LIFTING ALL BOATS? FINANCE

LITIGATION, EDUCATION

RESOURCES, AND STUDENT

NEEDS IN THE POST-ROSE ERA

David P. Sims

Economics Department

Brigham Young University

Provo, UT 84602

davesims@byu.edu

Abstrait
Rose v. Council for Better Education (1989) is often con-
sidered a transition point in education ﬁnance litigation,
heralding an era of increasing concern for measur-
able adequacy of education across a broad spectrum of
student needs. Prior research suggests that post-Rose
lawsuits had less effect on the distribution of school
spending than older litigation. This article suggests that
this focus on the raw resource distribution masks the
important effect of contemporary lawsuits in redistribut-
ing money to districts with greater student needs. My
ﬁndings suggest that a successful lawsuit does raise rev-
enues to a variety of districts but provides more money
to those districts with higher plausible indications of
student needs.

c(cid:2) 2011 Association pour le financement et la politique de l'éducation

455

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

FINANCE LITIGATION AND STUDENT NEEDS

INTRODUCTION

1.
In its 1819 decision in Commonwealth v. Dedham the Supreme Judicial Court
of Massachusetts ﬁrst articulated a principle that has been a key to the modern
development of education ﬁnance, namely, “It is the wise policy of the law to
give all the inhabitants equal privileges, for the education of their children in
the public schools.”1 Although the nineteenth-century development of public
education in the United States led to a mostly local system of education ﬁnance,
the persistent idea that federal and state governments have a responsibility to
ensure some level of public education for all students led to extensive court
interventions from 1971 to the present that have effectively overridden more
than a century of local ﬁnance precedents in many states.

This series of lawsuits, spanning more than thirty-ﬁve years since the
landmark Serrano v. Priest case in California (1971), articulated arguments for
increased school funding based on state constitutional provisions that require
equity in school funding among districts as well as adequate funding for all
students. Although equity and adequacy claims often coexist within the same
court case, education ﬁnance researchers commonly cite Kentucky’s 1989
Supreme Court ruling in Rose v. Council for Better Education (1989) as an impor-
tant landmark in the development of legal claims based on adequacy grounds
and the beginning of the contemporary era of education ﬁnance litigation.2

The importance of these lawsuits has led to a large education ﬁnance litera-
ture that considers the average effect of successful ﬁnance litigation on school
ressources. A common ﬁnding of this literature is that more recent plaintiff vic-
tories have failed to compress the measured distribution of district resources in
a manner similar to earlier lawsuits of the late 1970s and early 1980s. Plutôt
this litigation has resulted in broad funding increases across most districts
(Berry 2007; Corcoran and Evans 2007; Springer, Liu, and Guthrie 2009).
Cependant, in this focus on the shape of the funding distribution, prior research
has largely downplayed the relationship between lawsuit-induced funding in-
creases and observable indicators of student need. Unlike an equity goal,
educational adequacy cannot be deﬁned simply by reference to the funding
level of other districts, so it is unclear why we should expect lawsuits based on
adequacy claims to compress the funding distribution.

This article extends the existing literature on the effect of courtroom vic-
tories in the post-Rose era (1989–2002) on the distribution of district-level
education resources by showing how the increases in district resources from
ﬁnance litigation correspond to measurable indicators of student needs. Dans

The description of the Dedham case comes from the National Access Network Web site,
www.schoolfunding.info/states/ma/lit ma.php3.
Par exemple, Corcoran et al. (2004).

456

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

autres mots, it considers whether a rising tide of lawsuit-induced education
spending lifts all boats or selectively targets high-need districts.

I ﬁnd that a disproportionate share of lawsuit-induced resource gains ac-
crues to the districts with the highest fraction of free lunch and special educa-
tion students, though the targeting of resources to the highest-need districts is
imperfect. I also attempt to reconcile this ﬁnding of selectivity in the funding
distribution with evidence of similar proportional gains across the resource
distribution. Due to the complex nature of ﬁnance lawsuits, the lessons of this
era may prove a better guide to the likely consequences of contemporary legal
action than decisions from the 1970s.

The remainder of the article is organized as follows: section 2 brieﬂy exam-
ines the most important institutional details and the data, section 3 examines
the average resource effects of court decisions in my time frame, section 4 con-
siders the effects of lawsuits on the distribution of resources among districts,
and section 5 concludes.

2. BACKGROUND AND DATA
Rationale and Research
Beginning with Serrano, over one-third of U.S. states have seen their courts rule
that their education ﬁnance systems violate state constitutional obligations.3
Several other states spent time in court and ultimately prevailed. The main
legal theory underlying early state court ﬁnance cases was based on duties
of the state to provide its citizens with an equitable public education. Comme
the prevailing system of local property tax funding for public schools led
to a structural funding disadvantage for lower property wealth districts, le
most common remedy in school funding cases involved state intervention to
break the link between property wealth and school spending through some
combination of supplementation and redistribution.

Given the wide availability of data on school resources and expenditures
and the measurable nature of claims about funding disparities, it seems in-
evitable that a literature would emerge to examine the effects of school ﬁnance
lawsuits on the level and equitable distribution of school resources. Études
by Evans, Murray, and Schwab (1997) and Murray, Evans, and Schwab (1998)
use district-level panel data from most states in ﬁve-year intervals from 1972
à 1992 to evaluate the average effects of successful ﬁnance litigation during
this time period. Using various index measures of inequality, they conclude
that court-ordered ﬁnance reforms reduced within-state inequality by 19–34

3. Although the initial Serrano decision was primarily based on the equal protection clause of the U.S.
Constitution, the Supreme Court ruling in San Antonio Independent School District v. Rodriguez (1973)
forestalled this line of argument. Thus the focus of future litigation centered on state constitutions.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

457

FINANCE LITIGATION AND STUDENT NEEDS

pour cent. This is primarily the result of lower-spending school districts increas-
ing spending vis-`a-vis the higher-spending districts. En outre, the authors
contrast the effects of court-ordered ﬁnance reform with legislative reform
absent a court mandate, concluding that the latter is ineffective in increasing
education resources or reducing resource inequality.

While early court-ordered ﬁnance reforms resulted in more education
spending and a narrower distribution of spending among districts, it is less
clear that the increasing equity in ﬁnancial resources translated into more
equal education. Studies of the link between school ﬁnance reform and stu-
dent outcomes fail to provide a consensus with some documenting positive
achievement effects (voir, par exemple., Downes and Figlio 1998; Card and Payne 2002)
and others ﬁnding negative or zero effects (voir, par exemple., Husted and Kenny 1997;
Hoxby 2001). Observers, such as Hanushek (2003), raise questions about the
weak correlation of raw spending and desired student outcomes, incentive
structures that might discourage schools from achieving an efﬁcient conver-
sion of money into student outputs, and failure to account for the differential
needs of students across districts.

The idea that education ﬁnance should recognize need-based differences
across districts and should aim to provide adequate, rather than simply equal,
funding for all students had already been advanced in earlier court cases (par exemple.,
Robinson v. Cahill 1973; Pauley v. Kelly 1979). Cependant, the Rose decision is
commonly referenced as a watershed moment in education ﬁnance litigation
because it placed adequacy considerations into the center of the education ﬁ-
nance debate. In the Rose case the court accepted the plaintiff’s argument that
the state constitution guaranteed all Kentucky students an adequate educa-
tion as determined by the court. Since all state constitutions give the state
some responsibility for education, the ruling set an important precedent.
The adequacy doctrine combined with the expanding standards movement
in education to expand the ground on which ﬁnance lawsuits were contested.
Though lawsuits were still argued under both equity and adequacy grounds
post-Rose, it seems reasonable that such changes in institutional setting might
lead to different litigation effects on the distribution of education resources.
Even in the absence of adequacy principles, the effects of court decisions
on spending distributions might be different in an environment of evolving
awareness of student needs brought on by the movement to increase school
accountability.

En effet, recent research from Corcoran et al. (2004) and Corcoran and
Evans (2007) suggests that extending the time horizon of the earlier national
studies to include all years from 1972 à 2002 causes the estimates of court
ﬁnance reform’s effect on spending equity to attenuate considerably, cependant
a positive effect on average spending remains. When only the 1989–2004

458

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

period is considered, Corcoran and Evans (2007) ﬁnd that the effect of court
rulings on various inequality indices are not signiﬁcantly different from zero.
Berry (2007) and Springer, Liu, and Guthrie (2009) attempt to differentiate
the resource effects of separately categorized adequacy and equity lawsuits,
especially in the 1990–2000 period. Using standard inequality indices, le
latter concludes that adequacy reforms were less successful on average at
reducing inequality than were equity suits. This result is borne out by the
ﬁndings of Corcoran and Evans (2007), who move beyond the simple use of
single indices of inequality to consider the effect of recent lawsuits on some
unconditional percentiles of the spending distribution. They conclude that
spending increases due to lawsuits look to have increased spending at the
median, 5th percentile, and 95th percentile by comparable amounts.

Another branch of the literature looks at the effects of litigation-induced ﬁ-
nance reform on speciﬁc state spending distributions in the post-Rose period.4
In contrast with the national studies, these articles often ﬁnd some level of
district ﬁnance equalization associated with litigation success. En effet, Clark
(2003) and Flanagan and Murray (2004) both ﬁnd that the Rose decision itself
led to a substantial increase in state aid for lower-spending districts and a
consequent fall in a number of index measures of resource inequality. Dee
and Levine (2004) and Downes (2004) ﬁnd more modest compression in
spending due to the reforms in Massachusetts and Vermont, respectivement. Dans
general these studies also ﬁnd an increase in average education spending.

These contrasting results are not necessarily difﬁcult to reconcile. En effet,
given the institutional differences involved across state ﬁnance reforms and
the large standard errors presented in the national studies, there may be small
positive average equalization effects of post-Rose lawsuits, or the effects in
these particular states may be unusually large.5

While both the state- and national-level studies have continued to focus on
the equity of resources, the effect of lawsuits on the distribution of spending
across districts with different types of students has received less attention.
While Manwaring and Sheffrin (1997) and Downes and Shah (2006) argue

4. There are also several single-state studies of older reforms, such as the Serrano case in California
(Silva and Sonstelie 1995). While valuable, these are not directly comparable to the present study.
This state-level literature points out that in national level studies, “any attempt to classify ﬁnance
reforms will be imperfect” (Downes 2004, p. 290). It is certainly true that studies such as the present
one can miss important differences in institutional detail that might modify our prediction of the
effects of certain reforms. Cependant, single-state studies have their own attendant problems, most
notably constructing a counterfactual based on non–time series variation and obtaining statistical
precision (districts in the same state may not be independent observations). Given this, both types
of studies have important roles to play, and a synthesis is often informative. This study considers
whether court-ordered reform on average results in funding to address student needs. Studies of
how individual state reforms attempted to meet student needs would clearly address an important
complementary question.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

459

FINANCE LITIGATION AND STUDENT NEEDS

that the nature of state ﬁnance reform depends on the demographic proﬁle
of the state, and Aaronson (1999) and Downes and Figlio (1999) suggest that
ﬁnance reform can change that proﬁle, there is still little evidence on whether
these reforms systematically increase revenues in line with preexisting district
indicators of student need (besides preexisting revenues) beyond the work of
Card and Payne (2002), which I discuss later. This study begins to ﬁll that gap
by using free lunch status, racial composition, and special education status
to measure how district resource gains from ﬁnance litigation match student
needs.

Although court-ordered reform might well induce a distribution of re-
sources that respects student needs, there are reasons to suspect that this
might not occur in practice. Par exemple, while there is some agreement that
an adequate education should be based on standards set by the state, there is
no uniform standard of how to ﬁgure what adequate funding should mean.
Though a series of methods for “costing out” has developed, there is sufﬁ-
cient variation in results to support an enormous range in what might be
considered adequate spending. This has led critics of these methods, tel
as Hanushek (2005), to refer to them as “alchemy.” This uncertainty about
the details of adequate funding might affect the decisions of legislators and
state boards of education. While recent court decisions may provide them
with greater latitude to determine where to send money to meet student
needs, the removal of an equity constraint might instead motivate them to
add funding for their local, possibly high-resource, districts as a condition of
voting for a ﬁnance measure. Alternativement, expected cutbacks in local fund-
ing among high-resource districts in response to increased state aid may not
materialize.

Tableau 1 lists states that had a binding school ﬁnance court decision be-
tween 1989 et 2002 as well as the case name and whether the state
or plaintiff prevailed. In a few states where the court reconvened to issue
additional reinforcing rulings on the initial case, only the initial ruling is
noted. The geographical reach of the ﬁnance litigation is striking, with al-
most two-thirds of U.S. states experiencing a decision in a fourteen-year
time period. Also notable is the apparent lack of a clear regional or geo-
graphic pattern in state versus plaintiff rulings. De plus, the balance between
plaintiff and state victories appears to have been close to even during this
period.

Données
I draw data from a couple of governmental sources. Yearly data on per student
revenues and spending for school districts in the United States for school
years from 1989–90 to 2001–2 come from the Longitudinal School District

460

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

Case

Year

Decision

David P. Sims

Tableau 1.

Initial High Court Rulings on Education Finance during 1989–2002

State

Alabama

Alaska*

Arizona

Arkansas

Alabama Coalition for Equity v. Hunt
Ex Parte James

Kasayulie v. State

Roosevelt Elementary School District No. 66 v. Bishop

Lake View School District N. 25 v. Huckabee

Connecticut

Sheff v. O’Neill

Florida

Illinois

Kansas

Kentucky

Louisiana

Maine

Coalition for Adequacy v. Chiles

Committee for Educational Rights v. Edgar
Lewis v. Spagnolo

Uniﬁed School District No. 299 v. Kansas

Rose v. Council for Better Education

Charlet v. Legislature of State of Louisiana

Maine School District #1 v. Commissioner

Massachusetts

McDuffy v. Secretary of the E.O. of Education

Minnesota

Missouri*

Montana

Nebraska

Skeen v. State

Committee for Educational Equality v. State

Helena v. State

Gould v. Orr

New Hampshire

Claremont v. Governor

New Jersey

Abbot v. Burke

New Mexico*

Zuni School District v. State

North Dakota

Bismarck Public School District No. 1 v. North Dakota

Ohio

Oregon

Pennsylvania

DeRolph v. State

Coalition for Equitable School Funding v. Oregon

Marrero v. Commonwealth
Pennsylvania Assoc. of Rural and Small Schools v. Ridge

Rhode Island

Pawtucket v. Sundlun

South Dakota*

Bezdicheck v. South Dakota

Tennessee

Small Schools v. McWherter

Texas

Vermont

Virginia

Edgewood v. Kirby

Brigham v. State

Scott v. Commonwealth

W. Virginia*

Tomblin v. Gainer

Wisconsin

Kukor v. Grover
Vincent v. Voight

Wyoming

Campbell County School District v. State

1993
1997

1999

1994

2002

1996

1996
1999

1994

1989

1998

1995

1993

1989

1993

1997

1990

1999

1994

1997

1991

1998
1998

1995

1994

1993

1989

1997

1994

1995

1989
2000

1995

Plaintiff
State

Plaintiff

State

State
State

State

Plaintiff

State

Plaintiff

State

Plaintiff

State

Plaintiff

State

Plaintiff

State

State
State

State

Plaintiff

State

Plaintiff

State
State

Plaintiff

*The Alaska, Missouri, New Mexico, and West Virginia decisions were trial court verdicts that were
acted on by legislatures before any high court conﬁrmatory rulings. The South Dakota trial ruling
was not appealed. The table excludes repeat decisions (returns to court) on the same case (par exemple.,
Claremont III–VII in New Hampshire) listing only the ﬁrst in a series of high court decisions on
funding.
Source: Author’s compilation based on his reading of data at the National Access Network Web site
and Corcoran and Evans 2007, appendix table 19A.1.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

461

FINANCE LITIGATION AND STUDENT NEEDS

Fiscal-Nonﬁscal Detail File (LFNF) maintained by the National Center for Ed-
ucation Statistics (NCES) and collected in collaboration with the U.S. Census
Bureau.6 I combine these data with elements of the NCES Common Core of
Data series tracking districts’ racial composition, free lunch eligibility, and spe-
cial education eligibility. The resulting data set is ideal for examining changes
in spending while controlling for certain district characteristics that may also
change over time.

In order to track responses to school ﬁnance decisions I match the above
data with the information from table 1 about the timing of court decisions.
I generate two dummy variables: one if a district’s state has had a plaintiff
verdict within my time frame but before the year in question, and a second
if the state has had a defendant verdict. Any permanent differences in those
states’ ﬁnancing due to court decisions preceding the Rose era are accounted
for by the inclusion of state ﬁxed effects in my models. I further generate
variables that measure the number of years since a verdict and a dummy
variable to track states that have experienced a legislature-induced systemic
ﬁnance reform.7

To represent district resources, I use the natural logarithm of per student
total district revenues as the dependent variable in most of my analysis.8
Prior studies (Murray, Evans, and Schwab 1998; Card and Payne 2002) have
found that school ﬁnance litigation increases both state aid revenues and
overall revenues. While my primary interest is investigating how litigation
increases total resources and their distribution among districts, regardless of
the ultimate source of the funding, I also examine speciﬁcations that consider
revenues from state sources as the dependant variable. Using revenues rather
than current spending also prevents an inconsistent treatment of capital items
from the analysis, since the revenue streams used to pay off capital debt may
more uniformly approximate the use of capital in education than the varying
state cost depreciation methods.

As with all data, there are potential pitfalls in this data set. The NCES
identiﬁes 405 district-year observations in the data set that are considered to
be ﬁnancial outliers due to uncommonly high reported per pupil resource
ﬁgures that vary signiﬁcantly from prior years. In most cases this is due to
large changes in the number of students in the district and a lag in reappor-
tionment of resources. Par exemple, one district loses almost 90 percent of its

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

Irregular districts, such as those run for institutional populations, are excluded from the data by the
NCES.
For the remainder of the article, school year designations refer to the spring. Thus the 1992 school
year began in fall 1991 and concluded in spring 1992.

8. All revenue data are transformed to constant 1990 dollars using the Consumer Price Index before

taking logarithms.

462

David P. Sims

students but very little of its funding, resulting in per student ﬁnancing of over
$50,000 for a year. I exclude these observations from the analysis as well as observations for which race, free lunch, and special education data are missing. Since I wish to address the distribution of resources among districts within a state, I also exclude the state of Hawaii and the District of Columbia because each represents a single district.9 These exclusion standards are actually more conservative in some respects than those in the previous literature. Another potential pitfall is the presence of sample- as opposed to universe- level data for a few states in three early years of the LFNF record. In the 1991, 1993, et 1994 school years between seven and twelve states provided survey ﬁnancial data on samples of school districts covering roughly between 20 et 50 percent of the universe. Thus the NCES has imputed values for non-sampled districts in those states for those years. Heureusement, most of these states were unaffected by school ﬁnance litigation in these years. En outre, some basic testing has shown that excluding the non-sampled district-year observations has little effect on the estimation results, so I leave them in the sample. Descriptive statistics for the data are provided in table 2. Approximately 22 percent of the districts in my sample have experienced a plaintiff ruling in their state, while only 14.5 percent have experienced rulings that favor the state. Also notable is the small fraction of minority students: 6.9 percent Hispanics and 6.4 percent blacks associated with the use of school district as opposed to student-level averages. Nevertheless an average district has a high proportion of economically disadvantaged students, with a quarter eligible to receive free school lunches. Almost an eighth of the average district’s students have an individualized education plan, the contemporary designation for special ed- ucation status. I also use the modal school locale information in the data to generate dummy variables to mark urban and suburban districts. 3. FINANCE LITIGATION AND AVERAGE RESOURCES Investigating the distribution of funding gains due to school ﬁnance litigation across different types of school districts presumes there is some plausibly causal positive link between court decisions and average school resources. This section demonstrates that court decisions of this era are positively correlated with average district resources and that the variation they induce is plausibly exogenous. 9. Il y a 183,962 observations in the LFNF ﬁle for which there is at least one school and ten students. Of those I lose 405 observations to outliers, 26 from excluding Washington, CC, and Hawaii, 582 observations with missing race information, 5,153 observations with missing free lunch information, et 655 observations with missing special education information. I also exclude 10 observations that report more minority or special education students than total students. This leaves me with N = 177,131 observations. These exclusions total about 3.7 percent of the original sample. l Téléchargé à partir du site Web : / / direct . m je t . F / / e du e d p a r t i c e – pdlf / / / / / 6 4 4 5 5 1 6 8 9 2 9 4 e d p _ a _ 0 0 0 4 4 pd . f f par invité 0 7 Septembre 2 0 2 3 463 FINANCE LITIGATION AND STUDENT NEEDS Table 2. Descriptive Statistics Variable Name Log per student revenues Has a court ruling favoring plaintiff (1989–2002) Has a court ruling favoring state (1989–2002) Has a legislature-initiated reform (1989–2002) Students Fraction Asian Fraction black Fraction Hispanic Fraction Amerindian Fraction free lunch eligible Fraction special education Number of schools Uniﬁed school district Elementary/middle school district Urban district Suburban district District 1989 median household income Mean (s.d.) 8.600 (0.332) 0.220 (0.414) 0.145 (0.352) 0.073 (0.262) 3,117.17 (13,851.18) 0.014 (0.035) 0.064 (0.153) 0.069 (0.157) 0.025 (0.105) 0.258 (0.187) 0.118 (0.055) 6.05 (16.82) 0.762 (0.426) 0.165 (0.166) 0.061 (0.239) 0.218 (0.413) 28,961.00 (11,361.52) Remarques: Revenues are based on constant 1990 dollars. N = 177,131 except for 1989 median household income, which is calculated from the 150,526 districts with relevant data. A commonly posited statistical model of the relationship between educa- tion resources and ﬁnance lawsuits is: Ri j t = α j + τt + δ Pj t + X (cid:2) i j t γ + εi j t , (1) where i indexes district, j state, and t year. R is the natural logarithm of per pupil revenues, P is a variable that represents lawsuit decisions, and X is some 464 l Téléchargé à partir du site Web : / / direct . m je t . F / / e du e d p a r t i c e – pdlf / / / / / 6 4 4 5 5 1 6 8 9 2 9 4 e d p _ a _ 0 0 0 4 4 pd . f f par invité 0 7 Septembre 2 0 2 3 David P. Sims vector of district-level controls. Also included are state and year ﬁxed effects to control for permanent state-level differences and common disturbances across states, as well as an error term.10 Table 3 presents estimates of this average relationship between lawsuit out- comes and resources. As with all subsequent analysis, the reported standard errors are corrected for clustering at the state level.11 In addition, all regres- sions are weighted by the number of students enrolled in a district. My base parameterization of P is a single dummy variable for all district years in a state after a successful plaintiff lawsuit. Column 1 of table 3, using this parameterization and controlling only for state and year ﬁxed effects, suggests that a plaintiff victory raises subsequent district resources by an average of 6.2 percent relative to all other states. Given an average per student expense during this period of about $5,430,
this represents a roughly $335 per pupil increase in revenues. Column 2 adds
controls for district characteristics such as number of students, student race,
special education, and free lunch status and ﬁnds roughly the same coefﬁcient
of interest. Of note, districts with high fractions of minority or special education
students are associated with higher resource levels, ceteris paribus, while those
with high free lunch percentages see less funding.

Column 3 adds controls for other district characteristics such as grade levels
included, number of schools, and whether the district is urban or suburban as
opposed to rural. Here the measured relationship between plaintiff victories
and district resources remains the same. In all three cases the measured effect
is signiﬁcantly different from zero. A comparison of R2 measures across the
three regressions suggests that the addition of controls does not add much
explanatory power to the model.

Do the reported coefﬁcient estimates plausibly have a causal interpreta-
tion? The key identifying assumption of this model is the absence of omitted
variables that affect both P and R. If some other, unmeasured factor leads to
both ﬁnance lawsuits and resource increases in a particular state, the resulting
omitted variable leads to an upward bias in the estimated effect of lawsuit vic-
tories on resources. Par exemple, perhaps lawsuits reﬂect a public desire for
more education spending that will come to pass regardless of the court’s de-
cision. While the assumption that court decisions may be treated as plausibly
exogenous variation in resources is ubiquitous in the literature and is sup-
ported by studies stressing the unpredictable nature of court ﬁnance rulings

10. Similar models form the basis of much of the national-level literature on court-induced reform,
including Murray, Evans, and Schwab 1998, Corcoran et al. 2004, and Corcoran and Evans 2007.
11. This follows the recommendation of Bertrand, Duﬂo, and Mullainathan (2004) for panel data

situations with a high probability of serial correlation.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

465

FINANCE LITIGATION AND STUDENT NEEDS

)
9
(

0
5
0
.
0

)
7
1
0
.
0
(

)
8
(

0
6
0
.
0

)
9
1
0
.
0
(

)
7
(

0
6
0
0

)

9
1
0

(

1
8
2
.
0

)
4
5
0
.
0
(

4
4
2
.
0

)
0
6
0
.
0
(

)
1
5
0
.
0
(

1
0
2
.
0
−

5
1
2
.
0

)
1
4
0
.
0
(

7
6
1
.
0

)
5
3
0
.
0
(

)
9
5
0
.
0
(

9
9
0
.
0
−

1
3
0
0

)
6
3
0

(

4
6
2

)

5
5
0

(

8
1
2

)

9
5
0

(

)

0
5
0

(

8
6
1

0
−

)
6
(

2
5
0

)

9
1
0
0

(

4
3
0

0
−

)

3
2
0

(

5
6
2

)

5
5
0

(

7
1
2
0

)

9
5
0

(

7
6
1

0
−

)

1
5
0
0

(

)
5
(

3
5
0

)

8
1
0

(

1
0
0
0

)

3
0
0

(

)
4
(

3
6
0
0

)

9
1
0

(

)
3
(

8
5
0

)

9
1
0

(

)
2
(

8
5
0
.
0

)
9
1
0
.
0
(

)
1
(

2
6
0
.
0

)
9
1
0
.
0
(

n
w

F
F
je
t
n
un
P.

e
c
n
s

s
r
un
e
Oui

n
w
e
t
un
t
S

4
6
2

)

5
5
0

(

7
1
2

)

8
5
0

(

)

0
5
0

(

6
6
1

0
−

5
6
2
0

)

8
5
0

(

0
2
2

)

0
6
0

(

)

2
5
0

(

5
7
1

0
−

3
3
0

)

4
1
0

(

5
3
0
0

)

3
1
0

(

6
2
2

)

6
4
0
0

(

6
8
1

)

1
5
0
0

(

)

9
3
0

(

7
2
1
0
−

4
6
2
.
0

)
5
5
0
.
0
(

7
1
2
.
0

)
9
5
0
.
0
(

)
0
5
0
.
0
(

7
6
1
.
0
−

t
c
je
r
t
s
d

n
un
b
r
toi
b
toi
S

t
c

je
r
t
s
d

n
un
b
r
U

k
c
un
B

c
n
un
p
s
H

h
c
n
toi

e
e
r
F

m
r
o
F
e
r

e
v
je
t
un
s
g
e
L

s
e
toi
n
e
v
e
R.

je
je

p
toi
P.
r
e
p

un
t
o
T

F
o
m
h
t
je
r
un
g
o
L

un
r
toi
t
un
N

:
e
b
un
je
r
un
V

t
n
e
d
n
e
p
e
D

s
e
c
r
toi
o
s
e
R.
t
c
je
r
t
s
D
e
g
un
r
e
v
UN

n
o

oui
r
o
t
c
je
V

t
je
toi
s
w
un
L

F
o

t
c
e
F
F
E

3
e
b
un
T

466

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

4
7
5
.
0

)
4
0
1
.
0
(

4
6
5
.
0

)
3
2
1
.
0
(

6
6
4

)
6
1
1

(

4
6
4

)

2
1
1

(

2
5
4

)

1
1
1

(

6
3
4

)

9
1
1
0

(

0
5
4

)

8
1
1

(

5
5
4
.
0

)
2
1
1
.
0
(

s
e
Oui

o
N

s
e
Oui

o
N

s
e
Oui

o
N

s
e
Oui

o
N

4
6
.
0

5
6
.
0

3
6

s
e
Oui

o
N

s
e
Oui

3
6

s
e
Oui

o
N

s
e
Oui

3
6

s
e
Oui

o
N

s
e
Oui

o
N

s
e
Oui

o
N

s
e
Oui

3
6

4
6

s
e
Oui

o
N

s
e
Oui

3
6
.
0

s
e
Oui

o
N

s
e
Oui

0
6
.
0

n
o
je
t
un
c
toi
d
e

un
c
e
p
S

s
t
c
e
F
F
e

d
e
X
ﬁ

e
t
un
t
S

s
t
c
e
F
F
e

d
e
X
ﬁ

r
un
e
Oui

s
d
n
e
r
t

e
m

je
t

e
t
un
t
S

e
z
je
s
/
e
p
oui
t

t
c
je
r
t
s
D

e
m
o
c
n

n
un
d
e
M.

e
p
m
un
s

je
je

toi
F

2
R.

je
je

t
un

n
o
s
je
c
e
d

t
r
toi
o
c

o
n

h
t
je

s
e
t
un
t
s

e
r
e
h
w

,
4

n
m
toi
o
c

t
p
e
c
X
e

1
3
1
,
7
7
1
=
N

.
je
e
v
e

e
t
un
t
s

e
h
t

t
un

g
n
je
r
e
t
s
toi
c

r
o
F

d
e
t
c
e
r
r
o
c

e
r
un

s
e
s
e
h
t
n
e
r
un
p

d
e
t
r
o
p
e
r

s
r
o
r
r
e

d
r
un
d
n
un
t
S

:
s
e
t
o
N

.
s
t
c
je
r
t
s
d

e
m
o
s

s
e
m
o
c
n

n
un
d
e
m

r
o
F

oui
t
je
je
je

b
un

je
je

un
v
un

un
t
un
d

F
o

k
c
un

o
t

e
toi
d

6
2
5
,
0
5
1
=
N
e
r
e
h
w

,
8
n
m
toi
o
c

d
n
un

,
s
t
c
je
r
t
s
d

5
8
2
,
5
6
1

F
o

e
p
m
un
s

g
n
je
v
un
e

,
d
e
t
t
je

m
o

e
r
un
2
0
0
2

h
g
toi
o
r
h
t

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

467

FINANCE LITIGATION AND STUDENT NEEDS

favoring the plaintiff (voir, par exemple., Figlio, Husted, and Kenny 2004; Baicker and
Gordon 2006), some speciﬁcation checks with this particular data might help
buttress the case.

To test whether states that have lawsuits are systematically different than
those that do not, I conduct a number of speciﬁcation checks. Par exemple, il
could be that states that have never experienced a ﬁnance lawsuit are funda-
mentally poor controls to use in estimating the counterfactual. Thus column
4 repeats the regression from column 2 while omitting from the sample all
states that have never had a ﬁnance ruling as of 2002. This results in only a
slight change in the coefﬁcient of interest.

I also test different parameterizations of lawsuit decision measures.12 Col-
umn 5 allows a plaintiff success to change resources through both a one-time
shock to the intercept and a change in the slope of future increases, measured
as years since the lawsuit victory. The results suggest that the effect of lawsuits
is best represented as an intercept shift since the plaintiff dummy coefﬁcient
attenuates slightly while the variable measuring the years since a plaintiff de-
cision has a small and imprecisely estimated coefﬁcient. In column 6 I instead
add a dummy variable equal to one if there has been a court decision favoring
the state in prior years. While the coefﬁcient on plaintiff victories still suggests
that they increase funding by 5.2 pour cent, the results suggest that a state victory
actually decreases future funding by about 3.4 pour cent, although the coefﬁcient
is not signiﬁcantly different from zero. This makes a story linking all lawsuits
and ﬁnance increases to an omitted factor unlikely.

Column 7 tests the effects of legislative reforms versus those of court-
ordered reform. It adds a dummy variable equal to one if there has been a
legislature-induced ﬁnance reform in that state in prior sample years. Le
coefﬁcient is positive, although quite imprecisely estimated, and if taken as
given would imply an effect that is only about half the magnitude of the
court-induced reform. In all three of these speciﬁcation checks, the inclusion
of alternative parameters to control for ﬁnance reform does not signiﬁcantly
reduce the effect of a court-ordered reform, which remains between 0.05 et
0.06.

Another possible explanation is that states with a plaintiff victory are already
on a different ﬁnance trajectory than other states.13 Columns 8 et 9 consider
this possibility. Column 8 adds a 1989 measure of district median income to

12. Berry (2007) and Corcoran and Evans (2007) ﬁnd that adequacy lawsuits are associated with higher
average increases in revenues than equity suits. Cependant, the latter claim that this is not true in
the immediate post-Rose period. Cependant, the sorting of lawsuits into clear categories is not always
straightforward, as Corcoran and Evans (2007), National Access Network (2009), and Springer,
Liu, and Guthrie (2009) differ in which lawsuits are considered adequacy versus equity lawsuits.

13. This could be due, Par exemple, to previous ﬁnance decisions in some states.

468

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

Tableau 4. Relationship between Lawsuit Decisions and Past Revenues

Natural Logarithm of Total per Pupil Revenues

Dependent Variable

Lawsuit decision (at t + 1)

Lawsuit decision (at t + 2)

(1)

0.0001
(0.0167)

(2)

−0.0148
(0.0188)

Remarques: N = 177,131. The regressions use an indicator variable for future passage only in
the relevant year. Standard errors reported in parentheses are corrected for clustering at
the state level. Both columns control for state and year ﬁxed effects, district-level student
demographics, free lunch, and special education status as well as number of students.

the regression to see if the results can be explained by thus distinguishing
high-income districts. Column 9 instead allows each state to follow a separate
time trend. In both cases the plaintiff coefﬁcient estimate is quite robust.14

Tableau 4 investigates whether revenues predict a future lawsuit decision.
Such a correlation would support a reverse causality story, whereby states
experience court-ordered reform because of abnormal levels of school ﬁnance.
Alternatively it might indicate that changes in educational revenues caused
by the mere existence of the lawsuit might predate the ﬁnal court decision.
In each regression of log revenues on district characteristics, future court
rulings have been inserted as a dummy variable in the relevant state-year
observations. The estimated coefﬁcient of column 1 effectively indicates zero
correlation between revenues and lawsuit decisions that will happen one year
in the future. Considering lawsuits that will have decisions two years in the
future actually produces a negative estimated coefﬁcient for the effect on
contemporary revenues, but it is small and statistically insigniﬁcant.

While there is a large amount of variation in the duration of ﬁnance law-
suits, the key point for identiﬁcation is that the timing of the reform inducing
décision, the court ruling, is not correlated with unobserved factors that drive
spending. These regressions show that funding levels do not predict afﬁrma-
tive rulings. The regressions in table 3 also suggest that the plaintiff victory
effect on resources is robust to the inclusion of preexisting trends in state
spending. While there may be state funding disparities that could affect the
timing of lawsuit ﬁlings, if they have nothing to do with the timing of the court
rulings, the strategy holds.

14. There are signiﬁcant differences in the estimated time trend coefﬁcients across states (conditional
on controls), with a few states registering essentially zero real growth once demographics are
accounted for and others with real positive resource trends as high as 3.5 pour cent. Cependant, their
failure to move the lawsuit coefﬁcient suggests that the correlation of plaintiff decisions with base
spending trends is low.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

469

FINANCE LITIGATION AND STUDENT NEEDS

Although it is impossible to rule out all sources of bias, these simple
speciﬁcation checks provide more conﬁdence that the average increase in
school resources is a real phenomenon and that the table 3 estimates are a
good approximation of its magnitude.

4. HOW ARE THE GAINS FROM FINANCE LITIGATION DISTRIBUTED?
Resource Increases and Observable Student Need
The most important previous national study to consider how court ﬁnance
reform affects districts with variable student needs is Card and Payne (2002).
Speciﬁcally they sought to determine whether plaintiff victories in law-
suits changed the linear relationship between district spending and median
household income. Using district-level data from 1977 et 1992, they found
that a plaintiff victory in that time span reduced the size of this coefﬁcient,
decreasing the relationship between income levels in a district and district
ressources, while a state court victory had no effect. To examine whether court
rulings direct resources toward particular needs, I adopt an analogous, albeit
simpler, framework.15

Their choice of a median income measure likely reﬂected both the data at
hand—Card and Payne had two years of data that match up reasonably closely
with Census Bureau income estimates—and a lingering focus on funding
equity. A median household income measure, through its relationship to
median property values, separates property poor and property rich districts.
Because education funding was traditionally tied to property taxes, lawsuits
might affect spending equity by bridging gaps in spending between districts
that arise from differences in property tax base levels.

Cependant, median household income may be less useful as an indicator
of student needs. Le plus important, median incomes do not measure a char-
acteristic of students actually in the public schools but rather of all people
living in a particular area. Because of this, some districts with high median
incomes, such as those in dense urban areas, may have a high proportion of
poor or disadvantaged students in public schools.16 I attempt to address this
by measuring the relationship between school resources and three potential
measures of student need: the fraction of students in a district eligible for

15. The Card and Payne analysis was done in two stages, allowing each state its own coefﬁcient
describing the relationship between district resources and income. Cependant, my decision to cluster
correct standard errors at the state level rules out such an approach.

16. Wilson, Lambright, and Smeeding (2006) ﬁnd that studies that use district averages to calculate the
equity of the resource distribution across children overstate the true inequity found in household
income data. Cependant, such school-level resource distributions actually understate the inequity in
the gap between public spending and student needs.

470

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

Tableau 5. The Relationships among Court-Ordered Reform, Ressources, and Demographic Indicators of Need

Demographic
measure

Plaintiff win

Demographic
main effect

Interaction

Free
Lunch

(1)

0.026
(0.024)

−0.179
(0.050)

0.114
(0.052)

All district revenues

Fraction
Minority

Fraction
Noir

Fraction
Hispanic

Fraction
IEP

(2)

0.063
(0.027)

0.260
(0.056)

−0.020
(0.061)

(3)

0.051
(0.024)

0.202
(0.046)

0.044
(0.081)

(4)

0.061
(0.020)

0.149
(0.068)

−0.030
(0.045)

(5)

0.013
(0.039)

0.361
(0.109)

0.370
(0.298)

177,131

Remarques: The dependent variable for columns 1–5 is the natural logarithm of district revenues. All
columns control for state and year ﬁxed effects, district-level student demographics, free lunch,
and special education status as well as number of students and district urbanicity. Standard errors
reported in parentheses are corrected for clustering at the state level. IEP = individual education
plan.

free lunch, the fraction of minority students, and the fraction of students with
special education status.

In statistical terms I estimate:

Ri j t = α j + τt + δ Pj t + λDi j t + m(P × D)i j t + X

(cid:2)
i j t

c + εi j t ,

(2)

where Di j t is a continuous variable measuring student needs in district i in
state j at time t, and I also include an interaction of the need variable Di j t and
the lawsuit indicatorPi j t . In this framework the coefﬁcient estimate of λ gives
the linear relationship between student needs and resources in the absence
of a successful ﬁnance lawsuit, while the coefﬁcient estimate of μ indicates
how a plaintiff victory changes the relationship, and δ represents the effect of
a plaintiff victory in a district with a zero value for the chosen demographic
measure. Tableau 5 presents estimates of equation 2 for each demographic need
variable. The speciﬁcation follows the base speciﬁcation in column 3 of table 3
except for the addition of an interaction term between a particular demographic
variable and the plaintiff victory dummy variable.

Column 1 of table 5 considers the fraction of students eligible for free
lunch as a measure of the economic status of the students in the district.
The estimated coefﬁcient for the fraction free lunch eligible variable sug-
gests a signiﬁcant negative relationship between districts with high-need
students and revenue levels. Cependant, the predicted resource disparity of
roughly −0.18 log points between an entirely free lunch district and a dis-
trict with no free lunch eligible students seems small compared to commonly

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

471

FINANCE LITIGATION AND STUDENT NEEDS

held beliefs about massive funding inequality. Néanmoins, coefﬁcient es-
timates for the interaction term suggest that success in a plaintiff law-
suit mitigates the negative relationship by almost two-thirds. This result is
roughly analogous to the Card and Payne estimates for median income and
ressources.

Although free lunch eligibility might be considered an excellent proxy
for student need and is the closest approximation to a true student poverty
measure available, it has some potential limitations. Cruse and Powers (2006)
highlight data issues with the free lunch measure in the NCES data including
district nonparticipation, partial district nonresponse, differential patterns of
enrollment, and effort at enrollment and conclude that the resulting prediction
error is sufﬁciently high to preclude their use in making formal school district
poverty estimates.

Thus I also present estimates using the fraction of minority students as
my demographic measure of need in column 2. Here the effect appears to be
reversed. The estimated coefﬁcient on fraction minority is large and positive,
suggesting that districts with a high proportion of minorities have larger per
student resources. Of course many of these may also be in urban areas with
high relative costs for teacher salaries and other expense categories. La plupart
interestingly, while the coefﬁcient is small and not statistically different from
zero, the point estimate actually suggests that court reform slightly ﬂattens
the positive relationship between minority percentage and log revenues for a
district.

One possibility is that minority percentage is actually an amalgam of tradi-
tionally disadvantaged groups with other groups such as Asian students that
traditionally perform above state averages. Under these conditions it might not
serve as a good indicator of student needs. To check this possibility, columns 3
et 4 repeat the analysis using the district fraction of black and Hispanic stu-
dents as the respective need indicators. In both cases the positive preexisting
relationship between minority students and funding levels remains. Cependant,
in the case of black students the point estimate suggests that plaintiff ﬁnance
decisions may have a small positive effect on resource progressivity, cependant
the standard errors are too large to draw ﬁrm conclusions.

I repeat the above exercise in column 5 using one other potential indicator
of student need, the fraction of students that have an individual education
plan (IEP).17 The resource requirements of these students likely differ in fun-
damental ways from those of economically disadvantaged students. The results

17. This category is analogous to the traditional category of special education, though it has been

broadened to incorporate a wider variety of students with special individual needs.

472

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

suggest that districts with a large proportion of IEP students actually receive
à propos 0.36 log points more funding in the absence of a ﬁnance lawsuit. Le
interaction term further suggests that a lawsuit doubles the funding advantage
of these districts.

Taken at face value, these regressions suggest that lawsuits likely have a
disproportionately large and positive impact on the funding for districts with
a higher proportion of free lunch or IEP students. Cependant, just as the es-
timation of a conditional mean relationship between resources and lawsuits
might obscure differential lawsuit effects across the funding distribution, le
imposed linearity of the demographic lawsuit interaction term may hide im-
portant patterns. Consider, Par exemple, the patterns in minority enrollment
among districts, where the transition from very small rural schools to medium
suburban schools is forced to have the same effect as the transition to very
large urban districts.

To relax this restriction, I replace the plaintiff intercept and linear slope
interaction terms from equation 2 with four interactions, one between each
quartile of a particular demographic variable and the plaintiff victory dummy
variable. Thus each regression from panel A, columns 1–4 of table 6 reports
four interaction coefﬁcients, one for each quartile of districts sorted by student
need within each state-year combination. Quartile order is increasing in the
fraction of free lunch students.

In column 1 an interesting pattern emerges: a plaintiff lawsuit appears to
provide more resources to districts in the extremes of the free lunch student
distribution compared with those in the center. Thus lawsuits are associated
with resource increases of 6–7 percent for districts with extremes of eco-
nomically disadvantaged students, while those districts in the middle of the
distribution of economic need receive only 4 percent increases. A test across all
four coefﬁcients rejects the null hypothesis of quartile effect equality, and pair-
wise tests across coefﬁcients conﬁrm that the top and bottom quartile effects
differ signiﬁcantly from those on middle quartiles. Although the difference
in not statistically signiﬁcant, the point estimates also suggest that schools
with the most free lunch students may gain slightly relative to those with the
fewest.

Pursuing the same exercise for the fraction of minority students in column
2 suggests, perhaps unexpectedly, that the greatest gains from court reform
occur in schools with the fewest minority students. En fait le 10 pour cent
gain in resources for schools in the lowest minority quartile is signiﬁcantly
different from the effects in all other quartiles, 4 percentage points higher
than those in the second quartile and around 6 percentage points greater
than those in the 3rd and 4th quartiles. While not reported, similar regres-
sions using fraction black or fraction Hispanic students as the need indicator

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

473

FINANCE LITIGATION AND STUDENT NEEDS

Tableau 6. Court-Ordered Finance Reform and Student Need: Nonlinear Effects

Contemporary Need Measures

1989 Need Measures

Free
Lunch Minority

Fraction Fraction Fraction
IEP

IEP

Free
Lunch Minority

Fraction Fraction

(1)

(2)

(3)

(4)

(5)

(6)

IEP

(7)

UN. Total revenues

Need main effect

Plaintiff win x

Need quartile 1

Need quartile 2

Need quartile 3

Need quartile 4

p-value [quartile

equality] − F(3,48)

−0.156
(0.051)

0.288
(0.061)

0.385
(0.110)

0.210
(0.116)

−0.114
(0.056)

0.289
(0.060)

0.315
(0.141)

0.066
(0.022)

0.043
(0.021)

0.044
(0.024)

0.069
(0.020)

0.108
(0.030)

0.064
(0.030)

0.045
(0.021)

0.047
(0.022)

0.042
(0.020)

0.047
(0.019)

0.057
(0.021)

0.102
(0.021)

0.028
(0.022)

0.037
(0.020)

0.050
(0.021)

0.096
(0.022)

0.061
(0.023)

0.051
(0.022)

0.042
(0.023)

0.081
(0.019)

0.094
(0.031)

0.062
(0.027)

0.045
(0.021)

0.043
(0.021)

0.055
(0.020)

0.055
(0.022)

0.098
(0.026)

[0.014]

[0.000]

[0.004]

[0.001]

[0.027]

177,131 177,131

177,131 148,008

170,248 170,248

170,248

B. State revenues

Need main effect

Plaintiff win x

Need quartile 1

Need quartile 2

Need quartile 3

Need quartile 4

0.866
(0.179)

−0.038
(0.097)

0.825
(0.381)

0.545
(0.487)

0.093
(0.058)

0.133
(0.054)

0.098
(0.046)

0.165
(0.102)

0.206
(0.047)

0.121
(0.042)

0.081
(0.042)

0.126
(0.059)

0.018
(0.048)

0.132
(0.039)

0.186
(0.051)

0.219
(0.070)

−0.029
(0.065)

0.086
(0.057)

0.161
(0.063)

0.196
(0.078)

p-value [quartile

equality] − F(3,48)

[0.000]

[0.001]

[0.017]

[0.035]

177,131 177,131

177,131 147,887

Remarques: Each panel-column represents the results of a separate regression. The dependent variable
for panel A is the natural logarithm of district revenues, except in column 4 where revenues directed
speciﬁcally to special education categories are omitted. Panel B uses as its dependant variable
only the natural logarithm of state revenues to each district. Quartile order is increasing in the
respective demographic variable. All columns control for state and year ﬁxed effects, district-level
student demographics, free lunch, and special education status as well as the number of students
and district urbanicity. Standard errors reported in parentheses are corrected for clustering at the
state level. Numbers in brackets are p-values for F-tests of coefﬁcient equality across all quartiles.
IEP = individual education plan.

also produce decreasing coefﬁcients from quartiles 1–3 with a slight uptick in
quartile 4.

Curiously, the only demographic measure where resources increase mono-
tonically with the size of the potentially disadvantaged population is the fraction

474

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

of special education. En effet, the results of column 3 suggest that districts with
the most IEP students receive almost 2.5 times the funding increase of those
with the fewest such students.

Such a strong result suggests that categorical special education funding
may be driving this ﬁnding. Thus column 4 repeats the analysis after remov-
ing targeted funds for special education from the dependent variable resource
measure.18 The pattern from column 3 persists, while the difference between
funding in the top and bottom quartiles is actually larger and remains sta-
tistically signiﬁcant. This suggests that ﬁnance lawsuits disproportionately
increase funding to districts with large IEP populations in categories beyond
targeted IEP funding itself.

One potential worry is that these results are artifacts of incentives that
might be embodied in court-ordered ﬁnance reform for districts to increase
the number of students designated as high need. This might work in a similar
way to the pattern of increased disability diagnoses incentivized by high-stakes
accountability documented by Figlio and Getzler (2006). Par exemple, if there
are ﬁnancial rewards to having more IEP students, districts might be more
likely to give marginal students that designation. A quick check with my data
shows that a lawsuit victory has no predictive power on the future reported
number of IEP students.19 As further evidence against this possibility, columns
5–7 of table 6 show that the same pattern of results emerges even when district
need indicators from 1989, the beginning of my study period, are used to
sort districts into need quartiles. This suggests that district reclassiﬁcation of
students is unlikely to explain my results.

Another explanation for this pattern of results is that the legislative re-
sponse to court-ordered ﬁnance reform operates under political or legal con-
straints, which make it infeasible to perfectly target increased money to high-
need districts. If this is the case, a similar pattern of results should emerge in
the effects of court-ordered reform on state revenues to school districts. Panel B
of table 6 considers this possibility by substituting revenues from state sources
as the dependent variable. The resulting estimates for baseline need effects
suggest that state aid follows a different pattern than total revenues in the
absence of a plaintiff lawsuit victory. C'est, states give much more money to
districts with a large proportion of IEP students or free lunch students, alors que
there is little state aid difference across minority status measures. En outre,

18. This includes both federal and state funding for special education and handicapped education as
well as any level of revenue categorized as speciﬁcally for IEP students. Because these categories
are available only in the revenue data following the 1991 school year, the ﬁrst two years of the FNF
data are dropped from the column 3 regressions. This mechanically increases standard errors.
19. The relevant coefﬁcient estimate on the plaintiff dummy is 0.004 with a resulting 95 pour cent

conﬁdence interval of (−0.010, 0.020).

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

475

FINANCE LITIGATION AND STUDENT NEEDS

the results suggest that the average effect of lawsuits on state dollars is greater
than that on total revenues, as we might expect.

Cependant, many of the effects of student need on the distribution of court-
ordered spending increases show patterns similar to those seen with total
revenues. While there are some differences from the exact pattern of panel A
(par exemple., the 2nd quartile of free lunch districts receives more revenue than either
of its neighbors, and there is a slight uptick in revenues to the highest minority
districts), the results continue to suggest that there is some targeting of money
to the highest-need districts but that the targeting is imperfect.

Some of the results of table 6 might be due to the collinear nature of
the variables. Peut-être, Par exemple, the slower growth in resources for high
minority districts might be an artifact of the greater resources given to districts
with high proportions of IEP students. Tableau 7 considers this possibility by
examining the simultaneous effect of court ﬁnance reform on districts across
two jointly considered demographic measures. The procedure is much the
same as above, except instead of forming four interaction effects deﬁned by
the quartiles of one demographic variable, each regression considers sixteen
effects from a four-by-four grid of two demographic variables.20 Thus we can
look at how lawsuit-acquired resources accrue to high-minority, low free lunch
districts versus low-minority, high free lunch districts.

Each panel of table 7 considers the results for the different possible pairings
of potential student need indicators in a separate regression, with included
level controls for other demographic variables, district urbanicity, and state
and year effects, though these are not reported. Panel A, Par exemple, shows
that a plaintiff court decision increases resources for districts with the lowest
proportion of both free lunch and IEP students by 6.6 pour cent. Cependant,
the effect on the districts with the highest proportion of special needs and
free lunch students is signiﬁcantly greater at 12 pour cent. While this suggests
that post-Rose lawsuits promoted some matching of resources to needs, là
appear to be more complex patterns at work, as evidenced by the fact that fairly
high-need districts in the third quartile of both special education and free lunch
students only saw a 3 percent increase in resources. Although this resource
change cannot be statistically differentiated from zero, it is signiﬁcantly less
than the resource gains of either the lowest- or highest-needs districts.

En effet, these results largely support the patterns seen in table 6. With the
exception of districts in the lowest quartile of free lunch students, resource
gains are monotonically increasing in special education students, and even in
this exceptional case the districts with the most special education students see

20. Although it is possible to repeat the same process for a sixty-four-cell grid considering all three
demographic variables, the use of state-level clustered standard errors makes this impractical.

476

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

Tableau 7. The Effects of Court Reform on Resources across Joint District Demographic Characteristics

Free Lunch Quartile

Minority Quartile

IEP

quartile

1 0.066
(0.023)
[0.026]

0.016
(0.023)
[0.000]

0.031
(0.023)
[0.000]

0.024
(0.019)
[0.000]

IEP

quartile

1 0.103
(0.035)
[0.949]

0.064
(0.032)
[0.127]

0.035
(0.022)
[0.000]

0.021
(0.022)
[0.001]

2 0.041
(0.020)
[0.001]

0.038
(0.022)
[0.000]

0.033
(0.027)
[0.000]

0.072
(0.024)
[0.021]

3 0.066
(0.025)
[0.031]

0.045
(0.022)
[0.000]

0.033
(0.026)
[0.000]

0.075
(0.025)
[0.016]

4 0.099
(0.033)
[0.501]

0.084
(0.027)
[0.033]

0.082
(0.028)
[0.031]

0.120
(0.022)
[N/A]

2 0.077
(0.030)
[0.287]

0.038
(0.024)
[0.002]

0.031
(0.023)
[0.000]

0.049
(0.019)
[0.002]

3 0.105
(0.030)
[0.996]

0.058
(0.031)
[0.059]

0.040
(0.020)
[0.000]

0.051
(0.023)
[0.015]

4 0.152
(0.027)
[0.067]

0.100
(0.033)
[0.852]

0.069
(0.025)
[0.059]

0.105
(0.020)
[N/A]

Free Lunch Quartile

Minority

quartile

1 0.109
(0.033)
[0.148]

0.080
(0.037)
[0.610]

0.081
(0.030)
[0.498]

0.156
(0.029)
[0.000]

Noir

quartile

1 0.114
(0.037)
[0.239]

0.097
(0.036)
[0.479]

0.086
(0.030)
[0.585]

0.128
(0.022)
[0.013]

2 0.062
(0.029)
[0.925]

0.051
(0.033)
[0.837]

0.057
(0.035)
[0.948]

0.091
(0.033)
[0.374]

3 0.036
(0.022)
[0.318]

0.036
(0.023)
[0.173]

0.044
(0.029)
[0.633]

0.069
(0.036)
[0.829]

4 0.066
(0.028)
[0.730]

0.028
(0.024)
[0.103]

0.030
(0.024)
[0.197]

0.059
(0.022)
[N/A]

2 0.086
(0.032)
[0.605]

0.050
(0.031)
[0.744]

0.054
(0.030)
[0.810]

0.071
(0.026)
[0.845]

3 0.051
(0.024)
[0.715]

0.029
(0.022)
[0.142]

0.039
(0.022)
[0.352]

0.048
(0.020)
[0.613]

4 0.042
(0.027)
[0.189]

0.037
(0.022)
[0.198]

0.036
(0.026)
[0.395]

0.064
(0.026)
[N/A]

Remarques: Each panel reports the results of a separate regression, analogous to those of table 6.
Coefﬁcients reﬂect the marginal resource response of districts to court-ordered ﬁnance reform for
each cell of a quartile grid in two demographic indicators. The dependent variable is the natural
logarithm of district revenues. Quartile order is increasing in the respective demographic variable.
All panels control for state and year ﬁxed effects, and level values of other demographic controls.
Standard errors reported in parentheses are corrected for clustering at the state level. Numbers in
brackets are p-values for F-tests of coefﬁcient equality with the highest-need quartile pair in each
panel. IEP = individual education plan.

the greatest gains. D'autre part, the free lunch demographic measure
generally appears to generate a U-shaped pattern of higher resources for the
districts in the extremes of the free lunch distribution than for those in the
middle.

Panel B considers the combination of a district’s IEP level and fraction
minority. While the lowest- and highest-need districts appear to make roughly
equal gains by these measures, it is noteworthy that the largest gains by far are
made by the districts with the highest fraction of special education students
and the lowest fraction of minority students. Although the relationship is

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

477

FINANCE LITIGATION AND STUDENT NEEDS

not as clear as that in panel A, it seems generally true that looking along
the rows of the table the highest-minority districts receive smaller funding
increases than low-minority districts, conditional on special education needs.
Looking in the other direction, conditional on the fraction of minority students,
districts with a higher fraction of students with IEPs experience larger resource
gains.

The evidence of panel C largely conﬁrms these patterns. In most cases there
appears to be a U-shaped relationship between free lunch eligible students and
district resource gains and a negative relationship between fraction minority
and resource gains. The point estimates further suggest that the lowest-need
districts receive more resources from ﬁnance litigation than the highest-need
districts, yet a signiﬁcantly greater increase is reserved for the districts with the
fewest minorities and the most free lunch students. Panel D shows that these
relationships persist when need is deﬁned in terms of a particular minority
group presence, in this case black students.21

Discussion
The effectiveness of post-Rose lawsuits in selectively increasing funding to the
districts with higher indications of student need appears mixed. Court-ordered
reform appears to successfully allocate more resources to schools with more
IEP students and does target more resources to districts with the highest free
lunch needs. Cependant, it appears that the lawsuits also funnel signiﬁcant
resources to districts with few free lunch students and that the redistribution
of resources is negatively related to the fraction of minority students. Le
greater responsiveness of resources to poverty or special education measures
than to race likely reﬂects legal and political factors that limit the use of race as
an explicit component in ﬁnance decisions. It is also noteworthy that minority
students are almost equally likely to come from a district in the top versus
bottom half of the revenue distribution.

While it is possible that the U-shaped pattern of resource response across
the distribution of economic need reﬂects problems with the data or factors
unrelated to ﬁnance litigation, free lunch eligibility was the most often used
criteria for determining at-risk students in state education ﬁnance decisions
in the 1990s (Thompson and Silvernail 2001). The most likely explanation for
this peculiar pattern of resource distribution is rooted in the legislative process.
Support for increasing funding to the neediest districts likely requires legisla-
tive compromises that raise support, sometimes selectively, for other districts.

21. A regression using Hispanic percentage as the racial need indicator also produces monotonically
decreasing effects as Hispanic fraction increases. Regressions using revenues from state sources
as the dependent variable produce similar patterns but are much less precise and are not reported.

478

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

The estimates of table 6, panel B, tend to support this story of legislative
exigency.

The positive resource effect for districts with a high proportion of IEP stu-
dents is also curious, because the lawsuits considered in this case are distinct
from those ﬁled on behalf of special education students.22 One possibility is
that the fraction of IEP students in a district is correlated with other types of
student needs or district organization and should be regarded as a broader
proxy of need.

Cependant, there is some evidence to suggest that this effect may be real.
A periodic analysis of state special education ﬁnance systems suggests that
six of the eight states with plaintiff decisions in the 1989–1993 period had
undertaken major reforms of their special education ﬁnance systems by 1994
(Parrish et al. 1997). During the 1995–2000 period several states facing court
rulings also reformed their ﬁnance systems. Wyoming increased its cost reim-
bursement from special education funding from 85 à 100 pour cent, Arizona
increased its pupil weights substantially, and states like North Carolina, Ohio,
New Jersey, and Alaska chose to adopt completely new funding mechanisms
(Parrish et al. 2003). Because many of these programs work through some
sort of general or census aid grant, they are likely to show up in these data
outside targeted special education funds. In either case, the ﬁnding supports
the notion that adequacy lawsuits are directing more funding to districts that
show objective signs of greater need.

Reconciliation with the Literature
A ﬁnal issue is reconciling the results that indicate nearly equal resource
increases across the spending distribution found by Berry (2007), Corcoran
and Evans (2007), and Springer, Liu, and Guthrie (2009) with my results
suggesting a shift in funding toward the districts with the highest free lunch
and special education needs. I begin in table 8 by showing that these seemingly
contrary results are not artifacts of data set differences.

Panel A presents the results of two regressions of state by year-level Gini
and Theil indices of spending inequality on a dummy for a plaintiff school
ﬁnance decision and the control variables used in table 3, column 3.23 Though

22. There were literally hundreds of special education lawsuits that received judgments in the 1990s
(Zirkel 1997). Cependant, these lawsuits in this era are most often ﬁled against a speciﬁc district
or the state for failing to follow appropriate procedures or to assign the child to the appropriate
program rather than over the general state funding of programs. The most notable exception is
the Michigan decision Durant v. State of Michigan (1997), where the state was ordered to increase
funding reimbursements for special education.

23. The derivation of the Gini and Theil indices can be found in Murray, Evans, and Schwab (1998).

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

479

FINANCE LITIGATION AND STUDENT NEEDS

Tableau 8. Effect of a Lawsuit Victory on the Distribution of School Resources

UN. Inequality index

B. Mean effect

C. Conditional quantile

Gini

Theil

.05

.10

.25

.50

.75

.90

.95

−0.0023
(0.0047)

−0.0020
(0.0027)

0.058
(0.019)

0.062
(0.023)

0.062
(0.023)
[0.97]

0.065
(0.022)
[0.78]

0.061
(0.023)
[0.72]

0.047
(0.023)
[0.13]

0.072
(0.025)
[0.12]

0.060
(0.034)
[0.57]

Remarques: Each reported coefﬁcient is from a separate regression. In panel A
the regressions reﬂect state-level observations, and the dependent variable
is the stated equality index multiplied by 100. Panel B restates ordinary least
squares regression results from table 3, column 3. Standard errors reported
in parentheses in panels A and B are corrected for clustering at the state
level. Panel C presents the results from a series of quantile regressions for
the listed quantiles using log revenues as the dependent variable. Standard
errors are derived through a 100 repetitions bootstrap that incorporates the
clustered data design. All regressions include student levels, course, free lunch
status, and special education controls as well as state and year ﬁxed effects.
Numbers in brackets are p-values of an F test with null hypothesis that the
coefﬁcient equals the coefﬁcient of the preceding quantile.

both coefﬁcients are negative in sign, they are small with relatively large stan-
dard errors. By comparison, these coefﬁcient estimates are very close to those
of Corcoran and Evans (2007) for the post-Rose era and suggest a comparable
lack of evidence that these lawsuits compressed the spending distribution.

Another approach in these studies, most notably the recent work of Cor-
coran et al. (2004) and Corcoran and Evans (2007), uses values for different
centiles of the unconditional state spending distribution instead of a state mean

480

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

or inequality index as the dependent variable in their regressions. Combining
data from the post-Rose period with this methodology, the latter study is unable
to reject equal positive resource effects at various points in the unconditional
resource distribution. Panel C of table 8 presents a slightly different take on the
same question. Because I wish to control for the possible effects of potential
confounding variables, I use quantile regression to look at the effect of school
ﬁnance litigation on the conditional quantiles of the resource distribution.24

The quantile regression estimates are reported in table 8, panel C.25 The
bracketed numbers are p-values for a test of coefﬁcient equality with the pre-
ceding quantile. While my methodology differs slightly from those of previous
études, the take-home message is the same. The coefﬁcient estimates for the
.05, .10, et .25 quantiles suggest that a court verdict raises resources in the
lower tail of the distribution by about 6 pour cent. Cependant, in percentage terms
these gains are nearly identical to the gains at the distribution median and
le .95 quantile. From this baseline, the point estimates for the .75 quantile
appear to be somewhat lower and those for the 0.9 quantile somewhat higher,
though we cannot statistically reject the hypotheses that neither is signiﬁcantly
different from the effects on other quantiles. Thus the evidence seems to favor
a story under which court-ordered ﬁnance reform post-Rose leads to roughly
equal gains for the districts at almost all points in the resource distribution.

Tableau 9 provides a crucial insight in reconciling these ﬁndings with my
results on student needs. It presents the percentage of districts from each of
le 1989 revenue quartiles in my data that fall into a particular quartile of
student need. For reference, a completely uniform distribution would have
all cell values equal to 25 pour cent. The missing piece of the puzzle here is
the fact that the initial distribution of student needs across revenue quartiles
is surprisingly ﬂat. Thus among the highest revenue districts, 26.3 pour cent
of them come from the lowest free lunch quartile while 32 percent of them
are from the highest. In no case is a need quartile representative of less than
19 percent of the observations of its revenue quartile, and a full half of the

24. The algorithm proposed by Koenker and Bassett (1978) chooses parameters to minimize the sum,
across observations, of absolute deviations between the dependent variable and the linear combi-
nation of independent variables and parameters. Each observation is weighted by a check function
that scales for the proper quantile and ensures that all deviations are taken in the correct direction.
The estimated coefﬁcients from a quantile regression model are commonly interpreted as marginal
effects on the conditional distribution of the dependent variable measured at a particular conditional
quantile. Thus quantile regression allows me to compare the effect of a lawsuit victory at the top of
the conditional resource distribution—say, the 95th quantile—with the effect on the bottom of the
distribution—for example, the 5th quantile.

25. The standard errors for the reported quantile regressions are obtained via a 100-iteration bootstrap,
where the sampling procedure accounts for the clustered nature of the data. The estimated variance
covariance matrix allows testing of coefﬁcient equality across quantiles. For more about interpreting
these quantile regression results in education policy contexts, see Eide, Showalter, and Sims (2002).

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

481

FINANCE LITIGATION AND STUDENT NEEDS

Tableau 9. Distribution of Preexisting District Need Given District Revenues

UN. Revenues versus free lunch

Free lunch quartile

Revenue Quartile

28.7%

27.7%

24.4%

19.3%

23.0%

26.7%

28.3%

21.9%

22.3%

25.2%

26.1%

26.4%

B. Revenues versus special education (IEP)

IEP quartile

Revenue Quartile

25.9%

26.2%

26.9%

20.9%

22.6%

26.5%

24.4%

22.3%

25.4%

26.9%

25.5%

26.3%

20.2%

21.4%

32.0%

30.7%

20.7%

19.8%

28.8%

Remarques: The table shows the percentage of year 1989 district observations from a given revenue
quartile that fall into the indicated quartile of student need. Quartiles are from smallest values (1)
to largest (4). Numbers may not sum to 100 percent due to rounding. IEP = individual education
plan.

cells in the table are within 2 percentage points of the uniform 25 pour cent
measure. This relatively even distribution across revenue quartiles suggests
that selective funding increases for the districts with the greatest needs can be
consistent with the ﬁndings of minimal changes to the shape of the resource
distribution since those high-need districts are drawn from across the resource
distribution rather than from one particular section.

5. CONCLUSION
School ﬁnance litigation has been a deﬁning feature of the education world for
the past thirty-ﬁve years. In the wake of the Rose decision, the role of adequacy
considerations became increasingly important. This article tests whether there
is evidence that lawsuits of this era direct resources to districts with plausible
indicators of high student need. While I present quantile regression estimates
that conﬁrm earlier cross-state research that these post-Rose court rulings do
not appreciably change the shape of the school resource distribution, je discute
that the raw spending distribution as a relative measure of school resources is
not the object of primary importance when the goal is providing an adequate
éducation. The idea that adequacy lawsuits are lifting all boats by increasing
school resources across the distribution is an overly simplistic description
of the situation. En effet, I show that plaintiff victories appear to lead to a

482

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

redistribution of resources that favors districts with the most high-need stu-
bosses. This targeting is imperfect because most districts appear to gain some
ressources, and districts with the lowest level of free lunch students also capture
a disproportionate share. While further research is needed to better understand
the legislative mechanism by which this occurs, this study gives reason for lim-
ited optimism about the retargeting of resources brought about through school
ﬁnance litigation.

This article has beneﬁted from the helpful suggestions of Lars Lefgren, Mark Showalter,
Eric Eide, participants at the 2007 AEFA Annual Conference, and editorial and referee
remarks. I am also grateful for the diligent research assistance of Chad Lee and Biff
Jones. Errors remain the author’s responsibility.

RÉFÉRENCES
Aaronson, Daniel. 1999. The effect of school ﬁnance reform on population heterogene-
ville. National Tax Journal 51: 1–29.

Baicker, Katherine, and Nora Gordon. 2006. The effect of state education ﬁnance
reform on total local resources. Journal of Public Economics 90: 1519–35.

Berry, Christopher R. 2007. The impact of school ﬁnance judgments on state ﬁscal pol-
icy. In School money trials: The legal pursuit of educational adequacy, edited by Martin R.
West and Paul E. Peterson, pp. 213–42. Washington, CC: Brookings Institution Press.

Bertrand, Marianne, Esther Duﬂo, and Sendhil Mullainathan. 2004. How much
should we trust differences-in-differences estimates? Quarterly Journal of Economics
119: 249–75.

Card, David, et un. Abigail Payne. 2002. School ﬁnance reform, the distribution of
school spending, and the distribution of student test scores. Journal of Public Economics
83: 49–82.

Clark, Melissa A. 2003. Education reform, redistribution, and student achievement:
Evidence from the Kentucky Education Reform Act. Unpublished paper, Princeton
University.

Commonwealth v. Dedham. 1819. 16 Mass. 141, 146 (Mass. 1819).

Corcoran, Sean P., and William N. Evans. 2007. Equity, adequacy and the evolving
state role in education ﬁnance. In Handbook of research in education ﬁnance and policy,
edited by Helen F. Ladd and Edward B. Fiske, pp. 332–56. New York: Routledge.

Corcoran, Sean P., William N. Evans, Jennifer Godwin, Sheila E. Murray, and Robert
M.. Schwab. 2004. The changing distribution of education ﬁnance, 1972 à 1997. Dans
Social inequality, edited by Kathryn Neckerman, pp. 433–65. New York: Russell Sage
Fondation.

Cruse, Craig, and David Powers. 2006. Estimating school district poverty with free
and reduced-price lunch data. Washington, CC: U.S. Census Bureau, Small Area
Estimates Branch.

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

483

FINANCE LITIGATION AND STUDENT NEEDS

Dee, Thomas S., and Jeffrey Levine. 2004. The fate of new funding: Evidence from
Massachusetts’ education ﬁnance reforms. Educational Evaluation and Policy Analysis
26: 199–215.

Downes, Thomas A. 2004. School ﬁnance reform and school quality: Lessons from
Vermont. In Helping children left behind: State aid and the pursuit of educational equity,
edited by John Yinger, pp. 283–313. Cambridge, MA: AVEC Presse.

Downes, Thomas A., and David N. Figlio. 1998. School ﬁnance reforms, tax limits,
and student performance: Do reforms level-up or dumb down? Unpublished paper,
Tufts University.

Downes, Thomas A., and David N. Figlio. 1999. Economic inequality and the provision
of schooling. Economic Policy Review 5: 99–110.

Downes, Thomas A., and Mona P. Shah. 2006. The effect of school ﬁnance reforms on
the level and growth of per-pupil expenditures. Peabody Journal of Education 81: 1–38.

Durant v. State of Michigan. 1997. 566 N.W.2d 272 (Michigan Supreme Court).

Eide, Eric R., Mark H. Showalter, and David P. Sims. 2002. The effects of secondary
school quality on the distribution of earnings. Contemporary Economic Policy 20:
160–70.

Evans, William N., Sheila E. Murray, and Robert M. Schwab. 1997. School houses,
court houses and state houses after Serrano. Journal of Policy Analysis and Management
16: 10–31.

Figlio, David N., and Lawrence Getzler. 2006. Accountability, ability and disability:
Gaming the system? In Advances in microeconomics, edited by Timothy Gronberg, pp.
35–49. Amsterdam: Elsevier.

Figlio, David N., Thomas A. Husted, and Lawrence W. Kenny. 2004. Political economy
of the inequality in school spending. Journal of Urban Economics 55: 338–49.

Flanagan, Ann E., and Sheila E. Murray. 2004. A decade of reform: The impact of
school reform in Kentucky. In Helping children left behind: State aid and the pursuit of
educational equity, edited by John Yinger, pp. 195–313. Cambridge, MA: AVEC Presse.

Hanushek, Eric A. 2003. The failure of input-based schooling policies. Economic
Journal 113: 64–98.

Hanushek, Eric A. 2005. The alchemy of costing out an adequate education. Paper
presented at the Conference on Adequacy Lawsuits, Université Harvard, Octobre.

Hoxby, Caroline M. 2001. All school ﬁnance equalizations are not created equal.
Quarterly Journal of Economics 116: 1189–231.

Husted, Thomas A., and Lawrence W. Kenny. 1997. Efﬁciency in education: Evidence
from the states. In Proceedings of the Eighty-Ninth Annual Conference on Taxation, pp.
358–65. Washington, CC: National Tax Association.

Koenker, Roger, and Gilbert Bassett. 1978. Regression quantiles. Econometrica 46:
33–50.

484

D
o
w
n
o
un
d
e
d

F
r
o
m
h

t
t

:
/
/

d
je
r
e
c
t
.

je
t
.

e
d
toi
e
d
p
un
r
t
je
c
e
–
p
d

F
/

6
4
4
5
5
1
6
8
9
2
9
4
e
d
p
_
un
_
0
0
0
4
4
p
d

b
oui
g
toi
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

David P. Sims

Manwaring, Robert L., and Steven M. Sheffrin. 1997. Litigation, school ﬁnance reform,
and aggregate educational spending. International Tax and Public Finance 4: 107–27.

Murray, Sheila E., William N. Evans, and Robert M. Schwab. 1998. Education-ﬁnance
reform and the distribution of education resources. American Economic Review 88:
789–812.

National Access Network. 2009. Litigation. Available www.schoolfunding.info/
litigation/litigation.php3. Accessed 28 Avril 2009.

Parrish, Thomas B., Jennifer Harr, Jennifer Anthony, Amy Merickel, and Phil Esra.
2003. State special education ﬁnance systems, 1999–2000. Palo Alto, Californie: Center for
Special Education Finance.

Parrish, Thomas B., Fran O’Reilly, Ixtlac E. Duenas, and Jean Wolfman. 1997. State
special education ﬁnance systems, 1994–1995. Palo Alto, Californie: Center for Special Education
Finance.

Pauley v. Kelly. 1979. 255 S.E.2d 859 (WV Supreme Court of Appeals).

Robinson et al. v. Cahill et al. 1973. 303 A.2d 273 (NJ Supreme Court).

Rose v. Council for Better Education. 1989. 790 S.W.2d 186, 60 Ed. Law Rep. 1289.

San Antonio Independent School District v. Rodriguez. 1973. 411 U.S. 1.

Serrano v. Priest. 1971. 487 P.2d 1241 (Cal. Supreme Court 1971).

Silva, Fabio, and Jon Sonstelie. 1995. Did Serrano cause a decline in school spending?
National Tax Journal 48: 199–215.

Springer, Matthew G., Keke Liu, and James W. Guthrie. 2009. The impact of school
ﬁnance litigation on resource distribution: Comparing court-mandated equity and
adequacy reform. Education Economics 17: 421–44.

Thompson, UN. Mavourneen, and David L. Silvernail. 2001. States’ provisions of
extra funding for economically-disadvantaged students. Gorham, ME: Education Policy
Research Institute.

Wilson, Kathryn, Kristina Lambright, and Timothy M. Smeeding. 2006. École
ﬁnance, equivalent educational expenditure and the income distribution: Equal dollars
or equal chances for success. Education Finance and Policy 1: 396–424.