usps-test-t31.pdf

RECEIVED
Jnn12 3 03PH‘00
USPS-T-31
BEFORE THE
POSTAL RATE COMMISSION
WASHINGTON,
D.C. 202680001
POSTAL RATE AND FEE CHANGES, 2000
DIRECT TESTIMONY
OF
ANTHONY M. YEZER
ON BEHALF OF
UNITED STATES POSTAL SERVICE
Docket No. R2OOC-1
USPS-T-31
CONTENTS
B%Fs
Autobiographical
Pages ii - iii
Sketch
I.
Purpose and Background
Pages 1 - 3
II.
Modeling
Pages 3 - 9
III.
Data Preparation
IV.
Estimation
Approach
Results
Adopted
Pagesg-12
Pages 12 - 13
Page i
USPS-T-31
Page ii
Direct Testimony
Df
Anthony M. Yezer
Autobiographical
5
M. Yezer.
I am currently
University
and special consultant,
6
at the George Washington
7
Economic
8
since 1985
9
Professor
IO
Washington
University
11
economics.
During my tenure at George Washington
12
as a consultant
13
government.
14
projects,
15
sponsored
Research Associates,
and special
(I 972-l
976)
n/e/r/a.
consultant
since 1995.
and Associate
I have also worked
i.e. research
performed
including
985)
at George
on regional and urban
University,
I have served
agencies of the United States
substantial
sponsored
include
Fellowship,
an National
18
and finalist status in the Rhodes Scholarship
19
states.
Athletic Association
University
Defense
Education
Scholar-Athlete
competition
but
since 1991.
I currently
Act
Fellowship,
for the middle Atlantic
I have been a fellow of the Homer Hoyt School of Advanced
Real Estate and Urban Economics
research
including the National Science Foundation.
17
National Collegiate
I was Assistant
at and by George Washington
My
21
awards
(1977-l
concentrates
on several
by external organizations,
academic
of Economics
to the National
Previously
Professor
where my teaching
to many organizations,
Professor
I have been a Professor of Economics
16
20
--
My name is Anthony
Sketch
Studies in
serve on the editorial
1
boards
of four journals,
2
American
Real Estate and Urban Economics
3
I have published
4
papers in other venues.
5
urban economics.
In 1966,
6
and edit the monograph
an M.Sc.
over thirty
earned
in economics
8
Political Science
9
from the Massachusetts
in 1967.
articles in refereed
a B.A. degree from
from the London
My doctoral
Institute
Page iii
sponsored
by the
Association.
journals
Much of this research concentrates
I was awarded
7
series
USPS-T-31
degree in economics
of Technology
on real estate and
Dartmouth
School
and numerous
College and
of Economics
and
and urban studies
was conferred
in 1974.
USPS-T-31
Page 1
Direct Testimony
Of
Anthony 1111.
Yezer
Estimation
of Postal Service Facility Rental Costs
I. Purpose and Background
6
Pi
The purpose
project
conducted
is to present
by the Center
results
from
7
research
8
Department
9
Postal
10
appropriate
11
procedures
12
rental value of space used by the Postal Service to provide
13
service.
14
space
15
States.
16
data on property
17
real estate rents.
18
,--
of this testimony
for Economic
of Economics at the George Washington
Service
and, ultimately,
fees for post office
provision
Research
University
box service.
in the
for use by the
by the Postal Rate Commission
used to provide a statistically
I understand
a sponsored
This testimony
in setting
explains
valid means of predicting
the current
post office
that the Postal Service will use my results to allocate
costs to post office
boxes located
the
throughout
box
its
the United
The academic novelty of this project is the creation from Postal Service
leases the first standardized
nationwide
The first task of this project was to construct
19
determinants
20
States.
21
facilities
of rent per square
estimates
a statistical
foot at postal facilities
across
of current
model of the
the United
A second task was to predict expected rent per square foot for specific
assuming
that the leases were signed for a specific
term at a given
USPS-T-31
1
time.
The research
2
until September,
3
delivered
was undertaken
1999
when
1998
and continued
rents per square
foot was
to the Postal Service.
5
the academic
6
problem,
7
available.
8
in selected cities.
9
with that literature,
10
scale because predictions
11
of the effort.
literature.’
rents in real estate markets
Substantial
concentrating
on rented
A significantly
statistical
effort
residences
where
is far from new in
has been applied to this
massive
model developed
but the overall estimation
effort is conducted
to differentiate
the predicted
characteristics
15
even if the rental price for identical
16
square foot generally decreases with facility size. Therefore,
17
equal, we expect the predicted
18
square
19
estimates
20
alternative
21
can be used to predict either rent per square foot of specific
of rental
price
differences
of postal facilities differ, predicted
in the facikty.
product
rent per square foot reported
14
over space.
Because
the
rent per square foot can vary
units is constant.
For example,
rent per
other things being
rent per square foot to fall with the number of
In contrast,
a rental
of the rent per square foot for a standard
locations.
at a far larger
for the entire United States are a necessary
here from
feet
office rents
in this research is consistent
13
an index
data sets are
smaller group of papers has considered
The predictive
It is important
12
August,
a set of predicted
The problem of predicting
4
beginning
Page 2
While the econometric
price
index
is based on
size and type of facility in
model developed
in this research
facilities
or an
?
USPS-T-31
1
index of rental price differences
2
former for presentation
3
over space, the Postal Service chose only the
to the Postal Rate Commission.
The next section of this testimony
4
and estimate
5
section discusses
6
section,
an econometric
addresses the approach taken to develop
model of facility
rent determinants.
data sources, and data preparation
estimation
Page 3
The third
procedures.
In the fourth
estimates
of the rent per
results are discussed.
7
8
9
.-.
I
II. The Modeling
Approach
The objective
Adopted
in thii Study
of the modeling effort is to produce
10
square foot at Postal Service facilities,
particularly
11
post office box service.
taken to modeling
12
follows efforts
13
and property
14
and previous
15
space rented,
16
payments
17
space, and the physical
18
factors,
19
The approach
in the literature to account
characteristics.
research,
include
measures
location
of annual
and term
and maintenance
costs,
of the property.
and vary with market conditions
rent per square foot
for rents in terms of lease provisions
Data on leased properties,
date of endorsement
of utility
those facilities that provide
used in both this study
rent paid, square
of the lease, provisions
physical
characteristics
the time when
21
mechanism
the last lease was
in the model to project
for
of the
Rents are based on all these
at the time that the lease is signed.
Because this study must project rent per square foot into the future,
-,
20
feet of
endorsed,
it is necessary
rents per square foot forward
beyond
to provide
based on
a
USPS-T-31
1
recent market trends.
2
on variables
3
example,
4
such as the Consumer
5
providing
6
forward,
7
estimation
8
any explanatory
9
of future
IO
Therefore,
that are observable
it is important
at the time that the prediiions
projections
Price Index, or the Producer
of future
thus introducing
process.
rents also requires
Accordingly,
Price Index.
projecting
the model developed
variables that require projection forward
For
However,
these
indexes
in the rent
here does not include
to produce predictions
rent per square foot.
The variables
used in the statistical
12
used to get them ready for statistical
15
are made.
a potential source of error and controversy
from Postal Service data.
14
be based
the model could have included readily available national price indexes
11
‘13
that the predictions
Page4
this testimony.
analysis presented
A detailed discussion
here are all taken
of the data and the procedures
analysis is provided in the next section of
It is useful to group the variables
into the following
general
analysis is the quotient
of total
categories.
1. The dependent
variable
in the statistical
16
annual rent for the entire facility divided by size of the rented space in square
17
feet, for facility
18
j; it is useful to refer to this variable as R/SQFTj.
2. A vector of seven dummy variables, noted as M,, reflects responsibility
19
maintenance
20
sewage,
21
Service is responsible
and utilities
snow removal,
including:
maintenance,
and custodial
services.
electricity,
heating,
for
trash,
To the extent that the Postal
for providing these functions,
the rent should be lower.
-,
,-,
USPS-T-31
1
2
setting
3
front,
4
effect on rent per square foot.
5
higher cost.
6
.-.
including:
business
park, office
general retail, supermarket,
noted as S,
building,
mall, shopping
and other setting.
For example,
indicating
facilities
Setting
the faciiii
center,
store
should have an
within malls are generally
4. Variables that indicate the time at which the current lease was endorsed.
7
Time is measured annually
8
time measure and the other, T95i, is equal to zero before 1995 and equal to
9
time thereafter.
10
and hence the effect of time, Ti, should be positive.
11
time variable, T95i, is the difference
12
the entire period and the annual rate of increasesince
13
positive
(negative)
14
A third
endorsement
15
endorsement
16
_-.
3. A vector of eight dummy variables,
Page 5
with 1960 = 1. One time variable, Ti, is simply the
It is expected
that rents have risen over time in most markets
between
the annual rate of increase over
if rents have been increasing
time
variable,
The effect of the second
1995,
which
faster (slower)
DTi, is. a dummy
could be
since 1995.
indicating
that
the
date is missing.
5. Variables that reflect lease length, T4 and 0%.
level payments
to compensate
It is expected
17
leases have higher
18
inflation.
19
equal to one if the lease is less than two years or length was missing.
20
term leases likely reflect special needs, including
21
at facilities
Lease length in years is measured
by %,
owners
that longer
for the effects
DTLi is a dummy
seasonal demands
that are unlikely to provide post office box service.
of
variable
Short
for space
USPS-T-31
1
6. Dummy
2
branch offices,
3
interior
4
Lli is also included.
5
are not clear, but there is a strong expectation
6
fall with increasing
7
at a decreasing
8
of interior
9
literature.
10
variables
that indicate
physical features
such as:
Bi; presence of a loading dock, 4; and missing information
space, DI,. The natural logarithm
7. Two
of the facility
Page 6
on
of the square feet of interior space,
The effects of the dummy variables on rent per square foot
that cost per square foot should
facility size. Because rent per square foot should decrease
rate as size increases,
the estimation
space.
The use of this functional
dummy
variables,
is based on the logarithm
form is not uncommon
NoPi and Pi, that reflect
specific
11
parking space in the lease. The first dummy variable indicates
12
provision
for parking
13
indicates
that
14
thousand
square
15
provides
16
parking variables
17
8. A vector
parking
feet.
or that
provision
is provided
Compared
but total
should be associated
Two different
L,+ that
18
facility.
19
foot equations.
20
largest MSAs),
21
center of the central business district
parking
to the reference
for more than one thousand
of variables,
is missing
provision
for
,-
that there is no
and the ,second
space
in the
variable
is less that
case in which
square feet of parking,
one
the lease
both of these
with lower rent per square foot.
measures
the physical
location
of the
versions of these variables appear in the rent per square
For facilities
located
within
larger MSAs
location was based on the distance between
(specifically
the 65
the facility and the
(CBDJ as well as distance
north-south
or
?
USPS-T-31
1
east-west
2
automatically
3
Map Info.
4
CBD although
5
For facilities
6
groups of states where
7
location
8
that the county
9
county
10
quantitatively,
11
Essentially
12
activity
The CBD was located
using geographicinformation
The expectation
computed
systems (GIS) software-specifically
this effect may not be significant
from the
in cities with multiple
centers.
located outside larger MSAs, the data were grouped by state or by
grouping
was characterized
was needed to increase sample size.
by dummy variables for individual counties
had at least ten facilities.
dummy
variables
when
Experimentation
were generally non-significant,
the
number
of facilities
in the
Then
provided
indicated
that
either statistically
county
rents tend to be low unless there is some concentration
which
estimate
and distances
is that rents should decline with distance
was
or
small.
of economic
tends to increase Postal Service activity.
Based on the previous discussion,
13
14
from the CBD.
Page 7
the general form of the equation
used to
rent per square foot is given below:
15
16
I)
R/SQn;.
= a,, + t,=,
17
QnTb
+ e,DTLj
18
Lz~1~Jicj + $
u,M, + I+,
+ O,Bj + B&j
/3,S,
+ 0,T; + 0,,T95i
+ f&DTj
+
+ e,,lDj + 8uLli + YN,NOPj + YJ’j +
19
20
-,
21
The variables
endorsement
reflect: utility and maintenance
(M,), facility setting (S,), lease
time (Ti, T95,, and DT,), lease length (TLi and DTJ), branch
(9i),
USPS-T-31
Page 8
1
loading dock Dj, interior space (IDi and Ll,), parking (NoP, and P,), and location
2
(L,,J. $ is an identically
3
other Greek letters -indicate parameters
4
used to predict
5
measured
6
and independently
random error term.
to be estimated
rent per square foot for facilities
statistically
with specific
The
and are
characteristics
by the variables.
Equation
(1) includes a number of dummy variables
7
in cases where a particular
8
Such variables
9
that independent
10
technique
11
practice
12
record are missing or take on improbable
13
distributed
independent
variable is missing and zero otherwise.
only appear in equations
variable are actually
where at least some observations
missing.
Measurement
to include observations
all observations
variables
reflects a general
even if some items in the data
values..
error in the dependent
by including
variable,
14
resolved
15
deletion of facilities
16
are not really rental facilities and the certainty
17
market rents.
18
are addressed
19
Specifically,
20
statistical
software
21
constructs
ordinary least squares estimates
R/SOFT,, in equation
for which
R/SQFTr
> 0.
by the
the robust
regression
package,
robust
error in the dependent
regression
technique,
version
that these
that these rents could not reflect
issues of measurement
use of
(1) is
Casewise
with rent equal to zero is based on the possibility
Additional
on
The use of this dummy variable
to deal with missing values of independent
of attempting
which are equal to unity
5.0,
estimation
rreg, available
is used.
of equation
This
variable
techniques.
in the STATA
technique
first
(1) and then deletes any
.-
USPS-T-31
1
ownership
2
as leased and 11,608
3
could not be determined.
4
!
Examination
rent and lease terms,
27,467
were classified
as owned, while 434 were so incomplete
that their status
of the lease records indicated
that there were cases in which
5
multiple leases had been endorsed for the same physical address.
6
lease cases were examined
7
various aspects
8
for a branch
9
and/or an annex attached
10
rents associated
11
parking,
12
facilii,
13
Of course,
14
lease that covers the branch, loading dock, parking,
etc.
15
facilities,
Presumably
16
acquisition
17
with separate
18
activity.
19
--
and fields containing
Page 10
and found to reflect
of a single facility.
office,
For example,
separate
with these individual
Furthermore,
it is evident
leases are not independent.
that the
Rent for the
to rent for the main
that facility rent includes an implicit payment for parking, etc.
the vast majority
multiple
records
or construction
Accordingly,
for
parking at the branch,
loading dock, or annex is often trivial compared
indicating
lease contracts
separate leases might appear
the loading dock of that branch,
to the branch.
The multiple
of facilities
were
have one rental payment
not uncommon.
of different
for a single
Similarly
parts of the facility
records entered for each phase of the acquisition
for owned
this
at different
reflects
times
or construction
multiple leases at the same physical address are combined,
20
the lease payments
21
physical characteristics,
are summed
to a total
facility
lease payment
such as square feet of space, are aggregated.
i.e.
and the
This
USPS-T-31 Page 11
1
results
2
records are eliminated
3
facilities
4
of the same facility)
5
407 different
6
in a total
(1,611
of 25,606
distinct
by combining
records are eliminated
redundant
leases for a single facility),
based on ownership
lease
9,997
owned
of different
aspects
to reflect only
locations.
used in the statistical
analysis
are all taken,
directly
or
from the records in this data set. In some cases, such as square feet
7
indirectly,
8
of space, the data record includes duplicate
9
the variable is recorded
10
indicating
11
logically
12
feet of space, then the third field, indicating
13
in the previous
14
in which
15
equal to zero.
16
(1,801
and the 434 unusable records are combined
physical
The variables
leased facilities
in more than one field.
a measure,
precluded.
Where there are multiple fields
the median value is used except
section,
when
the median is
a series of dummy
positive space, is used.
variables
As noted
is used to indicate
cases
was missing or took on an illogical value, e.g. lease length
of the
system
i.e.
For example, if two of the three fields indicate zero square
a variable
Location
measures of a given variable,
facility
is determined
17
information
(GIS) to the physical
18
facilities
19
is based on county.
20
(Map Info) is used to locate the latitude
21
using the latitude
by application
of a geographic
address of the facility.
located outside the 65 large MSAs identified
In the case of
in this analysis,
location
For facilities in one of the 65 large MSAs, the GIS program
and longitude
and longitude
of the facility.
Then
of the center of the CBD for that MSA, as
USPS-T-31 Page 12
1
determined
2
facility
3
computed.
Note that, for facilities located south of the CBD center, north-south
4
is negative
and similarly the value of east-west
5
CBD center
6
I
in the GIS program, the radial distance from the CBD center to the
Then the distances
is computed.
north-south
for facilities
and east-west
are
to the west of the
is negative.
The time measure used in this analysis is annual and time is measured
in
7
years with 1960 equal to 1. For observations
8
such as time since lease endorsement
9
measure is set equal to zero but a dummy variable equal to unity when time is
10
missing and zero when it is present is included in the data set to distinguish
11
these observations,
12
casewise
deletion
of leases where a time measure
or lease length
This is consistent
is missing,
with my general practice
the time
of avoiding
as a response to missing observations.
13
14
I
IV. Results from Estimates of the Rent per Square Foot Equation
15
This section
discusses the results of estimating
16
II of this testimony
17
discussed
18
routine
19
models fitted to the 65 largest MSAs and the “non-metro”
20
location
21
by county
using the 25,606
in section
from
STATA
Ill.
The estimation
version
is characterized
for the latter.
observations
5.0.
technique
The estimating
Equation
(1) from section
of Postal Service
is the robust
equation
differs
leases
regression
between
state areas beCaUSe
by precise distance measures for the former case and
USPS-T-31
1
I gave close consideration
to determining
2
the data into geographic
3
is determined
4
geographic
5
makes uniform the effects of the independent
6
thereby
7
against extreme
8
with sample size. Thus the final level of geographic
9
these estimates
10
sizes below
11
estimates
12
But smaller states and MSAs are grouped
13
areas.
the best method
On the one hand, because
at a local level, one could estimate
areas. Moreover,
suppressing
estimating
the spatial variation
geographic
variables
equations
However,
disaggregation
results.
in equation
separately.
in order to enlarge sample sizes.
14
all variables
(I) into the regression.
15
are not appropriate
16
variable
17
data.
18
dummy
19
present always
20
and utilities
21
Alaska and South Carolina, the robust regression
is not appropriate
sample
As a result,
results begin with an attempt
for a given equation.
to fall
adopted in
in which
for large states and very large MSAs are often performed
areas, estimation
mitigating
for precision
is based on a round of initial experimentation
For all geographic
for small
across large regions
is the tendency
200 were judged to yield rather imprecise
pricing
on rent per square foot,
being examined.
disaggregation
for aggregating
real estate
separate
a single equation
Page 13
However,
The most
to force
some of the variables
common
is that the measure is not present
reason
that a
in a subset of the
For example, time is not always missing and hence a time zero or missing
variable
is not always
and, occasionally
appropriate.
Some of the setting
some of the lease provisions
are not present - e.g. snow
removal
in Florida.
types
are not
for maintenance
For two
states,
results do not converge
USPS-T-31
1
2
properly
and ordinary
The econometric
3
be reviewed
4
agreement.
lease squares estimates
estimates,
supporting
by a party requesting
Page 14
are used.
data, and resulting
,access and willing
rental values can
to sign a non-disclosure