Cohenetal_CreamSkimming.txt

A Preliminary Version of this Paper was Presented at
The Evolving Structure of Postal and Delivery Industries June
11-14, 1997, Helsingør, Denmark
Sponsored by Center for Research in Regulated Industries at
Rutgers University


AN ANALYSIS OF THE POTENTIAL FOR CREAM SKIMMING IN THE U.S.
RESIDENTIAL DELIVERY MARKET
Robert H. Cohen William W. Ferguson John D. Waller Spyros S.
Xenakis
Office of Rates, Analysis and Planning
U.S. Postal Rate Commission
October 1998
Published in Emerging Competition in Postal and Delivery
Systems, edited by M.A. Crew and
P.R. Kleindorfer. Boston: Kluwer Academic Publishers, 1999.
An Analysis of the Potential for Cream Skimming in the U.S.
Residential Delivery Market

A. Introduction
Postal administrations often claim that without a legal
monopoly1 to protect them from cream skimming, they would not be
able to continue providing universal service at uniform and
affordable prices. Because they serve delivery points with a range
of profitability, postal administrations fear that without monopoly
protection cream skimmers would capture their high profit routes,
leaving them with their less profitable routes. The question is,
would cream skimming divert so much volume that universal service
at affordable prices becomes infeasible?
Sweden is the only industrial nation which actually has had
competition in its letter mail market after abolishing its
monopoly.2 A competitor, City Mail, has served selected portions of
the country but was twice forced into bankruptcy.3 City Mail has
again emerged from bankruptcy and is competing with Sweden Post.
Perhaps on City Mail's third attempt, we will finally have a valid
test of whether Sweden Post actually needs monopoly protection in
order to provide universal service at reasonable prices. As far as
the authors know, the economics of the competition between Sweden
Post and its competitor has not been studied. If City Mail finally
succeeds in avoiding bankruptcy, such a study might determine the
reasons why Sweden Post was vulnerable to cream skimming. In
addition, it might describe the impact that a viable competitor has
had on
1
In the U.S., as in most industrial nations, the postal monopoly
extends only to letter mail (including addressed advertising).
2
Finland has abolished its monopoly and in so doing created a
licensing system for competitors to Finland Post. No other operator
has yet begun to offer service in competition with Finland Post.
New Zealand has also just recently abolished its monopoly, but it
is too early to know if competitors will emerge and be
successful.
3
A recent study found that the Swedish Postal Service had engaged
in a variety of anti-competitive practices with respect to its
competitor. See "Deregulation of the Postal Service's Market;
report on the abuse by the Postal Office of Its dominant market
position and on the Competition Authority's handling of the
question of competition in relation to the Post Office," Professor
Erik Nerep (Stockholm School of Economics), February 1995.
the dominant provider. It should be noted that the presence of
City Mail has already prompted Sweden Post to lower prices for
large volume mailers in areas served by its competitor.4
Cream skimmers take advantage of an incumbent's price when it is
based on an average of heterogeneous costs. They serve the low cost
portion of the market and price below the incumbent. Conceptually
cream skimming has two basic dimensions - product and geography.
Product cream skimming is where a competitor tries to capture the
most profitable portion of the market for a product with
heterogeneous costs. Geographic cream skimming is where a
competitor tries to provide a service only to selected, low cost
areas.5 City Mail is an example of geographic cream skimming. Its
carriers provide twice weekly delivery to various areas in Sweden
for large volume customers.6 This paper addresses only geographic
cream skimming. 7
Postal monopolies, like most other legal monopolies, are thought
by many observers (1) to be technically inefficient, and (2) to
have delivery functions which exhibit economies of scale.8 This
paper addresses several questions which arise from these two points
under the assumption that the monopoly and any other barriers to
entry were to be removed. It asks how much more efficient would a
potential competitor have to be to overcome the scale economies of
the U.S. Postal Service? Assuming potential competitors are able to
achieve the necessary level of efficiency, it then asks how much
volume might they be able to capture and in which markets? Finally,
it asks what effect
4
It would be interesting to learn: How much it cost Sweden Post
to serve the areas where it lowered prices relative to its average
cost, how important the abandonment of universal pricing is to
Sweden Post's competitive position, and are the circumstances in
Sweden applicable to the U.S.? Section J of this paper examines
this issue further.
5
Cream skimming may have a third dimension having to do with
transactions costs. In practice cream skimmers may restrict
themselves to large volume customers since the transaction costs of
dealing with smaller volume customers and individuals may be
prohibitive.
6
Sweden Post delivers five days per week.
7
A report by the General Accounting Office (GAO) at least
partially analyzed the effect product cream skimming would have on
the U.S. Postal Service. Its assessment was as follows: Priority
Mail (First Class over 11 ounces) - high risk of loss of
significant volumes, but a low financial impact, if it did occur;
First Class - low risk of loss of significant volumes, but a high
financial impact, if it did occur; Advertising Mail low risk of
loss and a medium financial impact, if it did occur. See "Postal
Service Reform; Issues Relevant to Changing Restrictions on Private
Letter Delivery," General Accounting Office, September 1996.
8
In the U.S., legal monopolies include the local delivery of
electricity, gas, water and until recently telephone.
would the loss of volume to cream skimmers have on the resulting
prices the Postal Service would have to charge in order to maintain
universal service?
Many observers believe that cream skimmers in the United States
would serve cities and leave rural areas to the universal service
provider because they assume that the cost of delivery to rural
areas is much higher than the cost of delivery to cities. An
earlier study9 challenged that assumption. It estimated that in
1989 city delivery cost per piece was only 8 percent lower than
rural delivery, but that city delivery cost per delivery point was
actually 7 percent higher than rural delivery cost. It concluded
that there was no cross subsidy of rural delivery by city delivery.
In theory, cream skimming of residential service could occur in
either city or rural areas. This paper focuses on city delivery
because of the more detailed data available on this type of
delivery.
We first provide a brief description of the current competition
in delivery. We then describe the data used in our analysis and the
costs and profitability of the Postal Service's city residential
delivery routes. We next examine the volume vulnerable to capture
by potential cream skimmers. We describe characteristics of routes
most likely to be profitable to competitors and provide a range of
estimates for the cost of cream skimmers. After examining the
consequences of successful cream skimming on Postal Service rates,
we explain why cream skimming is not likely to be successful on
business routes in the U.S. Finally, we provide some observations
on the relevance this analysis may have for other countries.


B. Current Competition In Delivery
Competitors of the U.S. Postal Service may legally deliver only
periodicals, catalogues over 24 pages, parcels, and unaddressed
letters.10 Alternative delivery is particularly difficult because
of the so called mail box law which prohibits any one but the
9 See Cohen, Ferguson & Xenakis, "Rural Delivery and the
Universal Service Obligation: A Quantitative Investigation."
Regulation and the Nature of Postal Delivery Services, Ed. Crew
& Kleindorfer, Kluwer Academic Publishers, 1993.
Technically, the U.S. monopoly is primarily a revenue monopoly.
A third party may deliver a letter if proper postage is affixed and
canceled. The monopoly also has an urgent mail exception for
letters which are charged twice normal First-Class postage with a
minimum of $3.00. Unlike most industrial nations, there is no
exemption from the monopoly based on weight or price.
Postal Service from accessing a private mail box.11 The U.S.
Postal Service broadly interprets "letter" to include any addressed
information recorded on a physical object. Thus, for example, it
considers an addressed grocery store advertisement to be a
letter.
In 1989, Publishers Express was founded by Time Inc., Meredith
Corp., American Express Publishing, New York Times Co., Times
Mirror Co., and R. R. Donnelley, among others, to provide
alternative delivery.12 In spite of its prestigious sponsorship,
the company ceased operations in 1996. Many other small
entrepreneurs provide alternative delivery, but they have a very
small share of the market. They are scattered around the country
and mostly deliver unaddressed "saturation" advertisements and
small quantities of periodicals and small parcels.
The real alternative delivery system in the United States is the
newspaper industry which delivers advertising preprints or
inserts.13 The Postal Service has the advantage of being able to
deliver these advertisements to all addresses, while newspapers
usually (but not always) deliver inserts just to their subscribers.
A typical newspaper in the U.S. has less than 50 percent coverage
in its service area. This puts it at a disadvantage vis a vis the
Postal Service, which delivers to all households. Newspapers, on
the other hand, have the advantage of charging lower prices.
Because marginal costs are very low, a newspaper price for
preprints might be as low as 5 or 6 cents per piece. The lowest
rate currently charged by the Postal Service is 11.0 to 11.4 cents
for saturation mail weighing up to 3.3 ounces and drop shipped at
the delivery office.14 Newspapers deliver about 86 billion inserts
annually,15 while the U.S. Postal Service delivers at least 11
billion competing items.16
11
As far as we know, the U.S. is the only country that has a mail
box rule.
12
For a more comprehensive discussion of alternative delivery in
the United States, see the General Accounting Office Study, n. 7,
supra.
13
Newspapers also deliver product samples.
14
The Postal Service has an additional disadvantage because city
carriers deliver only addressed advertising mail. This requires the
mailer to have an address list kept in delivery sequence and to
have address labels for each piece.
15
Newspaper Association of America's WEB Site;
www.naa.org/marketscope/databank/preppriyr.htm.
16
This was the saturation and 125-piece walk sequence Enhanced
Carrier Route mail volume in 1996. Many of these items are so
called "marriage mail" pieces which contain several individual
advertisements combined into a single piece. These individual
advertisements could each be a stand-alone insert. Thus, 11 billion
pieces is an extreme lower bound on the number of preprints carried
by the Postal Service.


C. City Carrier Data
Each year the Postal Service conducts a City Carrier Survey
(CCS) of delivered mail volume by route. Between 1993 and 1996, the
Postal Service conducted the survey using a relatively small panel
of about 400 routes.17 Prior to that, the Postal Service randomly
sampled a much larger cross section of its routes over the course
of a year. The last set of cross sectional data filed with the
Postal Rate Commission in 1989 consisted of a sample of 16,092
different routes.18 Because of the large sample size, these data
are far more useful for our analysis than the subsequent panel
data.19
Each record in the 1989 CCS contains the date of the
observation, the 5-Digit ZIP Code in which the route is located,
the route type,20 the number of pieces (by subclass and by shape)
per stop, the number of possible deliveries and actual deliveries
per stop, and the stop type ("single delivery residential,"
"multi-delivery residential," and "business-and-mixed"). Generally,
data are recorded for every tenth stop on a route. The authors were
able to make use of the records for 14,884 routes.21 This analysis
is based on the data set of 13,212 usable residential routes22 and
476,953 stops. 23 About 89 percent of all city routes are
residential routes. The rest are business or mixed (residential and
business).


D. Postal Service Delivery Cost And Profitability
This analysis uses delivery data from 1989, but uses 1996 cost
levels for wages, fringe benefits, and other associated delivery
costs. Total volume grew 13.5 percent
17
A panel survey collects data repeatedly from the same
sample.
18
There were 157 thousand city delivery routes in FY 1989. There
were also 46 thousand rural routes, most of which had costs
comparable to city delivery routes. See Cohen, Ferguson &
Xenakis, n. 9, supra.
19
The 1997 CCS data were not available when the analysis for this
paper was conducted.
20
Business foot, business motorized, residential foot, residential
park and loop, residential curb, mixed foot, mixed park and loop,
and mixed curb.
21
The data are organized by the 26 pay periods in a year. Portions
of the data from a few pay periods were not usable. Comparisons
with previously published summaries indicate that the missing data
did not bias our results.
22
To a much lesser extent this analysis makes use of data from 717
business routes.
23
Ninety-four percent (94%) of these stops are residential, while
six percent (6%) are either business, or mixed
business-and-residential stops. Not all stops on a residential
route are residences. Some are business stops and some are mixed
(business and residential).
between 1989 and 1996, and the number of routes increased 7
percent. We make no attempt to adjust for these increases.
When calculating city delivery cost, we include the entire
(fixed and volume variable) in-office and out-of-office cost24 for
all city routes, plus the cost of overtime, supervision, space, and
vehicles. Starting with a productive hourly rate of $25.25 for city
carriers, we calculate a cost per residential delivery route of
$266 per delivery day.25 The cost we calculate can be considered to
be the avoided cost or the incremental cost of the city delivery
function. Alternatively, it can be considered a lower bound on the
stand-alone cost of city delivery.26
We calculate the daily profit of residential delivery routes for
the Postal Service by totaling the revenue minus collection,
processing and transportation costs of the mail delivered on each
route and subtracting the delivery cost of $266.
Table 1 displays a variety of mean route statistics for
residential routes separated into quartiles based on their profit.
less As can be seen, the mean daily profit of each quartile ranges
from positive $248 to negative $110. The average profit for all
residential routes is $41 with 46.5 percent of the routes operating
at a loss. The route cost per piece ranges from 7.6 cents to 23.5
cents, and averages 12.5 cents. Route volume ranges from 3,485
pieces to 1,131 pieces, and averages 2,128 pieces. The number of
possible deliveries (which ranges from 670 to 411) is dependent on
route terrain. 27 All the statistics presented are monotonic
across28 quartiles except for route type. This is because of the
close relationship between cost (per piece, delivery and stop),
volume
24
"In-office" refers to the in the office activity of a letter
carrier (primarily sequencing mail to be delivered),
"out-of-office" refers to the activity of the carrier while on the
street.
25
This is based on the city delivery carrier total cost of $33.20
per hour. This, in turn, is based on a productive hourly rate (pay,
overtime pay, and benefits for hours worked) increased to account
for other costs that can be associated with carriers. These other
costs include systemwide labor related costs (e.g., worker's
compensation, civil service retirement unfunded liability),
indirect labor such as carrier supervision, vehicle costs, and
space related costs (rents, fuel, utilities, custodial
maintenance). The standard work day for a city carrier is eight
hours.
26
It is a lower bound because the cost includes no marketing or
administrative costs which a stand-alone firm would normally
incur.
27
By "physical terrain" we mean the physical proximity of the
delivery points and whether they are most advantageously served by
a curb, park & loop, or foot route.
28
A monotonic series is one whose every term is greater than (less
than) or equal to the previous term. 6
(per piece, delivery and stop), and the number of deliveries
(stops) on a route. With respect to route type, we see that the
percentage of curb routes is greatest in the most profitable
quartile.29 These routes are largely a suburban phenomenon, they
reach more stops in a day and hence, have more volume. The appendix
(Table A1) presents demographic characteristics related to the
quartiles in Table 1.
Table 1
SELECTED AVERAGES FOR RESIDENTIAL ROUTES WHEN ROUTES ARE SORTED
BY PROFIT PER ROUTE
Profit Cost (cents) Per (possible) Volume Pieces Per (possible)
Quartile (dollars) Piece Delivery Stop (pieces) Deliverya Stop


a
In the paper referenced in n. 9, supra, pieces per possible
delivery were weighed by stop and produced a slightly higher pieces
per delivery. Here they are unweighted.
The revenue for 46.5 percent of the routes (all of quartile 4
and most of quartile 3) is not enough to cover the costs of the
mail delivered on those routes. Thus, it might be argued that a
monopoly is necessary to ensure service to those households. For
several reasons, it would seem the burden of proof for this
position should fall on its proponents. The routes which are
unprofitable are dispersed across the country in a great number of
cities. A competitive postal service that refused to serve
households on unprofitable
Curb routes deliver directly from the vehicle to a curbside mail
receptacle. Park and loop routes involve parking, covering an area
by foot, then parking the vehicle in a new area.
routes would impose large transaction costs on its customers to
separate mail and to secure suppliers to deliver its remaining
mail. Such firms would be at a considerable disadvantage to those
providing universal service. Larger transaction networks (be they
mail, packages, overnight or telephone) are more valuable to
customers and providers than smaller networks. It is for sound
business reasons that Federal Express and United Parcel Service
provide universal delivery service. A major business asset of the
U.S. Postal Service and all other national posts is that they
provide service to every address in the countries they serve. Thus,
there is a business incentive for any postal provider (who is not
simply a cream skimmer) to offer universal service within the
territories it serves.
Routes which are unprofitable when serviced six days per week
would become profitable when serviced less frequently. In a
competitive environment, the Postal Service could easily retain
universal service, but perhaps not a universal service standard.30
Delivery less than six days per week would certainly be preferable
to abandoning delivery altogether. Of course, if the Postal Service
were to simply abandon delivery to unprofitable routes, it would
not have to refuse, return or destroy mail destined to these
routes. Delivery firms with lower costs than the Postal Service
would emerge to serve them and the Postal Service could hand off
its unprofitable mail to them. 31 Under any foreseeable
circumstances, universal delivery would not cease.32


E. Volume Available For Cream Skimmers
It is unlikely that cream skimmers would base their selection of
areas to serve on the Postal Service's delivery profits. Postal
Service delivery profitability depends on the revenue for all mail
delivered, and it is not likely that cream skimmers could capture
all categories of mail.
30
It is interesting to note that United Parcel Service recently
began delivery less frequently than daily to certain residential
areas.
31
Costs would be lower by dint of lower wages and/or less frequent
delivery.
32
The concept of a universal service obligation may really mean
more frequent delivery than economically warranted for households
living on unprofitable routes. This, of course, is in addition to
providing retail services to rural communities.
Residential cream skimmers would serve bulk mailers with
sufficient volumes of mail for the areas they serve. We assume that
only mail that is presorted to the carrier route level would be
vulnerable to cream skimming. If a bulk mailer did not have
sufficient volume to sort to the carrier route level, the
transaction costs for both mailer and cream skimmer would be
prohibitive. Of course, third parties might become intermediaries
between these bulk mailers and cream skimmers. Such parties would
have to engage in processing and transportation. They would in
effect be a parallel postal system for bulk mailers. An analysis of
separate postal systems for bulk mailers is beyond the scope of
this paper.
The U.S. Postal Service requires First-Class and advertising
mail to have at least ten pieces for a carrier to qualify for a
carrier route discount. Publications may qualify with as few as six
pieces for a carrier route. The amounts being presorted to the
carrier route level in 1996 are shown in the right hand column:

First-Class Mail is unzoned and receives no discount for being
deposited near its destination. Most First-Class carrier route
volume is prepared by utilities and may be local depending on the
billing operations of the utility. For purposes of this analysis we
assume that all First-Class carrier route mail is available to
local cream skimmers.
Publications Mail. While only 3.2 billion pieces of the carrier
route volume is entered locally or entered close enough to its
destination to be available to cream skimmers, we assume that all
4.5 billion pieces are available to cream skimmers.
Advertising Mail. Eight-five (85) percent or 25.2 billion pieces
of carrier route advertising mail are drop shipped or entered
locally. Notwithstanding, we assume here that all carrier route
advertising mail would be available to cream skimmers.
To summarize, 37.7 billion pieces of carrier route volume are
potentially available to cream skimmers.33 This represents 20.6
percent of the total volume of 182.7 billion pieces in 1996.
Geographic cream skimmers would concentrate on areas with the most
carrier route volume because their cost per piece would be lowest
and their potential profit would be the greatest.
Table 2 displays mean daily route statistics for routes when
divided into quartiles reflecting their volumes of carrier route
mail. We see a wide divergence in carrier route volume among the
quartiles. The range is from 1,263 pieces to 87 pieces with an
average of 551. The first quartile routes have more than twice the
volume of carrier route mail as the second quartile. In fact, the
first quartile has more than half of all carrier route volume
delivered on residential routes. Thus, the first quartile of routes
are much more attractive targets for cream skimmers than the other
routes. We again see a monotonic behavior of the other variables
across quartiles.34 Route profitability by quartile in Table 2
changes by a sizable amount as compared to route profitability in
Table 1. Less than half the routes in quartile 1, Table 1, are also
in quartile 1, Table 2. The appendix (Table A2) presents
demographic characteristics related to the quartiles in Table
2.
33
Some of this mail is delivered by rural carriers. For purpose of
this analysis, we assume that rural delivery is subject to the same
degree of cream skimming as city delivery. However, some carrier
route mail is delivered on business routes and some is delivered
via post office boxes and, thus, would not be available to cream
skimmers. Consequently, 37.7 billion represents an upper bound of
the volume available to cream skimmers.
34
The only deviation from the monotonic behavior in Table 2 occurs
for deliveries per route between quartiles 2 and 3.
Table 2
SELECTED AVERAGES FOR RESIDENTIAL ROUTES WHEN ROUTES ARE SORTED
BY VOLUME OF CARRIER ROUTE MAIL PER ROUTE
Profit Cost (cents) Per (possible) Quartile (dollars) Piece
Delivery Stop
Volume (pieces) All Carrier Pieces Per (possible) Mail Route
Delivery Stop




F. Estimating The Delivery Cost For Cream Skimmers
Cream skimmers undercut prices charged by an incumbent that has
heterogeneous costs but averaged prices. Cream skimmers profit by
selling low cost items at below the incumbents average price.
Delivery cream skimmers would attempt to serve those areas where
the cost to deliver carrier route pieces is the lowest (i.e., where
the volume of carrier route pieces is the highest). The first
quartile of routes in Table 2 would make the best targets. It is
this case we examine.
In an important sense, this is an unrealistic case since
quartile 1 represents 40,000 routes which are scattered over more
than half the 30,000 5-Digit ZIP Code areas in the U.S. It is
extremely unlikely that they constitute all the residential routes
in any 5-Digit ZIP Code nor is it likely that they make up the
entire suburban ring around cities. It is unrealistic for cream
skimmers to serve such a fragmented market. From both an
operational and marketing standpoint, cream skimmers would have to
serve markets with at best a large proportion of routes in quartile
1. Their markets would undoubtedly contain a large number of routes
in the other quartiles. Notwithstanding, for analytical purposes,
we assume the best possible case for cream skimmers; that they
would be able to serve only routes that fall within quartile 1.
A potential competitor must decide the level of service to
provide and this decision would affect the amount of volume that it
could actually capture. Many magazines, newspapers, and
advertisements must be delivered within a one, two or three day
window. A competitor delivering only one day a week would have to
forego much potential volume. On the other hand, delivery six days
a week would certainly raise the cost per delivered piece and make
it less likely that a cream skimmer would be viable.35 City Mail in
Sweden chose to deliver two days per week (Monday/Thursday in some
areas and Tuesday/Friday in other areas).
In all probability, a cream skimmer could not prosper unless it
could deliver mail at a price less than the incumbent postal
administration. Its cost per delivered piece would be a function of
its cost of labor and productivity. It would be expected that a
cream skimmer would have a significant labor cost advantage. Its
productivity would be influenced by the in-office and out-of-office
technology it employed. In 1989, Postal Service carriers spent more
than 40 percent of their time in-office preparing for delivery on
the street. Many mailstreams must be merged and sorted into
delivery sequence.36 It is not clear what technology cream skimmers
would use. Moreover, we do not know whether the personnel of a
cream skimmer would be more productive on the street than
U.S. Postal Service carriers. This analysis will combine these
variables (labor cost, technology, in-office and out-of-office
productivity) into a single efficiency parameter.
35
In a sense, each piece of advertising mail competes with other
advertising mail for the recipient's attention. If a cream skimmer
could capture a large portion of the available carrier route
volume, its customers might prefer more frequent delivery to
minimize this type of competition.
36
This function is rapidly being automated for letter size mail in
the U.S. Delivery point barcoded mail is increasingly being
sequenced on sorting machines, saving significant in-office
delivery time. From 1989 to 1996, the percentage of in-office time
for city carriers has declined from 41 percent to 34 percent. It is
expected that this trend will continue. Unless a potential cream
skimmer had enough capital to employ sequencing automation, the
Postal Service might have a significant advantage in sequencing and
merging letter-shaped mail. First Class is almost all letter
shaped, publications are almost all flat shaped, and carrier route
advertising is 43 percent letter shaped.


G. Cream Skimmers Cost Per Piece
Table 3 has four matrices, each reflecting a different
percentage of available volume assumed to be captured by cream
skimmers. Each matrix presents cream skimmer unit or per piece
costs based on frequency of delivery and relative efficiency of the
cream skimmer. We measure relative efficiency in terms of cream
skimmers cost relative to the Postal Service's cost.
The shaded cell in the top matrix, 13.1 cents, is the Postal
Service cost per piece for delivering only carrier route presorted
mail in the least expensive quartile. This number is derived from
the 8.7 cent cost of the Postal Service delivering all mail in the
least expensive quartile shown in Table 2. If only the carrier
route mail is delivered, the average delivered volume decreases
from 3,051 to 1,263 in this quartile. The variable costs drop as
the volume drops, but the fixed costs are spread over less volume
which increases the fixed costs proportionately. The result is an
increase in the unit costs to
13.1 cents shown in Table 3. Going down that column, the costs
per piece are displayed assuming a cream skimmer had the same cost
as the U.S. Postal Service, but delivers fewer days per week. The
costs decrease by the amount of fixed costs in delivery. The
variable cost remains the same because the volume does not change.
Most other national postal administrations would have a greater
percentage of fixed costs because they have fewer pieces per
possible delivery.37
The columns display cream skimmers' costs as a function of U.S.
Postal Service efficiency, expressed as a percentage of the Postal
Service's delivery cost per piece. The 120 percent column displays
the cream skimmer's per piece cost if its cost were 120 percent of
the U.S. Postal Service.38 The 60 percent column displays the cream
skimmer's per piece cost if its cost were 60 percent of the U.S.
Postal Service.
Table 3
CREAM SKIMMER'S COST PER PIECE IN CENTS AS A FUNCTION OF
EFFICIENCY RELATIVE TO POSTAL SERVICE, DAYS OF DELIVERY AND PERCENT
OF CARRIER ROUTE MAIL
100% OF CARRIER ROUTE VOLUME CAPTURED

75% OF CARRIER ROUTE VOLUME CAPTURED

50% OF CARRIER ROUTE VOLUME CAPTURED

25% OF CARRIER ROUTE VOLUME CAPTURED

For all mail the Postal Service city carrier cost (fixed and
variable) is 12.5 cents per piece. (See Tables 1 and 2, cost per
piece all routes.) This is the cost for the delivery function. Each
postal product has a variable cost for mail processing,
transportation and delivery. Each product also has an average
incremental (avoidable) cost. In the case of advertising carrier
route mail, the average incremental cost is about 7 cents.39 If
threatened by cream skimmers, the Postal Service could respond by
lowering its price towards average incremental costs. If the Postal
Service maintained uniform prices but had competition only in
selected areas, it would sacrifice revenue in those areas without
competition. These circumstances would probably call for selective
discounts (with average incremental costs as a floor) to large
volume mailers40 who would otherwise become customers of cream
skimmers. Thus, cream skimmers would most likely have to have costs
no higher than approximately 7 cents per piece in order to compete
on a cost basis alone.41
The Postal Service's productive hourly wage of more than $25 may
make it likely that cream skimmers could obtain some cost advantage
on that basis alone. We do not know how reducing the frequency of
delivery would affect demand for alternative delivery. A two-day
frequency could satisfy a three-day window, but would miss many
delivery dates for time value publications. It should also be noted
that some customers may not be willing to leave the Postal Service
for a variety of reasons including the sense of security of dealing
with a government agency.
If a cream skimmer were to capture only 25 percent of the
available market, it would have difficulty succeeding. At that
level of market penetration, its costs would have to be 40 percent
of the Postal Service's at a two-day frequency in order to be below
Postal Service incremental costs. With 50 percent of the average
volume at a two-day frequency, a cream skimmer's costs would have
to be 60 percent or less of the Postal Service's in order to remain
below the Service's average incremental costs.
39
The variable cost of carrier route advertising mail comes
predominantly from the delivery function (in-office and
out-of-office). The delivery function is about 50 percent variable.
Thus, the variable cost of the advertising carrier route mail
product can be much lower than the average (fixed and variable)
cost of delivery. The average incremental cost of advertising
carrier route mail is only slightly higher than the variable cost
since we estimate it here to consist of variable costs plus single
subclass stop costs for 1996.
40
Except for a customer with a very high elasticity of demand,
there is no reason for a monopoly to grant volume based discounts
since customers have no alternative supplier. We understand this
was the case with UPS, which had a virtual monopoly in surface
parcel delivery in the U.S., and did not give discounts to surface
parcel customers until Roadway Package Service began to target its
most lucrative customers in the 1980s. UPS then began offering
discounts to certain large volume customers.
41
Some mailers currently use cream skimmers because of service
considerations.


H. Consequences Of Successful Cream Skimming
Table 4 shows the required increase in the U.S. Postal Service
rates based on the percentage of carrier route volume that would be
captured by cream skimmers targeting mail in quartile 1, Table 2.42
The maximum effect is an increase of about 1.2 cents in the
First-Class rate (with corresponding increases in the other
classes). This would be an increase of nearly 4 percent; a
significant but limited effect.
Lifting the letter mail monopoly and allowing cream skimming
would likely have an important impact on rate setting for the
Postal Service's large volume carrier route customers, if our
assumptions are correct concerning the amount of volume available
for capture by cream skimmers. The maximum impact on overall postal
rates is, however, likely to be limited. The negative impact on
Postal Service's finances would likely be offset to a great extent
by the effect of limited competition on the Postal Service's
efficiency, service performance and product innovation.
Table 4 IMPACT OF CREAM SKIMMING ON



I. Comparison Of Business And Residential Routes
We do not analyze cream skimming on business routes because only
7 percent of mail volume delivered on them is carrier route as
compared to 26 percent on residential routes. This makes them
unlikely candidates for cream skimming based on our analysis. We
compare several statistics for city residential and business routes
in Table 5.
Table 5 COSTS BY TYPE OF CITY DELIVERY ROUTE

It can be seen that business routes have slightly more pieces
per route with slightly higher cost per piece.43 The cost per
delivery point and stop on business routes is much higher, but the
profit per piece remains higher.44
We are not aware of any study that has identified the reasons
why business routes have a higher cost per piece and much higher
cost per stop than residential routes. Several factors may
contribute, including the high cost of serving each individual
business in office buildings. In contrast, residential apartment
houses generally have mail boxes in one central and convenient
location for carriers. Other factors include the larger proportion
of business mail requiring signatures and the greater difficulty of
getting around in congested business areas as opposed to
residential and suburban neighborhoods. Because business routes
have much higher cost per delivery point, they
43
Business routes get much more assistance from routers who
sequence the mail in the office, thus saving in-office time for the
carrier. The cost per piece and per delivery point reflects these
extra costs.
44
Profit is higher on business routes than on residential routes
because of greater volumes of high profit mail, such as on Priority
and Express. Cream skimming delivery networks might be developed
for such mail in business areas, but are beyond the scope of this
paper.
are unlikely candidates for geographic cream skimmers who could
be expected to have far fewer pieces over which to spread the cost
per delivery point.45


J. Some Differences Between the U.S. and Other Countries
The classification structures in industrial countries have
varying degrees of efficiency (defined as adherence to cost).46
Efficient classification discourages cream skimmers because it
leaves less cream to skim. The less efficient the classification
structure, the more the administration needs monopoly protection.
The U.S. Postal Service's classification structure contains several
score of cost based discounts and rate differentials, probably more
than any other administration.47 This should tend to make the U.S.
less vulnerable to cream skimming. Moreover, the U.S. has more
pieces per capita than virtually any other country. Because the
delivery function has important economies of scale, ceteris
paribus, the U.S. should have lower per piece costs than virtually
every other country. This factor also makes the U.S. less
vulnerable to cream skimming. In addition, the proportion of fixed
cost in the delivery network is largely a function of volume per
capita (per delivery). The larger the volume per capita, the larger
the proportion of variable costs and the smaller the proportion of
fixed costs. Since virtually every other postal administration has
less volume per capita than the U.S., a larger proportion of their
delivery costs are fixed relative to the U.S. A cream skimmer can
reduce its fixed cost relative to an incumbent simply by reducing
the number of days of delivery. Thus, a cream skimmer would obtain
less cost advantage by cutting delivery frequency in the U.S. than
in countries with less per capita volume. This also tends to make
the U.S. less vulnerable to cream skimming than other
countries.48
45
Of course, messenger services are common in central business
districts, but this is for reasons of service and not cost.
46
The most efficient classification structure would be based on
costs alone and would offer worksharing discounts wherever
practical. These discounts would equal the cost savings. No postal
administration meets this standard nor do the authors recommend
that they do.
47
Advertising mail alone has over 60 different rate categories
with rate differentials based on shape and discounts for
presorting, barcoding and drop shipping.
48
A consequence of this is that an administration which delivers
five days per week, ceteris paribus, is a more difficult target for
cream skimmers than one which delivers six days per week.
On the other hand, some factors appear to make the U.S. Postal
Service more vulnerable than other postal administrations to cream
skimming. Few, if any, countries have as large a proportion of
carrier route volume as does the U.S. This is the mail most
susceptible to diversion. In addition, few postal administrations
pay their employees as large a wage premium as does the U.S. Postal
Service.49 A wage premium allows cream skimmers to obtain an
efficiency/cost advantage simply by paying the prevailing wage.
49 See Cohen, Chu, Ferguson & Xenakis, "A Cross Sectional
Comparison and Analysis of Productivity for 21 National Postal
Administrations." Managing Change in the Postal and Delivery
Industries, Ed. Crew & Kleindorfer, Kluwer Academic Publishers,
1997.


APPENDIX
This paper's analysis is based on volume and delivery statistics
from 13,212 residential carrier routes. These routes were extracted
from the Carrier Cost Survey (CCS), conducted by the U.S. Postal
Service in 1989. In addition to the volume and delivery statistics
for each of the 13,212 residential routes, CCS provides the
associated 5-Digit ZIP Code served by each route. A ZIP Code may be
served by many routes. The data presented in this appendix are
based on the demographic data for the 5-Digit ZIP Code for each
route in a quartile. Thus, the demographic data for the same
5-Digit ZIP Code could be averaged into the totals for two
different quartiles. This would happen if two routes from the same
5-Digit ZIP Code were in separate quartiles.
The United States Bureau of the Census has grouped the 1990
Census of Population and Housing data using 5-Digit ZIP Codes. The
file (STF 3B) is available in three CDs of about 1.4 gigabytes
(GBs) and contains socio-economic data for 29,467 5-Digit ZIP Codes
served by the U.S. Postal Service. The Puerto Rico ZIP Codes are
not included in STF 3B file.
By merging the STF 3B file with CCS data, we have obtained
economic and demographic data for the ZIP Codes in which 12,876
residential routes reside. We could not obtain any socio-economic
information for the ZIP Codes of 336 residential routes (most of
them in Puerto Rico) because their ZIP Codes were not included in
the STF 3B file.
Tables A1 through A3 provide statistics for age, household
income, and education attainment for the 5-Digit ZIP Code for the
12,876 residential routes in CCS. Table A1 presents the statistics
by quartile when routes are sorted by profitability. Tables A2 and
A3 present the statistics by quartile and decile respectively when
routes are sorted by carrier route volume. The statistics in Table
A1 show that, on average, routes in the more profitable quartiles
reside in ZIP Codes with higher income households and more educated
adults. Likewise, the statistics in Tables A2 and A3 show that
routes with high carrier route volume reside in ZIP Codes with
higher levels of household income and education attainment.