BEFORE THE POSTAL RATE COMMISSION
WASHINGTON, DC 20268-0001 Postal Rate and Fee Changes, 2006
)) Docket No. R2006-1
BRIEF OF MAGAZINE PUBLISHERS OF AMERICA, INC.,
AND ALLIANCE OF NONPROFIT MAILERS ON PERIODICAL RATES
David M. Levy Ronald S. Flagg Richard E. Young SIDLEY AUSTIN LLP 1501 K Street, N.W. Washington, DC 20005-1401 (202) 736-8000 Counsel for Magazine Publishers of America, Inc., and Alliance of Nonprofit Mailers
December 21, 2006
Postal Rate CommissionSubmitted 12/21/2006 4:22 pmFiling ID: 55488Accepted 12/21/2006
CONTENTS
I. INTRODUCTION...................................................................................................... 1
A. Interests of MPA and ANM ................................................................................... 1
B. Summary of Brief.................................................................................................. 2
1. Attribution of mail processing costs................................................................... 2
2. Attribution of city carrier costs ........................................................................... 2
3. Cost classification of payments into Postal Service Retiree Health Benefits Fund under Section 803 of PAEA ............................................................................ 3
4. Worksharing cost avoidance estimates............................................................. 3
5. Rate design for Periodicals Outside County Mail .............................................. 4
II. THE COMMISSION SHOULD ADOPT THE POSTAL SERVICE’S ECONOMIC ESTIMATES OF THE MAIL PROCESSING COSTS ATTRIBUTABLE TO PERIODICALS OUTSIDE COUNTY CLASS MAIL ......................................................... 5
A. In this docket, USPS witness McCrery has provided a response to the Commission’s request for more empirical information on setup and takedown time in operations. This response shows that setup and takedown time is better characterized as fixed than as variable with respect to volume changes..................... 5
B. With improved IOCS data, USPS witnesses Bozzo and McCrery have shown the substantial consistency between Bozzo’s econometric estimates and estimates based on operational analyses of fixed costs in mail processing. .......................................... 7
C. Witness Elliott has shown that the IOCS data can be used to update the Commission’s method of estimating volume variability and that the result is inconsistent with the econometric estimates offered by witnesses Roberts and Neels. 8
D. The Commission should accept the USPS econometric estimates of variability after a decade’s worth of econometric refinement and converging operational evidence. ................................................................................................................... 10
E. Witness Haldi’s argument that mail processing is often primarily for a single class of mail is factually incorrect. His argument for continuing to treat non-variable costs as variable is both incomplete and demonstrably irrelevant. ..................................... 11
III. THE COMMISSION SHOULD ADOPT THE ATTRIBUTION OF CITY CARRIER COSTS PROPOSED BY THE POSTAL SERVICE....................................................... 13
A. The Postal Service’s Analysis Of City Carrier Street Time Is The Best Evidence Of Record. ................................................................................................................. 13
B. The Testimony Of OCA Witness Smith (OCA-T-3) Does Not Provide A Reliable Basis For Any Adjustment To the Postal Service Study. ........................................... 16
1. OCA Witness Smith’s Critique of the CCSTS Is Superficial and Provides No Reasonable Basis for Changing Dr. Bradley’s Model............................................. 17
a. Dr. Smith is mistaken: the density variable should be included in the model. 17
b. Dr. Bradley’s Models Are Developed In The Best Tradition Of Operational Cost Modeling..................................................................................................... 17
c. Dr. Smith’s own CCSTS modeling efforts are superficial and result-oriented. 18
d. Dr. Smith recognizes the need for further analysis of the CCSTS data and models. ............................................................................................................... 20
2. The DOIS Data Used By Dr. Smith Are Too Flawed to Be Used For Modeling Now, and May Never Be Useful. ............................................................................ 22
C. The PRC Should Not Give Any Evidentiary Weight To Dr. Smith’s Responses To POIR 25..................................................................................................................... 24
IV. COST CLASSIFICATION OF PAYMENTS INTO POSTAL SERVICE RETIREE HEALTH BENEFITS FUND UNDER SECTION 803 OF PAEA..................................... 31
V. THE COMMISSION SHOULD ADOPT THE IMPROVED COST AVOIDANCE METHODOLOGIES AND MODELS PROPOSED BY MPA/ANM WITNESS GLICK. ... 33
A. The benchmarks used by MPA/ANM to estimate the Carrier Route and nontransportation destination entry cost avoidances represent the mail that is most likely to convert to worksharing.................................................................................. 33
B. The Postal Service’s flats mail processing cost model (USPS-LR-L-43) understates presort cost avoidances in general and the Carrier Route presort cost avoidance in particular and thus should be rejected. ................................................. 37
1. Two clearly wrong assumptions made by USPS witness Miller dramatically understate the Carrier Route cost avoidance. ........................................................ 37
2. In general, Mr. Miller understates flats presort cost avoidances by excluding some costs that are incurred sorting flats and bundles as well as costs that, under USPS costing methods, have been found to vary proportionally with these costs, from his proportional CRA adjustment. .................................................................. 40
C. The MPA/ANM flats mail processing cost model (MPA/ANM-LR-2) should be adopted because it is more accurate than the USPS model (USPS-LR-L-43). ......... 43
D. MPA/ANM-LR-2 appropriately reflects mail processing cost differences between non-Carrier Route flats and Carrier Route flats in the flat preparation cost pool........ 46
E. MPA/ANM-LR-2 appropriately treats all costs that are incurred sorting Periodicals Outside County flats and bundles as proportional costs. ........................................... 46
F. MPA/ANM’s estimates of the nontransportation destination entry cost model (Glick Direct (MPA/ANM-T-2) at 34) should be used to estimate the nontransportation cost avoided through destination entry. ..................................................................... 47
G. Witness Glick’s estimates of the cost of pallets and sacks and the unit cost avoided through palletization should be used because they include container-handling costs and cost differences at both destination and non-destination facilities. ............ 48
H. Witness Glick’s estimate of the 5-Digit pallet cost avoidance is reasonable. ...... 48
VI. RATE DESIGN FOR PERIODICALS OUTSIDE COUNTY MAIL ....................... 49
A. The Governing Ratemaking Standards .............................................................. 49
B. The Commission should reject the Postal Service’s proposed rate design for Periodicals Outside County Mail. ............................................................................... 51
1. The USPS proposal only modestly increases existing incentives to commingle and dropship periodicals. ....................................................................................... 51
Table 1 .......................................................................................................................... 52
Percentage Increase in Postage Incentive .................................................................... 52
to Co-Mail Publication ................................................................................................... 52
2. The USPS proposal would create a disincentive to the efficient preparation of periodicals on 5-Digit pallets. ................................................................................. 54
3. The Postal Service’s proposed container charge would create rate shock for some small publishers and perverse incentives for inefficiency. ............................ 55
4. The proposed Carrier Route and destination entry discounts are too small because they are based upon inaccurate and understated cost avoidance estimates as well as inappropriate assumptions about future costs....................... 58
5. The proposed disproportionate 18 percent increase in the Ride Along rate is an example of mechanistic ratemaking at its worst. .................................................... 59
C. The Commission Should Adopt The Periodicals Outside County Rate Design Proposed By MPA And ANM. .................................................................................... 59
1. The MPA/ANM rate design provides stronger incentives for efficient mail preparation than does the USPS rate design. ........................................................ 59
2. Publishers and printers will respond to these greater incentives: hence, discounts are more than just a “reward” for existing worksharing. ......................... 60
a. Increased incentives for worksharing will change the behavior of publishers and printers at the margin................................................................................... 62
b. That properly-sized worksharing discounts would incidentally “reward” mailers who would be willing to workshare even with smaller discounts is no justification for departing from ECPR-compliant prices....................................... 66
3. Increased commingling and dropshipping will benefit the subclass as a whole, because cost avoidances still exceed the discounts proposed by Mr. Glick. ......... 67
4. The MPA/ANM proposal will not cause undue rate shock for small publications. 67
a. Even for small mailers that do not change their worksharing behavior, the largest increase under the MPA/ANM proposal is only about half of the largest increase under the USPS proposal. ................................................................... 68
b. ABM has understated the extent to which publishers can offset the rate increases through greater commingling and destination entry. .......................... 72
5. Special issues raised by 5-digit pallet discount ............................................... 75
CONCLUSION .............................................................................................................. 77
BEFORE THE POSTAL RATE COMMISSION
WASHINGTON, DC 20268-0001 Postal Rate and Fee Changes, 2006
)) Docket No. R2006-1
BRIEF OF
MAGAZINE PUBLISHERS OF AMERICA, INC., AND ALLIANCE OF NONPROFIT MAILERS
ON PERIODICAL RATES
The Magazine Publishers of America, Inc. (“MPA”) and the Alliance of Nonprofit
Mailers (“ANM”) respectfully submit this brief, which covers rates for Outside County
Periodicals Mail.
I. INTRODUCTION
A. Interests of MPA and ANM
MPA is the national industry association of publishers of consumer magazines.
It represents more than 240 domestic publishing companies, with approximately 1,400
titles; more than 60 international members; and more than 100 associate members.
MPA members are substantial users of the Outside County subclass of Periodicals Mail.
ANM is a nonprofit corporation, chartered in the District of Columbia,
representing the interests of nonprofit organizations in postal rate matters. ANM's
members include many of the Nation's best-known charitable, religious, educational,
scientific and other nonprofit organizations, as well as smaller community organizations
and their umbrella groups. Many of ANM's members publish periodicals, and rely
heavily on the Nonprofit Outside County subclass of Periodicals Mail.
B. Summary of Brief
1. Attribution of mail processing costs
The Commission should adopt the Postal Service’s econometric estimates of
mail processing volume variability. The empirical data supplied in this docket by USPS
witness McCrery show that setup and takedown time in mail processing is better
characterized as fixed than as variable with respect to volume changes. As a result,
alternate estimates of mail processing variability, which are derived from the new IOCS
data and properly categorize setup and takedown time as fixed, demonstrate a
remarkable consistency with the Postal Service’s econometric estimates. The IOCS
data can be used to update the Commission’s method of estimating volume variability,
and produce results inconsistent with the econometric estimates by OCA witness
Roberts and UPS witness Neels. After a decade of refinements in the USPS
econometric analysis, the demonstrated consistency between those results and the
operational evidence presented in this case warrant acceptance of those econometric
estimates. Finally, Valpak witness Haldi has failed to provide a valid economic
justification for treating mail processing costs as attributable that the current record
indicates to be largely fixed with respect to volume changes.
2. Attribution of city carrier costs
The Commission should adopt the attribution of city carrier costs proposed by the
Postal Service as the best evidence of record. While the Commission’s frustration with
lack of progress in this area since R2005-1 is understandable, the record offers no
better alternative to the Postal Service study and results. The criticisms of the Postal
Service study by OCA witness Smith are mistaken: Dr. Bradley properly included a
density variable in his models; the models are consistent with modern economic
research techniques; Dr. Smith’s own CCSTS models are superficial and result-
oriented; and Dr. Smith himself recognizes the need for further work in this area.
- 2 -
Moreover, the DOIS data used by Dr. Smith are too flawed to be used for modeling, and
may prove incapable of being made useable.
The Commission should give no evidentiary weight to Dr. Smith’s 11th-hour
responses to POIR 25. Apart from the due process issues raised by MPA, ANM and
thirteen other intervenors in their December 18 objection to designation of these
responses into the record, Dr. Smith’s analysis raises serious and unresolved
methodological questions that have not been properly investigated.
3. Cost classification of payments into Postal Service Retiree Health Benefits Fund under Section 803 of PAEA
The Postal Accountability and Enhancement Act of 2006 (“PAEA”) became law
on December 20, 2006, one day before the filing deadline of this brief. Although
complete analysis of the ramifications of the Act for this case is not possible, one title of
the Act warrants brief discussion: Title VIII – Postal Service Retirement and Health
Benefits Funding. In particular, Section 803 requires the Postal Service to make fixed
annual payments over the next ten years into the Postal Service Retiree Health Benefits
Fund, including $5.4 billion in the Test Year of this rate case. For the reasons explained
below, the Test Year payment into the Postal Service Retiree Health Benefits Fund
under PAEA should be treated entirely as an institutional cost.
4. Worksharing cost avoidance estimates
The Commission should adopt the improved cost avoidance methodologies and
models proposed by MPA/ANM witness Glick:
• The benchmarks used by MPA and ANM to estimate the Carrier Route and nontransportation destination entry cost avoidances represent the mail that is mostly likely to convert at the margin to worksharing and they should be adopted.
• The Postal Service’s flats mail processing cost model (USPS-LR-L-43) understates presort cost avoidances in general and the Carrier Route presort cost avoidance in particular, and thus should be rejected.
- 3 -
• The MPA/ANM model of flats mail processing costs (MPA/ANM-LR-2) should be adopted because it is more accurate than the USPS model (USPS-LR-L-43).
• MPA/ANM-LR-2 appropriately reflects mail processing cost differences between non-Carrier Route flats and Carrier Route flats in the flat preparation cost pool.
• MPA/ANM-LR-2 appropriately treats all costs that are incurred sorting Periodicals Outside County flats and bundles as proportional costs.
• MPA/ANM’s estimates of the nontransportation destination entry cost model (Glick Direct (MPA/ANM-T-2) at 34) should be used to estimate the nontransportation costs avoided through destination entry.
• Witness Glick’s estimates of the cost of pallets and sacks and the unit cost avoided through palletization should be used because they include container-handling costs and cost differences at both destination and non-destination facilities.
• Witness Glick’s estimate of the 5-Digit pallet cost avoidance is reasonable.
5. Rate design for Periodicals Outside County Mail
In deciding which rate design to adopt for Periodicals Mail, the Commission
should seek to provide stronger incentives for presorting, dropshipping, palletizing and
other cost-efficient measures, while protecting small publications from rate shock. The
rate design proposed by the Postal Service fails to accomplish either of these goals. It
would increase only modestly the existing incentives for commingling and dropshipping;
it would create a disincentive for preparing periodicals efficiently on 5-digit pallets; and it
would create rate shock for some small publishers. The proposed Carrier Route and
destination entry discounts are too small because they are based upon inaccurate and
understated cost avoidance estimates as well as inappropriate assumptions about
future costs. And the proposed 18 percent increase in the Ride Along Rate is an
example of mechanistic ratemaking at its worst.
Instead, the Commission should adopt the Periodicals Outside County rate
design proposed by MPA and ANM. This rate design would provide stronger incentives
- 4 -
than the USPS rate design for efficient mail preparation. Publishers and printers will
respond to these incentives by engaging in additional co-mailing, co-palletization,
dropshipping, and presorting; hence the proposed discounts would be more than just a
“reward” for existing worksharing. At the same time, the MPA/ANM proposal will not
inflict undue rate shock on small publications. Even for small mailers that do not
change their worksharing behavior, the largest increase under the MPA/ANM proposal
is only about half the largest increase under the USPS proposal. Moreover, ABM has
understated the extent to which publishers can offset even these rate increases through
greater commingling and destination entry. Finally, the 5-digit pallet discount proposed
by MPA/ANM will not cause excessive leakage; and the concerns of USPS witness
McCrery that the discount could encourage the entry of lightly loaded pallets at entry
points distant from the destination of the mail can be allayed by conditioning eligibility
for the discounts on DDU/DSCF entry and mailpiece loadings of at least 250 pounds per
pallet.
II. THE COMMISSION SHOULD ADOPT THE POSTAL SERVICE’S ECONOMIC ESTIMATES OF THE MAIL PROCESSING COSTS ATTRIBUTABLE TO PERIODICALS OUTSIDE COUNTY CLASS MAIL
A. In this docket, USPS witness McCrery has provided a response to the Commission’s request for more empirical information on setup and takedown time in operations. This response shows that setup and takedown time is better characterized as fixed than as variable with respect to volume changes.
The Commission’s existing methodology for estimating the volume variability of
mail processing implicitly treats setup and takedown time as volume variable. In Docket
No. R2000-1, however, the Commission acknowledged that USPS witness Degen’s
argument that “scheme changes, not volumes, drive the number of setups and
takedowns, particularly in secondary operations,” could be “partly valid.” The
Commission noted, in particular, that “higher volume will sometimes lengthen runs
- 5 -
within a scheme without multiplying set-up and tear-down cycles.” R2000-1 Op. & Rec.
Decis., App. F, at 18-19.
The Commission raised the possibility, however, that “[n]arrow processing
windows can severely restrict the opportunity to lengthen runs for a given scheme” and
that “higher volumes are likely to cause the same scheme to be replicated.” Id. The
Commission ultimately concluded that the record was “not developed well enough to
support definitive findings on what the ratio of fixed set-up and tear-down time to
runtime is in any of the operations modeled by witness Bozzo.” Id. at 19. The
Commission asked that “[s]ome attempt to quantify the amount of fixed setup/shutdown
time … be provided in future proceedings.” R2000-1 Op. & Rec. Decis. ¶ 3033.
The record in the current case provides an answer to the Commission’s request.
In response to an interrogatory by MPA and ANM, USPS operational witness McCrery
has provided a nationwide snapshot of the sort schemes running at five different times
of day on May 18, 2006, on the Postal Service’s DBCS and AFSM 100 machines. The
snapshot shows that the majority of schemes are run on a single machine for all types
of sorts. For incoming secondary sorts—the majority of sort schemes—essentially all
schemes are run on a single machine. 11 Tr. 2896-97 (response of USPS witness
McCrery to MPA/USPS-T42-22(e)).
Mr. McCrery confirms that most sort schemes are incoming secondaries, which
are almost always run on a single machine at a facility. 11 Tr. 2896 (response of USPS
witness McCrery to MPA/USPS-T42-22(c), (d)). For letters, of which 79 percent of the
incoming secondary volume is sorted to DPS, the DPS sorting procedure requires that
each sort scheme be run on only a single machine. McCrery Direct (USPS-T-42) at 12,
n. 10; Id. at 36, lines 15-18. Even non-DPS incoming secondary sort schemes are
almost always run on only a single machine. 11 Tr. 2896 (response of USPS witness
McCrery to MPA/USPS-T42-22(d)).
- 6 -
Mr. McCrery’s testimony shows that the structure of Postal Service sorting
operations is such that the number of sort schemes run—and therefore the time spent in
setup and takedown—are better characterized as fixed rather than variable with respect
to volume changes. This is particularly true for the incoming secondary schemes—i.e.,
the majority of schemes—because they are rarely run on multiple machines.
B. With improved IOCS data, USPS witnesses Bozzo and McCrery have shown the substantial consistency between Bozzo’s econometric estimates and estimates based on operational analyses of fixed costs in mail processing.
With improved data available with the IOCS redesign, USPS witness Bozzo has
provided information on the composition of activities in the sorting cost pools that have
been estimated econometrically. Bozzo Direct (USPS-T-12) at 26-27, Table 2. The
data show substantial amounts of time in setup and takedown activities, as well as in
waiting, activities that are best characterized as fixed with respect to changes in volume.
In addition, the data show substantial amounts of time in container and other handling,
which an operational analysis suggests is likely to have a volume variability substantially
below 100 percent. Bozzo Direct (USPS-T-12) at 29-30; McCrery Direct (USPS-T-42)
at 38.
USPS witness McCrery provides a “bookkeeping” estimate of cost variability
using the IOCS figures on waiting and scheme changes (setup and takedown) supplied
by Dr. Bozzo. McCrery Direct (USPS-T-42) at 37-40. Dr. Bozzo provides estimates
based on the IOCS figures that alternately include waiting and scheme changes alone
or that combine waiting, scheme changes, and container handlings. Bozzo Direct
(USPS-T-12) at 77-80. These comparisons show a close consistency between the
econometric estimates and the estimates based on operational analyses of the fixed
costs in mail processing. For the letters composite, Dr. Bozzo’s IOCS-derived
estimates of fixed activity time range from 88 to 93 percent, compared to an
econometric composite of 88 percent with a standard error of 4 percentage points. For
- 7 -
the flats composite, Dr. Bozzo’s IOCS-derived estimates of fixed activity time range
from 87 to 92 percent, compared to an econometric composite of 92 percent with a
standard error of 5 percentage points. 10 Tr. 2528 (response of USPS witness Bozzo
to MPA/USPS-T12-3a). These are impressive levels of consistency between the
econometric results and IOCS-derived estimates that reflect operational understanding
of activities that are best characterized as relatively fixed with respect to volume
changes (waiting and scheme changes) or as having an intermediate level of volume
variability that is substantially below 100 percent (container handling).
C. Witness Elliott has shown that the IOCS data can be used to update the Commission’s method of estimating volume variability and that the result is inconsistent with the econometric estimates offered by witnesses Roberts and Neels.
MPA et al. witness Elliott has shown how the Commission’s method of estimating
volume variability can be updated to reflect the new IOCS data on setup and takedown
time that have been made available in this docket. This approach represents a
conservative change that is likely to overstate the level of volume variability because it
still excludes consideration of container handling.
The Commission’s method of estimating variable costs in mail processing rests
on an operational analysis that classifies activities as either “fixed” or “variable” based
on operational considerations, and then uses operational data from IOCS tallies to
calculate the proportion of costs in the “variable” cost categories. The Commission
identifies a specific set of activity codes as fixed and a larger category of mail
processing activities that are defined to be variable. See R97-1 Op. & Rec. Decis.
3010-3012; R97-1 USPS-LR-H-1; USPS-LR-L-100, file PRCACTV.rtf.
The updated approach relies on witness Bozzo’s responses to MPA and ANM
interrogatories that provide information about the costs associated with setup and
takedown time for all mail processing cost pools for which the redesigned IOCS
- 8 -
provides appropriate codes. 10 Tr. 2508-2526 (response of USPS witness Bozzo to
MPA/USPS-T12-1); 10 Tr. 2545-2546 (response of USPS witness Bozzo to
MPA/USPS-T12-4, MPA-ANM-4.xls, worksheet “data”). Dr. Elliott adds these setup and
takedown costs to the fixed costs of the Commission method, to derive a revised
Commission estimate of volume variability for all mail processing cost pools that have
non-zero setup and takedown costs. Elliott Reb. (MPA et al.-RT-2) at 4-7 and 13 (36 Tr.
12358-12361 and 12367), Table 1 and Exhibit 1.
Dr. Elliott further shows that the econometric estimates of UPS witness Neels
and OCA witness Roberts are inconsistent with the variability estimates of this revised
version of the Commission approach that correctly includes setup and takedown costs
as fixed with respect to volume changes. Id. at 7-9 (36 Tr. 12361-12363).
UPS witness Neels produces plant-level estimates of variability of 114 percent
(with a 95 percent confidence interval from 101 to 126 percent) and 103 percent (with a
95 percent confidence interval from 96 to 110 percent). He claims that “these
preliminary results are consistent with the Commission’s established position that mail
processing costs are 100 percent volume variable.” Neels Direct (UPS-T-1) at 57, lines
5-7. Dr. Elliott shows that witness Neels’ plant-level estimates are inconsistent with the
Commission approach when it is adjusted to reflect fixed character of setup and
takedown costs with respect to volume changes. The revised Commission estimates
produce plant-level volume variability of 92.2 percent, clearly below the econometric
estimates of Dr. Neels. Elliott Reb. (MPA et al.-RT-2) at 7-8 (36 Tr. 12361-12362).
Similarly, the econometric estimates of OCA witness Roberts for letters are
inconsistent with the adjusted Commission approach.1 Dr. Roberts produces 1 Dr. Roberts provides several estimates for the variability of flats but does not recommend that any of them be used for allocating postal costs because of their sensitivity to the data sample chosen and the imprecision of the estimates. OCA-T-1 at 44-50, especially at p. 50, lines 5-18.
- 9 -
econometric estimates of variability for letters of 127.6 percent and a 95 percent
confidence interval from 115.6 to 139.6 percent. 23 Tr. 8300-8301 (response of OCA
witness Roberts to USPS/OCA-T1-8(b)). Dr. Elliott shows that Dr. Roberts’ letter
estimates are also not consistent with the Commission approach, when that approach is
adjusted to reflect the fact that setup and takedown costs are best characterized as
fixed with respect to volume changes: the revised Commission estimates produce letter
volume variability of 92.3 percent, which clearly falls below the econometric estimates of
Dr. Roberts. Elliott Reb. (MPA et al.-RT-2) at 8-9 (36 Tr. 12362-12363).
D. The Commission should accept the USPS econometric estimates of variability after a decade’s worth of econometric refinement and converging operational evidence.
The econometric estimates of mail processing volume variability provided by
USPS witness Bozzo in the current case reflect a refined approach that has benefited
from nearly a decade’s worth of improvement in econometric approach and attention to
data quality, starting with the work provided by USPS witness Bradley (USPS-T-14) in
R97-1. The Commission noted in its decision in R2005-1 that the model of witness
Bozzo incorporated “several important improvements over the models offered in
previous dockets.” R2005-1 Op. & Rec. Decis. ¶ 4097.
With the current case, the USPS has shown that the econometric estimates of
variability using MODS data are remarkably consistent with alternate estimates using
the new IOCS data and informed by an operational understanding of the volume
variability of different activities. The latter comparison makes use of the same type of
operational analysis that underlies the Commission’s current approach to estimating
volume variability within mail processing. The substantial convergence of the
econometric and operational approaches underlines the improvement of the USPS
econometric approach after a substantial period of development.
- 10 -
In contrast, the econometric approaches provided in this case by UPS witness
Neels and OCA witness Roberts conflict with the operational evidence. The lack of
congruence with the operational evidence for the UPS and OCA econometric
approaches adds further weight to the substantial technical critiques of those
approaches by USPS witness Bozzo. Bozzo Reb. (USPS-RT-5), passim.
Given the congruence between the USPS econometric approach and the
operational evidence, and the conflict between the alternate econometric approaches
on the record and the operational evidence, the Commission should adopt the USPS
econometric estimates as the best currently available estimates of mail processing
variability.
However, if the Commission holds that its remaining concerns about data quality
and econometric method preclude the adoption of the USPS econometric estimates,
then it should adopt witness Elliott’s augmented version of the Commission method so
that setup and takedown costs are correctly approximated as fixed. The record of the
current case indicates clearly that these costs are better approximated as fixed than as
variable (as they are treated currently in the Commission’s approach).
E. Witness Haldi’s argument that mail processing is often primarily for a single class of mail is factually incorrect. His argument for continuing to treat non-variable costs as variable is both incomplete and demonstrably irrelevant.
ValPak witness Haldi argues that certain mail processing costs, even when fixed,
should be attributed to individual mail classes because the costs are often incurred
primarily for a single class of mail. Haldi Direct (VP-T-2) at 43-46. Dr. Haldi
misapprehends the record, however, and his position is inconsistent with USPS witness
McCrery’s responses to ValPak’s interrogatories. MPA et al.-RT-2 at 9-12 (Tr. 12363-
12366). The record shows that single-class sortation runs occur only for a minority of
sort schemes.
- 11 -
Dr. Haldi further argues that even for cases where multiple classes of mail are
processed together that it is acceptable to continue to treat non-variable costs as
variable in order to avoid the burden of figuring out how to construct an incremental cost
test according to shape:
“[S]o long as the Commission continues to treat mail processing costs as essentially 100 percent volume variable, it has no need to study how to treat non-volume variable mail processing costs, nor does it need to resolve the thorny issues raised thereby.”
Haldi Direct (VP-T-2) at 54, lines 15-18.
Dr. Haldi is effectively recommending that non-volume variable costs be
attributed in the manner appropriate only for volume variable costs. As USPS witness
Bozzo has pointed out, the burden thus is on Dr. Haldi, the proponent of using an
incorrect allocation method, to demonstrate that the end result would coincide with the
results of using the correct method. This Dr. Haldi has not done. 36 Tr. 12512, lines 1-
9 (Bozzo).
In any event, Dr. Haldi has failed to show that the attribution of non-volume
variable costs is necessary to ensure that each class and subclass, and each
combination of classes and subclasses, covers the relevant incremental costs. The
premise of Dr. Haldi’s argument for extended attribution is that the non-volume variable
costs caused by a single class, or a handful of classes, is so large that failure to
attribute them to the relevant classes would result in rates that fail to cover all of the
incremental costs of those classes. For the cluster of mail classes with which
Periodicals Mail is processed, however, this concern has no practical significance
because the revenue generated by the classes exceeds their combined incremental
costs by an amount far in excess of any conceivable measure of the fixed costs of mail
processing. As Dr. Bozzo has noted:
- 12 -
As a practical matter, the magnitudes of non-volume-variable costs in sorting operations are insufficient to lead to problems even if Dr. Haldi’s arguments are given maximum credence. At best, Dr. Haldi describes a theoretical problem that careful analysis shows not to have any practical significance for the ratemaking scheme’s cost tests.
Bozzo Reb. (USPS-RT-5) at 10-11 (36 Tr. 12402-12403).
The Commission should not be confused by Dr. Haldi’s incomplete arguments
into thinking that it should continue to treat incorrectly the fixed costs of setup and
takedown time according to the procedure for variable costs rather than according to the
procedure specified for fixed costs.
III. THE COMMISSION SHOULD ADOPT THE ATTRIBUTION OF CITY CARRIER COSTS PROPOSED BY THE POSTAL SERVICE
A. The Postal Service’s Analysis Of City Carrier Street Time Is The Best Evidence Of Record.
In this proceeding, the Commission is faced with an unusual situation with
respect to the attribution of city carrier street time. In Docket No. R2005-1, the Postal
Service presented an entirely new city carrier street-time study based on time and
volume data collected from city carriers in 2002. The study was complex, involving
considerable cooperation of carriers in the field, complicated data collection procedures,
considerable data processing and manipulation, and econometric modeling. Despite
extensive discovery on the data and models and the support of interveners for the
study, the Commission recognized that settlement of the case had prevented full-blown
adversarial testing of the study. In its Opinion, the Commission also identified several
shortcomings in the data and modeling effort. PRC Op. R2005-1 ¶¶ 4005-4018, 4030-
4091 and App. I. Ultimately, however, the Commission recommended rates based on
the new study, and the 2002 data underlying it, on the grounds that the study was a
major improvement over the previous street-time analyses, and the Postal Service
seemed likely to continue working to improve the analysis:
- 13 -
The Commission is gratified that the Postal Service has reviewed its approach to carrier street time variability analysis from the “ground up” and collected new data on carrier street time activity that are designed to support improved econometric modeling of that variability. It urges the Postal Service to continue its analytical work to improve the quality of the data gathered, and to explore additional econometric models that will yield more robust results.
PRC Op. & Rec. Decis. R2005-1, ¶ 4091.
The Postal Service filed its Request in R2006-1 on May 3, 2006. To attribute city
carrier street-time costs, the Service again relied on the city carrier street-time variability
results derived from the 2002 survey data and the econometric model that the
Commission had accepted in Docket No. R2005-1. USPS witness Bradley explained
the Postal Service’s failure to make any changes in its analysis as follows:
The Postal Service is very appreciative of the time the Commission took to analyze the study and for the extensive and detailed nature of the Commission’s criticisms and suggestions for improvement. Unfortunately, the Commission’s Opinion and Recommended Decision was issued just one week before the base year city carrier costs were due for constructing the product costs in the current docket. One week is not sufficient time just to review and comprehend over eighty pages of Commission comments and criticisms, let alone respond appropriately. There was simply insufficient time to refine the city carrier costing study along the lines indicated by the Commission. Because of this time limitation, Postal Service decided that the best available city carrier street time costing methodology was the one chosen by the Commission in Docket No. R2005-1.
Bradley Direct (USPS-T-14) at 10.
The Commission’s frustration over the lack of progress in refining the study since
R2005-1 is entirely understandable. The study and its results, however, clearly remain
the best available evidence of the record in this case. Despite the lack of litigation in
R2005, there has been very little discovery on the Postal Service’s direct case in this
proceeding. Only the OCA directly asked questions concerning the Postal Service’s
- 14 -
CCSTS data collection, modeling methodology, and econometric features and even
OCA asked relatively few questions.2
Moreover, OCA was also the only participant to file a direct case on the city
carrier street time analysis. The OCA witness, Dr. Smith, developed an assortment of
models, some developed from the CCSTS data and some developed from a set of
Delivery Operations Information System (DOIS) data requested from the Postal Service.
And, he recommends one CCSTS and one DOIS model. Yet, with respect to both
those models, Dr. Smith admitted that his efforts were very preliminary and that more
work is required on both (CCSTS and DOIS) data and modeling issues.3 Moreover, as
explained below, Dr. Smith’s CCSTS analyses are superficial and misguided, providing
no improvement over the Postal Service’s proposal, and the data underlying his DOIS
models may be fatally defective.
The Commission has already announced its intention to open a proceeding
focused on the carrier street time analyses.4 Such a proceeding would be an
appropriate forum for developing a full and thorough record and making informed and
tested decisions regarding city carrier street time data and modeling. In the present
case, however, there is no record evidence demonstrating the superiority of any
2 The OCA issued 12 interrogatories to Dr. Bradley on the subject. One of those concerned a request for DOIS data and received an institutional response. Valpak issued 17 interrogatories on the subject, but principally asked questions on data collection (specifically with respect to saturation and sequenced mail volumes) and did not probe on specific modeling or econometric issues. 3 With respect to his superficial investigation, see Smith’s responses to ADVO/OCA-T3-1, 2, 15, 24 (22 Tr. 8127-8129, 8079-8081, 8147). With respect to his recognition that further work is required, see Smith Direct (OCA-T-3) page 2 (lines 11-16), page 3 (lines 9-10), page 8 (lines 23-24), page 15 (lines 6-8), page 21 (lines 7-11), page 22 (lines 18-19), page 23 (lines 7-15), and responses to ADVO/OCA-T3-10 and 15 (22 Tr. 8135, 8079-8081), USPS/OCA-T3-15 (22 Tr. 8111). With respect to DOIS, see Smith Direct (OCA-T-3), page 16 (lines 7-10). 4 Docket R2006-1, PRC Order No. 1482 at 6.
- 15 -
alternative to the Postal Service study and Postal Service results. Under the
circumstances, and given the far-reaching effects that a change in the attribution of city
carrier costs would have on the rates for nearly all postal services and most postal
products in this case, prudence dictates adhering to the Postal Service study until an
alternative approach has been fully tested and verified.
B. The Testimony Of OCA Witness Smith (OCA-T-3) Does Not Provide A Reliable Basis For Any Adjustment To the Postal Service Study.
OCA witness Smith’s carrier cost testimony has two parts: one concerning the
CCSTS data and model and one concerning the DOIS data that the OCA recently
acquired from the Postal Service and models developed from those data.
In the first part, he criticized Dr. Bradley’s CCSTS data and model. In particular,
he simply noted problems with the CCSTS database: those discussed by the
Commission in its R2005-1 Opinion, including the presence of significant
multicollinearity. He also claimed that Dr. Bradley was wrong to include a density
variable in the CCSTS model5 and that Dr. Bradley’s models are “ad hoc” and not
based on economic theory. Finally, he presented 24 models developed from the
CCSTS data, briefly discussed them, and recommended one to the Commission, should
it decide not to use the DOIS data.
In the second part, Dr. Smith briefly discussed the DOIS data. He then
presented a variety of models developed from that data, briefly discussed each,
recommending one to the Commission, should it decide to use the DOIS data rather
than the CCSTS data.
5 In response to POIR No. 25, he also expands on that claim. Unfortunately, Dr. Bradley and Ms. Crowder have no opportunity to provide any rebuttal to his expansion.
- 16 -
1. OCA Witness Smith’s Critique of the CCSTS Is Superficial and Provides No Reasonable Basis for Changing Dr. Bradley’s Model
a. Dr. Smith is mistaken: the density variable should be included in the model.
Both Dr. Bradley and Ms. Crowder agreed that Dr. Smith was simply wrong to
claim that (1) density is an output of (endogenous to) the city carrier cost minimization
process and (2) therefore should be eliminated from the model. To begin with, it
appears that Dr. Smith may have confused route density with zip code density. The
former may be the result of management determination of a minimum-cost set of routes
within a zip but not the latter. Bradley Reb. (USPS-RT-4) at 7-9 (34 Tr. 11560-11562)
and Crowder Reb. (MPA et al-RT-1) at 8 n. 13 (34 Tr. 11647).
Density is a key explanatory variable in the street time model and is required to
control for the effects from variations in distances among delivery points. Eliminating
this variable will automatically bias the coefficients for all volume variables. Crowder
Reb. (MPA et al-RT-1) at 8-10 (34 Tr. 11647-11649) and Bradley Reb. (USPS-RT-4) at
8-9 (34 Tr. 11561-11562). For this reason, many carrier cost studies performed by
postal authorities have included the density variable. Bradley Reb. (USPS-RT-4) at 1-5
(34 Tr. 11554-11558). The street time model requires a density variable.
b. Dr. Bradley’s Models Are Developed In The Best Tradition Of Operational Cost Modeling.
Dr. Smith asserted that Dr. Bradley’s models are ad hoc and lack a theoretical
basis. However, they are, in fact, fully consistent with modern economic research
techniques. Bradley Reb. (USPS-RT-4) at 10-12 (34 Tr. 11563-11565). Dr. Bradley
explained the economic basis for his specifications of the CCSTS models in his R2005-
1 testimony, explicitly outlined the process to be used in specifying and estimating the
econometric cost function, clearly identified the variables to be included in the model,
and explained why he chose the quadratic functional form for the estimation process.
- 17 -
Bradley Reb. (USPS-RT-4) at 12-13 (34 Tr. 11565-11566). His choice of zip-code level
modeling and selected explanatory variables clearly recognizes cost effects from route
management and route adjustments made within zip codes. Id. Finally, he discussed
and addressed the known problems of multicollinearity and heteroskedasticity.
This is not to say that Dr. Bradley’s models are perfect and need no
improvement. Even he recognizes that there are outstanding issues that should be
investigated and alternative specifications that should be explored, leading to possible
model improvements. But, Dr. Bradley has provided a theoretical basis for his models.
Dr. Smith may not like Dr. Bradley’s basis, but his claims regarding Dr. Bradley’s lack of
theory should be ignored.
c. Dr. Smith’s own CCSTS modeling efforts are superficial and result-oriented.
Despite his criticisms of the CCSTS data and Dr. Bradley’s regular delivery
model, Dr. Smith provided no independent analysis of the data and offered no
alternative conceptual basis for his models. Dr. Bradley and Ms. Crowder both noted
the superficiality of his efforts. Bradley Reb. (USPS-RT-4) at 13-20 (34 Tr. 11566-
11573); Crowder Reb. (MPA et al.-RT-1) at 5-8 (34 Tr. 11644-11647). With respect to
the CCSTS data, Dr. Smith simply relied on the analyses in the Commission’s R2005-1
Opinion and performed no independent data culling, cleaning, segmentation, outlier
analysis, or other database modifications in an attempt to improve the data quality.
Smith Responses to USPS/OCA-T3-21 (22 Tr. 8118), ADVO/OCA-T3-1, 2, 24(c) (22 Tr.
8127-8129, 8147). Thus, despite his criticism of the CCSTS database, he used the
same CCSTS data as Dr. Bradley.
With respect to his modeling efforts, Dr. Smith simply adopted Dr. Bradley’s
quadratic structural model and, for each of his alternatives, deleted and/or made
- 18 -
changes in the original variables or mix of variables.6 Dr. Bradley called this the
“kitchen sink” approach – it mechanistically re-estimates versions of the model using
different permutations of the independent variables. Bradley Reb. (USPS-RT-4) at 14
(34 Tr. 11567). By his own admission, Dr. Smith did not develop an appropriate
theoretical basis to determine the most important cost causal variables and how they
should be combined to explain street time in a model specification.7 Smith Responses
to ADVO/OCA-T3-15(e) and 24 (22 Tr. 8081, 8147). Further, lacking a theoretical
basis, Dr. Smith was unable to test his specified models using the appropriate statistical
indicators to determine if model results comport with expected cost behavior – thus his
extremely brief discussion of each of his models. Crowder Reb. (MPA et al.-RT-1) at 6
(34 Tr. 11645). And, he provided no econometric tests for including or excluding the
variables he used. Bradley Reb. (USPS-RT-4) at 14 (34 Tr. 11567). This is especially
surprising in light of his criticism of Dr. Bradley for having no theoretical basis.
With respect to statistical issues, Dr. Smith identified autocorrelation in the data
but did nothing to resolve it and did not even discuss the known heteroskedasticity
problems in the data.8 Crowder Reb. (MPA-RT et al.-1) at 7 (34 Tr. 11646). With
respect to the multicollinearity issue, his only attempt to deal with the problem was
simply to offer restricted versions of his full quadratic models that effectively assign zero
6 Several of the models appear to be developed to address concerns on the impact of the differences between the CCSTS and DOIS data rather than specific attempts to develop an improved CCSTS model. Crowder Reb. (MPA et al.-RT-1) at 6, fn. 7 (34 Tr. 116454). 7Dr. Smith did, however, make some positive contributions to the modeling of carrier street time. These should be explored in future investigations. Bradley Reb. (USPS-RT-4) at 18 (34 Tr. 11571) and Crowder Reb. (MPA et al.-RT-1) at 10 (34 Tr. 11649). 8 Dr. Smith did not even provide the heteroskedasticity consistent statistical indicators (t-statistics) that are necessary to consistently evaluate all model results. Thus, his model recommendations are based on an inconsistent and, more importantly, incorrect set of indicators. Crowder Reb. (MPA-RT et al.-1) at 7 (34 Tr. 11646).
- 19 -
values to all cross-product variables. Id. However, he did not recommend a restricted
quadratic but instead simply searched for a full quadratic that appeared reasonable:
i.e., with no “sign problems” and meeting his a priori expectations. None of this is
accepted econometric procedure. It completely disposes of the need for a theoretical
basis and test statistics and it simply ignores the fact that there is a multicollinearity
problem causing large variances in the estimated coefficients. Bradley Reb. (USPS-RT-
4) at 13 n. 26 (34 Tr. 11566).
Because Dr. Smith offered no theoretical basis for a model and did not truly
address any of the econometric problems found in the data, he cannot, provide
legitimate evidence that any of his models are an improvement over Dr. Bradley’s
model. Crowder Reb. (MPA-RT et al.-1) at 7-8, 11 (34 Tr. 11646-47, 11650). Further,
both Dr. Bradley and Ms. Crowder agreed that Dr. Smith’s “kitchen sink” approach
cannot possibly provide a rational foundation for preferring any one model over the
others. Bradley Reb. (USPS-RT-4) at 20 (34 Tr. 11573); Crowder Reb. (MPA-RT et al.-
1) at 6, 11 (34 Tr. 11645, 11650).
d. Dr. Smith recognizes the need for further analysis of the CCSTS data and models.
To his credit, Dr. Smith recognized that he provided only a superficial analysis
and that much necessary investigation of the CCSTS data and models is still required.
Indeed, he agreed that all of the models he presented are inadequate or possibly
inadequate and that “it is not clear that meaningful conclusions can be obtained.” Smith
Direct (OCA-T-3) at 3, 15. Clearly, he had little confidence in his models and,
throughout his testimony, he demonstrated his understanding that more work is needed:
• “My testimony on the methodology of the original City Carrier cost model . . . concludes that additional improvements in the estimation of City Carrier volume variability and data availability are needed.” Smith Direct (OCA-T-3) at 2, lines 11-16.
- 20 -
• “Further specification or explanation of an economic model would be appropriate.” Id., at 8, lines 23-24. “Future work could consider whether some type of economic model, involving minimization of costs subject to some type of constraint could be developed. I have not yet used or examined all of the variables which could be considered, and whether currently unused variables could be combined with alternative models is an interesting issue.” Id., at 21, lines 7-11.
• “Depending on further research and development of postal delivery economic analysis it is possible that additional variables may be found to be appropriate.” Response to ADVO/OCA-T3-10(b) (22 Tr. 8135). ”. . . [W]e need more consideration of the underlying theoretical justification of the modeling effort as well as additional consideration of statistical and econometric issues.” Response to ADVO/OCA-T3-15(a) (22 Tr. 8079).
• “However, given the problems of the underlying database as evidenced by the types of results obtained it appears that the Carrier Cost analysis presented in Docket No. R2005-1 is flawed and that additional analysis is needed.” Smith Direct (OCA-T-3) at 15, lines 6-8.
• “An area for future research will be autocorrelation issues.” Id., at 22, lines 18-19.
• “In performing the modeling effort I considered a wide variety of alternatives to the equation proffered by witness Bradley in the modeling of City Carrier activities. These efforts frequently encountered sign problems, probably due to the underlying deficiencies of the database. Collinearity of the database is a problem, apparently making the application of a full quadratic model very difficult. In consideration of restricted quadratic models, one frequently obtains relationships among the costs that, on an a priori basis, do not appear to be reasonable. Accordingly, I advocate that the Commission view Carrier Cost volume variability as an open question: improvement is needed.” Id., at 23, lines 7-15.
• “The estimation of econometric models using ZIP Code-Day data is consistent with optimization taking place at the ZIP Code level. Whether a better or different model could be developed and how such a model would be estimated has not been determined.” Response to USPS/OCA-T3-15(e) (22 Tr. 8111).
All of these statements highlight the need for much more exploration of city
carrier costing data and modeling issues. In that respect, we agree with Dr. Smith’s call
- 21 -
for more investigatory work. And, until that is done, for the reasons explained in Part A
above, the Postal Service’s proposed city carrier cost attribution should be used.
2. The DOIS Data Used By Dr. Smith Are Too Flawed to Be Used For Modeling Now, and May Never Be Useful.
The OCA has developed a considerable interest in being able to use data from
the ongoing USPS Delivery Operations Information System (DOIS). This interest is
understandable because DOIS apparently has daily information on zip codes and city
routes, covers the vast majority of delivery zones and city carrier routes, has been in
existence now for several years, and is continually updated. As noted by Ms. Crowder,
DOIS provides an opportunity to use an extensive cross-sectional and time-series panel
without having to burden data collectors or carriers with considerable sampling and
testing. Such a database seems attractive and should be carefully investigated.
Crowder Reb. (MPA et al.-RT-1) at 13 (34 Tr. 11652).
At the current time, however, little is known about the DOIS data9 and Dr. Smith
supplied virtually no documentation on the DOIS system and how and why the data are
collected. Indeed, he had precious little information on DOIS. Bradley Reb. (USPS-RT-
4) at 22-25 (34 Tr. 11575-11578); Crowder Reb. (MPA et al.-RT-1) at 13-15 (34 Tr.
11652-11654). It is clear that the OCA did not provide sufficient information on the
DOIS data to satisfy the Commission’s documentation requirement and support their
use for ratemaking purposes. Bradley Reb. (USPS-RT-4) at 21 (34 Tr. 11574).
9 For example, there is no information on the type of city carrier routes included in DOIS (e.g., letter, special purpose, or both), how the DOIS data collection has changed over time, DOIS standardization and quality control procedures, or how to interpret zero time or volume data for a zip-route-day not a holiday or Sunday. See Smith responses to ADVO/OCA-T3-33-35, 37-40 (22 Tr. 8157-8159, 8161-8164). Also, there is little information in the record concerning how DOIS route-day time and volume variables are now, and have been in the past, measured, collected, defined, standardized or handled for quality control. Id. at 47, 53 (22 Tr. 8171, 8181).
- 22 -
Separately, the DOIS data appear to have serious, perhaps fatal deficiencies:
• There is no differentiation between SPRs and large packages, which clearly are handled differently by city carriers.
• Priority Mail is a single volume variable within DOIS although it is, in fact, composed of mixed shapes.
• Collection volume data are lacking.
• Accountable volume data are lacking.
• There is no information on the quality of the data or on the quality controls applied to the data (both dependent and independent variables). In total, Dr. Smith’s own preliminary quality control procedures eliminated over 33% of the route-day observations and that does not bode well for the quality of the remaining data.10
These deficiencies mean that Dr. Smith’s DOIS models are incomplete, biased,
and unsuitable for developing marginal cost and volume variability estimates. The
missing volume data also mean that variabilities for these volumes cannot be estimated
from a DOIS model and that estimated coefficients for the included volume variables are
likely biased as a result. Crowder Reb. (MPA et al.-RT-1) at 15 (34 Tr. 11654); Bradley
Reb. (USPS-RT-4) at 25 (34 Tr. 11578). Consolidation of priority mail, large packages,
and SPRs into one variable also biases the model results. Crowder Reb. (MPA et al.-
RT-1) at 15 (34 Tr. 11654). Further, because the DOIS data include all city carrier
street time, variabilities derived from DOIS models are relative to this total time; but the
missing collection and accountable variables mean direct application of this procedure
is incorrect. Crowder Reb. (MPA et al.-RT-1) at 16 (34 Tr. 11655); Bradley Reb.
(USPS-RT-4) at 25-26 (34 Tr. 11578-11579). And, finally, there are potentially other
data quality/reliability issues that have not even been identified yet. For example, Dr.
10 This is calculated from Table 3 of Dr. Smith’s Direct (OCA-T-3) (0.334 = 1 - 492,097/739,396).
- 23 -
Smith’s handling of missing volume data may have created a well-known bias and
inconsistency problem called errors-in-variables.11 Crowder Reb. (MPA et al.-RT-1) at
16 (34 Tr. 11655).
Despite Dr. Smith’s development of models from the DOIS data and his
recommended DOIS model, the DOIS data are clearly “Not Ready For Prime Time.”
C. The PRC Should Not Give Any Evidentiary Weight To Dr. Smith’s Responses To POIR 25.
On November 17, just three days before rebuttal testimony was due, the
Commission issued POIR No. 25 to Dr. Smith. It solicited some correlation data from
Dr. Smith and asked him to run two Commission-specified CCSTS models. It also
asked his opinion on whether its new specified street time models were useful in dealing
with the problems associated with multicollinearity in the CCSTS data. In this POIR, the
Commission focused on only one of several possible treatments for the problem: the
option of dropping certain independent variables selected by the Commission from the
full quadratic regression equations. The POIR was clearly designed to provide the
Commission with support for its own specified models.
Dr. Smith answered the POIR on November 27—two days before the first day of
the rebuttal round of hearings. He included the two correlation matrices and results
from two new models requested by the Commission. In response to the Commission’s
11 When there was a volume or delivery point variable with no value, Dr. Smith set it to zero. Smith Responses to MPA/ANM/OCA-T3-6-10 (22 Tr. 8088-8091). This is an example of the well-known error-in-variables problem that afflicts models with random errors in explanatory variables. Specifically when there are such errors, estimates for coefficients (regressors) will be biased because of correlation between the observed (uncorrected) independent variable and the random error term explaining variations in the dependent variable. See Pindyck and Rubinfeld, Econometric Models and Economic Forecasts, Third Edition, McGraw-Hill, 1991, pages 159 –161.
- 24 -
questions concerning his opinion of the results, he stated that correlation exists between
the various pairs of independent variables and implied that the multicollinearity related
to the “SPR” variable and the SPR cross-product terms is the major source of
collinearity in the model. Smith Response of to POIR 25 at 2-9. He also stated that the
Commission’s requested CC3 model (a quadratic without the density and SPR cross-
product variables), given that one chooses not to use the three-bundle approach,
appears to be superior to witness Bradley’s model,“. . . being more of a full quadratic,
having more reasonable marginal cost relationships, and not being burdened with an
incorrectly specified density variable.” Smith Response to POIR 25 at 10-13.
On December 13, 2006—more than three weeks after the deadline for filing
rebuttal testimony, six days after the close of hearings on that testimony, and only eight
days before the due date for post-trial briefs—the Presiding Officer designated Dr.
Smith’s responses to POIR No. 25 for inclusion in the evidentiary record. Presiding
Officer’s Ruling No. R2006-1/124, Attachment A at 2. On December 18, 2006, ANM,
MPA and thirteen other intervenors filed a joint objection to the designation. On
December 19, the Presiding Officer certified the issue to the full Commission for
decision. Presiding Officer’s Ruling No. R2006-1/129. To date, however, the
Commission has not issued a decision on the merits of the issue. In the interim,
Presiding Officer’s Ruling No. R2006-1/129 requires the parties to submit briefs on the
assumption that the objection will not be sustained. Id. Accordingly, the following
analysis discusses the merits of Dr. Smith’s response to POIR 25, to the extent
permitted by the existing (and incomplete) record.12
Wholly apart from the due process issues noted above, Dr. Smith’s response to
POIR No. 25 should be given no weight because his analysis of the alternative models
12 Needless to say, our discussion of the issue is not intended as a waiver of the due process issues raised in the December 18 objection.
- 25 -
proposed by the Commission raises serious and unresolved methodological questions
that have not been properly investigated. Dr. Smith’s responses to POIR No. 25 (and
the Commission’s own POIR language) lead one to believe that the SPR and SPR
cross-product variables are the principal cause of the multicollinearity problems in the
CCSTS model and, therefore, dropping the Commission-selected independent variables
is an appropriate treatment of the problem in the CCSTS model. However, these are
not only debatable assumptions, but also very likely incorrect.
The unresolved issues raised by Dr. Smith’s responses begin with a key premise
of both POIR No. 25 and Dr. Smith’s responses: that the SPR and SPR cross-product
variables are the main cause of the multicollinearity problems in the CCSTS model.
Another disputable assumption underlying Dr. Smith’s responses is that dropping the
Commission-selected independent variables is an appropriate solution to
multicollinearity in the CCSTS model. As noted by Dr. Bradley, MPA et al. witness
Crowder—and elsewhere by Dr. Smith himself—dropping independent variables leads
to biased results, and therefore should be undertaken only after thorough analysis and
careful deliberation.13 Moreover, as noted by Dr. Bradley, even if the estimated
coefficients from a restricted model “appear” reasonable, multicollinearity still will affect
the results.14 Further, a better remedy for multicollinearity often is simply to add more
information to the model, a step that may or may not entail the use of more
observations.15 For ratemaking, this alternative is superior to simply dropping model
(footnote continued on next page)
13 See Smith response to POIR No. 25, Item 1; Crowder Reb. (MPA et al.-RT-1) at 7, lines 8-17 (34 Tr. 11646). See also Bradley Reb. (USPS-RT-4) at 13 n. 26 (34 Tr. 11566); and Docket No. R2005-1, Response of USPS Witness Bradley to POIR No. 9, Question 9 (filed July 6, 2005). 14 Bradley Reb. (USPS-RT-4) at 13 n. 26 (34 Tr. 11566). 15 William H. Greene, Econometric Analysis (5th ed. 2003) at 58-59. One way to add information would be the three-bundle approach, which witnesses Smith, Bradley and
- 26 -
variables because it can (1) significantly reduce variances in coefficient estimates and
(2) preserve non-bias or limit the amount of introduced bias in the estimates. Such an
approach should be seriously investigated.
Equally debatable is the assumption, clearly shared by POIR No. 25 and Dr.
Smith, that multicollinearity related to the SPR cross-product terms is the major source
of multicollinearity problems in the model. See Response of OCA to POIR No. 25 at 8-
9. The reported data certainly suggest otherwise:
• The SPR variables are not the most highly correlated with all other variables. Rather, the correlation coefficients reported by Dr. Smith suggest that the correlation of sequenced volume with the cross-product terms is the most widespread, followed by that for letters and small parcels, and then flats and collection volume. In the Equation 2 matrix, five of the correlation coefficients involving sequenced volume and the other terms are over 80 percent, followed by three each for letters and SPRs, and two each for flats and collection volume.16
• Correlation coefficients for volume variables and own volume-PD cross-product terms are over 80 percent in all cases (the five volume variables). This indicates that the possible deliveries (PD) variable by itself is also highly collinear with each of the volume variables. If so, distinguishing between the effects on modeled street time from each volume shape and PD variable may be difficult.17 This is an example of a system problem, not something isolated to the SPR variable.
Crowder all agree should be considered. See, e.g., 19 Tr. 6745 (Bradley). Other approaches would be using a data set that extends over a longer period of time, introducing more information in the form of coefficient restrictions and normalizing the model with respect to one of the variables. These are all acceptable approaches that avoid causing bias. 16 Kennedy suggests in his text that correlation coefficients over 80 percent indicate serious multicollinearity between any two independent variables. Peter Kennedy, A Guide to Econometrics 209 (5th ed. 2003). 17 Dr. Bradley noted this particular point in his response to POIR No. 4, Question 11. There, he explored the possibility of dropping the PD cross-product terms. See 19 Tr. 6748 (Bradley).
- 27 -
• In CC3A (the full quadratic with no density variable), the variance inflation factors (“VIFs”) for letter and flat volumes are higher than that for small parcels. Letter, flat, small parcel and sequenced volume VIFs are all over 10 percent, indicating harmful collinearity between the indicated variable and all other independent variables as a group.18 The comparable model requested by the Commission (a restricted quadratic without SPR cross-product terms or density variable) shows a lower VIF for the small parcel variable. However, based on the VIFs from the CC3A model and correlation coefficient results, there is an equal rationale to drop the cross-product terms for the other volume variables individually as well in order to test a range of results. This demonstrates that the SPR variables were singled out inappropriately.
These questions about the soundness of the alternative models suggested by the
Commission in POIR No. 25 are underscored by the equivocal tenor of Dr. Smith’s
comments on those models. In response to Question 1(b), which solicited Dr. Smith’s
views on the “relative merits of omitting or retaining” certain cross-product variables, Dr.
Smith stated that
the interaction terms are drivers of carrier time and should be retained if one is modeling delivery time as a function of the shapes. However, there has been some consideration of modeling the delivery process in terms of three major bundles – (1) DPS, (2) Cased Mail, and (3) Sequenced Mail.
OCA Reponses to POIR No. 25 at 7. In other words, to avoid biasing the results,
alternatives to simply dropping relevant variables should also be considered as
treatment for the multicollinearity issue.
Although Dr. Smith ultimately opined that “from an estimation viewpoint a strong
case can be made for the dropping of the ‘spr’ based cross-product terms,” he
acknowledged that “one might not then have a flexible functional form”; “the record
contains no discussion of the dropping of cross products as related to the estimation of
flexible functional forms”; “from an empirical viewpoint, the dropping of the ‘spr’ variable
appears to be inappropriate”; “collinearity and its resulting problems could be reduced
18 Kennedy, supra, at 213.
- 28 -
through the use of a data set extending over a longer period of time”; and “from a
theoretical viewpoint all cross products should be retained in the current model.” Id. at
8-9.
Similarly, while he asserted in response to Question 2(c) that the Commission’s
requested CC3 model without SPR cross-product terms “appears to be superior to
witness Bradley’s model, being more of a full quadratic, having more reasonable
marginal cost relationships, and not being burdened with an incorrectly specified density
variable,” id. at 12, he acknowledged that whether “the modified CC3A model would
apply in today’s environment, given the increased use of DPS mail . . . is not clear.” Id.
at 12-13.
Dr. Smith’s previous testimony in this case raises further questions about his
judgment regarding (1) the merits of deleting only the SPR cross-product terms as a
remedy for multicollinearity and (2) the reasonableness of the “marginal cost rela-
tionships.” Throughout his direct testimony, Dr. Smith severely criticized Dr. Bradley’s
CCSTS regular delivery model. He questioned both the reliability of the dataset and the
adequacy of the model. See, e.g., Smith Direct (OCA-T-3) at 3-8 and 15. He also
admitted, however, that he had not performed any independent review of the data or
developed an appropriate conceptual structure in which to determine the most important
cost causal variables and how these variables should be combined to explain city
carrier street costs. In his responses to ADVO/OCA-T3-15(e) and 24 (22 Tr. 8081,
8147), for example, he acknowledged that he had not developed either an appropriate
economic specification for a city delivery cost model or even a suggested approach to
the development of such a specification. Without those basic analytical steps, one has
- 29 -
no basis for knowing what are the reasonable marginal cost relationships or which
model is “superior.”19
Further, as noted above, Dr. Smith is mistaken in assuming that the
multicollinearity problem is due mainly to the SPR cross-product variables, and that
eliminating only the SPR cross-product variables makes the model coefficients superior.
As Dr. Bradley explained in his rebuttal testimony (USPS-RT-4) at 13 n. 26 (34 Tr.
11566)
Dr. Smith appears to think that if a version of the full quadratic model does not produce any negative variabilities, then he can ignore the fact that multicollinearity is a problem. (See, the discussion of models on pages 12-14 of OCA-T-3.) This is erroneous. A negative variability is an extreme symptom of multicollinearity, but it is not a necessary condition for its existence. The existence of substantial multicollinearity renders the estimated coefficients unreliable even if they are not negative. Thus, it is not appropriate to deal with multicollinearity by picking and choosing among the full quadratic results based upon whether or not one produces a negative coefficient.
The same criticism also applies to using assumed reasonable marginal cost
relationships to choose among restricted quadratic models. The multicollinearity
problem with the CCSTS data is a system problem (as clearly demonstrated in the
correlation matrices requested by the Commission). Without serious analysis of the
sources of the multicollinearity and or consideration of alternative specifications that add
19 Notably, he did not recommend a restricted quadratic, or independently develop any restricted quadratic models of his own, but simply presented in his direct testimony some models that were based on Dr. Bradley’s own restrictions. Smith Direct (OCA-T-3) at 9-14. Further, as noted by MPA et al. witness Crowder, Dr. Smith’s model recommendation was based on obtaining results that avoid “sign” problems and satisfy his a priori expectations. These are inappropriate criteria. Lacking the appropriate conceptual analysis, he cannot determine whether a model’s results comport with reality. Crowder Direct (MPA et al.-RT-1) at 6, lines 3-13; Bradley Reb. (USPS-RT-4) at 13 n. 26 (34 Tr. 11566-67).
- 30 -
more information, isolating one variable (SPR) for special treatment is a makeshift and
inappropriate remedy for the problem.20
Finally, Dr. Smith’s position that a density variable should be omitted from the
model (Response of OCA witness Smith to POIR No. 25 at 10-12) has been vigorously
challenged by the other two witnesses in the case who have testified on city carrier
costs. See Bradley Reb. (USPS-RT-4) at 1-9 (34 Tr. 11554-62) and Crowder Reb.
(MPA et al.-RT-1) at 8-10 (34 Tr. 11647-49). Under the circumstances, it would be
grossly unfair to enter into the record what amounts to additional testimony by Dr. Smith
on this issue, at a point when no opportunity remains for meaningful rebuttal by the
parties sponsoring Bradley or Crowder.
IV. COST CLASSIFICATION OF PAYMENTS INTO POSTAL SERVICE RETIREE HEALTH BENEFITS FUND UNDER SECTION 803 OF PAEA
The Postal Accountability and Enhancement Act of 2006 (“PAEA”) became law
on December 20, 2006, one day before the filing deadline of this brief. A complete
analysis of the ramifications of the Act for this case is not possible.
One title of the Act warrants discussion, however: Title VIII – Postal Service
Retirement and Health Benefits Funding. Section 802 of the Act relieves the Postal
Service of its obligation to make Civil Service Retirement System (“CSRS”) payments
for the next ten years. Section 804 relieves the Postal Service of its obligation to
escrow the savings resulting from enactment of P.L. 108-18. Section 803, however,
20 In Docket No. R2005-1, the Commission asked Dr. Bradley whether dropping only the SPR cross-product terms would be an improvement over dropping all cross-product terms, as he chose to do. He responded that he believed that “dropping all of the cross product terms”—the equivalent of his “restricted quadratic model”— would produce better results than the Commission’s alternative of dropping only selected independent variables. Response of USPS Witness Bradley to POIR No. 9, Question 11(d) (filed July 6, 2005). Bradley thus clearly understood that the multicollinearity was a system problem, not one principally caused by the SPR variable.
- 31 -
requires the Postal Service to make fixed annual payments over the next ten years into
the Postal Service Retiree Health Benefits Fund. The fixed payment is specified at $5.4
billion in the Test Year of this rate case.
Because the net impacts of Sections 802, 803 and 804 on Test Year costs are
largely offsetting, their primary effect is on cost attribution levels. Specifically, while the
Commission has treated the payment into the escrow fund as an institutional cost
(R2005-1 Op., ¶ 4028), the CSRS payments – roughly $1.7 billion in the Test Year
(USPS-T-6 at 21) – have been treated as attributable to the same extent that all labor
costs are attributable.
The Test Year payment into the Postal Service Retiree Health Benefits Fund
under PAEA, however, should be treated entirely as an institutional cost. First, the cost
is neither volume variable nor class specific. If mail volume were to change by a given
percentage or, for that matter, if all mail volume in a subclass of mail were to disappear
entirely, the Postal Service would still be required to make the entire amount of the
payment into the Postal Service Retiree Health Benefits Fund mandated by PAEA for
that year. Moreover, the Act bars the Postal Service from using the amounts paid into
this fund until September 30, 2016. This is as fixed and institutional as costs get.
In Docket No. R2005-1, the Commission characterized the possible disposition of
payments into the escrow fund as follows:
The escrow is unique in that it is an operating expense that is not currently related to the operations of the Postal Service. Therefore, it cannot be causally related to any functional cost of the Postal Service. At some future time, Congress could specify that the escrow funds be used to fund Retiree Health Premium costs. In that case, the costs would be attributable to all mail according to the weighted average attributable cost of all labor.
PRC Op. R2005-1 ¶ 4027. We respectfully submit that the reasoning of Paragraph
4027 should not govern here. Given the magnitude of the existing unfunded retiree
- 32 -
health obligation, payments into the Health Benefits Fund established by PAEA will, for
all practical purposes, be used to pay down an unfunded obligation that was incurred in
years past. The obligation was incurred for work performed in years past, to provide
postal services for mail was entered and delivered in years past, and is attributable to
those past services alone. Nothing the Postal Service or its customers do in the Test
Year—or any future year—can cause those costs to change. Because those costs are
not causally related to the provision of postal services in the Test Year or any future
year, they cannot be attributed to Test Year or future services.
V. THE COMMISSION SHOULD ADOPT THE IMPROVED COST AVOIDANCE METHODOLOGIES AND MODELS PROPOSED BY MPA/ANM WITNESS GLICK.
A. The benchmarks used by MPA/ANM to estimate the Carrier Route and nontransportation destination entry cost avoidances represent the mail that is most likely to convert to worksharing.
On numerous occasions, the Commission has endorsed setting worksharing
discounts according to the efficient component pricing (“ECP”) rule – the principle that
worksharing discounts should be set equal to the unit costs avoided by the worksharing
activity. In other words, discounts should be based on a 100 percent passthrough of
avoided costs. As discussed by the Commission, ECP enhances efficiency because it
encourages mailers to workshare only if they can perform work less expensively than
the Postal Service. See, e.g., MC95-1 Op. & Rec. Decis. ¶ 3074.
Critical to the success of ECP is the choice of an appropriate benchmark mail
category from which to estimate cost avoidances and apply discounts. As described by
the Commission, the benchmark should represent the category of mail most likely at the
margin to convert to worksharing or revert from the worksharing category. R2000-1 Op.
& Rec. Dec. ¶ 5089.
- 33 -
The benchmark category of mail from which USPS calculates the Carrier Route
Basic cost avoidance is 5-Digit Nonautomation flats. 7 Tr. 1687 (Response of USPS
witness Tang to MPA/USPS-T35-4(f)). Although using a nonautomation rate category
as the benchmark has intuitive appeal because there is no requirement to barcode
Carrier Route Basic flats, nonautomation flats are not the category of mail most likely to
convert to the Carrier Route Basic rate category, or the category of mail to which Carrier
Route mail is most likely to revert. That category of mail is automation flats. Glick
Direct (MPA/ANM-T-2) at 14-16.
Two major requirements must be fulfilled to qualify a piece for the Carrier Route
Basic rate. First, the piece must be in a bundle that includes at least six pieces for the
particular carrier. Second, the piece must meet the following address hygiene
requirement:
Carrier route rates require the accuracy of the carrier route codes and sequence of mailpieces within the carrier route be updated within 90 days of the date of mailing using a CASS certified address matching software program.
11 Tr. 2880-81 (response of USPS witness McCrery to MPA/USPS-T42-13(a)(ii)).
Only pieces that currently qualify for automation rates are likely to meet this
address hygiene requirement. Glick Direct (MPA/ANM-T-2) at 15. Apart from an
inability to place barcodes on a publication (11 Tr. 2880-81 (response of USPS witness
McCrery to MPA/USPS-T42-13(b))), the major reason for entering a periodical at
nonautomation, non-Carrier Route rates is an inability to meet the CASS address
hygiene requirement for automation rates:
Automation rates require that the accuracy of the ZIP+4 code and delivery point code information be updated within six months of the date (sic) of mailing using a Coding Accuracy Support System (CASS) certified address matching software program.
11 Tr. 2880-81 (response of USPS witness McCrery to MPA/USPS-T42-13(a)(iii)).
- 34 -
A piece that does not meet the address hygiene requirement to qualify for
automation rates is also unlikely to be able to meet the higher standard to qualify for
Carrier Route rates. 11 Tr. 2880-81 (response of USPS witness McCrery to
MPA/USPS-T42-13(c)).
Automation flats are also most similar to Carrier Route flats in terms of important
cost causing characteristics. USPS/MPA/ANM-T2-31 (30 Tr. 10355-10356). Although
compliance with address hygiene requirements is an important cost driver for Carrier
Route flats, the presence of a barcode is not, because Carrier Route flats only
infrequently receive machine sorts. Carrier Route pieces are only candidates to be
sorted on machines when they are in broken bundles, which is only about ten percent of
the time. 30 Tr. 10369; 33 Tr. 10996. Further, a large portion of incoming secondary
sorts – the sortation from 5-Digit to Carrier Route – are manual. Glick Direct
(MPA/ANM-T-2) at 17-19. See, e.g., 11 Tr. 2875.21
Similarly, the benchmark for estimating the nontransportation destination entry
cost avoidance should be OADC/OSCF-entered pieces because almost all non-
destination entered pieces22 are entered at one of these types of facilities. On the other
21 Mr. Miller’s suggestion that nonautomation flats should be the benchmark for Carrier Route flats because rates for nonautomation flats in sacks are bundle-based (33 Tr. 10993, footnote 2) should be ignored because (1) it is of no practical importance; and (2) it is based on a misunderstanding of sack preparation rules for Carrier Route flats. Specifically, 95 percent of Carrier Route flats are palletized and, as noted in Mr. Miller’s footnote, rates are bundle-based for all palletized flats. USPS-LR-L-91, Table 3. Further, more than 99 percent of sacked flats in Carrier Route bundles pay the Carrier Route rate. USPS-LR-L-91, Table 3. This is because, similar to bundle-based rates for sacked automation flats, sacking rules for Carrier Route flats have been designed to “preserv[e] carrier route rate eligibility whenever six or more pieces are sorted to a carrier route.” 71 Fed. Reg. 1977 (January 12, 2006). 22 In the context of Periodicals, “non-destination-entered pieces” refers to all pieces that are not entered at a DADC, DSCF, or DDU. Glick Direct (MPA/ANM-T-2) at 30.
- 35 -
hand, hardly any are entered at the DBMC, which is the USPS benchmark. Glick Direct
(MPA/ANM-T-2) at 30-32.
The Postal Service proposes to calculate the nontransportation (i.e., cross
docking) costs avoided by entering Periodicals at destination facilities (i.e., DADC,
DSCF, and DDU) relative to non-dropshipped Zones 1 & 2 Periodicals. Mayes Direct
(USPS-T-25) at 7; 5 Tr. 872 (response of USPS witness Mayes to MPA/USPS-T25-3).
This choice of benchmark understates the nontransportation cost savings from
destination entry. Glick Direct (MPA/ANM-T-2) at 30-32; 5 Tr. 872.
This is because most non-dropshipped Periodicals are entered in higher zones
and require more container handlings than the USPS cost avoidance model assumes.
Glick Direct (MPA/ANM-T-2) at 31. Further, not only are non-destination-entered
publications entered in all of the zones, they “convert” to destination entry from higher
zones as well. Id. Based upon the non-dropshipped Zones 1&2 benchmark, the USPS
cost avoidance methodology amounts to estimating the cost savings from entering
Periodicals containers at the DDU, DSCF, and DADC relative to containers entered at
the DBMC. Id. Contrary to this assumption, the vast majority of Zones 1-8 containers
are entered at Origin ADCs, SCFs, and AOs. Id. at 31-32 and Table 7. Based upon
this entry profile, Zones 1-8 containers are typically handled at one or more facilities
before ever reaching the DBMC (or equivalent facility). 11 Tr. 2878-79 (response of
USPS witness McCrery to MPA/USPS-T42-12(a)-(b)).
To match better the entry profile of Zones 1-8 containers, Mr. Glick proposes
changing the benchmark for estimating the nontransportation destination entry cost
avoidance to Periodicals Outside County mail entered at Origin ADCs and SCFs. Glick
Direct (MPA/ANM-T-2) at 32.
- 36 -
B. The Postal Service’s flats mail processing cost model (USPS-LR-L-43) understates presort cost avoidances in general and the Carrier Route presort cost avoidance in particular and thus should be rejected.
1. Two clearly wrong assumptions made by USPS witness Miller dramatically understate the Carrier Route cost avoidance.
USPS witness Miller makes two clearly wrong assumptions that cause the
Carrier Route cost avoidance to be significantly understated. First, the Postal Service’s
flats mail processing cost model significantly understates the volume of flats that will
receive manual incoming secondary sorts, which understates the average cost of an
incoming secondary sort and thus the cost difference between 5-Digit Automation and
Carrier Route flats. Glick Direct (MPA/ANM-T-2) at 18; 30 Tr. 10314, 10366.
Mr. Miller assumes that flats will receive a manual incoming secondary sort only
if (1) the destination facility does not have the appropriate sorter;23 (2) the flat has been
sorted manually in a previous sort; or (3) the flat is rejected by a flat sorter. Tr. 3/280.
This assumption directly contradicts the testimony of the Postal Service’s own
operational witness, Mr. McCrery, who explains multiple additional reasons why flats
sometimes receive a manual incoming secondary sort. 11 Tr. 3091-3092. Perhaps the
most important reason ignored by Mr. Miller is that “small volumes of flats for [entire]
destination[s] are processed manually when the volume is insufficient to justify the fixed
costs of setting-up and sweeping a scheme for such a small volume.” USPS-T-42 at
19. Mr. Miller’s assumptions are also inconsistent with past USPS modeling practice –
in past cases, the Postal Service included incoming secondary factors to reflect the
23 For an AFSM 100 flat, the appropriate sorter is either an AFSM 100 or a UFSM 1000 beause AFSM 100 flats can be processed on either machine. For a UFSM 1000 flat, the appropriate sorter is only a UFSM 1000 because these pieces cannot be processed on an AFSM 100.
- 37 -
operational realities described by Mr. McCrery. 3 Tr. 256; Glick Direct (MPA/ANM-T-2)
at 17-18.24
Further, the result of Mr. Miller’s model – that about twenty percent of Periodicals
Outside County incoming secondary sorts are manual (3 Tr. 261) – is clearly
inconsistent with Mr. McCrery’s estimate that 44.7% of incoming secondary sorts are
manual and that (11 Tr. 2853, 2888)25 and McCrery’s assessment that the figure may
be even higher for Periodicals (11 Tr. 2875).26
24 In an interrogatory response, Mr. Miller provides several reasons why he excluded incoming secondary factors from the USPS cost model. Tr. 3/256, 262. Mr. Glick and Mr. Stralberg, however, explain in detail why Mr. Miller’s reasoning is flawed and his decision to exclude these factors reduces the accuracy of the USPS cost model. 30 Tr. 10312-10314; 31 Tr. 10581-10584. 25 Mr. McCrery describes the 44.7 percent figure as the percentage “of incoming secondary flats that were finalized in manual operations in the field.” 11 Tr. 2853. McCrery’s confirmation of how he calculated the 44.7 percentage figure, however, makes clear that the figure refers to the percentage of incoming secondary sorts that were manual. 11 Tr. 2888. Thus, for clarity, this brief refers to the 44.7 percent figure as the percentage of incoming secondary sorts that were manual. Similarly, this brief refers to the manual finalization rate presented by Mr. Miller in his rebuttal testimony (33 Tr. 11023) as the percentage of incoming secondary sorts that were manual. Regardless, manual finalization rates and the percentage of incoming secondary sorts that are manual are relatively similar. MPA/ANM-LR-2, worksheet “MANUAL IS CALC.” 26 Contrary to Mr. McCrery’s testimony, Mr. Miller attempts to suggest that the percentage of Periodicals Outside County incoming secondary sorts that are manual is lower than average because First-Class Mail single-piece flats would be expected to be manually sorted at an above-average rate. 33 Tr. 10998. This argument should be ignored because In-Office Cost System (IOCS) data indicate that First-Class Mail flats (which are primarily single-piece) are processed on automation at an above-average rate and thus are processed manually at a below-average rate. 30 Tr. 10361-10362. If anything, the lower than average rate of manual processing for First-Class Mail flats provides further support for Mr. McCrery’s belief that the rate of manual processing for periodicals is above average.
- 38 -
The result of Mr. Miller’s model is also inconsistent with Mr. Miller’s own
calculation (provided in his rebuttal testimony)27 that approximately 31 percent of
incoming secondary sorts are manual. 33 Tr. 11023. Further, during cross
examination, Mr. Miller’s 31 percent figure was shown to be an underestimate because
his calculation of the total number of incoming secondary sorts28 (which he used to
make a downward adjustment to McCrery’s estimate of the number of manual incoming
secondary sorts)29 ignored a large portion of Carrier Route volumes that required
incoming secondary sorts (33 Tr. 11044-11045) and the fact that some pieces (e.g.,
those that are rejected) require multiple sorts (33 Tr. 11055-11057). In fact, just adding
the portion of Standard Mail Commercial ECR Basic flats that requires incoming
secondary sorts into Mr. Miller’s calculations increases the manual percentage to 34%.
33 Tr. 11053.
Second, by classifying the Flats Preparation cost pool as fixed, the Postal
Service’s flats mail processing cost model implicitly assumes that Carrier Route and
non-Carrier Route flats will incur the exact same unit mail processing cost for preparing
flats for processing. 33 Tr. 11003. This is clearly inappropriate. As even Mr. Miller
admits, Carrier Route flats incur less of these costs than do Non-Carrier Route flats. 33
27 Mr. Miller dedicated a section of his rebuttal testimony (31 Tr. 10994-10999) to rebutting McCrery’s estimate that 44.7 percent of incoming secondary sorts are manual. The calculations he provided to rebut McCrery, which suggest (incorrectly) that 31 percent of incoming secondary sorts are manual (33 Tr. 11023), are distinct from and should not be confused with the manual percentage that results from the USPS flats mail processing cost model. 3 Tr. 261. 28 Mr. Miller refers to this figure as the “incoming secondary candidate RPW volume estimate.” 33 Tr. 10996. 29 33 Tr. 10996-10997. According to Mr. Miller’s method for adjusting Mr. McCrery’s estimate, a lower number for the total number of incoming secondary sorts results in a lower number of manual incoming secondary sorts, which (in turn) results in a lower estimate of the percentage of incoming secondary sorts that are manual.
- 39 -
Tr. 11059. The reason is that carriers, not clerks or mailhandlers, generally prepare
Carrier Route flats for processing. 11 Tr. 2877, 2886. Because all carrier costs for
Carrier Route (as well as all other) flats have been included in USPS witness Kelley’s
unit delivery cost estimates, assigning mail processing costs for these activities to
Carrier Route flats is an unfair double charge. Glick Direct (MPA/ANM-T-2) at 23.30
Also, while the flat preparation cost difference between non-Carrier Route flats
and Carrier Route flats may shrink somewhat by the Test Year due to the deployment of
the AFSM 100 – Automation Induction system (AFSM-ai) (33 Tr. 11004-11005), flats
preparation costs, and thus the flats preparation cost difference between Carrier Route
and non-Carrier Route flats, still will be significant. 30 Tr. 10321-22. Otherwise, the
Postal Service should have reduced its revenue requirement to a much greater extent to
reflect a much larger cost reduction from the AFSM-ai deployment.31
2. In general, Mr. Miller understates flats presort cost avoidances by excluding some costs that are incurred sorting flats and bundles as well as costs that, under USPS costing methods, have been found to vary proportionally with these costs, from his proportional CRA adjustment.
The Postal Service uses a hybrid method to estimate mail processing costs by
rate category (and thus to determine cost avoidances between rate categories). With
this method, the Postal Service first estimates costs by rate category using an
engineering mail flow model. To tie the mail processing costs from the mail flow model
30 To attempt to rebut Mr. Glick on this point, Mr. Miller attacks the proxy – Standard Mail delivery costs – that the Postal Service uses to estimate Periodicals unit delivery costs. 33 Tr. 11003-11004, footnote 31. Given that Mr. Miller is not an expert on delivery costs, Mr. Miller’s criticism of the delivery cost proxy should be ignored. 31 The total projected savings from the AFSM-ai deployment (which will affect costs in the AFSM 100 cost pool as well as costs in the flat preparation cost pool) is $88 million, which represents a little more than twenty percent of the $400 million Base Year cost in the flat preparation cost pool. 30 Tr. 10321.
- 40 -
back to mail processing costs generated by the Cost and Revenue Analysis (“CRA”),
witness Miller (USPS-T-20 at 6) performs a “CRA adjustment.” The first step in this
process is to classify each CRA cost pool as either proportional or fixed. The second
step is to develop a ratio of (1) the CRA unit cost for Periodicals Outside County flats in
proportional cost pools to (2) the weighted average modeled cost from the mail flow
models. The ratio is then multiplied by the modeled costs by rate category. Finally, the
unit cost in the fixed CRA cost pools is added to the cost for each rate category. See
USPS-LR-L-43, worksheet “CRA ADJ UNIT COSTS.”
Mr. Miller defines proportional cost pools as follows:
The proportional cost pools contain the costs for piece or bundle distribution operations that have actually been modeled. The flat sorting machine (“AFSM100”) cost pool is an example of a proportional cost pool.
USPS-T-20 at 6.
In other words, costs that according to the In-Office Cost System (IOCS)32 are
incurred sorting flats or bundles of flats should be treated as proportional. Miller,
however, classifies many costs that meet this definition as fixed. This inappropriately
understates the proportional CRA adjustment, which (in turn) understates presort cost
avoidance estimates.
First, despite Mr. McCrery’s confirmation that Periodicals Outside County costs
for flats in letter sorting, parcel sorting, Priority Mail sorting, and Express Mail cost pools
are “most likely incurred sorting” flats or bundles of flats (11 Tr. 2885), Mr. Miller
classifies these pools as fixed. Second, despite the logical conclusion that almost all
costs that IOCS classifies as Periodicals Outside County parcels are actually incurred
32 IOCS is the CRA costing system for mail processing costs.
- 41 -
for handling flats, Miller does not include these “parcel” costs in his CRA adjustment.
Glick Direct (MPA/ANM-T-2) at 20.
Specifically, the only reasonable explanation for the $26 unit cost for Periodicals
Outside County parcels is that some costs for flats are being recorded by IOCS as
parcel costs. Glick Direct (MPA/ANM-T-2) at 20-21. This is because IOCS classifies
costs for flats over ¾ inch thick (“UFSM 1000 flats” (USPS-LR-L-33 at 5)) as parcel
costs. USPS-T-13 at 34; 13 Tr. 3635-3636, 3629-3631. According to the Periodicals
mail characteristics study, UFSM 1000 flats comprise nearly ten percent of Periodicals
Outside County volume. USPS-LR-L-91, Table 3. “True” parcels, on the other hand,
comprise only .02% of subclass volume. 14 Tr. 4270-4271; Glick Direct (MPA/ANM-T-
2) at 21. Third, Mr. Miller treats the entire Non-MODS Allied cost pool as fixed despite
the fact that 37 percent of the costs in the pool are for bundle sorting. Glick Direct
(MPA/ANM-T-2) at 22; 10 Tr. 2474-75.33
Finally, according to Postal Service costing methods, the costs in the Mail
Processing Support (“1SUPPF1”) cost pool are estimated to vary in proportion to the
33 In his rebuttal testimony, Mr. Miller criticizes Mr. Glick’s proposal to disaggregate the Non-MODS Allied cost pool as one-sided because Mr. Glick does not treat container-handling costs in opening unit cost pools as fixed. 33 Tr. 11005-11007. Mr. Miller’s criticism is erroneous because it is based upon an inappropriately narrow interpretation of the activities entailed in sorting bundles. Specifically, Mr. Miller’s interpretation is at odds with the definition of the manual bundle sorting operation that was used in developing the productivities used by Mr. Miller in USPS-LR-L-43. As explained in Docket No. R2000-1, USPS-LR-I-88, the manual bundle sorting operation includes associated container-handling costs. 30 Tr. 10328. Further, the broad definition of manual bundle sorting operations used in the study is consistent with the Postal Service’s definition of the activities performed in opening units. 30 Tr. 10329. Further, as discussed by Mr. Stralberg, the Postal Service’s estimate that 37 percent of the Non-MODS Allied cost pool is incurred sorting bundles may be an understatement. This is because, as discussed above, a large portion of bundle sorting costs are actually incurred handling containers. It is unclear whether these container handling costs were included in the 37 percent estimate provided by the Postal Service or not. 31 Tr. 10594.
- 42 -
direct costs in supported cost pools. USPS-T-11 at 19; 11 Tr. 2454-55. The distribution
approach in USPS-LR-L-43 is completely at odds with this approach because it treats
the costs in this cost pool as fixed across rate categories despite the significant
difference in modeled costs by rate category in the supported cost pools. 33 Tr. 11007.
USPS-LR-L-43, PER OC Flats.xls, worksheet “CRA ADJ UNIT COSTS.”34
In sum, Mr. Miller criticizes Mr. Glick for recommending “completely one-sided”
changes to the Postal Service cost model. 33 Tr. 10994. Based upon the discussion
above, MPA and ANM believe that Mr. Glick is, in fact, correcting the Postal Service’s
“completely one-sided” cost model that is based upon a series of assumptions that
systematically understate cost avoidances in general and the Carrier Route cost
avoidance in particular.
C. The MPA/ANM flats mail processing cost model (MPA/ANM-LR-2) should be adopted because it is more accurate than the USPS model (USPS-LR-L-43).
Unlike USPS-LR-L-43, MPA/ANM-LR-2 appropriately includes incoming
secondary machine/manual factors. Glick Direct (MPA/ANM-T-2) at 19. The use of
incoming secondary machine/manual factors is consistent with past practice. In the last
two fully litigated rate cases (Docket Nos. R97-1 and R2001-1) as well as Docket No.
R2001-1 (the last rate case that was not based upon an “across-the-board” proposal),
the Postal Service itself appropriately included incoming secondary factors in the cost
models. R2000-1, USPS-LR-I-90, R2000_1_Flats Cost Model_Final USPS.xls,
34 In his rebuttal to Mr. Glick’s recommended treatment of the 1SUPPF1 cost pool, Mr. Miller states, “the fact that a cost pool could be impacted by worksharing does not necessarily mean that those costs would vary by rate category such that the cost pool should be classified as proportional.” 33 Tr. 11007. While it is hard to argue with this very general statement, it has no applicability in this particular instance. The Postal Service’s costing method is clear that costs in the 1SUPPF1 cost pool indeed do vary with costs in the supported pools. Further, costs in the supported pools (e.g., the sorting cost pools) do in fact vary by rate category.
- 43 -
Worksheet ‘Data’, Items (12)-(14); R97-1, USPS-LR-H-134, 2CREG.xls, Worksheet
‘inputs’, cell E27; 30 Tr. 10368. In fact, the Postal Service’s incoming secondary
manual factor in Docket No. R2001-1 was 35 percent, 30 Tr. 10368, much higher than
the 20 percent figure Mr. Glick proposes in this case. Glick Direct (MPA/ANM-T-2) at
19.
Based upon the appropriate use of incoming secondary machine/manual factors,
MPA/ANM-LR-2 produces a reasonable, indeed conservative, estimate (36%) of the
percentage of incoming secondary sorts that are sorted manually. Glick Direct
(MPA/ANM-T-2) at 19. This estimate is certainly conservative compared to Mr.
McCrery’s estimation that, on a systemwide basis, approximately 44.7% of incoming
secondary sorts are manual (11 Tr. 2853) and that the figure is likely to be higher for
Periodicals (11 Tr. 2875). This figure is also quite reasonable even compared to the
systemwide percentage (31%) estimated by Mr. Miller in his rebuttal testimony. 33 Tr.
11023. Further, as discussed above in Section V.B.1, Mr. Miller’s rebuttal estimate of
the systemwide average clearly understates the “true” average.35
Mr. Miller’s rebuttal attack on Mr. McCrery’s estimate of the share of incoming
secondary sorts that are manual constitutes improper rebuttal, and should be given no
weight by the Commission. 33 Tr. 11032-11034. It is a basic principle of due process
that a party cannot use a rebuttal witness to attack its own direct testimony – especially
after adverse parties have relied on that direct testimony in formulating their reply
position.
35 Casting further doubt on the accuracy of Mr. Miller’s estimate of the percentage of incoming secondary sorts that are manual is that Mr. Miller’s estimate of the number of manual incoming secondary sorts is only about half of the number estimated by a data system – FLASH – that USPS uses for management purposes. 33 Tr. 11036-11044; 11 Tr. 2888.
- 44 -
In any event, the credibility of Mr. Miller’s position is undermined because it
conflicts with the Postal Service’s own operational management. Mr. McCrery’s figures
were calculated from USPS data systems that are used by the Postal Service for the
purpose of evaluating flats performance and running the business, and thus are values
relied on by Postal Service management to make decisions in the ordinary course of
business. 33 Tr. 10994, 11035-11044; 11 Tr. 2888.36
Moreover, comparing the results of cost models to operational reality is an
appropriate way to determine the accuracy of intermediate cost model inputs. Cf. Miller
(33 Tr. 10993-10994). This is particularly true when, as for incoming secondary factors,
there are no data on the intermediate inputs. 30 Tr. 10363-10364. In the absence of
such data, the best approach is to use informed judgment and check the results against
the best available data (as Mr. Glick did) to develop reasonable estimates. 30 Tr.
10315. This approach is certainly better than Mr. Miller’s approach – basing his model
on the clearly wrong implicit assumption that the manual incoming secondary factor is
zero. 30 Tr. 10368.
Contrary to the position of USPS witness Miller (see, e.g., 33 Tr. 10999),
improvements to the Periodicals Outside County cost model need not wait for a full
review of the mail processing cost models for other shapes and mail classes. Having
said that, improvements similar to those proposed by Mr. Glick would most likely be
appropriate for other mail processing cost models. 30 Tr. 10327.
36 In his rebuttal testimony, Mr. Miller attempts to downplay this fact by stating, “While these data are appropriate for measuring performance, they are not appropriate for cost modeling purposes.” 33 Tr. 10994. This statement is clearly incorrect. Mr. Miller’s primary criticism of the use of the 44.7 percent figure provided by Mr. McCrery is that “THE ORIGINAL FINALIZATION RATES ARE NOT ACCURATE.” 30 Tr. 10995-10997, 11023. This criticism is equally applicable to the use of the data for performance measurement purposes.
- 45 -
D. MPA/ANM-LR-2 appropriately reflects mail processing cost differences between non-Carrier Route flats and Carrier Route flats in the flat preparation cost pool.
As discussed above, even Mr. Miller agrees that Carrier Route flats currently
incur less mail processing costs for flat preparation than do non-Carrier Route flats 33
Tr. 11059. Mr. Glick’s approach – distributing half of the flat preparation cost pool only
to non-Carrier Route flats (Glick Direct (MPA/ANM-T-2) at 22) – is a reasonable (if not
conservative) way to reflect this cost difference. First, consistent with Mr. Miller’s belief
that there is not a significant flat preparation cost difference between non-Carrier Route
flat in different non-Carrier Route rate categories (33 Tr. 11003), it assumes that all non-
Carrier Route flats will incur the same flat preparation cost. Second, although Mr. Miller
criticizes Mr. Glick’s decision to distribute fifty percent of the cost pool to non-carrier
route flats as “arbitrary” (33 Tr. 11003-11004), Mr. Glick explains that he judgmentally
selected this figure out of conservatism. 30 Tr. 10369. This explanation is supported by
the facts that only ten percent of Carrier Route bundles break, 30 Tr. 10321, and the
remaining 90 percent of Carrier Route bundles will generally be prepared by carriers,
not by mailhandlers, and thus their costs are already reflected in Mr. Kelley’s unit
delivery cost estimate (Glick Direct (MPA/ANM-T-2) at 22-23).
The results of Mr. Glick’s approach to distributing flat preparation costs also
appear to be conservative in comparison to estimates provided by another witness
based upon a detailed analysis of flat preparation costs. The MPA/ANM-LR-2 flat
preparation cost difference between 5-digit flats and carrier route flats is significantly
less than that estimated by Time Warner witness Stralberg. 30 Tr. 10319.
E. MPA/ANM-LR-2 appropriately treats all costs that are incurred sorting Periodicals Outside County flats and bundles as proportional costs.
As described in Section V.B.2. above, costs incurred sorting flats or bundles of
flats should be treated as “proportional” costs for the purpose of the CRA adjustment.
- 46 -
MPA/ANM-LR-2 includes several improvements to the USPS-proposed CRA
adjustment process to bring the cost model better in line with this definition.
First, consistent with Mr. McCrery’s testimony and Mr. Miller’s definition of
proportional cost pools, MPA/ANM-LR-2 classifies letter, parcel, Priority Mail, and
Express Mail sorting cost pools as proportional. Glick Direct (MPA/ANM-T-2) at 21; 11
Tr. 2885. Second, because the vast majority of costs that IOCS identifies as
Periodicals Outside County parcel costs are likely to be costs for handling UFSM 1000
flats, MPA/ANM-LR-2 appropriately uses the combined unit cost for all Periodicals
Outside County nonletters in determining the CRA adjustment. Glick Direct (MPA/ANM-
T-2) at 20-21. Third, consistent with Mr. Miller’s definition of a proportional cost pool,
MPA/ANM-LR-2 treats the costs for bundle sorting activities in the Non-MODS allied
cost pool as proportional costs. Glick Direct (MPA/ANM-T-2) at 22. Fourth, consistent
with the Postal Service’s method for distributing support costs to subclass of mail,
MPA/ANM-LR-2 treats a portion of support costs as proportional costs. Glick Direct
(MPA/ANM-T-2) at 22; 30 Tr. 10333.
F. MPA/ANM’s estimates of the nontransportation destination entry cost model (Glick Direct (MPA/ANM-T-2) at 34) should be used to estimate the nontransportation cost avoided through destination entry.
As discussed above, the MPA/ANM model is more accurate than the Postal
Service’s model because it is based upon a more accurate benchmark – OSCF/OADC-
entered containers (as compared to the Postal Service’s DBMC benchmark). Glick
Direct (MPA/ANM-T-2) at 30-33. Despite the change in benchmark, MPA/ANM notes
that Mr. Glick’s cost avoidance estimate is still conservative because it assumes that
OSCF/OADC-entered containers are handled at only one facility before arriving at the
DBMC. Glick Direct (MPA/ANM-T-2) at 33. In reality, they frequently will be handled at
multiple facilities before reaching the DBMC. Glick Direct (MPA/ANM-T-2) at 33,
footnotes 22 and 23. 11 Tr. 2878-2879; Tr. 18-C/6288.
- 47 -
G. Witness Glick’s estimates of the cost of pallets and sacks and the unit cost avoided through palletization should be used because they include container-handling costs and cost differences at both destination and non-destination facilities.
The Postal Service’s Periodicals Container Cost Analysis (USPS-LR-L-85) model
should not be used for evaluating the cost difference between sacks and pallets
because it includes only cost differences at destination facilities. 13 Tr. 3611-3612.
Instead, the Commission should use Mr. Glick’s estimate because it appropriately
includes cost differences at both destination and non-destination facilities. Glick Direct
(MPA/ANM-T-2) at 25-28. MPA and ANM would note that, despite including cost
differences at both types of facilities, Mr. Glick’s estimate is still conservative for three
reasons. Glick Direct (MPA/ANM-T-2) at 27-28. First, it completely ignores the
significant savings that result from the higher bundle integrity of palletized flats. Glick
Direct (MPA/ANM-T-2) at 28. Second, when estimating the per-piece container
handling cost for periodicals is sacks, it assumes that sacks contain an average of 45
pieces. Witness Tang, on the other hand, estimates that they contain an average of 42.
7 Tr. 1703; Glick Direct (MPA/ANM-T-2) at 27. Using Ms. Tang’s lower estimate of the
number of pieces per sack would have resulted in a higher per-piece sack-handling cost
estimate and thus a larger per-piece cost difference between Periodicals Outside
County flats entered in sacks and those entered on pallets. Third, Mr. Glick’s estimate
assumes that sacks will be handled at only two non-destination facilities while USPS
estimates this figure to be higher – in the 2.2 to 2.4 range. Glick Direct (MPA/ANM-T-2)
at 27; 18C Tr. 6291.
H. Witness Glick’s estimate of the 5-Digit pallet cost avoidance is reasonable.
Mr. Glick bases his 5-Digit pallet discount as the cost savings from avoiding one
bundle sort. Glick Direct (MPA/ANM-T-2) at 29. This is a conservative estimate
because it completely ignores the cost savings from the fact that 5-Digit pallets only
- 48 -
require a crossdock at the destination SCF. Glick Direct (MPA/ANM-T-2) at 29-30.
Further, as discussed above, requiring that 5-Digit pallets be entered at DDUs or
DSCFs and contain at least 250 pounds of mail will ensure that these pallets do not
impose any offsetting container-handling costs. 34 Tr. 11460-11461.
VI. RATE DESIGN FOR PERIODICALS OUTSIDE COUNTY MAIL
A. The Governing Ratemaking Standards
The Commission has recognized that the encouragement of efficient mailer
behavior is an important goal of Periodicals rate design. In Docket No. C2004-1, for
example, the Commission stated:
Cost-based rates and encouraging efficiency in postal operations have been two frequent hallmarks of Commission rate recommendations. The Commission has recommended – and the Postal Service has adopted – rates that move all classes and subclasses of mail toward more efficient preparation through discounts for presorting, dropshipping, palletizing, and other cost-efficient measures. The Commission continues to support an economically efficient approach that encourages the optimal use of society’s resources to process and deliver mail to its ultimate consumers.
Docket No. C2004-1, Complaint of Time Warner Inc. et al. Concerning Periodicals
Rates, Order No. 1446 (Oct. 21, 2005) at 45.
These economic concerns, the Commission noted, also have a legal dimension.
The sixth ratemaking factor of the Postal Reorganization Act directs the Commission to
give weight to “the degree of preparation of mail for delivery into the postal system
performed by the mailer and its effect upon reducing costs to the Postal Service.” Id.
(quoting 39 U.S.C. § 3622(b)(6)).
In Docket No. C2004-1, the Commission found that the complainants had
“presented a well-documented case that there are significant inefficiencies in the
Periodicals rate structure, and that some structural and administrative improvements
are appropriate and desirable.” Order No. 1446 at 4. “[P]rogress toward a more cost-
- 49 -
based structure is both possible and necessary to increase efficiencies in the
Periodicals rates.” Id. at 6.
The Commission also found, however, that changes in rate design as far-
reaching as those proposed by the complainants “would have substantial adverse
impact on thousands of small publications.” Id. at 4 & 6. Many smaller-circulation
publishers, the Commission suggested, lacked the minimum address density needed to
qualify for discounts for dropshipping or palletizing. Id., ¶¶ 5008, 5011. This, the
Commission found, was a relevant concern under the Act.37 “[E]ncouraging enhanced
efficiency – while a significant concern in devising a rate structure – is only one of many
considerations that enter into this task. The nature and extent of the potential impact on
rates and service are also important factors.” Id. at 4.
For this reason, the Commission did not issue a recommended decision or
otherwise order any changes in existing Periodicals rates. Instead, the Commission
contented itself with suggestions for improving the Periodicals rate design in future rate
cases, id., ¶¶ 5001-5012, and a general admonition to “update these [cost] analyses,
evaluate the impact of potential rate changes on various categories of Periodicals mail,
and suggest incremental changes that will foster efficient mailer practices without undue
disruption” or an “unreasonabl[e] impact” on “any segment of [Periodical] class” mail.
Id. at 5, 6, 41 and 45. The Commission left the choice of a specific “path for improving
37 The fourth and fifth ratemaking factors of the Act, codified at 39 U.S.C. § 3622(b)(4) and (5), direct the Commission to consider: “the effect of rate increases upon the general public, business mail users, and enterprises in the private sector of the economy engaged in the delivery of mail matter other than letters,” and “the available alternative means of sending and receiving letters and other mail matter at reasonable costs.” For periodicals, Section 3622(b)(8), commonly known as the “ECSI” factor, directs the Commission to consider “the educational, cultural, scientific, and informational value to the recipient of mail matter.”
- 50 -
the efficiency of Periodicals” to the judgment of the Postal Service in the first instance.
Id. at 6.
These standards have direct application to the appropriate rate design for
Periodicals Outside County Mail. Efficient mailing practices, such as commingling and
dropshipping periodicals and preparing mail on efficient 5-Digit pallets, are critical to
controlling Periodicals Outside County costs and thus should be encouraged. Although
the Commission should temper rates to mitigate rate shock, significant movement
towards greater cost recognition and incentives for efficiency is essential.
B. The Commission should reject the Postal Service’s proposed rate design for Periodicals Outside County Mail.
1. The USPS proposal only modestly increases existing incentives to commingle and dropship periodicals.
The Postal Service’s proposed rate structure does not go far in increasing
worksharing incentives, as reflected by a variety of different benchmarks and scenarios.
For example, as Table 1 shows, the Postal Service’s proposed rates would offer only
modest increases in the existing postage incentives to co-mail publications. As the
following table shows, the proposed increases in incentives would be similar to the
subclass average rate increase, and well below the increase in incentives offered by the
MPA/ANM proposal:
- 51 -
Table 1 Percentage Increase in Postage Incentive
to Co-Mail Publication
Publication Number of Pieces/Issue
USPS Proposal
MPA/ANM Proposal
Farm Collector 38,036 8.6% 19.3%
Gas Engine 15,192 13.7% 21.7%
Harper’s 155,472 11.8% 25.9%
Herb Companion 23,632 12.7% 22.4%
Interweave Knits 33,637 14.7% 24.2%
Mother Earth News 217,676 9.1% 20.2%
Natural Home and Garden 27,760 13.1% 25.2%
Source: Library Reference MPA/ANM-LR-4; Glick Direct (MPA/ANM-T-2) at 10.
Further, the USPS rate design offers an even smaller increase in the incentive to
co-palletize periodicals. Table 2 below simulates the impact of co-palletization on the
publications shown in Table 1. The “Co-pal” billing determinants reflected in Table 2
assume that bundle presort for these publications is the same as if mailed solo and that
two-thirds of the dropshipped pieces are entered at the DADC and the other one-third
are entered at the DSCF.38 Table 2 also shows that the percentage rate increases
under the USPS proposal often would be similar to or higher under the Co-pal scenario
than if those publications had chosen not to co-palletize and dropship their
publications.39 Table 2 also shows that the incentives offered by the USPS proposal to
co-palletize and dropship increase much less than under the MPA/ANM proposal.
(footnote continued on next page)
38 As Mr. Glick explained, the assumption that two-thirds of the dropshipped pieces are entered at the DADC stems from the fact that, in FY 2005, approximately 80 million pieces qualified for the DADC copalletization experimental discount and about 40 million at the DSCF. 30 Tr. 10405. Indeed, Mr. Glick testified that “the part that’s comailed is getting more to the SCF and so probably the percentage of pure copal that’s only getting to the DADC is probably higher than two-thirds.” Id. 39 The same is true for comail. Specifically, the percentage rate increases under the USPS proposal often would be similar to or higher under the Comail scenario than if
- 52 -
Table 2 (Simulated Co-Palletization Scenario)
% Rate Increase (USPS) % Increase in Incentive
Publication Number of
Pieces/Issue Solo Copal USPS MPA/ANM Farm Collector 38,036 11.6% 13.2% 2.0% 20.1%
Gas Engine 15,192 13.2% 13.8% 10.7% 26.9%
Harper's 155,472 10.6% 11.1% 6.5% 32.7%
Herb Companion 23,632 15.5% 17.0% 7.5% 34.2%
Interweave Knits 33,637 11.5% 11.3% 12.2% 26.3%
Mother Earth News 217,676 10.0% 10.7% 6.3% 19.7%
Natural Home and Garden 27,760 12.8% 13.7% 8.4% 30.1%
Source: ABM/MPA/ANM-T2-33 (30 Tr. 10281-82).
The Postal Service did introduce new rate elements focused on improving
containerization and entry practices -- a container charge and editorial dropship
discounts. Tang Direct (USPS-T-35) at 4-8. As evidenced by the marginal increases in
incentives to commingle and dropship, however, the incentive provided by these new
elements was largely offset by the elimination of pallet and co-pallet discounts. See
Tang Direct (USPS-T-35) at 11 (discussing USPS proposal to eliminate these
discounts).
The paltry increases in incentives proposed by the Postal Service are
unacceptable. As noted above, the Commission specifically found in Docket No.
C2004-1 that “significant inefficiencies” exist in the Periodicals rate structure, and that
“progress toward a more cost-based structure is both possible and necessary.” Order
No 1446 at 4 & 6. Moreover, the Postal Service itself acknowledged in this case that
improved containerization and increasing dropshipping, both of which result from
commingling, must be high priorities in this case. Tang Direct (USPS-T-35) at 8.
Further, current discounts in these areas are based upon excessively low passthroughs.
30 Tr. 10347 (Glick response to USPS/MPA/ANM-T2-22); 30 Tr. 10362-10363 (Glick
those publications had chosen not to comail and dropship their publications. MPA/ANM-T-2 (Glick Direct) at 9, Table 2.
- 53 -
response to MH/MPA/ANM-T2-1); Cohen Direct (MPA/ANM-T-1) at 15. The halting
change embodied in the Postal Service’s actual rate proposals does not begin to
represent adequate movement toward these goals
It is quite possible that the modesty of the increases proposed by the Postal
Service in incentives to commingle and dropship periodicals was at least partially the
result of an analytical error on Ms. Tang’s part. USPS witness Tang, in analyzing the
impact of the USPS proposal on co-pal incentives, failed to adjust for the impact of the
24-piece sack minimum rule. Second Revised Response to MPA/USPS-T35-13 (7 Tr.
1700). See also response to MPA/USPS-T35-28 (7 Tr. 1716) and Revised Response to
MPA/USPS-T35-15 (7 Tr. 1702). Specifically, her original analysis of the incentive to
co-palletize provided by the USPS proposal, which presumably influenced the USPS
proposal,40 suggested that the proposal would increase the incentive for many
publications to co-palletize by more than 30 percent. Original response to MPA/USPS-
T35-13 (not on record). It is quite likely that the Postal Service would have proposed
more far-reaching improvements in incentives if it had realized how modest its rate
design proposal was before it hardened into the Service’s formal Request to the
Commission.
2. The USPS proposal would create a disincentive to the efficient preparation of periodicals on 5-Digit pallets.
Preparation of mail on 5-digit pallets is a valuable and beneficial form of
worksharing. As discussed by USPS witness McCrery and quantified by Mr. Glick, this
40 After all, a goal of the Postal Service proposal in general and the container rate in particular was not just to maintain current mail preparation practices, but to promote “efficiency improvements,” make “progress towards more efficient mail preparation,” and “encourage better mail preparation and more worksharing.” Tang Direct (USPS-T-35) at 11, 16.
- 54 -
preparation option generates considerable savings for the Postal Service, particularly
when, as nearly always occurs, pallets are dropshipped and contain at least 250 pounds
of periodicals. Glick Direct (MPA/ANM-T-2) at 28-30; 11 Tr. 2865-2866; 34 Tr. 11460-
11461. Moreover, 5-Digit pallets will continue to provide these efficiencies after the
rollout of the Flats Sequencing System (FSS). Response of McCrery to MPA/USPS-
T42-5(c) (11 Tr. 2866).
The USPS rate proposal in this case, however, would create a postage
disincentive for preparation of mail on 5-digit pallets. Suppose that a 2,000-pound
periodical mailing had the address density that allowed its entry on either a one 3-Digit
pallet or four 5-Digit pallets. Under the USPS proposal, if the 5-Digit pallet option were
chosen, the container charge would be $3.40 (4 x 85 cents), $2.55 more than if the mail
were entered on one 3-Digit pallet. See USPS Request, Attachment A, p. 33 (Proposed
Rate Schedule 421).
Although the amount of the disincentive appears modest when expressed on a
per-piece basis, the direction of the incentive would clearly cause some movement
away from 5-Digit palletization. Wherever possible, publishers and printers avoid
preparation options that increase the combined cost of mail preparation, transportation,
and postage. The record shows that publishers react to financial incentives. Cohen
Direct (MPA/ANM-T-1) at 6-8, 15-16; Response of MPA/ANM witness Cohen to
MH/MPA/ANM-T1-5 (30 Tr. 10163); 30 Tr. 10217-10218. And, undeniably, there is a
cost to preparing and transporting these pallets.
3. The Postal Service’s proposed container charge would create rate shock for some small publishers and perverse incentives for inefficiency.
The Postal Service’s proposed container charge also creates perverse incentives
for inefficiency. While the general idea underlying the proposal—that Periodical rates
- 55 -
should give better signals about container-related costs—is sound, the container rate
design as actually implemented is poorly designed.
First, the proposal would unfairly overcharge the outside-county portion of in-
county publications by charging 85 cents for all containers with any Periodicals Outside
County pieces. Glick Direct (MPA/ANM-T-2) at 25; 39 Tr. 13539-13540.
Second, the proposal would discourage the efficient practice of mixed-class
comail, again by charging 85 cents for all containers with any Periodicals Outside
County pieces. Glick Direct (MPA/ANM-T-2) at 25; Cohen Direct (MPA/ANM-T-1) at 19.
Third, the proposal would have a disproportionate impact on small publications
entered in small sacks, or without containers at all. POIR No. 19 response (30 Tr.
10373). MH/MPA/ANM-T2-2-4 (30 Tr. 10304-10306); MH/MPA/ANM-T1-6-7 (30 Tr.
10164-10165). See also 30 Tr. 10348 (on situations where there will be small sacks);
30 Tr. 10440 (explaining that the container charge will likely have a significant impact on
a large number of publications without much benefit in terms of efficiency because
those publications do not enter a significant volume of mail); 30 Tr. 10219-10220
(benefit of container charge in terms of increasing sack size is small given the 24-piece
sack minimum rule).
Several members of the Commission expressed concern about these potential
effects during the appearance of USPS witness Tang on the witness stand.
Commissioner Hammond, for example, expressed concern that the proposed container
charge could greatly increase the postage costs of isolated outside county copies of in-
county newspapers. 7 Tr. 1878-79. Ms. Tang conceded in response that:
In terms of the container rate, it’s the scenario as you just described, yes, I would suppose so. But as I mentioned, it’s part of the rate design and it’s part of the rate structure.
7 Tr. 1880 (Tang). Commissioner Tisdale expressed concern about the levying of
virtual container charges for periodical mail entered in volumes small enough to be
- 56 -
uncontainerized. Id. at 1882-83. And Chairman Omas ordered the Postal Service to
provide additional information after the hearing concerning the impact of the proposed
container charge on the “thousands of very small periodical publications.” Id. at 1886.
“That is a very big concern to all of us, the small publications.” Id.
During Ms. Tang’s redirect examination, the Postal Service attempted to show
that one-piece sacks will be rare. Although the Postal Service is correct that the use of
sacks containing only one Outside-County piece will likely be rare, the responses to
POIR No. 19 make clear, however, that some publications (particularly in-county ones
similar to the local newspaper about which Commissioner Hammond expressed
concern) do indeed enter only a small number of Outside County pieces per sack. And
the rebuttal testimony of USPS witness Taufique confirms that these publications are
the ones most likely to receive very large increases under the USPS proposal. See 39
Tr. 13539-40 (Taufique) (acknowledging that the four small publications in the POIR 19
sample that would receive larger percentage increases from the USPS rate design than
the highest increase resulting from the MPA/ANM rate design are likely to be the
Outside County portions of Within County publications).
Fourth, the Postal Service’s container charge proposal would also create major
administrative problems for commingled publications. For comail and co-palletization
pools, the entire container charge would have to be paid on a single postage statement.
7 Tr. 1679 (Tang). This requirement would create practical difficulties in allocating the
charge among the individual publications in the container (e.g., by piece vs. pound).
Moreover, in states that apply sales tax to postage except when it is paid directly to the
Postal Service, a container charge could subject the participants in a co-pallet or co-
mailing pool to additional tax liability unless the Postal Service agreed to charge each
participant directly for an allocated share of the container charge.
- 57 -
4. The proposed Carrier Route and destination entry discounts are too small because they are based upon inaccurate and understated cost avoidance estimates as well as inappropriate assumptions about future costs.
As discussed in greater detail in Section V.A., the Postal Service’s proposed
nonautomation benchmark is inappropriate. See Glick Direct (MPA/ANM-T-2) at 14-16.
As discussed in Section V.B., the USPS proposed cost avoidance model is also
inaccurate, causing its cost avoidance estimates to be understated. See Glick Direct
(MPA/ANM-T-2) at 16-23.
Further, USPS witness Tang expresses the concern that Carrier Route
preparation may have less value in the future. 7 Tr. 1687-88 (Response of USPS
witness Tang to MPA/USPS-T35-4(f)). This concern is misplaced. Glick Direct
(MPA/ANM-T-2) at 14 n. 5. Although Carrier Route presort may not have value for
some locations in the future flat sequencing environment, McCrery Direct (USPS-T-42)
at 22, having a large number of pieces per ZIP Code still will. 11 Tr. 2865-66 (response
of USPS witness McCrery to MPA/USPS-T42-5(b)-(c)). In other words, FSS may
change the method of preparing large mailings (e.g., from preparing these mailings in
Carrier Route bundles to preparing them in much larger 5-Digit groupings) without
reducing the costs that these mailings avoid for the USPS.41
Furthermore, the Postal Service has made at least two errors in estimating
Nontransportation destination entry cost avoidances. First, the Postal Service
calculates the nontransportation costs avoided by entering Periodicals at destination
facilities by using non-dropshipped Zones 1 & 2 Periodicals as the cost benchmark. As
described in Section IV.A., this benchmark understates the nontransportation cost
savings from destination entry. Glick Direct (MPA/ANM-T-2) at 30-32; 5 Tr. 872.
41 The same is true for comailing: the method of preparation may change, but its value for the USPS will remain. See Glick response to USPS/MPA/ANM-T2-25 (30 Tr. 10349-10350).
- 58 -
Second, Ms. Tang made an arithmetic error when calculating the DSCF
advertising pound rate. Specifically, in an intermediate step in her calculation of
advertising pound rates, she incorrectly subtracted the DSCF container handling cost
avoidance (net of the DADC container handling cost avoidance), rather than the entire
DSCF container handling cost avoidance. Glick Direct (MPA/ANM-T-2) at 4-5;
Response to USPS/MPA/ANM-T2-34 (30 Tr. 10357).
5. The proposed disproportionate 18 percent increase in the Ride Along rate is an example of mechanistic ratemaking at its worst.
The Postal Service is proposing to increase the Periodicals Ride-Along rate from
13.1 cents to 15.5 cents, or by approximately 18 percent. See USPS-LR-L-126,
worksheet “Rate Comparison.” The Postal Service proposal is based upon strict
adherence to the original (albeit meaningless) formula that was used to set the Ride
Along rate. A rate increase in line with the subclass average would be much more
appropriate because, even at that level, the Ride-Along rate would make a significant
contribution to institutional costs. Glick Direct (MPA/ANM-T-2) at 35-38. Even after
covering the applicable pound rate, there is plenty of room to cover additional mail
processing and delivery costs, if any, that result from inclusion of a ride-along piece.
Glick Direct (MPA/ANM-T-2) at 37. Any additional costs caused by the greater use of
polywrap for pieces containing ride-alongs will be eliminated by the test year. Glick
Direct (MPA/ANM-T-2) at 37-38.
C. The Commission Should Adopt The Periodicals Outside County Rate Design Proposed By MPA And ANM.
1. The MPA/ANM rate design provides stronger incentives for efficient mail preparation than does the USPS rate design.
As shown by Mr. Glick, the MPA/ANM rate design provides more incentive to
commingle and dropship periodicals than does the USPS proposal. Glick Direct
- 59 -
(MPA/ANM-T-2) at 9-10; Response to ABM/MPA/ANM-T2-33 (30 Tr. 10258).42
Moreover, unlike the USPS proposal, it provides an incentive to prepare 5-Digit pallets,
which will increase efficiency with little revenue leakage to existing volume. Glick Direct
(MPA/ANM-T-2) at 28-30.
2. Publishers and printers will respond to these greater incentives: hence, discounts are more than just a “reward” for existing worksharing.
As previously noted, the Commission has emphasized the importance of using
rate incentives to encourage economic efficiency:
Cost-based rates and encouraging efficiency in postal operations have been two frequent hallmarks of Commission rate recommendations. The Commission has recommended – and the Postal Service has adopted – rates that move all classes and subclasses of mail toward more efficient preparation through discounts for presorting, dropshipping, palletizing, and other cost-efficient measures.
Docket No. C2004-1, Complaint of Time Warner Inc. et al. Concerning Periodicals
Rates, Order No. 1446 (Oct. 21, 2005) at 45.
Efficient price signals require that rate incentives for co-mailing and co-
palletization reflect the costs that the Postal Service would avoid from one more unit of
co-mailing or co-palletization – i.e., the marginal unit. ABM/MPA/ANM-12-2(c) (30 Tr.
42 On rebuttal, Mr. Taufique suggests that USPS proposal would bring about more changes than would the MPA/ANM proposal. See 39 Tr. 13455 (discussion of MPA/ANM proposal as only a reward); 39 Tr. 13467 (container charge is “more direct way of changing that behavior”); 39 Tr. 13472 (about MPA proposal’s encouragement to palletize – “Not in as direct fashion as much as the USPS proposal does”); 39 Tr. 13468 (Taufique notes that change is important and that USPS proposal will bring about more change where it counts). If Mr. Taufique wants change where it counts – which in his mind is only in the areas of palletization and dropshipping – he should prefer the MPA/ANM proposal because it provides more incentive than does the USPS proposal for changes in these areas. 39 Tr. 13469-13470. Further, Taufique’s argument that the container charge is a “more direct way of changing that behavior” is nonsensical. The way in which mailers react to postage incentives is unaffected by whether the incentive is direct or indirect.
- 60 -
10245); Cohen Direct (MPA/ANM-T-1) at 8-9; Panzar Direct (PB-T-1) at 16-26; Buc
Direct (PB-T-2) at 4-5; accord, MC95-1 Op. & Rec. Decis. ¶ 4256. An individual mailer
will respond to these signals by co-mailing and co-palletizing when the benefits to the
mailer (including the rate discounts offered by the Postal Service) from one additional
unit of co-mailing or palletization exceed the costs to the mailer. ABM/MPA/ANM-T2-
2(c) (30 Tr. 10245); Cohen Direct (MPA/ANM-T-1) at 12-16. Stated otherwise, the
efficient pricing signals are designed to influence mailer behavior for the marginal piece
of mail. Panzar Direct (PB-T-1) at 36-39. The effect of such price signals is to
encourage mailers to buy a mix of services from the Postal Service and private sector
suppliers that minimizes the combined cost of those services. 30 Tr. 10172-73 (Cohen).
The Commission has repeatedly found that marginal cost pricing, and ECPR, the
particular application of the marginal cost pricing rule to the pricing of intermediate
outputs that can also be supplied by the private sector, promote both fairness and the
minimization of combined Postal Service/private sector costs. See, e.g., PRC Op. &
Rec. Decis. MC95-1 (Jan. 26, 1996), ¶3074; PRC Op. & Rec. Decis. R2000-1 at ¶ 5060.
The Commission has reiterated its support for ECPR in this Docket. In its Second
Notice of Inquiry, the Commission stated:
It is well-established that cost-saving characteristics that are the result of worksharing should be specifically reflected in the rates. The Commission has adhered to the principle that within a subclass, worksharing rate differences should fully reflect the costs that the Postal Service would avoid (or incur) if the mailer were to move from one workshared category of mail to another. Rates that fully reflect avoidable costs (rates that pass 100 percent of the differences in avoidable cost through to the mailer) satisfy the cost recovery and fairness policies of the Postal Reorganization Act.
Second Notice of Inquiry (July 21, 2006) at 2-3. Accord, Third Notice of Inquiry (July 26,
2006) at 2; 30 Tr. 10258 (Glick); Glick Response to ABM/MPA/ANM-T2-10 (30 Tr.
10257).
- 61 -
The participants in this case who seek to block or slow the movement toward
greater incentives for efficiency in Periodical rate decision do not challenge directly the
merits of marginal cost pricing and the ECPR. Instead, they argue (simultaneously) that
the bigger rate incentives proposed by MPA/ANM are (1) smaller than needed to
overcome the added costs of and practical obstacles to additional worksharing, yet (2)
bigger than needed to compensate mailers that already engaging in worksharing, and
thus merely confer a “reward” or windfall on those mailers.43 These objections are as
unfounded as they are self-contradictory.
a. Increased incentives for worksharing will change the behavior of publishers and printers at the margin.
The economic case for the marginal cost pricing rule and its corollary for
vertically integrated firms, ECPR, does not assume that the entire potential supply of
worksharing activity be at the margin of cost-effectiveness. Such an assumption is
neither realistic nor necessary. Some mail is so costly or difficult to co-mail or co-
palletize that no economically feasible worksharing discount would induce this mail to
be commingled. At the other extreme, some mail is so cheap and easy to commingle
that even smaller discounts would suffice to induce the worksharing. But the mail
volume relevant to efficient pricing is not the mail at the extremes, but the mail at the
43 See, e.g., 35 Tr. 12061, lines 8-10 (comail incentives are adequate); 35 Tr. 12067-12068 (“it serves little purpose to create a demand for services well in excess of the ability of the industry to provide them. Great progress in co-mailing and co-palletizing has been accomplished and will be accomplished with a level of incentives at or near the level in today’s rates. Pushing the incentive accelerator to the floor will do more harm than good.”); 35 Tr. 12180 (“I think that this shows that the present incentives, which will be increased as a result of this case no matter which proposal or combination of proposals the Commission adopts, are doing their job. It is unnecessary to expose those that cannot move from sacks to punishing rate levels in order to increase the incentives and reward for palletizing too much an too soon.”).
- 62 -
margin of co-mailing or co-palletizing. If mail volume exists at the margin of
worksharing, an increase in the existing incentives would induce additional worksharing.
The record makes clear that this condition holds. The cost and feasibility of
additional co-mailing and co-palletizing is a continuum, not the discontinuous, all-or-
nothing, or bipolar state assumed by the defenders of the status quo. As a result,
increased incentives will most definitely result in improved mail preparation. See
MPA/ANM witness Cohen (30 Tr. 10146-47); MPA/ANM witness Glick (30 Tr. 10246-
10247, 10300); Cohen Direct (MPA/ANM-T-1) at 11-16.
Begin with the demand side. It is undisputed that the advent of pallet and co-
pallet discounts has increased the demand of publishers for co-mailing and co-
palletizing. Sixty-five of the 100 medium-size publications analyzed by MPA/ANM
witness Glick in response to POIR No. 19 already comail or co-palletize. 30 Tr. 10373
(Glick). Likewise, a recent ABM survey of 360 publications shows that nearly three-
quarters of the 251 million pieces of mail covered by the survey “are associated with
publications that are not predominantly sacked.” Bradfield Reb. (ABM-RT-1) at 3 & Exh.
LB-2 and LB-3 (35 Tr. 12061, 12080-90).
Prices also influence behavior on the supply side. The notion that increased
incentives will merely “create a demand for services well in excess of the ability of the
industry to provide them,” Bradfield Reb. (ABM-RT-1) at 9 (35 Tr. 12067), assumes a
static and unchanging world that is wholly at odds with the dynamic behavior of actual
suppliers in actual competitive markets. The free market is the reverse of Field of
Dreams: when customers come, suppliers build it. The printing industry has responded
to palletizing incentives by making substantial investments in added capacity to co-mail
and co-palletize. Cohen Direct (MPA/ANM-T-1) at 11-16.
Publisher demand for relief from “postal cost pressures” is why “the printing
industry is stepping forward with more co-mailing, co-palletization and other programs.”
Mailing Options for Small and Medium-Sized Publishers, Folio Magazine (Jan. 2006) at
- 63 -
24 (reproduced at 30 Tr. 10159). Publisher demand for co-mailing is why Quebecor
World Logistics, a major printer, announced last month that it will increase the co-
mailing capacity of its plant in Bolingbrook, Illinois, by 50 percent in the first quarter of
2007. See Quebecor News Release (Nov. 15, 2006) (reproduced at 35 Tr. 12119-20).
“Increasing customer demand, as well as considerable additional interest as a result of
pending postal rate increases have accelerated QWL’s expansion plans.” Id.
This surge of investment by printers and logistics companies in technology and
capacity has produced direct benefits for their publisher customers. The minimum
circulation size for these services to be cost effective has been declining. See Cohen
Direct (MPA/ANM-T-1) at 11-16; 30 Tr. 10152-53 (Cohen); id. at 10159 (Folio Magazine
article entitled “Mailing Options for Small and Medium-sized Publishers”). In addition,
smaller printers lacking sufficient volume density to make commingling profitable on a
stand-alone basis can increasingly achieve the necessary densities by teaming up with
larger printers or third-party logistics companies. Cohen Direct (MPA/ANM-T1) at 12;
30 Tr. 10441-10442 (Glick); 30 Tr. 10213, 10230 (Cohen).
But this progress should not be interpreted (as ABM witnesses have claimed) as
evidence that all or most of the potential gains from commingling have been realized.
Cf. Bradfield Reb. (ABM-RT-1) at 3 (35 Tr. 12061) (asserting that “the co-mail incentives
then in place” in 2004 “were adequate, for those that are able to participate,” and
“nothing has happened since to diminish those incentives”). To the contrary, insufficient
rate incentives have clearly limited participation in commingling and dropshipping
programs; and changing those incentives will cause participation to increase. The
record evidence makes clear that the cost and feasibility of commingling and
dropshipping range over a continuum, and that marginal changes in the rate incentives
will cause changes in the volume of mail that is workshared in these ways. 30 Tr.
10184-86 (Cohen).
- 64 -
For example, the survey responses supplied by ABM in response to TW/ABM-2
make clear that the level of copalletizing and comailing by the survey respondents will
very much depend on the rates set in this case. 30 Tr. 10194 (Cohen); id. at 10421
(Glick).
The failure of any publishers or printers to make use of the co-palletization
discounts established in Docket No. MC2004-1, Experimental Periodicals Co-
Palletization Dropship Discounts For High Editorial Publications (“Co-pal II
Experiment”), also confirms the sensitivity of co-palletizing to marginal changes in its
economics. The experiment failed because unexpectedly high fuel prices made
transportation more costly than anticipated, rendering the discounts uneconomic. 30 Tr.
10427 (Glick); see also MC2004-1 Data Collection Reports (filed Nov. 27, 2006, May
10, 2006, October 26, 2005, and June 14, 2005).
The level of incentives also has a demonstrable effect on the rate of adoption of
co-mailing. While co-mailing generally results in net postage savings, some publishers
just break even. See 35 Tr. 12066, 12101 (ABM witness Bradfield). Hence, marginal
changes in co-mail incentives clearly will affect co-mail volume. Id. at 12102 (conceding
that level of incentives affects tradeoff point for publisher). Accord, Mailing Options for
Small and Medium-Sized Publishers, Folio Magazine (Jan. 2006) at 24 (reproduced at
30 Tr. 10159)
On cross-examination, ABM’s witness Bradfield, an executive with VNU Business
Media, a publisher of 45 tiles (35 Tr. 12060), ultimately conceded that larger incentives
would logically result in more comailing and copalletizing. “[L]ogically if the incentives”
for comailing and copalletizing “were increased, then there would probably be more,” id.
at 12100. For Mr. Bradfield’s own company, a widening in the incentives to palletize
“would change the tradeoff point for us.” Id. at 12102. Changes in incentives for co-
palletizing and co-mailing could also affect the extent to which VNU divides its titles into
- 65 -
multiple versions, which in turn would affect the amount of volume entered by the
company on pallets. Id. at 12103-04.44
b. That properly-sized worksharing discounts would incidentally “reward” mailers who would be willing to workshare even with smaller discounts is no justification for departing from ECPR-compliant prices.
ABM’s alternative (and inconsistent) objection—that deepening worksharing
discounts will incidentally “reward” worksharing that would occur even at lower
discounts—is true but irrelevant. That result is in fact the normal outcome in any market
with an upward sloping supply curve when the price for a good or service is set equal to
the cost of the marginal unit supplied. Under these circumstances, suppliers of the
“inframarginal” (i.e., lower-cost) outputs normally receive “rewards” by definition. That is 44 Mr. Bradfield’s suggestion that co-mailing and co-palletizing can only be cost-effective if the pool of publications includes large publications (35 Tr. 12063-64) reflects a misunderstanding of the economics of commingling. The size of the benefit from commingling depends on the extent to which commingling changes the characteristics of the mail. While pools comprised solely of small publications may not achieve the same ultimate depth of presort as pools with larger publications, the former publications would also be likely to achieve lower depths of presort if mailed solo than would the larger publications mailed solo. The size of the payoff in reduced postage from the commingling depends not on the level of presort ultimately achieved, but the incremental improvement in presort depth. 30 Tr. 10422-23 (Glick).
For example, a comail pool comprised of small publications might upgrade mail from the 3-Digit automation to the 5-Digit automation rate category and from sacks to pallets. Although the comail pool would not ultimately be as finely sorted as a pool consisting of larger publications, the starting point would also be less finely sorted, and the incremental postage savings from upgrading the mailing can be just as large. For example, under the MPA/ANM proposal, the postage savings would be significant – 2.7 cents for palletizing plus 7.2 cents for upgrading from the 3-Digit to 5-Digit automation rate category. Calculated from Glick Direct (MPA/ANM-T-2) at 8, Table 1. Destination entry discounts would increase the savings further. This is more than the 7.4-cent savings that would result under the MPA/ANM proposal from upgrading a palletized mailing from 5-Digit automation to Carrier Route. Calculated from Glick Direct (MPA/ANM-T-2) at 8, Table 1. This presumably is the model followed by Ovid Bell and available to other smaller printers. See 35 Tr. 12065 (Folio Magazine article).
- 66 -
how ECPR, and efficient competitive markets, treat low cost suppliers. Panzar Direct
(PB-T-1) at 17-19, 22 and 35. Even the economic witness for APWU (ABM’s
counterpart in opposing ECPR-compliant worksharing discounts for First-Class Mail)
conceded this on cross-examination. See 20 Tr. 7206-10 (APWU witness Kobe).
3. Increased commingling and dropshipping will benefit the subclass as a whole, because cost avoidances still exceed the discounts proposed by Mr. Glick.
To avoid rate shock for small publications, the MPA/ANM rate design stops well
short of passing through 100 percent of worksharing cost avoidances. This means that
increases in commingling and dropshipping will reduce Periodicals Outside County
costs more than Periodicals Outside County revenues. For example, the passthrough
underlying the Carrier Route discount proposed by MPA/ANM is only 80 percent. Glick
Direct (MPA/ANM-T-2) at 14, footnote 5. Further, as Mr. Glick explained in response to
an interrogatory, for multiple reasons, commingling and dropshipping periodicals will
reduce Periodical Outside County costs much more than revenue. USPS/MPA/ANM-
T2-16(d)(ii) (30 Tr. 10336).
4. The MPA/ANM proposal will not cause undue rate shock for small publications.
ABM’s claim that the MPA/ANM rate design would have an excessive rate impact
on small publishers is unfounded. First, even for small mailers that do not change their
worksharing behavior, the largest increase under the MPA/ANM proposal is only about
half of the largest increase under the USPS proposal. Second, ABM has understated
the extent to which small publishers can offset the rate increases through greater
commingling and destination entry.
Moreover, some of the rate impacts that ABM has focused on in this case are
simply the proper recognition of the cost effects of business practices voluntarily
engaged in by some publishers. These practices—e.g., airshipping to accommodate
- 67 -
the latest possible editorial deadline and versioning—while intended to increase
subscription and advertising revenue, also hinder palletization. MPA and ANM do not
criticize publishers that choose to engage in these practices, which are both widespread
and legitimate. But efficiency and fairness dictate that such publishers bear the costs
that these choices create. Efficiency requires that publishers, when weighing the
potential costs and benefits of tighter editorial deadlines or additional versions, receive
price signals that allow appropriate decisionmaking. Perpetuating a postal rate
structure that fails to recognize the full efficiencies of pallets over sacks encourages
overuse of mailing practices that tend to increase sack usage.
Likewise, it is unfair to ask the Postal Service—and thus other publishers—to
subsidize the additional costs that the Postal Service must incur when publishers
shorten their editorial deadlines and multiply their versions. Individual publishers do not
expect their trucking companies to provide truck transportation, their paper suppliers to
furnish subsidized paper, or their other vendors to provide goods and services below
cost. And no publisher is required to share with other publishers the additional income
generated by offering fresher editorial content and multiple versions. By the same
token, however, it is unfair to require one publisher to subsidize another by maintaining
a postal rate structure that fails to recognize the full additional costs of mail entered in
sacks.
a. Even for small mailers that do not change their worksharing behavior, the largest increase under the MPA/ANM proposal is only about half of the largest increase under the USPS proposal.
MPA and ANM agree with USPS witness Taufique that “there are a number of
small mailers that will be affected by what we propose,” and the participants and the
Commission need to be “mindful of the impact on the smaller mailers.” 39 Tr. 13467. It
is appropriate for the Commission to balance efficiency concerns against rate shock,
- 68 -
and to compare the “worst case scenario” for smaller mailers under each of the
competing rate proposals. 30 Tr. 10174-75 (Cohen).
Because of the importance of this issue, the Commission should be commended
for requiring each of the proponents of Periodical rate designs in this case to undertake
an impact analysis on the POIR 19 sample of publications. These analyses clearly
illuminate the relative impact of each rate design on small publications, and enable the
Commission to provide the protection that it deems appropriate. The POIR 19
comparisons also demonstrate that the MPA/ANM proposal does the best job of
protecting small publications.
The comparisons show that, despite the larger incentive that the MPA/ANM
proposal offers for commingling and dropshipping, the maximum rate increase under
the proposal for any the small publications in the POIR 19 data set is only 22.6
percent—a much smaller value than the 43.7 percent maximum increase produced by
the USPS proposal. See 30 Tr. 10373 (Glick response to POIR 19). Indeed, several
small publications in the POIR No. 19 sample would receive rate increases under the
USPS proposal that exceed the maximum rate increase under the MPA/ANM proposal.
39 Tr. 13484.45 Even ABM witness Bradfield, whose prefiled rebuttal testimony took no
position on the relative merits of the USPS and MPA/ANM rate designs (see 35 Tr.
12094, lines 22-25), conceded that the USPS rate design is “more destructive” because
of its higher maximum increases for small publications. Id. at 12098, line 12, to 12099,
line 18.
The maximum rate increase is lower under the MPA/ANM proposal because its
designers made a conscious effort to provide adequate protection to small publications
45 While the MPA/ANM proposal produces a bigger average rate increase for small mailers than does the USPS proposal, the difference is small: 16.5 percent under the MPA/ANM proposal vs. 16.0 percent under the USPS proposal. Id.
- 69 -
that do not commingle or dropship. The MPA/ANM proposal would replace the
container charge with a pallet discount. In addition to providing a larger average
incentive to palletize, the proposal would limit maximum postage impact of using sacks.
30 Tr. 10304, 10164-65 (Glick).46
The proposal would also protect the unzoned editorial pound rate and rates for
Basic Presort and Nonautomation pieces. Glick Direct (MPA/ANM-T-2) at 6; Cohen
Direct (MPA/ANM-T-1) at 21-22. Specifically, the proposal would increase the Carrier
Route discount by lowering the proposed automation discounts by 0.5 cents per piece,
and also lower the passthroughs that underlie existing automation discounts.
USPS/MPA/ANM-T2-26 (30 Tr. 10351). The proposal would also lower the 3-Digit
presort discount by 1.5 cents per piece, and shift a portion of the editorial discount from
piece side to pound side.
In his rebuttal testimony, USPS witness Taufique tried to brush off the higher
maximum impact of the USPS rate design on the ground that only four small
publications in the sample have impacts exceeding the 22.6 percent maximum impact of
the MPA/ANM proposal. The obvious and quite applicable rejoinder is that a sample
represents the universe from which it was drawn. At the sampling rate of the 251-
publication POIR 19 sample, the four publications with impacts ranging from 23 to 43.7
percent under the USPS proposal represent a universe of about 1,000 publications. 39
Tr. 13487. Indeed, the single publication in the sample that would receive a 43.73
percent rate increase under the USPS proposal alone represents about 250
publications. Id.47
(footnote continued on next page)
46 That the pallet discount does not encourage use of larger sacks will not have significant efficiency implications because the use of small sacks is no longer a common practice given the 24-piece sack minimum rule. Glick Direct (MPA/ANM-T-2) at 11. 47 During redirect examination, USPS witness Taufique suggested that the largest increases under USPS proposal are unimportant because the publications receiving
- 70 -
Mr. Taufique also tried to dismiss the impact analysis on the theory that the
standard deviation of all the rate increases is a better measure of rate shock than is the
maximum percentage of any individual increase. 39 Tr. 13457, 13478-13484
(Taufique). This is a bizarre position. The standard deviation is a statistical concept:
the square root of the variance, which in turn is the weighted average of the squares of
the deviations of the individual outcomes from the mean value. Id. at 13481. The
standard deviation thus is essentially a measure of the average distance of all of the
points from the mean value, not the maximum distance. Id. at 13483. If the results for
most mailers are tightly clustered around the mean value, “you could have a low
standard deviation even if some mailers get really big increases.” Id. (emphasis added).
Mr. Taufique was unable to identify any Commission decision adopting the standard
deviation of all of the rate changes as a test for rate shock. Indeed, the term appears
nowhere in the discussion of rate shock in the Commission’s merits decision in C2004-
1, or the testimony filed by the USPS in that case. 39 Tr. 13489-13492.
them represent the Outside County portion of In-County Periodicals. Taufique Redirect (39 Tr. 13539-13540). We strongly disagree. As noted above, the publications receiving rate increases larger than the MPA/ANM maximum represent approximately 1,000 actual publications (39 Tr. 13487), some of which are the Outside-County portion of In-County publications and some are not. The impact of the USPS rate design on the former publications underscores the unfairness of the USPS container charge, which is imposed in its entirety on the Outside County of In-County publications, even when the content of the container is predominantly In-County copies. Moreover, the huge rate increases that the Postal Service’s proposal (particularly the container charge) would load onto In-County publications is in addition to the 24 percent average rate increase proposed by the USPS for the in-county portion of these publications. See Tang Direct (USPS-T-35) at 3.
- 71 -
b. ABM has understated the extent to which publishers can offset the rate increases through greater commingling and destination entry.
In this docket, ABM has suggested that some publications simply cannot be
commingled or dropshipped irrespective of any cost-based rate incentive.48 In
particular, ABM singles out five kinds of publications: low-circulation titles, tabloids,
polywrapped publications, time-sensitive weeklies, and publications that have been
divided into multiple versions. We discuss each in turn.
Low-circulation publications. ABM contends that comailing requires a
minimum circulation of 5,000 copies. Because there are no physical barriers to
comailing smaller-circulation publications, however, common sense suggests that
increased incentives would reduce the minimum circulation at which comailing is cost-
effective. As noted by Mr. Bradfield, Quad/Graphics co-mailed publications down to
1,500 pieces even back in 2004. 35 Tr. 12063. Likewise, in his testimony, Mr. Cavnar
notes that two ABM publications with less than 5,000 copies might begin comailing or
co-palletizing soon. ABM-T-1 at 3. Accord, Cohen Direct (MPA/ANM-T-1) at 13 (noting
that Ovid Bell is already offering co-mailing to publications with only a few thousand
copies per issue).
Tabloids and polywrapped publications. The difficulties created by special
publication formats are not insurmountable, contrary to the claims of ABM. Cohen
response to ABM/MPA/ANM-T2-19 (30 Tr. 10148). First, it should be noted that there
48 See, e.g., ABM/MPA/ANM-T2-19(a) (30 Tr. 10148) where ABM asked “Do you agree that there are some types of Periodicals mail that cannot now be co-mailed or co-palletized irrespective of any reasonable cost-based rate incentive that might be adopted (assuming that the mailer is unwilling to alter the basic nature of the publication)?” Although there are more logistical impediments to commingling and dropshipping some types of publications than others, these impediments are not insurmountable. Further, larger postage incentives would provide more encouragement to overcome any obstacles. Cohen Response to ABM/MPA/ANM-T2-19, 21 (30 Tr. 10148,10150).
- 72 -
are no impediments to co-palletizing tabloids and polywrapped publications. Glick
response to ABM/MPA/ANM-T2-3 (30 Tr. 10249). In fact, at least one publisher of
tabloid publications has been co-palletizing its publications since 1987. McGarvy Reb.
(ABM-RT-2) at 4 (35 Tr. 12177). Further, the co-mailing of tabloids is technically
feasible. 30 Tr. 10150 (Cohen). R.R. Donnelley was willing to co-mail Crain’s tabloid
publications and other tabloids are currently being co-bound. 35 Tr. 12178-12179,
12186-12187 (McGarvy). Also, although polywrapped publications are not generally
comailed with other non-polywrapped publications, polywrap-only pools are available,
according to an executive at Quad/Graphics. Docket No. C2004-1, Schick Direct (TW
et al.-T-4) at 7.
Weekly Publications With Tight Editorial Deadlines. ABM gives particular
emphasis to the difficulties of commingling and palletizing for weekly publications with
tight editorial deadlines. ABM reasons that preserving the freshness of the weekly
editorial content requires editorial deadlines that are too tight to wait for the assembly of
a comailing pool. See 35 Tr. 12176-12177 (McGarvy). Several responses are
warranted.
First, a sufficiently motivated publisher can, through proper coordination,
assemble comailing pools quickly. Cohen response to ABM/MPA/ANM-T2-20 (30 Tr.
10149). Moreover, it is also possible to co-bind weeklies. Cohen Direct (MPA/ANM-T-
1) at 14. Significantly, Crain, despite the time sensitivity of its weekly editorial content,
is able to co-palletize the portion of its circulation that can be shipped by surface carrier.
35 Tr. 12177 (McGarvy). For a centrally located publisher such as Crain, about 80
percent of total circulation can be shipped by surface transportation. Id. at 12190-
121911.
There is no doubt that the tight editorial deadlines of some weekly publications
may make comailing difficult, and that airshipped publications generally cannot be
- 73 -
palletized. When airshipping occurs, however, other mailers should not be required to
subsidize the additional costs that ensue.
Airshipping and other Practices that publishers may take to maintain the tightest
possible editorial deadline are voluntary business decisions that involve deliberate
economic tradeoffs. Tighter editorial deadlines increase the costs of a publication (e.g.,
by requiring air shipping of copies to distant subscribers, foreclosing the use of pallets
for those copies, and reducing the window of time for coordinating and assembling
comailing pools; but it also increases the attractiveness of a publication to potential
subscribers (and therefore advertisers), 35 Tr. 12188-89 (McGarvy). There is nothing
illegitimate about choosing to incur these extra costs if the anticipated extra profits
appear to justify them. But there is no reason, economic or equitable, for requiring other
publishers to subsidize the additional costs that these voluntary choices cause. Id. at
12187-89. Publishers do not expect air freight providers, trucking companies, or
printers to subsidize such decisions. Id. at 12197. There is no reason for different
treatment of postal costs.49 Moreover, efficiency requires that tradeoffs of this kind be
made with accurate information about the relative costs of each option. Cost-based
prices are a mechanism by which the market provides this information.
Publications with Multiple Versions. Multi-version publications raise the same
issues as weekly publications with tight editorial deadlines. ABM contends that
versioning, by effectively splintering a publication into multiple publications, makes co-
mailing and co-palletizing more difficult or impossible. Versioning, however, does not
49 The added postage expense that publishers would incur under the MPA/ANM proposal when using airshipping are far smaller than air shipping costs. Cf. 35 Tr. 12177, 12196-12197 (acknowledging that air transportation is expensive); id at 12189-90 (the MPA/ANM proposal would cost only a fraction of a penny for most issues than would the USPS proposal).
- 74 -
necessarily disqualify a publication from comailing. Even Quebecor World, the printer
cited by ABM witness Bradfield, allows publications with multiple versions to participate
if the each version exceeds a minimum required size. 35 Tr. 12065. Further, as
discussed above, larger incentives are likely to lower the minimum size requirements.
Furthermore, versioning, like the implementation of tight editorial deadlines, is a
business decision undertaken to generate revenue, either through additional advertising
revenue generated by the ability to offer a demographically selected audience, or
through the targeting of attachments (“cover tips”) designed to “renew people or track
new subscribers.” 35 Tr. 12064-65, 12102-04. While publishing practices of this kind
are reasonable and customary, it is not unfair to expect these publishers to bear the
extra costs that their decisions impose on the Postal Service. Moreover, as with
choices concerning the tightness of editorial deadlines, decisions about how many
versions to offer should be made with correct price signals about the costs imposed on
society by additional versions.
5. Special issues raised by 5-digit pallet discount
USPS witness Taufique, in his rebuttal testimony, singled out the 5-digit pallet
discount proposed by MPA/ANM for criticism on several grounds: (1) that the discount
would just reward existing behavior; (2) that revenue leakage would be large because 5-
digit palletizing is already widespread; (3) that encouraging the use of 5-digit pallets is
less important than encouraging movement from sacks to pallets. Moreover, he argued,
if the pallets are lightly loaded or entered into the postal network far from the destination
of the mail, container-handling costs could increase. Taufique Reb. (USPS-RT-12), at
3-6, 10-11, 39 Tr. 13455-13456, 11460-11461. These criticisms are unfounded.
First, the 5-digit pallet discount would not just reward mailers for existing
behavior. Additional 5-digit pallets can be created through comailing:
- 75 -
Under USPS-LR-189 it's clear that some comails result in much higher five digit pallet percentages. I think five of them had over 20 percent and one had over 50 percent on five digit pallets from a comail.
30 Tr. 10407 (Glick). And this is without any additional incentive to submit 5-digit
pallets. See also Glick response to USPS/MPA/ANM-T2-20(c)-(d) (30 Tr. 10345); 39
Tr. 13476-13477 (Taufique).
Second, the minimal volume currently entered on 5-digit pallets should allay any
concerns about leakage. Pieces on 5-digit pallets comprise less than ten percent of
Outside County Periodical volume. Further, the revenue leakage resulting from the
MPA/ANM proposed 5-digit pallet discounts would represent less than one-half of one
percent of Periodicals Outside County revenue. MPA/ANM-LR-2, worksheet “Rates”;
Glick response to USPS/MPA/ANM-T2-19 (30 Tr. 10343); 39 Tr. 13474-13476
(Taufique).
Third, the concerns about the adverse incentives potentially created by a 5-digit
pallet discount, 34 Tr. 11460-61 (McCrery), can be easily resolved by adopting the
eligibility restrictions proposed by Mr. McCrery. MPA and ANM would support limiting
eligibility for the 5-digit pallet discount to pallets that are DDU/DSCF-entered and weigh
more than 250 pounds. The vast majority of 5-digit pallets meet this standard. Almost
90 percent are DDU/DSCF-entered. The average 5-digit pallet contains about 1,100
pieces and thus, assuming an average piece weight of about one-half pound, weighs
about 500 pounds. Glick Direct (MPA/ANM-T-2) at 29, footnote 19; USPS/MPA/ANM-
T2-20 (c)-(e)(30 Tr. 10345); 34 Tr. 11461.
- 76 -
CONCLUSION
For the foregoing reasons, MPA and ANM respectfully request that the
Commission recommend rates for Outside County Periodicals Mail as set forth herein.
Respectfully submitted,
David M. Levy Ronald S. Flagg Richard E. Young SIDLEY AUSTIN LLP 1501 K Street, N.W. Washington, DC 20005-1401 (202) 736-8000 Counsel for Magazine Publishers of America, Inc., and Alliance of Nonprofit Mailers
December 21, 2006
- 77 -
Top Related