EssaysontheEconomicsofCrime By AaronJamesChalfin Adissertationsubmittedinpartialsatisfactionofthe requirementsforthedegreeof DoctorofPhilosophy in PublicPolicy inthe GraduateDivision ofthe UniversityofCalifornia,Berkeley Committeeincharge: ProfessorStevenRaphael,Chair ProfessorRuckerJohnson ProfessorJustinMcCrary ProfessorJesseRothstein Spring,2013 EssaysontheEconomicsofCrime Copyright2013 by AaronJamesChalfin Abstract EssaysontheEconomicsofCrime by AaronJamesChalfin DoctorofPhilosophyinPublicPolicy UniversityofCalifornia,Berkeley ProfessorStevenRaphael,Chair Thisdissertationconsiderstherolethatofvariousinputsininformingthemarketfor crimes. Chapter1considersthe“national”effectofimmigration. Usingpaneldataon U.S.citiesandaninstrument thatleveragestemporalvariation inrainfall indifferent regions of Mexico and persistence in regional Mexico-U.S. migration networks, my findingsindicatethatMexicanimmigrationisassociatedwithnoappreciablechange intherateofeitherviolentorpropertycrimesinU.S.cities. Chapter 2 leverages a natural experiment created by recent legislation in Arizona to estimate the impact on crime of an extremely large and discrete decline in the states foreign-born Mexican population. I show that Arizonas foreign-born Mexi- can population decreased by as much as 20 percent in the wake of the states 2008 implementation of the Legal Arizona Workers Act(LAWA), a broad-based E-Verify lawrequiringemployerstoverifytheimmigrationstatusofnewemployees,coupled with severe sanctions for employer noncompliance. In order to isolate the causal effectofthepassageandimplementationofLAWAoncrime,Ileverageasynthetic differences-in-differences estimator, using a new method of counterfactual estimation proposedbyAbadie,DiamondandHainmuller(2010). Incontrasttopreviousliterature, IfindsignificantandlargeeffectsofMexicanimmigrationonArizonaspropertycrime rate. Results are driven, in large part, by the fact that LAWA resulted in especially disproportionatedeclinesamongMexicanmigrantswhoareyoungandmaleand,as such,theeffectsarepredominantlycompositional. The final chapter, coauthored with Justin McCrary, considers the responsiveness of crime to police manpower. Using a new panel data set on crime in medium to largeU.S. citiesover 1960-2010, we show that(1) year-over-yearchanges in police per capita are largely idiosyncratic to demographic factors, the local economy, city budgets, measures of social disorganization, and recent changes in crime rates, (2) year-over-yearchangesinpolicepercapitaaremismeasured,leadingmanyestimatesin theliteraturetobetoosmallbyafactorof5,and(3)aftercorrectingformeasurement errorbiasandcontrollingforpopulationgrowth,aregressionofwithin-statedifferences inyear-over-year changesin citycrimes onwithin-statedifferences inyear-over-year changes in police yields economically large point estimates. Our estimates imply that each dollar spent on police is associated with approximately $1.60 in reduced victimizationcosts,suggestingthatU.S.citiesemploytoofewpolice. 1 Acknowledgements Iwouldliketothankmyprincipaladvisors: SteveRaphael,RuckerJohnson,Justin McCrary and Jesse Rothstein. I would also like to acknowledge helpful comments fromthe followingindividuals whohave providedfeedbackon individualchapters: OrleyAshenfelter,EmilyBruce,ShawnBushway,DavidCard,RajChetty,BobCooter, RachaelCroson, JohnDinardo, JohnEck, AaronEdlin, BenjaminHansen,Hans John- son, Louis Kaplow, Mark Kleiman, Tomislav Kovandzic, Prasad Krishnamurthy, Ron Lee, Thomas Lemieux, Morris Levy, John MacDonald, Jeff Miron, Denis Nekipelov, EmilyOwens,AlexPiquero,JimPowell,KevinQuinn,DanielRichman,SethSanders, David Sklansky, Geno Smolensky, Gary Solon, Kathy Speier, Hosung Sohn, Sarah Tahamont,EricTalley,JohnZedlewskiandFrankZimring. i In his seminal 1968 work, Gary Becker proposed that anindividual’s propensity to commitcrimesisprincipallyafunctionofthreefactors: (1)theindividual’sprobability of capture for a given crime, (2) the severity of the sanction if captured and (3) the opportunitycostoftheindividual’stime. Since1968,avastliteraturehasarisenwithin economicsthatusesaggregatedatatotesteachofthethreepredictionsofBecker’seco- nomicmodelofcrime. Withregardtotheprobabilityofcapture,alargeliteraturecon- siderstheresponsivenessofcrimetopolice,tovariouspolicepracticesandmodesofde- ployment. Withregardtotheseverityofthesanction,acorrespondinglylargeliterature considers the responsiveness of crime to prison along both the intensive and extensive margin. Withregardto theopportunitycost ofcrime, theliterature hasfocused exten- sivelyontheresponsivenessofcrimetowages,unemploymentratesandemployment conditions. Allthreeliteraturesfind,tovaryingdegrees,supportforBecker’stheory. This dissertationconsiders the role thatvariousinputs play in informingthe supply ofcrimesintheUnitedStates. Thefinalchapter,Chapter3whichisco-authoredwith JustinMcCrary,istightlylinkedtotheextantliteraturethathasdevelopedinresponse toBecker’seconomictheoryofcrime. Thischapterconsiderstheresponsivenessof crimetopolicemanpowerinU.S.cities. Theideaisthatanincreaseinthenumberof police raisean individual’sprobability of capture, thusdicincentivizingand deterring crime (as well as lowering crime via the incapacitation effect of prison). On this questiontheliteraturehasequivocatedovertimewithmorerecentpaperstendingto findevidenceofmodestelasticitiesofcrimewithrespecttopolice. Theseelasticities tend to be larger in magnitude for violent crimes than for property crimes, though thishas remainedan itemof considerable controversy. Sinceviolentcrimes arecon- siderably morecostly tosociety than property crimes, there remains a troublinglack ofconsensus onthe cost-beneficialityofpolice and, assuch, onthe appropriateness ofpublicinvestmentinpoliceasopposedtoprisonsastheprimaryoutletforcrime controldollars. Thisisafirstorderpublicpolicyquestion. Ourcontributiontothisliteratureistogeneratemorepreciseestimatesoftheelastic- itiesofindividualcrimetypeswithrespecttopolicemanpower. Usinganewpaneldata setoncrimeinmediumtolargeU.S.citiesover1960-2010,JustinandIshowthat(1) year-over-year changes in police per capita are largely idiosyncratic to demographic factors,thelocaleconomy,citybudgets,measuresofsocialdisorganization,andrecent changes in crime rates, (2) year-over-year changes in police per capita are mismea- sured, leading many estimates in the literature to be too small by a factor of 5, and (3)aftercorrectingformeasurementerrorbiasandcontrollingforpopulationgrowth, a regression of within-state differences in year-over-year changes in city crimes on within-statedifferencesinyear-over-yearchangesinpoliceyieldseconomicallylarge pointestimates. Ourestimatesaregenerallysimilarinmagnitudeto,butareestimated with a great deal more precision than, those from the quasi-experimental literature. Ourestimatesimplythateachdollarspentonpoliceisassociatedwithapproximately $1.60 in reduced victimization costs, suggesting that U.S. cities employ too few police. Theestimatesconfirmacontroversialfindingfromthepreviousliteraturethatpolice reduceviolentcrimemoresothanpropertycrime. Chapters 1 and 2 of this dissertation consider the effect of immigration on crime in U.S. cities. Immigration can be seen as a shock to the crime market which may beviewedasbeingexogenoustotheeconomicmodelofcrime. Inparticular,while immigrantsmayfacethesameincentivesasnatives,theymaybeselectedquitediffer- 1 ently. Thatis,immigrantsmaybeheavilyselectedwithregardtotheirresponsiveness topolice,prisonsandtheopportunitycostoftheirlabor. Totheextentthatthisistrue, immigrationmayhaveadirecteffectonthecrimeratethatoperatesonlyindirectly withrespecttotheeconomicmodelofcrime. InChapter1,Iconsiderthe“nationaleffectofimmigrationoncrime. Thechapter identifiesacausaleffectofMexicanimmigrationoncrimeusinganinstrumentthat leverages temporalvariationin rainfall in different regions inMexicoas wellas persis- tenceinregionalMexico-U.S.migrationnetworks. Theintuitionbehindtheinstrument is that deviations in Mexican weather patterns isolate quasi-random variation in the assignmentofMexicanimmigrantstoU.S.cities. MyfindingsindicatethatMexican immigration isassociated withno appreciable changein the ratesof either violent or property crimes in U.S. cities. Notably, this is a precisely estimated null effect as I canrejectthataonepercentagepointincreaseintherainfall-inducedshareofMexican migrantsleadstogreaterthanaonepercentincreaseinthecrimerate. Chapter 2 leverages a natural experiment created by recent legislation in Arizona to estimate the impact on crime of an extremely large and discrete decline in the states foreign-born Mexican population. I show that Arizona’s foreign-born Mexi- can population decreased by as much as 20 percent in the wake of the states 2008 implementation of the Legal Arizona Workers Act(LAWA), a broad-based E-Verify lawrequiringemployerstoverifytheimmigrationstatusofnewemployees,coupled with severe sanctions for employer noncompliance. In order to isolate the causal effectofthepassageandimplementationofLAWAoncrime,Ileverageasynthetic differences-in-differences estimator, using a new method of counterfactual estimation proposedbyAbadie,DiamondandHainmuller(2010). Incontrasttopreviousliterature, IfindsignificantandlargeeffectsofMexicanimmigrationonArizonaspropertycrime rate. Results are driven, in large part, by the fact that LAWA resulted in especially disproportionatedeclinesamongMexicanmigrantswhoareyoungandmaleand,as such,theeffectsarepredominantlycompositional. Each of the three chapters contributes to an already large and growing literature withineconomicsthatseekstounderstandthecausesandconsequencesofcriminal activityinurbanareas. Whilemuchremainsunexplained,theliteraturesuggeststhat police manpower is an important predictorof crime. The role of immigration is less clear. Themajorityoftheliteraturesuggeststhatimmigrationisanunimportantfactor intheaggregatecrimeproductionfunction. However,eveniftheeffectsarelargely compositional, recent evidence from Arizona motivates caution in interpreting the aggregateevidence. 2 Chapter 1: What is the Contribution of Mexican Immigration to U.S. Crime Rates? Evidence from Rainfall Shocks in Mexico May14,2013 Abstract This paper identifies a causal effect of Mexican immigration on crime using an instrument that leverages temporal variation in rainfall in dif- ferent regions in Mexico as well as persistence in regional Mexico-U.S. migration networks. The intuition behind the instrument is that devi- ations in Mexican weather patterns isolate quasi-random variation in the assignment of Mexican immigrants to U.S. cities. My findings indicate that Mexican immigration is associated with no appreciable change in the rates of either violent or property crimes in U.S. cities. Notably, this is a precisely estimated null effect as I can reject that a one percentage point increase in the rainfall-induced share of Mexican migrants leads to greater than a one percent increase in the crime rate. 3 I. Introduction Since 1980, the share of the U.S. population that is foreign born has doubled, rising from just over 6 percent in 1980 to over 12 percent in 2010. Compounding this demographic shift, the share of the foreign born population that is of Mexican originalsodoubled,leadingtoaquadruplingofthefractionofU.S.residentswhoare immigrantsfromMexico.1 Overthesametimeperiod,crimeratesincitiesacrossthe UnitedStateshavedeclinedconsiderably,inmanycases,reachinghistoriclows. While theaggregatetimeseriessuggeststhatincreasesinimmigrationfromMexicohavehada protectiveeffectoncrime,publicopinionhasgenerallyreachedtheoppositeconclusion, with a majority of U.S. natives indicating a belief that immigration is associated withincreasesincriminal activity(Espenshade andCalhoun1993). Meanwhile,the consensusintheacademicliteratureisthatimmigrantstotheUnitedStatesare,atworst, nomorelikelytoparticipateincriminalactivitythanU.S.nativesand,atbest,may befarlesslikelytoparticipateincrime(ButcherandPiehl1998a;ButcherandPiehl 1998b;Reid,Adelman,WeissandJaret2005;MoehlingandPiehl2007;Butcherand Piehl2009;Stowell,Messner,McKeeverandRaffalovich2009;Wadsworth2010).2 Whilerecentempiricalwork suggests ananswertotheconundrum, theliterature remains unsatisfyingin severalways. First, whilehistorical researchhas successfully disaggregatedtheeffectofearly20thcenturyimmigrationbynationality,thereislittle researchthataddresses thecriminalparticipationof recentMexicanimmigrants. Since thesearetheimmigrantswhohavebecomesuchasalientissueinrecentpolicydebates, addressing the degree to which Mexican immigration is (or is not) associated with crimewouldappeartobeanissuethatisoffirstorderimportance. Second,whileprior research has employed a variety of plausible identication strategies, chief among them the useof ethnicenclaves asan instrumental variable, concernsregarding the internal validityofthisstrategymotivatesfurtherinvestigation. Finally,themajorityoftheleast squares literatureidentifies an effect ofimmigration on crimeusing long differences, generallyemployingdecennialCensusdata. Whilethisstrategyplausiblyaddressesthe problem ofmeasurement errors in immigrationdata, such analyses lackthe granularity of research designs that employ annual data and are subject to concerns regarding internalmigration ofU.S. nativesin responseto immigrationor otherchangesin local conditions(Borjas2003;2006). Thisresearchaddstotheliteratureonimmigrationandcrimeinseveralimportant ways. First,byutilizingannualratherthandecennialdataontheshareofimmigrants and the crime rate, I am able to estimate the relationship between the two variables atasubstantiallymore granularlevel thanhasbeendoneinpastresearch. Second, by limitingmyanalysistoMexicanimmigration,Iamabletoisolatethespecificmigration 1Asrecentlyas1970,theshareofMexicanimmigrantsintheUnitedStateswasonly1.5percent (HansonandMcIntosh(2010). 2SeeBuonanno,BianchiandPinotti2011forsimilarresearchinasampleofItalianmunicipalities. 4 flows (young, low-skilled immigrants of Mexican origin) that have become such a salientissueinthepolicydebatesurroundingimmigrationreform. Finally,withregard tothecrimeliterature,Iintroduceanovelsourceofidentifyingvariationinconstructing an instrumental variable for the cross-city stock of immigrants in the United States. Specifically,IfollowthegeneralapproachofPugatchandYang(2011)andconstructan instrumentthatcombinesdataonthepermanent(longrun)componentofMexicanstate- U.S.citymigrationrelationswithdataontime-varyingrainfallshocksindifferentMex- icanstates. TheintuitionbehindtheinstrumentisthatdeviationsinMexicanweather patternsisolatequasi-randomvariationintheassignmentofMexicanimmigrantstoU.S. cities. Indeed,IfindstrongevidencethatMexicanimmigrationtotheUnitedStatesis responsivetoMexicanrainfall. Myfindingsindicatethat,onnet,Mexicanimmigration isassociatedwithnoappreciablechangeintheratesofeitherviolentofpropertycrimes inU.S. cities. Notably,thisis apreciselyestimated nulleffectas Icanreject thataone percentagepointincreaseintherainfall-inducedshareofMexicanmigrantsleadsto greaterthana1percentincreaseinviolentcrimesora1.5percentincreaseinproperty crimes. Finally, though I do find evidence that an increase in the share of Mexican migrantsleadstoamodestincreaseinpercapitarobberies,theresultissensitivetothe inclusionofLosAngeles,underscoringtheenormousheterogeneityinthetreatment effectaswellasthedifficultyinidentifyinga“nationaleffect”ofMexicanimmigration. Theremainderofthepaperisorganizedasfollows. SectionIIprovidesadiscussion of identification problemsin this literature as wellas a brief literature review. Section IIIprovidesadiscussionofmechanismsunderlyingthedecisiontomigrate. Section IVpresentstheeconometricframeworkusedtoestimateanaveragecausalresponse ofcrimetoimmigrationandincludesadiscussionoftheidentifyingassumptionsof themodel. SectionVdescribesthedataandsample. SectionVIpresentstheempirical resultsandincludesadiscussionthatlinkstheresultstothoseestimatedintheprior literature. SectionVIIconcludes. II. Conceptual Background A. EmpiricalChallenges Findings in the extant literature arise from two strains of research that attempt to identify the criminal participation of the foreign-born. The first examines the demographiccharacteristicsofinstitutionalizedpopulationsandfindsthatrecentim- migrants are substantially underrepresented among those individuals who reside in aninstitutionalized settingatthe timeofthe decennialcensus. Inparticular,Butcher andPiehl(1998a)findthattheforeign-bornareapproximatelyfivetimeslesslikely tobeinstitutionalizedthannatives,furtherdemonstratingthatthisfigureisunlikely tobe driven substantiallybyselectivedeportation. Theadvantageofresearch designs that compare theinstitutionalization rates of foreign-bornto the native-born is that the descriptivenatureoftheexercisedoesnotrequireaconvincingsourceofidentifying 5 variation.3 However, for several reasons, this line of research may not provide an internally-validandpolicyrelevantestimateofthecontributionofimmigrationtocross- city crime rates. First, since it is not possible to disaggregate the incarcerated from theotherwiseinstitutionalizedusingrecentdata,thevalidityoftheresultingestimates requires anassumption that immigrantsand natives havethe same relativepropensities tobe incarceratedconditionalupon institutionalization.4 Second, theinstitutionalized population, by definition, includes only those individuals who were apprehended, arrestedandsubsequentlyincarceratedforacrime,apotentiallyhighlyselectedsample offoreign-bornoffenders.5 Finally,totheextentthatimmigrationchangesthecalculus ofoffendingamongU.S.natives,anexaminationoftheinstitutionalizationratesofthe foreign-bornfailstocapturegeneralequilibriumeffectsassociatedwithimmigration. Thus, while the approach tostudying the relationship betweenimmigration and crime using individual-level microdata provides an important benchmark of the criminal involvement of the foreign born, this research is not a substitute for an empirical estimateoftheeffectofimmigrationoncrimederivedfromaggregatedata. Asecondstrainofresearchexploitscross-cityvariationinthestocksandflowsofthe foreignbornandreportsassociationsbetweenchangesinthesizeofacity’simmigrant populationanditscrimerate. Thisresearchdesignoffersakeyadvantageinthatthe researcherisable toobserve associationsbetweenimmigrationandcrime thatarenot contingentonanassumptionofequalapprehensionoradjudicationprobabilitiesamong immigrantsand natives.6 However, toachieveidentification, the designnecessarily reliesontheexogeneityofimmigrantlocationdecisions. Totheextentthatimmigrants endogenouslyselectdestinationcitieseitheraccordingtocity-specificcrimeratesor according to other unobserved city and time-varying amenities that are themselves correlatedwithcrime,thetreatmenteffectuncoveredbythisresearchwillbebiased. To see this, consider the following two-way fixed effects model that we might be interestedinestimating:7 CRIME =α+βIMM∗+φ +ρ +ε (1) it it i t it In (1), IMM∗ is the true measure of the immigrant share and φ and ρ are city and it i t 3Moreover, it is important to note that such analyses plausibly capture an effect which is due tosolelytothecriminalityofimmigrants, ratherthananeffectofimmigrationthatisamixtureof immigrantcrimesandcrimescommittedbynatives. 4TheU.S.Censuslastdifferentiatedbetweentheincarceratedpopulationandthepopulationthat isotherwiseinstitutionalizedin1980. 5That said, empirical evidence supports the idea that immigrants may be more likely to be institutionalizedconditionaluponarrest. Inparticular, immigrantsaremorelikelytofacepre-trial detainmentwhich,inturn,increasesthelikelihoodofaconviction(HaganandPalloni1999). 6Theseareastudiesarealsoabletocapturethe“generalequilibrium”effectsofimmigrationinsofar asthesedesignscapturechangesinthebehaviorofnativesthatariseasaresultofimmigration. The cost is that the treatment effect that is captured by such designs may not isolate the criminality of immigrantsthemselves. 7ThisisessentiallythemodelestimatedbyStowell,Messner,McKeeverandRaffalovich(2009). Wadsworth(2010)pursuesasimilarapproach,differencing(1)toremovethefixedeffects. 6