Embed Size (px)
Citation preview
The Effects of School Desegregation on Crime
David A. Weiner, NYC Dep’t of Ed.
Byron F. Lutz, Federal Reserve Board
Jens Ludwig, Georgetown & NBER
Topic: Effects of School Desegregation on Crime
• First credible (we think) estimates of court-ordered school desegregation impacts on crime – specifically homicides
• Use variation in timing of court desegregation orders among districts subject to orders 60s-80s
• Desegregation orders generate large declines in homicide rates to blacks and whites at time of desegregation and persistent, long-run declines in homicide offending for blacks
• Due in part to increase in racial integration for blacks & increases in education spending for whites
Why Might Segregation Matter for Crime?
• Peer effects– Standard Theory
– Specific to Desegregation: alter perceptions of opportunities for minorities in American society & increase self esteem
• School quality– Sort black students into higher quality schools and increase overall
school spending
– Direct Effect: produce students with lower crime propensity
– Indirect Effect: increased educational attainment alters opportunity cost of crime
• Ambiguous Effect on Crime : might also intensify racial tension
Historical Background
• Brown v. Board of Education issued in 1954
• Little desegregation occurred in the 1950s and early 1960s
• Larger southern districts began desegregating after 1968 Green decision
• Non-Southern districts began desegregating in large number after 1973 Keyes decision
Relevant Literature
• Lafree and Arum (2006), Criminology– Incarceration rates higher for blacks born in states with more racially
segregated schools– Effect is identified from those who migrate from their state of birth
Link
Data
• Vital Statistics (VS) data on mortality
– Census of homicide victims from 1959 to 1988 by age and race
– But only data on victims
• Supplemental Homicide Report (SHR) data on homicide offending– Data from 1976 to 2002
– Directly measures behavior, but measures arrests
• Numerous other data sources as well (census, school-level data collected by Welch and Light, etc.)
• Unit of Observation = County-year
Link
Research design
• 125 large districts identified in Welch and Light (1987)
– 1% of all districts but 20% all students, 45% of minority students
• Most desegregation orders between 1968 and 1978 (some earlier, later)
• Identification comes from plausibly random timing of these court orders among set of districts ever subject to such orders
• Compare changes in crime across districts around time of desegregation orders
link
Research design
,it p p it i rt itp
y D
Distribution of 1975 Black Age 15-24 Homicide Rates per 100,000
0
.005
.01
.015
0 50 100 150 200
Histogram Kernel Estimate
Research Design
• OLS with Level of the Homicide Rate as Dep Var.
• Proportional Response Model: Count Model
– Quasi-maximum likelihood (QML) count model
– Good consistency properties relative to other count models (Wooldridge 1999)
– Estimates a proportional response: interpret coefficients as estimating the percent change in the dependent variable
– Standard errors clustered using the robust variance estimator proposed by Wooldridge (1999)
Black, Count 15-19, Region-Year
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Black, OLS Level 15-24, Region-Year
-30
-25
-20
-15
-10
-5
0
5
10
15
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Black, OLS Level 15-24, Region-Year
-30
-25
-20
-15
-10
-5
0
5
10
15
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Black, Count 15-24, Region-Year
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Black, Count 15-24, Region-Year
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
White, Count 15-19, Region-Year
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
White, Count 15-19, Region-Year
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Link
Link: Falsification Exercise
Levels
(1) (4) (7)
-8.91(2.76)
-10.55(3.81)
Number of observations 3039
Region * Year Effects XCounty-Specific Linear Trends
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
Black Age 15 - 24
Table 3Black Homicide Victimization
Proportional ResponseOLS Log Dummy
OLSQML Count
Levels
(1) (4) (7)
-8.91 -0.14(2.76) (0.04)
-10.55 -0.23(3.81) (0.06)
Number of observations 3039 3039
Region * Year Effects X XCounty-Specific Linear Trends
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
Black Age 15 - 24
Table 3Black Homicide Victimization
Proportional ResponseOLS Log Dummy
OLSQML Count
Levels
(1) (4) (7)
-8.91 -0.14 -0.13(2.76) (0.04) (0.05)
-10.55 -0.23 -0.19(3.81) (0.06) (0.08)
Number of observations 3039 3039 3039
Region * Year Effects X X XCounty-Specific Linear Trends
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
Black Age 15 - 24
Table 3Black Homicide Victimization
Proportional ResponseOLS Log Dummy
OLSQML Count
Levels
(1) (4) (6)
-0.48(0.50)
-2.22(0.82)
Number of observations 3040
Region * Year Effects X
Table 4White Homicide Victimization
Proportional ResponseOLS Log Dummy
OLSQML Count
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
A. White Age 15 - 19
Levels
(1) (4) (6)
-0.48 -0.05(0.50) (0.06)
-2.22 -0.23(0.82) (0.09)
Number of observations 3040 3040
Region * Year Effects X X
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
A. White Age 15 - 19
Table 4White Homicide Victimization
Proportional ResponseOLS Log Dummy
OLSQML Count
Levels
(1) (4) (6)
-0.48 -0.05 -0.07(0.50) (0.06) (0.05)
-2.22 -0.23 -0.24(0.82) (0.09) (0.07)
Number of observations 3040 3040 3040
Region * Year Effects X X X
Table 4White Homicide Victimization
Proportional ResponseOLS Log Dummy
OLSQML Count
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
A. White Age 15 - 19
(1)
A. Log of White Children
Post Desegregation Years 1 - 5 -0.023(0.031)
Post Desegregation Years +6 0.000(0.047)
B. Log of Black Children
Post Desegregation Years 1 - 5 0.053(0.033)
Post Desegregation Years +6 0.074(0.050)
Number of Observations 420
Region *Year Effect X
Table 3Effect of Desegregation Plan on County Population
(3) (4) (5) (6)
Number of observations 2643 2643 2643 2643
Region * Year Effects X X X X1960 County characteristics * Year Effect
X X
BlackPost Desegregation Years 25+
Post Desegregation Years 20 - 24
Post Desegregation Years 25 - 29
QML Count
Table 9School Desegregation and Long Run Homicide Offending - Age 35 - 44
Proportional Response
Post Desegregation Years 30+
(3) (4) (5) (6)
-0.14 -0.13(0.07) (0.06)
Number of observations 2643 2643 2643 2643
Region * Year Effects X X X X1960 County characteristics * Year Effect
X X
BlackPost Desegregation Years 25+
Post Desegregation Years 20 - 24
Post Desegregation Years 25 - 29
QML Count
Table 9School Desegregation and Long Run Homicide Offending - Age 35 - 44
Proportional Response
Post Desegregation Years 30+
(3) (4) (5) (6)
-0.14 -0.13(0.07) (0.06)
-0.03 0.02(0.04) (0.05)
-0.17 -0.10(0.08) (0.06)
-0.18 -0.15(0.13) (0.09)
Number of observations 2643 2643 2643 2643
Region * Year Effects X X X X1960 County characteristics * Year Effect
X X
QML Count
Table 9School Desegregation and Long Run Homicide Offending - Age 35 - 44
Proportional Response
Post Desegregation Years 30+
BlackPost Desegregation Years 25+
Post Desegregation Years 20 - 24
Post Desegregation Years 25 - 29
(3) (4) (5) (6)
-0.16 -0.18(0.08) (0.09)
0.01 0.02(0.05) (0.05)
-0.16 -0.17(0.09) (0.09)
-0.10 -0.09(0.13) (0.13)
Number of observations 2643 2643 2643 2643
Region * Year Effects X X X X1960 County characteristics * Year Effect
X X
WhitePost Desegregation Years 25+
Post Desegregation Years 20 - 24
Post Desegregation Years 25 - 29
QML Count
Table 9School Desegregation and Long Run Homicide Offending - Age 35 - 44
Proportional Response
Post Desegregation Years 30+
Mechanisms
• School Racial Integration – Reber (2004)
• Local Publics Goods
Link
Changes in Local Public Goods
• Census Bureau’s Census of Governments Finances
• Panel: 1972, 1977, 1982, 1987
• All expenditures on given type of public good within the county (i.e. includes county governments, school districts, etc.)
(1) (2)
Post Desegregation Years 1 - 5 0.17(0.09)
Post Desegregation Years +6 0.16(0.09)
Post Desegregation 0.18(0.09)
Number of Observations 419 419Region * Year Effect X X
Table 8Effect of Desegregation Plan on Local Public Good
A. Ratio of Education Expenditures to Population age 5 -
(1) (2)
Post Desegregation Years 1 - 5 0.19(0.28)
Post Desegregation Years +6 -0.23(0.44)
Post Desegregation 0.25(0.27)
Number of Observations 419 419Region * Year Effect X X
Table 8Effect of Desegregation Plan on Local Public Good
B. Ratio of Police Expenditures to Population
Connection Between Mechanism and Homicides
• Calculate change in the mechanism from one year prior to desegregation to four years after - 5 year changes
• Interact these changes with the post vector
• Cannot be interpreted in a causal manner: suggestive evidence
(1) (2) (4)
Region * Year Effects XDesegregated after 1972Number of observations 3039
Post Deseg. Years 6 + * Δ Ed. Expend. Per
QML CountBlack Homicide Victimization Interactions
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Dissimilarity Index
Post Deseg. Years 6+ * Δ Dissimilarity Index
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per Pupil
(1) (2) (4)-0.12(0.05)
-0.19(0.06)
-0.47(0.18)
-0.69(0.20)
Region * Year Effects XDesegregated after 1972Number of observations 3039
Post Deseg. Years 6 + * Δ Ed. Expend. Per
QML CountBlack Homicide Victimization Interactions
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Dissimilarity Index
Post Deseg. Years 6+ * Δ Dissimilarity Index
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per Pupil
(1) (2) (4)-0.12 -0.10(0.05) (0.05)
-0.19 -0.18(0.06) (0.08)
-0.47(0.18)
-0.69(0.20)
0.12(0.11)
0.24(0.16)
Region * Year Effects X XDesegregated after 1972Number of observations 3039 3039
QML CountBlack Homicide Victimization Interactions
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Dissimilarity Index
Post Deseg. Years 6+ * Δ Dissimilarity Index
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per PupilPost Deseg. Years 6 + * Δ Ed. Expend. Per
(1) (2) (4)-0.12 -0.10 -0.08(0.05) (0.05) (0.07)
-0.19 -0.18 -0.20(0.06) (0.08) (0.09)
-0.47(0.18)
-0.69(0.20)
0.12(0.11)
0.24(0.16)
-0.03(0.11)
0.11(0.14)
Region * Year Effects X X XDesegregated after 1972
X
Number of observations 3039 3039 1433
Post Deseg. Years 6 + * Δ Ed. Expend. Per
QML CountBlack Homicide Victimization Interactions
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Dissimilarity Index
Post Deseg. Years 6+ * Δ Dissimilarity Index
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per Pupil
(1) (2) (3)-0.06(0.05)
-0.23(0.09)
0.20(0.34)
-0.04(0.31)
Region * Year Effects XDesegregated after 1972Number of observations 3040
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per PupilPost Deseg. Years 6 + * Δ Ed. Expend. Per Pupil
White Homicide Victimization Interactions
Post Deseg. Years 1 - 5 * Δ % white in deseg schoolPost Deseg. Years 6+ * Δ % white in deseg
QML Count
(1) (2) (3)-0.06 -0.06(0.05) (0.07)
-0.23 -0.21(0.09) (0.10)
0.20(0.34)
-0.04(0.31)
-0.04(0.06)
-0.20(0.10)
Region * Year Effects X XDesegregated after 1972Number of observations 3040 3040
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per PupilPost Deseg. Years 6 + * Δ Ed. Expend. Per Pupil
White Homicide Victimization Interactions
Post Deseg. Years 1 - 5 * Δ % white in deseg schoolPost Deseg. Years 6+ * Δ % white in deseg
QML Count
(1) (2) (3)-0.06 -0.06 -0.05(0.05) (0.07) (0.07)
-0.23 -0.21 -0.17(0.09) (0.10) (0.16)
0.20(0.34)
-0.04(0.31)
-0.04(0.06)
-0.20(0.10)
-0.30(0.14)
-0.44(0.13)
Region * Year Effects X X XDesegregated after 1972 XNumber of observations 3040 3040 1433
White Homicide Victimization Interactions
Post Deseg. Years 1 - 5 * Δ % white in deseg schoolPost Deseg. Years 6+ * Δ % white in deseg
QML Count
Post Deseg. Years 1 - 5
Post Deseg. Years 6+
Post Deseg. Years 1 - 5 * Δ Exposure Index
Post Deseg. Years 6+ * Δ Exposure Index
Post Deseg. Years 1 - 5 * Δ Ed. Expend. Per PupilPost Deseg. Years 6 + * Δ Ed. Expend. Per Pupil
Implications, part 1
• Can school desegregation orders starting in late 1960s explain part of the decline in the black homicide rate?
0
20
40
60
80
100
120
1950 1960 1970 1980
Year
Dea
ths/
100,
000
peop
le
Black Males Black Females White Males White Females Cumulative Desegregation Orders
Homicide rates for people aged 15-24
0
20
40
60
80
100
120
1950 1960 1970 1980
Year
Dea
ths/
100,
000
peop
le
0
20
40
60
80
100
120
Cum
ulat
ive
Des
egre
gatio
n O
rder
s
Black Males Black Females White Males White Females Cumulative Desegregation Orders
Homicide rates for people aged 15-24
Implications, part 2
• Previous results might misstate distributional and social welfare implications of desegregation– Our results: 2 fewer homicides per 100,000 whites– Around 10 per 100,000 fewer homicides to blacks– Cohen et al. (2004), WTP per homicide $9.7m– So benefits of around $200 per white youth in county and $1,000 per
black youth in county • By comparison, avg. per pupil spending in US over our study period
probably around $4,500
Contributions
• Non-Academic outcomes
– Could significantly alter conclusions about welfare arising from literature on academic outcomes:
• Perry Preschool, crime impacts 2/3 of benefits
– Also could change assessment of distributional consequences of policy
• Econometric Identification
• Long-run Effects of Desegregation
• Mechanisms
back
Data
• Unit of Observation = County-year– Note school districts were subject to court orders
– But Vital Statistics data at county level
– For around 1/3 of cases, district=county
– District enrollments account for around ½ of school-age population in our counties
• We examine homicide rates and therefore require annual population counts by county– In non-census years, these are imputations, with adjustments, produced
by the Census Bureau
– If these fail to capture “white flight” then our estimates will be biased downward – we will understate the homicide rate because of mismeasurement in the denominator
back
Full Sample 1960 1970 1980
Total 10.8 6.6 11.3 14.0
Total white 6.3 3.3 6.1 9.1
Total black 37.8 31.1 44.7 41.2
White 15-19 5.7 2.3 5.0 9.7
Black 15-19 29.0 20.3 37.1 25.8
White 15-24 7.6 3.4 5.8 12.4
Black 15-24 45.2 29.2 60.0 47.1
Table 1Descriptive Statistics
Homicide rates per 100,000
Full Sample 1960 1970 1980
Total 10.8 6.6 11.3 14.0
Total white 6.3 3.3 6.1 9.1
Total black 37.8 31.1 44.7 41.2
White 15-19 5.7 2.3 5.0 9.7
Black 15-19 29.0 20.3 37.1 25.8
White 15-24 7.6 3.4 5.8 12.4
Black 15-24 45.2 29.2 60.0 47.1
Table 1Descriptive Statistics
Homicide rates per 100,000
Full Sample 1960 1970 1980
Total 10.8 6.6 11.3 14.0
Total white 6.3 3.3 6.1 9.1
Total black 37.8 31.1 44.7 41.2
White 15-19 5.7 2.3 5.0 9.7
Black 15-19 29.0 20.3 37.1 25.8
White 15-24 7.6 3.4 5.8 12.4
Black 15-24 45.2 29.2 60.0 47.1
Table 1Descriptive Statistics
Homicide rates per 100,000
back
White, OLS Level 15-19, Region-Year
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
White, OLS Level 15-19, Region-Year
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
back
Segregation Measures
• Dissimilarity Index – assess sorting of students
• bit and wit refer to the number of black and white students, respectively, at school i at time t
• Bt and Wt refer to the total number of black and white students, respectively, in the school district
• ranges from 0 to 1, with 1 denoting complete segregation
• Interpretable as the percent of black students who would need to be reassigned to a different school for perfect integration to be achieved given the districts overall racial composition
• Racial Composition changes measured by enrollment by race at district level (`reverse white flight’)
t
itn
i t
itt W
w
B
bD
1
*2
1
Segregation Measures
• Exposure Index
• ti is the total enrollment at school i
• Interpretable as the percent of white students in the average black student’s school
• For a given district, it ranges from 0 to the percent of white students in the district as a whole
• It can be viewed as a measure of the extent of contact between the two races
• Changes reflect the net effect of changes in both sorting between schools and changes in the district racial composition
it
itn
iit
dtdt t
wb
BE *
1
1
Black-White Dissimilarity Index, Region-Year
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Black-White Exposure Index, Region-Year
-0.1
-0.05
0
0.05
0.1
0.15
0.2
-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
back
Motivation
• 1954 Brown decision launches one of most important social policy changes in 20th century America
• Most of the existing literature focuses on academic outcomes; limited research on the impact of court-ordered desegregation on economically important outcomes other than education
– Very large social costs of crime (~$2 trillion / year)
• Important because of recent wave of dismissals of court-ordered desegregation plans
0
5
10
15
1960 1965 1970 1975 1980 1985
Implementation Year
Num
ber
of C
ount
ies
South
link
0
5
10
15
1960 1965 1970 1975 1980 1985
Implementation Year
Num
ber
of C
ount
ies
South Non-South
Falsification Check
• Mortality from Illness
• Should not be affected by the mechanisms (peer effects, school quality, police spending, etc.) by which desegregation influences homicides
• General check on identification strategy and specifically on issue of possible measurement error in the denominator for whites
• Age 15 – 19 mortality from illness is 13.0 per 100,000, compared to 10.7 for homicides
LevelOLS Level QML Count OLS Log
Dummy(1) (3) (3)
-0.70 -0.06 -0.06(0.57) (0.04) (0.03)
-0.51 -0.03 -0.02(0.85) (0.06) (0.05)
Number of observations 3040 3040 3040
-0.32 -0.04 -0.01(1.71) (0.04) (0.03)
2.49 0.04 0.04(2.87) (0.05) (0.05)
Number of observations 3039 3039 3039
Region * Year Effects X X X
B. Black 15 - 24
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
Post Desegregation Years 6+
Post Desegregation Years 1 - 5
Table 5Falsification Test, Death From Illness
Proportional response
A. White 15 - 19
link
