
Delinking-based rates would violate that principle. In this section of our Brief, we show
why.
ECP is comprehensively discussed in the testimonies of NAA witness
Sidak (NAA-T-1 at 6 ff) and Pitney-Bowes witness Panzar (PB-T-1 at 16 ff).
Significantly, it is barely mentioned by Postal Service witness Taufique.3
1.
The Postal Service position.
As Mr. Taufique presents it, the Postal Service’s rationale for delinking
seems to be only that (i) cost data exist which allow it to be done, (ii) worksharing
mailers believe that their rates should reflect cost characteristics “not expressly
associated with the worksharing activity” (USPS-T-32 at 13), and (iii) various debates
over the proper way to construct worksharing rates can be “silenced.” (Id. at 15).
Whether the resulting rates would be economically efficient is not considered.
Postal Service rebuttal witness Robinson, on the other hand, argues that
various opposing parties overemphasize ECP, when they should “appropriately
balance[ ] the economic efficiency goal which underlies ECP with a full consideration of
the context in which postal rates are established.”4 This “context” appears to be the full
gamut of statutory directives. More specifically, after acknowledging that the § 3622(b)
3
Mr. Taufique mentions ECP once, in a quotation from a Commission opinion, when
discussing a detail of his Presort rates. USPS-T-32 at 31. He does not discuss
whether bottom-up costing as the basis of a single-piece/presort division violates ECP.
(His quotation from the Commission’s R2001-1 decision tends to show that it does.)
4
USPS-RT-10 at 1. She later disagrees with APWU, NAA, Pitney Bowes, and ValPak
witnesses, whom she characterizes as urging that “if the theory of Efficient Component
Pricing is not strictly adhered to, then the resulting prices will not be economically
efficient, and therefore, in some sense, they would be inappropriate for the Commission
to recommend.” (Id. at 3).
12
pricing criteria “are applied at the subclass level,” Ms. Robinson goes on to argue that,
even so,
. . . once markups are established, the factors enumerated in the
pricing criteria do not cease to be relevant in rate design. At every
step, the Commission must consider not only the cost justification
for a given price, but also the entire context within which the price
was developed.
(Id. at 6). This may or may not be true5, but it would not excuse beginning the process
by adopting a costing procedure guaranteed to produce inefficient rates. Ms.
Robinson’s argument boils down to insisting that, because ratemaking as a whole
invokes all the statutory factors, ECP should not determine the cost basis from which
the setting of particular rates starts out. But – unless efficiency is to be disregarded in
5
Ms. Robinson’s example from Docket R2000-1 certainly does not support her
argument. The Commission did indeed discuss the matters she lists, and discussed
them in connection with the single-piece First-Class rate, but – as she omits to mention
– it did so explicitly on the basis that “[a]ll of the First-Class letters and flats
worksharing discounts are set in relation to this rate.” PRC Op. R2000-1, ¶ 5058. In
other words: the Commission was proceeding in a manner entirely consistent with ECP
(as it made clear in ¶ 5060 of the same Opinion). Ms. Robinson mistakenly treats
“single-piece First-Class Mail rates” as though the single-piece first-ounce element
which the Commission discussed had been a coordinate subdivision of First-Class
Letters, rather than – as the Commission made plain – the basis for constructing all the
First-Class worksharing rates.
A similar distortion of the R2000-1 Opinion occurs at USPS-RT-10 at 4.
Ms. Robinson cites a Commission statement that “[e]conomic efficiency is neither the
exclusive nor even the paramount ratemaking objective under the Act” without
acknowledging that the Commission was there discussing a pure pricing issue –
specifically, whether Ramsey pricing was the proper way to assign markups to the
various subclasses. PRC Op. R2000-1, ¶¶ 4041-4042. This step is completely
separate from determining attributable costs (including costs avoided by worksharing).
Ms. Robinson has confused allocative efficiency (catered for by Ramsey pricing and
similar relative-demand-driven schemes for assigning non-attributable costs) with
productive efficiency (the goal of ECP). For the distinction, see PB-T-1 at 47-50
(Panzar).
13
the end result6 – there is no plausible reason for the Commission to begin with an
inherently inefficient costing step and then (try to) “adjust away” its undesirable
consequences by tinkering with the final rates.
2.
Efficient component pricing as explained by witnesses Sidak
and Panzar.
Unlike the Postal Service witnesses, witnesses Sidak and Panzar both
present orthodox definitions and discussions of ECP. (Their evident divergence of
views as to how it should relate to worksharing discounts is dealt with below, fn. 13.)
They show that, in a situation where the firm is a monopolist as to one subservice but
not others, the proper access price for the monopoly subservice is derived by “pricing”
the other subservices at their average incremental cost (AIC). In postal practice, this
amounts to calculating the AIC of an operation which need not be performed by the
Postal Service if the mailer performs it, and setting the worksharing discount equal to
that cost. A mailer that can provide the subservice (sorting, transportation, etc.) at less
than the Postal Service’s AIC will do so, and will benefit by any excess of the discount
over its own cost. A mailer that cannot provide the subservice more cheaply than the
Postal Service, on the other hand, will buy it from the Postal Service. In either case,
the lower-cost provider will provide the subservice. If the competitive subservice is
“overpriced” (i.e., if the discount exceeds its AIC), then efficiency suffers because
mailers perform work the Postal Service could do more cheaply. If it is “underpriced,”
the reverse is true: efficient private-sector activity is discouraged in favor of provision
by the Postal Service at higher cost.
6
Ms. Robinson does not argue that it should be. (See USPS-RT-10 at 4).
14
This is the principle adopted by the Commission in Docket MC95-1 and
repeatedly endorsed thereafter.7
3.
Delinking violates ECP.
The CRA-based delinking proposal cannot meet the ECP test, because it
fails – indeed, appears to fail deliberately – to distinguish between (i) costs saved
(avoided) by the mailer’s performing a function the Postal Service would otherwise
have to perform for it, and (ii) reductions (or cost differences) in the cost of a function
the Postal Service must still perform, which result from certain differences in the
intrinsic characteristics of the mailpieces in question.8
This distinction is of central importance. If the Postal Service does not
“compete” in the production of a mail characteristic, because the related function is not
one the Postal Service performs, then no efficiency-related purpose is served by
offering a price reduction for the presence of that characteristic.9 The Postal Service
does not formulate addresses or apply them to mail. Address quality therefore is not a
case where mailer activity can – and should, if lower in cost – be substituted for Postal
Service activity. It is a mailer activity pure and simple. But both Mr. Taufique and Dr.
Panzar advocate counting address quality as a source of cost reductions for which
7
See, e.g., PRC Op. MC95-1, ¶¶ 3074 ff., 4261 ff. (especially ¶¶ 4268-4272); PRC Op.
R2000-1, ¶¶ 5535 ff., 5805.
8
In other words: if the mailer performs on a given mailpiece an operation the Postal
Service would otherwise perform, the relevant comparison is with the same mailpiece
absent the extra mailer activity, and the result is cost avoidance. If the mailer simply
prepares the piece in such a way that it is easier or cheaper for the Postal Service to
handle via the same operations it would perform were the piece not so prepared, the
relevant comparison is with the average mailpiece in the category, and the result –
measured by that standard – is cost reduction.
9
This has been well explained by witness Sidak. Response to GCA/NAA-T1-1;
Tr. 26/9139-40.
15
worksharing mailers should be given reduced rates – Mr. Taufique through his delinked
rates, and Dr. Panzar through an explicit discount.10 (The question of why mailers in
worksharing categories should be compensated for this characteristic while singlepiece mailers are not is discussed below, pp. 18 ff.)
Dr. Panzar’s presentation is helpful here, both because it is particularly
detailed in its explanation of ECP and because it illustrates so clearly why delinking
and ECP are irreconcilable. For that reason, we begin with it rather than the Postal
Service’s much less analytical presentation. In defining ECP, Dr. Panzar states that:
Since it [i.e., the Postal Service] always has the option to do the
work itself, offering a presort or other worksharing discount is
tantamount to a “make or buy” decision (i.e., determining whether
to perform an activity “in house” or pay outsiders to do it).
(PB-T-1 at 16). He then gives an algebraic example of efficient worksharing in which
all the attributable costs he considers are either delivery costs or “upstream”
transportation and sorting costs. He demonstrates (what no one would dispute) that if
the mailer can, profitably to itself, provide these upstream services in return for a
discount equal to the attributable cost avoided, the Postal Service should let it do so;
while if the mailer would pay out more than these same functions cost the Service, it
should let the Service perform them.
But because all the costs in Dr. Panzar’s example are costs of
subservices which either the Postal Service or the mailer can provide11, he ducks the
issue of how to reflect “intrinsic” differences in mailpieces that may affect – but cannot
10
USPS-T-32 at 12-14, and PB-T-1 at 14-16, respectively.
11
His algebraic example (PB-T-1 at 20 ff.) shows the “consolidator” hired by the mailer
performing all the “upstream” functions (represented in the example by CU).
16
avoid – the cost of Postal Service operations. And because he has designed his
example to ignore any cost-reducing attributes that nevertheless do not eliminate or
substitute for any Postal Service function, he fails to address the fallacy on which the
delinking proposal is based.
It follows that his definition of worksharing – i.e., of activity that would
supply grounds for a discount – is not supported by his discussion, or his practical
example, of ECP. In fact, it is incompatible with ECP. Dr. Panzar defines worksharing
as “any private sector activity which reduces the costs of the Postal Service.”12 This
formulation includes activities that are not performed by the private sector in lieu of the
Postal Service, and thus do not avoid any Postal Service cost. That Dr. Panzar intends
his definition to be this broad is clear from his examples: drop-shipping, non-stamp
evidencing of postage, and address quality. While a theoretical argument certainly
could be made that drop-ship discounts do not violate ECP, a discount for address
quality plainly does. Improving address quality may reduce the unit cost of certain
Postal Service operations, but the situation is not one where the Service “always has
the option to do the work itself.”13
12
PB-T-1 at 7 (italics in original). He explicitly presents it as an “expanded” definition
of worksharing.
13
PB-T-1 at 16. To clarify still further the intended reach of Dr. Panzar’s definition,
GCA asked him (and NAA witness Sidak) whether an activity would qualify as
worksharing if (1) it were the mailer’s lowest-cost option, (2) it were a function the
Postal Service could not perform, and (3) it were both of these. Dr. Panzar responded
that the activity would be worksharing regardless of whether it was the mailer’s low-cost
option or one the Service could not perform. GCA/PB-T-1; Tr. 26/9148. Mr. Sidak,
however, stated that, as to (1), the answer would be “yes” if the activity was one in
which the Postal Service and the mailer competed; and, as to (2) and (3), the absence
of competition between them would dictate that the activity not count as worksharing.
He explicitly based his answer on ECP principles. GCA/NAA-T1-1; Tr. 26/9139-40.
17
Unlike Dr. Panzar, the Postal Service does not explicitly start from an
expanded – and theoretically unacceptable – definition of “worksharing.” But the effect
of the delinking proposal is the same: lower prices for workshared mail are claimed to
be justified on the basis of both worksharing in the strict sense and cost characteristics
“not expressly associated with the worksharing activity.” (USPS-T-32 (Taufique) at 13).
Delinking, just like Dr. Panzar’s definition of “worksharing,” treats full cost differences
and cost avoidance as if they were the same thing. Both approaches jumble together
(i) activities which might avoid a postal operation but are not now associated with a
discount (e.g., entry at a postal facility) and (ii) “activities” which cannot possibly avoid
a postal operation, but may serve to reduce the cost of an operation that must be
performed whether or not the mailpiece in question falls in a worksharing rate category.
The result, in each case, would be inefficient prices. For this reason alone, the
Commission should reject delinking and adhere to its established system of
constructing worksharing rates.
C.
Delinking-Based Rates Would Be Unfair And Inequitable.
Section 3622(b)(1) of the Postal Reorganization Act,
39 U.S.C. § 3622(b)(1), requires the Commission to recommend a “fair and equitable
schedule” of rates.14 First-Class rates based on delinking would not meet this
standard.
It is also worth noting in this connection that an earlier proposal to create a
discount for mail paid by meter strip was rejected by the Commission, in part because
the alleged cost reductions were not true worksharing savings. PRC Op. R2000-1,
¶¶ 5218 ff.
14
Likewise, § 403(c) prohibits undue or unreasonable discrimination among mail users.
18
An elementary component of fairness is like treatment of like mail.
Delinking treats pieces having cost-reducing – but not cost-avoiding – intrinsic
characteristics (like the address quality discussed by Dr. Panzar) as eligible for
reduced rates if they meet certain other tests. Quantity requirements are an obvious
example. Delinking would secure a lower rate for mailing with a minimum of 500
pieces15 but would fail to recognize, through a lower rate, the same characteristics in a
mailing of 490 physically indistinguishable pieces. This is unfair on its face; and, as we
have shown above, the unfairness is not excused by any gain in efficiency.
Postal Service witness Taufique recognizes that some single-piece FirstClass Mail is automation-compatible:
The Single-Piece Letters category includes mail with a broad array
of characteristics: various sizes, different levels of automation
compatibility and machine readability. A significant percentage has
handwritten addresses. A substantial percentage consists of
courtesy reply envelopes used for bill payment, are of standard
size, and bear a machine-generated address, Facing Identification
Mark, and a barcode.
As noted in many previous dockets, furthermore, SinglePiece Letters include bulk metered mail (BMM) letters which are,
for the most part, of uniform size with machine-generated
addresses. . . .
(USPS-T-32 at 13) (italics added).
15
Domestic Mail Classification Schedule, § 221.22. OCA witness Thompson correctly
criticizes, as based on merely anecdotal evidence, Mr. Taufique’s assertion that pieces
not resembling the traditional Bulk Metered Mail benchmark are now being handled by
presort bureaus. (OCT-T-4 at 10-12). Her doubt that presort bureaus will solicit
business from “each and every home office, small business and household mailer” is
fully justified. Id. at 12). Consequently, mail fully compatible with automated
processing will continue to be excluded from worksharing rates because it is entered
singly or in small quantities.
19
In Docket No. MC95-1, the Commission pointed out that “splitting
subclasses between large and small users could ultimately violate the requirement of
fairness of § 3623(c)(1).”16 The same is true of a division within – or nominally within17
– a single subclass where a prime determinant of eligibility is how many pieces are
entered at a time. Similarly, with reference to the same courtesy reply mail mentioned
by Mr. Taufique, the Commission noted (at PRC Op. MC95-1, ¶ 5031) that placing it in
the “Retail” (i.e., single-piece) subclass
. . . belies any logical division of First-Class Mail into groupings
based on the mail’s susceptibility to automated processing. It also
raises grave concerns about the fairness and equity of the
proposed subclasses by denying entry into the Automation
subclass to less than bulk quantities of mail that could, in fact, be
equally compatible with automated processing.
The same problem arises here. Since worksharing is already fully
recognized18 in rates constructed on the existing model, the fairness of the rate
schedule would be diminished by according only workshared mail the benefit of costfavorable intrinsic features significantly present in single-piece mail as well.
D.
The Delinking Proposal Is Indistinguishable In Practice From Bifurcation
Of First-Class Letters Into Workshared And Single-Piece Subclasses.
The preceding subsection referred to the principles of fairness set out in
the Commission’s MC95-1 decision. It may be objected that that case is not in point,
16
PRC Op. MC95-1, ¶ 3046. The Commission was referring to its earlier ZIP+4
decision, Docket No. MC83-2.
17
For the insubstantiality of any practical distinction between this case and the
separate-subclasses proposal of Docket MC95-1, see subsection D, below.
18
In fact, more than fully recognized: the First-Class Automation letter rates established
by the settlement in Docket R2001-1 (and updated, across the board and without
regard to particularized costs, in Docket R2005-1) entailed passthroughs of as much as
120 percent of estimated cost avoidances. See PRC Docket R2001-1, USPS-T-29
(Robinson) at 20.
20
since the Postal Service here proposes not separate subclasses, but only separate
Presort and Single-Piece revenue requirements based on CRA costs, and envisions
equal unit contributions for these categories.
This is a distinction without a difference. The cost bases of the two
proposed categories are precisely what they would be in a separate-subclasses
proposal, and the Postal Service’s suggestion – there is nothing that could reasonably
be called a commitment – that unit contributions would be equal is wholly inadequate to
prevent delinking from producing a de facto two-subclass structure in First-Class
Letters.
This is how witness Taufique expresses it:
The Postal Service proposes that the rate design process
begin with establishment of separate revenue requirements for
Single-Piece letters and Presort Letters, with the goal of obtaining
similar unit contributions from Single-Piece Letters in the aggregate
and from Presort Letters in the aggregate. The objective of the
approach introduced here is to gradually achieve a rate design
paradigm in which both workshare and single-piece mail contribute
equally to institutional costs on a unit contribution basis. The goal
of similar unit contributions from these two mail categories 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-31 at 15-16 (italics added)).
If two First-Class letter “categories” (i) have separate attributable cost
bases extracted from the CRA, and (ii) may under (unspecified) rate design and rate
impact circumstances be assigned unequal unit contributions, then they are pretty
clearly separate subclasses in all but name. “Rate impact,” for example, is a non-cost
pricing factor applied at the subclass level, via § 3622(b)(4). If, as Mr. Taufique
21
suggests, it is to be a reason for departing from the equal-unit-contribution model, then
that model, pro tanto, comprises effective subclasses.
As a result, the objections to bifurcating First Class Letters into
workshared and single-piece subclasses, explained by the Commission in Docket
MC95-1, apply here as well. While GCA certainly does not endorse witness Thress’s
elasticity estimates in this case (see Section V of this Brief), a further condition for
subclass status is a difference in demand. Demand elasticities for first class single
piece and workshared mail in this case are so close that no differences in demand are
apparent.
Again, it may be objected (i) that in Docket MC95-1 the Commission opted
for a structure that produced equal unit contributions for workshared and single-piece
mail; and (ii) that because the Postal Service proposal in this case at least aims at such
unit contributions, it is consistent with that decision.
The MC95-1 decision does recognize that equal unit contributions result
from efficient rates.19 It does not follow, however, that efficient (or fair) rates result from
equal unit contributions. Quite apart from the evident instability of the Postal Service’s
equal-unit-contribution policy, as described in its own testimony, it is entirely possible
to construct a set of highly inefficient rates that happen to exhibit equal unit
contributions. We showed in subsection 1, above, why this would be the outcome of
delinking.
19
If the rates are efficient, the Postal Service is indifferent as to whether it or the mailer
provides the competitive subservice. (See, e.g., NAA-T-1 at 9). If the Postal Service is
indifferent as to who provides the competitive subservice, it follows that it receives the
same unit contribution in either case.
22
The emphasis the Commission placed on equal unit contributions in
Docket MC95-1 was in part necessary because the Postal Service proposal in that
case looked toward equalized percentage markups, as between workshared and
single-piece mail, as the goal of reclassification. The Commission had to explain that
disparate percentage markups are an unavoidable by-product of efficiently-constructed
worksharing rates, and are neither inefficient nor unfair.20 But merely layering equal
unit contributions on top of cost bases which virtually guarantee the inefficiency of the
resulting rates neither satisfies the principles the Commission laid down in Docket
MC95-1 nor justifies ignoring them.
V.
THE COMMISSION SHOULD LIMIT THE PROPOSED CHANGE TO FIRST
CLASS LETTER MAIL FIRST OUNCE RATES TO NO MORE THAN 41 CENTS
A.
The Relevant Market Realities Have Changed Fundamentally.
Since R2000-1, the last fully litigated rate proceeding, market realities
central to postal ratemaking have changed. First Class mail volumes have eroded
substantially and continuously. Standard mail volumes have increased substantially
and continuously. There has been explosive growth in broadband deployment and
broadband usage. Importantly, these facts are not in dispute.
Significantly, the Postal Service projects these trends to continue: First
Class mail volumes will continue to erode and Standard mail volumes will continue to
increase. The Postal Service’s estimate in this proceeding is that in 2008 Standard mail
will become the dominant mail class and in 2008 the volume of Standard mail will
exceed the volume of First Class mail. This relationship is expected to continue. The
uncontroverted record evidence is that the primary cause of the decline in First Class
20
PRC Op. MC95-1, ¶¶ 3057 ff., especially ¶¶ 3069-3079.
23
mail volumes is (and will continue to be) electronic diversion. The collective expert
opinion in this case is that First Class mail volumes lost to electronic diversion will not
subsequently return.
Standard mail has grown in volume and is expected to continue to grow
notwithstanding the growth in Internet, and in particular, broadband usage.
Despite these market realities, the Postal Service has proposed FCLM
first ounce rates that continue the disparity in institutional cost coverage between
FCLM and Standard mail. This disparity has grown considerably since 1995. In light of
the present and expected market realities, the Commission should recommend rates
that ameliorate this disparity so that the unfair and economically irrational trend toward
ever-greater institutional cost burdens on FCLM is at least moderated.
In addition, unrebutted forensic-quality market survey research shows that
the frequency and magnitude of FCLM rate increases are likely to be important
behavioral triggers as regards consumer decisions to shift FCLM volumes to electronic
substitutes. Stated simply, price matters.
The record evidence also shows that electronic diversion is positively
correlated with education and income. Whether described as “infra-marginal”
customers, or “hard-core” mailers, the Postal Service should not be looking to saddle
the poor and less educated with inequitable rate increases.
The Postal Service’s attempt to justify its proposed rate increase to 42
cents based upon its proffered elasticity evidence is without merit. As explained further
below, the Commission cannot properly rely upon the Postal Service’s FCLM elasticity
estimates. GCA believes its elasticity study correctly shows that the elasticity for FCLM
24
is higher than Standard mail. At the end of the day, however, the elasticity battle is
subsumed by considerations of the broader market realities. Regardless of whether the
Commission credits GCA’s or the Postal Service’s measures of First Class elasticity, or
instead, relies upon its own discretion and ratemaking expertise, principles of
ratemaking policy informed by the new market realities warrant an increase in FCLM
rates to (at most) 41 cents.
B.
In The Face Of Electronic Competition First Class Mail Volumes Have
Significantly Eroded While Standard Mail Volumes Have Increased
Dramatically.
First Class volumes for government fiscal year (“GFY”) 2002 were some
102 billion pieces. (USPS-T-7 at 9). Regular rate (“commercial”) Standard mail volumes
for that same time period were some 73 billion pieces. (Id.) By GFY 2005, First Class
volumes had fallen to some 98 billion pieces, while commercial Standard mail volumes
increased to almost 86 billion pieces. (Id.) The Postal Service’s own projections in this
case show a crossover point in 2008 where commercial Standard volumes will exceed
First Class volumes (whether measured by before rates or after rate volumes). (Id.) A
graphic representation of this phenomenon for the years 1997 through 2008 appears at
Figure 1 of NAA witness Sidak’s rebuttal testimony. (NAA-RT-1 at 9).21 Based upon
actual volumes, and the Postal Service’s estimates in this case, over the ten-year
period from 1998 to 2008, First Class volumes will have decreased by some 9 billion
pieces; commercial volumes will have increased by more than 36 billion pieces. (Id.).
21
Note that Figure 1 visually compresses the time period 2005 to 2008 as compared to
the scale for years 1997 through 2005.
25
Commercial standard mail is expected to be the dominant class of mail in the mail
stream during the period of time the proposed new rates are in effect.
The differences between the present market realities, and those
prevailing at the time of R2000-1, are night and day. First Class volumes have fallen
steadily over the four year period from 2001 to 2005. This is unprecedented in recent
historical times. Equally striking is the powerful growth in commercial Standard mail
volumes over the same time period. This upheaval in market realities leads to two
important observations. First, data concerning earlier time periods is of little if any
relevance in assessing for ratemaking purposes the present postal phenomenon.
Second, the electronic diversion that is responsible for the significant and continued
decline in First Class volumes, is not contributing to a decline in commercial standard
volumes. To the contrary, notwithstanding the deployment and expansion of the
Internet, and the upsurge in broadband usage in the United States, commercial
standard mail volumes are flourishing.
GCA witnesses Clifton and Martin, NAA witness Sidak, and Postal
Service witnesses Bernstein and Thress all agree that electronic diversion is the
principal factor accounting for the loss in First Class volumes over the recent four year
period. According to witness Thress, the Postal Service’s volume forecasting witness,
“[t]he Internet has had a very strong negative effect on First-Class single-piece letters
volume, explaining annual losses that have averaged more than 4.5 percent per year
over the past decade.” (USPS-T-7 at 60). GCA witness Clifton’s testimony focuses on
the ratemaking implications “of the growth of competing electronic substitutes for First
Class single piece letters since the last litigated rate case in R2000-1” (GCA-T-1 at 2).
26
The testimony of Postal Service witness Bernstein examines the issue of
electronic diversion at some length. (See generally, USPS-T-8). The Bernstein
testimony shows that not only does the decrease in First Class volume coincide with an
increase in Internet and broadband usage, but there is substantial evidence, anecdotal
and otherwise of just such diversion. As Bernstein shows, each of us in our own day-today experience can identify countless examples of electronic communications that
substitute for communications previously transmitted by mail. (USPS-T-8 at 17-18;
Table 4). Relying upon Household Diary Study data, Bernstein shows that substantial
volumes of household bill payments have been electronically diverted over the recent
four year time period. “Two-thirds of households paid at least one bill electronically in
2005, compared with barely more than one-third doing so in 2000.” (Id. at 54). “In 2005,
households that used electronic methods paid 36.5 percent of their bills electronically
and 58.2 percent by mail. In 2000, electronic bill paying households paid 26.6 percent
of their bills electronically and 66.1 percent by mail.” (Id. at 56).
[W]ithin the two-thirds of households that use some kind of
electronic payment, online bill payment is the fastest growing
alternative to the mail, now used by one in four households. These
households pay fewer than half their bills through the mail. Further
exacerbating the problem is the fact that households that pay bills
online pay more bills than other households. Therefore, the fastest
growing source of electronic diversion of bill payments is affecting
the households that would otherwise be paying the most bills by
mail.
(Id. at 63).
Household bill payments make up only a portion of transactions mail, i.e.,
bills, payments, and statements mailed between or among households, governments
and businesses. “‘Payments, bills, and statements make up about half of all First Class
27
mail.’” (GCA-T-2 at 11 [quoting Schmid, Greg, Two Scenarios Of Future Mail Volumes,
President’s Commission on the United States Postal Service, May 2003, at 8]). There is
evidence of electronic diversion affecting all forms of transaction mail. (Tr. 38/13085).
However, even accepting Postal Service witness Thress’s narrow definition of a “bill
payments delivery market,” and relying upon the Household Diary Study bill payment
data (which does not include, among other things, payments between and among
governments or businesses) the real world implications of electronic diversion with
respect to the all important category of First Class transactions mail should be clear.
The Postal Service has reached the point (or soon will) where it has no meaningful
market power in the U.S. payments system. (Compare GCA-T-1 at 16 with
Tr. 38/13105-06). When Postal Service witness Thress (an economist and
econometrician), was asked on cross-examination whether the Postal Service had price
setting power in the payments market, he could not offer an opinion. (Tr. 38/13106).
Indeed, Postal Service witness Thress could not even say whether the Postal Service
was competitive in the payments delivery market. (Tr. 38/13103-05). His inability to
address either of these points was apparently strongly influenced by the rapid, recent
and substantial erosion of First Class mail’s market share in the payments market due
to electronic diversion. (Id.)
Although there appears to be little or no dispute that electronic diversion
is responsible for the recent declines in First Class volumes, the Bernstein testimony
examines and rejects the possibility of other explanations, such as economic or
demographic factors. (USPS-T-8 at 9-13). As his testimony points out, the overall state
of the economy is not a credible explanation because “over the past four-year period,
28
real GDP has grown … yet First Class Mail volumes are lower than they were in 2001.”
(Id. at 9). To like effect, “Increases in population … should lead to increases in mail
volume.” (Id. at 12). Although the immigrant and native born populations of the United
States increased over the recent four year period, First Class mail volumes declined.
(Id. at 12).
The marked erosion of First Class volumes in the last four years, and the
substantial growth of commercial Standard volumes is an incontrovertible fact. It is
likewise not in dispute, that the Postal Service projects a crossover in these mail
volumes, with the continued growth of commercial Standard volumes and continued fall
of First Class volumes. There is likewise an agreement of opinion, and no credible
argument to the contrary, that First Class volumes have fallen, and will continue to fall
due to electronic diversion. “[A]lthough there is debate in this proceeding regarding the
extent to which First-Class Mail has been diverted toward electronic communication,
there is no debate over the fact that electronic diversion exists and has had, and will
continue to have, an effect on First-Class Mail that is significant in both an economic
and statistical sense.” (NAA-RT-1 at 18; Tr. 32/10838).
C.
There Is Good Reason To Believe That Electronic Competition Will
Continue To Erode First Class Mail Volumes Given The Rapid Growth In
Broadband Deployment And Users.
The recent decline in First Class mail volumes coincides with a period of
explosive growth in broadband Internet deployment and use. According to NAA witness
Sidak’s testimony, and based upon FCC data, “the adoption of residential broadband
Internet access, as measured by residential broadband lines into individual
households, has increased from only 1.79 million in December 1999 to 42.94 million by
29
December 2005, an increase of nearly 2,300 percent.” (NAA RT-1 at 11; Tr. 32/10831
[emphasis supplied]). As regards the number of users, “[t]he United States now has
102.5 million broadband users, whereas there were only 50 million users in 2003.” (Id.
at 11-12; Tr. 32/10831-32). Witness Sidak, a noted telecommunications expert,
contends that the tremendous, sudden and continuing growth of broadband deployment
and use is “the most important … trend[] in [the] diversion of First-Class mail.” (Id. at
17; Tr. 32/10837). “The substantial recent deployment, and continuing consumer
adoption, of broadband communications has increased the availability and variety of
electronic communication substitutes for First-Class Mail and these trends are likely to
continue.” (Id. at 25-26; Tr. 32/10845-46). The conclusion to be drawn is stark: “The
emergence of broadband communications as a substitute for First-class Mail means
that the Postal Service’s old business model can no longer work.” (Id. at 5; Tr.
32/10825).
Broadband Internet access is qualitatively different than slow speed
Internet access. According to Postal Service witness Thress, as “connection speeds
increase, the Internet becomes a more feasible substitute for more things.” (USPS-T-7
at 27). Indeed, Mr. Thress acknowledges “that the historical pattern of the adoption of
broadband internet access has mirrored electronic substitution out of certain types of
mail.” (Id.) And while Thress identifies “electronic bill presentment” as one form of
substitution that broadband has facilitated and apparently spawned (id. at 27), it is also
the case that the growth of on-line bill payment has coincided with the explosive growth
in broadband use. Relying on comscore Networks data, GCA market research expert
Martin’s testimony notes that “the number of online banking customers grew to nearly
30
40 million during the fourth quarter of 2005, a 27-percent increase over the previous
year. Usage of online bill payment services grew 36 percent during the same period.”
(GCA-T-2 at 10).
Acknowledging the difference between broadband and slower speed
Internet access, Postal Service witness Bernstein stated that broadband users are
likely more comfortable with the Internet (and thus Internet communications and
transactions) than non-broadband users.
I think that as more and more people get broadband, it represents
a different kind of internet access than dial-up access. It’s always
on, it’s faster, it says something about the user, that they would
actually pay more in many cases to get this. So to me it’s
representative of someone who’s more connected literally and
figuratively with the internet and more reachable that way, more
accepting of it, more comfortable with it.
(Tr. 6/1449-50, Aug. 9, 2006).
The survey research of GCA witness Martin likewise shows a marked
correlation between consumer access to broadband technology and on-line bill
payment. Witness Martin identified four separate classes of consumer bill payers: hard
core mailers (pay regular bills only by mail); major mailers (pay more than 50% of their
bills by mail); minor mailers (pay more than 50% of their bills electronically); and
electronic payers (pay regular bills only electronically). (GCA-T-2 at 19). The Martin
survey data showed that, “[n]ot surprisingly the hardcore segment has the least access
to broad band, or high speed internet technology (26.8%). In the critical major mailers
segment there are 40.3% with such access. Remember these are the bill payers using
the mail for more than 50%, but less than 100%, of their regular payments. In the minor
31
mailers segment there are 53.4% who have access to broadband/high speed
technology.” (Id. at 34).
The evidentiary picture is clear and irrefutable. First Class mail volumes
have eroded in the face of the new and emerging competitive substitute of broadband
Internet communication. There are compelling reasons to believe that this will continue.
Recent regulatory and business developments are likely to spur broad band adoption.
Unlike First Class mail rates, the price of broadband continues to fall. (NAA-RT-1 at 13;
Tr. 32/10833). In addition, “[i]t is widely believed that a major stimulus to investment
and competition in broadband adoption began in 2005, when the FCC deregulated DSL
service and the Supreme Court affirmed that cable modem service was an unregulated
information service.” (Id. at 17 [footnote omitted]; Tr. 32/10837).22 These are the very
factors that Postal Service witness Bernstein testified could lead to greater electronic
diversion. “What conditions would be consistent with greater electronic diversion? One
driver would be greater than projected Internet penetration or broadband adoption.
Competition between Internet providers could intensify, access rates could fall, and
adoption levels would in turn be greater than projected in the baseline forecast.”
(GCA/USPS-T8-8; Tr. 6/1399).
D.
The Commission Should Ameliorate Rather Than Continue To Increase
The Disparity In Institutional Cost Coverage Between First Class And
Standard Mail.
Notwithstanding the fact that First Class volumes are declining (and are
expected to continue to decline) in the face of broadband competition, while
22
“I’m concerned you could have another wave of volume losses in first class like you
did in the earlier part of this decade as you go through this S-shaped growth curve of
broadband.” (Tr. 29/9964) (GCA witness Clifton).
32
commercial Standard volumes continue to grow (and are expected to continue to
grow),23 the Postal Service has proposed a rate increase that looks to First Class mail
to assume an even greater share of responsibility for institutional cost coverage. This is
inequitable (if not ultimately self-defeating). It is worth quoting at length on this point
the testimony of GCA economic witness Clifton.
There is a longstanding inequity in institutional unit cost
contributions between First Class and Standard Mail that calls for
such a redistribution of unit cost contributions even in the absence
of the own price elasticity comparisons presented in Section VII.
above. In GCA/USPS-T31-1, USPS witness O’Hara was asked to
confirm that the gap between the unit cost contributions of First
Class single piece mail and Standard A Regular mail had grown
“from a 12.7 cent difference in R2000-1 to a 13.5 cent difference in
R2006-1,” a 0.8 cent increase in the gap. Witness O’Hara used
revised USPS data in his response, which showed that the gap in
unit cost contributions between First Class single piece and
Standard A Regular has grown even more between R2000-1 and
R2006-1, from 12.7 cents to 14.2 cents, a 1.5 cent increase in the
gap. On these grounds, I could justify cutting the single piece rate
proposed by the Postal Service in this case by two cents. My
proposal to raise rates on Standard A Regular mail to maintain
revenue neutrality for my proposed one cent reduction in the
USPS rate proposal for single piece letters from 42 to 41 cents is
therefore reasonable, and would still leave the relative unit cost
contributions ½ cent less favorable for single piece mail in this
case than it was in the last litigated case in R2000-1!
(GCA-T-1 at 60).
NAA witness Sidak reaches the same conclusion. “Someone must pay the
[Postal Service’s] institutional costs, and it is reasonable that the burden of institutional
costs borne by Standard Mail - the new largest class – should not decrease but, if
anything, should increase.” (NAA-RT-1 at 5; Tr. 32/10825).
23
And notwithstanding the significant decrease in the elasticity for Standard mail since
the last litigated rate case in 2000.
33
First-Class Mail has long borne a significant majority of the
institutional costs burden, and under the Postal Service’s proposal
it will continue to fund a majority of institutional costs. … Given the
recent decline in the use of First-Class Mail at the hands of
electronic communication, this is the wrong approach and the
Commission should consider lessening, not increasing, the
institutional cost burden placed on First-class Mail . . . .
(Id. at 19-20; Tr. 32/10839-40). Like GCA witness Clifton, NAA witness Sidak shows,
based upon the Postal Service’s discovery responses in this proceeding, that the
Postal Service’s approach would raise First Class institutional cost coverage. (Id.; see
also O’Hara response to VP/USPS-T31-7, 9; Tr. 17/5113, 5115-25.) By comparison,
“the cost coverage index of Standard mail has remained relatively unchanged at slightly
below the system average cost coverage since 1996. In particular, the cost coverage
index of Standard mail was 0.98 in 1996 and would be the same in 2008 under the
Postal Service’s proposal and costing methodologies.” (Id. at 21, Tr. 32/10841).
Relative to the system wide average, the cost coverage for First Class mail in 1996 was
1.09. (Id. at 20, Tr. 32/10840) Under the Postal Service’s proposal the cost coverage
index of First Class mail relative to the system wide average in 2008 is expected to be
1.21. (Id.) This is unfair and misguided. It is also a perverse pricing policy. As witness
Clifton observed, “What kind of business would be raising rates in its product areas
where they are having competitive problems?” (Tr. 29/9969).
E.
Price Matters: Market Research Shows That Future Postal Rate Increases
Play A Significant Trigger Role In Further Consumer Diversion To
Electronic Substitutes.
The testimony of GCA market research witness Martin examined the
impact of future FCLM postal rate increases and “looked at the specific price levels
where they would seriously consider switching from mailed bill payments to competitive
34
electronic alternatives.” (GCA-T-2 at 2). In addressing the behavioral impact of future
postal increases, the Martin testimony and underlying survey work, differs from the
approach of many other surveys that have looked at the issue of electronic diversion.
(Id.) Dr. Martin is a respected figure in the area of market research.24
To examine consumer attitudes Dr. Martin oversaw a national telephone
survey. The consumer survey was conducted and documented in accordance with
accepted practices and methodology applicable to forensic survey research. (Id. at 15
and ff.; see also Appendix A discussing Research Design and Execution).
Based upon the survey data, Dr. Martin concludes that postal rate
increases are a significant behavioral trigger. (Id. at 2). For example, discussing his
category of major mailers (consumers who use the mail to pay more than 50% of their
bills), Dr. Martin finds that “[p]ostal rates seem to be a future trigger for diverting from
mailed bill payments. The data suggest that, once motivated to switch from the mailed
payments by the price (or changes in the price) of postage, major mailers then use
other competitive attributes for the switching or diversion decision.” (Id. at 23). Stated
simply: price matters.
The survey questioned consumers specifically as to postal price level
where they would consider switching. Consumers were read a list of the following
24
He is the Isadore and Leon Winkelman Professor Emeritus of Retail Marketing and
Professor Emeritus of Marketing at the School of Business, University of Michigan. (Id.
at 1). He has authored some 65 articles in peer reviewed journals in the area of market
research and served as a market research consultant for numerous major enterprises,
including e.g., the United States Treasury. (Id. at Appendix B4-8). He has also “served
as a consultant to numerous law firms representing more than 80 different
organizations over the past 35 years offering consulting concerning marketing,
advertising and consumer behavior research issues in litigation matters.” (Id. at 7).
35
prices: 42¢; 45¢; 50¢; 60¢; 75¢; $1.00; and “don’t know.” (Id. at 29). The survey results
are dramatic and telling. “An indicator of immediate USPS vulnerability is the fact that
18.7% of all [consumer] mail bill payers will seriously consider switching to some form
of electronic payment at a stamp price of 42¢.” (Id. at 32) Based upon data in witness
Bernstein’s testimony, and “factoring in that many payers only use the mail for a portion
of their payments, nevertheless there are 2.15 billion annual pieces of bill paying mail
at serious risk of diversion given an increase to the requested 42¢ in the first class
postal rate.” (Id.) (Original emphasis). Not surprisingly, the survey data shows that the
amount of bill payments at risk increases at higher postal rates. “In the longer run, an
increase to 45¢ yields 3.2 billion annual pieces of bill paying mail at serious risk.” (Id.)
Again based upon the Bernstein-HDS data, “[t]his compares to the already reported 3
billion piece diversion stretched out over the period 1999-2005.” (Id.).
The Martin survey testimony is unrebutted. There is no substantial
evidence in the record to the contrary.25 On redirect, following his rebuttal crossexamination, Postal Service witness Thress made a speech to the effect that changes
in First Class rates of a few cents would likely have no impact on the rate of electronic
diversion (as regards workshared mail). (Tr. 38/13147). This testimony is rank
speculation based upon nothing. Although witness Thress’s econometric model is
capable of providing quantitative estimates of electronic diversion, witness Thress
grudgingly conceded on cross-examination that the amount of estimated diversion does
25
On cross examination, Dr. Martin explained that the mean survey response of 4.0 on
a 7 point Likert scale indicated that future postal rate increases were an important
consideration for respondents in deciding whether to switch to electronic means of
payment. (Tr. 21/7627; 7637). The fact that survey respondents rated other non-rate
considerations as also being important does not detract from this fact. Brain function,
like cardiac function is important, but this does not make liver function unimportant.
36
not change regardless of the proposed FCLM rate. In other words, given a proposed
FCLM first ounce rate of 42 cents, witness Thress’s model estimates that some
3,919,518,000 pieces of First Class mail will be diverted in 2008. (USPS-T-7 at 51).
However, given a proposed FCLM first ounce rate of say 75 cents, witness Thress’s
model would still estimate the exact same amount of 3,919,518,000 pieces of First
Class mail will be diverted in 2008.26 The Thress model estimates diversion based upon
the operation of certain (unsound) Internet variables that function independent from the
proposed FCLM rates. Quite simply, unlike Dr. Martin, witness Thress has not even
looked at the issue of whether and to what extent consumers are likely to be influenced
to move to electronic substitutes in the face of proposed postal rate increases, and in
particular, the Postal Service’s proposed increase in First Class rates to 42 cents.27
The implications of the Martin survey evidence are clear: the greater the
price increase, the greater the number of consumers who will likely be behaviorally
prompted to seriously consider electronic substitutes to paying bills by mail. In an effort
to deter and forestall future diversion the Commission and Postal Service should
attempt to minimize future FCLM rate increases.
26
“I do not include a specific variable which interacts the internet and price. That’s
true.” (Tr. 38/13095). Under the Thress model there would be additional volumes lost
due to own price elasticity, but that loss is independent of the amount of mail volume
lost by diversion.
27
Dr. Clifton’s examination of the payments market shows that postal payments mail
“may . . . be absolutely price elastic.” (Tr. 29/10053; see also GCA-T-1 at 17-23).
37
F.
The Commission Should Not Look To An Infra-Marginal Class Of Less
Educated, Older, And Less Wealthy Consumers To Bear Further
Revenue Increases.
A further inequity in the Postal Service’s proposal to raise FCLM rates to
42 cents, and increase FCLM’s relative share of institutional cost coverage, is that the
proposed increase will likely be borne by a class of First Class single piece mailers
disproportionately comprised of the less educated, the elderly, and the less wealthy.
GCA market research witness Martin examined the issue of which bill
payers are most vulnerable to electronic diversion. According to his survey data, “[t]he
highly vulnerable tend to be younger and better educated.” (GCA-T-2 at 32). In
reaching this conclusion, Dr. Martin assessed the demographics of survey respondents
who had the technological means to switch to electronic bill payment and had provided
survey answers indicating a substantial likelihood of switching.
The Bernstein testimony presents further evidence that the consumers at
risk to diversion tend to be wealthier and more educated. Bernstein looked at the issue
of Internet and non-Internet households. “[W]hen comparing mail volume received by
Internet households to volume received by non-Internet households, one is really
looking at households that have twice the median income, three times as likely to be
headed by a college graduate, and hold 60 percent more financial accounts.”
(USPS-T-8 at 46). Postal service witness Bernstein further examined the households
that paid bills electronically and those that did not. He found that the wealthier
households (i.e., households that paid more bills) were “the fastest growing source of
electronic diversion of bill payments.” (Id. at 63). (See also Tr. 38/13101-102). “This
finding is consistent with the Diary Study finding regarding mail received, showing that
38
households that receive the most mail are receiving less, driving the overall decline in
First-Class Mail volumes.” (Id.)
NAA witness Sidak examined the issue of broadband subscribers. Based
on research conducted by the Pew Internet and American Life Project, he looked at
marginal broadband subscribers and posited that the demographic characteristics of
marginal broadband subscribers are probably more akin to non-broadband users than
to infra-marginal broadband subscribers. (That hypothesis seems fully consistent with
the Bernstein findings). Sidak observed that “[r]elative to all broadband subscribers,
marginal broadband subscribers had less income and less education. Such a finding
implies a consumer’s willingness to pay for broadband is positively related to his
income and education.” (NAA-RT-1 at 15 n.9; Tr. 32/10835). Sidak explains that a
ratemaking pricing policy that seeks to increase cost coverage by recovering revenues
from the less wealthy and less educated is “regressive,” i.e., unfair. (Id.)
If demographic characteristics of inframarginal customers of Firstclass Mail resemble those of marginal (rather than inframarginal)
consumers of broadband access, then it would follow that
increasing the cost coverage of First-class Mail would be
analogous to imposing a regressive tax. Conversely, reducing that
cost coverage would be analogous to imposing a progressive tax.
(Id.).
In the face of emerging electronic competition and the attendant erosion
of First Class volumes, the Postal Service should not look to the remaining captive
mailers to shoulder ever increasing revenue requirements. “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.” (NAA-RT-1 at 15; Tr. 32/10835). Dr. Martin found that some 75% of the
39
hardcore mailers (those who pay all their regular bills by mail) do not have broadband
access. (GCA-T-2 at 34). To paraphrase witness Sidak, “a downward spiral by which
the remaining customers of First-Class Mail [are] forced to pay continually increasing
cost coverage [is] inconsistent with ‘the establishment and maintenance of a fair and
equitable schedule’ [as prescribed by] section 3622(b).” (NAA-RT-1 at 15;
Tr. 32/10835).
G.
The Postal Service’s First Class Elasticity Estimates Are Not Credible.
Pursuant to Commission Rule 54(i) the Postal Service is required to
provide as part of its case-in-chief “[a]n econometric demand study relating postal
volumes to their economic and noneconomic determinants.” 39 C.F.R. § 3001.54(i).
Postal Service witness Thress has presented such a study and concluded that the
own-price elasticities of First Class single piece mail and First Class workshared mail
are -0.184 and -0.130 respectively. (USPS-T-7 at 9). Witness Thress calculates the
own-price elasticity of Standard regular mail as -0.296. (Id.). As a practical matter these
elasticities are a significant factor in the Postal Service’s pricing proposal. Postal
Service witness O’Hara conceded that if the Thress elasticity figures were substantially
different that he would feel obliged to revisit his pricing recommendations.
(Tr. 17/5179-81).
Problems with the Thress First Class elasticity estimates warrant a
revision in the proposed pricing of FCLM. The Thress First Class elasticity estimates
are unreliable: they are the product of methodologically unsound work. In the language
of Daubert (the Supreme Court’s seminal decision concerning expert evidence) and its
progeny, witness Thress’s First Class estimates are the product of junk econometrics.
40
GCA witness Clifton conducted a simpler and sounder elasticity study and estimated
the own-price elasticity of First-Class single piece mail to be -0.456, (GCA-T-1 at 3),
but rising over time as Internet and broadband penetration deepen. (Id. at 54). There is
every reason to believe that Standard regular mail can support the requisite increase in
institutional cost recovery. Indeed, the Postal Service’s own estimates of standard
regular mail show a decrease in elasticity over the past several rate cases.28
1.
Daubert analysis.
In Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993), the
Supreme Court charged federal courts with the duty of ensuring that the reasoning or
methodology underlying proffered scientific testimony was scientifically valid. The
Supreme Court extended the reach of Daubert from scientific testimony to expert
testimony in general in Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137 (1999). The
court must “make certain that an expert, whether basing testimony upon professional
studies or personal experience, employs in the courtroom the same level of intellectual
rigor that characterizes the practice of an expert in the relevant field.” Id. at 152.
The Daubert Court identified four nonexclusive factors that courts should
use to determine the reliability of expert opinion: 1) whether the theory has or can be
tested; 2) whether the theory has been subjected to peer review; 3) the known or
expected rate of error; and 4) whether the theory and methodology employed is
generally accepted in the relevant expert community. Daubert, 509 U.S. at 592-94. The
28
This fact is consistent with the robust growth in Standard mail volumes as
advertisers have apparently found certain other forms of advertising to be less effective
than direct mail.
41
court’s focus is “on principles and methodology, not on the conclusions they generate.”
Id. at 595.
Rule 31(a) provides for the admission of “relevant and material evidence
which is not unduly repetitious or cumulative.” 39 C.F.R. § 3001.31(a). Because the
Commission does not employ the Federal Rules of Evidence, the Daubert tenets are
not strictly applicable. Nevertheless, the “spirit of Daubert,” is applicable. Pasha v.
Gonzalez, 433 F.3d 530, 535 (7th Cir. 2005). “‘Junk science’ has no more place in
administrative proceedings than in judicial ones.” Id. (quoting Niam v. Ashcroft, 354
F.3d 652, 660 (7th Cir. 2004). See also Lobsters, Inc. v. Evans, 346 F. Supp.2d 340,
345 (D. Mass. 2004) (court should look to Daubert in assessing evidence under NOAA
evidentiary standard).
Applying the “spirit of Daubert” to test the methodological soundness of
Postal Service witness Thress’s First Class econometric modeling is particularly
appropriate in light of Commission Rule 31(k)(iv)(e) governing the admissibility of
econometric studies. Every econometric study must be accompanied by:
A reference to a detailed description in a text, manual, or technical
journal for every econometric technique used in the estimation
process and the reasons for selecting the technique, or, in the
alternative, a description and analysis of the technique that is
sufficient for a technical evaluation.
39 C.F.R. § 3001.31(k)(iv)(e). Consistent with Daubert, Rule 31(k)(iv)(e) places a
premium on the use of peer reviewed methodologically accepted practices. Failing that
the proponent of the econometric study must present a testable analysis sufficient to
justify any departure from accepted peer reviewed practice.
42
2.
Witness Thress’s test for model selection has no basis in the
literature or the econometric community.
Witness Thress generated some 20 different First Class elasticity models
employing a wide variety of variables. In particular, Mr. Thress experimented with a
wide variety of variables for purposes of estimating the electronic diversion of First
Class single piece and workshared mail. In Library Reference USPS-LR-L-65 (at 128),
witness Thress explained that the standard he employed for choosing between and
among his 20 odd models was to select the model that exhibited the least mean
squared error. (Tr. 6/1331-32). On cross examination (concerning his direct testimony)
Mr. Thress was asked to identify what econometric authorities supported the use of a
mean squared error diagnostic to choose among competing econometric models. He
could not identify any: “I don’t have anybody off the top of my head.” (Tr. 6/1333)
On cross examination concerning his rebuttal testimony witness Thress
was asked whether it was true that his least mean squared error test had no formal
basis in econometric literature. Mr. Thress vigorously denied this contention and stated
that the 3rd edition of the econometric textbook Econometric Analysis by William
Greene identified the least mean squared error metric as a test for choosing between
and among econometric models. (Tr. 38/13112). Upon further questioning Mr. Thress
acknowledged that all or most of his 20 odd experimental models were “nonnested”
models. (Tr. 38/13114). Counsel for GCA then handed witness Thress a copy of the
3rd edition of William Greene’s Econometric Analysis and asked him to review the
specific section of the text devoted to choosing between nonnested econometric
models. Indeed, witness Thress acknowledged the section was titled “Choosing
Between Nonnested Models.” (Tr. 38/13116). (Tr. 38/13114-15; 13112). Witness
43
Thress acknowledged that the only tests Greene identified for choosing between and
among nonnested models (i.e., the type of First Class econometric estimations models
that Mr. Thress generated) were: the (1) J-Test; (2) the Cox test; and (3) a Bayesian
test approach. (Tr. 38/13115). Contrary to his prior statement, witness Thress
acknowledged that the 3rd edition of the Greene textbook did not list a least mean
squared error approach as an appropriate form of model selection. (Tr. 38/13116).
Witness Thress had ample time to review the Greene text on the stand, and was further
afforded the opportunity to consult with his attorney following cross-examination.
(Tr. 38/13143). He retook the stand for re-direct by counsel for the Postal Service but
did not identify any passage in the Greene textbook that supported his use of a least
mean squared error metric as a means of model selection. (Tr. 38/13143-13147).
As evidenced by Mr. Thress’ testimony as to what did in fact appear in the
Greene textbook, there is econometric literature addressing accepted formal
techniques of model selection, including tests specific to choosing between and among
nonnested models. The testimony of GCA’s expert econometric witness, Dr. Kelejian29
further confirms and underscores this point.
There are, of course, many formal procedures which relate to
model selection. Some of these are nicely described in the
econometric text by Greene on pages 152-160. One of these is the
Bayesian method which involves posterior odds. This method is
described in more detail in a classic text by Zellner in chapter 10.
This method is particularly appealing because it accounts for
different numbers of parameters in the various models which are
29
Dr. Kelejian teaches graduate level econometrics at the University of Maryland. He
is a respected figure in his field, has published many peer reviewed papers in the field,
and served as guest editor of a special edition of the Journal of Econometrics.
(GCA-T-5 at 2; CV at Appendix A).
44
being considered, as well as for other model complications in a
unified approach.
In concluding I note that I have raised serious concerns concerning
Thress’s model selection procedure, as well as his procedures for
estimating the parameters of those models. I would therefore
seriously doubt the validity of his estimated elasticities which are in
terms of his estimated models.
GCA-T-5 at 15 (footnotes omitted). As evidenced by Dr. Kelejian’s reference to the
Zellner text, scholarly literature addressing tests for econometric model selection dates
back some 35 years to at least 1971. (Id. at 15 n.7) (citing Arnold Zellner, An
Introduction to Bayesian Inference in Econometrics, New York (John Wiley and Sons,
Inc., 1971)). In response to a discovery request, Dr. Kelejian elaborated on the lack of
any formally accepted basis in the literature for a mean squared error text. Examining
the newer 5th edition of the Greene textbook, Dr. Kelejian noted that “I also do not see
a statement which indicates that the mean squared error has a firm foundation in theory
and can be applied to models which have various complications, let alone models like
Thress’s which contain improper procedures.” (USPS/GCA-1-12-7; Tr. 24/8758-59). Dr.
Kelejian further identified the J Test, a similar comparative test involving the nesting of
2 models, and the Bayesian posterior odds approach as formally accepted tests for
model selection and explained why these models have explanatory power. (Id.).
Witness Thress admitted that each of the three tests identified by Dr. Kelejian was a
formal model test. (Tr. 38/13106).
GCA economic witness Clifton, also criticized witness Thress’s model as
the product of any unrecognized form of model selection. “Because Mr. Thress did not
employ any formally accepted procedure in his choice among models and instead used
an intuitive approach, one cannot rely on the model he chose as being the best model
45
estimating the single piece demand equation, even if we accepted his non-linear
approach to modeling with several problematic transformations.” (GCA-T-1 at 34).
Witness Thress’s model selection process fails the Commission’s
evidentiary standard and the Supreme Court’s Daubert analysis. How Mr. Thress
decided to choose between and among his 20 odd experimental models was a critical
step in his First Class elasticity estimation process. A number of Mr. Thress’s elasticity
test models estimated significantly higher First Class mail single piece elasticities than
his selected model’s estimated value of -0.184. For example, Mr. Thress’s test model
number 6 estimated First Class single piece elasticity to be -0.319. (USPS-LR65 at
65-215). Perhaps more important, is that formal model testing might have shown that
none of Thress’s test models were any good. As Dr. Kelejian explained, “if Z is jointly
significant, I would stop my testing procedure and go back to the drawing board to
reconsider by theory.” (USPS-GCA-7(c); Tr. 24/8759).
Although Mr. Thress had numerous opportunities to do so, he was unable
to provide a “detailed description in a text, manual, or technical journal for [the]
econometric technique [of model selection] used in [his] estimation process.”
39 C.F.R. § 3001.31(k)(iv)(e). That should be the end of the inquiry. Witness Thress’s
failure to rely upon an accepted form of model selection, a subject addressed in the
peer reviewed literature for some 35 years is an unacceptable departure from accepted
econometric practice. His First Class elasticity estimates are properly described as the
product of “junk econometrics” and cannot be relied upon as substantial evidence in the
record. Although Thress made a conclusory statement that his least mean squared
error metric was an accepted form of model selection, he could not identify anyone else
46
who relied upon the practice other than his “colleagues” (apparently those he works
with) in “doing Postal work.” (Tr. 6/1333). Tellingly, for purposes of preparing his
rebuttal testimony, witness Thress conducted a (purported) J Test (and not “least mean
squared error” tests) to test aspects of Dr. Clifton’s approach. Mr. Thress did not
perform any J Tests between any of his own 20 odd experimental models: he simply
presumed that he had selected the correct one based on his least mean squared error
criteria. (Tr. 38/13109-11).
Witness Thress also failed to provide “a description and analysis of the
[least mean squared error] technique that is sufficient for a technical evaluation.”
39 C.F.R. § 3001.31(k)(iv)(e). Given the state of the literature this would entail an
explanation as why an unorthodox (and not formally recognized) least mean squared
error test was an acceptable alternative to the long-vetted and peer reviewed J Test,
Bayesian approach or Cox test. Mr. Thress nowhere even acknowledged his departure
from accepted practice, let alone attempted to formally justify it.
“The Daubert ‘requirement that the expert testify to scientific knowledge –
conclusions supported by good grounds for each step in the analysis – means that any
step that renders the analysis unreliable under the Daubert factors renders the expert’s
testimony inadmissible.’” Amorgianos v. Nat’l R.R. Passenger Corp., 303 F.3d 256, 267
(2d Cir. 2002) (original emphasis) (quoting In re Paoli R.R. Yard PCB Litig.,
35 F.3d 717, 745 (3d Cir. 1994)). To be clear, GCA is not moving to strike witness’s
Thress’s First Class elasticity estimates; they have been admitted in the record. But as
the product of junk econometrics they cannot constitute competent evidence. Judged
against the applicable literature, Mr. Thress’ technique of model selection was no better
47
than throwing darts. This fundamental error renders all of his First Class elasticity
estimates beyond the bounds of consideration – they are entitled to no weight.
3.
There is no theoretical or real world basis supporting Witness
Thress’ choice of variables for purposes of estimating electronic
diversion of First Class single piece or workshared mail.
Witness Thress employed two different variables in his elasticity model
for purposes of estimating the electronic diversion of First Class single piece and
workshared mail. For purposes of estimating workshared related diversion, Mr. Thress
used a measure of broadband subscribers lagged one-year. (USPS-T-7 at 68). For
purposes of estimating First Class single piece diversion, he used an internet
consumption variable interacted with a linear time trend. (USPS-T-7 at 60). Witness
Thress’ basis for selecting these particular variables was “econometric
experimentation.” (Tr. 6/1329). In other words, witness Thress chose variables that
worked best in producing a model with a low mean squared error. But as discussed
above, it is not an accepted approach to select models (and their attendant different
variables) based upon a desire to minimize mean squared error. What witness Thress
did not do was to look out in the real world and decide what variable and what model
structure made the most real world sense as an explanatory factor to measure and
predict electronic diversion and simply see what results such an empirically grounded
model would produce.
Given the explosion in broadband deployment and use, witness Thress’s
failure to use a broadband variable to model and estimate the diversion of First Class
single piece mail is inexplicable. GCA witness Clifton proffers a compelling critique.
It would have made sense on economic grounds for witness Thress
to include a broadband variable in his single piece demand
48
equation as part of his never-ending experimentation to model the
impact of Internet diversion correctly, but not in his worksharing
equation as he has done. Mr. Thress’ associate, Mr. Bernstein,
states that broadband deepening of Internet usage by households
is in fact one of the major reasons online banking and payment of
bills generally online has been increasing since 2000. (Tr., op cit.,
page 1449, line 6 through page 1451, line 1.) Mr. Thress tried but
rejected inclusion of the broadband variable in his single piece
demand equation evidently because it did not produce a lower
MSE. The reasons it did not produce a lower MSE may be many,
but the fact remains that on economic grounds, it should be
included in the single piece equation. What witness Thress did do
was to attempt to capture Internet deepening through a very
complicated set of changes to his Internet variable in the single
piece demand equation, changes which may have improved his
MSE relative to straightforward application of broadband data, but
which create other problems whenever a time trend dummy
variable capturing everything and nothing is re-introduced into a
demand equation.
(GCA-T-1 at 7).
Witness Clifton also critiques the suitability of Mr. Thress’s reliance upon
a broadband variable for purposes of estimating workshared diversion because
businesses turned to broadband some years ago. (Id. at 6). But even if a broadband
variable is appropriate for purposes of estimating workshared diversion, witness
Thress’s particular broadband variable is seriously flawed. NAA witness Sidak explains
that “the lagged variable trails reality by a year.” (NAA-RT-1 at 16; Tr. 32/10836). This
one year lag fails to capture the extraordinary growth in broadband use: in 2004 there
were 64.1 million broadband users as compared to 102.5 million broadband users in
2006. (Id. at 16-17; see also Figure 2 at 11; Tr. 32/10836-37 and 10831). Accordingly,
witness Sidak opines, “as a telecommunications expert, that it is unlikely that Witness
Thress has adequately measured the most important of the recent effects of, and
trends in, electronic diversion of First-class Mail.” (Id.; Tr. 32/10837) In doing so,
49
“witness Thress has likely underestimated the true effect of electronic diversion on
First-class Mail volumes.” (Id. at 16; Tr. 32/10836).
As discussed above, First Class mail volumes are eroding because of
emerging electronic, and in particular, broadband competition. An elasticity model that
attempts to estimate future First Class models but is not properly designed to account
for the effect of electronic diversion, such as witness Thress’s model, is fundamentally
flawed and unreliable.
4.
Other problems in witness Thress’s estimating of First Class
elasticities.
Witness Thress’ elasticity model includes data going back to 1983, i.e.,
before the advent of the Internet, let alone the deployment of broadband technology.
(GCA-T-1 at 35). His elasticity model also uses logarithmic (log) values. The data input
value for Mr. Thress’ Internet variables prior to the Internet is “0” (zero). (Id. at 51) It is
not possible to take the log of zero. (Id.) One recognized econometric approach for
handling this situation is to utilize a Box-Cox transformation. (Id. at 50-51).
Witness Thress purported to employ a Box-Cox transformation in
developing his First Class elasticity estimation model. However, in response to GCA
witness Kelejian’s criticism, witness Thress acknowledged that his purported Box-Cox
transformation was mathematically incorrect. (USPS-RT-2 at 53; Tr. 38/13062).
Witness Thress admits that his alleged Box-Cox transformation did not include two
terms it should have included: “[t]he two terms on the middle line of the above equation
… are not included in my First-Class single-piece letters equation in this case.”
(USPS-RT-2 at 54; Tr. 38/13063). On cross-examination, witness Thress
acknowledged that there was a mathematical relationship (which he himself pointed
50
out) which should hold in his model if it accords with a Box-Cox transformation. (Tr.
38/13134-35; see also USPS-RT-2 at 54 n.14). Witness Thress admitted that he could
have tested whether this relationship did obtain in his model mathematically, but he
failed to do so. (Tr. 38/13134-35) Instead, witness Thress ran a different test
purporting to show that the missing terms were irrelevant.30 Significantly, however,
under questioning, witness Thress conceded that he should not have described his
data manipulation as a Box-Cox transformation. (Tr. 38/13134).
Witness Thress’ admission that he incorrectly described his one of a kind
data manipulation as a Box-Cox transformation, combined with his misstatement that
his least mean squared error test is a formally recognized method of model selection,
30
On cross-examination, Mr. Thress claimed that the regression output appearing at
pages 78 – 79 of LR-L-191 shows that his selected model did not need to include the
variables TREND and T02Q4, despite the fact that a mathematically correct Box-Cox
transformation would have required the inclusion of these variables. (Tr. 38/13135; see
also USPS-RT2 at 54-55). However, this test regression appears to be fatally flawed. If
one is running multiple regressions, in this case a single piece demand equation, the
right hand side includes several independent variables for which there is data. One
solves the equation using least squares techniques from which there emerges a unique
coefficient representing the influence of each variable. Before numerical values are
estimated for these coefficients, we could, for example, just call them b0, b1, b2, b3, b4,
etc. from page 78 of LR-L-191. Unfortunately, Mr. Thress has in essence estimated the
same coefficient, b1, for two different variables where he should have estimated b1 for
the first variable and b3 for the other variable. He has repeated this error for another
coefficient, in essence estimating b2 for two variables instead of estimating b2 for one
variable and b4 for the other variable. His equation is not correctly estimated as a
result of these errors. The actual coefficients that are repeated twice in his equation are
C01SP(26) and C01SP(25). What is happening when his computer software runs such
a mis-specified equation as that on pages 78-79 of LR-L-191? The software program
may estimate the problem coefficients the first time it encounters them but throw them
out the second time. In this case, two of Mr. Thress’s variables are simply ignored. Or,
it may not estimate the coefficients at all, in which case four of Mr. Thress’ variables are
ignored. It is hard to interpret what the statistical output shown on page 79 for those
two coefficients even represents. Is it an “average” from commingling the two variables
to which each coefficient is attached? What meaning can be attached to such a
coefficient?
51
evidence a core problem of credibility or competence. As prescribed by Commission
Rule 31(k)(iv)(e), a witness presenting an econometric model must provide an accurate
description of the econometric techniques used in constructing the model: “a
description and analysis of the technique that is sufficient for a technical evaluation.”
39 C.F.R. § 3001.31(k)(iv)(e). Witness Thress failed to provide an accurate description
of his work in two fundamental regards: incorrectly attributing his novel least mean
squared error selection test to the Greene textbook and mischaracterizing his novel log
transformation technique as a formal Box-Cox transformation.
As may be painfully obvious to the Commission after sitting through the
examination of econometric witnesses Kelejian and Thress, the vetting and testing of
an econometric model, particularly a complex model like witness Thress’s First Class
estimation model is a significant undertaking. If you cannot trust the modeler’s
description of his modeling procedures, as is the case here, then the task is
unmanageable. There is no reason to credit witness Thress’s statements concerning
his modeling and testing procedures and this renders the results of his First Class
elasticity model inherently unreliable.
A further problem with the Thress First Class elasticity model is his
reliance upon an inappropriate data set. The Postal Service faces a problem of recent
onset: the emergence of electronic competition, particularly in the form of broadband
Internet communication. As GCA witness Clifton points out, witness Thress’s model
drowns out the important near term data with data going back over 20 years.
In my view, witness Thress’ single piece demand equation does not
properly capture the “way the world is today” because the
phenomena of Internet diversion and electronic payments
substitutes for single piece mail are relatively recent – first
52
modeled by Mr. Thress in the 2000-2001 period, whereas his
model data goes all the way back to 1983, well over a decade
before the impact of these competing substitutes began to be felt in
single piece mail volumes.
The influence of emerging competing substitutes in recent years is
largely “washed out” of the USPS-sponsored research because the
recent data is simply homogenized by being added to all prior time
series data in Mr. Thress’ model, which includes prolonged periods
in the 1970s and 1980s where there were few if any competing
substitutes for FCLM.
(GCA-T-1 at 35-6, footnote omitted). Witness Thress admitted that “the further back in
time you go you get data that may be less applicable to the way world is today, so that
there’s a trade off.” (Tr. 6/1338).
In addition to methodological modeling problems, descriptive problems,
and data problems, the results of Mr. Thress’ First Class elasticity estimation model are
also problematic. Witness Thress’s elasticity estimate for workshared mail in R2005-1
was 253% higher than his elasticity for workshared mail in this proceeding. (GCA-T-1
at 5). As GCA witness Clifton pointed out, it cannot be that the elasticity of workshared
mail has changed that dramatically over the course of three extra quarters of data.
“Both estimates cannot be correct.” (Id.) Witness Thress acknowledged the problem. “‘I
do not believe that First-Class workshared mail has become increasingly inelastic
between the R2005-1 and R2006-1 rate cases.’” (Id.) (quoting Thress response to
ABA-NAPM/USPS-T7-2). Witness Thress attributes the discrepancy to flawed modeling
in R2005-1 and that the new one-year lagged broadband variable he used to model
workshared diversion in this proceeding produces the correct result of -0.130 (as
opposed to -0.329 in R2005-1). (Id. at 5-6; see also Tr. 6/1325-26). However, as
explained above, witness Thress’s one-year lagged broadband variable is no good, it is
53
essentially stale data. There is simply no reason to believe that either -0.130 or -0.329
is a correct estimate of the elasticity of First Class workshared mail. Thress’s model
and the results it produces are deeply flawed.
5.
Witness Clifton’s elasticity estimates are a better measure of
market realities.
GCA witness Clifton developed a linear demand model to estimate First
Class single piece and workshared elasticities. A guiding principle in the construction
of his model was to explain much with little. (GCA-T-1 at 44). Dr. Clifton used a
straightforward demand curve from textbook economics. This linear specification is the
most simple functional form for exploring changes in elasticity along the demand curve
(a focus of his analytical effort). Unlike witness Thress, witness Clifton’s linear model
does not employ a quasi-Box-Cox transformation (or anything of the kind). (Id.) Dr.
Clifton avoided what Dr. Kelejian labeled as the symmetry problem that arises in
witness Thress’s model because Clifton (unlike Thress) adopted an endogenously
determined value and sign to worksharing discount value. (Id.) The results of
Dr. Clifton’s model show that single piece First Class mail has become significantly
more elastic in the face of Internet competition, Standard A regular mail has not.
Over the entire time period, the elasticity for single piece mail I
estimated using the R2006-1 rate case materials with my VES
approach was -0.456 … compared to witness Thress’ estimate of 0.184. My estimate for Standard A regular mail using, the VES
approach was -0.254 … compared to witness Thress estimate of 0.296.
(GCA-T-1 at 54).
54
Dr. Clifton’s study shows that in the more recent time period First Class
single piece mail is becoming more elastic and Standard A regular mail is becoming
more inelastic.
[I]n the face of Internet diversion, while the elasticity of Standard A
Regular mail is becoming more inelastic, the elasticity of single
piece mail is becoming less inelastic. At the margin for the R20051 rate case, the difference appears to be -0.765 in 2005 for First
Class single piece compared to -0.190 for Standard A Regular.
Using the approach from R2006-1 with more data available for
2005, the difference appears to be -0. 565 for single piece
compared to -0.173 for Standard A Regular.
(Id.)
On rebuttal, witness Thress launched a host of challenges at witness
Clifton’s work while attempting to rehabilitate his own work. A marked feature of the
Thress rebuttal testimony are arguments that Thress himself does not credit. For
example, Thress suggested that it is plausible for First Class single piece mail to
increase in elasticity in the face of broadband competition because in other industries
price sensitive customers had departed for new competitive substitutes. (USPS-RT-2
at 13, Tr. 38/13022). But on cross-examination, Mr. Thress refused to say that the First
Class customers shifting to electronic substitutes were price sensitive. (Tr. 38/1309899; 13102). In other words, in order for Mr. Thress’ comparison to have any purpose,
he has to admit the very thing he adamantly refuses to acknowledge: that postal prices
are material to the scope of electronic diversion. As discussed above, the Martin
survey evidence clearly establishes that postal prices are central to consumer
decisionmaking as regards electronic diversion.
In response to Dr. Clifton’s criticism that it would make good business
sense for the Postal Service to attempt to compete on price given its eroding market
55
share, Mr. Thress resorts to the non sequitur that the Postal Service is competing on
non-price terms, e.g., by selling stamps over the Internet. Witness Thress then makes
the wholly conclusory and unsupported assertion that “[t]his is precisely the sort of
behavior that can help the Postal Service to remain competitive within the payments
delivery market.” (USPS-RT-2 at 12, Tr. 38/13021). But on cross-examination, witness
Thress would not and could not say that he believed the Postal Service was
competitive within the payments market. (Tr. 38/13104-05).
In Table 3 of his testimony, GCA witness Clifton presents a chart of
descriptive statistics estimating arc elasticities for single piece mail and electronics
payments. From the (markedly strong and consistent) results on cross-elasticities
Clifton infers “that the own-price elasticity for postal payments mail is likely high.”
(GCA-T-1 at 19).31 In response, witness Thress presented what he labeled “a more
correct analytical approach to Dr. Clifton’s Table 3.” (USPS-RT-2 at 26; Tr. 38/13035).
Remarkably, however, Mr. Thress qualified his regression analysis by stating that “[t]his
is not an analysis that I would endorse as being particularly useful in a rate case
setting, and, in fact, this is almost certainly not an analysis that I would be inclined to
perform. As such, my inclination would be to place extremely little confidence in these
results.” (USPS-RT-2 at 26-7; Tr. 38/13035). Recognize that Thress is raising these
qualifications against his own proffered analysis. On cross-examination, and in light of
his own caveats, witness Thress could not meaningfully explain why he had included in
his rebuttal testimony (his singularly mischaracterized) “more correct analytical
approach.” (Tr. 38/13139-40).
31
This conclusion is grounded in accepted theory. (GCA-T-1 at 18 n.11).
56
Mr. Thress also faults Dr. Clifton for using a linear demand specification
as opposed to a log-log specification. Mr. Thress performed a J-Test to compare
Dr. Clifton’s linear demand model with a log-log specification. Mr. Thress contends that
the J test finds in favor of the log-log specification. It is important to note that witness
Thress did not run a J-Test to compare his model against Dr. Clifton’s. Indeed, Mr.
Thress did not even run a J-Test that compares the form of his model against the form
of Dr. Clifton’s model. Mr. Thress’ model is not a log-log specification model. It is a log
model transformed by a bastardized Box-Cox transformation.
Witness Thress makes clear that he did not test Dr. Clifton’s linear model
against a log-log model as modified by a Box-Cox transformation: “[f]or this initial test,
a pure log-log specification is considered as the alternate hypotheses.” (USPS-RT-2 at
48 n.10; Tr. 38/13057).32 Thus, the J-test that Mr. Thress ran compares Dr. Clifton’s
model to something which does not exist in the record, i.e., a log-log specification
model. Stated otherwise, the record contains a Thress apple and a Clifton orange. Mr.
Thress claims to have compared the Clifton orange with a banana. Mr. Thress’s J-test
in no way shows that witness Thress’ model is superior to Dr. Clifton’s.
32
It is not clear how Mr. Thress could have done this. The Clifton model contains an
Internet variable, and for the time period before the Internet (when the input data is
zero) it is not possible to take the log of the Internet variable. In his own model,
Mr. Thress went to some length to justify his use of the Box Cox transformation on
these very grounds that one cannot take the log of a variable whose value is zero.
Now, however, in his J-test of Dr. Clifton’s linear specification, we find that Box Cox
was not necessary after all. He is able to take the log of Dr. Clifton’s Internet variable
by, in essence, entering as a quasi-dummy variable a value of -89.3 for the period
1983:1 – 1988:1,as found on page 5 of LR-L-191, and the log of the actual positive
values of the Internet for the period thereafter.
57
Most remarkably, even if the Commission were to credit witness Thress’s
J-test of linear versus a log – log version of Dr. Clifton’s model, it would need to accept
the own price elasticity of demand associated with the log – log version. This elasticity
is -0.53, the sum of the coefficients c(10) and c(11) on page 50 of LR-L-191. It is higher
than Dr. Clifton’s linear model elasticity of -0.456 and nearly three times higher than
witness Thress’s own model!
In an attempt to defend his use of a quasi-Box Cox transformation,
Mr. Thress introduces a Box-Cox transformation of the Internet variable into
Dr. Clifton’s equation and tests it via nonlinear estimation. (USPS-RT-2 at 49; Tr.
38/13058). This is, quite simply, bizarre. The very reason for employing a Box-Cox
transformation is to address problems that do not exist in a linear model such as Dr.
Clifton’s. Although Mr. Thress contends that Dr. Clifton should have included a BoxCox transformation in his model, this is methodologically pointless and thus unsound.33
Mr. Thress seems to not understand that introducing unnecessary complexities into a
model defeats the sound principle of explaining much with little. The introduction of
these modeling complexities has serious repercussions. According to Dr. Clifton, “[i]n
the context of the R2005-1 data and model structures, what factors explain the
difference between Thress’ results and mine? It turns out that 74% of the difference is
explained by Thress’ use of his so-called ‘Box-Cox’ transformation . . . .” (GCA-T-1 at
50). Mr. Thress does not dispute this finding.
33
Also problematic is Mr. Thress’s application of a Box-Cox transformation to only one
variable, rather than re-writing the equation to transform all of the variables in order to
test each of them empirically.
58
The most important observation that can made with respect to witness
Thress’ rebuttal testimony is what he failed to do. Witness Thress failed to run any
formally recognized model selection tests on his own 20 odd experimental models.
(Tr. 38/13109).
Mr. Thress also criticizes Dr. Clifton’s use of a positive coefficient with
respect to workshared mail. On cross-examination, witness Thress admitted that if you
run his model without imposing a restriction on the sign of the coefficient, the coefficient
turns out to be positive, i.e., turns out to be exactly the same sign that is in Dr. Clifton’s
model. (Tr. 38/13082). Dr. Clifton explained that he used an endogenous coefficient;
i.e., that he utilized the positive coefficient value that the model itself produced.
(GCA-T-1 at 44; 50).
Again, should the Commission consider witness Thress’s J-test of the
Clifton linear specification, it is noteworthy that Thress’s own finding in the log – log
specification of Dr. Clifton’s model is that the worksharing discount variable is positive
and highly significant. In Dr. Clifton’s linear model it was positive but marginally
significant for data from the R2006-1. (Thress’s t statistic for coefficient c(9) using log –
log is 3.29 on page 50 of LR-L-191, compared to a t statistic of 1.67 using a linear
specification on page 49 of LR-L-191.) This finding from Mr. Thress’s own J-test
strengthens Dr. Clifton’s argument that Mr. Thress’s a priori presumption is
fundamentally wrong that the worksharing discount must have a negative sign. The
findings from Mr. Thress’s own J-test suggest to the contrary that his own model’s
symmetry condition between worksharing and single piece reflecting a pre-Internet view
59
of conversion and reversion between the two mailstreams is fundamentally at odds with
contemporary market realities.
By comparison, Mr. Thress imposes a negative value on the workshared
coefficient in his model. (Tr. 38/13078). This adds further complexity in the modeling
process and, as Dr. Kelejian explains, creates serious symmetry problems. (GCA-T-5 at
6-12). Witness Thress attempts to defend this imposed negative sign on the grounds
that if First Class single piece prices rise, and First Class single piece volume falls,
then some of that single piece volume “may convert to First-Class workshared letters,
as indicated by my estimated in my negative own-price and negative discount
elasticities.” (USPS-RT-2 at 41; Tr. 38/13050). However, statements submitted by the
Major Mailers Association in this case (i.e., by those in the worksharing business) are
at odds with Mr. Thress’ understanding of the current market realities. There is very
little, if any, single piece mail that is likely to convert to worksharing. (Tr. 38/13086).
Moreover, the little that does will likely soon divert to electronic substitutes. (Tr.
38/13090-91).
The Internet offers another avenue of choice for mailers in responding to
any reduction in the worksharing discount. Mailers need not revert back to sending
such mail by single piece. They can in many instances simply opt out of the mail
altogether in favor of electronic substitutes. As set forth in the comments of MMA,
“unlike decades past, simplistic assumptions about the potential for workshared mail to
revert to single piece simply are not relevant to informed rate making in the electronic
age.” (Tr. 38/13090 (quoting comments)) In the view of MMA, any reversion to single
60
piece mail in today’s competitive marketplace will likely be very temporary, and that
single piece mail will likely soon divert to electronic substitutes. (Tr. 38/13090-91).
Mr. Thress cannot properly defend his imposed restriction on the sign of
the worksharing coefficient based upon an outdated “conversion/reversion” assumption
that does not comport with present circumstances. His imposition of a negative sign on
the workshared coefficient is further proof that he has not correctly modeled the impact
of Internet competition and electronic payments substitutes for mail in his demand
equation for First Class single piece mail. What Mr. Thress completely
mischaracterizes in his rebuttal testimony as being predictions from Dr. Clifton’s model
that are “at odds with economic theory” are in fact perfectly consistent with economic
theory once it is recognized that worksharing mailers can forgo using the mail by using
electronic substitutes. Dr. Clifton’s finding that the sign attached to the worksharing
discount variable is not in fact negative (as Mr. Thress pre-supposed it was), is
consistent with the statements of worksharing mailers that increases in the worksharing
discount do not lead to an automatic and mechanistic conversion of single piece into
workshared mail, and decreases in the worksharing discount do not lead to an
automatic and mechanistic reversion away from worksharing to single piece mail.
6.
The Commission does not have to resolve the battle of
elasticities in order to adopt GCA’s position.
As GCA has shown, there is a fundamental hole in the Postal Service’s
requisite case-in-chief. The Postal Service was required to provide an econometric
estimate of First Class elasticities and after rate volume demands, but has not in any
meaningful sense done so. As a general rule, an agency is bound by its own rules.
Here, however, the Commission has other resources. It can: (a) credit GCA witness
61
Clifton’s elasticity estimates; or (b) decide the issue of FCLM rates based upon the
record evidence as a whole recognizing the deficiencies of the Postal Service’s case
and discounting it accordingly.
“Ratemaking is . . . much less a science than an art.” Village of
Bensenville v. FAA, 376 F.3d 1114, 1122 (D.C. Cir. 2004) (quoting Nat’l Ass’n of Secs.
Dealers, Inc. v. SEC, 801 F.2d 1415, 1419 (D.C. Cir. 1986)). The Commission does
not, and should not resort to mechanistic approaches such as Ramsey pricing to
allocate proposed rate increases. There is good reason to ameliorate, and not
exacerbate the disparity in institutional cost coverage between First Class single piece
and standard regular mail. There is good reason to minimize any price increases in a
product that is losing volume (such as FCLM). There are good reasons to shift greater
institutional costs onto standard mail, the soon to-be dominant class of mail. As an
exercise in ratemaking discretion, the Commission, should it choose to do so, can
properly adopt GCA’s First Class rate proposal without deciding the battle of the
elasticities.
H.
The Commission Cannot Properly Look At The Instant Proposed Rate
Increase In A Vacuum
1.
Recently enacted postal reform legislation reduces uncertainty
and alleviates the need for a 1% contingency.
GCA did not present (or sponsor) testimony dealing with the Postal
Service’s overall revenue requirement or the level of the total $4 billion proposed rate
increase. A number of intervenors jointly sponsored testimony of Lawrence G. Buc
(DMA-T-1) that contends that the Postal Service has substantially overstated the Test
Year revenue requirement on several grounds, including the ground that the Postal
62
Service has overstated its contingency needs and contingency request (1%, or $767
million). Recently enacted and signed postal reform legislation (the “Postal
Accountability and Enhancement Act,” HR 6407, 109th Cong. (2006)) will, inter alia,
result in greater cost certainty by eliminating the obligation of the Postal Service for
pension liabilities for certain retired military employees who become postal workers.
The Act includes as well a requirement to use funds currently in escrow to pre-fund
retiree health benefits.
The fact that there is a new Act is a significant, substantial known change
that should be recognized as a known change by the Commission in its recommended
decision.
The record does not contain evidence that quantifies the effect of the
Postal Accountability and Enhancement Act on the Postal Service’s Test Year revenue
requirement, but does contain evidence raising issues with respect to the overall
revenue requirement. Irrespective of how the Commission evaluates that evidence, at
a minimum it should eliminate the 1% contingency allowance because there is greater
certainty in the revenue requirement which alleviates the need for a separate
contingency.34
2.
The Commission should not look at the instant proposed rate
increase in a vacuum.
There is a general consensus that the First Class volumes that have left
the postal system due to electronic diversion will not return. For example, GCA market
research witness Martin reports that “there is a high degree of satisfaction with the
34
Cf. PRC Op. R-2005-1, ¶¶ 5013-5015, where the Commission recognized that where
the amount of an obligation is known, there is no need to provide for a contingency.
63
electronic methods used by these 100% electronic payers offering little potential for
reacquisition by USPS.” (GCA-T-2 at 30). Erosion to electronic competition will
continue for the foreseeable future.
The long-term forecast is sobering, even measured by estimates prepared
for the Postal Service by witness Thress and his associates. Witness Thress and his
colleagues prepared a report entitled “2005 Mail Volume Forecast Scenarios.”
(GCA/USPS-T8-8; Tr. 6/1397). The Postal Service referenced information from that
report in its September 2005 Strategic Transformation Plan. (Id.) The Forecast report
presents three different scenarios: a baseline case, a pessimistic case and an
optimistic case. The baseline case presents First Class volume estimates not very
different from the Postal Service’s estimates in this case. The Forecast report estimates
some 90 billion pieces (89,547,000,000) of First Class mail in 2008 as compared to the
Postal Service’s after rate estimate in this case of some 91 billion pieces
(91,291,000,000) of First Class mail in 2008. (GCA/USPS-T8-8; Tr. 6/1397; USPS-T-7
at 9). As shown at Table A1, under the baseline scenario the Forecast Report
estimates a decrease in First Class mail volume to 80 billion pieces by 2014.
The pessimistic case, which the Forecast report describes as the next
likeliest case, paints a particularly discouraging picture of First Class volumes. A
central driver in the pessimistic case is an increase in electronic diversion as compared
to the baseline case. (GCA/USPS-T8-8; Tr. 6/1397). As discussed above, there are
very good reasons to believe that the Postal Service’s estimates of electronic diversion
in this proceeding are significantly understated. The pessimistic case predicts that by
2008 First Class mail volumes will have fallen to 84 billion pieces, and by 2014 to 70
64
billion pieces. (GCA/USPS-T8-8, Table A.2; Tr. 6/1435). First Class mail volumes in
2002 were some 102 million pieces. (Id.) Thus, under the baseline scenario First Class
mail volumes are projected to fall by some 22 billion pieces over the twelve year period
from 2002 to 2014, and under the pessimistic scenario by some 32 billion pieces over
the same time frame.
Although the postal reform legislation will likely provide welcome relief by
reducing some otherwise necessary outlays, the question remains as to where will the
revenues come from? In this case, the Postal Service has provided a business as usual
approach and, indeed, seeks to increase the institutional cost coverage responsibility
of First Class mail. In addition to being unfair, this makes no sense.
Change must begin somewhere, even if that change is a penny (or two).
In this proceeding, as set forth in the Martin survey material, ordinary consumers have
voiced their opinion that pennies matter to them. It is time for the Commission to take
the lead and begin a course correction. The Commission should adopt GCA’s proposal
and recommend a proposed increase in FCLM first ounce rates to no more than
41 cents.
VI.
THE SHAPE-BASED RATES PROPOSAL NEEDS IMPROVEMENT
A.
Introduction.
Arguing that today shape rather than weight is the main influence on
First-Class Letter costs, the Postal Service has proposed separate first-ounce rates for
Letters, Flats, and Parcels. In general, this proposal makes excellent sense and
deserves favorable consideration by the Commission. But it is not problem-free. If
65
recommended, it should be recommended only with an important – though simple, and
far from costly – adjustment.
This adjustment would create a sub-category of “low-aspect ratio Letters”
– single-piece Letters with an aspect ratio less than 1.3:1 – and attach to it the existing
13-cent non-machinable surcharge, rather than charging such pieces the 62-cent rate
for a single-piece Flat. The latter course, which the Service proposes, would levy an
unjustified (and, as will appear below, unjustifiable) 54-percent increase in the extra
cost of sending a square greeting card. Given the importance of square designs to
greeting card users, and the likelihood that taxing the square shape even more heavily
than it is taxed today will depress mail volume, the Commission should make this
simple adjustment to the shaped-based rate proposal.
The maxim35 that “everything should be made as simple as possible, but
no simpler” is apposite here. The Postal Service’s design violates it. Its proposal
maladroitly combines the existing definition of machinability36 – which controls liability
for the non-machinable surcharge – with a new rate distinction by basic shape (Flat,
Letter, or Parcel). This misalliance produces a structure in which a Letter-shaped piece
that fails any one of the current machinability tests is treated, for rate purposes, as it
were a Flat. Such a piece would still be processed and delivered as a Letter. But the
result of the Postal Service’s proposal is a 54 percent increase37 in the (effective)
35
Ascribed to Albert Einstein.
36
Domestic Mail Manual (“DMM”) § 101.1.1 (dimensional maxima for Letter shape) and
1.2 (machinability standards for Letters). This section of our Brief focuses solely on
single-piece Letters and Flats.
37
The current surcharge is $0.13; the proposed Letter-vs.-Flat first-ounce rate
differential is $0.62 - $0.42 = $0.20; $0.20 ÷ $0.13 = 1.5385.
66
surcharge, even though, as regards processing and delivery38, nothing has changed.
By confusing the concept of “non-machinable” in the sense demanded by the DMM
rules for machinable Letters with the much more fundamental difference between basic
shapes, the Postal Service’s proposed design incorporates a damaging
oversimplification.
The issue is important for consumers because the square or near-square
greeting card is a distinctive, highly valued medium of personal communication.
Penalizing it through an unjustified 54 percent increase in the extra cost of sending
what is still a Letter, though a non-standard one, can be expected to discourage its use,
and in the process to deprive the Postal Service of high-contribution First-Class
volume. This effect is not merely unfair to consumers; it also demands the
Commission’s attention under 39 U.S.C. § 3622(b)(4) (“effect of rate increases upon
the general public”). And, because, as GCA witness Liss demonstrates, the square
shape has special value for greeting card users and recipients, § 3622(b)(8)
(“educational, cultural, scientific, and informational value to the recipient”) also
counsels adoption of GCA’s proposed modification.
Consequently, the Commission should recommend the tripartite shapebased rate structure with a modification whereby single-piece low-aspect-ratio Letters
weighing one ounce or less continue to pay the existing non-machinable surcharge,
rather than being charged as Flats.39
38
Processing and delivery are the cost centers on which the Service bases its entire
proposed differential between Letters and Flats. (USPS-T-32 (Taufique) at 18, 22-23).
39
Details of this modification appear at pp. 9-10 and Exh. C of GCA-T-3 (Morrissey).
67
B.
Charging Low-Aspect-Ratio Letters As Though They Were Flats Is Bad
Mail Classification Practice.
1.
Introduction.
We start from the plain fact, which the Postal Service’s proposal tends to
obscure, that a low-aspect-ratio Letter is still a Letter. It is processed on letter-mail
sorting machinery40 and is delivered as a Letter, not a Flat.41 The cost of low-aspectratio Letters is combined with that for all other Letters to produce the Cost and
Revenue Analysis statistics which the Postal Service uses to underpin its proposed
rates (the significance of this fact is discussed in subsection B.2 below). Witness
Taufique has confirmed that for classification purposes a low-aspect-ratio letter would
not be a Flat.42
Since the Postal Service proposes not merely to maintain the Letter/Flat
shape distinction but also to increase its importance by embodying it in a large rate
differential, one would expect that some extraordinary necessity would have to be
alleged, in order to justify charging a Letter the rate for a Flat. But none has been. The
explanation given is that a low-aspect-ratio Letter is to pay a 48-percent-higher firstounce rate simply because it is non-machinable. (Id.) But the result is to create an
unreasonable and unfair classification.
40
See responses to GCA/USPS-T13-2 (Smith), Tr. 14/4230, and GCA/USPS-T42-1, -2
(McCrery), Tr. 11/2797-99. This fact is demonstrated not only by the testimony of these
witnesses, but by the fact that the Merrifield test relied on by Postal Service witness
Laws was run on letter-mail cancelling equipment. (Tr. 21/7692).
41
See responses to GCA/USPS-T30-1, -2 (Kelley), Tr. 12/3343-45.
42
Response to GCA/USPS-T32-1, Tr. 16/4812.
68
2.
Cost justification impossible.
Postal Service witness Taufique introduces his shape-based rates by
observing that “[p]reliminary cost studies suggest that lighter weight flats and parcels
may not be covering their costs.”43 But in fact the proposed treatment of low-aspectratio Letters cannot be defended on cost-recovery grounds.
First, any extra costs these pieces impose are double-counted in the
Postal Service proposal. Low-aspect-ratio Letters (along with other pieces that fail one
or more of the machinability tests of DMM § 101) are present in the mailstream today,
and the costs of processing and delivering them are combined with those of fully
machinable letters to produce the CRA statistics used by Postal Service witnesses
Smith and Kelley to estimate costs by shape. Witness Taufique has used these
aggregated costs to calculate a differential of 37.1 cents between Letters and Flats44
(which he then reduces by nearly half to avoid rate shock). But since Mr. Taufique
develops his proposed Letter rate with Letter costs that already include any extra costs
43
USPS-T-32 at 17. It is significant that Mr. Taufique here refers to “flats and parcels”
but not to non-machinable Letters. The cost studies referred to likewise concerned
only Flats and Parcels in the ordinary sense of those terms. See his response to
GCA/USPS-T32-2(b) (Tr. 16/4813) (“. . . if the aspect ratio was less than 1.3 or greater
than 2.5, but all other letter dimension requirements were met, the mail piece was
classified as a letter, not a flat.”).
44
USPS-T-32 at 23 (particularly the table).
69
imposed by low-aspect-ratio letters, he cannot justify charging these letters the Flats
rate in order to “recover” those same costs again.45
Even if this were not true, it still could not be argued that the cost of a lowaspect-ratio Letter even approaches that of a Flat. For example, the MODS data used
by witness McCrery show that while Flats make up about 2.7 percent of all collection
mail, they account for 60 percent of the hand cancellations.46 In GCA’s experiment,
very few of the low-aspect-ratio test pieces lacked evidence of automated sorting.47
The delivery cost of a Letter is $0.077 and of a Flat $0.143.48 Since for low-aspectratio Letters the existing surcharge and the proposed assimilation to Flats rest on the
notion of non-machinability – not on claims of additional delivery cost – it can scarcely
be argued that the delivery cost of a low-aspect-ratio Letter is anywhere close to that of
a Flat.49
45
This question is quite separate from the issue seemingly raised by Postal Service
rebuttal witness Laws (USPS-RT-16; Tr. 34/13428), who apparently believed that GCA
wishes to abolish any distinction between machinable and non-machinable Letters.
GCA, of course, is proposing in this case to maintain the existing non-machinable
surcharge for low-aspect-ratio Letters; see GCA-T-3 at 9-10 and Exh. C (Morrissey).
On cross-examination, Mr. Laws acknowledged this fact. (Tr. 39/13442-43).
46
Responses to GCA/USPS-T42-7 and -8, and to VP/USPS-T42-7(a); Tr. 11/2804-07,
3106.
47
All 504 square pieces were delivered back to GCA, and 480, or 95.24 percent,
showed evidence of at least some automated processing. Postal Service rebuttal
witness Laws did not controvert this fact; insofar as his testimony rests on experimental
findings, it covers only the cancellation stage. See generally USPS-RT-16, and the
Merrifield test report printed at Tr. 21/7681 ff. (particularly p. 7692, defining the scope
of the aspect ratio test).
48
USPS-T-30, Table 1 (Kelley), used by Mr. Taufique at USPS-T-32 at 23.
49
It is true that at one point Mr. Taufique does not limit his reasoning to Flats and
Parcels in the ordinary sense. At p. 22 of USPS-T-32, he states that
70
The main function of a rate category is to reflect substantial differences in
cost between one type of mail and another – here, Flats and Letters. Thus it is
irrational to include in one rate category mail, such as low-aspect-ratio Letters, whose
costs are much closer to – and for reporting purposes are included in – those of a type
forming a different rate category.
3.
An appropriate incentive?
In its presentation, the Postal Service seems to emphasize the alleged
incentive effect of shape-based rates: mailers who can convert their pieces from Flat to
Letter shape will now have a powerful motive (20 cents per piece) to do so.
(USPS-T-32 at 22-23 (Taufique)). GCA does not doubt that this is true, and in many
cases desirable. The question remains whether this objective justifies placing low-
. . . It is expected that, in those instances where the costs of conversion to
lower cost shapes are relatively low, and the converted mail pieces are
consistent with the needs of our customers, the Postal Service will benefit
from a conversion to lower cost items. . . . [italics added]
We suppose it might be argued that “lower cost shapes” was meant to include
standard-aspect-ratio (in place of low-aspect-ratio) Letters. The context proves
otherwise. This part of his discussion is titled “Flat and Parcel Shaped Pieces.” His
explanation of the cost differentials (Id. at 23) explicitly relies on costs provided by
Messrs. Smith and Kelley, who have confirmed that their studies do not count lowaspect-ratio Letters as Flats. Finally, after observing that he is proposing first-ounce
rates “for flat-shaped pieces and . . . for parcel-shaped pieces” (Id. at 18), Mr. Taufique
states:
Pieces that do not meet the letter machinability criteria (defined by
length, height, width, thickness, rigidity, variation in thickness, or aspect
ratio) become eligible for the next higher rate element, i.e., the first ounce
rate for flat shaped pieces. . . .
Id. at 19. All this demonstrates that Mr. Taufique does not assume that low-aspect-ratio
Letters cost the same as Flats; his proposal is to attach the Flats rate to a low-aspectratio Letter more or less regardless of its cost characteristics.
71
aspect-ratio Letters under the Flats rate schedule without cost justification. On this
point, some comments are in order.
First: a rate design change that motivates mailers to change from one
shape to another may not work in the same way when the question is between a
standard and a non-standard Letter-shape configuration. The Postal Service’s
approach seems to assume that the same number of pieces will be mailed as before
the change, but that – in many cases – they will become Letters rather than Flats.
Witness Taufique made no inquiry into the potential reaction of particular subtypes of
Letters before preparing his rate design. (Tr. 18/4965-69). The Postal Service’s
approach neglects the possibility that fewer (Letter-shape) pieces will be mailed if the
low-aspect-ratio configuration is penalized even more than it is today. GCA witness
Liss’s unrebutted testimony shows how this can happen.
Second: an “incentive” that, so far as rates are concerned, treats a singlepiece Letter like a Flat does not necessarily address the customer behaviors the Postal
Service wants to change.
Let us first assume that the Postal Service’s main object is to replace
Flats with Letters. The prospective sender of a low-aspect-ratio greeting card – unlike,
say, the law firm considering whether to start mailing documents folded, in Letter-size
envelopes – has never contemplated sending a Flat. The only effect of the “incentive,”
therefore, is to reduce the greeting card user’s options; and, as witness Liss has
shown, doing that may also reduce the Postal Service’s First-Class volumes and
revenues.
72
If, on the other hand, the Postal Service is equally interested in replacing
low-aspect-ratio, and other non-standard, Letter configurations with those qualifying as
machinable under DMM § 101.1.2, it has failed to show (or, apparently, even to inquire)
whether charging them the rate for Flats (rather than the existing 13-cent surcharge)
will have that effect. Mr. Taufique acknowledged that he had not studied the effect of
his proposal on particular subtypes of Letters. (Tr. 18/4965-67). No witness has
attempted to rebut Ms. Liss’s testimony that increasing the effective surcharge could
reduce mail usage.
In short, the Postal Service has presented a useful concept, but reduced it
to practice in so simplistic a way as to produce damaging side effects. Attaching to a
Letter which is “non-machinable” because it fails one of the DMM § 101.1.2 tests the
rate for a Flat, which is “non-machinable” because it is, dimensionally, not a Letter at
all, and can be handled only on Flats processing equipment, is a seriously deficient
approach to restructuring First-Class Mail.
C.
Why Proper Treatment Of Low-Aspect-Ratio Letters Is Important.
1.
Low-aspect-ratio letters are important to mailers.
It may be objected that (i) even granting what we have said above, much
the same might be said on behalf of various other Letters that fail one of the DMM
73
§ 101.1.2 machinability criteria; (ii) if an “exception”50 were made for low-aspect-ratio
letters, one would have to be made for all such mail configurations (e.g., “high” aspectratio letters, or letters more than 0.25 in. thick); and (iii) doing so would dilute the
shape-based rate structure. To argue in this way would be to ignore the record made
by GCA, and essentially uncontroverted by others.
GCA witness Liss drew on more than 20 years of experience as a
greeting card designer, publisher, and bulk mailer51 to demonstrate at least two
propositions of basic importance for this case:
·
that the square shape is highly favored by consumers and mailers
seeking a distinctive and aesthetically attractive design, and
·
that surcharging this shape adversely affects would-be purchasers
– and mailers – and could easily lead to reductions in mailed
volume.
The square shape is appealing in large part because it can “break through the clutter
and get noticed.” (Id. at 6). That its appeal is not a transitory fad is demonstrated not
only through witness Liss’s explanation of its historic importance for designers, artists,
and architects in a variety of cultures (Id. at 8-12), but also by her own experience –
recent experience, in particular – as a designer and industry awards judge:
50
Witness Laws’ term (Tr. 39/13442). It should not be necessary to point out that
calling GCA’s proposal an “exception” puts the cart a good many yards before the
horse. The Postal Service’s shape-based rates have not been recommended by the
Commission or adopted by the Governors. They are not an existing set of rules to
which an exception is sought, but a proposal before the Commission, on which the
Postal Service bears the burden of proof. 39 U.S.C. § 3624(a), incorporating 5 U.S.C.
§ 556(d). And witness Morrissey is in fact proposing no change in the existing
treatment of low-aspect-ratio letters.
51
GCA-T-4 at iii-iv, 1-3, 12-15.
74
. . . It has been my observation that over the last five years, the
number and variety of square shaped envelopes has increased, as
has the demand for clients for square formats. I believe that one
important reason is that the novelty of the square, non-rectangular
mailing piece attracts attention, interest, and sets it apart. It stands
out from the rest, and therefore has desirability in the
marketplace. . . .
(Id. at 7).
* * *
Based on the popularity of the square shape with my clients,
it is my belief that if a square shaped card were available to more
people without a surcharge to mail, I would expect that more of
these cards would be noticed for their novelty and inherent appeal,
picked up by consumers, purchased, and mailed. An important
factor in this sales cycle and addition to the mail stream, however,
is that the surcharge on this shape not be excessive, or even
better, if there were no surcharge to discourage designers and
consumers from purchasing and mailing them.
(Id. at 12-13).
Witness Liss has surveyed winners and finalists in GCA’s LOUIE
Awards52 competition, from the award’s inception in 1988 to the present, and has found
that the number of square pieces (or pieces fitting a square envelope) has nearly
tripled. (GCA-T-4 at 15, and Exhs. 1 and 2).
In short, Ms. Liss has demonstrated that the low-aspect-ratio Letter is
uniquely important to users of First-Class Letter mail. Her presentation is unrebutted
by any party. No comparable showing has been made on behalf of any other “nonmachinable” Letter-shape configuration. Hence it cannot be argued either (i) that
52
These awards are the only competition honoring creativity in the greeting card and
stationery industry. Entries are judged on criteria including imagination, impact,
artistry, harmony, sendability, and value.
75
GCA’s proposal addresses an insignificant question, or (ii) that, if approved, it would
entail a series of parallel exceptions for other Letter-shape pieces.
2.
Price matters.
Here again, Ms. Liss’s experience can help the Commission improve the
Postal Service’s proposed rate structure. As she explains the market situation:
. . . It has been my experience that when informed of the
surcharge, some clients decide to reduce the number of pieces
they mail, or to select a more standard mailing size, which is not
their first choice. They want to break through the clutter, but are
unwilling to pay the excess premium in order to do so. I am
concerned that by adding an additional surcharge through the new
rate treatment of square shaped envelopes as if they were flats, the
problem will be amplified. . . .
(Id. at 13 (italics added). This assessment of potential market effects should be of
concern to the Postal Service as well as the Commission. First-Class Letter volumes
are already declining, and doing so for reasons having much more to do with price than
the Postal Service appears to recognize.53 The Commission, at least, should not allow
an oversimplified rate design based on a confused appreciation of underlying costs to
compound the problem.54
53
See generally section V of the present Brief.
54
Some Postal Service interrogatories to Ms. Liss suggest that the argument may be
made (on brief, since no rebuttal to her testimony has been filed) that her predictions
should be discounted because not based on a quantitative study. Since the Postal
Service has made no study – quantitative or otherwise – of this problem, such a
criticism would be misplaced. Even if the Postal Service’s proposed rate design were
not inherently defective, regardless of its precise volume effect, the Commission would
be justified in acting in reliance on Ms. Liss’s practical industry experience as against
an objection that it is not an econometric analysis. As matters in fact stand, she has
supplied additional cogent reasons to make an improvement that should be made
anyway as a matter of sound mail classification.
76
D.
GCA’s Case For Its Proposal Has Not Been Rebutted.
As pointed out above, GCA witness Liss’s testimony in support of the
GCA proposal stands unrebutted. The Postal Service has filed testimony critical of
witness Morrissey’s presentation, but despite its label it is not rebuttal in any effective
sense. Postal Service rebuttal witness Laws has set out to rebut a “proposal” which
GCA has not made. That he presents an interpretation of how low-aspect-ratio pieces
behave in (some aspects of) the automation process, and criticizes GCA’s field test for
not being something it was not intended to be, and for practical reasons could not have
been, cannot change this fact.
We emphasize once more that GCA is not proposing, in this case, to do
away with the present non-machinable surcharge on low-aspect-ratio Letters or to
change the DMM definition of low aspect ratio. But Mr. Laws appears, in his prefiled
testimony, to believe that GCA does seek to change the existing standards of
machinability55 and (presumably) to relieve low-aspect-ratio Letters of the surcharge.
Mr. Morrissey, however, made it clear that GCA proposes nothing of the sort. GCA
urges only that a single-piece subcategory of Low-Aspect-Ratio Letters be established,
and that a one-ounce piece falling within it pay the existing 13-cent surcharge in
55
USPS-RT-16 at 4, 5; Tr. 39/13431, 13432. Mr. Laws agreed on oral cross
examination that GCA’s proposal did incorporate a surcharge and that GCA’s proposed
aspect ratio definition was the same as the present rule. (Tr. 39/13442-43).
77
addition to ordinary first-ounce single-piece postage.56 A piece with an aspect ratio
lower than 1 to 1.3 would pay this surcharge, just as it would today.
GCA does not, however, merely ask the Commission to recognize Mr.
Laws’s testimony addresses a non-issue, and discount it accordingly. Some comments
on his treatment of the GCA field experiment are also in order.
Mr. Laws dismisses GCA’s field test of 500 square and 500 “standard
shape” (rectangular) pieces, but in doing so he misstates its scope and purpose. “The
fatal flaw in this [GCA] experiment,” he says,
is that witness Morrissey attempts to determine how a process
works, and whether or not there are additional costs, by only
observing the input and the output of the process. . . . [57]
56
See GCA-T-3 at 9-10, and Exh. C (for classification language). If GCA’s rate
proposal described in section V of this brief is adopted, the resulting rate would be
$0.41 + $0.13 = $0.54. If it is not, the rate would (presumably) be $0.42 + $0.13 =
$0.55.
57
USPS-RT-16 at 9; Tr. 39/13,436. Subsequently, he paraphrases an interrogatory
response in which Mr. Morrissey partially confirmed various propositions about
numbers the GCA test did not reveal. Mr. Laws’s paraphrase, however, is not entirely
accurate. For example, he states that Mr. Morrissey admitted that the GCA test does
not reveal “the number of pieces I.D. tagged on an AFCS or DBCS but rejected within
subsequent automated mail processing steps due to low aspect ratio and the
propensity of pieces to tip over[.]” What Mr. Morrissey actually said was:
Confirmed in part. The GCA Square Card Test data does not reveal how
many (if any) square or rectangular test letter pieces were ID tagged on
an AFCS or DBCS but rejected within subsequent automated mail
processing steps. I cannot confirm that such rejections, if any occurred,
were due (in whole or in part) to the low aspect ratio and/or any propensity
of the piece to tip over, or instead to human error or mechanical failure.
The Square Card Test was not designed to study this issue.
78
Significantly, Mr. Laws does not cite any page in Mr. Morrissey’s testimony where such
claims appear – for good reason, since they are nowhere made. Mr. Morrissey stated
plainly that the object of the test was to examine “the degree to which square
envelopes can not be processed on Postal Service automated cancellation and sorting
equipment.” (GCA-T-3 at 2). In other words, his project was to see how many (what
proportion) of the square cards58 did not bear evidence of machine cancellation or
sorting. Particularly as regards sortation, this proportion turned out to be quite small.
Again, Mr. Laws compares the GCA test to “claiming that one can determine the route
someone took to travel from point A to point B and how much it cost to make that trip by
only looking at the outside of the vehicle to see if there is mud on it.” (USPS-RT-16 at
9-10; Tr. 39/13436-37). This is picturesquely put, but irrelevant. GCA has never
claimed to know, experimentally or otherwise, the precise cost of handling a square
envelope. (It may perhaps be presumed that the Postal Service does not know it
either; at all events, Mr. Laws has not suggested a figure.) GCA’s claim is that there is
nothing in this record to support – but a good deal to refute – the notion of charging a
low-aspect-ratio Letter the same rate as a Flat, rather than the present surcharge.
Response to USPS/GCA-T3-5 (italics added); Tr. 21/7780. Mr. Laws’s paraphrase (like
the original question) ignores the distinction between rejections due to some
characteristic peculiar to square envelopes and rejections having some other cause.
That not every processing failure is caused by a problematic aspect ratio is strongly
suggested by GCA’s results from its Boulder, CO, test deck. Mr. Trumble, the Boulder
test participant, mailed 63 square and 63 rectangular pieces, and all126 pieces arrived
at Mr. Morrissey’s office bearing no cancellations whatever. (GCA-T-3 at 5).
58
And, of course, of the rectangular (“standard”) cards included as a control group.
79
There is also a practical point. The correspondence concerning the
Merrifield test relied on by Mr. Laws59 shows that GCA and the Postal Service were
discussing in May 200560 how the test would be carried out. The final report is dated in
April 2006 – some eleven months later. A test revealing all the details cited by
Mr. Laws as missing from GCA’s experiment would obviously have to have been an
observed test, like that conducted at Merrifield. It would have involved both cancelling
and sorting equipment. To achieve the same geographic scope as the GCA
experiment, it would have had to be conducted in several facilities around the country.
It is reasonable to conclude that such a test would have consumed at least as much
time as the eleven-month Merrifield experiment.
But GCA became aware of the proposal to raise the effective surcharge
on low-aspect-ratio letters when the Postal Service filed its Request (May 3, 2006); and
GCA’s responsive evidence was due on September 6, 2006. Thus GCA had
approximately four months to plan, conduct, and interpret its experiment, and embody it
in testimony. The Postal Service, in short, appears to suggest that the GCA test should
be discounted because it did not embody procedures impossible to carry out within the
statutory time schedule for a rate case.
E.
Conclusion.
The Postal Service has made a sufficient case for the general proposition
that shape rather than weight is the main driver of First-Class processing and delivery
59
It is worth noting that the Merrifield test, insofar as it addressed the aspect ratio
question, studied only the cancellation operation, and so provides no data at all
regarding automated sorting. See Tr. 21/7692.
60
See Tr. 21/7675-76 (Valerie B. Cooper to Nicholas F. Barranca).
80
costs. If its tripartite, shape-based rate design had been more analytically arrived at,
GCA might well be supporting it without significant qualification.
In fact, the proposed rate design irrationally treats low-aspect-ratio Letters
as Flats, for rate purposes, and does so without cost justification or any reasoned
expectation that the resulting 54 percent increase in the effective surcharge will
produce (what may or may not be) the hoped-for change in mailer behavior. Hence it is
necessary to adjust the proposal. And it is necessary not just because the proposal as
it stands is not rational mail classification, but also because it would cause
unreasonable harm to consumers and other greeting card users in defiance of the
requirements of §§ 3622(b)(1), (4), and (8), and with no promise of an offsetting benefit
to the Postal Service.
GCA’s proposed remedy is simple. The existing non-standard surcharge
should be continued for low-aspect-ratio Letters. It would be added to whatever firstounce single-piece Letter rate is adopted in this proceeding. The DMM definition of low
aspect ratio would remain the same. The dollars involved are practically insignificant.61
The Commission should adopt GCA’s proposed adjustment of the shape-based rate
design and recommend it as so adjusted.
61
Witness Morrissey pointed out that only 28 percent of single-piece mail subject to the
surcharge is Letter-shape (the rest being Flats and a few Parcels). According to
Mr. Taufique (response to GCA/USPS-T32-4(a), Tr. 16/4815), 124.3 million Letters –
as against 303.3 million Flats and 12.0 million Parcels – paid the 13-cent surcharge for
any reason in FY 2005. This was about 0.3 percent of the total single-piece, nonpermit-imprint volume of Letters, Flats, and Parcels of 41,384.0 million (see USPS-LR87). Letters liable to the surcharge because of low aspect ratio are of course only a
fraction of the 124.3 million. GCA-T-3 at 7 (Morrissey).
81
VII.
CONCLUSION
The Postal Service’s rate increase proposal substantially and, with one
exception, adversely affects the interests of the citizen mailers who mail First Class
single piece mail. The Commission’s recommended decision should include the
following determinations:
·
Recommend acceptance of the Postal Service’s proposal to introduce a
Forever Stamp which will provide an important benefit for citizen mailers,
particularly users of greeting cards.
·
Recommend rejection of the Postal Service’s proposal to delink
workshared and single-piece First Class mail rates because such
delinking is inefficient, unfair, inequitable, and contrary to longstanding
Commission policy and precedent.
·
Find that the proposed increase to First Class Letter mail rates should be
limited to a first ounce rate of no more than 41 cents in order to moderate
the unfair and economically irrational trend toward ever-greater
institutional cost burdens on First Class letter mail, particularly
considering the ongoing erosion of First Class mail volumes in the face of
growing broadband competition and the ongoing increase of Standard
mail volumes.
·
Recommend rejection of the Postal Service’s proposal to reclassify low
aspect letters and thereby increase the surcharge from the present 13
cents per piece to 20 cents per piece, which reclassification is not
82
supported by any cost evidence and would discourage mailing of square
greeting cards and reduce mail volume.
Respectfully submitted,
/s/ James Horwood
James N. Horwood
Peter J. Hopkins
Spiegel & McDiarmid
1333 New Hampshire Avenue, N.W.
Second Floor
Washington, D.C. 20036
Telephone: (202) 879-4000
Facsimile: (202) 393-2866
E-mail: [email protected]
[email protected]
David F. Stover
2970 South Columbus Street
No. 1B
Arlington, Virginia 22206-1450
Telephone: (703) 998-2568
Facsimile: (703) 998-2987
E-mail:
[email protected]
Date: December 21, 2006
83