Do Police Reduce Crime? Estimates Using the Allocation of Police Forces after a Terrorist Attack
Author(s): Rafael Di Tella and Ernesto Schargrodsky
Source: The American Economic Review , Mar., 2004, Vol. 94, No. 1 (Mar., 2004), pp. 115- 133
Published by: American Economic Association Stable URL: https://www.jstor.org/stable/3592772
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to The American Economic Review
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
All use subject to https://about.jstor.org/terms
Do Police Reduce Crime? Estimates Using the Allocation of Police Forces After a Terrorist Attack
By RAFAEL DI TELLA AND ERNESTO SCHARGRODSKY*
An important challenge in the crime literature is to isolate causal effects of police on crime. Following a terrorist attack on the main Jewish center in Buenos Aires, Argentina, in July 1994, all Jewish institutions received police protection. Thus, this hideous event induced a geographical allocation of police forces that can be
presumed exogenous in a crime regression. Using data on the location of car thefts before and after the attack, we find a large deterrent effect of observable police on crime. The effect is local, with no appreciable impact outside the narrow area in which the police are deployed. (JEL K42)
Classical criminology assumes that criminaplaspers surveyed researchers found either a pos- are rational beings who weigh the costs anitdive effect of police presence on crime or no benefits of their actions. Gary Becker (196re8l)ationship between these variables. More re- produced the first fully fledged theory of cricmenet surveys by Thomas Marvell and Carlisle
based on rational behavior. His research led to
an upsurge of interest in the economics of crim-ire (2000) reach similar conclusions.
inal behavior [see, for example, Isaac Ehrlich There is, however, a serious endogeneity (1973), Ann Witte (1980), Ehrlich and Georgeproblem with these studies that arises from the Brower (1987), James Andreoni (1991), Rich-simultaneous determination of crime and police ard Freeman (1996), Steven Levitt (1997), presence (see Franklin Fisher and Daniel Nagin, Pablo Fajnzylber et al. (2000), inter alia]. One1978). It is likely that the government of a city of the central predictions of Becker’s theory isin which the crime rate increases will hire more
that crime will decrease when police presence police officers. Areas beset by high crime will increases. A basic problem with this predictionthus end up with more police officers than areas is that it has largely failed to find empiricawlith low crime rates, introducing a positive bias support. In a survey of the literature, Samuel in the police coefficient in a crime regression. A Cameron (1988) reports that in 18 out of 22central challenge in the crime literature has been
to break this endogeneity in order to identify
causal effects of police on crime.
* Di Tella: Harvard Business School, Boston, MA 02163 Two recent papers use a time-series strategy
(e-mail: Schargrodsky: Universidad Tor-
cuato Di Tella, Mifiones 2177, (C1428ATG) Buenos Aires,to address this problem. Using data for the
United States, Marvell and Moody (1996) find Argentina (e-mail: We thank a co-
Granger-causation between crime and police editor, an extremely constructive referee, Jushan Bai, Se-
bastian Galiani, Erzo Luttmer, Robert MacCulloch, Sam running in both directions. In a similar vein,
Peltzman, Andrea Rotnitzky, several key informants, and
Hope Corman and H. Naci Mocan (2000) ex-
seminar participants at the 2001 AEA New Orleans meet-
ings, the University of California-Berkeley, Stanford Uni-ploit high-frequency data for New York City to versity, Econometric Society, LACEA, UTDT, UdeSA, show that increases in the number of police
officers cause a reduction in one out of five Getulio Vargas, AAEP, and UNLP for helpful suggestions.
Moody (1996) and John Eck and Edward Magu-
The second author thanks SCID at Stanford University for
crime categories (specifically, burglary). Monthly
their hospitality. Matias Cattaneo, Luciana Esquerro, and
data are used because hiring and training delays
Magali Junowicz provided excellent research assistance.
The database and computer programs used in this paper arein the response of the police authority to an
increase in crime will mitigate simultaneity bias available at http://www.people.hbs.edu/rditella and www.
utdt.edu/-eschargr.
present in low-frequency data. In order to
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
All use subject to https://about.jstor.org/terms
116 THE AMERICAN ECONOMIC REVIEW MARCH 2004
validly address the simultaneity concerns,ttahtesoef the evidence leads Levitt (2002) to identification strategies depend, cruciallyw,oonnder: “If electoral cycles can provide no the assumption that the authorities are unmabolreettohan suggestive evidence of a causal im- forecast crime-fighting needs.’ pact of police on crime, are there other identi-
Levitt (1997) develops a different appfricoaatcihon strategies that can do better?”
using instrumental variables to break simultInanteh-is paper we present a different approach ity. He docum ents the presence of an eletcotoerstailm ate the causal effect of police on crim e cycle in police hiring and uses the timOingJoufly 18, 1994 terrorists exploded a bom gubernatorial and mayoral elections to inthstartud-estroyed the Asociacion Mutual Israelita ment for police presence in a panel of 59Arlagregnetina (A.M.I.A.), the main Jewish center in U.S. cities from 1970-1992. Using two-Asrtgagenetina. Eighty-five people died and mor
than 300 were wounded in the attack. One week least-squares (2SLS) techniques, Levitt finds a
negative and significant effect of policelaotnervtih-e federal government assigned police olent crime. The pattern across individuaplrcortiemcteion to every Jewish and Muslim building categories is surprising, with murder exhinbithine gcountry. Because the geographical distri- the largest (and the only significant) coeffbuictieonto,f these institutions can be presumed to andwithveryimpreciseestimatesforthbeeceaxtoeg-enousinacrimeregression,thishideous gories in which the rational model is prevsuenmtecdonstitutes a natural experiment whereby
the simultaneous determination of crime and to be more relevant (e.g., property crimes). Still,
the validity of the instrum ent m ight bepoqluicees-presence can be broken.3
We collected information on the number of tioned. The timing of elections may affect crime
by way of channels other than the nummbeortoorfvehicle thefts per block in three neigh-
borhoods in Buenos Aires before and after the police officers on the street. Levitt avoids some
terrorist attack. The information covers the of these concerns by controlling for the unem-
ployment rate and public spending, altnhinoeu-gmhonth period beginning April 1 and end- police effort and crime reporting (as winegllDaescember 31, 1994. We also collected in-
formation on the location of each Jewish police hiring) may also respond to the timing of
elections, particularly if the police are theintsatrigteution in these neighborhoods. We then es- of political m anipulation. Sim ilarly, the tbiemhavte-d the effect of police presence on car
theft. Our difference-in-differences estimates ior of judges and prosecutors may be affected
by elections, something that could logicalslhyowret-hat blocks that receive police protection duce criminal activity during such times.e2xperience significantly fewer car thefts than A more severe concern raised by JustitnheMrce-st of the neighborhoods. The effect is
Crary (2002) is that Levitt’s 2SLS estilmargaet.eRselative to the control group, car thefts suffer from a computational error (see alsfoallLbevy-75 percent in the blocks in which the itt’s reply, 2002). When the mistake is coprroetcetcetded institutions are situated. However, the the replication results show no effect ofefpfoeclitcies extremely local. We find no evidence on crime at standard significance levetlhs.atTphoelice presence in a given block reduces car
theft one or two blocks away from the protected buildings.
There has been considerable interest in iden- I Criminologists often emphasize the benefits of antici-
pating crime patterns. David Bayley (1998), for etxiafmyipnlge, the mechanisms by which police pres-
states “The key assumption behind smarter law enforcement
ence reduces crime. Is it that police presence
is that crime is not evenly scattered through time and space.
makes criminal activity less attractive (deter-
Police are not faced with meeting all crime threats every-
rence), or is it that police officers apprehend where all the time. Instead, each form of crime displays a
particular pattern which, if understood, provides opcproirmtuinai-ls leaving fewer of them around to com- ties for law enforcement” (Bayley, 1998, p. 174). On the
allocation of police resources to protect high crime areas,
often called “hot spots,” see Lawrence Sherman 3etOnaln.atural and randomized experiments, see the dis- (1989) and Sherman and David Weisburd (1995).cussions in Robert LaLonde (1986), Joshua Angrist (1990),
2 On the incentives faced by members of the juAdnigcirairsty and Alan Krueger (1991), Daniel Hamermesh see, for example, Richard Posner (1993). (1999), and Bruce Sacerdote (2001).
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
All use subject to https://about.jstor.org/terms
Code Help
VOL. 94 NO. 1 DI TELLA AND SCHARGRODSKY: DO POLICE REDUCE CRIME? 117
m it crim es (incapacitation)? Being bAairseesdpropner.6 Although providing this surveil- changes in crime levels in particular lolcanacteiroenqusiredthedistractionofanonnegligible (i.e., the protected blocks) our results aproepourtnio-nofthepoliceforcesprotectingthe likely to reflect changes in the num berasreoasfininw-hich these buildings are located, the carcerated criminals, which should affpeoclitceaflolrcesmadeaseriousefforttomaintain neighborhood blocks, not just those copnrtevaionusinlegvels of police presence in the rest of Jewish institutions.4 Thus, we believe thesaetnoeuigrhborhoods.Governmentofficials estimates are most appropriately interpwroertreieddatshatcompromisingpoliceprotection the causal deterrent effect of police statfhfroiungghouotnthe neighborhoods might generate in car theft. However, it is still possible theatrescidaernts ill feelings towards the Jewish thefts w ere displaced in a w ay thatcowmme uanritey.7 Because the personnel commit- unable to m easure, in w hich case the efmfenetctcouolfd not be met with the normal number
policing m ay be sm aller than our esotfipmolaicteeasssigned to these neighborhoods, the suggest. increased police presence was achieved with
The rest of the paper is organized as follows. officers reassigned from administrative tasks at
In Section I we describe our data. In Section II
we discuss the empirical strategy and presenctations Division, and the Mounted Police.8
our results. Section III concludes.
The data analyzed in this paper are from three noncontiguous, Buenos Aires neighborhoods
I. Data Description that collectively represent about 3.2 percent of the city’s area and account for 6.9 percent of its On July 18, 1994 a terrorist attack destroypedopulation. One police station is located in each the main Jewish center (A.M.I.A.) in Buennoesighborhood.9 The neighborhoods were se- Aires, Argentina.5 Seven days later, on July 2le5c,ted on the basis of three criteria: they were
the areas with the largest numbers of Jewish the federal government decided to provide 24-
hour police protection to more than 270 Jewiisnhstitutions in the city; significant portions of and Muslim institutions (including synagogueths,e neighborhoods were not close to a protected mosques, clubs, cemeteries, and schools) in Airn-stitution (more than 50 percent of blocks are gentina. Muslim institutions were protected fmorore than two blocks removed from a protected fear of potential retaliations after the Islamic
organization, Hezbollah, claimed responsibility
for the attack. Nearly ten years after the attack
6 Approximately 85 percent of the Jewish population of
this protection is still provided. the country lives in Buenos Aires and its suburbs.
7 Institutional information for this paper was gathered A significant proportion of the protected
through a series of interviews with key informants, includ- buildings are Jewish institutions within Buenos
4 Daniel Kessler and Levitt (1999) use California’s sednu-ring the period under consideration as well as a former tence enhancement laws for a selected group of crimesfetdoeral judge, a former federal prosecutor, and the director distinguish between incapacitation and deterrence. See aolfsoa nongovernmental organization devoted to protecting Levitt (1998). Articles studying responses to increasesciivnil rights.
detection probabilities include Avner Bar-Ilan and Sacer-8 For example, more than one-third of approximately 200 dote (2001) on red light violations, and Robert McCormpioclikce officers stationed in Once, one of the neighborhoods and Robert Tollison (1984), on fouls committed by baskweit-h the highest density of Jewish institutions, had to be ball players.
5This was the second terrorist attack in the city of Buenos Aires. The Israeli embassy had been destroyed on March 17, 1992. In the months immediately following this first attack, the most prominent Jewish centers, including A.M.I.A., had been given more attention by officers on patrol. But surveillance was not generalized and declined gradually. Information on these attacks can be found in www.atentado-amia.com.ar, www.daia.org.ar, and www. bnaibrith.org.
reassigned to protection duties. The personnel necessary to maintain the previous level of police presence in the rest of the neighborhood was pulled from outside of this police station.
9 There are 53 police stations in Buenos Aires. Adrian
Pelacchi (2000) provides an in-depth discussion of the in-
stitutional features of crime and the police force in Argen- tina.
10 There are no Muslim institutions in the neighborhoods considered in our study.
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
the Central Police Department, the Communi-
ing the Secretary of Security (third level of authority in the federal government, behind the president and ministers), the Chief of the Federal Police, and the Minister of the Interior
All use subject to https://about.jstor.org/terms
Computer Science Tutoring
THE AMERICAN ECONOMIC REVIEW
MARCH 2004
July 18 July 25
April May June Aug Sept Oct Nov Dec
PreP.TIMELINEost FIGURE 1. TIMELINE OF EVENTS
institution), providing a control group for oursurvey, Ministerio de Justicia, 2000). Second study; and three was the maximum number ofbecause criminals often use stolen cars in the
police stations for which we were able to con- commission of other crimes, victims who report
vince police authorities to provide us data.”1car thefts to police forestall confusion about There are a total of 876 blocks in these three their involvement in such crimes. The victim-
neighborhoods. The block constitutes the unit oifzation survey cited above reports that 87 per- observation for our study.12 cent of Buenos Aires car thefts are reported to
surveillance had not yet been introduced and wa hich they intend to commit crimes.
second week during which police began to im- Car theft information obtained from the po-
lice includes the address at which the stolen plement the protection policy. By the end of the
last week of July police protection was fullvyehicle was parked, make and year of the vehi- functioning and known to the public. Finally,cle, day and time of the report, and whether the August 1 to December 31 covers the period ofrobbery was violent. During the period of anal- police protection. ysis 794 nonarmed car thefts were reported in
Although victims’ tendency to underreportthese neighborhoods.14 Although they normally often results in official records underestimatinogccur in the middle of blocks, car thefts in many crime levels, this is a minor problem for cacrases are reported at corers so as to facilitate thefts in Buenos Aires for two reasons. First,victims’ verbal descriptions of crime locations police intervention is required to activate cart the time they file police reports. We assigned insurance against theft, a type of insurance caro-ne-quarter of each car theft reported at a corer
to each of the intersection’s four blocks.15 ried by most car owners in Buenos Aires (89
percent according to the official victimization
13 Ninety-four percent of Buenos Aires car robberies 1 The police stations’ daily records, which register autooccur in the street (Ministerio de Justicia, 2000).
We obtained all the information available to
the police, compared to only 29 percent for all the police (with the exception of the victim’stypes of crime. A further advantage of auto theft name) about each auto theft in these neighbord-ata is that this category of crime is expected to be hoods for the nine-month period starting Aprmilore sensitive to police presence.’3 Most robber- 1, 1994 and ending December 31, 1994. Figureies occur after a brief period of surveillance of the 1 presents a timeline of the events in our studyin.tended victim. Criminals concentrating their at- April 1 to July 17 constitutes the period befortention on mobile victims might miss the presence the terrorist attack. The interim period of Julyof police. A parked car, on the other hand, gives 18 to July 31 includes a first week during whichriminals time to gather information on areas in
thefts on the same pages as reports of every other type of 14 We exclude a small number (63) of armed robberies crime or incident, are not available to the public. The Chierfeported during this period as well as 86 misreports that of the Federal Police had to issue a special authorizationcorrespond to nonexisting or incomplete addresses or to car instructing police station personnel to transcribe the data ftohrefts that occurred outside of our sample neighborhoods us.
(i.e., that were reported to the wrong police station).
12 We consider a block as the segment of a street15bTe-his procedure assigns some fractions of thefts to tween two corners. With few exceptions, Buenos Aibreloscikssaoutside the boundaries of the neighborhoods under
perfect grid city, with streets crossing perpendiculsatruldyy,atwhich reduces the total number of car thefts from 794
to 778.75. corers. Each block is about 100 meters (110 yards) long.
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
All use subject to https://about.jstor.org/terms
VOL. 94 NO. I
DI TELLA AND SCHARGRODSKY: DO POLICE REDUCE CRIME? 119
TABLE 1-DEMOGRAPHIC CHARACTERISTICS OF CONTROL AND TREATMENT AREAS
Census tracts without Census tracts with
Demographic Jewish institutions Jewish institutions Difference characteristics (A) (B) (C) = (A) – (B)
Home ownership rate 0.696 0.663 0.032 (0.008) (0.017) (0.019) Overcrowding rate 0.014 0.017 -0.002 (0.001) (0.002) (0.003) Poverty rate 0.042 0.052 -0.010 (0.003) (0.008) (0.009)
Education of household 11.653 11.052 0.600
head (0.147) (0.300) (0.335) Number of household 2.719 2.685 0.034
m em bers (0.023) (0.054) (0.059) Female population 0.556 0.552 0.003
(0.001) (0.003) (0.003) Unemployment rate 0.053 0.059 -0.005 (0.001) (0.003) (0.003)
Age 38.005 37.690 0.315 (0.128) (0.223) (0.258)
Number of census tracts 53 14
Notes: Columns (A) and (B) present the mean of each v with Jewish institutions in our sample. Column (C) p Standard deviations are in parentheses. Home ownersh occupied houses. Overcrowding rate is the percentage people per room. Poverty rate is the percentage of hous need (overcrowding; four or more members per wo with low educational attainment; poor quality housing school; or no fecal evacuation system). Education of educational attainment of the household head in num the percentage of women in the total population. unemployment for the population of age 14 or hig population.
Source: 1991 Population Census.
The com pleted data segtatiionfcolruwdhicehdcenisnusfinoformatiaontisoavnailable o n t h e g e o g r a p h y o f tinh BeuseneosnAiereisgishcbenosurs htraoctosd(fsra,cciones particular, the precise cleoncsalaest),iwohnichocofvereaappcrhoximJateelwyeiigshhtto institution. There are 45tepnrconttiegcuotuesdhectianress.tTietstustofiomneasnsirneveal
no statistical differences between census tracts this part of the city. Thirty-seven of them are
that contain and do not contain Jewish institu- within these neighborhoods, while the rest are
near the boundaries (butitonslealosnsgtthehfaolnlowitnhgdrimeensibonls:ohcomkes aw ay).16 The geograopwhnerischipalrated,piesrctenrtaigbeoufotveirocrnowdeodf
blocks, institutions, anhdouscehaorlds,tpherecefntagseoifsposourmhoumsehoald-s, rizedinTableAlinthenumAbperpofehnoudsehioxld.members,percentageof
Using inform ation frowmoment,hemepl1oy9m9en1trcate,nansduagse,.TTheao-nly b l e 1 c o m p a r e s s o c i o e cdiomennsoionmaloincg wchihch athresae centsuesrtriacstst diifc-s potentially related to cfreriemdweasyveairscotfiemducaitzioantofiothnehoausnehdold
c a r o w n e r s h i p a c r o s s aherade: 1a1.s65 awndi1t1.0h5,oreusptectivaenly,dforwtraictsh
without and with Jewish institutions. We inter- Jewish institutions. The lowest level of aggre-
pret these results as evidence that the surveil-
lance policy was randomly assigned across
socioeconomic characteristics. Table A2 in the
16 None of the protected institutions in our sample is
located at a corer.
Appendix compares demographics and car theft
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
All use subject to https://about.jstor.org/terms
浙大学霸代写 加微信 cstutorcs
THE AMERICAN ECONOMIC REVIEW MARCH 2004
TABLE 2-MONTHLY EVOLUTION OF CAR THEFT
More than two One block Two blocks blocks from Jewish from nearest from nearest
nearest Jewish institution on Jewish Jewish Difference Difference Difference institution the block institution institution (E) = (F) = (G) =
Month (A) (B) (C) (D) (B) – (A) (C) – (A) (D) – (A)
April 0.09955 0.12162 0.12111 0.12278 0.02206 0.02156 0.02323 (0.248) (0.361) (0.287) (0.297) (0.060) (0.025) (0.022)
May 0.10840 0.08783 0.07763 0.09734 -0.02056 -0.03076 -0.01106 (0.235) (0.205) (0.181) (0.259) (0.035) (0.018) (0.020)
June 0.07853 0.12837 0.07763 0.06969 0.04983 -0.00090 -0.00884 (0.196) (0.286) (0.215) (0.186) (0.047) (0.019) (0.015)
July (1-17) 0.03926 0.02027 0.05900 0.03097 -0.01899 0.01973 -0.00829 (0.145) (0.069) (0.210) (0.141) (0.013) (0.017) (0.011)
July (18-31) 0.03926 0.02702 0.07298 0.06858 -0.01224 0.03371 0.02931 (0.146) (0.078) (0.217) (0.238) (0.014) (0.018) (0.017)
August 0.11836 0.04729 0.06677 0.12721 -0.07106 -0.05159 0.00884 (0.287) (0.175) (0.219) (0.304) (0.031) (0.021) (0.024)
September 0.10176 0.01351 0.09006 0.09845 -0.08825 -0.01170 -0.00331 (0.256) (0.057) (0.276) (0.248) (0.015) (0.024) (0.020)
October 0.12112 0.06081 0.09782 0.08849 -0.06031 -0.02330 -0.03263
(0.267) (0.215) (0.260) (0.236) (0.037) (0.024) (0.020) November 0.09623 0.02702 0.11024 0.10176 -0.06921 0.01400 0.00553
(0.240) (0.078) (0.288) (0.217) (0.017) (0.025) (0.018) December 0.10176 0.02702 0.11645 0.10619 -0.07474 0.01468 0.00442
(0.268) (0.078) (0.278) (0.225) (0.018) (0.025) (0.019) Number of 452 37 161 226
Notes: The first four columns present the mean and standard deviation (in parentheses) of the number of car thefts for each type of block per month. The average number of car thefts for July can be obtained by summing the subperiods. The last three columns present the differences of means of columns (B), (C), and (D) relative to column (A), with standard deviations in parentheses.
rates for the neighborhoods under study relative to the whole city.
A key dimension in our empirical exercise is the distance of each block in our sample to the nearest Jewish institution, whether or not the building is within our neighborhoods. We dis- tinguish among blocks that contain a Jewish institution, blocks that are contiguous in any direction to a block containing a Jewish institu- tion, and blocks that are two blocks away in any direction from a block containing a Jewish in- stitution. We then compare these with blocks that are more than two blocks away from a block containing a Jewish institution.
Table 2 presents means (and standard devia- tions) of auto thefts for each month for each type of block. The bottom row tallies the num- ber of blocks of each type. For the month of July we consider, separately, the period before and after the terrorist attack. For the post-July
period, the table shows that, relative to the con-
trol group (i.e., blocks more than two blocks
away from the nearest Jewish institution),
blocks occupied by a Jewish institution experi- enced a lower level of car theft. A similar re-
duction is not observed for blocks that are one
or two blocks away from the nearest Jewish
institution. In particular, differences of means
indicate that average car theft in blocks with
protected institutions is significantly less than
average car theft for the control group for every
month after July, with the exception of October.
Although casual inspection of the data for blocks that contain a Jewish institution also
suggests a decline for the first days of July
(before the attack), the difference with the
control group is not statistically significant
for this period. Indeed, for every period prior
to the terrorist attack we cannot reject that the car theft mean for the blocks with Jewish
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
All use subject to https://about.jstor.org/terms
VOL. 94 NO. 1 DI TELLA AND SCHARGRODSKY: DO POLICE REDUCE CRIME? 121
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 April May June July August September October November December
By Week (Left Axis) Means (Right Axis)
Jewish Institution in the Block – 0 – – Pre and Post Means for Jewish Institution in the Block
One Block from Nearest Jewish Institution – – – – Pre and Post Means for One Block from Nearest Jewish Institution
…… Two Blocks from Nearest Jewish Institution – – O – -Pre and Post Means for Two Blocks from Nearest Jewish Institution
– More than Two Blocks from Nearest Jewish Institution – * – -Pre and Post Means for More than Two Blocks from Nearest Jewish Institution
FIGURE 2. WEEKLY EVOLUTION OF CAR THEFTS
Notes: Per-week average of car thefts for blocks that contain a Jewish institution (37 blocks), blocks that are one block away from the nearest Jewish institution (161), blocks that are two blocks away from the nearest Jewish institution (226), and blocks that are more than two blocks away from the nearest Jewish institution (452). The horizontal lines are pre- and postattack averages (excluding car thefts that occurred between July 18 and July 31).
institutions is equal to the mean for the con- trol group.
for each block.17 We exclude car thefts that
occurred between July 18 and July 31.18 Having data on blocks with and without protected insti- tutions allows us to define a treatment and a
control group. We include month fixed effects that control for any aggregate shocks in the evolution of crime, and block fixed effects that control for time-invariant influences. Control-
ling for time and individual effects, we obtain the difference-in-differences estimators of the
effect of police on crime using the following model:
Car Theftit = aoSame Block Policeit + al One Block Policeit
+ a2 Two Blocks Policeit + M, + Fi + eit,
Figure 2 plots the same information at a more disaggregated level, namely, by week. The se-
ries (left axis) are obviously more volatile for
the aggregates that average a smaller number of
blocks (see the bottom row of Table 2). The
horizontal lines (right axis) represent the pre-
and postattack averages for the weekly data for
each block type. Prior to the attack there are no discernible differences in these averages across
the different types of blocks. After the attack,
however, average car thefts for blocks that con-
tain Jewish institutions evolve around a lower
mean. Instead, car theft levels for the other types of blocks show a slight increase over time.
II. The Effect of Police on Car Theft
A. Empirical Strategy
Our purpose is to identify the causal effect17ofOf course, our monthly level of aggregation is arbi-
trary. Similar results obtain when we aggregate the data, for police presence on car thefts. Using the total
example, at the weekly level. All results reported but not number of car thefts per block during each
month from April to December as the dependent
18 Including the period between July 25 and July 31 does
not affect our results. variable gives us a panel with nine observations
This content downloaded from 130.102.13.92 on Wed, 12 Apr 2023 02:57:52 UTC
presented are available upon request.
All use subject to https://about.jstor.org/terms
122 THE AMERICAN ECONOMIC REVIEW MARCH 2004
Car Theftit is the num ber of car thetfhtosugihnwe can calculate the reduced-form re-
block i for month t;
Same-Block Policeit is a dummy variable that
the distribution of police forces per block at any given ti