Full text of DOI:10.1080/01944360108976250

This article was downloaded by: [The University of Manchester Library]
On: 19 November 2014, At: 23:58
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK
Journal of the American Planning Association
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/rjpa20
Changing Water and Sewer Finance: Distributional
Impacts and Effects on the Viability of Affordable
Housing
Dick Netzer , Michael Schill & Scott Susin
Published online: 26 Nov 2007.
To cite this article: Dick Netzer , Michael Schill & Scott Susin (2001) Changing Water and Sewer Finance: Distributional
Impacts and Effects on the Viability of Affordable Housing , Journal of the American Planning Association, 67:4, 420-436, DOI:
10.1080/01944360108976250
To link to this article: http://dx.doi.org/10.1080/01944360108976250
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

Changing Water
and Sewer Finance
Distributional Impacts and Effects
on the Viability of Affordable
Housing
Dick Netzer, Michael Schill, and Scott Susin
Usage-based charges finance an in-
creasing share of water and sewer
costs in the U.S. The shift has been es-
pecially sharp in New York City, where
residential users traditionally did not
pay usage-based charges. In this arti-
cle, we examine the impact that a tran-
sition to universal metered billing
would have on multifamily housing.
We find that universal metering would
result in large increases in water and
sewer bills for many multifamily build-
ings in low-income neighborhoods.
Comparing the predicted increases
with the net operating income of land-
lords in low-income neighborhoods
indicates that universal metering might
pose a threat to the viability of much
affordable housing.
rior to the mid 1980s, the City of New York financed its water supply
and wastewater treatment systems without attempting to make these
systems financially self-sustaining or to employ usage-based charges
P
to encourage conservation. The operating costs of the water supply system
were met from water charges, which in the case of business structures were
based on meter readings, and in the case of residential structures on “front-
age” charges, a flat rate based on a formula that takes into account street
frontage, the number of housing units and the number of water fixtures in
the building.1 Operating costs of the sewer and wastewater treatment system
were met from usage-based charges set at a specified fraction of water
charges. Capital costs were financed by general obligation bonds, but from
the late 1960s until the 1980s, federal and state government grants covered
80% of the capital costs.
Since the mid 1980s, New York City, like many other cities, has changed
the way it finances its water and sewer systems. It now finances these expen-
ditures (including debt service on bonds issued to finance capital outlays) en-
tirely from usage-based charges. For all U.S. local governments, between
1972 and 1992 revenues from usage-based charges for water increased from
42% to 87% of current water supply operating and debt service costs; rev-
enues from usage-based charges for sewers and wastewater treatment in-
creased from approximately 38% to 77% (U.S. Census Bureau, 1974, 1997).
The result in New York City (as in many other places) has been very large in-
creases in water and sewer rates: a nearly 400% increase from 1979 to 1998
and a 300% increase from 1985. During the same period, the cost of living in
the New York metropolitan area rose by only 111% (U.S. Bureau of Labor
Statistics, 1999). Similarly, between 1981 and 1996 median rents, as mea-
sured by the New York City Housing and Vacancy Survey, rose by 161% (Lee,
1999).
Netzer, AICP, is a professor of economics
and public administration at the Wagner
Graduate School of Public Service, New
York University. Schill is a professor of law
and urban planning at the School of Law,
New York University. Susin is a postdoc-
toral fellow at the U.S. Census Bureau,
HHES Division; when this article was
prepared he was a research fellow at the
Center for Real Estate and Urban Policy,
School of Law, New York University.
Journal of the American Planning Association,
Vol. 67, No. 4, Autumn 2001. © American
Planning Association, Chicago, IL.
N
N
420 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
Moreover, the City is in the process of installing
meters in all residential buildings, partly because of the
drought of the mid 1980s (and forecasts that the safe
yield of the water system would provide only a small
margin over increasing use), but mostly in response to
litigation and court consent decrees designed to reduce
wastewater flows through the treatment plants. Under
New York’s metering program individual apartment
units would not be metered; instead one meter would
service the entire building.2 However, at present a series
of transitional measures have been put in place, and
most owners of multifamily buildings continue to pay
frontage rates.
The practice of charging consumers of public-util-
ity-type services prices that reflect the marginal costs of
the resources consumed has increased in recent decades,
in part because of the obvious advantages in conserva-
tion of resources. Ordinarily, this imposes relatively high
costs on low-income people, a drawback that has gener-
ated “lifeline” pricing for electric and telephone service,
under which modest levels of consumption are billed at
very low rates, with marginal-cost pricing only for
amounts beyond those levels. In this article, we focus on
the distributional impact of a shift to charging for water
and sewer service based entirely on actual water use mea-
sured by meters. In particular, we examined what the im-
pact of universal metering in New York City would be
on low- and moderate-income housing. We found that,
despite its possible positive effects on conservation, uni-
versal water metering would have a substantial and re-
gressive impact on both the providers and consumers of
the city’s low-income housing.
Unaffordable rents do not necessarily translate into
profitability for the city’s landlords. While owners of un-
regulated buildings in such areas as the core of Manhat-
tan may earn high rates of return, for landlords in many
of the city’s low-income communities real estate is not
an easy way to earn a living. Despite an upsurge in prop-
erty values since the early 1990s, in 1996 over 4% of the
residential buildings in the city had lien-to-value ratios
in excess of one (Scafidi et al., 1998). Since the mid 1980s,
owners have lost more than 100,000 units of housing to
the City for nonpayment of taxes. The reasons for dis-
tress among landlords are numerous and include exces-
sive debt, rent regulation, inefficient management prac-
tices, and the inability to raise rents to a level that would
cover operating expenses (Braconi, 1999).
It is worth noting that outside of core Manhattan
rent control and rent stabilization have little effect on
rents, since the controlled rent is often above the mar-
ket rent. Pollakowski (1997) found that the median reg-
ulated unit rented for just 7% less than comparable un-
regulated units, with smaller effects in poorer areas.
For some buildings in the city, prevailing rents are so
low that they do not cover the day-to-day expenses of op-
erating a building, much less operating expenses plus
debt service. For example, a report by Arthur Anderson
(1995) estimated that with respect to City-owned build-
ings, the rent collected from tenants amounted to only
half of the expense the City incurred in operating the
properties. Furthermore, according to data obtained
from the New York City Department of Finance, just
over 10% of building owners who challenged their as-
sessments in 1997 had net operating incomes that were
negative.3
The Affordable Housing Problem
in New York City
Transition to Universal Water
Metering
The major housing problem facing residents of New
York City concerns affordability. Despite the fact that
well over one half of the city’s housing stock is subject
to rent control or rent stabilization, the number of
households paying an excessive proportion of their in-
come for rent is staggering. According to the Housing
and Vacancy Survey (conducted by the U.S. Census Bu-
reau every 3 years since the 1950s), for nearly 526,000
households, or approximately one quarter of all renters,
rent consumed more than one half of their incomes in
1996 (Lee, 1999). For an additional 68,000 homeowners,
housing-related expenses consumed more than 60% of
their incomes. Close to 95% of all renter households with
these severe housing-cost burdens had incomes below
half of the area’s median income. Among owners, the
comparable proportion was 69% (Schill & Scafidi, 1999).
New York City’s transition to universal water me-
tering has not been an entirely smooth one. Initial expe-
rience with the new method suggested that water bills
would rise rapidly for some segments of the city’s hous-
ing stock. Mistakes in billing further fueled the belief
among owners that allowing meters to be installed in
their buildings would generate difficulties. The City re-
sponded to these concerns by enacting a transition pro-
gram that froze water rates for a couple of years and per-
mitted building owners to choose between paying by the
existing frontage-rate method or by metered use.
As Table 1 indicates, the majority of multifamily
rental buildings in New York City have chosen not to be
metered. There are 87,051 properties in New York City
with four or more housing units. Complete data were
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 421

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
TABLE 1. Multifamily buildings in study database.
obtained for 70,328 of these buildings. As of 1998, al-
most two thirds of the buildings—45,457 properties—
had no meter and their owners were being billed for
water at frontage rates. An additional 4,864 buildings
had meters, but the owners have chosen to continue to
be billed at frontage rates. Owners of just over 22% of the
properties paid usage-based, metered charges, and an
additional 11.8% paid a combination of metered and
frontage-rate charges.
Building type
Number
Buildings with four or more units
Buildings with complete dataa
Jamaica buildings
87,051
70,328
1,101
Balance of New York City buildings
Paying frontage rate
69,493
45,457
4,864
15,763
8,273
Metered, but paying frontage rate
Paying metered rate
Since the City’s intention to move to universal me-
tering for water and sewer charges was announced, own-
ers, tenants, and mortgagees have been concerned about
the impact such a change would have on the city’s af-
fordable housing stock. In particular, it has been argued
that low- and moderate-income housing will experience
disproportionate price increases because (1) the build-
ings and plumbing are older, making the probability of
leaks much higher; (2) low- and moderate-income ten-
ants are more likely to be home during the day using
their plumbing than middle- and upper-income tenants
and homeowners; and (3) the incidence of “doubling
up,” that is, occupancy of the housing unit by two or
more households, is likely to be higher in low- and mod-
erate-income households, thereby increasing usage.
The concerns of the affordable housing community
have received some support over the past decade. In 1992,
the Department of Environmental Protection (DEP)
monitored water use in several hundred buildings. In one
area of the Bronx, average bills were estimated to rise by
131% under metering as compared to frontage rates; in a
Queens community, however, average bills would in-
crease by less than 10%. Data provided by certain owners
and managers also showed substantial increases (Com-
munity Preservation Corporation, 1997).
In 1994, the DEP and the New York City Rent
Guidelines Board commissioned a report by Speedwell,
Inc. on the impact of metering on housing in New York
City. The author of the report (Blackburn, 1994) simu-
lated water use in buildings throughout the city, based
on a sample of 590 properties. He concluded that if water
use remained at current levels (i.e., no conservation by
users in response to metering) and if billing caps were re-
moved, approximately 23% of all buildings would have
annual metered bills of $500 or more per unit, 8% would
have bills of $750 per unit, and 3% would have bills ex-
ceeding $1,000 per unit. Once certain assumptions
about conservation were made, the author of the report
concluded that the average metered bill would be $300,
which is $10 less than the average bill at frontage rates.
However, even with conservation, 17% of all buildings
would have bills in the maximum range of $500 to $600
per unit. The report also found that increases were most
likely to occur in low-income neighborhoods.
Paying frontage & metered rates
Sources: New York City Departments of Environmental Protection
(DEP) and Finance.
a. Most buildings with incomplete data lacked it because
1) they could not be matched to the DEP data, or
2) they were matched but did not have water charges, or
3) the periods to which the water charges referred were not
ascertainable.
In addition, some buildings
1) lacked complete data in the Real Property Assessment Data
(RPAD) file, or
2) had extreme values of variables in the RPAD file, or
3) could not be matched to the census tract data.
Estimating the Impact of Water
Metering on Affordable Housing
For our study, estimates of the effect of metering on
water and sewer charges were made for multifamily
rental buildings in New York City. Two data sets on 1998
water use and combined water and sewer charges for par-
ticular groups of buildings in the city were used to esti-
mate the change in water and sewer charges per housing
unit that would be likely to occur if all such buildings
were to be billed on the basis of meter readings, as only
one fourth of the buildings were at the time of the study.
The first data set consists of 4,864 buildings that have
installed water meters that register water use, but that
still pay at frontage rates. These buildings are known as
“monitor-only-meter” (M-O-M) buildings, because the
meters are read, but not used to determine water and
sewer charges. Thus, the readings can be used to calcu-
late what their bills would have been in 1998 if these ac-
counts actually had been charged on the basis of the
meter readings. As more fully described below, by using
housing and real estate tax data, we ascertained the char-
acteristics of these buildings and their neighborhoods.
Through regression analysis of M-O-M buildings, we
were able to obtain estimates of the impact of these
neighborhood and building characteristics on water and
sewer charges, which could then be applied to all build-
ings in the city. Thus, these estimates for M-O-M build-
N
N
422 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
ings provide an approximation of water and sewer
charges in 1998 for all buildings if they had been charged
for actual water use. They reflect short-term changes, es-
timating what would happen to bills before building
owners had any opportunity to make plumbing im-
provements designed to conserve water.
water metering on affordable housing. In particular, we
were interested in the likelihood of severe impacts—the
possibility that some buildings would incur much
higher charges under metering.
Before discussing the details of the regression meth-
ods we used, it is worthwhile illustrating our empirical
strategy with simple unadjusted estimates. The pre-
dicted charges are the percentiles of metered bills for the
Jamaica buildings and metered use for the M-O-M build-
ings multiplied by the applicable rate. Current charges
are the simple percentiles for buildings currently paying
frontage rates. Change is the difference between the two.
These figures make no adjustment for the fact that
the M-O-M and Jamaica buildings do not have the same
characteristics as the frontage-rate buildings to which
they are being compared. For example, the Jamaica
buildings tend to be somewhat newer than buildings in
the rest of New York. Quantile regression techniques
allow us to adjust these figures for the observable char-
acteristics of these buildings and the census tracts in
which they are located.4 Specifically, we first estimated
quantile regressions in the Jamaica and M-O-M build-
ings samples. Then we calculated the predicted charges,
evaluated at the sample means of the frontage-rate
buildings. We estimated ordinary least squares (OLS) re-
gressions as well, and calculated the predicted mean.
Below, we discuss the possibility that these two groups of
buildings differ in unobservable ways from the frontage-
rate buildings and present details of the regressions.
To obtain long-term estimates of water and sewer
charges that take into account likely conservation efforts
made by owners in response to metering, we used a sec-
ond data set consisting of 1,101 buildings in southeast
Queens that for many years had been served by the Ja-
maica Water Supply Company and had always paid their
water charges on the basis of actual metered use. When
the City took over the Jamaica Water Supply Company
in 1996, these “Jamaica buildings” were not permitted
to switch to frontage rates, but continued to be charged
on the basis of their meter readings. As with the M-O-M
buildings, we ascertained the characteristics of these
buildings and their neighborhoods and used regression
analysis to estimate what would happen if all buildings
with similar characteristics used the same amount of
water as those buildings. These long-term estimates as-
sume that building owners had the opportunity to make
conservation-related improvements, as the Jamaica
building owners are assumed to have done over the years.
We examined not only the average effect, but also
the effect on a number of percentiles, as Table 2 shows.
Examining the entire distribution of water and sewer
charges allows for a much richer picture of the effects of
TABLE 2. Distribution of annual water and sewer charges per unit and changes at selected percentiles.
Water and sewer charges ($)
Percentile
Type of charge
10th
25th
Median
75th
90th
M-O-M buildings
Predicteda, b
Currentc
129
214
−85−
217
291
−74−
360
316
44
598
343
255
1,013
401
612
Changee
Jamaica buildings
Predicteda, d
Currentc
133
214
−81−
212
291
−79−
322
316
6
488
343
145
780
401
379
Changee
a. Predictions are not regression-adjusted
b. 4,864 M-O-M buildings
c. 45,457 frontage rate buildings
d. 1,101 Jamaica buildings
e. Predicted charges minus current charges
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 423

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
family buildings analyzed in this study, but after adjust-
ing for neighborhood poverty rates, buildings for which
protests were filed appeared to be similar to those that
had no protests filed. For example, buildings located in
high-poverty census tracts whose owners protested their
assessments had market values that were 9% higher than
those for which no protests were filed. Although this dif-
ference is statistically significant, it is rather small. Fur-
thermore, a comparison of income and expense filings
for the TCIE sample was found to be similar to aggre-
gate data for the stock of rent-stabilized buildings as re-
ported by the New York City Rent Guidelines Board
(1998).
Description of Data
We used data on three groups of multifamily build-
ings obtained from the DEP: water and sewer charges for
metered Jamaica buildings, water consumption data for
M-O-M buildings, and water and sewer charges for other
New York City multifamily buildings being charged
frontage rates. In some instances, problems in coding
billing adjustments (erroneous charges that were can-
celed and the accounts rebilled) led to extreme outliers—
bills with unrealistically high or low charges. These ex-
treme outlier bills, at the 1st and 99th percentiles, were
truncated from the database before it was further ana-
lyzed. A number of analyses, described in the Appendix,
indicated that this step was sufficient to produce accu-
rate measures of water and sewer charges.
The Regression Models
The analysis was restricted to buildings with four or
more units, and excluded public housing projects, co-
operatives, and condominiums. Not including the Ja-
maica buildings, there are 85,280 such buildings in New
York City. The sample we used in most of our analyses
includes 81% of these buildings, or 69,493 properties (see
Table 1). Most of the buildings that were eliminated
from the sample were deleted because they were not
identifiable in the DEP data, or they were identifiable but
did not have water charges, or the periods that the water
charges referred to were unascertainable.
The three sets of water billing and use data were then
merged with data from two additional sources. Infor-
mation on the number of units in a building, its size (in
square feet), its value as estimated by tax assessors, and
its age were obtained from the 1998 Real Property As-
sessment Data file (RPAD) maintained by the New York
City Department of Finance (1998a). In addition, infor-
mation from the 1990 decennial census on neighbor-
hood characteristics, such as median income and the
rental vacancy rate (U.S. Census Bureau, 1990), were
merged into the database. These neighborhood charac-
teristics are intended to serve as proxies for building
characteristics (such as the vacancy rate) and household
characteristics (such as income) for which direct data
were not available. Neighborhoods are defined as census
tracts.
Regression Analysis Based on M-O-M
Buildings
The water use patterns of M-O-M buildings can be
used to estimate water use and water and sewer bills for
all multifamily buildings in New York City. These 4,864
buildings have meters that are monitored by the City,
but are still charged frontage rates. Because water use is
currently known for these buildings, the City’s water and
sewer rates can be applied to learn what would have been
their water and sewer charges based on meter readings. A
variety of building and neighborhood characteristics are
known about each building. Through the use of OLS
and quantile regression techniques,5 we were able to es-
timate how much each of these characteristics contrib-
uted to water use. These regression coefficients were then
used to estimate water use and bills for all buildings in
the city. These estimates should be interpreted as short-
term effects of metering since they do not account for
the conservation response likely to occur once metering
is put into place.
The variable means and results of the M-O-M build-
ings and later regressions are shown in Appendix Tables
A-1, A-2, and A-3. The dependent variable is the loga-
rithm6 of annual water charges per housing unit. Inde-
pendent variables consist of three building character-
istics (the logarithm of the market value per unit as
estimated in RPAD, the logarithm of square feet per
unit, and the building age) and four neighborhood char-
acteristics (the census tract vacancy rate, the median
household income in logarithms, the proportion of
units with more than one person per room, and the av-
erage number of persons per household).
Data on the financial characteristics of buildings in
New York City were obtained from the Tax Commission
Income and Expense (TCIE) database (New York City
Department of Finance, 1998b). All owners of rental
properties with more than 11 units are required by law to
file income and expense statements with the Depart-
ment of Finance each year. These data are confidential.
However, when a building’s owner protests its tax as-
sessment, its income and expense data become publicly
available. Protests were filed for only 10% of the multi-
Because we were primarily concerned with predic-
tion, rather than the coefficients as such, we did not hesi-
tate to include variables that are closely related. For ex-
ample, we included both persons per household and the
N
N
424 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
fraction of “overcrowded” households with more than
one person per room (both measured at the census tract
level). Similarly, we included the log value per unit, log
square feet per unit, and log census tract median income,
even though all three variables might reasonably be in-
terpreted as proxies for the income of individual house-
holds. Below, we discuss alternative models with a more
parsimonious specification. The only coefficient that ad-
mits of a simple interpretation in these regressions is
building age, which is .0055 in the OLS regression, indi-
cating that an additional year increases water use by 0.5%
(SE=. 0013). In spite of our emphasis on prediction, the
individual coefficients were estimated with surprising
precision.
After estimating the regressions, we calculated the
average values of the explanatory variables (value per
unit, building age, vacancy rate, etc.), and computed pre-
dictions for the city as a whole. We also report predic-
tions for high-poverty neighborhoods and other sub-
groups, which are calculated using the average values of
the explanatory variables within each subgroup.
The main problem with this procedure is that the
buildings are self-selected: The owners must ask that the
installed meters not be used as the basis for billing and
must have cooperated in the installation of meters (while
installation of meters is compulsory, there are opportu-
nities for delay on the part of owners). Building owners
who believe that water use is very high may have avoided
metering altogether, so that estimates based on this sam-
ple may understate city-wide multifamily building water
use. On the other hand, once the meters were installed,
the owners may have learned that water use was very
high and requested that frontage-rate billing continue;
so these estimates may overstate city-wide multifamily
building water use.
for many years, the self-selection bias that exists for
M-O-M buildings is eliminated. Nevertheless, the Ja-
maica buildings sample does have one drawback: Rela-
tively few of the properties are located in high-poverty
neighborhoods (i.e., census tracts with poverty rates over
40%). This may have biased our results toward overesti-
mating the impact of conservation on water use.
Other Analysis
Before moving to the next section, which describes
the estimates we derived for the impact of water meter-
ing on water and sewer charges, there are two additional
methodological points that must be discussed. First, in
order to learn more about the implications of changes
in water and sewer bills, we examined their likely impact
on building finances. This was done by comparing our
estimates of changes in water and sewer bills to the net
operating income (NOI) of multifamily buildings in
New York City. These data, from the City’s TCIE data
file, are subject to self-selection problems based on the
owner’s view of the strength of his or her case for a real
estate tax reduction. However, we believe that the finan-
cial data in the TCIE data file are roughly representative
of the universe of multifamily buildings in New York
City. In addition, the data should not be biased with re-
spect to water use, although exceptionally high water
and sewer charges could conceivably be one of a number
of factors that lead to low NOI.
Second, our focus was on affordable housing, par-
ticularly housing for low- and moderate-income fami-
lies. However, the data we used for this analysis contains
neither rents nor income, nor any other household char-
acteristics of the occupants of particular buildings,
which are essential for precise measurement of afford-
able housing. The proxy we used for affordable housing
is the poverty rate of the census tract in which the build-
ing is located. Although housing that is affordable to
low- and moderate-income households may exist in low-
poverty tracts because of rent regulation, it is our as-
sumption that low-cost housing is more likely to be pre-
sent in neighborhoods with higher poverty rates.
Regression Analysis Based on Jamaica
Buildings
To obtain long-term estimates of water bills that in-
corporate likely conservation measures taken by build-
ing owners, we relied on data from the Jamaica buildings
sample. The Jamaica buildings regressions include all the
same variables as the M-O-M buildings, in addition to
controls for billing adjustments and missing dates in the
billing adjustment data. These variables are rarely statis-
tically significant and follow no clear pattern, an addi-
tional indication that errors in the billing adjustment
data are causing little, if any, bias. The main difference in
the Jamaica buildings results is that the OLS coefficient
on building age falls by a factor of three to .0016, which
is not statistically significant (SE=. 0014).
Estimates of the Impact of Metering
on Water Bills
Tables 3 and 4 show the distribution of water and
sewer charges actually billed at the frontage rate in 1998
to owners of multifamily buildings and the predicted
charges if all of those buildings had been charged on the
basis of the meter readings. In addition to presenting the
results by the poverty rate in the census tract in which
the building is located, the tables present distributions
of the charges at each poverty rate. That is, in addition to
One of the major advantages of the Jamaica build-
ings regressions is that because they have been metered
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 425

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
TABLE 3. Annual water and sewer charges per unit and short-term predicted changes, based on regression-adjusted charges
to M-O-M buildings.
Water and sewer charges ($)
Percentile
Poverty rate
in census tract (%)
Number
Meana
10th
25th
Median
75th
90th
Current chargesb
0–5
5–10
10–15
15–20
20–25
25–30
30–35
35–40
> 40
2,439
7,662
7,785
7,379
5,528
3,623
2,668
3,222
5,151
45,457
324
314
310
308
295
289
282
275
270
298
247
243
248
242
213
196
181
151
152
214
301
296
295
292
290
286
288
286
280
291
327
322
316
313
310
309
310
309
310
316
380
354
346
343
337
331
334
336
332
343
449
409
399
407
399
382
377
399
388
401
Total
Short-term predicted changes (predicted charges minus current chargesc)
0–5
2,439
7,662
7,785
7,379
5,528
3,623
2,668
3,222
5,151
45,457
−79−
−59−
−21−
18
−149−
−137−
−128−
−109−
−68−
−41−
4
−137−
−125−
−100−
−73−
−46−
−30−
9
9
57
−64−
−78−
−63−
−21−
23
64
83
135
146
201
30
4
47
216
264
359
450
525
576
659
675
812
463
5–10
10–15
15–20
20–25
25–30
30–35
35–40
> 40
112
178
239
273
342
364
457
191
62
87
147
152
212
36
20
45
−77−
Total
Note: Public housing, co-ops, and condominiums are excluded from the sample.
a. The means are geometric.
b. Charges for buildings with monitor-only meters are the observed water consumption multiplied by the combined water and sewer rate for
FY 1998 ($3.11/CCF).
c. Current charges reported are for buildings with only frontage-rate billing. Buildings with both metered and frontage-rate charges are
excluded.
the means7 and medians, predicted and actual charges
for the buildings at the 10th, 25th, 75th, and 90th per-
centiles show which buildings would be most and least
affected by the change in billing practice.
Table 5 shows the mean and median predicted
changes in annual water and sewer charges per housing
unit for all buildings, regardless of the poverty rate in
the census tract. Clearly, a shift to metering would not
present a severe financial problem for the average build-
ing that is now paying the frontage rate. There are, how-
ever, thousands of buildings that are well above the
average.
most research on water demand finds that although the
demand for water outside the home (e.g., for lawns) is
fairly sensitive to price, for water use inside the home it is
not.8 Several studies found that the transition from flat
rates to metering in Denver had almost no effect on in-
home water use (Brown & Caldwell, 1984). Hanke (1970),
however, found a 36% reduction in in-home use after
Boulder, Colorado, switched to metering. A drawback of
this literature for our purposes is that it focuses on sin-
gle-family dwellings. Multifamily buildings in New York
have only a single meter upon which bills are based,
rather than individual apartment meters. Since house-
holds have little direct incentive to conserve water (al-
though building owners do), multifamily buildings can
be expected to be less responsive to price, although the
magnitude of this effect has not been studied.9
Short-term Versus Long-term Impacts
There are two reasons to expect that water use will
not be very responsive to price in New York City. First,
N
N
426 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
TABLE 4. Annual water and sewer charges per unit and long-term predicted changes, based on regression-adjusted charges
to Jamaica buildings.
Water and sewer charges ($)
Percentile
Poverty rate
in census tract (%)
Number
Meana
10th
25th
Median
75th
90th
Current chargesb
0–5
5–10
10–15
15–20
20–25
25–30
30–35
35–40
> 40
2,439
7,662
7,785
7,379
5,528
3,623
2,668
3,222
5,151
45,457
324
314
310
308
295
289
282
275
270
298
247
243
248
242
213
196
181
151
152
214
301
296
295
292
290
286
288
286
280
291
327
322
316
313
310
309
310
309
310
316
380
354
346
343
337
331
334
336
332
343
449
409
399
407
399
382
377
399
388
401
Total
Long-term predicted changes (predicted charges minus current charges)
0–5
2,439
7,662
7,785
7,379
5,528
3,623
2,668
3,222
5,151
45,457
−57−
−44−
−23−
−7−
6
−141−
−135−
−128−
−109−
−81−
−62−
−51−
−15−
1
−132−
−121−
−101−
−78−
−67−
−57−
−56−
−40−
−2−
−61−
−53−
−27−
−3−
8
14
10
31
63
27
50
75
96
98
108
86
108
144
88
198
233
246
235
210
226
171
179
221
221
5–10
10–15
15–20
20–25
25–30
30–35
35–40
> 40
15
12
36
70
Total
−3−
−88−
−79−
−8−
Note: Public housing, co-ops, and condominiums are excluded from the sample.
a. The means are geometric.
b. Current charges are for buildings with only frontage-rate charges. Buildings with both metered and frontage-rate charges are excluded.
We can examine the amount by which water use can
be expected to decrease over time in response to large in-
creases in water bills by comparing our two regression
analyses. The short-term analysis based on the actual
water use by the M-O-M buildings represents the situa-
tion before meter-based billing and the long-term anal-
ysis based on water use by the Jamaica buildings predicts
the situation afterwards.
As shown in Tables 3 and 4, the average customer is
predicted to have reduced water use by about 11%, over
time. That is, bills are predicted to decline from $334 to
$296.10 The customer at the 90th percentile in Table 4
consumes 28% less water than the customer at that per-
centile in Table 3; for the 75th percentile, the difference
is 19%. Of course, it is important to note that the cus-
tomer in the 90th percentile in Table 3 may not be the
same customer as the one in the 90th percentile of Table
4. However, the change faced by particular customers (as
opposed to the overall change) is not especially relevant
for our purposes. Since frontage-rate charges tend to
vary over a fairly narrow range, the changes faced by a
particular customer will, in most cases, be similar to the
overall changes.
In the short term, water and sewer charges are pre-
dicted to decline for roughly 13,000 of the 45,457 front-
age-rate buildings when billing based on meter readings
is instituted, because their water use was already low.11 In
the long term, water and sewer charges are predicted to
decline for nearly 18,000 of these buildings, as building
owners find ways to conserve water. Across the entire
spectrum of data, charges in the long term are predicted
to decline from our short-term estimates.
Nonetheless, many buildings would experience
quite large increases. Approximately 9,000 buildings
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 427

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
TABLE 5. Mean and median changes in water and sewer
charges for all buildings.
ance, as is to be expected, because the frontage-rate
schedule varies mainly with the number of units in a
building. For all buildings (Table 2), the range of current
charges was $214 at the 10th percentile to $401at the
90th percentile, a ratio of less than 1:2—with a median of
$316. In the short-term model (Table 3), the predicted
range is from $138 to $864, a 1:6 ratio—with a median of
$346. In the long-term model (Table 4), the predicted
range is from $127 to $621, a ratio of 1:5—with a median
of $307.
Changes in water
and sewer charges
Short-term
(Table 3)
Long-term
(Table 4)
Mean change ($)
Mean change (%)
Median change ($)
Median change (%)
36.0
12.1
30.0
9.5
−3−.0
−1.0−
−8−.0
−2.5−
Similarly, the range from the 25th to the 75th per-
centile is considerably enlarged. In 1998, for all build-
ings (Table 2) the range of current charges was from
$291 to $343, a 1:1.2 ratio. In the short-term model, the
predicted range is from $227 to $534, a 1:2.35 ratio. In
the long-term model, the predicted range is from $212 to
$431, a 1:2 ratio.
would have short-term annual increases in excess of $200
per unit, and more than 3,700 buildings would have sim-
ilar increases even in the long term. Because of the pos-
sible bias in the Jamaica buildings sample, it is conceiv-
able that even these fairly modest reductions in water use
over time are overly optimistic about conservation, as
noted earlier. The direction of the bias with respect to
the M-O-M buildings sample is less clear.
Patterns by Census Tract Poverty Rates
The results of our analysis consistently demonstrate
that a transition to metering would have a dispropor-
tionately negative impact on multifamily buildings lo-
cated in areas with higher poverty rates. Table 3 predicts
short-term water and sewer charges that exhibit a dis-
tinct progression as the census tract poverty rate in-
creases. In 1998, the median charges were almost uni-
form among poverty-rate classes: The mean variation
from the median for all buildings was less than $6 per
unit. Median charges are predicted to increase steadily
as the poverty rate grows, from $249 in tracts with pov-
erty rates under 5% to $511 in tracts with poverty rates of
40% or more. At the 90th percentile, increases of between
$525 and $812 per unit are predicted for buildings in
tracts with poverty rates of 20% or more.
In 1998, the range from the 25th to the 75 th percen-
tile was quite close to $50 for all the poverty-rate classes.
The short-term predictions in Table 3 show a range of
about $150 for the lowest poverty rate-classes (15% or
less), rising sharply to as much as $400 in the higher-pov-
erty-rate classes. Thus, there would be large building-to-
building differentials within neighborhoods.
Severely Impacted Buildings
Although the predicted changes in water and sewer
charges for “average” buildings are not large, this is not
true for a significant number of buildings located in cen-
sus tracts with higher poverty rates. In the short-term
(see Table 3), in census tracts with poverty rates of 10% or
more, predicted annual increases would be more than
$300 per unit for the top decile—roughly 3,500 build-
ings. In tracts with poverty rates of 30% or more similar
increases would occur for the buildings between the 75th
and 90th percentiles—another 1,650 buildings. Given
that in 1998 most of these buildings were billed between
$300 and $400 per unit, a $300 increase is a large one.
Worse yet, about 2,000 buildings, all in tracts with pov-
erty rates of 20% or more, are predicted to face annual
increases of $500 or more per unit if the existing caps
were lifted.
In the long term (see Table 4), however, very few
buildings are predicted to have increases of as much as
$300 per unit. About 3,700 buildings are predicted to ex-
perience increases of more than $200 but less than $300
per unit; these are located across the poverty-rate spec-
trum. Of course, even a $200-per-unit increase can be a
serious problem for buildings whose financial state is
precarious. Furthermore, as noted above, the long-term
predictions may be too optimistic about the opportuni-
ties for water conservation.
According to Table 4, the long-term differences in
charges among the poverty-rate classes would be much
smaller than the short-term differences. The range from
the 25th to the 75th percentile is predicted to be greater
than at present (but much less than in the short term),
although it too does not vary much among poverty-rate
classes.
Impact on Building Finances
The TCIE data shown in Table 6 permit us to place
the predicted changes in water and sewer charges in the
context of building finances. Tables 3 and 4 predict that,
Dispersion of Water and Sewer Charges
In 1998, water and sewer charges per unit in build-
ings billed at frontage rates exhibited a fairly small vari-
N
N
428 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
TABLE 6. Annual net operating income per unit.
Net operating income ($)
Percentile
Poverty rate
in census tract (%)
Number
Mean
10th
25th
Median
75th
90th
0–5
5–0
859
2,249
1,640
1,137
904
544
442
482
600
4,799
4,003
3,152
2,318
1,877
1,909
1,959
1,573
1,365
2,948
2,626
177
153
1,838
1,638
1,215
978
667
606
561
461
289
952
4,240
3,462
2,585
2,116
1,697
1,698
1,587
1,299
1,004
2,274
2,035
7,133
5,643
4,426
3,143
2,653
2,776
2,536
2,226
2,036
4,179
3,728
10,460
8,597
7,134
4,750
4,161
4,225
4,392
3,889
4,028
7,004
6,228
10–15
15–20
20–25
25–30
30–35
35–40
> 40
−27−
−33−
−242−
−232−
−42−
−48−
−669−
−75−
−137−
Total
Total (weighted)
8,857
87,322
777
Source: Tax Commission Income and Expense Data from the Major Property File maintained by the New York City Department of Finance
(1998b).
Notes: Data is truncated at the 1st and 99th percentiles. Observations with zero or missing rental income data have been omitted. Weighted
total reflects the overall distribution of poverty in New York City.
in both the short term and long term, most buildings
(currently paying frontage rates) located in the low-pov-
erty-rate tracts—below 15%—would experience small re-
ductions, not increases in water and sewer charges as a
result of metering. But in the tracts with higher poverty
rates—15% and above—those charges would increase.
For each poverty-rate class of 15% or more, Table 7
compares the predicted increases in median water and
sewer charges with median net operating income per
unit. Clearly, the short-term increases in charges in cen-
sus tracts with higher poverty rates would likely pose a
significant financial problem for some affordable hous-
ing. Once again, we see that the long-term problem is
much less severe, but subject to the caveat that the ex-
tent of conservation implied by the long-term predic-
tions may be overly optimistic.
Even under these optimistic long-term predictions,
there would be hardship cases. This can be seen by com-
bining the data for buildings that are below the median
in net operating income with the data for buildings that
are below the median with respect to increased water and
sewer charges. Consider the predicted changes in Table 4
for buildings at the 75th percentile in the high-poverty
classes and the net operating income data in Table 6 at
the 25th percentile for those classes shown in Table 8. It
is clear that increases in water and sewer charges of those
amounts for buildings with the indicated net operating
income would impair financial viability. This is not to
say that most buildings in the high-poverty tracts would
experience the same difficulties, but surely the number
of such cases would be significant.
Of course, to some extent increased water and sewer
charges would be passed along to tenants in the form of
higher rents, although what little evidence currently ex-
ists suggests that this process would be very slow (Black-
ley & Follain, 1995). Currently, median rents are $545 per
month in the highest-poverty neighborhoods. If all of the
75th percentile increase in charges ($457 per year in Table
3 and $144 in Table 4) were passed on to the tenants,
rents would rise by 7% in the short run and 2% in the long
run. Although these are not enormous rent increases, it is
TABLE 7. Changes in water and sewer charges as percent
of net operating income.
Poverty rate in
Short-term
Long-term
census tract (%)
changes (%)
changes (%)
15–20
20–25
25–30
30–35
35–40
> 40
1.1
3.8
4.9
8.5
11.2
20.0
−0.1−
0.5
0.8
0.6
2.4
6.3
Note: Calculated from data in Tables 3, 4, and 6.
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 429

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
TABLE 8. Long-term changes in water and sewer charges as percent of net operating income in extreme cases.
Poverty rate
in census tract (%)
Change in charges
at 75th percentile ($)a
Net operating income
at 25th percentile ($)b
Change in charges as %
of net operating income
20-25
25-30
30-35
35-40
> 40
98
108
86
108
144
667
606
561
461
289
14.7
17.8
15.3
23.4
49.8
a. See Table 4.
b. See Table 6.
not clear that the tenants have the ability to pay them.
Currently 55% of the tenants in the highest-poverty
neighborhoods pay more than 30% of their income in
rent and 32% pay more than 50%. Because rents are al-
ready high relative to income in high-poverty neighbor-
hoods, landlords may have difficulty passing along the
rent increases.12 Another potential hazard for the stock of
affordable housing concerns the numerous City-assisted
affordable housing projects operated by private sponsors.
These projects were underwritten on the basis of operat-
ing costs at the time the contract was signed, trended for
inflation in general, but without provision for large in-
creases caused by unique events such as the change to
usage-based water and sewer charges. Although chang-
ing the regulatory and financial agreements to include
additional subsidies is possible, the complex financing
involved makes that anything but automatic. Thus, some
of these projects might face insolvency.
We also replicated the results in Tables 3 and 4,
deleting value per unit and the fraction of overcrowded
households from the regression models. These two var-
iables have the highest correlation (around .5) with two
other variables, square feet per unit and persons per
household. After this change, the predictions for me-
tered charges in M-O-M buildings were relatively un-
changed. The predictions for the Jamaica buildings show
the same general pattern as the results in Table 4, and
change very little in the lower-poverty neighborhoods.
In higher-poverty neighborhoods, the more parsimo-
nious specification predicts even larger increases. For the
highest-poverty neighborhoods, the less extensive spec-
ification predicts approximately double the mean and
median increase, compared to Table 4.14
Conclusion and Policy Implications
The results of our analysis consistently demonstrate
that if the City were to end its current transition pro-
grams, the negative impacts of metering would be borne
disproportionately by the owners of multifamily build-
ings and/or their tenants in the poorest neighborhoods.
Water and sewer bills would likely increase substantially
in neighborhoods with relatively high poverty rates and
actually decrease in those with low poverty rates. Since
affordable housing tends to be clustered in neighbor-
hoods with high poverty rates, these increases would be
borne by the landlords and tenants of affordable hous-
ing. In many of these communities, the incomes of resi-
dents are so low that owners would not be able to pass
the increased costs through to tenants in rent. Many of
these owners could be threatened with negative cash flow
and, ultimately, abandon their buildings.
Robustness of Estimates
In order to check the robustness of our results, we
performed several additional analyses. Tables 3 and 4
compare buildings that are currently paying frontage
charges to a comparison group and use regression anal-
ysis to adjust for any differences in characteristics. For
the M-O-M buildings, however, we can observe both
frontage charges and usage-based charges for each build-
ing. A simple analysis which takes advantage of this sit-
uation compares means and quantiles of frontage
charges and metered “shadow charges” for the M-O-M
buildings themselves.13 This makes regression analysis
unnecessary, although the results cannot be extrapo-
lated to the city as a whole. The median change for all
the M-O-M buildings using this method is $44, com-
pared to $30 in Table 3, and the 75th percentile change
is $252, compared to $191 in Table 3. The results for the
different neighborhoods are also quite similar to the re-
sults in Table 3.
It is important to recall that the rising burden of wa-
ter and sewer rates on low- and moderate-income hous-
ing is only in part linked to metering. Water and sewer
rates have risen extraordinarily rapidly over the past 20
years. While rate hikes in excess of inflation are common
N
N
430 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
in many parts of the nation (Beecher & Mann, 1997), the
rapid increase in New York City combined with the tran-
sition to metering has intensified problems for housing
providers.
groups. In a system with fixed and variable rates, the
same effect could be achieved by waiving some or all of
the fixed charge for the targeted groups. Although this
approach might be the most appealing on efficiency
grounds because it does not distort the relationship be-
tween water use and water charges, it might be too cum-
bersome for many municipalities to manage. This is es-
pecially true because many agencies charged with water
provision do not have experience with income transfer
programs and do not view such a function as part of
their mission.
Another alternative would be for a city to institute
caps on charges set at some level that would be afford-
able to the property owners. Rate caps are problematic
on two grounds: (1) they would benefit all property own-
ers, even those whom the government would not wish to
subsidize, and (2) although an incentive to conserve
would exist up to the cap, once the usage exceeded the
cap this incentive would disappear. These limitations
could be ameliorated (although not eliminated) by tar-
geting the cap to areas with large proportions of low-
and moderate-income households and requiring con-
servation efforts to be undertaken by those owners who
benefit.
In the end, the New York City Water Board decided
to respond to the distributional concerns of universal
metering by stepping back. It decided that all owners of
multifamily residential buildings with six or more dwell-
ing units who replace at least 70% of their water fixtures
with “low water consuming fixtures” will be entitled to
elect either metered charges or frontage charges. Under
this program, the Department of Environmental Pro-
tection maintains the right to audit water use from time
to time to ensure that conservation goals are being
achieved. With respect to 62% of all housing units in
New York City, the relief provided is neither targeted to
those who need it the most, nor is it in a form that pre-
serves incentives to conserve.
Thus, the transition to usage-based charges for
water and sewer service presents the standard efficiency-
versus-equity tradeoff typical of so many debates about
user charges. On the one hand, usage-based charges can
force consumers to take into account the cost of the
scarce resources they consume and provide appropriate
incentives for conservation. On the other hand, for cer-
tain services or commodities such as water, consumers
(landlords and tenants) may not be able to reduce con-
sumption below some level because they require the ser-
vice or commodity for survival, and/or they cannot fi-
nance conservation-related capital expenses. In addition,
metering applies not to the ultimate consumer, but to
the building owner who is not able to affect directly the
usage patterns of his or her tenants. Probably for these
reasons, we find that in New York City, as in other cities
that have been studied, user charges are likely to have
only a modest effect on conservation. The public sector
is faced with a challenge, then, if it wishes to ease the
hardships created by a metering system and, at the same
time, preserve incentives to conserve.
The water and sewer charge problem for multifamily
buildings in New York City is aggravated by the fact that
water use is the sole determinant of the charge, once me-
tering is instituted. Conventional utility rate structures
have a fixed and a variable component, with the fixed
part of the tariff (usually an equal amount per customer)
based on costs such as billing, meter reading, and other
costs that do not vary directly with actual use. Using a
fixed and variable rate structure would lower charges for
buildings with high water use and reduce some of the
projected increases in water bills for owners of buildings
in poorer neighborhoods.
Furthermore, many municipalities use “lifeline”
rates in which charges are set below cost for a set amount
of water, thereby ensuring that consumers can receive a
given level of service at a subsidized rate (Saunders et al.,
1998). These lifeline rates can be targeted to low- and
moderate-income households, landlords with low- and
moderate-income tenants, or geographic areas in which
the median income is low. The advantage of this type of
subsidy structure is that after the subsidized amount of
water is consumed, the price jumps up to a level approx-
imating marginal cost, thereby preserving incentives to
conserve.
NOTES
1. Until 1980, property-tax-exempt organizations were sup-
plied water without charge.
2. Submetering is sometimes used for residential water (Ro-
mano, 1998). However, the multistory nature of most
rental buildings, with several vertical risers bringing water
into each apartment, makes such a billing mechanism im-
practical.
3. Net operating income is calculated by subtracting oper-
ating expenses plus property taxes from gross revenue. It
does not include a deduction for debt service despite the
fact that many of these buildings have loans secured by
mortgages.
Alternatively, the rate structure could be maintained
at average or marginal cost, while the distributional ob-
jective is achieved through a cash subsidy from the gov-
ernment to low-income households or other targeted
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 431

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
4. Quantile regression was developed by Koenker and Basset
(1978). For an example of an application of quantile re-
gression methods, see Friedlander and Robins (1997).
5. The predicted values from OLS regressions are predicted
means. Quantile regression allows us to predict not only
what the mean water bill will be in the future, but also the
10th percentile water bill, the median water bill, etc. Use of
this estimator allows predictions of the future distribu-
tion of water bills without requiring any a priori assump-
tions about the shape of the distribution.
6. Box-Cox tests suggest that measuring the dependent vari-
able in logarithms improves the fit of the equation. In ad-
dition, the use of logarithms greatly reduces the hetero-
skedasticity of the data.
7. Because the dependent variable is measured in loga-
rithms, our predicted means are geometric means, rather
than the more common arithmetic means. The geometric
mean is the anti-logarithm of the arithmetic mean of the
logarithm of water charges. The advantage of geometric
means is that because they put less weight on extremely
high water charges, they are a better measure of central
tendency, in this case the water bill received by the “typi-
cal” customer.
1000 vs. about 4000 for the M-O-M buildings) and be-
cause there are few high-poverty tracts where the Jamaica
buildings are located.
REFERENCES
Arthur Andersen Inc. (1995). Breaking the cycle: Developing an ef-
fective strategy for dealing with New York City’s in rem housing
problem. New York: Author.
Baumann, D. B., Boland, J. J., & Hanemann, W. M. (1998).
Urban water demand management and planning. New York:
McGraw-Hill.
Beecher, J. A., & Mann, P. C. (1997, July 15). Real water rates
on the rise. Public Utilities Fortnightly, pp. 42–46.
Blackburn, A. J. (1994). The impact of metered billing for water and
sewer on multifamily housing in New York (Report). Boston:
Speedwell, Inc.
Blackley, D. M., & Follain, J. R. (1995). In search of empirical evi-
dence that links rent and user cost (Working Paper 5177).
Cambridge, MA: National Bureau of Economic Research.
Braconi, F. P. (1999.) In re in rem: Innovation and expediency
in New York’s housing policy. In M. H. Schill (Ed.), Hous-
ing and community development in New York City: Facing the
future (pp. 93–118). Albany: State University of New York
Press.
8. Howe (1982), for example, finds a very small elasticity of
demand for water: −.06 in the winter when little water is
used outside the home, and about −.5 in the summer.
These results are typical of the studies reviewed in chapter
2 of Baumann et al. (1998).
Brown & Caldwell. (1984). The effect of water meters on water use.
Washington, DC: U.S. Department of Housing and Urban
Development.
9. Hansen (1996) tests and fails to reject the assumption that
building owners have no incentive to conserve and that
residents “free-ride.” However, he does not focus on this
issue. Wong (1972), in one of the few studies of a city with
a mainly multifamily housing stock, finds a demand elas-
ticity of −.02 for Chicago, which indicates that price has
almost no effect on demand.
Community Preservation Corporation. (1997, December 8).
Water/sewer pricing in New York City (Draft report). New
York: Author.
Friedlander, D., & Robins, P. K. (1997).The distributional im-
pacts of social programs. Evaluation Review, 21(5), 531–
553.
Hanke, S. H. (1970). Demand for water under dynamic condi-
tions. Water Resources Research, 6(5), 1253–1261.
Hansen, L. G. (1996). Water and energy price impacts on resi-
dential water demand in Copenhagen. Land Economics,
72(1), 66–79.
Howe, C. W. (1982). The impact of price on residential water
demand: Some new insights. Water Resources Research,
18(4), 713–716.
10. The predictions are not shown in Tables 3 and 4, but can
be inferred from the tables.
11. Statements about the number of buildings that face in-
creases or decreases are helpful in summarizing the results
of our analysis. Strictly speaking, however, data limita-
tions only allow us to approximate these numbers.
12. The figures on rent in high-poverty areas are the authors’
tabulations from the 1999 New York City Housing and
Vacancy Survey (Lee, 1999) for “Sub-borough Areas” 8
and 9 in Harlem, 1 through 4 in the South Bronx, and 1
and 4 in Brooklyn. These areas contain 77% of the New
York City housing units in census tracts with poverty rates
of 40% or more. Because sub-borough areas are relatively
large (containing 130,000 people on average), not all of
the residents are in high-poverty tracts; 56% are in tracts
with poverty rates of 40% or more while a quarter are in
tracts with 10% or less. Rent is defined as “gross rent,”
which is rent plus utility payments.
Koenker, R., & Bassett, G., Jr. (1978). Regression quantiles.
Econometrica, 46(1), 33–50.
Lee, M. W. (1999). Housing New York City, 1996. New York: New
York City Department of Housing Preservation and De-
velopment.
New York City Department of Finance. (1998a). 1998 Real Prop-
erty Assessment Data [Electronic data tape]. New York:
Author.
New York City Department of Finance. (1998b). 1998 Tax Com-
mission Income and Expense Data [Electronic data tape]. New
York: Author.
13. A supplemental table containing these comparisons is
available from the authors.
14. The Jamaica buildings results are probably more sensitive
to the choice of explanatory variables than the M-O-M
buildings results because the sample size is smaller (about
New York City Rent Guidelines Board. (1998, April 7). Income
and expense study (Report). New York: Author.
Pollakowski, H. O. (1997). The effects of rent deregulation in New
York City (Working Paper 67). Cambridge, MA: MIT Cen-
ter for Real Estate.
N
N
432 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
Romano, E. (1998). Making water submetering easier to swal-
low. Journal of Property Management, 63(5), 38–44.
Saunders, M., Kimmel, P., Spade, M., & Brockway, N. (1998).
Water affordability programs. Denver: American Water
Works Association.
Scafidi, B. P., Schill, M. H., Wachter, S. M., & Culhane, D. P.
(1998). An economic analysis of housing abandonment.
Journal of Housing Economics, 7(4), 287–303.
Schill, M. H., & Scafidi, B. P. (1999). Housing conditions and
problems in New York City. In M. H. Schill (Ed.), Housing
and community development in New York City: Facing the fu-
ture (pp.11–52). Albany: State University of New York
Press.
U.S. Bureau of Labor Statistics. (1999). Consumer price index—all
urban consumers. New York all items (ID: CUURA101SA0)
[On-line]. Available: <http://146.142.4.24/cgi-bin/
surveymost>.
U.S. Census Bureau. (1974). 1972 Census of governments: Vol. 4,
No. 5, Compendium of government finances. Washington, DC:
Author.
U.S. Census Bureau. (1990). 1990 Census of population and hous-
ing, census tract tabulations [Electronic file]. New York: Com-
munity Studies of New York, Inc. [Distributor].
U.S. Census Bureau. (1997). 1992 Census of governments: Vol. 4,
No. 5, Compendium of government finances. Washington, DC:
Author.
date, based upon the origin date of recent cancellations.
However, for 72% of the buildings with rebillings, the
procedure was unable to assign an origin date. Billing
adjustments are of greatest concern for the Jamaica
buildings sample and a lesser concern for the frontage-
rate sample, but are not an issue for the M-O-M build-
ings sample since no bills were issued to these buildings.
We took several steps to ensure that problems
caused by billing adjustments were not affecting our re-
sults. First, we truncated water charges at the 1st and
99th percentiles, excluding extreme outliers that are
mainly due to problems with billing adjustments. After
truncating the samples in this way, a number of analyses
were performed to check whether billing adjustments
had affected the results of the study. For the Jamaica
buildings, where water bills are used to make long-term
predictions for the rest of New York City, little evidence
was found to indicate that water bills with billing ad-
justments or missing origin dates were unusually high
or low. We also replicated several of our key analyses, ex-
cluding buildings with missing origin dates, and found
that this change had very little effect on the results.
Billing adjustments matter less for the frontage-rate
buildings because adjustments are less frequent for these
buildings and because knowledge of the frontage-rate
schedule reassures us that most frontage-rate charges
fall within realistic ranges. However, some charges at the
10th percentile seem too low. These low charges are at-
tributable to several factors. First, a number of formerly
vacant buildings that have been renovated are likely
being billed less than the rate schedule would indicate
because DEP records have not yet been updated to re-
flect occupancy. Second, the difficulty we experienced in
assigning rebills to the correct origin date may have re-
sulted in mismeasurement of bills. Last, it is possible
that the unit counts contained in the RPAD database are
not always current. These low bills indicate that caution
is warranted in interpreting the results for the 10th per-
centile, but they are otherwise unlikely to have much im-
pact on our conclusions.
Wong, S. T. (1972). A model of municipal water demand: A
case study of northeastern Illinois. Land Economics, 48(1),
34–44.
APPENDIX
Some Data Problems
Extensive processing of the DEP data was necessary
to obtain accurate measures of water use and charges.
For purposes of our analysis, charges needed to be as-
signed to the year in which they were incurred. For ordi-
nary charges, this was relatively straightforward, since
the DEP data include the date on which each water bill
was issued. However, when billing adjustments (i.e., can-
cellations and rebills) were made, it was important to
know not only the date the adjustment was made, but
also the date of the bill that was being adjusted (the ori-
gin date). The DEP data usually indicate the origin date
of cancellations, but not of rebillings. To surmount this
problem, we created a procedure to estimate the origin
Table A-1 shows the means of regression variables
for the three samples (Jamaica, M-O-M, and frontage-
rate buildings). Tables A-2 and A-3 show the respective
regression results for the M-O-M and Jamaica buildings
samples.
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 433

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
TABLE A-1. Means of regression variables.
Building type
M-O-M
Variable
Jamaica
Frontage rate
Annual water and sewer charges per unita
Natural logarithm
Geometric mean ($)
5.76
319 00
434 00
5.84
344 00
503 00
5.70
298 00
319 00
Arithmetic mean ($)
RPAD variables
Value per unit (logs)
Square feet per unit (logs)
Building age (years)
10.40
6.72
60 00
26.40
10.20
6.70
74.20
19.70
10.20
6.68
73.40
19.30
Number of residential units
Census tract variables
Rental vacancy rate (%)
Average persons/household
Poverty rate (%)
Median household income (logs)
Gross rent (logs)
3.80
2.62
12.40
10.40
6.37
4.52
2.41
23.30
10.10
6.22
4.64
2.39
21.30
10.20
6.23
Households with >1 person/room (%)
21.60
15.30
15.50
Poverty rate in census tract (%)
0–10
10–20
20–30
30–40
> 40
51.00
36.50
10.20
0.09
23.60
26.60
17.80
14.00
18.10
22.20
33.40
20.10
13.00
11.30
13.60
Billing adjustments (% of accounts)
Cancellation
Rebill
Lost cancellation
Lost rebill
6.72
7.36
2.82
3.27
9.89
10.80
.003
8.12
7.49
.38
8.61
6.49
Sample size
1,101
4,864
45,457
a. For M-O-M buildings, water charges are implicit, based on metered use.
N
N
434 APA Journal Autumn 2001 Vol. 67, No. 4

CHANGING WATER AND SEWER FINANCE
TABLE A-2. OLS and quantile regression results for the M-O-M buildings sample.
Quantile regression coefficient at selected percentiles (t-statistic)
Percentile
OLS coefficient
at the mean
(t-statistica)
10th
25th
Median
75th
90th
RPAD variables
Log value per unit
−0.0332
(−0.797)
0.529
(8.95)
0.00546
(4.06)
−0.0885
(−1.59)
0.370
(4.74)
0.00942
(4.90)
−0.109
(−3.66)
0.612
(14.2)
0.00536
(5.34)
−0.0861
(−3.76)
0.566
(15.8)
0.00390
(4.77)
−0.0797
(−2.57)
0.572
(11.1)
0.00297
(2.54)
−0.0180
(−0.360)
0.642
(7.61)
−0.309E-3
(−0.161)
Log square feet per unit
Building age
Census tract variables
Rental vacancy rate
−0.981
(−1.21)
−0.105
(−2.04)
−0.0723
(−0.193)
0.377
−4.01
(−4.50)
0.0116
(0.159)
−1.17
−0.605
(−1.28)
−0.138
(−3.38)
−0.272
(−1.00)
0.346
0.324
(0.863)
−0.142
(−4.30)
0.313
0.412
(0.821)
−0.191
(−4.11)
0.491
0.499
(0.602)
−0.197
(−2.62)
0.526
Log median household income
Fraction w/ >1 person per room
Persons per household
(−2.44)
0.581
(1.46)
0.284
(1.62)
0.218
(1.00)
0.176
(6.51)
(7.36)
(8.00)
(8.30)
(4.65)
(2.21)
Constant
2.44
(3.72)
0.140
1.52
(1.79)
0.078
2.70
(5.67)
0.110
3.36
(8.51)
0.118
4.37
(7.78)
0.097
4.16
(4.67)
0.067
R2 / Pseudo R2
Sample size
4,864
4,864
4,864
4,864
4,864
4,864
Notes: M-O-M “charges” are observed consumption for buildings with monitor-only meters and frontage-rate billing, multiplied by the
combined water and sewer rate for FY 1998 ($3.11/CCF). Dependent variable is the log of annual water and sewer charges per unit truncated
at the 1st and 99th percentiles.
a. OLS t-statistics are adjusted for clustering within census tracts.
N
N
APA Journal Autumn 2001 Vol. 67, No. 4 435

DICK NETZER, MICHAEL SCHILL, AND SCOTT SUSIN
TABLE A-3. OLS and quantile regression results for the Jamaica buildings sample.
Quantile regression coefficient at selected percentiles (t-statistic)
OLS coefficient
at the mean
(t-statistica)
Percentile
Median
10th
25th
75th
90th
RPAD variables
Log value per unit
0.345
(5.64)
0.573
(9.03)
0.00159
(1.15)
0.213
(2.15)
0.807
(7.12)
−0.00335
(−1.37)
0.152
(2.10)
0.612
(6.69)
2.27E-04
(0.112)
0.265
(5.50)
0.610
(9.54)
0.00149
(1.11)
0.371
(8.81)
0.526
(8.88)
0.00156
(1.36)
0.615
(5.35)
0.335
(2.37)
0.00356
(1.16)
Log square feet per unit
Building age
Census tract variables
Rental vacancy rate
−1.41
(−2.07)
−0.343
(−3.77)
0.651
(3.12)
0.00876
(0.214)
−2.98
(−2.58)
−0.330
(−1.59)
1.21
(2.90)
−0.0608
(−0.856)
−1.03
(−1.06)
−0.353
(−2.34)
0.882
(2.97)
0.0137
(0.221)
−0.437
(−0.697)
−0.324
(−3.57)
0.712
(3.82)
0.0360
(0.930)
−0.812
(−1.51)
−0.296
(−3.81)
0.659
(4.39)
0.0186
(0.626)
−2.28
(−1.76)
−0.322
(−1.66)
0.505
(1.32)
0.0257
(0.391)
Log median household income
Fraction w/ >1 person per room
Persons per household
Billing adjustments
Cancellation
0.143
(0.451)
0.0266
(0.0941)
0.226
(0.838)
0.272
(0.914)
−0.492
(−2.47)
0.353
(1.57)
0.848
(3.90)
−0.156
(−0.633)
0.0745
(0.358)
0.0800
(0.350)
0.0385
(0.172)
0.170
0.0316
(0.176)
0.216
0.212
(1.25)
0.0231
(0.134)
0.0792
(0.475)
0.209
0.692
(1.34)
Rebill
−0.176
(−0.350)
0.0595
(0.124)
0.468
(1.17)
Lost cancellation
Lost rebill
0.0735
(0.409)
0.0669
(0.345)
(0.698)
(1.19)
(0.955)
Constant
1.66
(1.54)
0.240
0.980
(0.443)
0.108
3.16
(1.92)
0.096
1.95
(1.93)
0.121
1.56
(1.75)
0.162
0.900
(0.404)
0.203
R2 / Pseudo R2
Sample size
1,101
1,101
1,101
1,101
1,101
1,101
Note: Dependent variable is the log of annual water and sewer charges per unit truncated at the 1st and 99th percentiles.
a. OLS t-statistics are adjusted for clustering within census tracts.
N
N
436 APA Journal Autumn 2001 Vol. 67, No. 4