Asymmetry in Local Government Responses in Growing vs. Shrinking Counties: The Case of Education Finance

DOWNLOAD

Asymmetry in Local Government Responses in Growing vs.
Shrinking Counties: The Case of Education Financea
Abstract
Spending for k-12 education in the United States increased by more than 220% between 1972
and 2012, faster than can be explained by population growth (a 48% increase), growth in median
household income (a 32% increase), or changes in other economic, demographic, and
institutional variables. Importantly, school spending nearly doubled in places that experienced
ongoing population decline. In this paper, analysis reveals asymmetric responses in school
spending to changes in school age population growing and shrinking counties. This research
increases understanding of why education spending tends not to shrink in the face of ongoing
declines in school age population, a situation that exists in about 25% of counties.
Key Words: Education finance; asymmetry, declining; growing
JEL Code: H71
2
1. Introduction
Between 1972 and 2012 spending for k-12 education in the United States (US) grew by 220%,
much faster than can be accounted for by changes in income and demographics. Over the past
several decades numerous researchers have sought to explain the underlying factors that drive
the growth of government in industrialized countries. Berry, et al. (2012) conducted a detailed
empirical analysis of US local government spending growth over the 1962-2002 period using
data aggregated to the county level. The authors demonstrated that economic, demographic, and
institutional factors explain a significant portion of growth. Despite this, their evaluation reveals
that these factors do not fully explain growth in government over this period. In this sense,
Berry, et al. (2012) is similar to earlier empirical studies in that the typical socio-economic
variables motivated by models of government (Median Voter—Bowen and Black, 1957;
Leviathan—Brennan and Buchanan, 1980) as well as other considerations do not fully explain
the US local government growth experience. Interestingly, Berry, et al. (2012) also show that the
unexplained growth phenomenon exists even in places experiencing population decline.
The purpose of this paper is to offer an examination of the US local government growth
experience with a focus on k-12 education finances over the 1972-2012 period, where I test for
potential asymmetries in how education spending is influenced by changes in population and
school age population in counties where population is shrinking, stable, and growing. As a
prelude to the full analysis, I find significant asymmetries in how education spending responds to
changes in the proportion of school age children in the population, while controlling for a range
of economic, demographic and institutional factors. Education spending in growing places is
much more responsive to changes in school age population than in shrinking places. That is,
spending tends to increase rapidly with growth in school age population, but is unresponsive to
3
decreases in school age population. The evaluation offers insight for both urban core and rural
places experiencing long-term chronic decline, where local leaders must make difficult choices
in maintaining quality educational services affordably.
The next section offers a review of the most relevant literature on the growth of
government with a focus on education finance. Section 3 discusses the data and empirical
approach used in this evaluation. Section 4 presents the empirical analysis and findings, and
section five concludes.
2. Literature Review
In this section, I offer a review of research on the growth of government, emphasizing the
experience of local governments in the United States. I also discuss several of the most relevant
articles from the education finance literature. I conclude the section by offering a summary of
two primary explanations for why we might observe asymmetric responses to population change
in growing and shrinking places: 1) wages and employment tend to be unresponsive to the
downward pressures associated with population decline; and 2) upward pressure by bureaucrats
to increase spending during periods of growth and resistance to budget reductions during periods
of decline. Consider first the literature on local government growth.
2.1 Growth in Local Government
Economists often frame the demand for government services in the context of the median
voter model. Starting with Bowen (1943) and Black (1958), economists asserted that a
community’s choice of public services under majority rule depends on the median of the
individual demands: Under restrictive conditions, majority rule generates a political equilibrium
that reflects the preferences of the median voter. This general framework was used by
4
Borcherding and Deacon (1972), Bergstrom and Goodman (1973), and many others to
demonstrate that a jurisdiction’s demand for public services depends upon the income of the
median voter, the median (tax) price of the public good, and the preferences of the median voter,
as well as other variables that capture the demand side of the political process. A wide range of
empirical research has usefully applied the median voter model to examine government spending
levels and priorities. Changing community economic and demographic forces ought to play a
primary role in changing government spending levels and priorities.
The present work follows this general line of thinking by considering a number of socioeconomic
and demographic variables in an effort to explain education revenue/expenditure
growth, including median household income, household income in the top 10th percentile,
poverty rate, the proportion of adults with a BA degree, county population, the share of county
households with a single female head, the share of county population over the age of 65 and
under 18, and the share of county population that is white/Caucasian. Rising median incomes as
well as the rising incomes of the top 10% of income earners and higher levels of education may
lead to greater demand for educational services, and vice versa. Increasing single female-headed
households are expected to reduce education spending. Population change, as well as the share of
the population under the age of 18, is expected to be positively related to education spending,
whereas the share of the population over the age of 65 is expected to be negatively related to
education spending. I have no a priori expectation regarding the how the share of the population
that is Caucasian is related to spending.
Brennan and Buchanan (1980) offer another framework for thinking about growth of
government that is worthy of consideration. According to Brennan and Buchanan (2012),
government may have “leviathan” powers, and thus citizens call for legal constraints to limit
5
government power to tax and issue debt.1 Since the 1970s, legislative and referenda processes
have been used extensively across the states to enact new limitations on local governments’
ability to tax and spend.2 It is, therefore, important to include explanatory variables that
characterize the imposition of newly imposed constraints on local government spending.
However, as noted by Blankenau and Skidmore (2002), the imposition of tax and expenditure
limits (TEL) often coincides with school finance reform (SFR), which significantly reduced local
control over education spending and increased reliance on intergovernmental transfers. In fact, a
number of new TELs on schools were imposed with the specific purpose of reducing local
control over education taxes and spending. Taking these developments into consideration, I
incorporate information on TELs as well changes in school finance that occurred during the
period of analysis. Public sector employees may also seek to increase bargaining power over
citizens and thus create “leviathan” powers through the support of strong public sector unions.
To counteract such pressures, a number of states have enacted “Right to Work” (RTW) laws,
which weaken the negotiating power of public sector unions; state and local government
employees are not required to pay union dues in RTW states (Reed 2003). As discussed in more
detail in the next section, I control for these three institutional features as well as changes in the
number of school districts when analyzing the growth of k-12 revenues and spending. While this
body of research informs the types of variable that help to explain government growth, it does
not offer context for assessing the asymmetry issue, which is the focus of the present paper.
Of particular interest is the idea that the responsiveness of local government spending
may differ in shrinking places relative to growing places. Berry, et al. (2012) have documented
the tendency for local governments to grow, even when population is in decline. Further, there
1 See Mueller, chapter 21 (2003) and Oates (1989) for more detailed discussions.
2 See Skidmore (1999) for a review of the literature on TELs.
6
are numerous cases across the country in which this tendency has resulted in dire fiscal
conditions. The goal of the present research is to improve understanding of this asymmetry: Why
is it that shrinking places often fail to correspondingly reduce the size and scope of government?
One possible explanation is proposed by Niskanen (1975): Bureaucrats seek to maximize their
own personal benefits by seeking ever-larger budgets. In this context, bureaucrats may place
upward pressure to increase spending during periods of growth, and to resist budget reductions
during periods of decline. Baumal’s “cost disease” (1993) may also be a contributing factor in
driving the costs of education services higher, even in shrinking places.
The present research expands our understanding of this phenomenon by: 1) Considering a
wide array of socioeconomic factors within the long-term 1972-2012 timeframe, with a focus on
changes in population and school age population, 2) examining the growth of five education
revenues and expenditure categories, and 3) using a flexible empirical specification that allows
coefficient estimates on total population and school age population to differ across shrinking,
stable, and growing counties. Before turning to the data and empirical analysis, it important to
consider the several elements of the more specific literature on education finance.
2.2 Education Finance
The discussion here focuses on two aspects of an expansive education finance literature:
1) Effects of changing demographic factors on education spending, and 2) the effects of
changing institutions such as tax and expenditure limits and school finance reform on education
spending. While it is beyond the scope of this paper to offer a comprehensive review of this large
literature, I discuss a subset of research that is most relevant to the present work.
7
Several articles examine the implications of changing the demographic make-up of
communities on education spending. 3 For example, Harris, et al. (2001) consider the role of the
changing age structure of the population in education spending. Using a panel of public school
districts, they find that an increasing proportion of the elderly have modest negative effects on
local education spending. Epple, Romano, and Sieg (2012), Figlio and Fletcher (2011) also
examined the role of demographic change in school spending. Epple, et al. (2012) focus on
intergenerational conflict, emphasizing the importance of the older generation’s mobility. Figlio
and Fletcher (2011) also consider the role of the growing elderly population in school spending,
finding that increases in the number of the elderly aging in place is associated with reduced
education spending. The majority of studies such as these focus on the impact of an aging
population on education spending, though they consider other changing demographic trends as
well.
Imazeki and Reschovsky (2003) discuss the challenges of financing education in rural
areas, given the small size and often shrinking populations in rural school districts. They estimate
cost functions across rural and non-rural places in Wisconsin and Texas, concluding that, though
the cost structures are similar across rural and non-rural school districts, small district size, high
poverty rate, and a high burden of special needs all lead to higher costs in many rural areas.
Finally, Corcoran and Evans (2010) consider the role of income inequality in the support of
public education, finding that 12% to 22% of the increase in school spending over the 1970-2000
period was attributable to rising income inequality.
There is also a large literature on how changing institutions affect education spending.
First, there is a body of research on how the “tax revolt” and the emergence of new limitations
3 See, for example, Poterba (1998), Harris, Evans, and Schwab (2001), Ladd and Murray (2001), and Grob and
Wolter (2007).
8
on local government tax and spending powers beginning in the 1970s affected education
spending. Much of this literature is summarized in the aforementioned article by Blankenau and
Skidmore (2002), as well as Mullins and Wallin (2004).4 While there is a significant challenge in
identifying causal relationships between the imposition of tax and expenditure limitations (TELs)
and changes in education spending, research generally supports the idea that the imposition of
new TELs on local governments corresponds with reductions in local broad-based taxes
(property taxation) and increased reliance on state aid, as well as other types of revenue such as
user charges.
Beginning in the 1970s, the majority of states experienced legal challenges to their school
finance systems on the basis that inequities in funding violated state constitutions. Beginning
with a major ruling in California (Serranno v. Priest, 1971 and 1976), a series of court rulings
across the nation regarding equity in school finance led to significant changes in school funding.
The primary goal of the rulings was to reduce disparities in funding per pupil across school
districts. Generally, existing research concludes that school finance reforms (SFR) led to
reductions in reliance on local property taxes, and to increased reliance on state government
resources in funding local schools.5 In addition, researchers such as Evans, Murray, and Schwab
(1998) show that SFR significantly reduced disparities in per pupil spending across school
districts. However, as noted by Yinger (2004) and Hoxby (1998) the nature of reforms and their
impacts differ greatly across the states. Researchers such as Fahy (2008) also examined the role
of education finance reform in determining education spending in particular states. Fahy (2008)
4 Kenyon (2008) offers an excellent discussion of the interrelationships between, and evolution of, property taxes
and school finance.
5 See Yinger (2004) for an excellent summary of the impacts of school finance reforms across the nation.
9
considered the role of state aid in improving equity across schools, finding limited overall
effectiveness.
Within the context of the changing landscape of education finance, a relevant and open
question is the degree to which changes in education spending affected school performance and
longer-term student outcomes. The recent works of Jackson, Johnson, Persico (2016), Hyman
(2017), and Lafortune, Rothstein, Schanzenbach (2018) offer compelling evidence using
exogenous variation in school spending to show significant positive effects on student outcomes.
Increases in school spending appear to result in significant improvements in student outcomes.
With school finance reforms, increases in education spending tended to occur in lower spending
school districts, and was the result of state level redistribution of resources. This discussion may
be of particular relevance to the present paper in that shrinking rural counties have increased
stressors associated with maintaining public service levels such as education—one policy option
that may help to avert negative educational outcomes in declining places is to ensure an adequate
level of school spending from higher levels of government.
With the exceptions of Berry, et al. (2012) and Das and Skidmore (2018), researchers
have not considered potential asymmetry in local government spending across growing and
shrinking places. Nevertheless, there is a rationale for the idea that we ought to observe
asymmetries. My primary hypothesis is that education spending is less responsive to declines in
school-age population than to school-age population growth. When overall and school age
population is growing, both operating and capital spending increase in order to meet increased
demand for educational services. However, when overall and school-age population (and thus the
demand for educational services) is in decline, operational spending, such as labor costs, may
become unresponsive as the number of households and students shrink. Further, capital
10
maintenance costs cannot easily be cut without risking depreciation/neglect. There may also be
other types of inertia that limit cuts to spending in shrinking places. For example, wages are
sticky downward, and eliminating excess labor is often difficult. Baumal’s cost disease
framework suggests that increasing costs on educational service provision, even in shrinking
places. Further, the work of Niskanen (1975) suggests that bureaucrats would push for spending
increases during periods of growth, but resist cuts during periods of decline. Thus, the
responsiveness of education spending to population change in shrinking places is likely to be less
than in growing places. For similar reasons, asymmetry is also expected with changes in the
overall population. For these reasons, I hypothesize that the analysis will demonstrate asymmetry
in responses to changes in population and school-age population in places that are expanding
versus places that are in decline. While the empirical analysis does not explicitly measure the
degree to which the aforementioned factors are driving the asymmetry, it is able to document the
degree to which asymmetry is present, which offers a significant contribution of our
understanding of this phenomena.
3. Data and Empirical Approach
Data on local government education revenues and expenditures come from the United States
Census of Governments. Local school fiscal data from independent school districts are
aggregated to the county level.6 In total, 2,752 counties are included in the analysis. The data are
6 School districts sometimes overlap multiple counties, and multiple school districts are often found within a single
county. Regarding school districts that overlap multiple counties, the Census of Governments makes no attempt to
pro-rate the data based on government boundaries. Instead, data for a school district is assigned to the county where
it is headquartered. Due to the nature of the data collected from The Historical Finance Data Base of Individual
Local Governments, only independent school districts are included in the analysis; these are a type of school district
that operates as an independent entity separate from county, municipality, township, special district, and state
governments. They possess their own taxing authority and provide local government finance data separate from
other government types. Therefore, we are unable to separate dependent school district revenue and expenditures,
especially intergovernmental revenues to schools in counties and municipalities that have direct authority over
11
available in five-year intervals (1972, 1977, 1982, 1987, 1992, 1997, 2002, 2007, and 2012. To
examine asymmetry in the impacts of the explanatory variables on education revenues and
expenditures, I create three indicator variables: The variable Shrink identifies counties with
declining population over the 1972-2012 period (about 24% of counties); the variable Stable
identifies all counties that had between -5% and +5% growth over the period (10% of counties),
and the variable Grow identifies counties with positive population growth greater than 5% over
the period (66% of counties).7 These indicator variables are then interacted with the population
and school age population variables . A limitation of using county level data is that the analysis
is unable to capture intra-county variation in education spending across school districts. Further,
we are not able to capture factors such as the advent of charter schools and school choice on
school spending; researchers such as Buerger and Bifulco (2019) and other have shown the
introduction of charter schools and school choice to have meaningful effects on both student
composition and costs; controlling or county trends in the first-difference specification should
help to address potential omitted factors. An advantage of the county level data, however, is that
the examination is nationwide, and is conducted over a long period of time. Further, we are able
to include a wide range of explanatory variables not available if one were to use school district
level data. Last, although we could potentially define growing, stable, and shrinking over shorter
periods, my primary objective is to examine the long-term responses to growth vs. decline. There
are trade-offs in decisions to use certain types of data and periods of analyses; despite the
inherent limitations of county level data, the analysis offers new insights into the dynamics of
school spending across space and over time.
school districts. In total 291 counties with dependent school districts, many of which are in Virginia, Tennessee, and
North Carolina are not included in the analysis, nor are dependent school districts that can be found within counties
that predominantly have independent school districts.
7 Eleven counties were omitted because their data were missing for some of the years included in the analysis.
12
The logarithmic specifications are based on the following equation:
Δ𝑅𝑒𝑣𝑖𝑡𝑗 = 𝑆ℎ𝑟𝑖𝑛𝑘 ∗ Δ𝑃𝑜𝑝𝑖𝑡𝛼1 + 𝑆𝑡𝑎𝑏𝑙𝑒 ∗ Δ𝑃𝑜𝑝𝑖𝑡𝛼2 + 𝐺𝑟𝑜𝑤 ∗ Δ𝑃𝑜𝑝𝑖𝑡𝛼3 + 𝑆ℎ𝑟𝑖𝑛𝑘 ∗
Δ𝑆𝑐ℎ𝑜𝑜𝑙𝐴𝑔𝑒𝑖𝑡𝛼4 + 𝑆𝑡𝑎𝑏𝑙𝑒 ∗ Δ𝑆𝑐ℎ𝑜𝑜𝑙𝐴𝑔𝑒𝑖𝑡𝛼5 + 𝐺𝑟𝑜𝑤 ∗ Δ𝑆𝑐ℎ𝑜𝑜𝑙𝐴𝑔𝑒𝑖𝑡𝛼5 +
Δ𝐷𝑒𝑚𝑖𝑡𝛽1 + Δ𝐸𝑐𝑜𝑛𝑖𝑡𝛽2 + Δ𝐼𝑛𝑠𝑡𝑖𝑡𝛽3 + 𝑐𝑖 + 𝑡𝑡 + 𝑒𝑖𝑡 (1)
where ΔRev represents the change in the natural logarithm of the education revenue (or
expenditure) for county i between periods t and t-1 for revenue (expenditure) category j, ΔPop
represents changes in the natural logarithm of population, ΔSchoolAge represents changes in the
proportion of school age population, ΔDem represents a vector of other demographic variables
that include the percentage of households headed by a single female, the percentage of the
population over the age of 65, and percentage of the population that is white, ΔEcon represents a
vector of economic variables that include the change in natural logarithm of median household
income, the change in the natural logarithm of the income of the top 10% of households, and the
change in the poverty rate, and ΔInst is a vector of institutional variables which includes
variables that indicate change in RTW status, the change in the number of tax and expenditure
limitations (TEL), the change in number of school finance reform efforts (SFR), and the change
in the number of school districts.8 t is vector of time indicator variables, and c represents a vector
of county fixed effects, which accounts for unobserved community trends that have effects on
education spending. This is a first-difference specification that controls for county-specific
trends with county fixed effects as well as overall national trends with time indicator variables.
Data sources and definitions are provided in Appendix Table A. Summary statistics for all
explanatory variables are presented in Tables 1 and 2 for declining and growing counties
8 Caution is warranted in the interpretation of the coefficients on the TEL, SFR, and RTW variables because policy
changes such as these are potentially endogenously determined. Unfortunately, identifying valid instruments within
a panel data framework is challenging. TEL, SFR, and RTW are included in the analysis primarily as control
variables because previous research demostrates their importance in determining the education spending growth.
13
respectively. Tables 3 and 4 present summary statistics of the education finance variables for
shrinking and growing counties.
Note that, because this is a first-difference estimation, the coefficient estimates are
formed by the within-county variation in the independent variables. Thus, it is the within-county
changes in the independent variables upon which the coefficients are generated. In the case of the
institutional variables, the coefficients are being estimated by the changes in the status of the
institutions; over this long period of time we have many changes in RTW, TEL, SFR, and the
number of school districts across the states. It should also be recognized that the nature of TELs
and SFR differ considerably from state to state. Amiel, et al. (2009) and Mullins and Wallin
(2004) catalog TELs and the major characteristics that define them for all states over time. The
approach I use is to identify every new TEL that is imposed on schools in every state. While I
identify every change in the status to TELs over time, this measure does not capture the various
TEL characteristics, and thus measures the average effect of TELs on school revenue and
spending growth. I also include the variable “State TEL”, which again measures every new TEL
on state government that is imposed in each state. To clarify, two TEL variables are included in
the analysis: State TEL and School TEL. Similarly, the SFR variable includes every courtordered
and legislative change in SFR status, but it does not capture the important differences
across states in SFR characteristics as cataloged by researchers such as Yinger (2004) and Hoxby
(2001). This variable also measures the average effect of SFR across the states and over time.
Note that these variables are primarily used as controls, though the estimates may reveal useful
interesting coefficient estimates.
To assess the differences in the effects of the population and school age population
variables on the dependent variables, I interact each variable with the Shrink, Stable, and Grow
14
indicator variables. More specifically, Shrink is an indicator equal to 1 if the county experienced
population decline of more than -5% over the period of analysis and zero otherwise. Stable is an
indicator variable equal to 1 if population change was between -5% and +5% over the period,
and zero otherwise. Grow is an indicator equal to 1 if the county experienced positive population
growth of more than 5% over the period of analysis, and zero otherwise. With this framework,
one can determine whether the coefficients for population and school age population differ
across shrinking, stable, and growing counties. The regression estimates use a technique in which
the standard errors are clustered at the county level to address both temporal autocorrelation and
cross-sectional correlation.9 Education expenditure/revenue categories included in j are: Total
education expenditure/revenue from all overlying jurisdictions (Table 5, column 1), own source
revenues (Table 5, column 2), intergovernmental transfers from state and federal governments
(Table 5, columns 3), expenditure on current operations (Table 5, column 4), and expenditures
on capital outlays (Table 5, column 5). These regressions enable one to see how the changing
population and school age population education finances differ across shrinking, stable, and
growing counties, while controlling for a wide range of economic, demographic, and institutional
factors.
Before turning to the econometric analysis, consider Figures 1 and 2, which illustrate
trends over time in per-capita local government revenue, own-source revenue, intergovernmental
transfers, median household income, population, and school age population. From the graphs it is
clear that median household income grew more slowly across both growing and shrinking
counties than did education revenues/expenditures. In 2002 median household income began to
fall in both growing and shrinking counties. Growth in education spending slowed greatly
9 The Stata procedure for panel corrected standard errors is used.
15
between 2007 and 2012 across the nation. Finally, population declined in shrinking counties, but
grew elsewhere. Figure 3 offers a map of population change in shrinking, stable, and growing
counties. As one might expect, many of the shrinking counties are found in the rural mid-section
of the country, whereas the growing counties are in the south and along the coasts. However,
with the exception of California, Florida, Utah and a few of the small east coast states, shrinking
counties exist in every state across the nation. Note that most shrinking counties experienced
growth in education revenues and expenditures despite experiencing reductions in population
and school age population, and only modest growth in median income over the period. This
descriptive summary information provides context for understanding the estimates generated
from the regression analysis, which is discussed next.
4. Empirical Analysis
Before considering the regression results, some caution is warranted in assigning causality to the
coefficient estimates due to potential endogeniety of the regressors. Changes in school spending
could very well lead to changes in population and school age population, or the imposition of
new fiscal rules. For reference, in specifications not presented but available upon request, I
estimated regressions similar to those presented except population and school age population
were introduced as lagged terms. These estimates are similar to those presented in the paper. In
sum,the evaluation offers a useful and informative evaluation of important trends across
shrinking, stable, and growing counties. Consider the estimates presented in Table 5, which
include regressions for total education revenues/expenditures (column 1), intergovernmental
revenue (column 2), own source revenue (column 3), operating expenditures (column 4), and
capital expenditure (column 5). The regressions explain between 4% and 25% of the variation in
16
the regressions. Note that a low adjusted R2 is not uncommon in this type of regression model.
The data are first differenced and then estimated using the fixed effects technique so that average
growth (decline) in each county is captured with the county fixed effects. The variables in the
regressions capture the remaining variation in growth (decline), and thus the low adjusted R2. An
advantage of this approach is that it offers very robust coefficient estimates that are unlikely to
be biased by omitted factors or spurious correlations.
In the total education revenue/expenditure regression, the coefficient on population is
similar for shrinking, stable, and growing counties, and this is generally true in the
intergovernmental revenue, own-source revenue, operating expenditures, and capital expenditure
regressions. This finding suggests that revenues and expenditures are not responding
differentially to changes in population across shrinking, stable, and growing counties. However,
we observe significant differentials across shrinking, stable, and growing places in response to
changes in school age population. Total revenue/expenditure is very responsive to changes in
school population within stable and growing places, but is unresponsive in shrinking places. The
drivers of this result appear to be own-source revenues and capital spending. In these regressions,
own-source revenue and capital spending appear to increase when school age population
declines. The negative coefficient on school age population for shrinking counties in the ownsource
regression is counter balanced by the positive coefficient (though not statistically
significant) on the same variable in the intergovernmental revenue regression. These findings
confirm the hypothesis that there is significant asymmetry in responses to changes in school age
population in shrinking vs. stable and growing counties. The findings regarding the school-aged
proportion of the population are consistent with the a priori expectations.
17
Consider coefficients on the control variables. In general, increases in median income
correspond with increases in education spending. However, this result appears to be driven by
associated increases in intergovernmental transfers. Growth in the income of the top 10% also
drives spending increases, but here it seems to be the result of increased own source or local
spending. Controlling for income neither increases poverty nor the percent of adults with at least
a college degree are significant determinants of education spending. However, increases in the
number of female-headed households and the elderly are associated with reductions in spending.
Turning to the institutional variables, the adoption of right-to-work laws is associated with
reduced spending. TELs imposed on state governments reduce intergovernmental transfers to
schools, but lead to increased own-source spending; however, the net effect on education
spending is negative. On the other hand, new TELs imposed on school districts reduce ownsource
spending but are associated with increases in intergovernmental transfers. The net effect
on total education spending is negligible. School finance reforms lead to increases in
intergovernmental transfers and reductions in own-source spending, but the net effect on overall
spending is positive. Again, I stress that researchers such as Hoxby (2001) have shown
significant differential effects depending on the nature of the reforms; our evaluation only offers
an estimate of the average effect. Finally, changes in the number of school districts are positively
associated with changes in spending.
The findings presented in the paper are robust to alternative specifications and estimation
methods. In the reported estimations, the Shrink, Stable, and Grow variables are not time
varying, rather they are based on population change over the entire period of analysis. In
estimations that are available upon request, Shrink, Stable, and Grow are allowed to vary over
Census periods; these estimates are similar to those presented and are therefore not discussed
18
further. I also considered estimates in which there were three separate categories of counties—
shrinking, moderate growth, and high growth. These estimates are consistent with those
presented. In terms of estimation methods, in addition to the first-difference specification—with
the added county indicator variables to control for county specific growth trends—I also
estimated a straight first-difference specification as well as a two-way fixed effects specification
with county-specific time trends. These regressions are generally consistent with the primary
estimates discussed here. For reference, I report the two-way fixed effects regressions with
county-specific time trends in Appendix B. The estimates generated from the other approaches
are available from the author upon request.
5. Implications and Conclusions
This study offers an examination of the growth of K-12 education revenues and spending over
the 1972-2012 period using detailed fiscal data for most counties in the United States. A key
objective of the analyses is to increase understanding of why school spending continues to grow
even in the face of declining population and school age population. Over the period of analysis,
about 10% of US counties experienced population decline of more than -5%, where most
declining counties are rural. The evaluation presented in this paper offers some new insight.
First, education revenue/expenditure is more responsive to changes in school age population in
growing places than in declining places. This finding is consistent with the theoretical
discussion; eagerness in expansion and then resistance in making cuts to labor and capital make
it easier to increase spending during periods of growth, but more difficult to cut spending in the
face of decline.
19
Overall, the analysis offers new information that increases our understanding of the
dynamics of education finance in growing and shrinking counties. Are the asymmetries
identified in this analysis an efficient pattern of education finance dynamics? Should education
spending fall when school age population declines as rapidly as it increases during periods of
growth? While the empirical analysis does not offer clear answers to these questions, it sheds
light on important education spending patterns that have been masked within standard regression
analysis thus far. Considerations such as difficulties in cutting wages and employment to the
downside, eagerness of bureaucracies to expand during periods of growth but resistance to cuts
during periods of decline, and intergovernmental assistance formulae help to explain some of the
observed asymmetries.
Local leaders in declining areas must deal with an ever-present tension. On the one hand,
they may feel compelled to devote public resources in order to ensure the children in their
communities receive high quality education. On the other hand, they must be careful not to overburden
decreasing numbers of households with rising taxes to pay for those services, which
could further hasten the decline. The evaluation shows that, on average, local leaders were
willing to increase spending as the number of school age children declined, presumably in an
effort to maintain the quality of educational services. This tension between maintaining the
quality of educational services and being sensitive to tax burdens is always present in shrinking
communities; creativity and perhaps some intergovernmental assistance are required to
affordably maintain essential public services during periods of chronic decline.
20
Figure 1: Declining Population Variable Trends
Figure 2: Growing Population Variable Trends
21
Figure 3: Percent Change in Real Per-Capita Total Education Revenue from 1972 to 2012
22
Table 1: Declining Population Control Variables
1972 1982 1992 2002 2012
Economic
Median Income 30,740 35,152 39,366 44,638 39,230
(7,211) (6,538) (7,033) (7,383) (7,246)
Top Ten Income 65,558 70,382 81,362 104,265 142,669
(11,824) (9,111) (11,855) (14,458) (18,048)
Poverty Rate 0.167 0.138 0.139 0.127 0.168
(0.096) (0.069) (0.072) (0.063) (0.066)
Pct BA Degree 0.071 0.103 0.122 0.149 0.171
(0.023) (0.029) (0.034) (0.044) (0.054)
Demographic
Population 48,575 46,630 44,365 43,910 42,595
(246,942) (232,161) (225,892) (229,363) (222,006)
Female HH Rate 0.067 0.076 0.088 0.097 0.105
(0.031) (0.039) (0.049) (0.051) (0.056)
Pct Over 65 0.131 0.151 0.170 0.172 0.176
(0.037) (0.038) (0.038) (0.035) (0.035)
Pct Under 18 0.337 0.287 0.267 0.249 0.230
(0.039) (0.032) (0.029) (0.026) (0.027)
Pct White 0.915 0.902 0.891 0.871 0.862
(0.164) (0.166) (0.171) (0.177) (0.179)
Institutional
Right to Work 0.584 0.605 0.613 0.640 0.657
(0.493) (0.489) (0.487) (0.480) (0.475)
State TELs 0 0.147 0.260 0.405 0.461
– (0.355) (0.526) (0.623) (0.737)
School TELs 0.923 1.545 1.914 2.017 1.506
(0.458) (0.707) (0.892) (0.932) (0.766)
School Finance Reform 0.171 0.537 1.327 1.932 2.363
(0.377) (0.641) (1.096) (1.178) (1.368)
School Districts 5.917 5.241 4.877 4.295 4.098
(8.319) (7.609) (7.313) (6.994) (7.624)
Standard deviation in parentheses. Adjusted to 2009 dollars.
23
T able 2: Growing Population Control Variables
1972 1982 1992 2002 2012
Economic
Median Income 33,043 38,094 45,010 50,780 44,277
(8,450) (8,240) (10,964) (11,890) (11,503)
Top Ten Income 68,018 75,057 93,707 120,682 146,496
(12,596) (12,301) (17,832) (22,482) (17,373)
Poverty Rate 0.162 0.122 0.121 0.116 0.168
(0.087) (0.059) (0.063) (0.054) (0.060)
Pct BA Degree 0.084 0.122 0.145 0.175 0.199
(0.045) (0.060) (0.073) (0.085) (0.092)
Demographic
Population 65,186 76,596 87,944 100,363 110,383
(203,921) (231,462) (272,542) (305,225) (328,984)
Female HH Rate 0.078 0.088 0.101 0.108 0.122
(0.025) (0.028) (0.032) (0.034) (0.039)
Pct Over 65 0.107 0.118 0.129 0.131 0.140
(0.035) (0.035) (0.036) (0.034) (0.036)
Pct Under 18 0.339 0.291 0.266 0.252 0.236
(0.038) (0.034) (0.034) (0.032) (0.034)
Pct White 0.889 0.878 0.866 0.841 0.828
(0.146) (0.141) (0.143) (0.148) (0.151)
Institutional
Right to Work 0.517 0.538 0.555 0.580 0.611
(0.500) (0.499) (0.497) (0.494) (0.488)
State TELs 0 0.231 0.411 0.624 0.719
– (0.421) (0.599) (0.669) (0.823)
School TELs 0.834 1.594 1.985 2.225 1.594
(0.583) (1.004) (1.071) (1.213) (0.974)
School Finance Reform 0.064 0.584 1.363 2.024 2.408
(0.245) (0.666) (1.178) (1.230) (1.470)
School Districts 5.683 5.526 5.405 5.170 5.304
(7.378) (7.163) (6.933) (6.675) (6.917)
Standard deviation in parentheses. Adjusted to 2009 dollars.
24
Table 3: Declining Population Local Government Education Spending
1972 1982 1992 2002 2012
Total Revenue 50,358 50,840 65,758 82,495 93,206
(292,916) (262,250) (348,497) (440,048) (524,989)
Own-Source Revenue 29,327 25,674 33,449 37,608 42,584
(199,381) (144,470) (222,724) (265,181) (304,513)
Intergovernmental Revenue 21,031 25,166 32,309 44,887 50,622
(97,818) (120,434) (132,799) (185,455) (233,862)
Current Operations 41,547 42,242 55,848 67,530 74,505
(211,608) (207,846) (274,882) (346,760) (410,416)
Capital Outlays 3,162 2,146 3,926 8,302 7,424
(18,321) (8,127) (18,944) (50,847) (40,030)
Standard deviation in parentheses. Adjusted to 2009 dollars, in thousands.
Table 4: Growing Population Local Government Education Spending
1972 1982 1992 2002 2012
Total Revenue 59,773 74,302 117,966 179,262 206,626
(215,303) (240,112) (383,564) (591,457) (640,240)
Own-Source Revenue 32,122 32,789 52,807 75,628 90,314
(132,313) (95,120) (158,018) (238,219) (271,658)
Intergovernmental Revenue 27,651 41,513 65,159 103,634 116,313
(85,897) (162,563) (259,152) (388,536) (411,449)
Current Operations 49,891 61,065 97,440 144,708 167,260
(182,095) (197,928) (313,193) (466,797) (500,710)
Capital Outlays 5,236 4,632 10,956 22,979 17,252
(16,823) (14,035) (35,174) (81,975) (63,091)
Standard deviation in parentheses. Adjusted to 2009 dollars, in thousands.
25
Table 5: Local Government Education Regressions: First Differenced Variables with Fixed Effects
Total
Revenue
Own-Source
Revenue
Intergovernmental
Revenue
Current
Operations
Capital
Outlays
ln(Population)
Declining Units 0.931*** 1.250*** 0.640*** 0.685*** 2.501***
(0.140) (0.150) (0.231) (0.0870) (0.770)
Stable Units 0.909*** 1.322*** 0.663*** 0.797*** 0.984
(0.167) (0.265) (0.232) (0.125) (0.876)
Growing Units 1.008*** 1.395*** 0.654*** 0.886*** 2.537***
(0.0535) (0.0840) (0.0827) (0.0414) (0.270)
Pct Under 18
Declining Units -0.119 -1.316** 0.879 0.369 -3.969**
(0.324) (0.516) (0.726) (0.287) (1.835)
Stable Units 1.826*** 0.0497 3.661*** 1.448*** 7.885***
(0.430) (0.688) (0.765) (0.371) (2.569)
Growing Units 1.946*** 3.005*** 2.233*** 1.755*** 6.839***
(0.213) (0.430) (0.399) (0.186) (1.173)
Other
ln(Median Income) 0.122** -0.000512 0.250** 0.0353 0.668**
(0.0586) (0.0411) (0.106) (0.0246) (0.279)
ln(Top Ten Income) 0.106*** 0.170*** -0.0120 0.0373* 0.815***
(0.0279) (0.0446) (0.0531) (0.0206) (0.152)
Poverty Rate -0.0718 0.0347 0.194 0.00640 -0.141
(0.147) (0.188) (0.252) (0.0919) (0.757)
Pct BA Degree 0.0771 0.108 0.161 0.0827 0.256
(0.135) (0.115) (0.208) (0.0588) (0.641)
Female HH Rate -0.134*** -0.245** -0.0503* -0.131*** -0.671
(0.0478) (0.115) (0.0288) (0.0229) (0.538)
Pct Over 65 -1.115*** -0.619 -1.152** -1.404*** -0.571
(0.325) (0.437) (0.499) (0.289) (1.555)
Pct White -0.494*** 0.0554 -0.562*** -0.424*** -1.160**
(0.0815) (0.148) (0.146) (0.0683) (0.528)
Right to Work -0.0468*** 0.0393*** -0.109*** -0.0532*** -0.197**
(0.00867) (0.0139) (0.0107) (0.00796) (0.0797)
State TEL’s -0.00704* 0.0177** -0.0301*** -1.16e-05 -0.0756***
(0.00422) (0.00818) (0.00715) (0.00312) (0.0282)
School TEL’s -0.00222 -0.0572*** 0.0419*** 0.00108 0.00210
(0.00194) (0.00365) (0.00371) (0.00142) (0.0133)
School Finance Reform 0.0270*** -0.0105** 0.0635*** 0.00981*** 0.0157
(0.00235) (0.00426) (0.00422) (0.00182) (0.0162)
School District Number 0.0176*** 0.0130*** 0.0170*** 0.0193*** 0.00913
(0.00310) (0.00340) (0.00461) (0.00298) (0.00970)
Constant 0.102*** 0.0126 0.207*** 0.127*** 0.0832*
(0.00897) (0.0147) (0.0157) (0.00734) (0.0474)
Observations 21,976 21,976 21,973 21,975 21,867
R-squared 0.205 0.070 0.153 0.255 0.042
Number of Units 2,752 2,752 2,751 2,752 2,752
Dependent variables in log form. Cluster-robust standard errors in parentheses. Time fixed effects included.
***p<0.01, **p<0.05, *p<0.1
26
References
Amiel, L., Deller, S., and Stallman, J. (2009). “The construction of a tax and expenditure
limitation index for the US.” University of Wisconsin-Madison, Staff Paper Series No. 536.
Baumal, W.J. (1993). Health care, education and the cost disease: A looming crisis for public
choice. Public Choice: 17-28
Berry, C., Grogger, J., and West, M. (2102). The growth of government. University of Chicago
Working Paper.
Biehl, D. (1998). Wagner’s Law: an introduction to and a translation of the last version of
Adolph Wagner’s text of 1911. Public Finance= Finances publiques, 53(1): 102-11.
Bergstrom, T. C., & Goodman, R. P. (1973). Private demands for public goods. The American
Economic Review, 63 (3): 280-296.
Black, D. (1948). On the rationale of group decision making. The Journal of Political Economy,
56 (1), 133-146.
Blankenau, W., and Skidmore, M. (2004). School finance litigation, tax expenditure limitations,
and education spending. Contemporary Economic Policy, 22 (1): 127-143.
Borcherding, T. E., & Deacon, R. T. (1972). The demand for the services of non-federal
governments. The American Economic Review, 64 (5): 891-901.
Bowen, H. R. (1943). The interpretation of voting in the allocation of economic resources. The
Quarterly Journal of Economics, 58 (1): 27-48.
Brennan, G. and James Buchanan, J. (1980). The Power to Tax: Analytical Foundations of Fiscal
Constitution (Cambridge, MA: Cambridge University Press).
Buerger, C. and Bifulco, R. (2019). The effect of charter school districts’ student composition, costs,
and efficiency: The case of New York state. Economics of Education Review, 69: 61-72.
Corcoran, S. and Evans, W. (2010). Income Inequality, the Median Voter, and the Support for
Public Education. National Bureau of Economic Research Working Paper No. 16097.
Das, B. and Skidmore, M. (2018). Asymmetry in municipal government responses in growing
versus shrinking counties with focus on capital spending. Journal of Regional Analysis and
Policy, 48(4): 62-75.
Epple, D., Romano, R. and Sieg, H.. (2012). The intergenerational conflict over the provision of
public education. Journal of Public Economics, 96(3-4): 255-268.
27
Figlio, D. and Fletcher, D.. (2011). Suburbanization, demographic change and the consequences
of school finance. Journal of Public Economics, 96(1-2): 1144-1153.
Fahy, C. (2012). Fiscal Capacity Measurement and Equity in Local Contributions to Schools:
The Effects of Education Finance Reform in Massachusetts. Journal of Education Finance,
37(4): 317-346.
Grob, U. and Wolter, S.C. (2007). Demographic change and public education spending: A conflict
between young and old? Education Economics, 15(3): 277-292.
Hanushek, E., Rivkin, S., and L. Taylor (1996). Aggregation and the estimated effects of school
resources. NBER Working Paper No. 5548, http://www.nber.org/papers/w5548.
Hoxby, C.M. (2001). All school finance equalizations are not created equal. Quarterly Journal of
Economics, 116(4): 1189-1231.
Harris, A.R., Evans, W.N., and Schwab, R.M. (2001). Education spending in an aging America.
Journal of Public Economics, 81(3): 449-472.
Hyman, J. (2017). Does money matter in the long run? Effects of school spending on educational
attainment. American Economic Journal: Economic Policy, 9(4): 256-80.
Imazeki, J., and Reschovsky, A. (2003). Financing adequate education in rural settings. Journal of
Education Finance, 29(Summer): 137-156.
Jackson, K.C., Johnson, R.C., Persico (2016). The effects of school spending on educational and
economic outcomes: Evidence from school finance reforms. The Quarterly Journal of Economics,
131(1): 157-218.
Kenyon, D.A. (2008). The Property Tax-School Funding Dilemma: Policy Focus Report (Cambridge,
MA: Lincoln Institute of Land Policy).
Ladd, H.F and Murray, S. (2001). Intergenerational conflict reconsidered: county demographic
structure and the demand for public education. Economics of Education Review, 20(4): 343-357.
Lafortune, J., Rothstein, J., Schanzenbach, D.W. (2018). School finance reform and the
distribution of student achievement. American Economic Journal: Applied Economics, 10(2): 1-
26.
Mueller, D.C. (2003). Public choice, Vol. III. Cambridge University Press, Cambridge, UK.
Mullins, D.R., & Wallin, B.A (2004). Tax and expenditure limitations: Introduction and
overview. Public Budgeting and Finance, 24 (4): 2-15.
Murray, S.E., Evans, W.N., and Schwab, R.M. (1998). Education-finance reform and the
distribution of education resources. American Economic Review, 88(4): 789-812.
Niskanen, W. (1975). Bureaucrats and politicians. Journal of Law and Economics, 18
(December): 617-43
28
Oates, W. (1989). Searching for Leviathan: a reply and some further reflections. The American
Economic Review: 578-583.
Peacock, A., & Scott, A. (2000). The curious attraction of Wagner’s law. Public Choice, 102(1-
2): 1-17.
Poterba, J.M. (1998). Demographic change, intergenerational linkages, and public education.
American Economic Review Papers and Proceedings of the Hundred and Tenth Annual Meeting
of the American Economic Association, 88(2): 315-320.
Reed, W. R. (2003). How right-to-work laws affect wages. Journal of Labor Research, 24 (4):
713-730.
Skidmore, M. (1999). Tax and expenditure limitations and the fiscal relationships between state
and local governments. Public Choice, 88 (1-2): 77-102
29
Appendix A: Data Variables, Definitions, Sources and Methods
Variable Definition
1 Total Education Revenue Total revenue received by k-12 schools aggregated to the county level.
1 Own-Source Revenue Revenue raised directly by k-12 schools aggregated to the county level.
1 Intergovernmental Revenue
Revenue received by k-12 schools from other governmental units (primarily state and federal
governments) aggregated to the county level.
1 Operational Expenditures Expenditures used by a k-12 schools to operate its normal operations aggregated to the
county level.
1 Capital Expenditures Expenditures used by a governmental unit to acquire or upgrade capital assets aggregated to
the county level.
2 Population Total number of persons inhabiting a county.
2 Median Income Income level that divides the income distribution into two equal groups for a county.
2 Top Ten Income Income level that defines the lower bound of the top ten percent income bracket for a county.
2 Female HH Rate Percentage of households that are female-headed in a county.
2 Poverty Rate Percentage of households with income below the poverty line in a county.
2 Pct Over 65 Percentage of the population aged 65 years or older in a county.
2 Pct Under 18 Percentage of the population aged 18 years or younger in a county.
2 Pct BA Degree Percentage of the population that have earned a bachelor’s degree in a county.
2 Pct White Percentage of the population of the White race in a county.
3 Right to Work
Statute that prohibits union security agreements. This variable equal 1 if a RTW law exists 1
a state, and 0 otherwise
4,5 State TELs
Statutes that restrict the level of growth, or spending of a state governmental unit. This
variable increases by 1 every time a new TEL is imposed, and is reduced by 1 if a TEL is
eliminated.
4,6,7 School TELs
Statutes that restrict the level of growth, or spending of local education governmental units.
This variable increases by 1 every time a new TEL is imposed, and is reduced by 1 if a TEL
is eliminated.
8 School Finance Reform
Judicial or legislative acts that reform school funding rules. This variable increases by 1
every time a new SFR is imposed, and is reduced by 1 if a SFR is eliminated.
2 Independent School Districts The number of independent school districts within each county.
1 United States Census Bureau. “State and Local Government Finance Data” from Census of Government Finances and
Annual Survey of Local Government Finances.
2 Minnesota Population Center. National Historical Geographic Information System (NHGIS): Version 2.0. Minneapolis, MN:
University of Minnesota 2011.
3 United States Department of Labor. “State Right-to-work Laws and Constitutional Amendments in Effect as of January 1,
2009 With Year of Passage”.
4 Significant Features of the Property Tax. http://datatoolkits.lincolninst.edu/subcenters/significant-features-propertytax/
Report_Tax_Limits.aspx. Lincoln Institute of Land Policy and George Washington Institute of Public Policy.
5 National Conference of State Legislatures. Prepared by Bert Wasisanen. “State Tax and Expenditure Limits – 2010.”
6 Advisory Commission on Intergovernmental Relations. “Tax and Expenditure Limits on Local Governments”. Publication
M-195: 1995.
7 Amiel, Lindsay, Deller, S.C., and Stallman, J.I. “The Construction of a Tax and Expenditure Limitation Index for the US.”
University of Wisconsin-Madison, Staff Paper Series No. 536: 2009.
8 Jackson, C. Kirabo, Johnson, R., Persico, C. “The Effect of School Finance Reforms on the Distribution of Spending,
Academic Achievement, and Adult Outcomes.” National Bureau of Eonomic Research Working Paper No. 20118: 2014.
30
Appendix A (continued)
Variable Description and Method
Top Ten Income
Top ten income is defined as a top 10% (or 90th percentile) income level of U.S.
households. As the U.S. Census does not provide the full income distribution at
the local level, we restore an (approximate) income distribution using the
reported number of households in each of 10 income categories. First, the upper
limits of income distribution for each sample periods are estimated using the
historical national-level household income trends. Assuming households are
distributed uniformly within each income category, we get households
distribution function across income levels and using this function we calculate
the top ten percent income by targeting the income level where the area under
the households’ income distribution function above that income level is equal to
0.10*total households in a county.
Right to Work
Right to work statutes are defined as a dummy variable: 1 if a state has enacted a
statute or constitutional amendment, and 0 if the state has not. The dummy
variable applies to all types of local government units within a state.
Tax and Expenditure Limits
TELs are defined as account variables that capture the number of statutory
limitation changes that affect a government unit over the period. The type of
TEL or specific limits are not considered. The starting point in 1972 is 0. School
TELs apply to counties with independent local school districts; counties without
independent local school districts operate through counties, municipalities,
townships, and special districts, therefore the TELs imposed on these
jurisdictions are applied instead. State TELs apply to the state government.
School Finance Reform
The School Finance Reform variable is defined as a count variable that captures
the number of legislative or judicial reforms within a given state. The starting
point in 1972 is 0.
Independent School Districts
The School District variable is the number of independent school districts within
a county, not the number of schools; the reporting methods of these counties and
school districts vary. States, counties and municipalities that operate school
districts as part of their own expenditures, rather than as independent school
districts, are set to 0 because local government finance data does not provide this
information.
31
Appendix B: Local Government Education Regressions – Two Way Fixed Effects with County-specific Time Trends
Total
Revenue
Own-Source
Revenue
Intergovernmental
Revenue
Current
Operations
Capital
Outlays
ln(Population)
Declining Units 0.7675*** 1.1347*** 0.4890* 0.5980*** 1.2311
(0.1813) (0.2016) (0.2732) 0.0980 (0.8762)
Stable Units 0.7415*** 1.1848*** 0.8569** 0.8252*** 0.1353
(0.1826) (0.2690) (0.4101) 0.1811 (0.8305)
Growing Units 0.9765*** 1.4211*** 0.5989*** 0.8954*** 1.6803***
(0.0547) (0.0980) (0.0961) 0.0442 (0.2409)
Pct Under 18
Declining Units 1.0497*** -1.2744** 2.4011*** 1.1323*** 0.2994
(0.3798) (0.5802) (0.7984) 0.3281 (1.9342)
Stable Units 2.9049*** 0.2525 5.8002*** 2.5403*** 8.8420***
(0.4395) (0.7691) (0.8839) 0.3943 (2.6188)
Growing Units 2.8128*** 3.4711*** 3.5745*** 2.5047*** 7.8884***
(0.2471) (0.5268) (0.4623) 0.2071 (1.1503)
Other
ln(Median Income) 0.0867 -0.0369 0.2125** 0.0126 0.4919**
(0.0601) (0.0485) (0.1036) 0.0310 (0.2175)
ln(Top Ten Income) 0.1040*** 0.1767** -0.0530 0.0612** 0.7256***
(0.0310) (0.0557) (0.0567) 0.0250 (0.1502)
Poverty Rate -0.2286 0.0271 -0.1181 -0.0607 -1.0818
(0.1577) (0.2334) (0.2601) 0.1118 (0.6936)
Pct BA Degree 0.0221 0.1208 -0.1345 0.0824 0.0861
(0.1324) (0.1643) (0.2426) 0.0755 (0.6754)
Female HH Rate -0.0775 -0.2172 0.0251 -0.0400 -0.4786
(0.0688) (0.1530) (0.0591) 0.0309 (0.7477)
Pct Over 65 -1.0142*** -1.2482** -0.4746 -1.1242*** -2.0351
(0.3407) (0.5483) (0.7001) 0.2867 (1.4828)
Pct White -0.4251*** 0.1190 -0.4260** -0.4634*** -1.0874**
(0.0990) (0.1654) (0.1904) 0.0815 (0.4790)
Right to Work -0.0105 0.1605*** -0.1003*** -0.0200** -0.3007***
(0.0116) (0.0193) (0.0160) 0.0097 (0.0769)
State TELs 0.0060 0.0157 -0.0288*** 0.0063 -0.0834***
(0.0053) (0.0117) (0.0101) 0.0045 (0.0269)
School TELs -0.0093*** -0.0614*** 0.0290*** -0.0038* -0.0297**
(0.0025) (0.0052) (0.0052) 0.0020 (0.0135)
School Finance Reform 0.0235*** -0.0336*** 0.0656*** 0.0169 -0.0176
(0.0026) (0.0047) (0.0046) 0.0022*** (0.0153)
School District Number 0.0212*** 0.0094** 0.0268*** 0.0218 0.0216**
(0.0032) (0.0041) (0.0048) 0.0031 (0.0091)
Constant -9.681*** -61.021*** 16.271*** -20.263*** 42.478***
(1.662) (3.137) (2.891) (1.367) (7.309)
Observations 24,733 24,733 24,731 24,732 24,659
R-squared 0.993 0.981 0.985 0.996 0.834
Number of Units 2,754 2,754 2,754 2,754 2,753
Dependent variables in log form. Cluster-robust standard errors in parentheses. Time fixed effects included.
***p<0.01, **p<0.05, *p<0.1

Leave a Reply

Your email address will not be published. Required fields are marked *