Deportation, Crime, and Victimization

Transcription

1 Deportation, Crime, and Victimization Sandra V. Rozo Steven Raphael Therese Anders Abstract We study whether the forced removal of undocumented immigrants from the United States increases crime and victimization in Mexican municipalities. Using individual and municipal panel data on victimization, safety perceptions, and violent crime matched with annual deportation flows from the United States to Mexico, we assess whether municipalities and individuals in closer proximity to geographic repatriation points experience higher crime and perceived insecurity with surges in deportation flows. We also assess the degree to which victimization is concentrated on repatriated immigrants themselves. We find that individuals with greater geographic exposure to deportation flows have a higher likelihood of being victims of robberies and also experience lower perceived personal safety. Deportees themselves are most likely to be crime victims. JEL Classification: K37, O15. Keywords: Crime, Migration, Latin America. USC Marshall School of Business, sandra.rozo@marshall.usc.edu UC Berkeley Goldman School of Public Policy, stevenraphael@berkeley.edu USC, School of International Relations. We are grateful to Andrea Velasquez for useful suggestions. 1

2 1 Introduction During the 1990s, the global stock of international migrants grew by 2 million migrants per year on average and between 2000 and 2010, the average annual growth of international migration accelerated to roughly 4.6 million migrants annually (UN, 2013). The sharp rise in international migration has created growing concerns regarding economic effects on receiving-country economies, the social integration of immigrants in host countries, as well as concerns that international migration may increase crime and social disorder. Several researchers have studied the effects of migration on crime in destination developed countries. Research on the United States experience tends to find small to no effects of immigrants on crime and an immigrant population that offends at a lower rate relative to the native born (Butcher and Piehl, 1998; Reid et al., 2005; Moehling and Piehl, 2009; Wadsworth, 2010; Stowell et al., 2013; Chalfin and McCrary, 2014). Studies on the experience of Western European countries find that migrants in general do not have a higher propensity to engage in crime (Bell et al., 2013). Undocumented migrants, however, as a result of their legal status and consequent limited opportunities in the formal sector, may offend at a higher rate (Alonso-Borrego et al., 2012; Baker, 2014; Freedman et al., 2014; Pinotti, 2014; Mastrobuoni and Pinotti, 2015). There has been much less research focused on the effects of immigration enforcement policy in host countries on crime and victimization in the origin countries; the issue that we investigate in this article. 1 Identifying these effects is crucial for the adequate formulation of public policy to prevent higher levels of violence and successfully incorporate deported individuals back into society. A link between destination country enforcement policy and origin country crime and victimiza- 1 We are aware of only a few empirical studies of this topic. Using a panel of 45 countries for the years 1970 to 2004, Blake (2014) estimates that increases in deportation rates increase homicide rate in receiving countries. The author instruments deportation rates with the United States unemployment rate of the foreign-born population, immigrant education levels, the size of the immigrant population, and the proximity to the United States. However, it is hard to argue that the homicide rates in one s country of origin is only indirectly related to the education of its migrants through United States deportation policy. Similar exercises are carried out by Jakubowski (2010). Madjd-Sadjadi and Alleyne (2007) use data from a survey of deportees in Jamaica and find that deportations of criminal aliens increase the total number of murders and rapes by approximately 5%. However, since the authors do not have information on deportations, they predict its likelihood using observable covariates in the U.S. Census. 2

3 tion may occur through two main channels. First, selective removal may create concentrated flows of deportees with weak ties to their home country, few skills valued in the formal labor market, but substantial criminal and correctional experiences. Such deportees may turn to criminal activity to generate income, with enforcement policy effectively re-exporting crime back to the home nation. This issue has recently come to the fore with a surge in the migration of unaccompanied minors from Central America to the United States (Nakamura, 2014; Meyer et al., 2014). Press accounts of the unaccompanied minor flow stress the extremely high violence and in particular, the activity of criminal gangs in Central America as a major push factor (Markon and Partlow, 2015; Fisher, 2016). The criminal gangs, in turn, have direct connections to the United States, as many have their genesis in the deportation of criminal aliens (many with very weak social ties to their home nations) from the United States, and even bear the names of and have connections to violent United States street gangs (such as the 18th Street or the MS-13 gangs). Deportees with previous criminal backgrounds, consequently, may become perpetrators of crime upon their return to their home countries (Bain et al., 2006). 2 Second, large scale deportation and strict border enforcement may create a large pool of potential criminal victims among deportees and those trying to migrate illegally. Migrants are often returned to repatriations points located on the United States national border with Mexico, or to a select few repatriation points within the country s interior, not necessarily near home municipalities. Many migrants remain in these areas in the hope of returning to the United States and may fall prey to individuals seeking to rob deportees and perhaps extort their families. In a similar vein, migrants heading north that concentrate in key crossing areas with few social ties on the border may be easy targets and perhaps even attract individuals looking for easy victims. Migrants without criminal backgrounds, but who have lived for long periods of time outside of their home countries, 2 To the extent that the individual s experience in the United States was criminogenic (e.g., migrant youth growing up in and acculturating to the norms of very poor United States neighborhoods, criminality-enhancing experiences within the United States criminal justice system, developed formal ties with United States street gangs, etc.), the migrant may return to the source country more criminally prone than when he or she left for the United States (or relative to the person s counterfactual criminality in the absence of the trip north). Deportees with formal ties to United States street gangs may seek out gangs upon removal, start local divisions of their organization, and even help expand their activities transitionally. 3

4 hence, may be more likely to become crime victims (De León, 2013). This paper explores the effects of regional concentration of deportation flows on local crime rates and whether deportees themselves are at an elevated risk of becoming crime victims. We study deportations from the United States to Mexico between 1995 and While crime rates have been declining over the last two decades in most regions of the world, crime and violent crime in particular, have surged in Central America and Mexico (Di Tella et al., 2010). 3 Moreover, during our study period the United States drastically increased border and internal immigration enforcement activities. Total annual deportations, composed of voluntary returns and forced removals, increased from an annual total of approximately 225,000 in 1995 to a comparable total of 1.7 million in 2013 (DHS, 2014), with more than 95% of the total number of deportees being sent to Mexico and Central America. Policy makers in Mexico and Central America, consequently, continuously express ongoing concerns about the possible effects that forced deportations from the United States may have on the violent crime upsurge in this region (Seelke, 2014). To test for an effect of deportation from the United States on crime in Mexico, we employ annual repatriation data for the period 1995 through 2014 for the 28 Mexican repatriation points through which the United States returns Mexican citizens. We combine these data with panel data at the individual level on victimization (available through the Mexican Family Life Survey) and individual perceptions of safety (from the Mexican surveys on public safety perceptions). We also match these data to panel data at the level of Mexican municipalities on violent crime, including homicides, drug-related deaths, kidnappings, robberies, criminal convictions, cartel presence, and illegal-drugs seizures. Our basic empirical strategy involves testing for an interaction effect between proximity to repatriation points and variation in deportation levels on individual and municipality level crime outcomes. Our estimates include individual (or municipality) fixed effects as well as year fixed effects, and thus flexibly control for time-invariant individual/municipal heterogeneity as well as 3 According to the 2013 Global Study on Homicide, between 1995 and 2012, homicide rates in Central America increased by 51.4%, from 17.5 to 26.5 victims per 100,000 population; positioning the 8 countries of this region as the most violent in the world (UNODC, 2014). 4

5 time trends common to all cross sectional analytical units. Our estimates will be valid so long as there are no other time-varying covariates which are also affecting municipalities located closer and further away from the Mexican repatriation points in different ways. We address these concerns controlling for all the covariates that represented threats to the validity of our estimates, such as municipal outflows of migrants from Mexico to the United States and drug trafficking flows. We find little evidence of an effect of deportation levels on serious violent crime (such as homicides, drug-related deaths, illegal drug seizures, or an indicator of cartel presence). We do, however, find an effect for robbery both as measured in official municipality crime statistics and individual victimization reports. Although individuals more exposed to deportation events do not report changes in their current local safety perceptions, they do report lower safety perceptions for their future and for their country as a whole. We hypothesize, that the higher pessimism is not a consequence of being a direct victim of a crime, but of a stigma that deportees carry in their home countries. To explore whether deportees are the victims or the perpetrators of crime, we use individual data on the number of times that an individual has returned from the United States a measure available in the Mexican Life Family Survey. In our sample, approximately 80% of the individuals who report having returned from the United States did not have legal documents when traveling to the United States. Hence, we use the number of times that an individual has returned from the United States as a proxy for ever being deported. We then compare the victimization rates of individuals who have returned from the United States (and arguably were deported) to the victimization rates of individuals who have never been to the United States. The results of this exercise are consistent with our panel data results. Deportees are at an elevated risk of being robbed, with this risk increasing in the number of times an individual has been to the United States. These findings suggest that, at least in the Mexican case, deportees may contribute to crime rates through their elevated risk of victimization. Our findings may be a lower bound estimate of deportee victimization rates. According to 5

6 a representative survey of deportees from the United States to Central America collected by the Mexican Institute of Migration between October of 2011 and September of 2012, more than 90% of deportees neglect to report crime incidents in which they were the victims due to their distrust of local authorities (Calva et al., 2012). Thus, if anything, we find strong evidence of the high vulnerability of deportees upon their return to their home countries. 2 U.S. Enforcement Policy and the Impacts on Origin Nations For most of the 20th century, United States immigration policy pertaining to undocumented immigration involved border enforcement and limited enforcement of immigration law in the nation s interior. This changed in 1986 with the passage of the Immigration Reform and Control Act (IRCA). In addition to creating an amnesty program for undocumented immigrants with long-term ties to the United States, IRCA made it illegal to knowingly hire an undocumented immigrant and authorized the levying of sanctions on employers who violated either the hiring or paperwork requirements of the law. While very little resources have been devoted to workplace enforcement and investigation (Brownell, 2005), the 1986 law marked a shift in the stance of federal policy towards undocumented immigrants and represents an initial enhancement of enforcement of immigration law in the nation s interior. The 1996 passage of the Illegal Immigrant Reform and Immigrant Responsibility Act (IIRIRA) represents the beginning of what has been dubbed by Rosenblum and Meissner (2014) as the formidable deportation machine. IIRIRA introduced several reforms that increased the speed, volume, and perhaps qualitative composition of deportation activity. First, IIRIRA streamlined the formal removal process by eliminating the requirement for judicial review for individuals apprehended at the border or individuals with a prior removal order. 4 Second, IIRIRA redefined what is 4 The alternative to formal removal is voluntary return, whereby those apprehended simply agree to leave the country. The consequences of voluntary return are much less severe, as formal removal comes with a ban on reentry for a fixed period of time and renders subsequent attempts to enter the United States illegally a felony punishable by federal prison sentences. 6

7 an aggravate felony for the purposes of making one eligible for formal removal from the country. Prior to 1996, unauthorized immigrants and legal permanents residents convicted for relatively serious felonies receiving lengthy prison sentences were eligible for deportation upon completion of their prison terms. IIRIRA made the definition more inclusive and made possible the retrospective application of the law for past (i.e., pre-1996) convictions for both unauthorized immigrants and legal permanent residents. Following the September 11, 2001 terrorist attacks, resources devoted to enforcing immigration law increased greatly. With the creation of the new Department of Homeland Security, resources devoted to both border enforcement (the responsibility of the United States Customs and Border Protection) as well as interior enforcement (the responsibility of the newly-created Immigration and Customs Enforcement agency) expanded. Finally, enforcement of immigration law has come to rely more heavily on information gathered by local law enforcement. This began with the introduction of a memorandum of understanding between willing local law enforcement agencies and the federal government (287(g) agreements) that provided training to local law enforcement and delegates authority to localities to enforce immigration law within its jurisdiction. This was enhanced by the introduction and complete rollout across United States correctional institutions of the Secure Communities program between 2008 and Secure Communities forwards finger prints normally collected in the process of booking criminal defendants following an arrest to the Department of Homeland Security (DHS). When DHS identifies a deportable alien, they issue a 48-hour hold notice to the local authorities to facilitate detention and the commencement of formal removal proceedings. Participation is not voluntary and local agencies cannot opt out of participating. Secure Communities was discontinued in November 2014, and replaced with a more targeted deportation effort titled the Priority Enforcement Program. This constellation of policy changes greatly increased the volume of deportation as well as the qualitative nature of deportations. To start, a greater proportion of deportations in recent years 7

8 involve formal removal proceedings rather than voluntary returns. Second, the proportion of deportation resulting from interior enforcement has increased. Finally, criminal deportations (either due to an actual criminal conviction or an arrest and deportation through the Secure Communities program) have grown considerably. Figure I presents historical data on total deportations, deportations through voluntary returns, and deportations through formal removals. The figure reveals that through most of the 20th century, nearly all deportations took the form of voluntary returns. This changed in 1996 with the passage of IIRIRA and by 2012, over half of all deportations resulted from a formal removal order. Beginning in 1998, the data on removals can be separated into criminal and non-criminal removals. Figure II shows the increasing contribution of criminal removals to formal removal proceedings. The figure reveals a sharp increase between 2008 and 2013 with the introduction and complete rollout of Secure Communities. Criminal status includes those cases in which the DHS has evidence of a conviction. 5 When stratified by the nationality of deportees, we see the disproportionate representation of Mexicans and Central Americans among criminal deportees (Figure II). 3 Description of the Data and Methodological Strategy Our empirical strategy makes use of the fact that deportees from the United States to Mexico pass through a small number of repatriation points, most on the United States border but some in the interior. To be specific, we test for an interaction effect between surges in deportations and proximity to these repatriation points on reported crimes, victimizations, and perceived security. 5 It is important to emphasize that the increase in criminal deportations since 1996 does not necessarily imply more criminal activity among the undocumented as well a legal permanent resident populations. As we have already noted, IIRIRA expanded the definition of a deportable offense and made the definition retroactive. In addition, the increased propensity to use formal removal proceedings rather than voluntary returns renders all who reenter a felon under federal law. The increase in the use of formal removals is likely behind the rapid growth of federal inmates serving time for immigration violations since Finally, Miles and Cox (2014) show that many of the immigrants taken into federal custody under the Secure Communities are classified as low risk by ICE and that the program has had no measurable effect on crime in United States counties (see Blake 2014 and Dingeman and Rumbaut, 2010). 8

9 Our empirical strategy is inextricably linked to the institutional and data infrastructures of the United States and Mexico as they pertain to deportations and crime. Hence, we begin by describing the data sets used in this paper. We then detail our empirical strategy for testing for an effect of deportation activity on crime. We should mention from the outset an important qualification. Namely, we are studying the impacts of deportations on crime and victimization in a specific region within Mexico. Given the historically large volume of population movements between Mexico and the United States, it is not immediately obvious that the findings here generalize to other areas that have seen increases in criminal deportations from the United States (El Salvador, Guatemala, or Honduras for example). Mexico has a longer history of circular migration and strong ties between specific Mexican sending communities and receiving communities in the United States. Central American migrants on the other hand, may have been more disconnected from their homelands upon deportation given the greater physical distance from the United States and more recent migration history. 3.1 Data description We employ various data sources to measure deportation activity from the United States as well as crime rates and individual victimization in Mexico. Our principal data source for the geography of deportation activity comes from the Unidad de Política Migratoria (UPM) of the Mexican Secretaría de Gobernación. The UPM data measures repatriation events of Mexican nationals initiated by the United States for the years 1995 through Repatriation flows are counted as events rather than individuals since the same individual may be repatriated more than once. Most importantly for our purposes, the UPM data provide repatriation event counts by year and repatriation point, permitting measurement of variation over time and across location in the regional impact of deportation activity initiated by the United States. Between 1995 and 2014, we observe 28 separate designated repatriations points in the UPM data with the following distribution across Mexican 9

10 states: 6 5 in Baja California, 5 in Chihuahua, 2 in Coahuila, 7 in Sonora, 7 in Tamaulipas, 1 in the Distrito Federal (Benito Juárez Airport), and 1 in Jalisco (Guadalajara Airport). Of these 28 locations, 18 had non-zero repatriation event counts as of The location of the repatriation points along the U.S.-Mexico border is presented in Figure III. Comparison of repatriation event totals from the UPM to those measured by the Department of Homeland Security reveals similar time series patterns in overall deportation totals (for example, the sharp reduction on the repatriation events in 2001 and 2002 see Figure IV). To characterize who is deported from the United States to Mexican repatriation points, we present a series of descriptive statistics from the annual Border Survey of Mexican Migration, [Encuesta sobre Migración en la Frontera Norte de México (EMIF)]. The EMIF is a survey of deportees collected by El Colegio de la Frontera Norte, a government-funded research institution whose main campus is located in Tijuana. The EMIF is conducted at selected points and airports in the interior of Mexico. The objective of the EMIF is to provide unbiased estimates of the size, composition, and demographic characteristics of Mexican labor flows that cross the U.S.-Mexican border. EMIF s data corresponds to events rather than individuals since a migrant can migrate more than once. 7 Rounds of data collection are conducted regularly for each quarter of a year at bus and train stations, highways, airports, and ports of entry in 10 Mexican border cities 8 and 4 international airports (Benito Juárez, Guadalajara, Morelia, and León). EMIF allows identification of individuals 15 years of age or older who where returned or removed by United States authorities. We use annual data from the 1999 to 2014 to characterize the population of migrants who were deported and returned to Mexico by the United States border patrol. Figures V and VI present descriptive statistics for recent deportees for 1999 through Mexico is divided in 32 federal entities including 31 states and 1 federal district. The country could further be subdivided into 2,438 municipalities 7 Other studies that have used deportee surveys such as the EMIF to characterize deportees include Chort and de la Rupelle (2015), Rendall et al. (2009), and Madjd-Sadjadi and Alleyne (2007). 8 Including Tijuana, Mexicali, Nogales, Agua Prieta, Ciudad Juárez, Ciudad Acuna, Piedras Negras, Nuevo Laredo, Reynosa, and Matamoros. 10

11 Most deportees are men, 20 to 30 years of age, with low levels of education. By 1999 most repatriations events involve individuals who had spent less than a day in the United States, who had no family either in the United States or Mexico, had tried to cross the border more than once, and who were planning to attempt to cross the border again. Most report being deported from Arizona, California, and Texas. Curiously, most individuals in Mexico after deportation, still report the United States as their actual state of residence, perhaps indicative of their intentions to attempt to cross again and the strength of their existing ties to the United States. Regarding our key dependent variables, we employ individual-level data on victimization and perceived safety as well as municipal-level crime data. Individual-level data on victimization comes from the Mexican Family Life Survey (MXFLS) while we draw information on safety perceptions from the Survey on Public Safety Perceptions (ECOSEP, the acronym in Spanish). Municipal violent crime is measured in six main variables described below. The MXFLS is a panel multi-thematic survey representative of the Mexican population at the national, urban, rural, and regional level. The panel contains information for each individual over a 10-year period, collected in three rounds: 2002, , and The MXFLS collects information on socio-demographics, labor market, time allocation, tastes and habits, health measures, insurance, migration, fertility, and victimization. Concerning the victimization module, the MXFLS collects information on the number of times that an individual has been victimized by type of event (i.e., kidnapped, sexually abused or harassed, robbed, participated in physical fight, or was a victim of any other crime). Appendix A presents descriptive statistics of the sample for each of the waves. The ECOSEP was collected by the Mexican Statistics Department (INEGI, for its Spanish name) from 2009 to 2013 in the 32 biggest urban centers of Mexico. 9 The data set is a monthly rotating panel with information on socio-demographics, municipality, and safety perceptions. Each individual is observed over the course of three years during four consecutive months. It includes approximately 99,000 observations or 17,222 individuals across the five years available. Appendix 9 The data for 2013 is only available for March and June. 11

12 A presents descriptive statistics for the survey. Safety perceptions are measured through five variables that identify the individual s current and future safety perceptions for the country and the respondent. Perceptions are measured on a Likert-type scale from 0 to 5, where 5 described the highest level of perceived safety. Finally, we assemble municipal level crime data from various sources. The outcomes we use include i) homicides from INEGI (available between 1990 and 2013); ii) annual drug related deaths as reported by the Mexican Presidency (available between 2006 and 2011); iii) the number of reported kidnappings and robberies from the Mexican Procuraduría General de la República (available between 2006 and 2011), iv) annual number of criminal convictions also from the Mexican Procuraduría General de la República (available between 1997 and 2006), v) annual cartels presence as constructed by Coscia and Rios (2012) (available from 1990 to 2010), and vi) annual illegal-drug seizures as collected by the Mexican government (available between 2006 and 2011). Descriptive statistics of these variables are presented in Appendix B. Figure VII presents trends in homicide rates averaged over all municipalities as reported by INEGI. Mexican homicide rates decreased between 1990 to 2005, and then increased drastically between 2006 and According to several studies, violence in Mexico dramatically intensified due to three main factors (Robles et al., 2013): exogenous changes to the narcotics market, including above all the relative success of Colombia at counter narcotics operations and drug seizures (Castillo et al., 2014); the increased fragmentation of drug cartels into numerous smaller organizations and criminal cells; and the militarized fight against drugs and drug trafficking that began during the administration of President Felipe Calderón (Dell, 2015). 3.2 Research Methodology The sign of the causal effects of deportation from a host country (e.g., the United States) on the crime levels of an origin country (e.g., Mexico) is theoretically ambiguous and likely to vary across alternative national settings. To the extent that local authorities take the appropriate steps to rein- 12

13 tegrate their nationals, or to the extent that deportees have strong family and social ties in their country of origin, deportation flows may not impact crime at all. Moreover, it is entirely possibly that migrants, being a self-selected group oriented towards work, may be less likely to offend, and thus their repatriation may actually lower the number of crimes per capita. On the other hand, one can identify causal channels through which deportations may increase crime rates in the receiving country. First, United States policy selectively deports individuals who become involved with the criminal justice system either through convictions, or in more recent years, simple arrests and bookings. To the extent that those involved with the United States criminal justice system are more likely to commit crime in their home country (either due to individual heterogeneity or as a result of their criminal justice involvement), deportations could increase crimes per capita. Second, even if the deportees had no previous criminal history, those with very weak ties to their origin nation (for example, immigrants who migrated to the United States at a young age, who have been in the United States for many years, or whose extended familial network resides in the United States) may have scant employment opportunities in the home country and turn to crime as a result. Finally, deportees themselves may be at an elevated risk of victimization. To the extent that their social ties are weak due to time in the United States or being deported to a repatriation point away from their region of origin, deportees may be easy targets. Moreover, those immigrants with family in the United States may also be easy targets for extortion. To test for an effect of deportation flows from the United States on Mexican violent crime, we test for an interaction effect between physical proximity to repatriation points and repatriation volume on crime outcomes. To measure proximity, or alternatively stated, exposure to deportation events, we construct a municipality-level index (Dep mt ) which is a distance-weighted average of the repatriation events across the 28 Mexican repatriation points. The weights are constant in time and correspond to the inverse distance from the centroid of the municipality (m) where the 13

14 individual is located to each repatriation point (k). Formally: 28 Dep mt = [repatriation events kt (1/distance mk )] (1) k=1 We standardize the deportation exposure to be mean zero and have a standard deviation of one for ease of interpretation. Using this measure, we estimate the effects of deportation exposure on violent crime using the following specification: V imt = α 0 + α 1 Dep mt + γ i + γ t + X mt φ + ɛ it (2) where V imt is the violence indicator observed or reported by individual i, located at municipality m, and year t; Dep mt corresponds to the deportation exposure indicator; γ i and γ t are fixed effects by individual (or municipality) and year; and X mt represents a matrix of municipal time-varying controls. Standard errors were clustered by municipality and also corrected for heteroskedasticity. Given our identification strategy we are effectively comparing the changes in victimization and violent crime observed in municipalities closer and further away from the 28 U.S.-Mexican repatriation points when there are surges of deportation flows. Because our estimates include fixed effects by individual (or municipality) and year, they are not threatened by static differences between municipalities or individuals, nor by aggregate time trends. For instance, differences in violent crime that are explained by static characteristics such as being near the United States frontier are not a threat to our identification. Additionally, changes in macroeconomic trends in the United States or Mexican economy are accounted for by the time fixed effects. Our identification strategy will only be threatened if E[ɛ it Dep mt γ i, γ t, X mt ] 0 that is if, for example, there are other time-varying covariates changing in time, which also affect areas close and further away from the repatriation points in different ways. It may be the case, for example, that population outflows changed during 1995 and 2014 and 14

15 that these changes were different in areas closer to the repatriation points. To check for the sensitivity of our estimates to this issue, we introduced three types of municipal controls i) population, ii) migration outflows from Mexico to the United States by state, and iii) a municipal migration outflow index. 10 Our results are robust to the inclusion of these controls (see Table I). 11 Another possible threat is that changes in drug trafficking flows may affect violence near the repatriation points. To test for these possibility we include controls for cartels presence in all of our estimates. The results are presented in Table I and are robust to their inclusion. In addition to the exercises carried out above, for the estimates using MXFLS data, we further control for covariates such as age, marital status, dwelling ownership, number of rooms, education level, student status, income, if the individual carried valuables, and the number of times the individual saw the police. The results are robust to the inclusion of all these covariates. 4 Results 4.1 Deportation Effects on Victimization Estimation results for the specification described by equation (2) are presented in Tables I and II. Table I presents the estimates using as dependent variables individual-level victimization outcomes from the MXFLS data. These measures include the number of times the respondent has been i) kidnapped, ii) harassed, iii) robbed, iv) participated in physical fights, v) been a victim of a crime 10 Municipal population comes form the population censuses and was collected by the Instituto Nacional de Estadística y Geografía (INEGI), the Mexican statistics agency. Censuses were collected every five years beginning in The last population census was collected in Migration outflows by state were collected in 1995 and then from 1999 to 2014 in the Border Survey of Mexican Migration in the Northern Frontier collected by the migration Mexican agency (Instituto Nacional de Migración). The outflows correspond to the count of the number of events of international migration from Mexico into the United States. The migration outflow index was constructed by the Consejo Nacional de Población, the Mexican demographic planning agency in 2000 and It is constructed from the population censuses as an average of the percentage of households receiving remittances, the percentage of households with members in the United States, the percentage of households with circular migrants between the United States and Mexico, and the percentage of households with members returning voluntarily from the United States. 11 For the municipality estimates given the time availability of the results it was only possible to include controls for migration outflows by state. 15

16 other than the ones mentioned above, and vi) been a victim of any crime (i.e., cumulative crime victimization). The results suggest that deportation exposure has a robust significant effect on the number of robbery victimization. We also find a significant effect of deportation exposure on kidnappings, however, these effect sizes are relatively small. The results for robbery and kidnapping are robust to the inclusion of individual fixed effects, year fixed effects, additional individual sociodemographic time-varying covariates, and other controls to account for the base-line risk of being exposed to crime (such as whether individuals carried valuables, owned a weapon, or the number of times the respondent saw the police). The OLS estimate in column 18 indicates that a one standard deviation increase in the deportation exposure index increases the number of robberies at the individual level by 0.04 (equal to roughly 30 percent of the dependent variable mean). This is approximately the same effect observed on the total number of times an individual is victim of any crime (see column 36, equal to roughly one quarter of the dependent variable mean). Table II presents estimation results where the dependent variables are municipal-level homicides, the number of criminal convictions, a dummy variable for Mexican cartel presence, drugrelated deaths, average reported kidnappings, average reported robberies, and illegal-drug seizures (these variables are reported in absolute numbers but we control for municipal population size). The results for these outcomes find little evidence of an impact on local homicide levels, criminal convictions, and kidnappings. Interestingly, we find no evidence of an impact of deportation exposure on the degree of cartel presence, drug-related deaths, or drug seizures. Note, much of the increase in homicide in Mexico in the late 2000 s has been attributed to inter-cartel conflict and the effort by the Mexican government to combat major drug cartels (Dell, 2015). A priori, one might expect cartel activity to be linked to deportation levels to the extent that deportees themselves are involved with cartels or if cartels are involved in exploiting those who are deported. We find little evidence of this. We do however find a marginally significant effect of deportation on robbery levels. Note, 16

17 the robbery data here involve robberies reported to police and thus, provides a data source that is completely independent of the individual victimization data from the MXFLS. The fact that both show an impact on robbery is reassuring. 4.2 Deportation Effects on Security Perceptions Table III presents estimates of equation (2) using the Survey on Public Safety Perceptions. The estimates include municipality, annual, and monthly fixed effects, as well as individual controls for gender, age, marital status, education level, and income. Since the key dependent variable is measured at the level of the municipality, we cluster the standard errors at the municipality level. The results suggest that the individuals who are located in municipalities more exposed to deportation believe that their personal safety is lower (though this effect is not significant in all specifications), that their future safety is lower, and that the country as whole is less safe. Since safety perceptions are lower when individuals feel unsafe, the results suggests that deportation is associated with higher levels of risk. We hypothesize that this result may indicate that as a result of higher deportation inflows individuals have a greater awareness of violent crime (for example through the local news or conversations with other members of the local community), but that they are not direct victims of crime. Given that the dependent variables are measured on a Likert-type scale, it is difficult to interpret the cardinal value of the effect size. 4.3 Are Deportees Victims or Perpetrators? So far, our estimates suggest that individuals in close proximity to deportation events are at an elevated risk of becoming a robbery victim. We find little evidence of an impact of deportation on drug trafficking activity or crimes related to cartel activity. The question remains, however, as to whether the deportees are the victims or the perpetrators of these types of crimes. In this section, we try to address this question using information pertaining to immigration history available in 17

18 the MXFLS. In particular, the survey has detailed information on the past migration history of all interviewees. The information includes prior residential locations as well as whether they were legal residents when they resided outside of Mexico. Based on this information, we constructed a variable measuring the number of times that an individual has been returned from the United States. Most of the variation in this variable is between rather than within person. Hence, in these models we exclude individual fixed effects as their inclusion leaves little variation in this explanatory variable. We run a regression of our victimization variables (i.e., number of times an individual has been kidnapped, harassed, has participated in fights, has been a victim of other crime, and the cumulative number of times the individual has been a victim of any crime during the last year) on the number of times that an individual has returned from the United States to Mexico. 12 Only 2% of the sample has ever returned from the United States to Mexico, yet of this population, 78.68% of the individuals reported that they did not have legal documents to be in the United States. Hence, the number of times that an individual has returned from the United States may be a good proxy for being a deportee (it remains true, however, that we do not observe whether the individuals returned voluntarily or involuntarily from the United States to Mexico). The results of this exercise are presented in Table IV. They suggest that the probability of being the victim of a crime is increasing in the number of times that an individual has returned from the United States. The coefficients are only significant for the probability of being the victim of a robbery (see column 3), which is ultimately reflected in the cumulative number of times that an individual is a victim to any type of crime (see column 6). The estimates suggest that the deportees are actually the victims of crime, but not the perpetrators of these actions. Our findings are in-line with the findings of a recent survey collected by the Mexican Institute of Migration (INM) [Instituto Nacional de Migración] and El Colegio de la Frontera del Norte (CFN), a migration Mexican research think tank (see Calva et al., 2012). Between October of 12 The estimates include fixed effects by year and the same controls included in the estimates presented in Table I. Additionally, the standard errors are clustered at the individual level. 18

19 2011 and September of 2012, the INM and CFN collected a survey on aggression and abuse of migrants. The survey was applied to nationals from Guatemala, El Salvador, Honduras, and Mexico who were returned by the migratory authorities of the United States or Mexico. 13 Their sample includes approximately 514 thousand observations, of which 70% corresponds to Mexican nationals returned by the United States. Of these total, 12.3% of the migrants reported that they were victims of some of form of abuse or crime. In line with our results, approximately 75% of the crime events experienced by migrants correspond to robberies and extortion. The information in the survey also allows us to see who are the perpetrators of these crimes. The migrants report that the main perpetrators of these crimes are gangs (34%), the police (31%), and coyotes (16%). Finally 96% of the victims of abuse reports that they will not report these actions to the police, which suggests that our gauges (in particular the municipality-level crime rates) may be underestimating deportee victimization rates. Our results are also consistent with previous analysis carried out by Thomson and Cohen (2014) who explore 3.2 million individual actions taken by the United States Immigration and Customs Enforcement between 2007 and Thompson finds that approximately two thirds of the deportations involve people who had committed minor infractions, including traffic violations, or had no criminal record whatsoever. According to the article only 20% of the cases involved people convicted of serious crimes, including drug-related offenses. 5 Conclusions This paper assesses whether deportation flows affect crime in deportee receiving countries. We focus on the case of deportations from the United States to Mexico, the main recipient country of 13 The surveys of Mexicans returned from the US were collected in 8 cities of the northern frontier including Matamoros, Reynosa, Nuevo Laredo, Ciudad Acuña, Ciudad Juárez, Nogales, Mexicali, and Tijuana. The individuals returned from the United States to the other Central American countries were interviewed in the international airports of Guatemala, Tegucigalpa, and San Salvador. Finally, the individuals returned from Mexico to the other Central American countries were interviewed in the frontier cities of Tecin Uman, El Carmen, y la Mesilla (in Guatemala), El Corinto (in Honduras), and La Hachadura in El Salvador. 14 Their analysis excludes information on deportations handled exclusively by Border Patrol. 19

20 deportees from the United States. We investigate whether individuals (or municipalities) in closer proximity to each of the 18 U.S.-Mexican repatriation points experience higher crime victimization and perceived insecurity with surges in deportation flows. Our results suggest that municipalities and individuals more exposed to deportation have higher victimization rates operating primarily through an increase in robberies. We find little evidence of deportation flows on homicides, cartel activity, or illegal-drugs seizures. Interestingly, we find that those who have been deported in the past are at an elevated likelihood of becoming crime victims, with again most of this effect operating through robbery. This finding in conjunction with our deportation exposure results leads us to conclude that the victimization of deportees themselves is driving our principal exposure result. Our analysis here is limited to the contemporaneous effect of deportation activity on crime. One might hypothesize that there are longer term effects of sustained deportations on local crime rates. For example, to the extent that those seeking to victimize the vulnerable who migrate may relocate closer to repatriation points, the crime fundamentals of areas with greater exposure to deportees may be altered through sustained influx of deportees. Moreover, to the extent that deportees themselves turn to crime as a result of being socially isolated or imperfectly reintegrated into Mexican society, longer term effects may differ from the short term contemporaneous estimates here. We should also emphasize that the findings here may not necessarily generalize to other important examples where countries are on the receiving end of deportees. For example, the Central American nations of Guatemala, El Salvador, and Honduras, have recently received much attention due to elevated gang activity, having the highest homicide rates in the world, and the ensuing migration of unaccompanied minors to the United States fleeing violence. Moreover, the major gangs operating in these nations have direct ties to street and prison gangs in the United States, presenting a prima facie link between deportation activity and crime rates in receiving nations. To be sure, the volume and qualitative nature of migratory flows in both directions between the United States and 20

21 Mexico likely differs fundamentally from migratory flows between the United States and Central America. Future research on this relationship in other national settings is certainly merited. References Alonso-Borrego, C., N. Garoupa, and P. Vázquez (2012). Does immigration cause crime? evidence from Spain. American Law and Economics Review, Bain, B., S. Logan, and K. Kairies (2006, March). Deportation feeds a cycle of violence in Central America. Americas Program. International Relations Center. Published online. Baker, S. R. (2014). Effects of immigrant legalization on crime: The 1986 immigration reform and control act. Technical report, Stanford Law and Economics Olin Working Paper. Bell, B., F. Fasani, and S. Machin (2013). Crime and immigration: Evidence from large immigrant waves. Review of Economics and Statistics 21(3), Blake, G. O. (2014). America s deadly export: Evidence from cross-country panel data of deportation and homicide rates. International Review of Law and Economics 37, Brownell, P. (2005). The declining enforcement of employer sanctions. Migration Information Source 1, Butcher, K. F. and A. M. Piehl (1998). Cross-city evidence on the relationship between immigration and crime. Journal of Policy Analysis and Management 17(3), Calva, L., A. Castañeda, M.-L. Coubes, and M. D. Paris (2012). Principales resultados de la encuesta sobre agresion y abuso a migrantes devueltos por las autoridades migratorias. Technical report, El Colegio de la Frontera del Norte. Castillo, J. C., D. Mejia, and P. Restrepo (2014). Scarcity without leviathan: The violent effects 21

22 of cocaine supply shortages in the Mexican drug war. Technical report, Available at SSRN Chalfin, A. and J. McCrary (2014). Criminal deterrence: A review of the literature. Journal of Economic Literature. Chort, I. and M. de la Rupelle (2015, August). Determinants of mexico-us outward and return migration flows: A state-level panel data analysis. Working paper, Dauphine Université Paris. Coscia, M. and V. Rios (2012). Knowing where and how criminal organizations operate using web content. Technical report, Proceedings of the 21st ACM international conference on Information and knowledge management. De León, J. (2013). The efficacy and impact of the Alien Transfer Exit Programme: Migrant perspectives from Nogales, Sonora, Mexico. International Migration 51(2), Dell, M. (2015). Trafficking networks and the mexican drug war. The American Economic Review 105(6), DHS (2014) yearbook of immigration statistics. Technical report, U.S. Department of Homeland Security. Di Tella, R., S. Edwards, and E. Schargrodsky (2010). The economics of crime: Lessons for and from Latin America. University of Chicago Press. Dingeman, M. K. and R. G. Rumbaut (2010). The immigration-crime nexus and post-deportation experiences: En/countering stereotypes in Southern California and El Salvador. University of La Verne Law Review 31(2), Fisher, S. (2016, August). Join a gang or die: Deported Central American teens face harrowing choice. Newsweek. Freedman, M., E. Owens, and S. Bohn (2014). Immigration, employment opportunities, and criminal behavior. Technical report, Unpublished. 22

23 Jakubowski, J. R. (2010). Do Criminal Deportations Affect Homicide Rates in Central America. Master s thesis, Graduate School of Arts & Sciences at Georgetown University, Washington, D.C. Madjd-Sadjadi, Z. and D. Alleyne (2007). The potential Jamaican impact of criminal deportees from the U.S. Journal of Ethnicity in Criminal Justice 5(2-3), Markon, J. and J. Partlow (2015, December). Unaccompanied children crossing southern border in greater numbers again, raising fears of new migrant crisis. Washington Post. Mastrobuoni, G. and P. Pinotti (2015). Legal status and the criminal activity of immigrants. American Economic Journal: Applied Economics 7(2), Meyer, P. J., C. R. Seelke, M. Taft-Morales, and R. Margesson (2014). Unaccompanied children from Central America: Foreign policy considerations. Current Politics and Economics of South and Central America 7(3), 463. Miles, T. J. and A. B. Cox (2014). Does immigration enforcement reduce crime? evidence from Secure Communities. Journal of Law and Economics 57(4), Moehling, C. and A. M. Piehl (2009). Immigration, crime, and incarceration in early twentiethcentury america. Demography 46(4), Nakamura, D. (2014, August). Influx of minors across Texas border driven by belief that they will be allowed to stay in United States. Washington Post. Pinotti, P. (2014). Clicking on heaven s door: The effect of immigrant legalization on crime. Baffi Center Research Paper ( ). Reid, L. W., H. E. Weiss, R. M. Adelman, and C. Jaret (2005). The immigration crime relationship: Evidence across us metropolitan areas. Social Science Research 34(4),

24 Rendall, M. S., E. Aguila, R. Basurto-Dávila, and M. S. Handcock (2009, May). Migration between Mexico and the U.S. estimated from a border survey. Technical report, RAND Labor and Population program and Population Research Center. Robles, G., G. Calderón, and B. Magaloni (2013). The economic consequences of drug trafficking violence in Mexico. Poverty and Governance Series Working Paper, Stanford University. Rosenblum, M. R. and D. Meissner (2014). The deportation dilemma: Reconciling tough and humane enforcement. Washington, DC: Migration Policy Institute. Seelke, C. (2014). Gangs in Central America. Congressional Research Service. Stowell, J. I., S. F. Messner, M. S. Barton, and L. E. Raffalovich (2013). Addition by subtraction? a longitudinal analysis of the impact of deportation efforts on violent crime. Law & Society Review 47(4), Thomson, G. and S. Cohen (2014). More deportations follow minor crimes, records show. The New York Times. April UN (2013). World migration in figures. Technical report, United Nations, Population Division. UNODC (2014). Global study on homicide. Technical report, United Nations of Drugs and Crime. Wadsworth, T. (2010). Is immigration responsible for the crime drop? an assessment of the influence of immigration on changes in violent crime between 1990 and Social Science Quarterly 91(2),

25 Table I: Effects of Deportation on Individual Victimization (MXFLS Data) Kidnapped (N. Events) Harassed (N. Events) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Deportation Exposure (Dep) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Migration Outflow Index (Mun) (0.000) (0.000) Population (0.000) (0.000) Migration Outflow (State) (0.000) (0.000) Cartels Presence (0.001) (0.001) R-squared Obs Robbery (N. Events) Fight Participation (N. Events) (13) (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) Deportation Exposure (Dep) (0.004) (0.004) (0.005) (0.004) (0.004) (0.004) (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) Migration Outflow Index (Mun) (0.007) (0.001) Population (0.000) (0.000) Migration Outflow (State) (0.000) (0.000) Cartels Presence (0.021) (0.001) R-squared Obs Controls for all Panels Individual FE Y Y Y Y Y Y Y Y Y Y Y Y Year FE Y Y Y Y Y Y Y Y Y Y Y Y Individual Controls Y Y Y Y Y Y Y Y Y Y Notes: The estimates use data from the Mexican Life Family Survey. Deportation exposure is measured according to equation (1), it was transformed to standard deviations to ease interpretation. Individual controls include: marital status, if the individual owns a dwelling, number of rooms, education level, if the individual is a current student, income, if the individual carries valuables, the number of times the individual saw the police, and if the individual carries any weapons. Standard errors are clustered at the municipal level. : significant at the 1%, : significant at the 5%, and : significant at the 10% level. 25

26 Table I (cont d): Effects of Deportation on Individual Victimization (MXFLS Data) Victim of Other Crimes (N. Events) Cumulative Crime Victimization (N. Events) (25) (26) (27) (28) (29) (30) (31) (32) (33) (34) (35) (36) Deportation Exposure (Dep) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.004) (0.004) (0.005) (0.004) (0.004) (0.004) Migration Outflow Index (Mun) (0.000) (0.007) Population (0.000) (0.000) Migration Outflow (State) (0.000) (0.000) Cartels Presence (0.001) (0.021) R-squared Obs Controls for all Panels Individual FE Y Y Y Y Y Y Y Y Y Y Y Y Year FE Y Y Y Y Y Y Y Y Y Y Y Y Individual Controls Y Y Y Y Y Y Y Y Y Y Notes: The estimates use data from the Mexican Life Family Survey an individual panel survey collected in three rounds: 2002, , and Deportation exposure is measured according to equation (1), it was transformed to standard deviations to ease interpretation. Individual controls include: marital status, if the individual owns a dwelling, number of rooms, education level, if the individual is a current student, income, if the individual carries valuables, the number of times the individual saw the police, and if the individual carries any weapons. Standard errors are clustered at the municipal level. : significant at the 1%, : significant at the 5%, and : significant at the 10% level. 26

27 Table II: Effects of Deportations on Municipal Crime Homicides Convicted Criminals Cartels Presence Drug-related deaths Kidnappings Robbery (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Dep. Exposure (2.63) (1.88) (4.90) (4.07) (0.02) (0.01) (15.70) (15.68) (0.12) (0.12) (19.82) (14.68) R-squared Obs Period Available Cocaine Seizures Heroin Seizures Marijuana Seizures Meth. Seizures (13) (14) (15) (16) (17) (18) (19) (20) Dep. Exposure (4.72) (4.69) (0.93) (0.91) (190.56) (187.23) (9.83) (9.75) R-squared Obs Period Available Controls for all Panels Mun. FE Y Y Y Y Y Y Y Y Y Y Y Y Year FE Y Y Y Y Y Y Y Y Y Y Y Y Population Controls Y Y Y Y Y Y Notes: Homicides are obtained from INEGI. Convicted crimes come from the Mexican Procuraduría General de la República. Cartels presence comes from Coscia and Rios (2012). Illegal-Drug related deaths are reported by the Mexican Presidency; kidnappings and robberies come from the Mexican Procuraduría General de la República. Illegal-drug seizures come from the Mexican government. Seizures are expressed in kilograms. Deportation exposure is measured according to equation (1), it was transformed to standard deviations to ease interpretation. Standard errors are clustered at the municipality level. Estimates with are significant at the 1%, those with are significant at the 5%, and those with are significant at the 10% level. 27

28 Table III: Effects of Repatriations on Safety Perceptions Indv. Safety Today [1 to 5(highest)] Indv. Future Safety [1 to 5(highest)] Country s Safety Today [1 to 5(highest)] (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. Exposure (0.023) (0.208) (0.211) (0.012) (0.113) (0.116) (0.010) (0.119) (0.120) R-squared Obs Mun FE Y Y Y Y Y Y Month FE Y Y Y Y Y Y Year FE Y Y Y Y Y Y Controls Y Y Y Country s Future Safety [1 to 5(highest)] Safety Walking Alone [1 to 5(highest)] (10) (11) (12) (13) (14) (15) Dep. Exposure (0.011) (0.089) (0.092) (0.011) (0.103) (0.106) R-squared Obs Mun FE Y Y Y Y Month FE Y Y Y Y Year FE Y Y Y Y Controls Y Y Notes: The ECOSEP was collected by INEGI from 2009 to 2013 in the 32 biggest urban centers of Mexico. Individual controls include gender, age, marital status, education level, and income. Standard errors are clustered at the municipal level. : significant at the 1%, : significant at the 5%, and : significant at the 10% level. 28

29 Table IV: Victimization of Individuals that Returned from the United States (MXFLS data) Kidnapped Harassed Robbed Fights Participation Victim of Other Crimes Cumulative Crime Victimization (1) (2) (3) (4) (5) (6) N. of returns from the US (0.006) (0.006) (0.064) (0.013) (0.001) (0.068) Age (0.000) (0.000) (0.001) (0.000) (0.000) (0.001) Male (0.002) (0.002) (0.022) (0.003) (0.002) (0.023) Married (0.002) (0.002) (0.023) (0.003) (0.001) (0.025) Household Owns House (0.002) (0.002) (0.030) (0.002) (0.001) (0.031) Rooms (0.000) (0.001) (0.010) (0.001) (0.001) (0.010) Elementary education (0.002) (0.002) (0.024) (0.003) (0.002) (0.027) Secondary Education (0.003) (0.003) (0.024) (0.003) (0.002) (0.026) Current Student (0.005) (0.004) (0.056) (0.014) (0.006) (0.060) Income (thousands of pesos) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Out at night? [1(lowest) to 4] (0.001) (0.001) (0.015) (0.002) (0.001) (0.015) Carry valuables? [1(lowest) to 4] (0.001) (0.002) (0.014) (0.002) (0.001) (0.014) N Seeing Police (0.000) (0.000) (0.001) (0.000) (0.000) (0.001) Carry weapon (0.001) (0.010) (0.090) (0.009) (0.010) (0.088) R-squared Obs Year FE Y Y Y Y Y Y Notes: The estimates use data from the three waves of the Mexican Life Family Survey. Standard errors are clustered at the municipal level. : significant at the 1%, : significant at the 5%, and : significant at the 10% level. 29

30 Figure I: DHS Deportations (Number of Individuals) Notes: Removals are compulsory and account for the confirmed movement of an inadmissible or deportable alien out of the United States based on an order of removal. An alien who is removed has administrative or criminal consequences placed on subsequent reentry owing to the fact of the removal. Returns are the confirmed movement of an inadmissible or deportable alien out of the United States not based on an order of removal. Source: United States Department of Homeland Security. 30

31 Figure II: Removals by Country of Nationality and Criminal Status(Number of Individuals) Notes: Removals are the compulsory and account for the confirmed movement of an inadmissible or deportable alien out of the United States based on an order of removal. An alien who is removed has administrative or criminal consequences placed on subsequent reentry owing to the fact of the removal. Source: United States Department of Homeland Security. 31

32 Figure III: Repatriation Points in the US-Mexican Frontier Notes: According to the Unidad de Política Migratoria of the Mexican Secretaría de Gobernación, there were 28 repatriation points in Mexico between 1995 and 2014: 5 in Baja California, 5 in Chihuahua, 2 in Coahuila, 7 in Sonora, 7 in Tamaulipas, 1 in the Distrito Federal (Benito Juárez airport), and 1 in Jalisco (Guadalajara airport). Of these 18 remained open by Source: Authors illustration based on information by the Unidad de Política Migratoria of the Mexican Secretaría de Gobernación and the Centers for Disease Control and Prevention. Figure IV: Mexican Repatriations (Number of Events) Notes: The figure shows the annual number of events of repatriation of Mexican nationals by the United States migratory authorities between 1995 and Repatriation flows are counted in events since the same individual may be repatriated more than once. Source: Unidad de Política Migratoria (UPM) of the Mexican Secretaría de Gobernación. 32