PAYING FOR PROGRESS: CONDITIONAL GRANTS AND THE DESEGREGATION OF SOUTHERN SCHOOLS*

Elizabeth Cascio Nora Gordon Dartmouth College and NBER UC San Diego and NBER

Ethan Lewis Sarah Reber

Dartmouth College UCLA and NBER

October 2008

Abstract

This paper shows how conditional grants can induce government officials to undertake even the most controversial policy changes. Specifically, we find that the threat of withdrawal of federal funding played a causal role in desegregating Southern schools in the mid-1960s. We use newly collected archival data and exploit quasi-experimental variation in federal funding to show that school districts with more federal funding on the line were more likely to move away from “token” levels of desegregation, instead desegregating just enough to receive their funds; such districts were also more likely to desegregate their teaching faculties, another condition of the grants. We emphasize not only that Southern school boards had their price, but that it was a substantial one: our estimates imply that on average a district would need to be paid $1,000 per pupil to move beyond token desegregation. These financially-induced reductions in segregation were important in getting desegregation underway ten years after Brown and encouraging voluntary desegregation—as opposed to court-mandated plans—through 1970.

* For their helpful comments and questions, we are grateful to seminar participants at Northwestern, Stanford, UBC, UCD, UCI, UCSD, UCR, the All-UC Labor Workshop, the NBER Economics of Education and Development of the American Economy program meetings, and the annual meetings of the AEA, SSHA, and SOLE. We would especially like to thank Patty Anderson, Sandra Black, Leah Platt Boustan, Ken Chay, Julie Cullen, Roger Gordon, Jon Guryan, Sean Reardon, and Doug Staiger. Jeremy Gerst, Maria Kahle, Farah Kaiksow, Allison Kidd, Cyrus Kosar, Eric Larsen, Patricia Tong, and Courtney Wicher provided indispensable research assistance. This research was supported by grants from the National Science Foundation (Award Number 0519126), the Spencer Foundation (Award Number #200600131), and the University of Kentucky Center for Poverty Research Regional Small Grants Program. Cascio gratefully acknowledges support from a Junior Faculty Research Grant from the Institute of Governmental Affairs at UC Davis. Gordon gratefully acknowledges support from the National Academy of Education/Spencer Foundation postdoctoral fellowship. The data presented, the statements made, and the views expressed are solely the responsibility of the authors.

I. Introduction Because the U.S. Constitution reserves powers not explicitly delegated to the federal

government for the states, conditional grants are key levers for federal policymakers seeking to

affect a broad range of state and local policies. For example, states cannot receive federal highway

funding unless they implement federally-approved speed limits and drinking ages; universities

cannot receive federal research funding if they do not provide gender parity in athletic offerings;

states can lose federal highway and other funding if they do not satisfy certain requirements

under the Clean Air Act; and historically, states could not receive federal land grants for higher

education unless they offered a prescribed curriculum including agricultural science. More

recently, the threat of withdrawal of the same stream of federal education funding we study here

has been used to encourage states and districts to comply with the accountability requirements

under the No Child Left Behind Act. We find that federal grants in the mid-1960s were able to

induce Southern school boards to meet extremely unpopular conditions—desegregating students

and teachers in public schools.

Dismantling the dual system of education in the South was, to say the least, contentious.

Both the popular press and researchers have emphasized the strength of school board opposition

and the need for court mandates, in some cases enforced by police or the National Guard, as a

“stick” to desegregate Southern schools. Particularly salient cases, such as those in Little Rock

and New Orleans, highlighted extreme white resistance, perhaps contributing to a perception that

the average white Southerner was so opposed to desegregation that he would only accept it when

physically forced. While the courts played a critical role in desegregating Southern schools (e.g.,

Welch and Light, 1987; Reber, 2005), nearly half of Southern school districts desegregated

without court supervision (Cascio, et al., 2008). In this paper, we explore the role of “carrots”—

financial incentives for “voluntary” (not court-supervised) desegregation introduced in the mid-

1

1960s—in this process. We use newly collected data to show that Southern school boards had

their price, but that it was a substantial one: our estimates imply that on average a district would

need to be paid $1,000 per pupil to move beyond token desegregation.

In order to receive any federal funds, Southern school districts had to comply with the

non-discrimination provisions of the Civil Rights Act of 1964 (CRA) by desegregating their

schools. Although schools received little federal funding in 1964, Title I of the Elementary and

Secondary Education Act of 1965 (ESEA) created large federal grants for education that made

non-compliance costly. Figure I shows trends in segregation and court supervision for school

districts in the former Confederacy1 during the two decades following the Supreme Court’s 1954

ruling in Brown vs. Board of Education. 2 There was a burst of largely voluntary desegregation

activity around the time CRA and ESEA were first enforced, pre-dating landmark Supreme Court

rulings thought to have strengthened the hand of the courts in desegregating schools.3 We have

previously shown that high-poverty districts—which stood to gain the most from Title I—were

particularly likely to desegregate around the time of CRA and Title I implementation (Cascio, et

al., 2008). Other studies have also suggested that these policies may have been responsible for

this trend.4 However, past work has not been able to separate the influence of financial pressure

on school districts from the effects of other, concurrent changes to the policy environment, such

as the heightened threat of the Attorney General suing school districts (authorized under other

provisions of CRA) and the enfranchisement of blacks following the Voting Rights Act of 1965.

1 These states include Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia. Texas lacks segregation data and is therefore excluded from our sample and Figure I. 2 Almond, Chay, and Greenstone (2006) hypothesize that CRA, in conjunction with the potential of receiving Medicare payments beginning in 1966, contributed to the desegregation of rural Southern hospitals and conclude that increased access to medical care as a result reduced black infant mortality substantially. However, they do not explicitly test the extent to which incentives affected hospital desegregation itself, as we do for schools in this paper. 3 Green v. New Kent County (391 U.S. 430), 1968 and Swann v. Charlotte-Mecklenburg (401 U.S. 1), 1971. 4 See, for example, Rosenberg (1991), Boozer, Krueger, and Wolkon (1992), Clotfelter (2004), and Ashenfelter, Collins, and Yoon (2006).

2

We present a simple theoretical framework which yields an intuitive prediction: if

financial pressure under CRA caused desegregation, then as the potential grant to a district

(exogenously) increased, so did the likelihood of the district exerting the minimum desegregation

effort required to collect federal funds. To test this prediction, we exploit idiosyncratic variation

across school districts in the amount of federal funding at risk from non-compliance with

desegregation guidelines. In 1966, the amount of Title I funding a compliant district would have

received was based on child poverty as well as state-specific factors. In particular, the gap in

expected Title I receipts between poor and rich districts was larger in states with higher per-pupil

spending prior to ESEA; there was not a similar gap for other factors influencing desegregation

or for pre-program desegregation outcomes. We examine whether the relatively large difference

in funding was matched by a relatively large difference in the likelihood of just meeting the

required “threshold” level of student desegregation, desegregating teaching faculties, and being

supervised by the courts.

We investigate the effects of these financial incentives in 1966, the first year affected by

the policy for which appropriate data exist, and find that these incentives did prompt Southern

school districts to move beyond token desegregation. We estimate that the probability of having

only token desegregation (less than two percent of blacks in desegregated schools) fell by about

10 percentage points for each additional hundred dollars in potential per-pupil Title I funding (in

constant 2007 dollars), with districts moving to slightly higher levels of desegregation (two to six

percent of blacks in desegregated schools). This corresponds closely to what was required for

compliance with CRA for the average Southern district. Districts with larger grants were also

more likely to have desegregated faculties, a requirement for CRA compliance, and less likely to

be supervised by a court.

3

Financial incentives appear to have been an important impetus behind the long-overdue

movement away from purely token desegregation a decade after Brown. Our estimates imply that

Title I’s financial incentives explain about 40 percent of the shift away from the highest

segregation levels between 1964 and 1966. On the other hand, the changes in segregation during

this period were small relative to what would follow in later years as DHEW strengthened

guidelines and court supervision expanded. The estimates also suggest that Southerners’

willingness to pay for segregation—while not infinite—was substantial.

II. Financial Incentives and Desegregation in Theory

In this section, we present a framework for understanding the effects of conditional

federal funding on school desegregation using a modified version of the model Margo (1990)

used to understand black-white school spending gaps prior to Brown. Because spending on black

and white students had greatly converged before the period covered by our study (Card and

Krueger, 1992; Donohue, Heckman, and Todd, 2002; Margo, 1990), we depart from Margo and

assume that expenditure per pupil did not vary by race within district. 5 We assume that districts

faced a trade-off between expenditure per pupil, e, and student segregation, s. We measure

segregation as the fraction of black students attending all-black schools, so s ranges from zero to

one. We also assume that decision-making rested in the hands of Southern whites, as few

Southern school boards had any black members at this time,6 and that white school boards chose

e and s to maximize their utility given the school district budget:

( )seUU ,= ,

where the marginal utilities of both spending and segregation are positive and diminishing

5 On the other hand, differences in spending per pupil in black and white schools did persist through the mid-1960s in Louisiana (Reber, 2007). In any case, allowing for differential spending by race would not change the key results. 6 By 1966, between 12 and 19 blacks had been elected to local school boards in the 10 Southern states in our sample (U.S. Commission on Civil Rights, 1968; Appendix VI). There were very few majority black electorate districts in the South after the Voting Rights Act (Alt, 1994).

4

( 0, >∂∂∂∂ sUeU , and 0, 2222 <∂∂∂∂ sUeU ).7

Per-pupil expenditure could not exceed net per-pupil revenue:

)(sfmle τ−++= ,

where l , m , and f respectively represent revenue per pupil from local, state, and federal

sources. For simplicity, we assume that local and state revenue were fixed, though the substantive

implications of the model are unchanged if we relax this assumption by introducing local control

over taxation. )(sτ is the per-pupil expense to the district of segregationist policy s , which has

three components. First, segregation may have entailed logistical costs from diseconomies of

scale and transportation. Second, maintaining higher levels of segregation entailed costs

associated with deterring and fighting litigation; we denote these first two categories of costs by λ.

Third, requirements for compliance with CRA made receipt of federal funds conditional

on reaching some threshold level of student desegregation. Federal funds per pupil received by a

district can therefore be characterized as fsf =)( if ss ~≤ and 0)( =sf if , where

represents this threshold. Thus, the cost of segregation was

ss ~>

s~ λτ ss =)( if the threshold was

reached and fss += λτ )( if not, generating a discontinuity in the budget constraint at : s~

⎩⎨⎧

>−+≤−++

=ssifsml

ssifsfmle ~

~

λλ

.

The value of changed over time and was district-specific, as discussed below. s~

We are interested in how financial incentives influenced districts’ segregation policy

choices. We can examine this issue by showing how the probability of crossing the threshold

level of segregation varied with the amount of federal funding. Figure II plots the budget

constraints of two hypothetical school districts. Both districts have the same preferences over 7 The magnitude of ∂U/∂s depends on the racial composition of the district, which itself might be influenced by the choice of desegregation policy (due to white flight). The critical implication of the model for testing our hypothesis, however, is simply that ∂U/∂s is positive, not that it is constant.

5

segregation and spending, as represented by their identical indifference curves, but the district

represented in Panel A has a smaller Title I allocation and therefore a smaller increase in funding

at . The graphs show that the district facing a sufficiently large federal grant (Panel B) would

have desegregated—just to the point required by CRA,

s~

ss ~= —while the district facing the

smaller grant would have optimized by remaining fully segregated (s =1).

In practice, not all districts would have chosen ss ~= or s =1. For example, a district

with sufficiently weak tastes for segregation and/or sufficiently high costs of segregation may

have desegregated more than . It is also possible that conditional federal funding had a

perverse effect on segregation for districts that would have otherwise been segregated less than

s~

s~. Assuming segregation is a normal good, such a district would have consumed more

segregation due to the income effect of receiving the grant. Given the very low levels of

voluntary desegregation before 1966, we expect this effect is small.

This simple framework thus suggests that, all else equal, districts with larger grants would

have been more likely to just cross the threshold to receive their federal funds; our empirical

models are designed to look for the incentive to operate around . Variation in segregation

policy s may have also arisen due to heterogeneity across districts in preferences for segregation

as well as in the slope and position of the budget constraint (due to differences the threat of

litigation and per-pupil revenue). This observation points to the importance of using variation in

federal funding that is not correlated with these other key determinants of segregation to identify

the effect of conditional funding. We exploit quasi-random variation in funding under Title I in

1966 to approach this idealized comparison.

s~

III. Empirical Strategy

The model of interest is

(1) ( ) djdjdjdj Xfpptiy εθα +++= 1966

6

where ydj represents an indicator set to one if district d in state j met a particular desegregation

target; pptidj1966 represents potential Title I funding per pupil in the district in 1966; f(Xdj) represents

a function of observable district characteristics; and εdj captures unobserved determinants of the

segregation decision. The parameter of interest is θ , which gives the effect of potential Title I

funding on ydj, conditional on observables.8 If the requirement that districts meet desegregation

targets to receive federal funding affected segregation decisions, we expect θ to be positive.

Ordinary least squares (OLS) estimates of θ will be biased if unobserved determinants of

segregation are correlated with the per-pupil Title I grant. Because Title I was a compensatory

program and funding was based on the number of poor children living in a district, we are

particularly concerned about confounding effects of federal funding with effects of poverty.

Fortunately, the Title I funding formula provides a source of variation in federal funding that is

arguably uncorrelated with other determinants of a district’s segregation policy. A district’s 1966

Title I allocation was equal to the number of poor children counted in the 1960 Census (poord1960)

multiplied by a state-specific component – funds granted per poor child (stategrantj1966). 9 The per-

pupil Title I grant was therefore

(2) 1966

196019661966

d

djdj enr

poorstategrantppti ≡ ,

8 In fact, all federal funding was on the line, not just Title I funding. We use Title I funding rather than all federal funding due to data constraints. The parameter θ is thus appropriately interpreted as the reduced-form effect of the Title I program on desegregation. However, as long as the other categories of federal funding are uncorrelated with the identifying variation in Title I funding, our empirical strategy will also produce unbiased estimates of the effect of federal funding overall. If Title I and other federal funding were positively (negatively) correlated, our estimates will overstate (understate) the effects of overall federal funding (because more (less) funding is on the line than our specification accounts for). While we cannot test this directly, we think it is unlikely that other federal funding is correlated with Title I funding. We also note that while Title I funding was in theory categorical (for compensatory programs), in practice it was used to finance all types of current education spending (see Washington Research Project, 1969). 9 Specifically, poord1960 represents the estimated count of five to seventeen-year-olds living in families with incomes less than $2000, based on the 1960 Census. The number of Title I eligibles in a district also included children whose families had incomes above $2000 but received AFDC, neglected and delinquent children, and children in foster homes. In practice, the number of Title I eligibles in these categories was quite small in the South. See the Data Appendix.

7

where enrd1966 is the district’s enrollment in 1966.

Title I funding per poor child (stategrantj1966) in 1966 was equal to one-half 1964 average

per-pupil expenditure in state j.10 Within a state, therefore, districts with more poor children

were due more Title I funding, but two districts with the same poverty count would have

received different amounts depending on their state. Intuitively, our approach is similar to a

difference-in-differences framework, comparing outcomes for higher- and lower-poverty districts

in higher- and lower-spending states. That is, the gap in expected Title I receipts between poor

and rich districts was larger in states with higher per-pupil spending prior to ESEA; we test

whether the relatively large difference in funding was matched by a relatively large difference in

the likelihood of just meeting the “threshold” level of desegregation required to receive federal

funding.

Following the logic of the difference-in-differences estimation strategy, we include

functions of the district- and state-specific components of per-pupil Title I funding as controls in

our baseline model. In some models, we also control for other characteristics of school districts

that might have influenced segregation policy. Expanding the vector of controls in Equation (1),

we have:

(1′) ( ) ( ) ( ) djdjdddjdj Wlstategranthenrpoorgpptiy εθα +++++= 1966196619601966 .

The functions ( )⋅g and ( )⋅h are critical controls, as the district’s child poverty rate,

poord1960/enrd1966, and the state factor in Title I funding, stategrantj1966, both correlate strongly with

potential Title I funding and may be directly related to segregation outcomes. The vector Wd

represents other district characteristics.

10 Because the Title I program was not fully funded for the 1966-67 academic year, this figure was ratably reduced by a state-specific multiplicative constant. The main results are qualitatively similar if the non-ratably-reduced grant is used in constructing the simulated Title I grant instrument described below. Appendix Table I gives the values of stategrantj1966 by state.

8

While the total Title I grant was determined based only on pre-program district

characteristics, the per-pupil amount in 1966 depended on 1966 enrollment, which was potentially

endogenous to desegregation policy. The endogeneity of enrollment could bias our estimates in

either direction. For example, districts with an unobserved taste for segregation might have

desegregated less and experienced more “white flight” in response to desegregation. This would

generate a negative correlation between pptidj1966 and the error term in Equation (1′), biasing OLS

estimates of θ downward. Conversely, holding preferences constant, school desegregation may

have increased white flight, biasing OLS estimates of θ upward. Even in the absence of an

endogeneity problem, OLS estimates of θ will be attenuated if current enrollment is measured

with error.

We therefore instrument for the actual 1966 per-pupil Title I grant with the district’s

“simulated” Title I grant per pupil, , which holds enrollment constant at pre-Title I

levels:

SIMdjppti 1966

(3) pred

djSIMdj enr

poorstategrantppti

,

196019661966 ≡

The denominator of is average enrollment in district d in the years prior to Title I’s

introduction (i.e., through the early 1960s). Notice that the simulated grant is based entirely on

pre-program district characteristics and is itself another noisy measure of Title I funding per

pupil, allowing us to address biases from both the endogeneity of enrollment and measurement

error. The first stage reveals that is a strong predictor of , with F-statistics over

300.

SIMdjppti 1966

SIMdjppti 1966 1966djppti

11

Finally, current-year values of other district characteristics in equation (1′) are also

11 See Appendix Table II for first-stage regressions. The specifications are discussed in detail in Section V.

9

potentially endogenous. We therefore use pre-program, rather than 1966, characteristics as

controls and estimate the following equation by two-stage least squares (TSLS):

(1′′) ( ) ( ) ( ) djpredjpredddjdj Wlstategranthenrpoorgpptiy εθα +++++= ,,19601966 .12

Formally, TSLS estimates of θ in Equation (1′′) are consistent if the simulated grant is

uncorrelated with εdj, conditional on the chosen functions g( ⋅ ), h( ⋅ ), and l( ⋅ ). This assumption

is arguably more likely to be satisfied for more flexible g( ⋅ ) and h( ⋅ ). We therefore use the most

flexible functions that our data can accommodate maintaining reasonable precision: dummies for

quantiles of the pre-program child poverty rate, poord1960/enrd,pre, for g( ⋅ ) and state fixed effects for

h( ⋅ ). Note that state fixed effects account for effects of state-specific factors that directly

affected district-level segregation policy—for example, the lack of desegregation progress in

Alabama under the leadership of staunchly segregationist Governor George Wallace.

While we cannot directly test this identifying assumption, we present several pieces of

supporting evidence. First, we show that, in our chosen specification, is uncorrelated

with several observed proxies for segregationist preferences and the threat of litigation, and that

the coefficients of interest are not sensitive to the addition of key observables to the model.

Second, we estimate several “placebo” regressions with pre-program segregation-related

outcomes as dependent variables and find no significant “effect” of the program before it

existed, as might have been expected if were correlated with unobserved preferences.

SIMdjppti 1966

SIMdjppti 1966

12 We estimate a linear probability model for ease of implementation. The reduced-form, marginal effects of the simulated per-pupil grant (Equation (3)) on our dichotomous outcomes are similar when estimated using a linear probability or probit model.

10

IV. Data

We have created a uniquely comprehensive dataset for this analysis, linking a variety of

sources to obtain school-district-level data on student and teacher segregation levels, court

supervision, Title I allocations, official CRA compliance, and demographic and fiscal control

variables. This section provides a brief overview of these data; see the Data Appendix for more

detail and the Data References for specific publications.

IV.A. Title I and Other Explanatory Variables

Table IA shows summary statistics for the explanatory variables used in our analysis. The

key variable of interest is Title I funding per pupil in 1966, collected from Congressional reports.

In 1966, this figure was $277 for the average district, which was reasonably large relative to the

average pre-program (early 1960s) per-pupil current expenditure of $1,700 (both in 2007 dollars).

Recall that the “simulated” Title I grant per pupil is the product of the grant per eligible child,

which varied across states (see Appendix Table I) and the district’s pre-program child poverty

rate, which was on average 33.8 percent. It is this poverty rate that enters directly and flexibly

into Equation (1′′).

Equation (1′′) also incorporates a vector of district characteristics that may have been

related to its choice of segregation policy. We use per-pupil current expenditure, obtained from

annual state administrative reports and averaged over 1960 to 1963, to control for the size of the

budget, as suggested by the model. In practice, current expenditure may also reflect preferences

for spending versus segregation. We control for percent black enrollment, from the same state

reports and years, because it affects the amount of exposure white children would have faced

with any given share of blacks in desegregated schools.13 We also use the share of votes cast for

Strom Thurmond in the 1948 Presidential election, measured at the county level, as a proxy for 13 In some cases, we do not have data on enrollment by race for these years, so we use data from later in the 1960s; see Data Appendix.

11

segregationist preferences. As shown in Table IA, the average district in our sample was 37

percent black in the early 1960s and in a county where 36 percent of votes were cast for

Thurmond in 1948. District enrollment serves as measure of the threat of litigation (Cascio, et

al., 2008); the average district in our sample enrolled just over 6,100 students. Several

characteristics of the county population in 1960, taken from the City and County Data Book – the

percent of the population with a high school degree, the share of employment in agriculture,

median family income, and an urban indicator – round out our list of controls.

Recall the identifying assumption in our model: in a specification with sufficiently

flexible controls for its state and district-level components, the simulated Title I grant per pupil

should not be correlated with unobserved determinants of segregation policy. Table II shows

that, with the exception of pre-program expenditure per pupil and the county urban indicator, the

observed district characteristics described above are not significantly related to the instrument in

two such specifications. To mitigate any potential remaining biases, as well as to reduce residual

variation, we control flexibly for pre-program per-pupil funding and all district characteristics in

all specifications estimated below.

IV.B. Outcomes

The main prediction of the model presented in Section II was that school districts

with larger potential federal grants would have been more likely to choose levels of student

desegregation at or above the threshold for receiving federal funds. To identify where this

threshold was — and to develop outcome variables accordingly — it is critical to

understand the specific requirements of the law.

Districts with court-supervised desegregation plans were automatically in

compliance with CRA; other districts were required to submit so-called “voluntary”

desegregation plans satisfying policy guidelines set out by the Department of Health,

12

Education, and Welfare (DHEW). Nearly all desegregation during this period involved

moving black students to formerly all-white schools, and DHEW guidelines were specified

in terms of the share of blacks that had to be transferred. The first such guidelines, for the

1965-66 school year, were vague and ultimately required at most token desegregation. By

1966-67, the year of our main analysis, the guidelines were more specific and required

districts to transfer more than a handful of blacks from all-black schools, with higher

growth in “transfer rates” expected from districts which had desegregated fewer blacks the

prior year.

In theory, we could identify the threshold level of desegregation for each district in

1966 based on its transfer rate in the prior year, and use an indicator for exceeding that

threshold as our dependent variable. In practice, we do not have the data on 1965 transfer

rates needed to calculate the 1966 threshold for districts in all states.14 Moreover, the

guidelines for transfer rates provided by DHEW did not specify clear targets for all

districts.15 They were also not fully enforced, suggesting that districts probably had a

general idea of their thresholds in 1966 but faced some uncertainty about exactly what was

required to receive their funding. In fact, DHEW did not have the data required to enforce

its own guidelines literally (they lacked data on district-level black enrollment, the

denominator necessary to calculate the transfer rate), and the data we use in our analysis are

at least as good as what DHEW would have had at the time.

14 The lack of data for 1965 is also why we do not investigate the effects of financial incentives in the first year they were in force. 15 Districts with transfer rates (in practice, shares of blacks in desegregated schools) of eight to nine percent in 1965 were expected to double their transfer rates; taking the guidelines literally, in 1966 these districts would have been required to have 16 to 18 percent of blacks in desegregated schools. A tripling of transfer rates was expected from districts that had transferred 4 to 5 percent of blacks in 1965, a “proportionally larger” change for districts that had transferred fewer than 4 percent of blacks in 1965, and a “substantial start” from districts with no transfers in 1965 (U.S. DHEW, March 1966; p. 8).

13

In order to focus the analysis on the relevant part of the distribution of student

desegregation, our key dependent variables are indicators for whether a district fell into

specific “bins” defined by the share of blacks in desegregated schools.16 We define these

bins based on our analysis of estimated transfer rates for the two states where we do have

district-level segregation data in 1965 (South Carolina and Tennessee), aggregate state-level

desegregation data in 1965, and the DHEW guidelines. Aggregate data show that the South

Carolina desegregation experience was representative of our sample, while Tennessee was a

desegregation outlier.17 Ninety percent of districts in South Carolina had transfer rates

above zero but less than four percent in 1965; in fifty percent of South Carolina districts,

1965 transfer rates were less than one percent. According to the Guidelines, districts with

transfer rates in this range (greater than zero, but less than four percent) were expected to

more than triple their level of desegregation activity.

Thus, we expect that most districts in our sample would have needed more than

“token” levels of desegregation to meet the targets set out in the Guidelines, while very few

districts would have needed to transfer more than about 10 percent of black students in

desegregated schools. We therefore construct indicator variables for whether a district fell

into each of the following categories: less than two percent of blacks in desegregated

schools (our measure of token desegregation), two to six percent, six to 10, 10 to 20, 20 to

30, 30 to 50, and 50 to 100; we expect to see little action above the 10 percent threshold.

Table IB shows that 64 percent of districts did, in fact, have less than 10 percent of blacks

in desegregated schools in 1966; over half had less than six percent. 16 In Section II, we modeled school board preferences as a positive function of segregation; however, the guidelines for CRA compliance were presented in terms of desegregation. We specify the threshold for compliance in terms of the share of blacks in desegregated schools so that our empirical work matches the policy guidance. 17 The aggregate transfer rate for South Carolina in 1965 (1.65 percent) was comparable to those of Alabama (0.43 percent), Georgia (2.66 percent), Louisiana (0.85 percent), and Mississippi (0.59 percent). The other states in our sample had higher 1965 transfer rates in the aggregate, but they were still quite a bit lower than the Tennessee average of 16.31 (Southern Education Reporting Service, 1967).

14

To create these indicators, we began by calculating the fraction of blacks in desegregated

schools using data from two sources. Data on the number of black students in desegregated

schools (the numerator) come from publications of the Southern Education Reporting Service

(SERS), an organization of Southern newspaper editors funded by the Ford Foundation.18 We

estimate the total number of blacks in the district (the denominator) using current year fall

enrollment and percent black in enrollment in the early 1960s. 19 Roughly 18.1 percent of blacks

were in desegregated schools in 1966, compared to only 0.8 percent in 1964; we create the

analogous variables for 1964 for our specification checks.

The DHEW guidelines also required teacher desegregation on the extensive margin

starting in 1966. In addition to information on the number of black students in desegregated

schools, the SERS publications described above recorded the number of black teachers working

with white teachers. We use these data to estimate the effects of financial incentives on faculty

desegregation. As shown in Table IB, 73 percent of districts had desegregated their teaching

faculties by 1966, or had at least one black teacher in the same school with a white teacher.

Districts with more Title I funding per pupil may also have been less likely to comply

with CRA through submission of a court order, as opposed to a voluntary plan. District-level

data published by DHEW indicates whether a district complied with CRA by submitting a court-

ordered plan. Court orders likely indicate that the district was perceived to be a good target for

litigation, perhaps because it was resistant to desegregating on its own; 9.5 percent of districts in

our sample complied with CRA through submission of a court order in 1966.

18 In 1964 (and earlier), SERS reported data on the number of blacks in desegregated schools for districts that had any desegregation activity; districts not listed are assumed to be fully segregated. In 1966, the number of blacks in desegregated schools was mainly collected by the federal Office of Education (part of DHEW) and reported by SERS. 19 Many states stopped reporting enrollment by race in their administrative publications by mid-1960s, but we do have total enrollment. In a few cases, our estimate of the share of blacks in school with whites is more than one due to measurement error in black enrollment; we recode these to one.

15

The DHEW reports also indicate whether a district had its federal funds deferred or

terminated for failure to comply with CRA. While characterized by contemporaries as weak and

disorganized (Bailey and Mosher, 1968; Orfield, 1969), enforcement of CRA did result in fund

deferral or termination for about 20 percent of districts in our sample by 1966 (Table 1B, Panel

B), so the law was not an empty threat. The district-level South Carolina and Tennessee data

confirm that all districts whose change in transfer rates from 1965 to 1966 appear to conform to

the DHEW guidelines were in fact deemed compliant by DHEW and received their federal funds

on time. No district in either state had their federal funds terminated. Tennessee had unusually

high desegregation levels, so only two of 146 districts had their funds deferred. About 25-30

percent of South Carolina districts which appeared noncompliant had their funds deferred.

These data suggest that enforcement was generally accurate, but not all noncompliant districts

were pursued. Given these caveats, we examine the probability of having federal funds deferred

or terminated as a robustness check, rather than as a central outcome of interest.

IV.C. Sample

We restrict our sample to Southern states for which we could obtain the necessary

data. These include Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North

Carolina, South Carolina, Tennessee and Virginia. It would be interesting to examine the

effects of financial incentives in Border states, which also enforced a dual system before

Brown; unfortunately, the data are less complete for these states. Because school districts

both consolidate and split apart during our sample period, we use the state records

referenced above to establish a history of reorganizations and aggregate the raw data to the

largest unit to which a district was ever a party. Of these “aggregated” districts, we exclude

those for which desegregation was not relevant because they were one-race or nearly one-

16

race.20 We also exclude districts that were automatically in compliance with CRA in 1966

because they were supervised by a court in 1964. Our main estimation sample includes 912

school districts.21

V. The Effects of Financial Incentives in 1966

V.A. Financial Pressure and Student Desegregation

Table III presents estimates of the effect of potential Title I funding per pupil on two

measures of student segregation in 1966: the indicator for a “token” level of desegregation (less

than two percent of black students attending desegregated schools) and the indicator for having

moved just beyond that token level (two to six percent of blacks in desegregated schools). As

described above, for the average Southern district, conditional federal funding was likely to have

mattered most for student desegregation in this part of the distribution. The four specifications

presented for each outcome differ in the choice of function to control for child poverty and the

inclusion of additional pre-existing district characteristics, Wd,pre.22 All specifications include state

fixed effects.

20 In particular, we drop districts that were less than three percent or more than 97 percent black in the early 1960s. The three percent cutoff is arbitrary; results are not sensitive to using alternative cutoffs. 21 There are 1,476 aggregated districts in the raw data, for 1,288 of which we have an estimate of percent black in enrollment in the early 1960s. The vast majority of the 188 districts lost from this sample restriction would have been dropped had their racial composition been directly observed: 78 percent of these districts filed an “assurance of compliance,” or Form 441, with DHEW in 1966, signifying that they were uniracial districts. Of the 1,288 with data on racial composition, 1,083 were between 3 and 97 percent black, on average, in the early 1960s, and of these, 1,015 were not under court order in 1964. We drop 23 districts because they lack data on other key explanatory variables and another 80 districts because data on student desegregation in 1966 are not available. The resulting sample size is 912. See Appendix Table III. 22 Here and in all tables below, standard errors are robust to heteroskedasticity and correlation in unobservable characteristics across districts within counties. We cluster standard errors on county because some of our control variables vary at the county level. Note that our data are a cross-section, and the state fixed effects will account for any unobservable state-specific component of the error term. Of course, there remains the possibility that the remaining component of the error term is correlated across districts within the same state. When we cluster standard errors on state and use the critical values from the t-distribution with 8 degrees of freedom to establish statistical significance (following Monte Carlo simulations done in Cameron, Gelbach, and Miller (2007)), the statistical significance of our key results is largely unchanged. All estimates also give each district equal weight; weighting the analysis by early 1960s black enrollment yields very similar findings. These additional findings are available from the authors upon request.

17

Because the same specifications are shown below for other outcomes, we discuss them

here in some detail. We begin by estimating θ in Equation (1′′), controlling flexibly for the state-

and district-specific components of the simulated per-pupil Title I grant but omitting other pre-

existing district characteristics. The first specification, shown in columns (1) and (5), includes

state fixed effects and dummies for 20 quantiles of the pre-program district child poverty rate.

To improve the precision of our estimates moving forward, all subsequent models include a more

parsimonious set of quantile dummies for the child poverty rate (“restricted” quantiles).23 We

first show the more parsimonious model without additional controls (columns (2) and (6)). We

then add controls that capture preferences and components of the budget constraint (percent of

votes cast for Thurmond in 1948 and early 1960s expenditure per pupil, both using quintile

indicators, and early 1960s percent black in enrollment, using decile indicators) and the litigation

threat (log of early 1960s enrollment) (columns (3) and (7)). The final specification adds the

other socioeconomic indicators available at the county level in the 1960 Census (columns (4) and

(8)). If our instrument is valid, it should be the case that the inclusion of these characteristics

does not significantly change our point estimates, but may improve their precision if the controls

increase the explanatory power of the model.

TSLS estimates are shown in Panel A.24 The findings suggest that the requirement that

districts meet desegregation targets to receive their federal funds did affect behavior, shifting

districts from tokenism to somewhat more meaningful desegregation. In the specification with

the full set of controls, the TSLS estimates imply that a hundred-dollar increase in Title I funding

per pupil was associated with a 13.3 percentage-point increase in the likelihood of having two to

23 In the “restricted quantile model,” we retain dummies for the top two (of twenty) quantiles from the first specification, but replace the rest of the quantile indicators with decile indicators; the 90th percentile of the child poverty distribution is 58.4 percent, the 95th percentile is 66.1 percent, and the maximum poverty rate is 96.1 percent. Note that estimates with the full set of poverty dummies from the first specification are similar in magnitude, but less precise. 24 Estimates from the first-stage model are shown in Appendix Table II.

18

six percent of blacks in desegregated schools (column (8)), and a 9.7 percentage point decline in

the likelihood of having less than two percent of blacks in desegregated schools (column (4)).

We cannot reject the hypothesis that the reduction in the probability of being in the less than two

percent bin was the same as the increase in the probability of being in the two to six percent bin.

By comparison, the OLS estimates, shown in Panel B, are the same sign, but smaller in

magnitude and not statistically significant. This suggests that districts that desegregated less also

experienced larger enrollment declines. For the outcomes discussed below, this general pattern of

findings for the OLS and TSLS estimates persists.25

As expected, the controls, Wd,pre, change the TSLS estimates little. Notably, the controls

for per-pupil spending in the early 1960s – the one district characteristic correlated with the

instrument (Table II) – do not significantly improve the fit of the model.26 Otherwise, the

coefficients on the controls are consistent with our expectations.27 Furthermore, across

specifications, the coefficients on the poverty dummies indicate that higher poverty districts

desegregated less, all else equal. The direction of this correlation notably works against finding

any effect of financial incentives on desegregation. Our empirical approach thus tests whether

the relationship between poverty and desegregation is less negative in higher spending states,

where districts had larger grants holding poverty constant.

Panel A of Table IV presents TSLS estimates of the effect of potential Title I funding per

pupil from our preferred specification (columns (4) and (8) in Table III, with the most complete

set of district controls) for the full distribution of student segregation in 1966. The estimated

25 The TSLS estimates may be larger than the OLS estimates because OLS estimates are either attenuated by measurement error or biased due to endogeneity of enrollment, which is the denominator for per-pupil Title I funding. 26 F-statistics on the pre-program per-pupil current expenditure quintile dummies are 0.936 and 0.875 for the TSLS specifications shown in columns (3) and (4) and 1.274 and 1.242 for the TSLS specifications shown in columns (7) and (8). 27 For example, consistent with the findings of Cascio, et al. (2008), districts with the highest Thurmond vote shares in 1948 and districts with higher early 1960s black enrollment were significantly more likely to have engaged in only token desegregation in 1966.

19

effects of Title I funding per pupil on the likelihood of having between six and 20 percent of

black students in all-black schools are close to zero (columns (3) and (4)), and while uniformly

negative, the TSLS estimates for bins at or above 20 percent (columns (5) through (7)) are small

and not statistically significant. The final column of Table IV shows the effect of Title I funding

on the average percent of blacks in desegregated schools; the coefficient of interest is small and

statistically insignificant. Table IV suggests, therefore, that while the financial incentive moved

districts somewhat beyond tokenism, it did not have a large effect on blacks’ overall exposure to

whites.28

Figure III presents a more detailed version of the results in Panel A of Table IV

graphically, plotting TSLS coefficients on Title I funding per pupil from the same specification

shown in the table, but where the dependent variables are indicators for whether a district fell

into narrower, two-percentage point bins across the entire distribution of student desegregation.

The dashed lines represent the 95 percent confidence intervals on these estimates. The figure

shows that the financial incentive affected behavior only in the lower tail of the distribution of

desegregation.

If our empirical strategy has uncovered the causal effects of financial pressure on

desegregation, it should be the case that districts eventually offered relatively large per-pupil Title

I grants had similar desegregation patterns (reflecting similar unobserved propensities to

desegregate) as other districts prior to the introduction of financial incentives. Consistent with

this, there is no significant relationship between Title I funds at risk and student desegregation

across its entire distribution in 1964, as shown in Panel B of Table IV.

28 Of course, even a small movement from very low levels of desegregation to somewhat more desegregation would change exposure of blacks to whites. The estimates in the first two columns suggest that a $100 larger grant per pupil shifted about 10 percent of districts from less than two to two to six percent of blacks in desegregated schools; using the mid-points of the ranges, this amounts to an increase in the percent of blacks in desegregated schools of about 0.3 ((4-1)*0.1) percentage points. The estimate in column (8) is not precise enough to pick up such an effect.

20

V.B. Effects of Financial Pressure on Teacher Desegregation and Court Supervision

Receipt of federal funding rested not only on meeting threshold levels of student

desegregation, but also on desegregating faculties. In addition, CRA gave districts more incentive

to desegregate on their own without court supervision. The first set of columns in Table V shows

the results of estimating the same models with an indicator equal to one if the district had any

black teachers on faculties with white teachers as the dependent variable. The TSLS estimates are

quite stable across the four specifications, changing little in magnitude as controls are added. The

point estimates are positive and significant at the 10 percent level, suggesting that all else equal,

each additional $100 in potential per-pupil Title I grant increased the probability that a district’s

teaching faculty would be desegregated by 7.5 percentage points.

The second panel of Table V presents estimates of the effects of potential Title I funding

per pupil on whether a district complied with CRA through submission of a court order. Recall

that submission of a court order, as opposed to a voluntary plan, was a sign of resistance to

desegregation, signaling that a district was sued for failure to desegregate sufficiently on its own.

For this outcome, there is a strong negative effect of the financial incentive that changes little in

magnitude across specifications; each additional $100 in per-pupil Title I funding reduced the

probability of being under court order by 6.9 percentage points.

V.C. Additional Robustness Checks

If the size of the Title I grant at risk affected the likelihood that a district would do

enough to comply with CRA, we would expect districts with larger grants to have been less likely

to have had their federal funding deferred or terminated (an indicator that DHEW considered

them out of compliance). Table VI shows the effects of Title I funds on this outcome. The

coefficient of interest is negative, as expected, but not statistically significant. The imprecision of

these results is unsurprising considering the previously discussed inability of DHEW to defer or

21

terminate funds precisely according to its own policy guidance. We are reassured, however, to

see that there is no statistically significant positive relationship. School boards had little

information about the likely strength of enforcement ex ante, so it is not surprising that their

desegregation activity lined up with the DHEW requirements even though enforcement was

imperfect.

Above, we showed that potential Title I grants were uncorrelated with pre-program

student desegregation on the intensive margin, helping to rule out the possibility that the size of

grants was correlated with unobserved tastes for segregation. However, in 1964, most of the

variation in desegregation activity was on the extensive margin; as Panel B of Table IV shows, all

but the first bin was nearly empty in 1964. In Table VII, we present TSLS estimates from a

version of Equation (1′′) where the dependent variable is an indicator for whether a district had

any black students in school with white students in 1964.29 The estimate is marginally significant

in two specifications (columns (1) and (3)), but not in our preferred specification with the full set

of controls (column (4)). Moreover, the coefficient is negative in all specifications, implying that if

anything, districts with larger per-pupil Title I grants were less inclined to desegregate. This

suggests that we may be underestimating the effects of financial incentives.30 Following the same

logic, we also explore whether financial pressure had an “effect” on the likelihood that a district

was under court order in 1964. The results of this exercise are shown in the remaining columns

of Table VII. There was no significant relationship, suggesting that the Title I funding formula

did not target districts that were more likely to desegregate on their own.31

29 We would perform a similar robustness check for teacher desegregation, but no school districts in our estimation sample had any teacher desegregation in 1964. 30 This substantive finding holds for desegregation in each of 1961 through 1963 as well, suggesting that our estimates are not biased from reversion to the mean. 31 Because districts under court order in 1964 were automatically in compliance with CRA, they were not at risk to change their behavior from financial pressure and hence are omitted from our main analysis. The findings in Table VII thus show that this restriction in our sample does not lead to a sample selection bias.

22

The strength of opposition to desegregation—and therefore the “price” required to meet

desegregation targets—most likely varied across districts. Most notably, districts with stronger

support for Strom Thurmond or a higher share of black enrollment might be expected to have

responded less to the same incentive. Unfortunately, when we estimate the model separately for

“high” and “low” Thurmond vote share (or black share) subsamples, our results (not shown)

become sufficiently imprecise as to rule out rejection of even sizeable differences across

subsamples.32

VI. Long-Run Effects of Early Financial Incentives

The results presented in the previous section show that financial incentives in 1966

contributed to a shift away from token desegregation the same year as well as a striking increase

in “voluntary,” rather than court-supervised, desegregation. Did financial incentives under Title I

and CRA continue to matter for desegregation in the years that followed? Unfortunately, a

change to the Title I funding formula in 1967 eliminated the cross-state variation in grants per

poor child central to our identification strategy, making it impossible for us to study the direct

effects of later incentives. 33 However, we can ask whether having a bigger grant “early” (in 1966)

affected outcomes in later years. That is, did early incentives influence how much districts

desegregated and how they got there? Addressing this question helps to fill out our

understanding of how the “carrot” of federal funding interacted with the more well-studied

“stick” of court supervision in desegregating schools over a longer time horizon.

32 Although the results from the fully interacted (split sample) model (on either a high Thurmond vote share indicator or a high black enrollment share indicator) are too imprecise to be informative, we also estimate the main model adding in the same model terms for both (1) the grant interacted with an indicator for high Thurmond share and (2) the grant interacted with an indicator for high black enrollment share. The results for student segregation suggest that districts with stronger segregationist preferences (high Thurmond share) were less responsive to the incentive; there are no significant differences in responsiveness by Thurmond share for other outcomes or by black share for any outcome. 33 The formula was changed so that stategrant was half of lagged per-pupil expenditure in the state or the national average, whichever figure was larger; the latter applied to all but one state in our sample, eliminating essentially all cross-state variation in stategrant.

23

Table VIII shows results from estimating the preferred specification for segregation and

court supervision outcomes for selected years between 1968 and 1976. The measure of

segregation relevant in 1966—the share of blacks in desegregated schools—increasingly fails to

capture the relevant margins of desegregation in later years, as all-black schools were virtually

eliminated by 1970 (Figure I). We therefore use the dissimilarity index as our measure of

segregation for later years.34 Columns (1) through (4) show that the 1966 grant did not have a

significant effect on segregation in later years. However, districts with larger incentives to

desegregate voluntarily early were about 10 percentage points less likely to be under court order

in 1968 for each additional hundred dollars in the potential 1966 Title I grant per pupil (columns

(5) to (8)). This effect is also present in 1970, but faded over the 1970s; by 1976, the effect is

negative but relatively small and not statistically significant.

Thus, early financial incentives to desegregate voluntarily appear to have set some

districts on a path that reduced the likelihood of becoming the target of litigation in the medium

term. Yet, early incentives did not translate to lower segregation in either the medium or long

term. Said differently, early financial incentives had a somewhat lasting effect on how districts

fashioned desegregation plans – whether on their own or under the supervision of the courts –

but not these plans’ ultimate success.

These findings provide indirect evidence on the growing importance of the courts in

desegregating schools. Through 1970, districts with stronger financial incentives to desegregate

early achieved comparable levels of desegregation even though they were significantly less likely

to be court-supervised. During this period, DHEW guidelines and the legal standard applied by

the courts were similar; both required that steps be taken to eliminate the dual system of

34 The dissimilarity index is a measure of racial balance and can be interpreted as the share of students who would have to be transferred to another school in order to replicate the racial composition of the entire district in each school.

24

identifiably black and white schools. The 1971 Supreme Court decision in Swann fundamentally

strengthened the standard applied by the courts, calling for racial balance and authorizing the use

of busing, at the same time that the Nixon Administration backed off CRA enforcement

(Halpern, 1995; Orfield, 2000). In light of these changes, it is perhaps not surprising that the

share of districts under court supervision rose dramatically (Figure I). After 1970, districts with

stronger early financial incentives to desegregate continued to achieve levels of desegregation

comparable to those of their weaker early incentive counterparts but were no longer less likely to

be under court order.

VII. Conclusions

We have shown in this paper that the substantial federal funds offered through Title I of

the Elementary and Secondary Education Act of 1965 provided an effective incentive for school

districts to comply with the Civil Rights Act by the 1966-67 school year. Districts with more

federal funding on the line were more likely to change from behavior that would have clearly

been out of compliance with CRA—having less than two percent of their black students in

desegregated schools, and having fully segregated teaching faculties—to behavior for which most

districts would have been judged compliant—having two to six percent of their black students in

desegregated schools, and having at least some desegregation of their teaching faculties. Districts

facing larger incentives were also less likely to be under the supervision of the courts in 1966,

consistent with greater propensity to desegregate voluntarily. Our analysis examines school

district behavior on the margins upon which DHEW evaluated compliance; if school boards

were rational decision-makers, these are the margins which should matter.

These findings naturally prompt larger economic and historical questions. First, how

much of the massive change in school segregation can be explained by financial incentives under

Title I and CRA? Based on our results, we estimate that these incentives explain about forty

25

percent of the shift away from token desegregation between 1964 and 1966.35 While the

conditional nature of federal grants after CRA did not substantially change the exposure of the

average black student to white students in the short run, this policy shift required districts to

implement more than the nominal voluntary desegregation plans they had submitted earlier,

setting the stage for still more substantial reductions in segregation in the following years. The

“carrot” of financial incentives also reduced the need for the “stick” of court mandates through

the early 1970s. Over time, the importance of the financial incentives studied here faded as the

courts got tougher and the Nixon Administration relaxed CRA enforcement.

Our estimates indicate that school boards could be paid to desegregate, but their

willingness to pay for segregated schools was substantial: The average district was about 10

percent less likely to have only token desegregation for every $100 of potential grant per pupil;

thus, on average a potential grant of $1,000 per pupil would have been required per district

moved beyond token desegregation.36 This is large relative to school budgets at the time,

representing on average almost 60 percent of per-pupil expenditure in the South in the early

1960s.

The existing literature estimating willingness to pay for segregation has drawn inferences

from preferences revealed through the housing market. We calculate that our estimate of

willingness to pay implies that house prices in a school district with “just enough” (two to six

percent of blacks in desegregated schools) desegregation should have been about 1.3 percent

lower compared to a district with token desegregation.37 This estimate is strikingly similar to that

35 We use our estimates to arrive at this figure by comparing the likelihood of having zero to two percent of blacks in desegregated schools with the average potential Title I grant as opposed to no such grant. 36 Because of the sizeable uncertainty surrounding future DHEW policy guidelines for CRA compliance, the size of future Title I grants, and what the courts would require in future years, we interpret our results as identifying the effect of one year’s potential grant amount on one year’s segregation policy, rather than as the effect of an expected stream of future payments. 37 To arrive at this figure, we rely on additional data from the 1960 Census public use sample. We estimate that it took on average $1,000 per pupil (in 2007 dollars) to move a district beyond token desegregation. In the states in our

26

of Clotfelter (1975), the only other paper to date generating an estimate of willingness to pay

based on school segregation in the South in the 1960s. Using data from Atlanta, he found that a

three percentage point increase in black enrollment share in the assigned high school (roughly

comparable to the change in black share on the margin we examine) was associated with a 1.4

percent decline in house prices between 1960 and 1970. Kane, Riegg and Staiger (2006) also

investigate willingness to pay for school segregation in the South, exploiting court-ordered school

assignment boundaries in Mecklenburg County, North Carolina in the 1990s. Their results are

directly comparable and again quite similar in magnitude to Clotfelter’s.38 We refer the reader to

Boustan (2007) and Bayer, Ferreira and McMillan (2007) for estimates of willingness to pay for

segregation in the North.39

The fundamental question we address in this analysis, the power of conditional grants to

affect local behavior, remains important today. The No Child Left Behind Act continues to use

Title I grants as an important carrot, now to encourage states to adopt accountability systems and

districts to strive to meet the standards created by those systems; more generally, the federal

government continues to use conditional grants to affect a wide range of local and state policies

beyond education. This analysis shows that conditional grants can induce local governments to

take even the most politically controversial actions.

sample, there were 0.811 public school children per household, implying a cost of $811 per household. The median price of an owner-occupied dwelling in the South in 1960 was $61,291 (again in 2007 dollars). On average, therefore, the “price” of moving a district beyond tokenism was about 1.3 percent of the median home value. Absent the grant, if districts were forced to desegregate as much as they did, home prices would have fallen by that amount. If districts (correctly) anticipated imperfect enforcement of the DHEW guidelines, then our estimates will be lower bounds of willingness to pay for segregation. 38 They find that a 10 percentage point increase in the percent of black students at the assigned high school was associated with a 4.2 percent decline in home price. Their results are weaker for middle schools and insignificant for elementary schools. 39 Boustan uses data from 1960 through 1980 in 32 non-Southern metropolitan areas and estimates that mandatory busing was associated with a 5.3 percent reduction in housing values. Bayer, Ferreira and McMillan, using 1990 data from the San Francisco Bay Area, find that after controls for neighborhood quality are included, the racial composition of a neighborhood no longer affects its housing values. They do not examine the effects of the racial composition of schools.

27

References

Almond, Chay and Michael Greenstone. 2006. “Civil Rights, the War on Poverty, and Black-White Convergence in Infant Mortality in the Rural South and Mississippi.” MIT Department of Economics Working Paper No. 07-04.

Alt, James E. 1994. “The Impact of the Voting Rights Act on Black and White Voter

Registration in the South” in Chandler Davidson and Bernard Grofman, Eds., Quiet Revolution in the South: The Impact of the Voting Rights Act, 1965-1990. Princeton, NJ: Princeton University Press, 351-377.

Ashenfelter, Orley, William Collins, and Albert Yoon. 2006. “Evaluating the Role of Brown v.

Board of Education in School Equalization, Desegregation, and the Income of African Americans.” American Law and Economics Review 8(2): 213-248.

Bailey, Stephen K. and Edith K. Mosher. 1968. ESEA: The Office of Education Administers a Law.”

Syracuse, NY: Syracuse University Press. Bayer, Patrick, Fernando Ferreira, and Patrick McMillan. 2007. “A Unified Framework for

Measuring Preferences for Schools and Neighborhoods.” Journal of Political Economy 115 (4): 588-638.

Boozer, Michael, Alan Krueger, and Shari Wolkon. 1992. “Race and School Quality Since Brown

v. Board of Education.” Brookings Papers on Economic Activity, Microeconomics, 269-326. Boustan, Leah Platt. 2007. “Escape from the City? The Role of Race, Income, and Local Public

Goods in Post-War Suburbanization.” NBER Working Paper No. 13311. Cameron, A. Colin, Jonah B. Gelbach, and Douglas L. Miller. 2007. “Bootstrap-Based

Improvements for Inference with Clustered Errors.” NBER Technical Working Paper No. 344.

Cascio, Elizabeth, Nora Gordon, Ethan Lewis, Sarah Reber. 2008. “From Brown to Busing.”

Journal of Urban Economics 64(2): 296-325. Card, David and Alan B. Krueger. 1992. “School Quality and Black-White Relative Earnings: A

Direct Assessment.” Quarterly Journal of Economics 107 (1): 151-200. Clotfelter, Charles T. 1975. “The Effect of School Desegregation on Housing Prices.” Review of

Economics and Statistics 57: 395-404. Clotfelter, Charles T. 2004. After Brown: The Rise and Retreat of School Desegregation. Princeton, NJ:

Princeton University Press. Donohue, John J, III, James J. Heckman, and Petra E. Todd. 2002. “The Schooling of Southern

Blacks: The Roles of Legal Activism and Private Philanthropy, 1910-1960.” Quarterly Journal of Economics 117(1): 225-68.

28

Halpern, Stephen C. 1995. On the Limits of the Law: The Ironic Legacy of Title VI of the 1964 Civil Rights Act. Baltimore: the Johns Hopkins University Press.

Kane, Thomas J., Stephanie K. Riegg, and Douglas O. Staiger. 2006. “School Quality,

Neighborhoods, and Housing Prices.” American Law and Economics Review 9(2): 183-212. Orfield, Gary. 1969. The Reconstruction of Southern Education: The Schools and the 1964 Civil Rights

Act, New York: Wiley-Interscience. Orfield, Gary. 2000. “The 1964 Civil Rights Act and American Education.” In Legacies of the

1964 Civil Rights Act (Bernard Groffman, ed.), pp. 89-128. Charlottesville and London: University of Virginia Press.

Reber, Sarah J. 2005. “Court-Ordered Desegregation: Successes and Failures in Integration

Since Brown,” Journal of Human Resources, 40(3): 559-590.

Reber, Sarah J. 2007. “From Separate and Unequal to Integrated and Equal? School Desegregation and School Finance in Louisiana,” NBER Working Paper 13192.

Rosenberg, Gerald N. 1991. The Hollow Hope: Can Courts Bring About Social Change? Chicago:

University of Chicago Press. U.S. Commission on Civil Rights. 1968. Political Participation: A study of the participation by Negroes in

the electoral and political processes in 10 Southern States since passage of the Voting Rights Act of 1965. United States Commission on Civil Rights. Washington, DC.

Washington Research Project of the Southern Center for Studies in Public Policy and the

NAACP Legal Defense and Education Fund, Inc. 1969. Title I of ESEA: Is It Helping Poor Children? Washington, DC.

Welch, Finis and Audrey Light. 1987. New Evidence on School Desegregation. Washington, D.C.:

U.S. Commission on Civil Rights.

29

Figure I. Trends in Desegregation and Court Orders

CRA

/ESE

AShare of Districts withStudent Desegregation

Under CourtSupervision

Share of Blacks inDesegregated Schools

0.2

.4.6

.81

1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976

Notes: Authors’ calculations based on Southern Education Reporting Service, Department of Health Education and Welfare, and Office of Civil Rights data. Figures are weighted by school district. See Data Appendix for details.

30

Figure II. Theoretical Predictions

1=s * s~ s

e

1*~ ss = s

e

e* = l+m+f

e*= l+m

A. District with small Federal grant available B. District with large Federal grant available

Notes: e represents expenditure per pupil; l, m and f represent local, state, and federal revenue, respectively; s represents segregation; and presents the threshold level of student segregation at which federal funds are received.

s~ re

31

Figure III. Effect per $100 Grant/Pupil on Each Level of Desegregation, 1966

-.2-.1

0.1

.2

0 10 20 30 40 50

0-50% of Blacks in Desegregated Schools-.2

-.10

.1.2

50 60 70 80 90 100

(continued) 50-100% of Blacks in Desegregated Schools

Notes: The solid line connects TSLS coefficients on the per-pupil Title I grant (in 100s of $2007) from regressions where the dependent variables are a series of dummies for the corresponding two percentage-point range of the percent of blacks in desegregated schools in 1966. The underlying specification also includes state fixed effects, restricted quantile effects for the early 1960s child poverty rate, and the complete set of district and county-level controls. The dotted lines represent the 95 percent confidence intervals on these estimates; standard errors are clustered on county.

32

Mean Std. dev.

Title I Per Pupil (current enrollment, $2007) 277 149Simulated Title I Per Pupil (early 1960s enrollment, $2007) 274 138

Early 1960s Child Poverty % 33.8 17.6Early 1960s Black Enrollment % 37.2 20.21948 Thurmond Vote % 35.6 28.7Early 1960s Enrollment 6122 10999Early 1960s Expenditure per Pupil (100s of $2007) 1708 4261960 County Characteristics: % with High School Degree 26.37 7.36 % Employed in Agriculture 37.30 24.04 Median family income ($1960) 3273.54 966.04 =1 if Urban 0.20 0.40

Number of Districts 912

Table IA Descriptive Statistics: Title I Funding and Other District and County Characteristics

A. Title I Funding

B. Other District and County Characteristics

Notes: Sample includes school districts in Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia. For more information, see text and Data Appendix.

33

1964 1966

Percent Blacks in Desegregated Schools 0.8 18.1(std. dev.) (5.4) (28.7)= 1 if % Black Students in Desegregated Schools is:

Less than 2% 0.947 0.305At least 2% but less than 6% 0.021 0.247At least 6% but less than 10% 0.012 0.087At least 10% but less than 20% 0.011 0.121At least 20% but less than 30% 0.003 0.055At least 30% but less than 50% 0.002 0.05950% or more 0.003 0.127

Number of Districts 903 912

=1 if Any Black Teachers Work with White Teachers 0.000 0.730Number of Districts 912 873

Funds Deferred or Terminated not applicable 0.203Under Court Order not applicable 0.095Number of Districts 903 912

B. Title VI CRA Compliance

A. Segregation Choices

Table IBDescriptive Statistics: Measures of Segregation Policy

Notes: Sample includes school districts in Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia. For more information, see text and Data Appendix.

34

Dependent Variable: (1) (2)

1948 Thurmond % -1.625 -0.540(1.780) (1.530)

Early 1960s Black Enrollment % -1.464 0.769(1.633) (1.442)

Ln Early 1960s Enrollment -0.145* -0.115(0.0860) (0.0828)

Early 1960s Expenditure per Pupil (100s of $2007) 0.865*** 0.562**(0.233) (0.241)

1960 Share with a High School Degree -0.00107 -0.00661(0.00960) (0.00910)

1960 Median Family Income ($1960) 78.00 -43.41(108.4) (112.4)

1960 Share Employed in Agriculture -0.0346 -0.0165(0.0316) (0.0297)

1960 Urban Indicator 0.136*** 0.103**(0.0518) (0.0494)

Controls:State Fixed Effects X XEarly 1960s Child Poverty %:

Dummies for 20 Quantiles XRestricted Quantile Effects† X

D. County Characteristics

Title I Funding Per Pupil (in 100s of $2007)IV Coefficient (Standard Error) on

Table II Potential Title I Funding, 1966, and Other Determinants of Segregation Policy

A. Proxies for Preferences

B. Proxies for Litigation Threat

C. Potential School Budget

Notes: Each entry gives the IV coefficient (standard error) on Title I funding per pupil (100s of $2007) in a model predicting the district or county characteristic listed. The instrument for Title I funding per pupil is simulated Title I funding per-pupil (also in 100s of $2007); see text. All regressions contain 912 district-level observations. Standard errors are clustered on county. *** p<0.01, ** p<0.05, * p<0.1 † Dummies for nine deciles and the top two of twenty quantiles in the first specification.

35

(1) (2) (3) (4) (5) (6) (7) (8)

Mean of Dependent Variable

Title I Per Pupil (in 100s of $2007) -0.0999** -0.0798* -0.0983** -0.0965** 0.139*** 0.154*** 0.146*** 0.133***(0.0488) (0.0436) (0.0445) (0.0446) (0.0475) (0.0431) (0.0437) (0.0437)

R-Squared 0.371 0.368 0.399 0.400 0.083 0.069 0.159 0.167First Stage F -Stat 305.4 423.2 461.1 409.5 305.4 423.2 461.1 409.5

Title I Per Pupil (in 100s of $2007) -0.0561 -0.0476 -0.0601* -0.0575 0.0205 0.0357 0.0483 0.0397(0.0392) (0.0365) (0.0346) (0.0350) (0.0352) (0.0323) (0.0327) (0.0322)

R-Squared 0.372 0.369 0.400 0.401 0.094 0.081 0.166 0.174

Controls:State Fixed Effects X X X X X X X XEarly 1960s Child Poverty %:

Dummies for 20 Quantiles X XRestricted Quantile Effects† X X X X X X

Early 1960s Black Enrollment % (Decile Dummies) X X X X1948 Thurmond Vote % (Quintile Dummies) X X X XLn Early 1960s Enrollment X X X XEarly 1960s Exp. per Pupil (Quintile Dummies) X X X X1960 County Characteristics‡ X X

Number of Districts 912 912 912 912 912 912 912 912

B. Ordinary Least Squares

A. Two-Stage Least Squares

Table IIIPotential Title I Funding and Student Desegregation, 1966

Less Than 2%, 1966 At least 2% but less than 6%, 1966 = 1 if % Black Students in Desegregated Schools is:

0.305 0.247

Notes: Each column in each panel gives results from a different regression. The unit of observation is a school district; see text and Data Appendix for descriptions of the sample. In addition to the controls listed, all models include as an explanatory variable an indicator for whether the district had any student desegregation in 1964. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1 † Dummies for nine deciles and the top two of twenty quantiles in the first specification.‡1960 county characteristics include share with a high school degree, share employed in agriculture, median family income, indicator for urban.

36

< 2% 2 to < 6% 6 to <10% 10 to <20% 20 to <30% 30 to <50% 50%+(1) (2) (3) (4) (5) (6) (7) (8)

Mean of Dependent Variable 0.305 0.247 0.0866 0.121 0.0548 0.0592 0.127 18.1

Title I Per Pupil, 1966 (in 100s of $2007) -0.0965** 0.133*** 0.00840 0.00126 -0.0150 -0.00486 -0.0264 -1.69(0.0446) (0.0437) (0.0232) (0.0348) (0.0209) (0.0308) (0.0201) (1.45)

R-Squared 0.400 0.167 0.080 0.108 0.090 0.096 0.491 0.637

Number of Districts 912 912 912 912 912 912 912 912

Mean of Dependent Variable 0.947 0.0210 0.0122 0.0111 0.00332 0.00221 0.00332 0.8

Title I Per Pupil, 1966 (in 100s of $2007) 0.00227 0.00500 -0.00678 -0.00435 0.00551 0.00287 -0.00452 -0.21(0.0142) (0.00970) (0.00675) (0.00776) (0.00371) (0.00272) (0.00423) (0.31)

R-Squared 0.268 0.119 0.100 0.081 0.044 0.051 0.067 0.141

Number of Districts 903 903 903 903 903 903 903 903

A. 1966

B. 1964

Table IVTSLS Estimates of the Effect of Potential Title I Funding on the Distribution of Student Desegregation, 1964 and 1966

= 1 if % Black Students in Desegregated Schools is: % of Blacks in Desegregated Schools

Notes: Each column in each panel gives results from a different regression. The unit of observation is a school district; see text and Data Appendix for descriptions of the sample. The specification is the same as that shown in columns (4) and (8) of Table III. In addition to the controls listed, all models in Panel A (1966) include as an explanatory variable an indicator for whether the district had any student desegregation in 1964. The F -statistic on the instrument in the underlying first-stage regressions is 409.5 in both Panels A and B. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1

37

(1) (2) (3) (4) (5) (6) (7) (8)

Mean of Dependent Variable

Title I Per Pupil (in 100s of $2007) 0.0764* 0.0784* 0.0721* 0.0745* -0.0580** -0.0644** -0.0679** -0.0692***(0.0462) (0.0419) (0.0433) (0.0425) (0.0289) (0.0264) (0.0272) (0.0265)

R-Squared 0.339 0.332 0.353 0.356 0.277 0.271 0.328 0.334First Stage F -Stat 281.4 380.1 415.5 373.6 305.4 423.2 461.1 409.5

Title I Per Pupil (in 100s of $2007) -0.000526 0.00603 0.0155 0.0185 -0.0345* -0.0408** -0.0309* -0.0322*(0.0377) (0.0355) (0.0366) (0.0362) (0.0185) (0.0173) (0.0184) (0.0181)

R-Squared 0.343 0.336 0.356 0.358 0.278 0.272 0.330 0.336

Controls:State Fixed Effects X X X X X X X XEarly 1960s Child Poverty %:

Dummies for 20 Quantiles X XRestricted Quantile Effects† X X X X X X

Early 1960s Black Enrollment % (Decile Dummies) X X X X1948 Thurmond Vote % (Quintile Dummies) X X X XLn Early 1960s Enrollment X X X XEarly 1960s Exp. per Pupil (Quintile Dummies) X X X X1960 County Characteristics‡ X X

Number of Districts 873 873 873 873 912 912 912 912

A. Two-Stage Least Squares

B. Ordinary Least Squares

= 1 if Any Black Teachers in Same School as White Teachers

0.730

Table VThe Impact of Potential Title I Funding on Teacher Desegregation and Court Supervision, 1966

= 1 if Under Court Order

0.0954

Notes: Each column in each panel gives results from a different regression. The unit of observation is a school district; see text and Data Appendix for descriptions of the sample. In addition to the controls listed, all models also include as an explanatory variable an indicator for whether the district had any student desegregation in 1964. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1† Dummies for nine deciles and the top two of twenty quantiles in the first specification.‡1960 county characteristics include share with a high school degree, share employed in agriculture, median family income, indicator for urban.

38

(1) (2) (3) (4)

Mean of Dependent Variable

Title I Per Pupil (in 100s of $2007) -0.0515 -0.0290 -0.0334 -0.0339(0.0433) (0.0392) (0.0395) (0.0393)

R-Squared 0.232 0.225 0.257 0.262First Stage F -Stat 305.4 423.2 461.1 409.5

Title I Per Pupil (in 100s of $2007) 0.0023 0.0129 0.0093 0.0077(0.0326) (0.0310) (0.0308) (0.0300)

R-Squared 0.234 0.227 0.259 0.263

Controls:State Fixed Effects X X X XEarly 1960s Child Poverty %:

Dummies for 20 Quantiles XRestricted Quantile Effects† X X X

Early 1960s Black Enrollment % (Decile Dummies) X X1948 Thurmond Vote % (Quintile Dummies) X XLn Early 1960s Enrollment X XEarly 1960s Exp. per Pupil (Quintile Dummies) X X1960 County Characteristics‡ X

Number of Districts 912 912 912 912

B. Ordinary Least Squares

A. Two-Stage Least Squares

Table VIPotential Title I Funding and Recorded Compliance with Title VI of the Civil Rights Act, 1966

=1 if Funds Deferred or Terminated

0.203

Notes: Each column in each panel gives results from a different regression. The unit of observation is a school district; see text and Data Appendix for descriptions of the sample. In addition to the controls listed, all models also include as an explanatory variable an indicator for whether the district had any student desegregation in 1964. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1 † Dummies for nine deciles and the top two of twenty quantiles in the first specification.‡1960 county characteristics include share with a high school degree, share employed in agriculture, median family income, indicator for urban.

39

(1) (2) (3) (4) (5) (6) (7) (8)

Mean of Dependent Variable

Title I Per Pupil (in 100s of $2007) -0.0530* -0.0401 -0.0440* -0.0319 -0.0121 -0.0245 -0.0206 -0.00976(0.0281) (0.0260) (0.0257) (0.0255) (0.0227) (0.0219) (0.0212) (0.0195)

R-Squared 0.375 0.368 0.429 0.447 0.083 0.075 0.157 0.176First Stage F -Stat 309.9 426.8 463.3 410.7 313.0 438.8 471.0 412.1

Title I Per Pupil (in 100s of $2007) -0.0500*** -0.0426** -0.0301* -0.0220 -0.00674 -0.0136 -0.00123 0.00457(0.0184) (0.0173) (0.0183) (0.0176) (0.0153) (0.0152) (0.0149) (0.0145)

R-Squared 0.375 0.368 0.429 0.447 0.083 0.075 0.158 0.177

Controls:State Fixed Effects X X X X X X X XEarly 1960s Child Poverty %:

Dummies for 20 Quantiles X XRestricted Quantile Effects† X X X X X X

Early 1960s Black Enrollment % (Decile Dummies) X X X X1948 Thurmond Vote % (Quintile Dummies) X X X XLn Early 1960s Enrollment X X X XEarly 1960s Exp. per Pupil (Quintile Dummies) X X X X1960 County Characteristics‡ X X

Number of Districts 912 912 912 912 968 968 968 968

A. Two-Stage Least Squares

B. Ordinary Least Squares

Table VIIPotential Title I Funding and Pre-Program Outcomes

= 1 if Any Student Desegregation, 1964 =1 if Under Court Order, 1964

0.174 0.0579

Notes: Each column in each panel gives results from a different regression. The unit of observation is a school district. The sample in columns (5)-(8) includes districts under court order in 1964. The districts are not included in our main estimation sample; see text and Data Appendix. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1 † Dummies for nine deciles and the top two of twenty quantiles in the first specification.‡1960 county characteristics include share with a high school degree, share employed in agriculture, median family income, indicator for urban.

40

1968 1970 1972 1976 1968 1970 1972 1976(1) (2) (3) (4) (5) (6) (7) (8)

Mean of Dependent Variable 0.716 0.234 0.206 0.195 0.268 0.294 0.488 0.485

Title I Per Pupil 0.00914 -0.0155 -0.0135 -0.0212 -0.0968** -0.119*** -0.0653 -0.0242(in 100s of $2007) (0.0184) (0.0175) (0.0153) (0.0157) (0.0388) (0.0387) (0.0414) (0.0431)

R-Squared 0.634 0.462 0.491 0.521 0.616 0.514 0.473 0.454

Additional Controls:State Fixed Effects X X X X X X X XEarly 1960s Child Poverty %:

Restricted Quantile Effects† X X X X X X X XEarly 1960s Exp. per Pupil (Quintile Dummies) X X X X X X X X1960 County Characteristics‡ X X X X X X X X

Number of Districts 837 858 858 792 773 864 866 880

Two-Stage Least Squares

Table VIIIPotential Title I Funding and Long-Term Outcomes

Dissimilarity Index =1 if Under Court Order

Notes: Each column in each panel gives results from a different regression. The unit of observation is a school district. Districts that experienced a boundary change between 1969 and 1976 are excluded from the analysis. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1 † Dummies for nine deciles and the top two of twenty quantiles in the first specification.‡1960 county characteristics include share with a high school degree, share employed in agriculture, median family income, indicator for urban.

41

State: stategrant 1966 ($2007) Number of districts

Alabama 798 83Arkansas 871 121Florida 1161 56Georgia 906 146Louisiana 891 59Mississippi 578 97North Carolina 862 124South Carolina 647 86Tennessee 840 67Virginia 844 73

Total 912

Appendix Table IThe State Grant Component of Title I Funding, 1966

Notes: See Section III for a description of the Title I funding formula.

42

(1) (2) (3) (4)

Simulated Per-Pupil Title I Grant 1.128*** 1.128*** 1.127*** 1.125***(0.0646) (0.0548) (0.0525) (0.0556)

R-Squared 0.972 0.971 0.973 0.973F -Stat on Instrument 305.4 423.2 461.1 409.5

Controls:State Fixed Effects X X X XEarly 1960s Child Poverty %:

Dummies for 20 Quantiles XRestricted Quantile Effects† X X X

Early 1960s Black Enrollment % (Decile Dummies) X X1948 Thurmond Vote % (Quintile Dummies) X XLn Early 1960s Enrollment X XEarly 1960s Exp. per Pupil (Quintile Dummies) X X1960 County Characteristics‡ X

Number of Districts 912 912 912 912

Appendix Table IIFirst Stage Regressions: All Specifications, Main Estimation Sample

Per-Pupil Title I Grant

Notes: Each column in each panel gives results from a different regression. Both simulated and actual per-pupil Title I grants are in 100s of $2007. The unit of observation is a school district; see text and Data Appendix for descriptions of the main estimation sample. In addition to the controls listed, all models include as an explanatory variable an indicator for whether the district had any student desegregation in 1964. See Table III for further description of controls. Standard errors, in parentheses, are clustered on county. *** p<0.01, ** p<0.05, * p<0.1 † Dummies for nine deciles and the top two of twenty quantiles in the first specification.‡1960 county characteristics include share with a high school degree, share employed in agriculture, median family income, indicator for urban.

43

State: Full SampleWith Early 1960s % Black in Enrollment

Early 1960s % Black in Enrollment 3% to

97%Not Under Court

Order in 1964

Has Other Explanatory

Variables Has Outcomes

Alabama 113 113 100 92 92 83Arkansas 387 205 148 146 146 121Florida 67 67 67 61 61 56Georgia 193 193 171 167 161 146Louisiana 65 65 65 62 61 59Mississippi 134 130 117 113 108 97North Carolina 152 151 135 129 126 124South Carolina 93 93 93 88 88 86Tennessee 146 145 80 69 67 67Virginia 126 126 107 88 82 73

Total 1,476 1,288 1,083 1,015 992 912

Appendix Table IIIDefinition of Sample

Notes: See Data Appendix for description of variables and sources.

44

Data Appendix I. Sources and Key Variables Complete citations of all documents referenced below are provided under Data References. A. Data on Title I Funding and Child Poverty Title I funding allocations were initially made at the county level; states then allocated grants to districts within each county using data from one of a number of data sources, which included Census tract-level data tabulated for minor governmental subdivisions, supplied to state governments by the Office of Education, and administrative data from the AFDC program (Bailey and Mosher, 1968). We do not know which data source states used for these purposes, but we do observe the district-level allocation of funds in the first year of the program, 1965-66. We use these district-level allocations to estimate district-level allocations for 1966-67. In particular, district-level Title I allocations per pupil for 1966-67 were calculated in three steps. First, we calculated the share of the county-level Title I allocation that each district received in 1965-66. In the second step, this share was multiplied by the 1966-67 county-level allocation to arrive at an estimate the 1966-67 district-level allocation; thus, we assume that districts were entitled to a constant share of the county Title I allocation in two consecutive years.1 We divide this figure by current (fall 1966) enrollment to arrive at the Title I grant per pupil, pptid1966; sources of data on enrollment are discussed below. Data on both district-level Title I allocations in 1965-66 and county-level Title I allocations in 1966-67 were hand-entered from a U.S. Senate publication.2 This publication also gives the number of Title I eligibles in 1966-67 in each county in each of the following categories: (1) in low income families (5-17 year olds in families with incomes less than $2000 in 1960); (2) in families receiving AFDC in excess of $2000; (3) delinquent children; (4) neglected children; and (5) children in foster homes. We estimate the number of Title I eligibles at the district level, eligiblesd1966, as the number of county-level Title I eligibles in 1966-67 times the fraction of the county Title I entitlement for which the district was eligible in 1965-66, calculated as discussed above. The district-level components of Title I eligiblesd1966 are calculated in the same way. In 1965, counts under all categories above were based entirely on data collected prior to the introduction of Title I, but in 1966 only category (1), poord1960, was, and categories (2)-(5) used data from 1965 . For this reason, we calculate the using eligiblesd1965, but use eligiblesd1966 in the calculation of (endogenous) pptid1966:

SIMdppti 1966

1 This seems a reasonable assumption given the infrequency with which states updated the source of data used for the county sub-allocation. 2 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 89-750 (May 1967). Part I, Chapter 3 provided the district-level Title I allocations for 1965-66 and Part 2, Chapter 4 provided the county-level Title I allocations and counts of eligible children for 1966-67.

45

pred

djSIMd enr

poorstategrantppti

,

19601966 ≡ and

1966

19601966

d

djd enr

eligiblesstategrantppti ≡

In practice, this choice makes very little difference in the numerators of the actual and simulated Title I grants per pupil, as in nearly all Southern counties, eligiblesd1966 ≈ poord1960.3 Recall that enrd,pre represents average enrollment in the district in the early-1960s, and enrd1966 represents current (fall 1966) enrollment; see below for a discussion of sources and calculation. The pre-program (early 1960s) child poverty rate is poord1960/enrd,pre. B. Data on District Enrollment and Current Expenditure We have gathered data on total enrollment, enrollment by race, and current expenditure at the school district level prior to 1964 from annual reports of state departments or superintendents of education. Data were entered from print publications.4 The enrollment variable used in our analysis is average fall enrollment in the district between 1960 and 1963.5 The years in which enrollment by race are available differ by state, so we average district percent black across all available years between 1960 and 1963 to arrive at pre-program percent black in enrollment. For states where enrollment by race is not provided in these annual reports, we instead use the 1964 and 1966 SERS Statistical Summaries (North Carolina), 1960 county-level Census data on the racial breakdown of the population of 5 to 17 year olds (Florida, where district boundaries correspond to counties), or 1967 data on school enrollment by race collected by DHEW (Arkansas). Current expenditure was also reported in the state publications; pre-program per-pupil current expenditure is current expenditure divided by enrollment averages over all years from 1960-63 for which data are available. C. County-level data The county-level percent of votes cast for the States’ Rights Democrat, Strom Thurmond, in the 1948 Presidential election was drawn from Electoral Data for Counties in the United States: Presidential and Congressional Races, 1840-1974 (ICPSR Study No. 8611). Data on 1960 median family income, share with a high school degree, share employed in agriculture, and urban status at the county level are drawn from City and County Data Book Consolidated File, County Data 1947-1977 (ICPSR Study No. 7736).

3 In our estimation sample, average ratio of poord1960 to eligiblesd1966 is 0.984, the median ratio of poord1960 to eligiblesd1966 is 0.995, and the minimum ratio of poord1960 to eligiblesd1966 is 0.865. 4 Alabama Department of Education (various years), Arkansas Department of Education (various years), Florida State Superintendent of Public Instruction (various years), Georgia State Department of Education (various years), North Carolina Education Association (various years), South Carolina State Department of Education (various years), State Department of Education of Louisiana (various years), Tennessee Department of Education (various years), Virginia State Board of Education (various years). 5 States report slightly different enrollment concepts in their annual reports, usually fall enrollment or registration, average daily membership or average daily attendance. We use the measure that is most consistently reported within the state over time. Arkansas, Georgia, and North Carolina report average daily attendance rather than enrollment. To make the data more comparable with that for other states, we multiply the ADA by the statewide average ratio of fall enrollment to ADA reported in “Statistics of State School Systems, 1963-64.”

46

D. Desegregation and Plan Type Data by School District, 1964 and Earlier For 1956 through 1964, we have hand-entered data on desegregation and type of desegregation plan (court-ordered or voluntary) from print publications of the Southern Education Reporting Service (SERS), entitled A Statistical Summary, State By State, of School Segregation-Desegregation in the Southern and Border Area from 1954 to the Present.6 These publications give, for all districts desegregated in policy or in practice,7 the number of blacks attending public school with whites, the total number of black children enrolled in public schools, and whether desegregation was court-ordered or undertaken voluntarily by the local school board. Using these data, we are able to construct three variables: an indicator set to one if the district had a court-ordered desegregation plan, an indicator set to one if the district had any blacks enrolled in public schools with whites, and the percent of blacks attending desegregated schools. These figures are plotted in Figure I, and the first two variables for 1964 are used for a robustness check in Table VII. For districts not listed in these publications, we assume that all of these variables are equal to zero.8 It is difficult to assess the credibility of this assumption, since no other agencies collected data on desegregation over the period of interest. SERS data collection strategy is also unclear: according to the 1964 publication, data “were supplied by agencies of the respective states,” but the exact procedure is not described. However, since there were such low rates of desegregation during the period, it was most likely not very onerous to collect the data.9 Two previous uses of SERS data also suggest the credibility of our assumption and the data collected by SERS more generally. First, SERS supplied desegregation data to the U.S. Commission on Civil Rights by contractual agreement (U.S. Commission on Civil Rights, 1966; p. 30). Second, the state-level summaries of desegregation activity are considered the best available data by social scientists and have been previously cited in academic research (e.g., Rosenberg, 1991; Orfield, 2000).

6 We use the data presented in the following versions of this publication: April 15, 1957 (for fall 1956), November 1957 (for fall 1957), October 1958 (for fall 1958), May 1960 (for fall 1959), November 1960 (for fall 1960), November 1961 (for fall 1961), November 1962 (for fall 1962), 1963-64 (for fall 1963), and November 1964 (for fall 1964). Prior to 1963-64, the relevant publication is entitled A Statistical Summary, State-by-State, of Segregation-Desegregation Activity Affecting Southern Schools from 1954 to Present. 7 Districts desegregated in policy but not in practice had freedom of choice plans, where blacks’ option to apply to white schools was not exercised, or court orders that had not yet taken effect. 8 We compiled lists of districts by state and year from reports of district finances, enrollment, and other activities published by state departments of education (see I.B of this Appendix). 9 High rates of desegregation may be why data collected by SERS for the Border region are less complete. It may also be the reason why SERS relied on a U.S. Office of Education Survey in 1966 (see Section I.F of this Appendix).

47

E. Plan Type and CRA Compliance Data by School District, 1966 and Later For December 1966, we have hand-entered information on school districts’ compliance with Title VI of the Civil Rights Act from a print publication of DHEW entitled Status of Compliance: Public School Districts, Seventeen Southern and Border States. This publication provides the type of desegregation plan submitted and whether the plan was approved by DHEW for all school districts. We set the court order indicator to one for districts with approved court-ordered plans and to zero for all other districts. For districts where plans were not approved as of the date of publication, information is provided on whether federal funds to the district have been deferred or terminated. For 1968 and later, compliance and court-order status were provided on the Elementary and Secondary Schools Civil Rights Survey tapes, collected by the Office of Civil Rights. See Cascio, et al. (2008) for more detail on these data. F. Desegregation Data by School District, 1966 and Later For fall 1966, we have hand-entered information on the number of blacks attending public school with any whites from a print publication of SERS entitled Statistical Summary of School Segregation-Desegregation in the Southern and Border States 1966-1967. Most of the data reported were from computer printouts provided to SERS by the Office of Education (OE) from its first survey of school desegregation in the South. The universe of the OE survey is not clear from published documents, but presumably included all districts in the South; SERS reports the response rate at 80 percent. SERS correspondents were able to fill in data for some districts where data in the OE survey were missing. For all districts listed in this publication, we set the desegregation indicator equal to one if any blacks are reported to be in school with whites, and otherwise equal to zero. We also calculate the percent of blacks in desegregated schools by estimating the total number of blacks in the district with fall 1966 district enrollment times fraction black in the district in the early 1960s (see section I.B of this Appendix for sources); the total number of blacks in the district is not reported in the SERS publication. We assume that districts that had their federal funds deferred or terminated in 1966 but were not listed in the 1966 SERS publication had sufficiently high levels of segregation to fall into the first “bin” of student desegregation (less than two percent of blacks in desegregated schools). Although we expect that many of these districts had not desegregated at all (perhaps explaining why they did not report data to the Office of Education), imputing values for the extensive margin of desegregation would require much stronger assumptions than our imputation into the most segregated bin and we therefore retain the missing values on the extensive margin for these districts. Of the 992 districts that had all the other data required for the analysis, 132 did not report the share of blacks in desegregated schools in the 1966 SERS publication; of these, 39 had their funds deferred or terminated and were given imputed values of less than 2 percent of blacks in desegregated schools.10 Thirteen

10 Among districts with funding deferred or terminated for which we do observe the share of blacks in desegregated schools, almost 75 percent have desegregated less than 2 percent of blacks; non-reporting districts

48

districts did not appear in 1966 SERS but had not desegregated at all in later years; we assume these districts had not desegregated in 1966. Our estimation sample therefore includes 912 school districts. The estimates are not sensitive to dropping districts with imputed outcomes. These findings are available from the authors on request. The SERS publication also lists whether black teachers are teaching on desegregated faculties and in some cases how many; we use this information to construct an indicator equal to one if any blacks teach on desegregated faculties and zero otherwise. If a district was not listed in the 1966 publication, the teacher desegregation variable is coded as missing. Student segregation data for 1968-1976 (plotted in Figure I) are based on school-level surveys conducted by the Office of Civil Rights (OCR). See Cascio, et al. (2008) for more detail on these data. II. School District Reorganization Activities School districts both consolidate and, less commonly, split apart during our sample period. We use the state records referenced above to establish a history of these reorganizations. For each year, we construct a crosswalk between the district (id) and the largest (district) unit to which the district is party between 1961 and 1969 (agg_id). We then merge this crosswalk to each data set and collapse key variables (as described below) to the agg_id-year level. For example, if districts A and B merge in 1966 to form district C, we will observe A and B jointly as one observation, geographically identical to and with the same agg_id as district C (C is observed from 1966 on), prior to 1966. And if district X splits into districts Y and Z in 1964, we will observe Y and Z jointly as one observation, geographically identical to and with the same agg_id as X (X is observed only before 1964), beginning in 1964. We aggregate indicator variables to the agg_id-year level as follows. First, we set the value of the indicator for the agg_id equal to one if any districts associated with the agg_id are observed to have the indicator set to one. Second, if all districts associated with the agg_id are observed, and none have the indicator set to one, the indicator for the agg_id set to zero. Third, if not all districts associated with the agg_id are observed (as is sometimes the case in the 1966 SERS or OCR desegregation data), and all observed districts have the indicator set to zero, we code the indicator for the agg_id as missing. For example, in the 1966 SERS, we code an agg_id as desegregated if at least one of its component district ids is observed with any blacks in school with whites, as not desegregated if all districts are observed and none are desegregated, and as missing if no observed districts are desegregated, but not all districts are observed. Numerical variables (the percent of blacks in desegregated schools, enrollment, early 1960s percent black in enrollment, early 1960s child poverty rate) are coded as missing if not all ids associated with an agg_id are observed in the raw data. Where all districts are observed, we sum up all components of the variables (e.g., number of blacks attending desegregated schools, total enrollment, etc.), and calculate values for the agg_id accordingly. For example, in the 1966 SERS, we first calculate the percent of blacks in desegregated schools by first are likely to have less desegregation. We do not impute values for the extensive margin of desegregation for these districts.

49

calculating the number of blacks attending all black schools and the total number of blacks across all districts in the agg_id. We then take the ratio of these sums. Forty-four of the 912 aggregated districts in our estimation sample (4.82 percent) were involved in a consolidation or split at some point over 1961 to 1969. Our findings are robust to omitting these observations. Tables and graphs limited to districts not re-organized between 1961 and 1969 are available from the authors on request.

50

Data References Alabama Department of Education. Annual Report for the Scholastic Year Ending June 30, 1966

and for the Fiscal Year Ending September 30, 1966: Statistical and Financial Data. Montgomery, AL. Reports covering 1960/61-1965/66 school years.

Arkansas Department of Education. Report on House Concurrent Resolution No. 58 of the 63rd

General Assembly, Little Rock, AR. Reports covering 1960/61-1965/66 school years. Clubb, Jerome M., William H. Flanigan, and Nancy H. Zingale. Electoral Data for Counties in

the United States: Presidential and Congressional Races, 1984-1972 [Computer file]. Compiled by Jerome M. Clubb, University of Michigan, William H. Flanigan, University of Minnesota, and Nancy H. Zingale, College of St. Thomas. ICPSR08611-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [producer and distributor], 2006-11-13.

Florida State Superintendent of Public Instruction. Ranking of the Counties. Research Report,

Division of Research. Tallahassee, FL. Reports covering 1962/63, 1964/65, 1965/66 school years.

Georgia State Department of Education. Annual Reports of the Department of Education

to the General Assembly of the State of Georgia. Atlanta, GA. Reports covering 1961/62, 1963/64 and 1965/66 school years.

North Carolina Education Association. Per Pupil Expenditures for Current Expense: Information

Provided by Division of Statistical Services State Department of Public Instruction. Raleigh, NC. Reports covering 1961/62-1965/66 school years.

South Carolina State Department of Education. Annual Report of the State Superintendent

of Education of the State of South Carolina. Columbia, SC. Reports covering 1960/61 – 1965/66 school years.

Southern Education Reporting Service. A Statistical Summary, State-by-State, of Segregation-

Desegregation Activity Affecting Southern Schools from 1954 to Present, Together With Pertinent Data on Enrollment, Teacher Pay, Etc.. Nashville, TN: Southern Education Reporting Service, April 15, 1957.

----. A Statistical Summary, State-by-State, of Segregation-Desegregation Activity Affecting Southern

Schools from 1954 to Present, Together With Pertinent Data on Enrollment, Teacher Pay, Etc.. Second Revision. Nashville, TN: Southern Education Reporting Service, November 1, 1957.

----. A Statistical Summary, State-by-State, of Segregation-Desegregation Activity Affecting Southern

Schools from 1954 to Present, Together With Pertinent Data on Enrollment, Teacher Pay, Etc.. Fifth Printing. Nashville, TN: Southern Education Reporting Service, October 15, 1958.

----. A Statistical Summary, State-by-State, of Segregation-Desegregation Activity Affecting Southern

Schools from 1954 to Present, Together With Pertinent Data on Enrollment, Teachers, Colleges,

51

Litigation, and Legislation. Seventh Revision. Nashville, TN: Southern Education Reporting Service, May 1960.

----. A Statistical Summary, State-by-State, of Segregation-Desegregation Activity Affecting Southern

Schools from 1954 to Present, Together With Pertinent Data on Enrollment, Teachers, Colleges, Litigation, and Legislation. Eighth Revision. Nashville, TN: Southern Education Reporting Service, November 1960.

----. A Statistical Summary, State By State, of Segregation-Desegregation Activity Affecting Southern

Schools from 1954 to the Present, Together with Pertinent Data on Enrollment, Teachers, Colleges, Litigation, and Legislation. Tenth revision. Nashville, TN: Southern Education Reporting Service, November 1961.

----. A Statistical Summary, State By State, of Segregation-Desegregation Activity Affecting Southern

Schools from 1954 to the Present, Together with Pertinent Data on Enrollment, Teachers, Colleges, Litigation, and Legislation. Eleventh revision. Nashville, TN: Southern Education Reporting Service, November 1962.

----. A Statistical Summary, State By State, of School Segregation-Desegregation in the Southern and

Border Area from 1954 to the Present. Thirteenth revision. Nashville, TN: Southern Education Reporting Service, 1963-64.

----. A Statistical Summary, State By State, of School Segregation-Desegregation in the Southern and

Border Area from 1954 to the Present. Fourteenth revision. Nashville, TN: Southern Education Reporting Service, November 1964.

----. A Statistical Summary, State By State, of School Segregation-Desegregation in the Southern and

Border Area from 1954 to the Present. Sixteenth revision. Nashville, TN: Southern Education Reporting Service, February 1967.

State Department of Education of Louisiana. Financial and Statistical Report. Baton Rouge, LA.

Reports covering 1960/61 – 1965/66 school years. Tennessee Department of Education. Annual Statistical Report of the Department of Education.

Nashville, TN. Reports covering 1960/61 – 1965/66 school years. U.S. Department of Commerce, Bureau of the Census. County and City Data Book [United

States] Consolidated File: County Data, 1947-1977 [Computer file]. ICPSR version. Washington, DC: U.S. Dept. of Commerce, Bureau of the Census [producer], 1978. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 1999.

U.S. Department of Health, Education, and Welfare. Equal Educational Opportunities

Programs. Status of Compliance Public School Districts Seventeen Southern and Border States. Report No. 1. Washington, DC: U.S. Government Printing Office, December 1966.

US Senate, Committee on Labor and Public Welfare. Subcommittee on Education. Maximum

Basic Grants—Elementary and Secondary Education Act of 1965 (Public Law 81-874, Title II,

52

and Public Law 89-10, Title I). Washington, DC: US Government Printing Office, September 1965.

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 89-750. Washington, DC: U.S. Government Printing Office, May 1967.

Virginia State Board of Education. Annual Report of the Superintendent of Public

Instruction of the Commonwealth of Virginia. Richmond, VA. Reports covering 1960/61 – 1965/66 school years.

53