NBER WORKING PAPER SERIES
INCOME INEQUALITY AND LOCAL GOVERNMENT IN THE UNITED STATES,
1970-2000
Leah Platt Boustan
Fernando Ferreira
Hernan Winkler
Eric Zolt
Working Paper 16299
http://www.nber.org/papers/w16299
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
August 2010
We received useful comments from seminar participants at the UCLA and UC Berkeley Schools of
Law. We gratefully acknowledge financial support from the UCLA Ziman Center for Real Estate.
Fernando Ferreira would like to thank the Research Sponsor Program of the Zell/Lurie Real Estate
Center at Wharton for financial support. The views expressed herein are those of the authors and do
not necessarily reflect the views of the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.
© 2010 by Leah Platt Boustan, Fernando Ferreira, Hernan Winkler, and Eric Zolt. All rights reserved.
Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided
that full credit, including © notice, is given to the source.
Income Inequality and Local Government in the United States, 1970-2000
Leah Platt Boustan, Fernando Ferreira, Hernan Winkler, and Eric Zolt
NBER Working Paper No. 16299
August 2010
JEL No. H7
ABSTRACT
The income distribution in many developed countries widened dramatically from 1970 to 2000. Scholars
speculate that inequality contributes to a host of social ills by weakening the public sector. In contrast,
we find that growing income inequality is associated with an expansion in revenues and expenditures
on a wide range of services at the municipal and school district levels in the United States. These results
are robust to a number of model specifications, including instrumental variables that deal with the
endogeneity of local expenditures. Our results are inconsistent with models that predict heterogeneous
societies provide lower levels of public goods.
Leah Platt Boustan
Department of Economics
8283 Bunche Hall
UCLA
Los Angeles, CA 90095-1477
and NBER
lboustan@econ.ucla.edu
Fernando Ferreira
The Wharton School
University of Pennsylvania
1461 Steinberg - Dietrich Hall
3620 Locust Walk
Philadelphia, PA 19104-6302
and NBER
fferreir@wharton.upenn.edu
Hernan Winkler
Department of Economics
UCLA
Los Angeles, CA 90095-1477
hwinkler@ucla.edu
Eric Zolt
UCLA Law School
Box 951476, 2113 Law Bldg
Los Angeles, CA 90095-1476
zolt@law.ucla.edu
I. Introduction
Over the past thirty years, the income distribution has widened dramatically in the United
States and many other developed countries (Piketty and Saez, 2003; Smeeding, 2004). Income
inequality is correlated with several negative social outcomes—including high crime rates, low
levels of education achievement, and bad health.1 Yet, little is known about whether these
relationships are causal and, if so, the channels through which a widening income distribution
might translate into these social ills.
One frequently proposed mechanism is that income inequality may weaken the public
sector. Some political economy models suggest that, in heterogeneous societies, residents cannot
agree either on the composition of public goods or on the taxes and charges used to fund them
(Benabou, 1996, 2000). In particular, rich households may rely on private alternatives to public
goods and the poor may prioritize personal consumption over public contributions, generating
dissent between the ends and the middle of the income distribution (Epple and Romano, 1996).2
On the other hand, models based on the median voter theorem predict that a widening of the
income distribution will encourage greater use of progressive taxation for redistribution (Meltzer
and Richard, 1981; Alesina and Rodrik, 1994; Persson and Tabellini, 1994). Societies with
greater inequality may also have greater needs, leading altruistic voters to support social
programs.
Existing empirical work has not provided definitive evidence for the direction of the
relationship between income inequality and the size of the public sector.3 Two types of
1
See, inter alia, Kawachi, et al., 1997; Kennedy, et al., 1998; and Fajnzylber, Lederman and Loayza, 2002. For an
opposing view, see Deaton and Lubotsky, 2002.
2
Heterogeneity can also reduce social capital between residents, which may undermine trust, norms of reciprocity,
and support for local government activity (Putnam, 2000; Boix and Posner, 1998; Costa and Kahn, 2003).
3
In a cross-section of countries, countries with high levels of inequality, like the United States, engage in less public
spending (see, for example, Lindert, 1994, 1996; Moene and Wallerstein, 2005; Schwabish, Smeeding, and Osberg,
2006). In contrast, comparisons across US states and within states over time find that rising income inequality is
1
identification problems compound the lack of a consistent empirical relationship between income
inequality and public goods provision. Cross-country comparisons suffer from omitted variable
bias; that is, countries with high income inequality may also have other characteristics that could
limit the size of the public sector. Cross-state comparisons additionally suffer from endogenous
household sorting. If high-income families migrate to states with high public expenditures, the
positive association observed in the literature between state public expenditures and income
inequality may be spurious.
In this paper, we examine the relationship between income inequality and government
finances at the local level in the United States from 1970 to 2000. We focus mainly on
municipalities and school districts, but also present estimates for states. Local government
represent a large segment of the economy; in the 2009 fiscal year, local governments disbursed
more than $3 trillion in aggregate for such important services as education and public safety.4
Our study has several advantages over existing empirical work. First, large samples of
municipalities and school districts exhibit much greater variation in income inequality over time
than do the small number of countries or states used in previous studies. Secondly, we develop
an instrumental variable strategy to mitigate concerns about potential reverse causality from the
endogenous sorting of households across localities. Our procedure synthetically advances the
income distribution in a city or school district forward from 1970 by matching the initial income
distribution to national patterns of income growth over the next decades. By design, our
instrument cannot be influenced by mobility into and out of communities; rather, it isolates the
accompanied by higher government expenditures and increasing progressivity in the state tax code (Chernick, 2005;
Schwabish, 2008).
4
State governments accounted for $1.36 billion in expenditures in 2009, while all other local governments (cities,
school districts, etc.) accounted for $1.72 billion in that fiscal year. The federal government spent $3.52 billion in
2009. Beyond cities and school districts, counties and special districts provide local services, though these
governmental units represent a relatively small share of the total expenditures. These facts were compiled from the
website http://www.usgovernmentspending.com/.
2
component of change in the local income distribution that is driven by shifts in the return to skill
over time.
We find no evidence that an increase in income inequality reduces expenditures on public
services in cities or school districts; rather, as the income distribution widens, localities increase
their revenue collection and expenditures. Our best causal estimates suggest that the average
increase in the city-level Gini coefficient over this period (5 points) leads to a $63 increase in
expenditures per resident. These values imply that the widening of the income distribution from
1970 to 2000 can explain 15 percent of the growth in municipal expenditures over this period.
Among school districts, the average change in the Gini is associated with a $198 increase in
property tax revenue per pupil with a corresponding $190 decline in state transfers. Although
rising inequality can explain 29 percent of the growth in property tax revenue from 1970 to 2000,
state systems of school finance equalization appear to have undone much of the connection
between changes in the local income distribution and local revenue collection.
For municipalities, rising income inequality is not only associated with increased
expenditures on police services, which we may expect if inequality also leads to higher crime
rates, but also generates additional outlays for fire protection and road maintenance. In related
results, we find that growing racial fractionalization is associated with larger government
expenditures across a wide range of expenditure categories, casting doubt on earlier findings that
more racially fragmented cities spend a smaller share of their budget on public goods (Alesina,
Baqir and Easterly, 1999; see also Cutler, Elmendorf and Zeckhauser, 1993; Hopkins, 2009).
State level results also show a positive impact of inequality on local finances, but the estimates
are not precise enough to be statistically distinguished from zero.
3
Our results are consistent with recent work by Corcoran and Evans (2010), which
documents a positive relationship between income inequality and educational expenditures at the
school district level.5 Yet, a series of papers have found that, before World War II, unequal
communities raised less local revenue and provided fewer common goods and services (Goldin
and Katz, 1999; Ramcharan, 2009; Galor, Moav and Vollrath, 2009; Zolt, 2009). Taken together,
these results suggest that the relationship between income inequality and the size of local
government has changed over time. This change may be due to shifts in the sources of local
revenue away from property taxation toward more regressive revenue sources like sales taxes
and direct charges, or to the increasing role of state governments in funding (and, in some cases,
providing) goods that have historically been the responsibility of localities.
Our findings are inconsistent with models that predict that heterogeneous societies are
unable to compromise on common public goods and services. While our evidence is more
supportive of the median voter model, we caution that cities and school districts do not rely on
progressive forms of revenue, such as income taxation, and rarely engage in spending that is
explicitly redistributive. Therefore, it is unlikely that, in this context, rising inequality lowers the
tax price of public services for the median voter. With this caveat in mind, we prefer to
emphasize our substantive findings; reconciling these patterns with models of local political
economy provides a rich area for future research. Overall, our findings challenge the hypothesis
that income inequality reduces the provision of public goods from local governments in the
United States.
The remainder of the paper is organized as follows. The next section discusses our
measures of income inequality and government activity at the local level. Section III describes
Our results were generated independently of Corcoran and Evans’ recent study. We reach similar conclusions
despite using different methods to measure income inequality within school districts and developing a different
instrument for changes in inequality at the local level.
5
4
our panel estimation as well as an instrument for shifts in the local income distribution. Section
IV documents the positive relationship between changes in local inequality and growing
revenues and expenditures at the city, school district and state levels. Section V concludes.
II. Data on Income Distribution and Government Activity at the Local Level
II.A. Income Inequality
We collect decadal data on the income distribution and the levels of expenditures and
revenues from 1970 to 2000 for a large number of cities and school districts. The municipal
sample consists of a balanced panel of every Census-defined place (incorporated city or town)
with 2,500 or more residents in 1970. We exclude the 903 municipalities that were directly
responsible for providing education services, leaving us with a sample of 3,369 cities and towns.
The majority of our sample is made up of small towns: 65 percent of the municipalities in the
sample have fewer than 10,000 residents. Our school district sample contains the 9,024 districts
with more than 2,500 residents in 1970.6
Because of Census privacy restrictions, we cannot recover the full income distribution at
the local level. Instead, we use published Census reports, which indicate the number of
households in a jurisdiction in each of 15 to 20 income categories, to generate an (approximate)
income distribution. We assign each household an income level equal to the median income in its
6
The Census of Population provides demographic information for 11,687 and 14,405 school districts in 1970 and
2000, respectively. We use the School District Geographic Reference File for 1970 to combine the demographic
information with expenditure data from the Census of Governments (available at
http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/3515/detail). The sample consists of the 9,024 school districts
that could be matched between 1970 and 2000. This sampling rule eliminates school districts that eventually
disappear from the data due, for example, to consolidations with other districts. We choose not to aggregate districts
that eventually consolidate because the political economy mechanism that we have in mind pertains to the actual
voters and residents of a district. As a result, a component of the measured variation in income inequality over time
within a district will be due to mergers with neighboring districts.
5
bin by decade as calculated from Census micro-data. We then generate Gini coefficients at the
local level for this modified income distribution.7
In 1970, the average municipality in our sample had a Gini coefficient of 0.32, compared
to the national Gini coefficient of 0.39 (Table 1). By 2000, the Gini coefficient in the average
municipality increased by 5.5 points to 0.38. However, this average increase hides tremendous
variation across municipalities. The Gini coefficient increased by less than one point (or even
decreased) in one third of the cities in our sample, while in another third the Gini coefficient
increased by more than five points.
II.B. City Finances
The Census (and Surveys) of Governments provide information on municipal revenues
and expenditures by detailed category. The first panel of Table 1 contains summary statistics on
the sources of revenue and the categories of current expenditures at the municipality level. All
values are reported in year 2000 dollars. In the average municipality, expenditures per resident
doubled from $460 in 1970 to $870 by 2000. Municipalities allocate the majority of their budget
toward the maintenance of local infrastructure and on fire and police protection. Spending on
infrastructure, including roads, sewers, water and electricity, comprise 44 percent of average
municipal budgets and spending on police and fire protection make up another 21 percent. In
comparison, redistribution in the form of direct public welfare and expenditures on health and
public hospitals contribute a negligible amount (less than five percent) of the typical municipal
budget.
7
Without a full set of micro data at the municipal level, we are unable to calculate other measures of inequality,
such as the 90-10 ratio, with sufficient accuracy.
6
In 1970, property taxes were the largest source of municipal revenue, accounting for 33
percent of total proceeds. By 2000, the reliance on property taxes declined to only 22 percent of
the total budget, replaced in large part by inter-governmental transfers and direct charges for
services.8 Sales taxes also increased from a negligible portion of the budget in 1970 to 12 percent
of total revenue in 2000. Political economy models often assume that tax revenue is generated
through a progressive tax instrument, such as an income tax. However, most property and sales
taxes are regressive in the sense that they require higher tax payments as a share of total income
from poor households (Suits, 1977; Phares, 1985).9 Direct charges may be even more regressive
than property taxation because they are levied on a per house basis rather than tied to the value of
the home.10 On the other side of the ledger, inter-governmental transfers are often financed
through progressive state or federal income taxes; however, the tax burden for these transfers
disproportionately falls on households living outside of the locality in question.
We caution that higher government expenditures need not be synonymous with a higher
quality or quantity of public services for the average resident. First, the majority of government
expenditures cover the wages and salaries of municipal workers, an increase in which may not
translate into a higher quality of service provision. Secondly, anecdotal evidence suggests that a
greater share of city services are directed toward high-income neighborhoods; however, with
existing data sets, we cannot observe how municipal services are allocated within the
jurisdiction. Finally, we note that local governments may expand certain programs in order to
8
The relative decline in property taxes from 1970 to 2000 was part of a larger decline in the use of local property
taxes over the twentieth century (Oates and Schwab, 2004; Sokoloff and Zolt, 2007). This trend was accelerated in
the 1980s by statutory limits on the level or growth of property tax rates in some states.
9
Specific features of the tax system, including exemptions for food and other items from sales taxes or initial
threshold exemptions from property taxes, can affect the incidence of these instruments. There is significant
scholarly debate about the true incidence of the property tax (see Mieszkowski, 1972; Aaron, 1974; Musgrave, 1974
and Hamilton, 1976).
10
The largest categories of direct charges are for sewers (23 percent), hospitals (20 percent), airports (8 percent) and
sanitation services (8 percent).
7
combat new social problems associated with rising income inequality, thereby leaving the level
of public services unchanged. For example, inequality has been linked to higher rates of violent
crime (Fajnzylber, Lederman and Loayza, 2002). Cities may hire additional police officers to
combat the higher crime rates, resulting in more government spending without net improvements
in public safety.
II.C. School District Finances
The second panel of Table 1 presents the descriptive statistics for our school district
sample. In 1970, the typical district spent $4,140 per pupil. By 2000, this total nearly doubled to
$7,868 per pupil. The sources of school district revenue changed dramatically over this period.
While, in 1970, school revenues were evenly split between local property taxes and intergovernmental transfers, by 2000 state and federal transfers made up 70 percent of the average
school district budget.
The changing pattern of revenues in our sample reflects the increasing centralization of
K-12 funding over time. States began to supplement local revenues for education services in the
mid-twentieth century. At that time, state aid was typically disbursed as a flat grant per pupil,
with additional funds provided to poor districts (Hoxby, 2001). In 1965, the federal government
began providing school funding through Title I of the Elementary and Secondary School Act
(Cascio, et al., 2010). As a result, by 1970, locally-raised revenue only accounted for 60 percent
of school district budgets.
More recently, the use of local revenue sources, even as a supplement to state aid, has
been called into question. Property taxes allow wealthy districts to raise more revenue than poor
districts at the same tax rate, thereby generating an association between the level of wealth in a
8
district and its level of school funding. Starting with the Serrano v. Priest decision in California
(1971), many state supreme courts have ruled that existing systems of local school finance are
unconstitutional.11
In response to these legal challenges, states have adopted various plans to equalize school
funding across districts (Hoxby, 2001; Metzler 2003). The most common approach has been to
modify a state’s aid formula in order to directly supplement districts with smaller local property
tax capacity. Some states also guarantee that districts will be able to raise a certain level of
revenue at a given tax rate; the difference between locally raised revenue and the guaranteed
level is then made up by the state. Following this wave of reforms, the share of school revenues
raised through local property taxes declined from 60 percent in 1970 to 30 percent in 2004.
III. Estimating the Relationship Between Income Inequality and Government Activity
III.A. Basic Specification for Municipalities
The relationship between income inequality and public finances can be described by the
following equation:
yit = β(Gini)it + ΓXit + εit
εit =
i
+
it
(1)
where i indexes a city or town in Census year t, y is a local public finance outcomes such as total
expenditures, Gini is the Gini coefficient, and the coefficient β indicates the estimated effect of
income inequality on local finances. X contains a set of time-varying city characteristics,
11
Differences in school funding on the basis of local property wealth have been found to violate rights to equal
protection under some state constitutions (Briffault, 2006). In other states, local financing violates constitutional
provisions requiring that the state provide an adequate elementary and secondary education to all students. Claims
under the Federal equal protection clause were denied by the Supreme Court in San Antonio Independent School
District v. Rodriguez.
9
including total population, the share of the population that is black, Hispanic, or over 65 years of
age, and median household income. εit captures the unobserved determinant of local finances,
which depends on a permanent component
i
and a transitory component
it.
Pooling data from 1970 to 2000, we estimate the following equation in first differences to
absorb the permanent component of the error term ( i):
∆yit = β(∆Gini)it + Γ∆Xit + Rit + ∆
it
(2)
where equation (2) also includes the vector Rit to allow each Census region to have distinct time
trends in both patterns of government finances and income inequality. The coefficient of interest
(β) indicates the relationship between changes in the Gini coefficient and changes in government
revenue or expenditure within a municipality over time, holding constant changes in median
income and basic demographics. For the rest of the paper we refer to equation (2) as the OLS
specification.
III.B. Instrumental Variable for Income Inequality
Equation 2 is not sufficient, on its own, to establish a causal relationship between income
inequality and local government finances. On the one hand, the income distribution may affect
government activity through a number of channels: the preferences of local voters, compensatory
transfers from the state and federal government, or simply a mechanical relationship between
inequality and the size of the local tax base. However, it is also possible that changes in
government expenditures could induce shifts in the local income distribution. For instance, an
increase in local expenditures may attract wealthy households who prefer generous public
10
services even at the expense of higher taxes. These high-income arrivals would widen the local
income distribution.
To mitigate concerns about this form of reverse causality, we construct an instrumental
variable that is correlated with changes in an area’s Gini coefficient but is not otherwise
associated with changes in local revenues or expenditures. Our instrument is based on a
―synthetic‖ version of the income distribution in a municipality. Recall that the actual Gini
coefficient is calculated from counts of the number of households in a locality by income bin in
every decade. The first step in constructing our instrument is to replace these decade-specific
household tallies with the initial (1970) distribution of households by income bin. By freezing
the distribution of households across bins in 1970, we foreclose the possibility that richer or
poorer households move into a town in search of a given bundle of public goods.
We then allow the income level of households in the synthetic distribution to grow over
time according to the actual change in median income by income bin and decade from the
Census micro-data.12 As a result, time series variation in the synthetic distribution stems only
from national patterns of income growth by segment of the income distribution. In other words,
the initial income distribution in an area serves as a set of weights indicating how national
income growth likely affects each locality. For example, in the 1980s, the income level of
households in the top income bin grew faster than those for the rest of the distribution. The
instrument will therefore predict greater changes in the Gini coefficient over the 1980s in
municipalities that started out with a large number of high-income households in 1970.
12
To calculate the median income of a 1970 income bin in later decades, we convert the endpoints of each bin,
which are denominated in absolute income levels, into percentiles of the income distribution. Results are
qualitatively similar when we allow changes in median income by bin and decade to vary by region.
11
We present the first stage relationship between the actual and synthetic Gini coefficients
in graphical form in Figure 1 both in level and in changes. We find a strong positive relationship
between the two measures, suggesting that much of the change in local income distributions
from 1970 to 2000 was driven by trends in income growth, rather than by in- and out-mobility of
households from the top or bottom of the income distribution. The F-statistic on the relationship
between the actual and synthetic Gini coefficients is 975.77, surpassing the conventional
threshold for a strong instrument by two orders of magnitude.
III.C. Additional Specification for School Districts
Because of the substantial changes in the arrangement of school finance over this period,
analyzing the relationship between income inequality and school district revenues requires some
care. In particular, we want to allow for the possibility that an increase in income inequality may
have different effects in states with and without school financing equalization plans. Districts
that experience rising income inequality due to income growth for the rich may be heavily taxed
by state equalization plans, whereas districts with inequality driven by falling incomes among the
poor may be heavily subsidized.
We define SFR (―school finance reform‖), an indicator variable equal to one in states
whose systems of school finance have been deemed unconstitutional by the state supreme court.
This condition that applies to 14 states by 2000.13 Equation 3 interacts this state-level reform
indicator with changes in the school district-level Gini coefficient. We estimate:
∆yit = θ(SFR)it + β1(∆Gini)it + β2(∆Gini · SFR)it + Rit + Γ∆Xit +
13
it
(3)
We rely on Card and Payne’s (2002) taxonomy of school finance cases as updated by Baicker and Gordon (2006).
12
where i indexes school districts and t indicates the Census decade (t = 1970, 2000). The
coefficient β1 summarizes the relationship between changes in income inequality and changes in
revenues or expenditures per pupil in the average school district. The coefficient β2 tests whether
this relationship is different in states that fell under court order to reform their system of school
finance by 2000. We also allow the effect of district-level median income to vary according to a
state’s school finance regime.
We should note that some states that did not face a court order to equalize school
spending over this period might have reformed their school finance systems preemptively in
order to avoid the threat of litigation (Metzler, 2003). In this case, the two groups of states may
respond equivalently to changes in inequality, leading the coefficients on the interaction terms to
be indistinguishable from zero.
IV. Results
IV.A. Impact of Income Inequality on Municipalities
Table 2 presents results from equation 2, which estimates the relationship between
changes in income inequality and changes in government revenue or expenditure within a city
over time. We find that an increase in inequality leads to modest growth in municipal revenues
and expenditures. The coefficients imply that a five point increase in the Gini coefficient, the
average change in the Gini over this period, is associated with a $27 increase in expenditures per
capita. Police spending represents $3 of the total increase in municipal expenditures, while the
remainder is spent on other ―productive‖ public services including fire protection and local
roads. Income inequality has little effect on spending for either public welfare or health and
13
hospitals; however, together, these categories represent less than five percent of the typical
municipal budget.
The revenues required to fund these expenditures are collected by means of a range of
local tax instruments, including property taxes, sales taxes, and direct charges for services.
Higher property tax revenues could stem either from a decision to increase the tax rate or from a
more mechanical relationship between inequality and the property tax base. The one revenue
category that is not associated with a widening of the income distribution is federal and state
transfers. This may not be surprising because the majority of state transfers to local governments
are provided to school districts, which are examined in the next section, and because state
transfers to municipalities are based on formulas that often do not take into account the local
income distribution.
Table 3 considers heterogeneous effects of inequality on government revenue by initial
municipality size and by initial median income. We subdivide the sample first by median size in
1970 (6,500 residents) and then by median household income ($41,000 in 2000 dollars). An
increase in income inequality is associated with greater revenue collection in all cases. However,
the positive relationship between income inequality and government revenues is strongest in
smaller and richer towns. Residents of smaller towns may develop a stronger base of social
capital and therefore be willing to fund public goods even as the ends of the income distribution
begin to pull away from the middle. We note that, in particular, rich towns are more likely to
expand their property tax collection as the income distribution widens. This is consistent with a
greater increase in top-end inequality in rich towns; as the rich get richer, the property tax base
may increase.
14
Table 4 contains results from the second stage of our instrumental variables analysis, in
which we instrument for actual changes in the Gini coefficient with changes in national income
growth weighted by the initial income distribution in a locality. Most of the IV coefficients are
positive, statistically significant and, if anything, are larger than their OLS counterparts. In the
IV specification, rising inequality increases municipal revenue from all major sources, including
inter-governmental transfers. However, the relationship between inequality and the separate
categories of expenditures are not statistically significant in the IV regressions, with the
exception of fire protection.
If our OLS estimates were plagued by reverse causality – for example, because the rich
are attracted to towns with generous public services – we would expect the IV coefficients to be
smaller than OLS. The fact that the IV estimates are larger than OLS suggests that the
instrumental variables procedure may instead be correcting for measurement error, which can
bias estimates towards zero. By these estimates, a five point increase in the Gini coefficient leads
to a $63 increase in expenditures per capita. From 1970-2000, the average municipality
experienced a $410 increase in revenues per capita. The widening of the income distribution can
thus explain 15 percent of the growth in the size of local governments from 1970 to 2000 (=
63/410). Overall, the pattern of both OLS and IV results suggests that income inequality neither
reduces the demand for municipal goods and services nor does it limit residents’ ability to pay
for them.
IV.B. Impact of Change in Racial Heterogeneity on Municipalities
Table 5 examines the effect of another form of local heterogeneity, racial
fractionalization, on municipal budgets. Alesina, Baqir and Easterly (1999) argue that, although
15
cities with a racially diverse population spend more per resident, they devote a smaller share of
their budget to ―productive‖ public goods, such as roads, sewers and trash collection. We reestimate equation 2, replacing the separate measures of black and Hispanic population share with
an index of racial/ethnic fractionalization. Our index is based on four racial/ethnic categories:
white, non-Hispanics; black, non-Hispanics; Hispanics; and other races (which include Asians,
Pacific Islanders and American Indians).14 We improve upon the methodology used in Alesina,
et al. by using a panel of cities from 1970 to 2000, rather than a single cross-section in 1990, and
by extending the analysis to municipalities with fewer than 25,000 residents.
As in Alesina, et al., we find that an increase in racial heterogeneity is associated with
larger municipal expenditures. While half of the increase is due to higher police spending, we
also find large positive effects on fire protection and health and hospital spending. Because
spending on roads fails to keep pace with the overall increase in expenditures, the share of the
budget dedicated to roads does fall, which Alesina, et al. interprets as a decline in the share of
revenue dedicated to productive public goods. However, we contend that the interpretation of
these patterns are extremely sensitive to the classification of municipal spending into
―productive‖ versus ―non-productive‖ public goods. It is reasonable to believe that spending on
fire protection and public hospitals are equally as productive as spending on roads and,
conversely, that spending on roads is equally susceptible to corruption for patronage purposes.
On the revenue side, we confirm Alesina et al.’s finding that racial heterogeneity is
associated with an increase in inter-governmental transfers. However, we dispute the
interpretation that racially diverse cities are unwilling to raise their own revenue and therefore
need to be subsidized by the state ―to compensate…[for] the difficulties…in directing local
The racial fractionalization index is defined as 1 – Σi (Number of residents of race or ethnicityi)2. Separate counts
of Asian and Pacific Islanders do not exist at the municipal level in 1970 or 1980.
14
16
resources to the supply of public goods‖ (p. 1266). Instead, we find that an increase in racial
diversity is also associated with an increase in own-source revenue collection, including both
property and sales taxes.
IV.C. Impact of Income Inequality on School Districts
Turning to school districts, we begin in Table 6 by estimating the baseline specification
(equation 2), which relates decadal changes in income inequality to changes in government
activity, first in OLS and then using our instrument for district-level changes in inequality. As for
municipalities, we find that an increase in income inequality among residents of a school district
is associated with rising expenditures per pupil. However, the relationship between income
inequality and total expenditures per pupil is small. According to our IV estimate, a 1.1 point
increase in the Gini coefficient, the average increase at the district-level from 1970 to 2000,
would result in only $29 additional dollars of expenditure per pupil.
The total effect of inequality on school resources masks countervailing trends for the two
main sources of revenue. A 1.1 point increase in the Gini is associated with a $198 increase in
property tax revenue per pupil and a corresponding $190 decline in state transfers. This pattern is
consistent with the prospect that state systems of school finance equalization worked to undo the
association between the local income distribution and local revenue collection. The next table
tests this hypothesis more directly.
Table 7 presents coefficients from equation 3, which allows the effect of income
inequality on school district finances to vary with a state’s system of school finance. For this
specification, we consider long-run changes in school expenditures from 1970 to 2000 in order to
allow the reforms of the 1970s and 1980s time to take hold. Table 7 reports only OLS results
17
because the instrument is not sufficiently powerful to explain changes in inequality over this
thirty year interval. The first panel replicates the basic specification (presented in Table 6) over
this thirty year period; in the second panel, we allow the relationships between a district’s level
of income inequality and its median income to differ in states with and without court-ordered
school finance reform.
The first row of Table 7 shows that states under court-order to reform their system of
school finance provide a higher level of state transfers per pupil (see also Card and Payne, 2002).
By 2000, the average district under court order received an additional $474 of state funding per
pupil. However, a portion of this state transfer was reversed by a corresponding reduction in
local property tax revenue. Overall, we find no difference in the level of total expenditures per
pupil in states with and without court-ordered school finance reform, but instead see differences
only in the source of this revenue. As in Table 6, we document that school districts in which the
income distribution widened between 1970 and 2000 raise more revenue per pupil from property
taxation. The magnitude of these effects is somewhat smaller than the decadal IV estimates; a 1.1
point increase in the Gini coefficient is associated with $57 of additional property tax revenue
per pupil. In this specification, we do not find a corresponding decline in state transfers in the
average district. However, as the next panel shows, the relationship between inequality and state
transfers differs in states with and without court-ordered school finance reform.
Panel 2 demonstrates that the relationship between income inequality and school
expenditures is mediated by a state’s system of school finance. In districts whose state system of
school finance are not under court supervision, rising inequality is positively related to both
property tax revenue and state transfers, such that a 1.1 point increase in the Gini would lead to a
$61 increase in total resources per pupil. However, in states required to equalize school funding,
18
a 1.1 point increase in the Gini coefficient is associated with a $44 decline in state transfers per
pupil offset by a corresponding $94 dollar increase in property tax revenue. This pattern could be
generated by a rise in top-end inequality. As the rich get richer, the property tax base in a district
may expand, allowing for greater amounts of own-source revenue collection. However, in states
with strong equalization programs, most of the excess taxing capacity that accompanies rising
income inequality is offset by reductions in state aid.
Table 7 also reports the relationship between educational expenditures and the median
income of a school district’s residents. Not surprisingly, wealthier districts spend more on
education per pupil. On average, a five percent increase in median income is associated with a
$102 dollar increase in per-pupil expenditures. As in Card and Payne (2002), we find that the
presence of an equalization court order does not change the magnitude of the relationship
between local income and school expenditures but does alter the source of these additional funds.
In states that are not under court order, an increase in local median income is associated with
greater school resources from both own-source revenue and state transfers. In contrast, in
equalization states, state transfers do not increase with local median income. Instead, the loss of
state revenue is compensated with a stronger association between median income and property
tax revenue.
IV.D. Impact of Income Inequality on States
Lastly, we consider the relationship between changes in income inequality and changes in
revenues and expenditures at the state level. State-level tax systems vary greatly, both in their
choice of tax instruments and in their degree of progressivity. For example, while the average
state relies on personal income taxation for 12 percent of its revenue, nine states do not impose
19
an income tax at all. Although state income taxes are generally progressive, with the degree of
progressivity determined by the rate structure, the tax-free threshold, and the use of exemptions
and credits, the overall bundle of state taxes tend to be mildly regressive (Chernick, 2005; Davis,
et al., 2009).
In states with a progressive income tax system, the median voter model would predict a
stronger relationship between income inequality and state-level fiscal outcomes (Hayes and
Slottje, 1989; Fletcher and Murray, 2008). Descriptive statistics seem to confirm this intuition:
average state Gini increased from 0.357 in 1970 to 0.43 in 2000, while total revenues and
expenditures per capita increased by approximately $1,800 (on a basis of approximately $2,000)
over these decades.15 Yet, mechanisms like voter altruism and social capital may be less effective
for larger jurisdictions, which could attenuate (or even reverse) any positive relationship between
inequality and expenditures at the state level.
Table 8 presents state level results. Estimates are based on equation 2, with all municipal
fiscal outcomes replaced by state fiscal outcomes. We find that an increase in inequality has a
positive effect on state revenues and expenditures but all estimates are statistically
indistinguishable from zero at conventional levels. Consistent with median voter models, we do
see that a 7.3 point increase in the Gini coefficient, the average increase in this measure of
income inequality at the state level from 1970 to 2000, is associated with a $913 increase in
general tax revenue per capita. Similarly, changes in income inequality have a positive effect on
total state-level expenditures (but are statistically indistinguishable from zero). All expenditure
sub-items show a positive effect, with the exception of public welfare. Instrumental variable
estimates show similar patterns.16 Overall, Table 8 estimates suggest that increasing income
15
16
State fiscal data are collected from the Census of Governments. Appendix Table 1 provides summary statistics.
These estimates are available upon request.
20
inequality had, if any, a positive impact on the fiscal situation of state governments. However, as
opposed to our previous estimates at the city and school district level, the limited number of
observations at the state level does not allow us to draw precise inference about the impact of
income inequality on state finances.
V. Conclusion
The income distribution in the United States widened greatly from 1970 to 2000. We use
variation in income dispersion at the local level to examine the relationship between inequality
and the size of the public sector. Contrary to models that emphasize disagreements between
residents of heterogeneous societies over the optimal level of public expenditures, we find that
rising income inequality is associated with larger increases in tax revenues and faster growth in
public expenditures at municipal, school district and state levels.
Revenues and expenditures per resident increased in nearly all communities over this
period. Our best causal estimates suggest that a five point increase in the Gini coefficient,
roughly the change in the average locality from 1970 to 2000, leads to a $63 increase in
municipal expenditures per resident to cover services like police and fire protection and
infrastructure maintenance and a $190 increase in locally-raised school expenditures per pupil.
By the estimates, the widening of the income distribution can explain around 15 percent of the
growth in the size of local government over the period.
We conclude by noting that, although income inequality is associated with greater public
expenditures, it is not clear that additional funds necessarily translate into a larger quantity or
higher quality of public goods. Furthermore, the incidence of local taxation and the distribution
of local services need not be progressive and likely varies substantially across governmental
21
units. Hence, we stop short of claiming that local government activity wholly or partially
compensates for the potential social ills associated with income inequality. However, given the
empirical patterns documented here, we argue that it is unlikely that the social ills correlated with
inequality are due to a weakening of the public sector.
22
Bibliography
Aaron, Henry J., 1974. ―A New View of Property Tax Incidence.‖ American Economic Review.
64, 212−221.
Alesina, Alberto, Reza Baqir, and William Easterly, 1999. ―Public Goods and Ethnic Divisions.‖
Quarterly Journal of Economics. 114(4), 1243-1284
Alesina, Alberto, and Dani Rodrik, 1994. ―Distributive Politics and Economic Growth.‖
Quarterly Journal of Economics. 109, 465–490.
Baicker, Katherine, and Nora Gordon, 2006. ―The Effect of State Education Finance Reform on
Total Local Resources.‖ Journal of Public Economics. 90, 1519-1535.
Benabou, Roland, 1996. ―Heterogeneity, Stratification, and Growth: Macroeconomic
Implications of Community Structure and School Finance.‖ American Economic Review.
86, 584-609.
Benabou, Roland, 2000. ―Unequal Societies: Income Distribution and the Social Contract.‖
American Economic Review. 90, 96-129.
Boix, Carles, and Posner, Daniel N., 1998. ―Social Capital: Explaining its Origin and Effects on
Governmental Performance.’ British Journal of Political Science. 28, 686–93.
Briffault, Richard, 2006. ―Home Rule and Local Political Innovation.‖ Journal of Law and
Politics. 22(1).
Card, David, and Payne, A. Abigail, 2002. ―School Finance Reform, the Distribution of School
Spending, and the Distribution of Student Test Scores.‖ Journal of Public Economics.
83(1), 49-82.
Cascio, Elizabeth, Nora Gordon, Ethan Lewis, and Sarah Reber, 2010. ―Paying for Progress:
Conditional Grants and the Desegregation of Southern Schools.‖ Quarterly Journal of
Economics. 125(1), 445-482.
Chernick, Howard, 2005. ―On the Determinants of Subnational Tax Progressivity in the U.S.‖
National Tax Journal. 58, 93-112.
Corcoran, Sean P., and William N. Evans, 2010. ―Income Inequality, the Median Voter, and the
Support for Public Education.‖ NBER Working Paper No. 16097.
Costa, Dora L., and Matthew E. Kahn, 2003. ―Understanding the American Decline in Social
Capital, 1952-1998.‖ Kyklos. 56, 17-46.
Cutler, David M., Douglas W. Elmendorf, and Richard Zeckhauser, 1993. ―Demographic
Characteristics and the Public Bundle.‖ Public Finance. 48, 178-198.
23
Davis, Carl, Kelly Davis, Matthew Gardner, Robert S. McIntyre, Jeff McLynch and Alla
Sapozhnikova, 2009. Who Pays? A Distributional Analysis of the Tax Systems in All 50
States. Washington, D.C.: Institute on Taxation and Economic Policy.
Deaton, Angus, and Darren Lubotsky, 2003. ―Mortality, Inequality and Race in American Cities
and States.‖ Social Science and Medicine. 56(6), 1139-53.
Epple, Dennis, and Richard E. Romano, 1996. ―Ends Against the Middle: Determining Public
Service Provision When There Are Private Alternatives,‖ Journal of Public Economics.
62, 297-325.
Fajnzylber, Pablo, Daniel Lederman, and Norman Loayza, 2002. ―Inequality and Violent
Crime.‖ Journal of Law and Economics. 45, 1-39.
Fletcher, Jason M. and Matthew N. Murray, 2008. ―What Factors Influence the Structure of the
State Income Tax?‖ Public Finance Review. 36, 475-496.
Galor, Oded, Omer Moav, and Dietrich Vollrath, 2009. ―Land Inequality, the Emergence of
Human Capital Promoting Institutions and the Great Divergence.‖ Review of Economic
Studies. 76(1), 143-179.
Goldin, Claudia, and Lawrence F. Katz, 1999. ―Human Capital and Social Capital: The Rise of
Secondary Schooling in America, 1910-1940.‖ Journal of Interdisciplinary History. 26,
683-723.
Hamilton, Bruce W., 1976. ―Capitalization of Intrajurisdictional Differences in Local Tax
Prices.‖ American Economic Review. 66:5, 743-53.
Hayes, Kathy and Daniel J. Slottje, 1989. ―The Efficacy of State and Local Governments’
Redistributional Policies.‖ Public Finance Quarterly. 17, 304-322.
Hopkins, Daniel J., 2009. ―The Diversity Discount: When Increasing Ethnic and Racial Diversity
Prevents Tax Increases.‖ Journal of Politics. 71, 160-177.
Hoxby, Caroline M., 2001. ―All School Finance Equalizations Are Not Created Equal.‖
Quarterly Journal of Economics. 116, 1189-1231.
Kawachi, Ichiro, Bruce P. Kennedy, Kimberly Lochner, and Deborah Prothrow-Smith, 1997.
―Social Capital, Income Inequality, and Mortality.‖ American Journal of Public Health.
87(9), 1491-8.
Kennedy, Bruce P., Ichiro Kawachi, Deborah Prothrow-Stith, Kimberly Lochner, and Vanita
Gupta, 1998. ―Social Capital, Income Inequality, and Firearm Violent Crime.‖ Social
Science Medicine. 47, 7-17.
Lindert, Peter H., 1994. ―The Rise in Social Spending, 1880–1930.‖ Explorations in Economic
History. 31, 1-37.
24
Lindert, Peter H., 1996. ―What Limits Social Spending?‖ Explorations in Economic History. 33,
1-34.
Mieszkowski, Peter M., 1972. ―The Property Tax: An Excise Tax or a Profits Tax?‖ Journal of
Public Economics. 1, 13-96.
Meltzer, Allan H., and Scott F. Richard, 1981. ―A Rational Theory of the Size of Government.‖
Journal of Political Economy. 89, 914-927.
Metzler, Jeffrey, 2003. ―Inequitable Equilibrium: Education Finance in the United States.‖
Indiana Law Review. 36, 561
Moene, Karl O. and Michael Wallerstein, 2005. ―Social Democracy as a Development Strategy.‖
In Pranab Bardhan, Samuel Bowles, and Michael Wallerstein, eds. Globalization and
Egalitarian Redistribution. Princeton: Princeton University Press.
Musgrave, Richard A., 1974. ―Is a Property Tax on Housing Regressive?‖ American Economic
Review. 64, 222-229.
Persson, Torsten, and Guido Tabellini, 1994. ―Is Inequality Harmful for Growth?‖ American
Economic Review. 84, 600-621.
Oates, Wallace E. and Robert M. Schwab, 2004. ―What Should Local Governments Tax: Income
or Property?‖ In A. Schwartz, ed., City Taxes, City Spending: Essays in Honor of Dick
Netzer. Cheltenham, U.K.: Edward Elgar, 7-29.
Phares, Donald, 1980. Who Pays State and Local Taxes? Cambridge, MA: Oelgeschalager, Gunn
and Hain.
Piketty, Thomas, and Emmanuel Saez, 2003. ―Income Inequality in The United States, 19131998.‖ Quarterly Journal of Economics. 118, 1-39.
Putnam, Robert D., 2000. Bowling Alone: The Collapse and Revival of American Community.
New York: Simon and Schuster.
Ramcharan, Rodney, 2009. ―Inequality and Redistribution: Evidence from U.S. Counties and
States, 1890-1930.‖ Review of Economics and Statistics. Forthcoming.
Schwabish, Jonathan A., Timothy M. Smeeding, and Lars Osberg, 2006. ―Income Distribution
and Social Expenditures: A Cross-National Perspective.‖ In Dmitri B. Papadimitriou, ed.
The Distributional Effects of Government Spending and Taxation. Northampton, MA:
Edward Elgar Publishing, 247-288.
Schwabish, Jonathan A., 2008. ―The Effects of Earnings Inequality on State Social Spending in
the United States.‖ Public Finance Review. 36(5), 588
25
Smeeding, Timothy M., 2004. ―Twenty Years of Research on Income Inequality, Poverty, and
Redistribution in the Developed World: Introduction and Overview.‖ Socio-Economic
Review. 2: 149-63.
Sokoloff, Kenneth L., and Eric M. Zolt, 2005. ―Inequality and Taxation: Evidence from the
Americas on How Inequality May Influence Tax Institutions.‖ Tax Law Review. 59(2),
167-241.
Suits, Daniel B., 1977. ―Measurement of Tax Progressivity.‖ American Economic Review. 67,
747-752.
Zolt, Eric M., 2009. ―Inequality, Collective Action, and Taxing and Spending Patterns of State
and Local Governments.‖ Tax Law Review. 62, 445-539.
26
.1
.2
.3
.4
.5
.6
Figure 1: First stage regression, Relationship between actual and synthetic Gini coefficients
at the municipal level, 1970-2000
A. Levels
.1
.2
.3
.4
Predicted Gini
.5
.6
45 degree line
Notes: Each point in the scatter diagram represents a municipality’s actual and predicted Gini coefficients. Gini
coefficients are calculated using the income bins from Census reports and the median income of each bin from
Census Microdata. The computation of predicted Gini coefficients is described in section III.B.
-.2
-.1
0
.1
.2
B. Changes
-.1
-.05
0
Predicted Gini (Change)
.05
.1
45 degree line
Notes: Each point in the scatter diagram represents the residual change in a municipality’s actual and predicted Gini
coefficients over a decade after controlling for changes in population, share of black and Hispanic population,
median income, share of individuals older than 65 and regional trends.
27
Table 1: Summary statistics, municipal and school district revenues and expenditures,
1970-2000
I. Municipalities (per capita)
A. Revenue
1970 Mean
1970 SD
Δ 1970-2000
B. Expenditure
1970 (Mean)
1970 (SD)
Δ 1970-2000
Gini
coefficient
General
revenue
Property
Inter-gov
Sales
transfers
Direct
charges
tax
0.320
(0.054)
0.055
449.7
(300.7)
428.6
149.1
(116.9)
45.2
99.4
(143.2)
100.9
90.2
(143.5)
105.1
3.1
(7.1)
57.5
General
expenditures
Police
Fire
Highways
Public
welfare
Health &
hospitals
462.2
(436.5)
410.0
76.1
(45.1)
56.6
31.4
(34.4)
20.0
60.7
(34.7)
9.8
0.6
7.7
1.9
20.6
117.5
10.4
Inter-gov
Direct
charges
Tax
II. School districts (per pupil)
1970 (Mean)
1970 (SD)
Δ 1970-2000
Gini
coefficient
Total
revenue
Total
expenditure
Property
tax
0.370
(0.039)
0.011
4185.9
(1833.9)
3794.1
4138.9
(2122.2)
3730.3
2071.4
(1597.9)
679.8
transfers
1821.0
(1036.1)
2757.9
238.3
(246.5)
48.1
Notes: Revenues and expenditures are reported in 2000 dollars. We provide the mean of each variable in 1970, the
standard deviation in 1970 in parentheses and the average change from 1970-2000 in italics. The municipality
statistics are for the 3,369 cities and towns with at least 2,500 residents in 1970 that do not provide education
services. The school district statistics reflect the 9,024 districts with more than 2,500 residents in 1970.
28
Table 2: OLS estimates of the relationship between income inequality and municipal
revenue and expenditures per capita, 1970-2000
General
revenue
474.1**
[209.3]
Property tax
Direct charges
Sales tax
119.3***
[43.09]
Inter-govern
Transfers
52.46
[85.09]
289.5**
[120.4]
262.5**
[120.5]
General
expenditures
536.3**
[209.2]
Police
Fire
Highways
Public welfare
53.81**
[25.31]
63.16***
[17.48]
68.41***
[19.70]
-4.399
[71.06]
Other tax
73.53***
[26.50]
Health &
hospitals
42.13
[207.9]
Notes: Sample includes municipalities in Census years 1970-2000 that were not responsible for education services in
1970 (N = 13476, or 3369 municipalities per year). Cells report the estimated coefficient on the change in the Gini
coefficient from equation 2 in text. Standard errors in parentheses and are clustered by municipality.
Table 3: The relationship between income inequality and municipal revenue per capita by
initial population and median income, 1970-2000
General
revenue
Property tax
Intergovern
transfers
Direct
charges
A. By initial population
163.1
331.3**
[103.7]
[166.0]
Below med
578.3*
[298.3]
88.87*
[49.39]
Above med
170.0
[233.5]
196.6***
[65.08]
Below med
97.11
[341.5]
57.80
[42.70]
Above med
802.2***
[261.2]
127.3*
[74.51]
-205.8
[138.7]
201.2
[133.3]
B. By initial median income
-72.36
352.0*
[116.1]
[191.8]
176.8
[128.1]
231.9*
[121.1]
Sales tax
Other tax
237.0
[150.9]
102.3***
[33.59]
263.7*
[141.7]
-0.866
[39.32]
270.8**
[132.0]
59.10*
[32.50]
211.0
[216.8]
66.48
[42.14]
Note: Sample includes municipalities in Census years 1970- 2000 that were not responsible for education services in
1970 (N = 13476, or 3369 municipalities per year). The first and third rows are the estimated coefficients on the
change in the Gini coefficient from equation 2 for a subsample of municipalities whose population (Panel A) or
household median income (Panel B) are below the sample median in 1970. The second and fourth rows report the
same coefficients but for those municipalities whose population (Panel A) or household median income (Panel B)
are above the sample median in 1970. Median initial population in 1970 was 6,430. Median initial household
median income was $41,273 in 2000 dollars. Standard errors in parentheses and are clustered by municipality.
29
Table 4: IV estimates of the relationship between income inequality and municipal revenue
and expenditure per capita, 1970-2000
General
revenue
1079*
[646.8]
Property tax
Inter-govern
transfers
596.3***
[195.3]
Direct charges
Sales tax
498.0**
[237.6]
657.8*
[349.1]
151.8
[96.83]
General
expenditures
1260*
[701.5]
Police
Fire
Highways
-44.87
[160.2]
134.4**
[65.56]
93.02
[75.39]
Public
welfare
-103.2
[237.0]
Health &
hospitals
329.7
[425.6]
620.2***
[175.2]
Other tax
Notes: Sample includes municipalities in Census years 1970-2000 that were not responsible for education services in
1970 (N = 13476, or 3369 municipalities per year). Cells report the estimated coefficient on the change in the Gini
coefficient from equation 2 in text. The instrument for the actual Gini coefficient is based on a ―synthetic‖ version of
the local income distribution; see Section IIIb for details. Standard errors in parentheses and are clustered by
municipality.
Table 5: OLS estimates of the effect of racial fractionalization on municipal revenue and
expenditures per capita, 1970-2000
General
revenue
108.7**
[52.39]
Property tax
General
expenditures
103.0*
[58.83]
Police
58.12***
[14.43]
53.76***
[11.58]
Inter-govern
transfers
56.51*
[30.92]
Direct charges
Sales tax
Other tax
-9.451
[29.57]
13.49
[37.15]
-0.911
[13.14]
Fire
Highways
Public welfare
5.167
[6.877]
-52.91*
[31.68]
Health &
hospitals
139.8*
[83.55]
21.75***
[6.555]
Notes: Sample includes municipalities in Census years 1970-2000 that were not responsible for education services in
1972 (N = 13476, or 3369 municipalities per year). Cells report the estimated coefficient β from equation 2 in text
but replacing the Gini coefficient by the index of racial fractionalization. Standard errors in parentheses and are
clustered by municipality.
30
Table 6: OLS and IV estimates of the relationship between income inequality and school
district revenue and expenditure per capita, 1970-2000
Total
spending
946.3
[657.8]
OLS
Property
tax
2471***
[625.5]
State
transfers
-2086***
[555.6]
Total
spending
2678
[4405]
IV
Property
tax
18084***
[4298]
State
transfers
-17341***
[2762]
Notes: Sample includes school districts in Census years 1970-2000 (N = 36,096, or 9,024 school districts per year).
Standard errors in parentheses and are clustered by school district. Cells report the estimated coefficient on the
change in the Gini coefficient from equation 2 in text. The instrument for the actual Gini coefficient is based on a
―synthetic‖ version of the local income distribution; see Section IIIb for details.
Table 7: Effect of median income and income inequality on school district revenue and
expenditures per pupil, 1970-2000
Court order (SFR)
Gini coefficient
Total
spending
72.10
[262.0]
Panel 1
Property
tax
-147.0
[259.7]
State
transfers
473.9**
[218.4]
Total
spending
-4506
[6531]
Panel 2
Property
tax
-9619
[8460]
State
transfers
14884**
[7033]
6089***
[1871]
5232***
[1694]
1414
[1612]
5626**
[2251]
4069**
[1893]
3210*
[1681]
796.9
[4457]
4503
[4256]
-7213**
[3437]
2043***
[143.3]
281.5***
[69.68]
1422**
[77.57]
403.2
[526.1]
731.0
[684.0]
-1099*
[583.9]
Gini · SFR
ln(median income)
ln(median) · SFR
2036***
[138.9]
265.1***
[68.85]
1447***
[80.34]
Notes: Sample includes school districts in Census years 1970 and 2000 (N = 18,048, or 9024 school districts per
year). Cells report the estimated coefficients of equation 3 in text. SFR is an indicator variable equal to one in the
year 2000 for the 14 states whose systems of school finance were deemed unconstitutional by the state supreme
court. Standard errors in parentheses and are clustered by school district.
31
Table 8: OLS estimates of the relationship between income inequality and state revenues
and expenditures per capita, 1970-2000
General
revenue
Sales
tax
Income
tax
Intergov.
Transfers
Charges &
Misc.
12503
(8058)
1773*
(1050)
20052**
(10064)
2735
(1893)
-22579
(15156)
General
expenditures
Public
safety
Highways
Education
Health &
hospitals
Public
welfare
13099
(11229)
167
(282)
388
(1375)
3633
(5569)
329
(933)
-39
(1269)
Notes: Sample includes all US states in Census years 1970-2000. Cells report the estimated coefficient on the
change in the Gini coefficient from equation 2 in text. Standard errors in parentheses and are clustered by state.
32
Appendix Table 1: Summary statistics, State revenue and expenditures per capita,
1970-2000
A. Revenue
1970 Mean
1970 SD
Δ 1970-2000
B. Expenditure
1970 (Mean)
1970 (SD)
Δ 1970-2000
Gini
coefficient
General
revenue
Sales
tax
Income
tax
Intergov.
transfers
Charges &
Misc.
0.357
(0.024)
0.073
General
expenditures
2057
(2241)
1717
Public
safety
601
(188)
281
Highways
253
(202)
486
Education
506
(220)
525
Health &
hospitals
515
(2030)
321
Public
welfare
1849
(642)
1805
43
(23)
117
367
(162)
-28
751
(250)
549
113
(35)
140
241
(103)
569
Notes: Revenues and expenditures are reported in 2000 dollars. We provide the mean of each variable in 1970, the
standard deviation in 1970 in parentheses and the average change from 1970-2000 in italics. The state statistics are
for all US states.
33