From Separate and Unequal to Integrated and Equal?

From Separate and Unequal to Integrated and Equal?
School Desegregation and School Finance in Louisiana
Sarah J. Reber
University of California, Los Angeles
NBER
February 2009
Abstract
This paper examines the effects of one of the most important policies aimed at closing racial
school quality gaps: school desegregation. Desegregation was imposed on local school
districts by federal policy and might have induced unintended behavioral responses of white
families as well as state and local governments. This study examines the responses of white
families and state and local governments to desegregation and is the first to examine the
effects of desegregation on the finances of school districts. Importantly, desegregation did
induce white flight from blacker to whiter public school districts and to private schools, but
the local property-tax base and local revenue were not adversely affected. White flight likely
would have been even more pronounced if not for a significant “payoff” from the state to
districts wither high black shares of enrollment. The state legislature directed significant new
funding to districts where whites were particularly affected by desegregation, allowing quality
in desegregated schools to be “leveled up” to that previously experienced only in the white
schools. Desegregation appears to have achieved its intended goal of improving resources
available in schools attended by blacks.
_____________________________
I am grateful to Ken Chay, David Cutler, Claudia Goldin, Nora Gordon, Caroline Hoxby,
Larry Katz, Rob Lemke, Tara Watson and participants in seminars at Berkeley, the National
Tax Association, and the Institute for Research on Poverty for helpful discussions and
advice and to Margo Schlanger for help understanding the legal history. Chris Antunes, Kara
Sullivan, and Garett Trombly provided excellent research assistance. All errors are my own.
Funding from the National Science Foundation, the American Educational Research
Association, and the Spencer Foundation is gratefully acknowledged.
I. Introduction
Policymakers and advocates have long been interested in reducing socioeconomic
and racial inequality of school quality, but this can be difficult in light of the decentralized
nature of primary and secondary education in the United States. Individual families as well as
state and local governments can often undermine the efforts of higher levels of government.
The desegregation of Southern education was perhaps the largest policy aimed at eradicating
inequality in education in the 20th century. Improving school quality for blacks—by tying
their fate to that of whites—was an important motivation for desegregating public schools.1
This paper uses newly collected data to analyze a variety of responses—intended and
unintended—to school desegregation in Louisiana and is the first to examine fiscal
responses to this important policy.
The particularly detailed data on students, teachers, and school district finances
published by the state department of education make Louisiana an excellent context in
which to assess the effects of desegregation on a variety of outcomes. Further, the state’s
experience with segregation and desegregation was similar to that of other states in the Deep
South. Despite the potential for white flight, erosion of the local tax base, and white
dissatisfaction with desegregated schools to undermine the policy’s effectiveness,
desegregation—imposed by the federal government and courts in a state that was especially
resistant—was quite successful at improving resources in blacks’ schools.
Within the leadership of the National Association for the Advancement of Colored People (NAACP) and
later the NAACP Legal Defense and Education Fund (LDF), there was significant debate starting in the 1930s
about whether to pursue equality for black schools under the “separate but equal” doctrine or by attacking
segregation itself. By 1950, the LDF pursued the latter strategy; the legal argument in Brown proposed that the
stigma associated with state-sanctioned segregation harmed black children. Practical issues in enforcing
“separate but equal” were also at play: In a 1952 discussion of what legal strategy to pursue, Spotts Robinson
argued, “We had four cases…in each of the four cases…we had to go back to court because the injunctions—
equalization injunctions, mind you—were not abided by…Contempt was not the answer…To get the problem
solved we must go back and this time seek further relief which involves a direct attack on segregation itself”
(quoted in Greenberg, 1994). See Greenberg (1994) for an extensive discussion of the LDF legal strategy
before and after Brown.
1
1
A combination of federal policies, including the Civil Rights Act and court-ordered
busing, forced Southern school districts to dramatically increase the extent to which black
and white students attended the same schools and to do so very quickly between 1965 and
1970. These policies had two key goals: to increase interracial contact in schools and to
reduce black-white school quality gaps. But whites resisted these changes, and behavioral
responses by a variety of actors could have undermined these goals. First, white families might
have responded by leaving public schools, migrating from blacker to whiter public school
districts, or reducing their willingness to support public schools at the ballot box. Local and
state governments might also have responded by altering funding for public schools: On the
one hand, unhappiness with desegregated schools and white flight could have led to
reductions in funding; on the other hand, funding might have been increased so that whites
would not see school quality fall as black and white schools combined. Finally, prices—in
particular, property values—might have adjusted in response to the new policy.
Indeed, I show that whites did express dissatisfaction with desegregation by moving
from blacker to whiter public school districts and enrolling in private schools. Surprisingly
though, the local property tax base and local revenue raised were not adversely affected by
desegregation. Reductions in enrollment translated to increases in the per-pupil tax base in
districts more affected by desegregation. Finally, an influx of new state funding was directed
disproportionately to districts where whites were particularly affected, allowing funding in
integrated schools to be “leveled up” to the level previously experienced only in the white
schools. White flight likely would have been even more extensive without these
compensating increases in state aid. Overall, desegregation policy was quite effective at
reducing segregation and increasing resources devoted to blacks’ schools, despite the
potential for behavioral responses of white families and governments to produce unintended
2
consequences. In other work (Reber, 2007), I show that the increased funding for blacks’
schools that accompanied desegregation increased educational attainment for blacks.2
Previous work has shown that—despite substantial narrowing of black-white gaps in
teacher salaries, class sizes and overall expenditure—black-white inequality in Southern
schools remained entrenched at mid-century (Margo, 1990; Card and Krueger, 1992; Bond,
1934). The existing literature on school finance picks up in the 1970s, focusing on the school
finance equalization movements of the 1970s and later. The period of desegregation,
between the end of Margo’s study (ending in 1950) and the school finance equalization
studies has gone unexplored, largely because school finance data for this period was
unavailable. More than 50 years after Brown, many researchers and policymakers continue to
voice concerns about the quality of blacks’ schools.3 But the existing literature has not
addressed the question of whether desegregation ever helped improve resources for blacks’
schools at all. This study helps fill this gap, showing that desegregation was effective in
increasing funding for blacks’ schools by eliminating black-white school quality gaps in
Louisiana.
This analysis also has implications for the school-finance literature more generally.
Although the increasing role of state governments in school finance is widely acknowledged,
much of the existing literature on school finance employs a Tiebout-style local-finance
framework. This paper shows that there was sorting in response to desegregation (white
flight), but this did not translate to adverse affects on the finances of school districts. Rather,
the response of state funding was a more important consideration. Most work that does
examine the role of state funding focuses on state school finance “reforms” or law changes
2 Guryan (2004) also found beneficial effects of court-ordered desegregation plans on black high school
dropout rates; this paper suggests that improvements in school quality for blacks as a possible mechanism.
3 See, for example, Cashin (2004); Cottrol, Diamond, and Ware (2003); Ogletree (2004); Bell (2004); and
Clotfelter (2004).
3
(e.g. Card and Payne (2002) and Murray, Evans, and Schwab (1998)). I show that significant
changes in the distribution of state funding across districts can occur without a school
finance “reform,” as the state government was able to redirect substantial funding by subtly
changing the parameters of the existing school finance formula rather than with a major
revision of its approach.
The paper proceeds as follows: Section II provides a framework for thinking about
responses to desegregation and relevant background. Section III discusses the empirical
methods; the data are discussed in Section IV. The results are reported in Section V, and
Section VI concludes.
II. Framework and Background
A. Framework
Desegregation was a large-scale policy likely to have induced responses in a variety of
actors. Rather than develop a general model incorporating all of these actors, I simply
describe how they might be expected to respond to desegregation qualitatively.4 In
particular, I focus on the potential for unintended responses to undermine the primary goals
of the policy: increasing interaction of blacks and whites in schools and improving the
resources available in schools attended by blacks.
First, white families might respond to desegregation by enrolling their children in
private schools or moving from blacker to whiter school districts (thereby reducing—though
not eliminating—their children’s exposure to blacks). The possibility of “white flight” has
been studied extensively in the existing literature; in previous work (Reber, 2005), I found
that in a national sample of large school districts, white enrollment did decline in response to
implementation of court-ordered desegregation plans, though not enough to fully offset the
A Tiebout-style model of local public goods is often used in studies of school finances; however, local finance
was relatively unimportant compared to state finance in this context, so I do not pursue such a model here.
4
4
increase in exposure of blacks to whites. Although this is not the primary contribution of
this paper, I document patterns of white enrollment responses to desegregation in Louisiana.
In addition to “voting with their feet,” whites might change their literal votes or
pressure local and state governments to change funding or other policies in response to
desegregation. In this paper, I focus on how state and local funding for schools changed.
How funding is expected to respond to desegregation depends on the political economy of
school finance. On the one hand, desegregation likely reduced the value of the public
schools (net of taxes paid) to the median voter (who was typically still white during this
period). In a standard Tiebout-style framework, this would reduce property values and
therefore the local property tax base so that a higher tax rate would be required to raise the
same revenue locally. Perhaps more important, the composition of whites living in the
district might change if families with the highest taste for education flee for alternative public
or private schools. Both of these forces would tend to decrease local funding for
desegregated schools. The potential for desegregation to adversely affect school expenditure
through this channel was mitigated by the relatively small share of funding from local
sources: The median local share of revenue was 21 in 1965.
On the other hand, in the early 1960s, segregated black schools received less funding
than their white counterparts (this is documented in detail below). This meant that without
additional funding for schools, the average funding in whites’ schools would fall as they
combined with black schools. To the extent that large numbers of whites remained in the
public schools, this would have put pressure on state and local governments to increase
funding for schools in order to prevent whites from experiencing a drop in school quality. In
fact, it was hoped that this mechanism would operate to improve school funding for blacks.
5
Finally, desegregation changed the peer composition of schools, exposing whites to
more blacks, who were also lower-income on average. So, depending on whether peers and
funding are substitutes or complements in the production of education, voters’ demand for
funding could have gone up or down.
Starting in 1966, the new federal Elementary and Secondary Education Act (ESEA)
program provided grants to local school districts based on the number of poor kids in
residence. This program provided substantial funding state-wide and particularly in blacker
districts (due to the correlation between race and poverty). The co-incident introduction of
this program complicates the analysis of funding from other sources somewhat. While a
Neoclassical model of local public goods would predict that local governments would spend
out of federal grants as they would any other income (so reducing their own revenues and
increasing school spending by only a fraction of the grant),5 a substantial literature
documents so-called “flypaper effects” for federal education aid to lower levels of
government: Local governments tend not to offset increases in federal aid as much as the
Neoclassical model predicts (see Hines and Thaler, 1995, for a review) . In the analysis
below, I control for the share of students eligible for Title I to separate the effects of racial
composition from the effects of new funding under this program.
Overall, how school fundingis expected to have changed as a result of school
desegregation is ambiguous and depends on whether the unintended behavioral responses of
white families and state and local governments were significant enough to offset the
intended responses. After laying out the historical context in which these changes took place,
5 According to provisions of ESEA, school districts and the state were technically prohibited from reducing
their own spending on schools in response to receipt of funds under the program, and their were some
“maintenance of effort” requirements in the law. Such requirements are difficult to enforce (especially when
spending is trending up, since local governments could simply reduce the rate at which they increase their own
spending, offsetting the Federal grant over time.)
6
I analyze these responses to desegregation empirically. I examine the effects of desegregation
on a number of outcomes, including measures of segregation, white flight to public and
private schools, the local property tax base, per-pupil funding from different sources, current
expenditure, and class size. This paper is the first of which I am aware to study the effects of
desegregation on school district finances finance.
B. School Segregation Policy and Trends
Before Brown, Southern schools were completely segregated by race. Post-Brown,
New Orleans was the first Louisiana school district to desegregate, under court order in the
early 1960s. Most other districts followed, often under court order, in the second half of the
1960s. Figure 1 shows trends in school segregation for Louisiana, which were similar to
trends in other states in the Deep South. Segregation remained high more than a decade
after Brown: in 1964, less than 5 percent of Louisiana districts had any desegregated schools
and the average black was in a school that was only about 1.5 percent white. Following
passage of the 1964 Civil Rights Act (CRA) and the 1965 Elementary and Secondary
Education Act (ESEA)—which together gave resistant Southern school districts a financial
incentive to desegregate (Cascio, et al., 2008), most districts began the process of
desegregating. By 1966, 80 percent of districts had at least some desegregation; and between
1968 and 1970, segregation fell substantially as districts across the state implemented more
substantial plans following the Supreme Court’s decision in Green.6 The Figure points to the
rapid desegregation between 1965 and 1970; the empirical analysis will focus on analyzing
changes in outcomes during this period.
6 In Green vs. New Kent County (1968, 391 U.S. 430), the Supreme Court found that “freedom of choice” plans
did not produce sufficient desegregation. Following this decision, more desegregation plans with mandatory
reassignments, sometimes facilitated by busing, began to be required by the courts.
7
III. Methods
To identify the effects of school desegregation on the outcomes of interest, I need a
source of variation in desegregation policy that is not related to other factors that might
influence changes in enrollment, the local property tax base, revenue, and educational inputs.
Desegregation activity throughout the state was concentrated in the late 1960s, so identifying
effects from variation in the timing of desegregation is not possible. Instead, I use variation
in the intensity of desegregation due to initial conditions. In particular, whites in districts with
higher black enrollment shares were more affected by the large-scale desegregation that
occurred in all Louisiana school districts between 1965 and 1970, so I will examine how
changes in outcomes varied with initial black enrollment share, controlling for other preexisting characteristics. In the remainder of this section, I first discuss why the initial black
enrollment share determined the intensity of desegregation policy before describing the
simple models I will estimate.
A. Black Enrollment Share and the Intensity of Desegregation Treatment
From the perspective of whites, who continued to control Southern school boards in
the late 1960s, if the black share of enrollment in the district was higher, the effects of
desegregation policy were larger on two margins: All else equal, when schools desegregated,
both white exposure to blacks and whites’ class sizes would increase more in blacker
districts.
As school districts mixed blacks and whites in schools, whites in blacker districts
would see a larger increase in exposure to blacks. Figure 2 shows this graphically, relating the
1961 black enrollment share to the change in white exposure to blacks from 1964 to 1970;7
recall that white exposure to blacks was near zero for all districts in 1964. If all districts
7
Data for 1965 are not available, so I report the change from 1964.
8
integrated fully and their black share of enrollment did not change between 1961 and 1970,
the points would lie on the 45-degree line. Most are below the 45-degree line suggesting
integration was incomplete or there was more white flight from blacker districts, but the
black share of enrollment before desegregation is a strong predictor of whites’ subsequent
exposure to blacks.8
Combining black and white schools also meant that black-white gaps in financing
and class size would have to be closed. Due to the system of school finance operating in
Louisiana and throughout the South during the first half of the twentieth century, such gaps
were larger in blacker districts. This meant that, after desegregation, whites in blacker
districts would see larger declines in funding and increases in class size in their schools,
compared to their white counterparts in whiter school districts.
Bond (1934) and Margo (1990) examine the system of school finance prevailing in
the first half of the 20th century in the South in great detail, showing the importance of black
enrollment share in determining black-white school funding gaps. Because the historical
context was important in establishing the initial conditions that made desegregation more
intense in blacker districts, I briefly outline Margo’s and Bond’s argument and show that the
system Margo documents through 1950 still operated in Louisiana in the mid-1960s.
Following Reconstruction, funding for schools in the South was segregated by race,
with taxes from blacks funding the black schools and taxes from whites funding the white
Exposure of whites to blacks in a district will depend on two factors: the black share of enrollment and how
evenly blacks and whites are distributed across schools within the district. One might have expected that it
would be easier (politically and logistically) to spread blacks evenly across schools in districts with lower black
enrollment shares. If this were the case, whites in high black enrollment share districts might not see larger
increases in exposure to blacks as a result of desegregation. The dissimilarity index provides a measure of how
evenly blacks and whites are distributed across schools, given the black enrollment share (it can be interpreted
as the share of black students who would have to switch to another school to achieve a racial balance in every
school that is the same as the racial balance of the district). While there is variation in the change in the
dissimilarity index across districts, it is uncorrelated with initial black enrollment share. High black enrollment
share districts do not appear to have more or less intensive desegregation plans.
8
9
schools.9 This segregation of funding (but not of students) was disallowed by the courts over
time. These decisions, together with the Supreme Court’s “separate but equal” ruling, made
it difficult for state governments to discriminate in the allocation of funding to counties and
local school districts,10 and Southern state governments typically allocated revenue to
counties on a per-pupil basis, without regard to race. However, local school boards did
discriminate, disproportionately directing funding to the white schools. This meant that
whites in districts with many blacks could finance better schools without raising much
revenue locally by diverting a portion of state funding for black students to white schools;
this produced a strong positive relationship between the black-white funding gap and black
enrollment share.
During the 50 years leading up to Brown, these black-white gaps did narrow
substantially, from quite high levels (Margo, 1990; Card and Krueger, 1992). Still, even in
1959, the average Louisiana district spent 72 cents per pupil on instruction in the black
schools for every dollar spent in the white schools. After 1959, Louisiana did not report
expenditure separately by race, but black-white gaps in class size averaged about six more
students per teacher in the first half of the 1960s. As Bond and Margo showed for the earlier
part of the century, the black-white class-size gap was larger in blacker districts.
Figure 3A confirms that in the early 1960s, whites in blacker districts enjoyed
significantly lower class sizes, compared to whites in whiter districts, while black class size
was variable but unrelated to black enrollment share (Figure 3B). When black and white
schools merged, holding the number of teachers constant, black and white students would
Also see Kousser (1991) for a discussion of the history of school segregation and finance in Louisiana before
1900.
10 See Kousser (1980) for a discussion of the Court’s ruling against separate funding for black and white
schools in Kentucky. The response parallels the response to desegregation in Louisiana 80 years later. When
funds could no longer be segregated by race, Kentucky voters chose to increase spending for black schools
rather than decrease spending for white schools, approving an increase in statewide taxes that tripled state-level
expenditures for blacks (Kousser, 1980).
9
10
experience the average class size in the district. Thus the difference between white class size
and the average class size is a measure of the effect of desegregation on whites, were total
funding to remain constant, through the school resources channel.11 The final panel of
Figure 3 shows how this gap relates to initial black enrollment share. The slope of the
regression line is -7.9. The gap between the white student-teacher ratio and the average
student-teacher ratio was expected to be about 3.6 students larger for a district at the 90th
percentile of the black enrollment share distribution (64 percent black), compared to a
district at the 10th percentile (19 percent black). Upon desegregation, this is how much more
the blacker district would need to reduce average student-teacher ratios in the district if
whites were to avoid seeing their class sizes go up.
These graphs show results for 1961, but the same pattern continues to hold through
1965. In results not reported, I confirm that in the early 1960s, per-pupil revenue from state
and federal sources was unrelated to black enrollment share. Blacker districts raised less
revenue locally, so total revenue per pupil was negatively related to fraction black. Taken
together, these results show that the early 20th-century Jim Crow system of school finance
continued to operate in Louisiana in the early 1960s.
The fact that gaps between white school funding and black school funding were
larger in high black enrollment share districts, meant that all else equal, school funding for
whites in blacker districts would fall more upon desegregation. In addition, whites’ exposure
to blacks increased more in high black enrollment share districts after desegregation. As
discussed in Section II, these changes might have induced a variety of responses from
whites, local school districts, and the state government. The key point for the empirical
analysis is that such responses are expected to be larger in blacker districts because
Student-teacher ratio is just one measure of school quality gaps. Other race-specific measures are not
available for this period, but such measures would arguably be more unequal due to their lower visibility.
11
11
desegregation had a larger effect on whites (who also continued to control local
governments) in those districts.
B. Estimation
As just discussed, the intensity of desegregation policy depended on initial black
enrollment share. To estimate the effects of desegregation on outcomes, I first estimate the
simple correlation of changes in the dependent variables of interest around the time of
desegregation (1965 to 1970) with the 1961 black share of enrollment—early enough so that
it was plausibly unaffected by desegregation:
(1)
Δy i  β 0  β1 FractionBlack61i  ε i ,
where y is the change in an outcome, such as per-pupil revenue, from 1965 to 1970,
FractionBlack61 is the black share of public enrollment in 1961, and  is an error term.
These regressions establish descriptively whether changes in key variables differed by
black enrollment share. Of course, it is possible that blacker districts would have experienced
different changes in enrollment, the property tax base, financing and class size during this
period for reasons unrelated to the intensity of desegregation. To mitigate this concern, I add
controls to equation (1) for observable pre-existing characteristics of districts that could be
related to trends in the outcomes of interest. The new federal Title I program targeted
significant new funding to poor districts; to account for changes in outcomes related to the
introduction of this program, I control for the share of enrollment eligible for Title I.12 This
is particularly important for the revenue variables, as districts with more Title I eligibles per
pupil will see larger increases in federal, and therefore total, funding. I also include 1961 perpupil current expenditure and demographic and socio-economic indicators from the 1960
Census: per-capita income, the percent of households with complete plumbing, percent of
I use the number of Title I eligibles in 1966, which was determined based primarily on 1960 Census data, so
FractionTIEligible1966 was pre-determined.
12
12
households with income below $3,000, the (log of) total population, and the percent of the
population living in urban areas:
(2)
Δy i  β 0  β1 FractionBlack61i  β 2 FractionTIEligible i  β 3 InitialExpenditurei  X i θ  ε i .
This list of observables is far from exhaustive, but it is reassuring that most of the
coefficients on black enrollment share are not substantially affected by their inclusion.
If differential changes in outcomes by black enrollment share are indeed due to
differing intensities of desegregation policy for high and low black enrollment share districts,
those changes should be concentrated during the period of desegregation. I therefore pursue
two related approaches to examine the timing of changes and show that the coefficients on
FractionBlack61 equation (2) are not simply picking up longer-term differential trends for
blacker and whiter districts. I describe these specification checks after presenting the main
results below.
IV. Data
Previous researchers have not been able to examine the effects of desegregation on
district finances in part because of a lack of data for this period. To fill this gap, I compiled a
new annual panel dataset from administrative reports published by the Louisiana
Department of Education. These reports contain detailed data on students, teachers, and
finances of Louisiana’s 64 Parish and 2 City districts13 for 1955 to 1975.14 Annual data on the
following variables were taken from this source: number of classroom teachers by race,
public and private student enrollment by race, revenue by source, and current expenditure. I
also collected detailed data on program costs by category and required contributions under
the state minimum foundation program for key years, including state aid for transportation,
from this source. These data can be used to calculate student-teacher ratios by race for the
Before 1967, there were 3 City districts; Lake Charles City consolidated with its Parish.
A few years are missing or incomplete, but the largest gap is a single year. Missing data generally resulted
from difficulty in acquiring the relevant report for that year.
13
14
13
early 1960s. The Southern Education Reporting Service (SERS) reports that students and
teachers were completely segregated by race through 1965, so the student-teacher ratio for
blacks is the number of black students divided by the number of black teachers, and similarly
for whites. After students and teachers of different races are mixed, I can no longer calculate
the student-teacher ratio separately using these data.
Data for the two City districts was combined with that of their respective Parishes
(counties) and matched to data on total assessed valuation (the local property tax base), as
well as the value of real property and non-real property from Biennial Reports of the
Louisiana Tax Commission for 1955-1975.15 I link these new datasets with demographic
data—including population, the fraction of the population living in urban areas, and
measures of income and poverty for each Parish from the 1960 Census. Finally, I use data
on the number of students eligible for Title I funding in 1966 from a Congressional Report
(U.S. Senate, 1967). A district’s Title I eligible count was determined based primarily on data
from the 1960 Census, so this can be considered a “pre-program” variable.
Segregation measures—black exposure to whites, white exposure to blacks, and
whether the district had any desegregation—were calculated using school-level data from the
Office of Civil Rights (OCR) Elementary and Secondary School Survey data for 1967 to
1976. For earlier years, these are estimated based on data on the share of blacks in school
with any whites reported by the Southern Education Reporting Service.16
The City districts cannot be considered separately because the Tax Commission reports assessed valuation at
the Parish level. Excluding these from the analysis does not affect any of the results.
16 SERS reported the number of black students in school with any whites. When the share of blacks in school
with whites is small, black exposure to whites is approximately equal to (perblanywh)  (1-perbl)/[(1perbl)+(perbl  perblanywh)], and white exposure to blacks is approximately (perbl  perblanywh)/[(1perbl)+(perbl  perblanywh)]; where perblanywh is the percent of blacks in school with whites and perbl is the
percent of district enrollment that is black.
15
14
Summary statistics for the dependent variables are reported for 1965, 1970, and the
change from 1965 to 1970 in Table 1A. Table 1B reports summary statistics for the preexisting characteristic control variables.
V. Results
A. “White Flight”: Public and Private Enrollment Responses
Figure 4 shows state-wide trends in public and private enrollment by race.
Desegregation was associated with a noticeable shift of enrollment from public to private
schools. White public school enrollment, which had been steadily rising, dropped by nearly
30,000 students between 1968 and 1970, the period of most intense desegregation.
Conversely, the downward trend in white private school enrollment was reversed, rising by
26,500 students over the same period—a 25 percent increase.17 These students appear not to
have returned to the public system, as white private school enrollment remained elevated
throughout the 1970s. These trends can also be seen in Table 1A: White public enrollment
fell by 8 log points on average between 1965 and 1970, while the share of white students in
private schools rose by 8 percentage points.
If this decline in public and increase in private enrollment was in fact due to
desegregation, we would expect white flight to be more pronounced and private school
enrollment to rise more in districts where whites were more affected by desegregation—
those with high initial black enrollment shares. Indeed, Table 2 shows that this was the case.
Table 2 reports results of estimating equations (1) and (2) with the change from 1965 to
1970 in several enrollment measures as dependent variables. I report similar specifications
for other outcomes, so I walk through the specifications for the enrollment variables in
Private schools that did not want state accreditation were not required to report their enrollments and are not
included in these figures. The level of private school enrollment is therefore underestimated, and the increase
from 1968-1970 may be underestimated to the extent that new, unaccredited schools were opened in response
to desegregation.
17
15
some detail. The first column of each panel reports the results without controls; in the
second specification, controls for the share of students eligible for Title I are added; the
socio-demographic controls from the Census and initial per-pupil current expenditure (1961)
are added in the third specification. Finally, for the enrollment variables only, I include the
change in (the natural log of) lagged births in the Parish. Lagged births is the sum of births
(separately by race where appropriate) for all the cohorts who would be in 1st through 12th
grades during the current school year.18 These data are available for only 52 counties, so the
sample is slightly smaller in these specifications.
The results for the log change in white public enrollment are reported in Panel A.
The coefficients are consistently negative and significant, indicating that, relative to white
districts, blacker districts lost more white enrollment. In the first specification with no
controls, the coefficient on black enrollment share is -1.1; including all controls reduces the
coefficient to -0.78, indicating that a one standard deviation increase in 1961 black
enrollment share (16 percentage points) was associated with an additional reduction in white
public school enrollment of 14 to 18 log points between 1965 and 1970. I also experimented
with including the square of black enrollment share in some specifications to capture nonlinearities in the response of whites; the quadratic term was generally not significant or large,
so I do not report the results.
This differential decline of white enrollment in blacker districts represents white
flight to either private schools in the same district or a whiter public school district. If all the
response was on the private school margin, the coefficient on initial black enrollment share
would be zero with change in total (private plus public) white enrollment as the dependent
variable. Instead, the coefficient is negative and significant in all three specifications,
18
For example, lagged births for 1965 is the sum of births in the years 1948 through 1959.
16
suggesting that not all of the “white flight” was to private schools (Panel B). With all the
controls, the coefficient on black enrollment share is –0.33, compared to –0.78 for the
change in public enrollment; this suggests that increases in private school enrollment
accounted for about 60 percent of the differential decline in public school enrollment in
blacker districts.
Panel C indicates that the change in the share of white students enrolled in private
schools also increased more in blacker districts; a one standard deviation increase in initial
black enrollment share is associated with an additional 4.4 to 7.2 percentage point increase in
the fraction of students in private schools. The raw data indicate that ten percent of districts
had increases in the share of whites in private school of more than 33 percentage points, so
white flight to private schools was quite substantial in some districts.
These results suggest that there was white flight out of districts with high black
enrollment shares to both whiter public school districts and to private schools. On the one
hand, this raises concerns about the potential for declining property values and support for
public schools; on the other hand, lower enrollment mean that districts had fewer students
to support. Panel D shows results with the (log) change in the total (black and white) public
enrollment as the dependent variable. The coefficient on initial black enrollment share is
statistically significant and ranges from -0.35 to -0.61, confirming that the total number of
students supported in the public schools did decline more in high black enrollment share
districts.
B. Local Property Tax Base Responses
The substantial white flight from high black enrollment share districts raises
concerns about declining property tax bases in those districts. Table 3 reports results of
estimating equations (1) and (2) with (log) changes in total assessed valuation overall and
17
separately by real estate (real AV) and other property (non-real AV), 19 as well as total
assessed valuation per pupil. The standard errors are large, but the results do not point to
larger declines in the property tax base in districts more affected by desegregation. The
results for per-pupil AV (Panel B) show that the per-pupil property tax base increased more in
blacker districts, reflecting the relative decline in total enrollment.
Given the clear dissatisfaction with desegregation among whites, why didn’t the
property tax base suffer? One possibility is that residential property values did decline, but
the assessed valuation data are too noisy to pick up the effect. Indeed, the standard errors in
Table 3 are so large that I cannot rule out very large negative or positive effects. Changes in
residential real estate prices are reflected in assessed valuation imperfectly and with a lag;
assessments often do not reflect changes in market prices. Further, residential real estate is
only a portion of total assessed valuation. In 1965, real property—residential and
commercial—averaged about 42 percent of total assessed valuation.20
Even if property values and the local tax base were affected, but the estimates are
too imprecise to pick it up, any effects on total revenue were mitigated by the relatively small
local contribution to total revenue. As in many Southern states at the time, local sources
accounted for a relatively small share of total revenue; the average district raised 23 percent
of its revenue locally in 1965. And as shown below, desegregation did not reduce local
Real assessed valuation includes country and city lots and improvements on country and city lots, agricultural
lands, timberlands, marshland, and manufacturing plant buildings; country and city real estate and
improvements are the largest component of real assessed valuation. Non-real assessed valuation includes public
service corporations, manufacturing plants and equipment, inventories, machinery and equipment, business
furniture and fixtures, watercraft, aircraft, credits (insurance and finance companies), financial institutions,
leasehold improvements, drilling rigs, pipelines, oil and gas equipment, and miscellaneous property.
20 An increase in black demand might also offset a reduction in demand for housing by whites. As shown
below, blacker districts saw larger increases in spending on schools that was not offset by higher taxes,
increasing demand for housing. Because the assessed valuation data are so noisy and local revenue is a small
share of total revenue, I do not explore these possibilities further.
19
18
revenue per pupil. Overall, concerns about desegregation harming local districts’ funding
capacity seem not to have been borne out in Louisiana.
C. Revenue Responses: Per-Pupil Revenue by Source
Figure 5 shows trends in average per-pupil revenue for four mutually exclusive
categories: local (primarily property and sales taxes), state formula aid (the state minimum
foundation program is described in more detail in the appendix), federal Elementary and
Secondary Education Act (ESEA) program funds, and other state and federal revenue. 21
Local revenue and state formula aid began rising around 1963 and 1965, respectively, and
continued to increase in most years through the mid-1970s. The structure of the state
minimum foundation formula did not change during this period, although some of the
parameters of the formula and the total revenue distributed did change. Federal ESEA
revenue jumped substantially following the introduction of the program in 1965 and
remained around that level into the 1970s. The combination of these trends meant that total
revenue began to increase substantially after 1965, around the time districts took their first
steps toward desegregation. In the remainder of this section, I examine whether changes in
revenue were larger or smaller in districts more affected by desegregation.
Table 4 shows the results of estimating equations (1) and (2) for total revenue per
pupil and separately for the categories described above. Again, there is no evidence that
desegregation—and the accompanying white flight—harmed school district finances. If
anything, the opposite was true. Changes in revenue per pupil from every source were larger
in districts more affected by desegregation. Recall that without new spending, whites in more
affected (blacker) districts would have seen their class size rise significantly more as the
21 Federal ESEA includes all revenue received through ESEA programs; the largest component is Title I. Other
non-local includes revenue received through other state, federal or mixed programs and accounts for about 17
percent of non-local revenue in 1965.
19
better-funded white schools combined with black schools. The results in Table 4, discussed
in more detail below, suggest that state authorities—and to a lesser extent local school
districts—responded by raising more revenue.
Panel A shows the results for local revenue per pupil. The coefficients are not
precisely estimated, but are positive across specifications and marginally statistically
significant in only one specification; coefficients on black enrollment share range from 0.20
to 0.49.22 The dependent variable is measured in thousands of dollars, so this implies that
each one standard deviation increase in initial black enrollment share (0.16) was associated
with an additional $32 to $78 increase in per-pupil local revenue between 1965 and 1970.
Panel B shows the results for Federal ESEA revenue per pupil. With no controls, the
coefficient on black enrollment share is 0.67 (significant at the 1 percent level), but, as
expected, this is reduced to a statistically insignificant 0.14 when the share of students
eligible for Title I is controlled. That is, the relationship between black enrollment share and
the change in federal ESEA funding is due mostly to the correlation between race and
poverty and the fact that Title I eligibility was based on low-income status.
Panel C shows that changes in state formula aid were larger in districts more affected
by desegregation. The coefficient on black enrollment share is highly significant and little
affected by controls, ranging from 0.56 to 0.61. While the controls—particularly the control
for the share of students eligible for Title I—knocked out the coefficient on black
enrollment share for federal ESEA funding (as expected), this is not the case for state
formula aid. This suggests that the changes in state formula aid related to race rather than
In results not reported, I examined the major components of local revenue—property taxes, sales taxes, and
other local taxes—separately. The importance of the sales tax as a source of local revenue increased during this
period. In 1965, less than half of Parishes had any sales tax, and the sales tax accounted for only 7.5 percent of
local revenue for the average district. By 1970, more than three-quarters of districts had a sales tax, and the
sales tax accounted for 26.8 percent of local revenue for the average district. Here, the point estimates depend
more on whether controls are included, but it appears that the disproportionate increase in local revenue in
blacker districts came roughly equally from changes in sales taxes and local property taxes.
22
20
poverty; as discussed above, the intensity of desegregation, from the perspective of whites,
also depended on a district’s racial composition. A one standard deviation increase in black
enrollment share was associated with an additional $89 to $100 in state formula aid per pupil.
Panel D shows that other state and federal revenue also increased slightly more in
more affected (blacker) districts, but the coefficient is small and only marginally significant
without controls and insignificant when controls are included.
Finally, Panel F shows the results for total revenue per pupil—the sum of the four
categories already presented. The coefficient on black enrollment share is positive and
statistically significant in all specifications. The coefficient is reduced from 1.6 to 1.1 by the
inclusion of controls—this is due primarily to the fact that controls reduce the black
enrollment share coefficient in the federal ESEA regression. Overall, the differential increase
in total revenue for blacker districts was large. The estimates imply that a district at the 90th
percentile of the black enrollment share distribution (64 percent black) would have expected
to see an increase in total per-pupil revenue that was about $495 more than that experienced
by a district at the 10th percentile (19 percent black), controlling for other factors ((0.640.19) 1.11000), compared to the 1965 average of $3,060. The results below show that
much of the disproportionate increase in funding was used to reduce average class sizes
enough in blacker districts so that whites would not see their class sizes rise.
Controlling for other factors, the majority of additional revenue for blacker, more
affected districts came from state formula aid, so I discuss this category in more detail. What
accounts for this differential increase in state formula aid? Did the state legislature re-allocate
state aid in response to desegregation? The state aid formula did not consider racial
composition—or even the economic status—of districts explicitly, and there were no
identifiable “reforms” to the system of state aid during this period. Further, the legislature
21
did not make a single, identifiable change to the formula to provide more revenue to
predominately black districts. Instead, based on an analysis of each component of the
formula, I find that blacker districts benefited disproportionately from virtually every change
in the calculation of total grants between 1965 and 1970. A little more than half of the
disproportionate increase came from allotments for teachers. See the Appendix for a detailed
analysis of changes in state aid.
It would be useful to know the legislative history of the state aid program during this
period—for example, was desegregation discussed as the parameters of the formula were
determined. Unfortunately, the budget process in Louisiana at this time was secretive, and
records of legislative debates for this period do not exist.
D. Changes in Educational Inputs
Total revenue per pupil increased substantially more in districts that were more
affected by desegregation, but did this new funding improve the quality of education for
blacks? Data on exactly where the new money went are limited, but the results in Table 5
suggest that the additional revenue went to current expenditure, especially to reduce class
size, and somewhat for additional transportation costs. Panel A shows the results with the
change in current expenditure per pupil from 1965 to 1970 as the dependent variable; with
the full set of controls, the coefficient on black enrollment share is 1.1, the same as the
coefficient for the total revenue per pupil regressions, confirming that, not surprisingly, the
new revenue was spent on current operations.
To the extent that busing was required to integrate schools, transportation costs may
have risen more in more affected districts. Panel B shows the results with the change in per-
22
pupil state transportation aid as the dependent variable.23 The coefficients are statistically
significant and stable across specifications; with all controls, the coefficient is 0.17,
accounting for about 15 percent of the increase in current expenditure.
Panel C shows the results with the change in the teacher-student ratio as the
dependent variable. (I use the teacher-student rather than student-teacher ratio so that the
coefficient can be scaled by teacher salaries and compared to per-pupil expenditures.) The
coefficient is positive and highly significant, indicating that blacker districts increased the
ratio of teachers to students more. The average teacher salary according to the salary
schedule was about $40,000 in 1970 (in 2007 dollars). Translated into expenditure per pupil,
the coefficient would be about 0.45 (1.1340/100), so additional spending on teachers
explains about 40 percent of the differential increase in current expenditure.
Recall that if additional resources had not been disproportionately made available to
heavily black districts, whites in those districts would have seen their class sizes rise (teacherstudent ratios fall), relative to whites in less black districts due to the large gap between the
white class size and the average (black and white) class size. Regressing the gap between
white and average teacher-student ratios in 1965 on initial black enrollment share and the
other controls, the coefficient on black enrollment share is -1.10 (statistically significant at
the 1 percent level, not reported). The fact that the coefficient on black enrollment share
with the change in teacher-student ratios from 1965 to 1970 as the dependent variable is
similar in magnitude but opposite in sign (Panel C), suggests that the differential increase in
teacher-student ratios was enough to bring the average ratio up to the level previously
experienced only by whites.
23 This is not necessarily the entire transportation budget of the district but covers bus driver salaries for a
number of drivers determined according to the state aid formula by the number of total miles students were
transported.
23
It is interesting to note that before desegregation, there was significant inequality
between districts in spending on whites, with whites in high black enrollment share districts
enjoying smaller classes. After desegregation, this within-race, between-district inequality
remained, but students in high black enrollment share districts of both races had smaller class
sizes.24
E. Specification Checks
The results above show that between 1965 and 1970, more affected, blacker districts
saw larger declines in white public enrollment and increases in white private enrollment, as
well as larger increases in per-pupil revenue—especially from state formula aid, instructional
spending, and teachers per student. I argue that these disproportionate changes were due to
greater intensity of desegregation—from the perspective of white decision-makers—in
blacker districts. If this is the case, the observed differential changes should be concentrated
during the period of desegregation. I pursue two related approaches to pinpoint the timing
of these changes and show that they were indeed concentrated between 1965 and 1970.
First, I estimate versions of equation (2) with changes in the key outcome variables
before desegregation as dependent variables (“placebo” outcomes):
(3)
Δy i  η0  η1 FractionBlack57i  η 2 FractionTIEligible1966i  X i φ  ε i ,
24As
mentioned above, Card and Krueger (1992), Margo (1990), and others have shown that black-white
inequality of school quality in the South had narrowed substantially before desegregation. Consistent with Card
and Krueger’s work, the data from the Annual Financial and Statistical Reports show that class sizes for both
races were declining rapidly in the 1950s, and the black-white gap was closing somewhat. The decline for blacks
accelerated around 1964, when there was still virtually no desegregation. Anecdotal accounts suggest that
before Green (1968), whites hoped that improving resources for blacks might fend off more stringent
desegregation, which may account for the acceleration in 1964. In addition, ESEA programs were most likely
more often implemented in black schools, as black students had higher poverty rates. Class sizes for whites, on
the other hand, declined relatively steadily between 1955 and 1965; and assuming whites had a class size near
the average class size after 1970, the decline continued unperturbed through 1975. Over the longer term, other
forces were probably more important than desegregation in reducing black-white school quality gaps (although
the threat of desegregation may have contributed). Still, the period of desegregation in Louisiana does appear to
have finally ended the Jim-Crow era system of school finance in the state.
24
where y is the change in outcome from 1957 to 196325 and the black share of enrollment is
measured in 1957. Due to data limitations, I exclude lagged changes in the outcome; the
other controls are the same as in equation (2). 1 is expected to be small and insignificant.
The results of these “placebo” regressions are reported in Table 6. The coefficient
on 1957 fraction black is never statistically significant or large in magnitude. This suggests
that the results above do not reflect the continuation of a pre-existing trend.
To trace out changes in the relationship between outcome variables and black
enrollment share year-by-year, I estimate regressions of the outcomes on 1961 fraction black
and the same control variables for most years between 1955 and 1975 (data are missing for a
few years). The trend in the coefficient on fraction black over time shows the differential
trend in the outcome for higher and lower black enrollment share districts, controlling for
other factors. I plot the coefficients on fraction black from these regressions with 95 percent
confidence intervals for the key outcomes in Figures 6 to 8; because I am interested in how
the coefficient on fraction black changed over time, rather than its level, I normalize so that
the coefficient is zero in 1965.
Figure 6 shows the differential trend in the log of white enrollment and the share of
whites in private schools. These figures clearly demonstrate that the differential reduction in
white public and increase in white private enrollment were concentrated during the period of
desegregation, especially after 1968.
Figure 7 shows differential trends in per-pupil revenue by source. Although there
was a slight upward trend before and after the period of desegregation, the differential
increase in total revenue per pupil was particularly sharp during the period of desegregation.
Recall that, controlling for other factors, state formula aid accounted for most of the
25
I use 1957 to 1963 because it is the latest 5-year change before 1965 for which I have data available.
25
differential increase in total revenue between 1965 and 1970; Panel B shows that this does
not simply reflect a gradual trend; the differential increase between 1968 and 1970 is
particularly sharp. The relationship between local revenue per pupil and initial black
enrollment share was noisier, especially after 1965, but also suggests that funding from this
source increased more in blacker districts during the period of desegregation (Panel C).
ESEA revenue increased more in blacker districts between 1965 and 1966, but this effect
disappears by 1970—this is due to the control for the share of Title I eligibles (Panel D).
(The graph for other non-local revenue is omitted to save space.)
Finally, Figure 8 shows that differential increases in current expenditure per pupil
and teachers per student were also particularly sharp between 1965 and 1970.
F. Alternative Explanations
Alternative explanations for the disproportionate increase in funding for higher black
enrollment share districts include changing preferences for spending on black students
relative to whites and increased voting by blacks following the Voting Rights Act of 1965.
Given the persistent resistance to desegregation and other Civil Rights policies in the state, it
seems unlikely that changing attitudes explain the observed changes in funding, although I
cannot rule it out.
Increased voting among blacks is the most plausible alternative explanation: As
blacks gained political power, they may have been able to direct more state funding to
heavily black districts. Adding controls for the change from 1964 to 1968 in the share of the
black voting-age population registered to vote in each county, collected by Matthews and
Prothro (1963a, 1963b), does not alter the coefficients on black enrollment share much.
This is an admittedly imperfect measure of the change in black voting or political influence
overall. However, even considering that voting among blacks did increase substantially
26
during this period, it is unclear whether blacks would have gained enough influence in the
state legislature to bring additional revenue to their districts. Still, it is possible that it was the
combination of Civil Rights and War on Poverty policies—desegregation, the Civil Rights Act,
the Voting Rights Act, and the Elementary and Secondary Education Act—that produced
these changes in Louisiana’s school finance system.
V. Conclusions
Equalizing school quality has been an important goal of policymakers, and it was
hoped that desegregation would improve the resources available in blacks’ schools by tying
their fate to that of whites. While a significant literature examines the effects of
desegregation on white flight and other outcomes, I know of no study that examines the
fiscal effects of desegregation. In Louisiana, the Jim Crow system produced substantial
black-white gaps in school resources—which were especially large in blacker districts—
throughout the first half of the 20th century. This system finally unraveled in the course of a
few years as new funding was made available and average class size fell so that whites would
not see their class sizes grow as black and white schools combined.
Districts that were more affected by desegregation did experience more white flight
to both whiter public school districts and private schools, but the property tax base was not
adversely affected. Despite the potential for unintended responses on the part of whites—
voting with their feet or expressing their disapproval at the ballot box—to undermine the
ability of desegregation policy to increase resources available in the schools blacks attended,
desegregation instead led to large increases in per-pupil funding. The analysis suggests that
this new funding was used to level up the quality of schools for both races (now attending
the same schools) to that previously experienced only by whites.
27
The analysis points to a general increase in demand for resources in local school
districts, especially blacker districts, to which the legislature responded, if not transparently.
Given the large increases in class size—after years of decline—that whites in heavily black
districts would have seen in integrated schools, it is plausible that the state legislature would
be persuaded to increase aid to such districts, and the analysis—particularly the fact that
these changes were closely timed with desegregation—suggests that the disproportionate
increase in revenue from state formula aid were likely caused by desegregation.
Although there was no state school finance “reform,” there was substantial
redistribution of aid under the existing formula. Most of the literature on state school
finance has focused on school finance “reforms” or major changes in the law. The
experience of Louisiana during this period suggests that significant re-allocation of funding
across districts can occur more quietly, without a major reform.
28
References
Alexander, S. K. (1967). Public School Finance Programs, 1966-1967. Washington, D.C.,
U.S. Department of Health, Education, and Welfare.
Bell, D. A. (2004). Silent Covenants: Brown v. Board of Education and the Unfulfilled
Hopes for Racial Reform. Oxford University Press.
Bond, Horace Mann. (1934). The Education of the Negro in the American Social Order.
New York, Prentice-Hall.
Boozer, M. A., A. B. Krueger, et al. (1992). “Race and School Quality since Brown v. Board
of Education.” Brookings Papers on Economic Activity, Microeconomics: 269-326.
Card, D. and A. B. Krueger (1992). “School Quality and Black-White Relative Earnings: A
Direct Assessment.” Quarterly Journal of Economics: 107(1): 151-200.
Card, D. and A. Abigail Payne (2002). "School Finance Reform, The Distribution Of School
Spending, And The Distribution Of Student Test Scores." Journal of Public
Economics: 83(1): 49-82.
Cascio, Elizabeth, Nora Gordon, Ethan Lewis, and Sarah Reber (2008). “From Brown to
Busing.” Journal of Urban Economics: 64 (2): 296-325.
Cashin, S. (2004). The Failures of Integration: How Race and Class Are Undermining the
American Dream. New York. Public Affairs.
Clotfelter, C. T. (2004). After Brown: The Rise and Retreat of School Desegregation.
Princeton. Princeton University Press.
Clotfelter, C. T. (1975). “The Effect of School Desegregation on Housing Prices,” Review of
Economics and Statistics. 57: 446-451.
Coleman, J. S. (1966). Equality of Educational Opportunity Summary Report. Washington,
D.C., U.S. Dept. of Health, Education, and Welfare, Office of Education.
Cottol, R. J., R. T. Diamond, L. B. Ware (2003). Brown vs. Board of Education: Caste,
Culture, and the Constitution. Lawrence, Kansas. University of Kansas Press.
Greenberg, J. (1994). Crusaders in the Courts: How a Dedicated Band of Lawyers Fought
for the Civil Rights Revolution. New York. Basic Books.
Guryan, J. (2004). “Desegregation and Black Dropout Rates,” American Economic Review,
September 94(4): 919-943.
Hines, J. R. and R. H. Thaler (1995). “The Flypaper Effect.” Journal of Economic
Perspectives 9(4): 217-26.
29
Johns, T. L. (1969). Public School Finance Programs, 1968-69. Washington, D.C., U.S.
Department of Health, Education, and Welfare.
Johns, T. L. (1972). Public School Finance Programs, 1971-72. Washington, D.C,
Department of Health, Education, and Welfare.
Kousser, J. M. (1980). “Making Separate Equal: The Integration of Black and White School
Funds in Kentucky,” Journal of Interdisciplinary History. 10(3): 399-428.
Kousser, J. M. (1991). “Before Plessy, Before Brown: The Development of the Law of
Racial Integration in Louisiana and Kansas,” in Paul Finkelman and Stephen C.
Gottlieb, eds., Toward A Usable Past: Liberty Under State Governments; Athens,
GA: University of Georgia Press: 213-70.
Lynch, Bill. “Legislature: Study in Contrasts.” New Orleans States-Item. June 14, 1969.
Margo, R. A. (1990). Race and Schooling in the South, 1880-1950: An Economic History.
Chicago and London, University of Chicago Press.
Matthews, D. R. and J. W. Prothro (1963a). “Social and Economic Factors and Negro
Registration in the South,” American Political Science Review. 57(1): 24-44.
Matthews, D. R. and J. W. Prothro (1963b). “Political Factors and Negro Voter Registration
in the South,” American Political Science Review. 57(2): 355-367.
Munse, A. A. (1965). State Programs for Public School Support. Washington, D.C., U.S.
Department of Health, Education, and Welfare, Office of Education.
Murray, S. E., W. N. Evans, and R. M. Schwab (1998). “Education-Finance Reform and the
Distribution of Education Resources,” The American Economic Review: 88(4): 789812.
Ogletree, C.J. (2004). All Deliberate Speed. New York and London: W. W. Norton &
Company.
Orfield, G. (1975). Congressional Power: Congress and Social Change. Harcourt Brace
Jovanavich.
Reber, S. J. (2005). “Court-Ordered Desegregation: Successes and Failures in Integration
since Brown vs. Board of Education.” Journal of Human Resources, Summer.
Reber, S. J. (2007). “Desegregation and Educational Attainment for Blacks,” NBER
Working Paper 13193.
Ryan, J. E. (1999). “Schools, Race, and Money.” Yale Law Journal 109: 249-316.
Shipp, D. E. and C. W. Hilton (1962). School Finance in Louisiana: A Handbook on Local,
30
State, and Federal School Finance in Louisiana. Baton Rouge, Louisiana Education
Research Association.
Shipp, D. E., C. W. Hilton, et al. (1966). School Finance in Louisiana: A Handbook on
Local, State, and Federal School Finance in Louisiana. Baton Rouge, Louisiana
Education Research Association.
Southern Education Reporting Service (1967). Statistical Summary, State by State, of School
Segregation-Desegregation in the Southern and Border Area from 1954 to the
Present. Nashville, Tennessee, SERS.
U.S. Senate. Committee on Labor and Public Welfare. Subcommittee on Education. Notes
and Working Papers Concerning the Administration of Programs Authorized Under Title I of Public
Law 89-10, The Elementary and Secondary Education Act of 1965 As Amended by Public Law 89750. Washington, DC: U.S. Government Printing Office, May 1967.
31
Appendix: Detailed Analysis of Public School Fund Changes
A district’s allocation from the MFP (Minimum Foundation Program) was its
“program cost” less its required support, both of which were calculated based on formulas:
Program Cost = (Instructional Personnel Allotted)(Avg. Salary per Salary Schedule) +
(Transportation Funding per Schedule) +
(Flat Per-ADM Payment)  (Avg. Daily Membership) +
(Adjustment for Declining Enrollment)
Required Support = (5 Mill Constitutional Tax)  (Assessed Valuation) +
(Severance Tax Collections) +
(50% of Revenue from Rent or Lease of School Lands)
The overall structure of the formula above was not changed between 1961 and 1975,
although several of the parameters did change. Instructional personnel were allotted schoolby-school, based on enrollment, and then aggregated to the district level;26 the salary
schedule—which depended on education and experience—was then applied to the
personnel actually employed.27 The allotment for transportation depended on one-way miles
traveled, the type of vehicles used and the bus driver salary schedule. Districts also received a
flat per-pupil grant for other costs.28 Finally, in some years an adjustment was added to
The district was not required to employ the number of teachers allotted to a school in that school; they could
distribute their allotted teachers across the district as they saw fit.
27 Teachers, supervisors, and visiting teachers were actually allotted separately. Supervisor and visiting teacher
salaries did not depend on education and experience. I combine the three categories for ease of exposition;
expenses for instructional personnel are primarily for teachers. There was a single salary schedule for black and
white teachers.
28 Technically, districts receive a small “per-educable” distribution from the Public School Fund based on the
number of educable students in the district (including those in private schools); this payment is then considered
part of required support for purposes of the Minimum Foundation Program. Since these payments cancel out
from the perspective of the Public School Fund as a whole, I exclude them here.
26
32
program costs for districts with declining enrollment, presumably to ease the transition to
lower enrollment.29
In terms of required support, the proceeds from the 5 Mill Constitutional Tax
depend directly on the local property tax base. The severance tax is paid to a Parish when
certain natural resources (such as minerals) are harvested. Proceeds from this tax, as well as
from rent or lease of school lands, depend on the natural resources in the Parish and are
relatively small for most Parishes.
Both the schedules used in calculating the programs costs (set by the state)—such as
the teacher salary schedule and the number of teachers allotted for schools of different
sizes—and the inputs districts employed—such as the education and experience of teachers
and the number of miles students are transported—influenced a district’s program cost.
Both could have affected districts with different black enrollment shares differently. Ideally,
I would separate the effects of changes in the formula from changes in inputs to the formula
by recalculating the 1970 allocation using the formula in effect in 1965 and comparing this to
the actual allocation—the difference could then be attributed to changes in the inputs.30
Unfortunately, not all the information needed to apply the school finance formula is
available in the Annual Financial and Statistical Report. Still, more-detailed data on districts’
revenue from the MFP are available, and I use these to assess whether changes in the
formula (schedules), changes in inputs chosen by districts, or both account for the
disproportionate increase in revenue from this source for high-black-enrollment-share
districts. The Appendix Table presents a decomposition of the change in the coefficient on
initial black enrollment share from 1965 to 1970 into the parts attributable to program costs
29 There were sometimes other adjustments when a district was under- or over-paid in the previous year. These
were small, so I exclude them here.
30 Alternatively, one could evaluate the 1965 and 1970 formulas, holding inputs constant at their 1965 values.
33
and required contributions.31 I present the results for regressions with and without controls;
the results are similar, so I refer to the specification with controls in the discussion below.
Because the declining enrollment adjustments were only temporary (in 1970), I also show the
decomposition for the change from 1965 to 1973.
There is no single component of the formula that was adjusted to provide more
revenue to predominately black districts; instead, such districts benefited at least a little from
virtually every change in the calculation of total costs between 1965 and 1970. Allotments
for salaries (number of personnel allotted × average salary), transportation, and the declining
enrollment adjustments all contributed to the disproportionate increase in revenue from the
state formula aid, accounting for 57, 31, and 13 percent of the change, respectively. In fact,
these program-cost-side categories account for more than 100 percent of the increase (122
percent), as required support also increased somewhat more in blacker districts (reducing
their formula aid).
The allotment for instructional personnel increased more in high-black-enrollmentshare districts for two reasons: First, the number of teachers allotted per student grew more
in those districts. The teacher allotment schedule changed so that relatively large elementary
schools were allotted more teachers. I do not have information on school size, but this
finding suggests that blacker districts had larger schools and therefore benefited more from
this provision. Second, average salaries increased more in high-black-enrollment share
districts, which employed more-experienced and better-educated teachers before
desegregation and saw bigger increases in both the average education and experience of their
teachers between 1965 and 1970, relative to whiter districts. The average return to
31 Each Annual Financial and Statistical Report presents detail on Minimum Foundation Program allotments for
the following school year. “Other Adjustments” is the difference in the allotment reported in the main revenue
table and the amount reported in the MFP detailed table of the previous year’s report.
34
experience in the salary schedule actually fell slightly, but this was more than offset by the
higher experience and educational attainment and increases in the return to education in
high-black-enrollment-share districts.
Some of these increases are expected to be temporary (for example, the declining
enrollment adjustment). However, the differential increase in state formula aid for blacker
districts was maintained through the early 1970s (Figure 7B). The second set of results in the
Appendix Table shows the same decomposition for the change from 1965 to 1973. The
results are quite similar, suggesting the changes in state formula aid were maintained for
some time after desegregation.
Overall, the decomposition of changes in state formula aid funding does not point to
a single, identifiable change that was clearly designed to help districts with higher black
enrollment shares. Nevertheless, the sum of many changes added up to the large
disproportionate increases for “blacker” districts which were more affected by
desegregation.
35
Figure 1. Trends in Segregation, Louisiana School Districts
.8
Fraction
.6
.4
.2
Share of Districts with
Any Desegregation
Major C.O. Desegregation Begins
CRA/ESEA
1
White Exposure to Blacks
Black Exposure to Whites
0
1960
1965
1968
1970
1975
Notes: Authors calculations based on U.S. Department of Health Education and Welfare Office of Civil Rights data (1967-76) and Southern
Education Reporting Survey data (1960-66). Trends reported are averages for all districts; results are similar if sample is limited to the 47 districts with
data available in all years reported.
36
Figure 2. Black Enrollment Share and Change in Whites’ Exposure to Blacks
Change in White Exposure to Blacks, 1964-1970
.8
.6
.4
.2
0
0
.2
.4
Black Share of Enrollment, 1961
.6
Notes: Black share of enrollment from Louisiana Department of Education Annual Financial and Statistical Report. Change in white exposure to blacks
calculated from U.S. Department of Health Education and Welfare Office of Civil Rights and Southern Education Reporting Service data.
37
.8
Figure 3. Black Enrollment Share and Student-Teacher Ratios, 1961
A. White Student-Teacher Ratio
B. Black Student-Teacher Ratio
40
Students per Teacher
Students per Teacher
35
30
25
20
15
35
30
25
20
0
.2
.4
Black Share of Enrollment
.6
.8
0
.2
.4
Black Share of Enrollment
.6
.8
C. Average ST Ratio - White ST Ratio
Students per Teacher
5
0
-5
-10
0
.2
.4
Black Share of Enrollment
.6
.8
Notes: Black share of enrollment from Louisiana Department of Education Annual Financial and Statistical Report. Students and teachers were segregated
by race, so the white student-teacher ratio is the total number of white students divided by the total number of white classroom teachers at the district
level; black student-teacher ratio is calculated similarly. Panel C shows the white student-teacher ratio less the average student-teacher ratio (total
students divided by total teachers).
38
Figure 4. Trends in Public and Private Enrollment by Race, Louisiana
600
Thousands of Students
White Public
400
Black Public
200
White Non-Public
Black Non-Public
0
1960
1965
1970
Notes: Author’s calculation based on Louisiana Department of Education Annual Financial and Statistical Report.
39
1975
Figure 5. Trends in Per-Pupil Revenue by Source
5
Thousands of Dollars per Pupil (2007$)
Total
4
3
State Formula Aid
2
Local
1
Other Non-Local
Federal ESEA
0
1960
1965
1970
Notes: Author’s calculation based on Louisiana Department of Education Annual Financial and Statistical Report.
40
1975
Figure 6. Differential Trends in White Enrollment by 1961 Fraction Black
-1.5
.4
.2
-.2
0
-1
-.5
Log Points
Percentage Points
0
.6
.8
B. Share of White Enrollment in Private Schools
.5
A. Natural Log of White Public Enrollment
1955
1960
1965
1970
1975
1955
1960
1965
1970
Notes: Figure plots coefficient on 1961 fraction black (and 95% confidence interval) from year-by-year regressions of outcome variable on 1961
fraction black, Census controls, and the share of enrollment eligible for Title I. Coefficients are re-scaled so that the coefficient is equal to 0 in 1965.
41
1975
Figure 7. Differential Trends in Per-Pupil Revenue by 1961 Fraction Black
A. Total Revenue Per Pupil
B. State Formula Aid Per Pupil
3
Thousands of Dollars (2007$)
Thousands of Dollars (2007$)
3
2
1
0
-1
2
1
0
-1
1955
1960
1965
1970
1975
1955
C. Local Revenue Per Pupil
1965
1970
1975
D. Federal ESEA Revenue Per Pupil
3
Thousands of Dollars (2007$)
3
Thousands of Dollars (2007$)
1960
2
1
0
-1
2
1
0
-1
1955
1960
1965
1970
1975
1955
1960
1965
1970
Notes: Figure plots coefficient on 1961 fraction black (and 95% confidence interval) from year-by-year regressions of outcome variable on 1961
fraction black, Census controls, and the share of enrollment eligible for Title I. Coefficients are re-scaled so that the coefficient is equal to 0 in 1965.
42
1975
Figure 8. Differential Trends in Educational Inputs by 1961 Fraction Black
A. Current Expenditure Per Pupil
B. Student-Teacher Ratio
3
2
Student per Teacher
Thousands of Dollars (2007$)
5
1
0
0
-5
-10
-1
-15
1955
1960
1965
1970
1975
1955
1960
1965
1970
Notes: Figure plots coefficient on 1961 fraction black (and 95% confidence interval) from year-by-year regressions of outcome variable on 1961
fraction black, Census controls, and the share of enrollment eligible for Title I. Coefficients are re-scaled so that the coefficient is equal to 0 in 1965.
43
1975
Table 1A. Summary Statistics: Dependent Variables
1965
1970
Change
Mean
s.d.
Mean
s.d.
Mean
s.d.
LN (White Public Enrollment)
8.50
0.94
8.41
1.10
-0.08
0.25
LN (White Public and Private Enrollment)
8.60
1.02
8.62
1.09
0.02
0.16
White Private Share of Enrollment
0.09
0.11
0.18
0.14
0.08
0.14
LN (Total Enrollment)
9.05
0.81
9.04
0.89
-0.02
0.16
LN (Assessed Valuation)
12.30
1.01
12.33
1.02
0.03
0.13
LN (Real Assessed Valuation)
11.41
1.07
11.39
1.09
-0.02
0.09
LN (Non-Real Assessed Valuation)
11.72
0.99
11.77
1.00
0.05
0.19
LN (Per-Pupil Assessed Valuation)
3.24
0.38
3.29
0.40
0.05
0.18
Total Revenue Per Pupil (Thousands of 2007$)
3.06
0.35
4.23
0.49
1.17
0.51
Local Revenue Per Pupil (Thousands of 2007$)
0.71
0.27
1.07
0.36
0.36
0.26
State Formula Aid Per Pupil (Thousands of 2007$)
1.89
0.30
2.34
0.32
0.45
0.19
Federal ESEA Revenue Per Pupil (Thousands of 2007$)
0.06
0.10
0.33
0.22
0.27
0.22
Other State & Federal Aid Per Pupil (Thousands of 2007$)
0.39
0.11
0.49
0.13
0.10
0.11
Current Expenditure Per Pupil (Thousands of 2007$)
2.67
0.30
3.76
0.42
1.09
0.43
Teachers per 100 students
4.10
0.32
4.71
0.40
0.61
0.41
Transportation Aid Per Pupil (Thousands of 2007$)
0.27
0.11
0.33
0.13
0.06
0.05
Enrollment Variables
Property Tax Base Variables
Revenue Variables
Educational Inputs
63
Number of observations
Notes: Cameron Parish had unusually high local revenue due to the discovery of oil on school property and is excluded from the analysis.
44
Table 1B. Summary Statistics: Independent Variables
Mean
s.d.
Fraction Black, 1961
0.39
0.16
Per-Capita Income, Thousands of Dollars (1960 Census)
1.07
0.28
Share of Households with Complete Plumbing (1960 Census)
0.48
0.14
Share of Households with less than $3,000 Income (1960 Census)
0.46
0.14
LN (Total Population) (1960 Census)
10.34
0.86
Urban Share of Population (1960 Census)
0.36
0.25
Current Expenditure Per Pupil (1961)
2.32
0.32
Share of Enrollment Eligible for Title I (1966)
0.32
0.16
63
Number of observations
Notes: Cameron Parish had unusually high local revenue due to the discovery of oil on school
property and is excluded from the analysis.
45
Table 2. Effects of Desegregation on Public and Private Enrollment
(1)
Black Share of Enrollment (1961)
A. Ln (White Public Enrollment)
(2)
(3)
(4)
-1.108***
(0.141)
Fraction Eligible for Title I (1966)
Census Controls
Change in Ln (Lagged Births)#
Observations
R-squared
63
0.503
-0.858***
(0.163)
-0.792***
(0.191)
-0.756***
(0.213)
X
X
X
X
X
X
63
0.557
63
0.629
52
0.672
C. White Percent Private
Black Share of Enrollment (1961)
0.452***
(0.0901)
Fraction Eligible for Title I (1966)
Census Controls
Change in Ln (Lagged Births)
Observations
R-squared
63
0.291
0.301***
(0.105)
0.277**
(0.128)
X
X
X
63
0.360
63
0.418
B. Ln (White Public + Private Enrollment)
(5)
(6)
(7)
(8)
-0.521***
(0.111)
63
0.265
-0.466***
(0.135)
-0.426***
(0.156)
-0.337*
(0.168)
X
X
X
X
X
X
63
0.271
63
0.412
52
0.494
D. Ln (Black and White Public Enrollment)
-0.609***
(0.0976)
63
0.390
-0.391***
(0.109)
-0.351***
(0.129)
-0.443***
(0.139)
X
X
X
X
X
X
63
0.495
63
0.564
52
0.652
Notes: Robust standard errors in parentheses. Dependent variables are changes from 1965 to 1970. Lagged births is calculated by adding the number of
births for all cohorts who would be aged 6-18 in the fall of the school year born in the Parish, separately by race where appropriate. Birth data are only
available for 52 Parishes. Census controls are county-level variables from the 1960 Census: Per-capita income, percent of households with complete
plumbing, percent of households with less than $3,000 in income, the natural log of population, and the share of population in urban areas. Cameron
Parish excluded. * significant at 10%; ** significant at 5%; *** significant at 1%.
46
Table 3. Effects of Desegregation on the Property Tax Base
A. Ln (Total Assessed Valuation)
(1)
(2)
(3)
Black Share of Enrollment (1961)
-0.00366
(0.103)
Fraction Eligible for Title I (1966)
Census Controls
Observations
R-squared
63
0.000
0.161
(0.120)
0.212
(0.128)
X
X
X
63
0.089
63
0.375
C. Ln (Real Assessed Valuation)
Black Share of Enrollment (1961)
-0.111
(0.0703)
Fraction Eligible for Title I (1966)
Census Controls
Observations
R-squared
63
0.040
0.0122
(0.0813)
-0.00757
(0.100)
X
X
X
63
0.142
63
0.209
B. Ln (Per-Pupil Assessed Valuation)
(5)
(6)
(7)
0.606***
(0.122)
63
0.289
0.553***
(0.148)
0.563***
(0.177)
X
X
X
63
0.294
63
0.387
D. Ln (Non-Real Assessed Valuation)
0.0196
(0.148)
63
0.000
0.212
(0.176)
0.286
(0.192)
X
X
X
63
0.059
63
0.318
Notes: Robust standard errors in parentheses. Dependent variables are changes from 1965 to 1970. Real assessed valuation includes the value of
real property; non-real assessed valuation includes the value of all other property. Census controls are county-level variables from the 1960
Census: Per-capita income, percent of households with complete plumbing, percent of households with less than $3,000 in income, the natural
log of population, and the share of population in urban areas. Cameron Parish excluded. * significant at 10%; ** significant at 5%; *** significant
at 1%.
47
Table 4. Effects of Desegregation on Per-Pupil Revenue by Source
A. Local Revenue PP
Black Share of Enrollment (1961/62)
0.203
(0.207)
Fraction Eligible for Title I (1966)
Census Controls
Lagged Change (1961-1964)
Observations
R-squared
63
0.016
B. Federal ESEA PP
0.490*
(0.245)
0.437
(0.302)
X
X
X
X
63
0.081
63
0.154
0.667***
(0.151)
63
0.243
D. Other State & Federal PP
Black Share of Enrollment (1961/62)
0.158*
(0.0878)
Fraction Eligible for Title I (1966)
Census Controls
Lagged Change (1961-1964)
Observations
R-squared
63
0.051
C. State Formula Aid PP
0.142
(0.141)
0.133
(0.173)
X
X
X
X
63
0.560
63
0.596
0.609***
(0.126)
63
0.277
0.629***
(0.154)
0.554***
(0.160)
X
X
X
X
63
0.278
63
0.529
F. Total Revenue PP
0.0426
(0.104)
-0.0325
(0.130)
X
X
X
X
63
0.108
63
0.160
1.637***
(0.341)
63
0.274
1.303***
(0.410)
1.092**
(0.482)
X
X
X
X
63
0.298
63
0.411
Notes: Robust standard errors in parentheses. Dependent variables are changes from 1965 to 1970 in thousands of 2007 CPI-adjusted dollars. Census controls are countylevel variables from the 1960 Census: Per-capita income, percent of households with complete plumbing, percent of households with less than $3,000 in income, the
natural log of population, and the share of population in urban areas. Cameron Parish excluded. * significant at 10%; ** significant at 5%; *** significant at 1%.
48
Table 5. Effects of Desegregation on Educational Inputs
A. Current Expenditure PP
(1)
(2)
(3)
Black Share of Enrollment (1961/62)
1.371***
(0.290)
Fraction Eligible for Title I (1966)
Census Controls
Lagged Change (1961-1964)
Observations
R-squared
63
0.269
B. State Transportion PP
(8)
(9)
(10)
1.082***
(0.348)
1.091***
(0.408)
X
X
X
X
63
0.294
63
0.411
0.161***
(0.0344)
63
0.263
0.155***
(0.0421)
0.171***
(0.0458)
X
X
X
63
0.264
63
0.471
C. Teachers per 100 students
(5)
(6)
(7)
1.281***
(0.279)
63
0.257
1.122***
(0.339)
1.133***
(0.362)
X
X
X
X
63
0.266
63
0.491
Notes: Robust standard errors in parentheses. Dependent variables are change from 1965/66 to 1970/71 school year. Per-pupil current expenditure and state
transportation aid are in thousands of 2007 CPI-adjusted dollars. Student-teacher ratio is total classroom teachers divided by total enrollment. Census controls are countylevel variables from the 1960 Census: Per-capita income, percent of households with complete plumbing, percent of households with less than $3,000 in income, the
natural log of population, and the share of population in urban areas. Cameron Parish excluded. * significant at 10%; ** significant at 5%; *** significant at 1%.
49
Table 6. Effect of Fraction Black on Pre-Program Trends
Dependent Variable (Change 1957 to 1963)
Coefficient on
1957 Fraction Black
Ln (White Public Enrollment)
0.155
(0.107)
Ln (White Public and Private Enrollment)
0.149
(0.0897)
White Fraction Private
-0.00896
(0.0296)
Ln (Black and White Public Enrollment)
0.0733
(0.0820)
Ln (Total Assessed Valuation)
0.0606
(0.132)
Ln (Per-pupil Assessed Valuation)
-0.0127
(0.151)
Ln (Real Assessed Valuation)
-0.0305
(0.154)
Ln (Non-Real Assessed Valuation)
0.153
(0.226)
Total Revenue Per Pupil
-0.135
(0.194)
Local Revenue Per Pupil
-0.117
(0.143)
State Formula Aid Per Pupil
-0.134
(0.0889)
Non-Local Revenue Per Pupil
0.115
(0.138)
Current Expenditure Per Pupil
-0.121
(0.115)
Teachers per 100 students
0.00452
(0.212)
Notes: Robust standard errors in parentheses. Dependent variables are change
from 1957/58 to 1963/64 school year. Coefficient on 1957 fraction black is
reported; robust standard errors in parentheses. All regressions include the
fraction of enrollment eligible for Title I in 1966, per-capita income (1960), share
of households with complete plumbing (1960), share of households with less
than $3,000 income (1960), natural log of 1960 population, the 1960 urban share,
and per-pupil current expenditure in 1957. Cameron Parish excluded. *
significant at 10%; ** significant at 5%; *** significant at 1%.
50
Appendix Table. Decomposition of Differential Change in State Formula Aid
Change 1965-70
Change 1965-73
No Controls
With Controls
No Controls
With Controls
Coefficient on Fraction Black
0.609
0.554
0.577
0.527
Percent Attributable to:
Program Costs
112%
122%
96%
110%
55%
26%
8%
23%
0%
57%
31%
13%
21%
0%
67%
16%
1%
0%
12%
72%
20%
1%
0%
17%
Required Support (Negative)
-27%
-34%
-19%
-31%
5 Mill Constitutional Tax
Severance Tax
Rent or Lease of Lands
-7%
-12%
-8%
-6%
-14%
-14%
-8%
-10%
0%
-9%
-16%
-5%
16%
12%
24%
20%
Salaries
Transportation
Per-ADM Instruction Costs
Declining Enrollment
Salary Adjustment
Other Adjustments
Notes: Per-pupil revenue figures are reported are in thousands of constant 2007 dollars. Program costs include allotment
for salaries of teachers, visiting teachers, supervisors, and transportation, adjustments for districts with declining
enrollments (1970), and adjustments for teacher salaries (1973). Required support includes the proceeds of the 5 mill
Constitutional and severance tax and half of the funds received from rent or lease of school lands. Information on the
breakdown of program costs and required support for the following school year were reported in the Annual Report,
while actual payments from the state formula was reported in the Annual Report of the same year; "other adjustments" is
the difference between actual payments and program costs less required support (these are adjustments to aid allocations
made after the previous year's report was printed).
51