Silence of the Innocents: Illegal Immigrants Underreporting of Crime and their Victimization

Silence of the Innocents: Illegal Immigrants Underreporting of Crime and their Victimization Stefano Comino, 1 Giovanni Mastrobuoni, 2 Antonio Nicolò 3 1 University of Udine, 2 University of Essex, 3 University of Padova International Conference on Migration and Welfare Rome, 11-12 May 2017

Introduction Large number of undocumented immigrants: in 2014, in the US 11.3 million (3.5% of the pop.), up from about 3.5 million in 1990. How to deal with this high number of illegal immigrants?

Introduction Large number of undocumented immigrants: in 2014, in the US 11.3 million (3.5% of the pop.), up from about 3.5 million in 1990. How to deal with this high number of illegal immigrants? Public opinion polls: citizens tend to fear that illegal immigration brings lost jobs and rising welfare costs but also high rates of crime The debate centers around having an amnesty as the last major one happened in 1986 (IRCA, Immigration Reform and Control Act)

Contribution We contribute to this debate providing evidence on an important negative consequence of illegally residing in the country which is directly borne by the undocumented immigrants: The inability to protect their property and their human right to security. Out of fear of deportation undocumented immigrants drastically lower their propensity to report a crime to the police This generates an essentially unenforced space of action for criminals.

Literature: Immigration and Crime Butcher and Piehl (1998) and Piehl (2007) find little evidence of a correlation between (legal and illegal) immigration rates and incarceration rates (CENSUS data). Bianchi et al. 2012 (Italy) and Bell et al. (2013) confirm such findings. Evidence that legals status matters: less crime once legalized, Mastrobuoni&Pinotti (2015) and Pinotti (2016), more jobs Kossoudji Cobb-Clarke (2002), Lozano and Sorensen (2011), Barcellos (2010).

Literature: Underreporting of Immigrants Evidence on underreporting is scarce. Relies on correlational studies that do not measure legal status (Gutierrez and Kirk, 2015) or on studies that do measure legal status but only for small convenience samples (Bucher et al., 2010) In search for such evidence we use the National Crime Victimization Survey (measures reporting to the police) around the 1986 US immigration amnesty (Immigration Reform and Control Act, IRCA)

Model A society composed of two ethnic groups natives: poor or rich immigrants: legal or undocumented (all poor)

Model A society composed of two ethnic groups natives: poor or rich immigrants: legal or undocumented (all poor) Each individual chooses be honest or become criminal a criminals chooses which ethnic group to target cannot observe the characteristics of the victim (poor/rich or legal/undocumented)... but just the average of the group choice of the target subject to some error

Model Utility if honest increasing in wealth (larger for rich individuals) increasing in the reporting rate (better able to protect herself and her property)

Model Utility if honest increasing in wealth (larger for rich individuals) increasing in the reporting rate (better able to protect herself and her property) Utility if criminal increasing in the average wealth of the target group decreasing in the average reporting rate of the target group

Equilbrium Reporting larger for richer individuals (larger amount stolen) larger for legal individuals (no risk of deportation)

Equilbrium Reporting larger for richer individuals (larger amount stolen) larger for legal individuals (no risk of deportation) Honest or Criminal & target group trade-off: larger av. wealth of victims vs lower reporting rate

Effects of an Amnesty Legalized immigrants no risk of deportation reporting rate of legalized immigrants increases average reporting rate of the immigrant group (legal and undocumented) increases reduces the utility for criminals of targeting immigrants utility of being honest for legalized immigrants increases Effect of amnesties on crime reduced crime committed against immigrants overall reduction in crime (ˆθ increases): higher utility of being honest for legalized immigrants + lower utility of targeting immigrants shifting from immigrants to native target (θ reduces) uncertain effects on crime committed against natives shifting from immigrants to native target (θ reduces) errors in selecting the target natives benefit from the overall reduction in crime

Immigration Reform and Control Act (IRCA) 1986: US Amnesty Undocumented immigrants who were in continuous residence since January 1, 1982 (5 years) were eligible for temporary legal status Approximately 1.75 million people applied for legalization through the program and about 94% of applications were approved for temporary residency. Special Agricultural Worker (SAW) program provided permanent residency to aliens who could demonstrate they had 60 days of seasonal agricultural work experience in qualifying crops from May 1985 to May 1986. Nearly 1.3 million people applied for the SAW program.

Application years 87-88; largely hispanic applicants Number of applicants 0 200 400 600 800 1000 Total number of applicants (x1000) 1980 1982 1984 1986 1988 1990 Year Non-hispanic Hispanic

Shortly-lived effects Before the IRCA: at least 3 million undocumented immigrants resided in the U.S. By 1990: estimated stock of undocumented immigrants 3.5m These changes can also be seen in terms of deportations... Deportations of Illegal Immigrants 800000 1300000 1800000 Deportations of Illegal Immigrants 1978 1980 1982 1984 1986 1988 1990 1992 Year

Other data sources National Crime Victimization Survey (NCVS) information on victimization and reporting behavior in 41 Metropolitan Statistical Areas (MSAs) no information on legal status (undocumented/legal) of victims Hispanic vs. non-hispanic Census: year 1980 and year 1990... information on legal status and whether the individual resided in the U.S. 5 years in advance eligibility criteria allow us: estimate number of eligible individuals across MSAs allow us: estimate misclassification probabilities (see later)

IRCA Applicants against their Predicted Number According to the 1980 Census (similar with 1990) CA Log Number of Eligible Hispanics (CENSUS) 4 6 8 10 12 14 VT NDSD NJ MD MA VA NMWA NV CT CO LA MIPA GA OR RIIN DC OHWI KS OK UT ID NC MO HI IAMN NESC KY AK WY TNAR NH AL MS WV DE MT ME AZ FL NY IL TX 4 6 8 10 12 14 Log Number of Hispanic Applicants

Fraction of Hispanic Population by MSA (Illegal and Total) Fraction that would be eligible for the IRCA amnesty Predicted Fraction of Census Year 1980 Census Year 1990 MSA Eligible and Hispanic (NCVS) Non-Hispanics Hispanics Non-Hispanics Hispanics 1 Anaheim-Santa Ana, CA 2.25% 2.58% 21.35% 5.68% 29.81% 2 Atlanta, GA 0.09% 0.63% 15.06% 1.40% 19.35% 3 Baltimore, MD 0.02% 0.88% 8.71% 1.03% 10.04% 4 Boston, MA-NH 0.18% 2.48% 16.44% 3.34% 18.93% 5 Charlotte-Gastonia-Rock Hill, NC-SC 0.15% 0.50% 12.71% 0.78% 14.79% 6 Chicago, IL 0.98% 2.05% 22.48% 2.40% 26.55% 7 Cincinnati, OH-KY-IN 0.03% 0.45% 7.21% 0.57% 10.53% 8 Cleveland, Lorain, Elyria, OH 0.04% 1.48% 5.73% 1.32% 6.11% 9 Columbus, OH 0.00% 0.57% 3.83% 0.48% 1.85% 10 Dallas, TX 0.52% 0.62% 11.65% 1.78% 24.62% 11 Denver, CO 0.17% 0.94% 3.15% 1.38% 7.03% 12 Detroit, MI 0.12% 1.81% 10.61% 1.87% 7.62% 13 Fort Lauderdale, FL 1.00% 1.72% 18.96% 4.47% 23.78% 14 Fort Worth-Arlington, TX 1.54% 1.51% 17.90% 15 Houston, TX 1.26% 1.07% 13.22% 2.43% 26.18% 16 Kansas City, MO-KS 0.04% 0.35% 6.20% 0.49% 9.06% 17 Los Angeles-Long Beach, CA 5.54% 3.62% 28.75% 7.93% 35.50% 18 Miami, FL 13.46% 2.87% 42.36% 8.27% 37.44% 19 Minneapolis-St. Paul, MN-WI 0.04% 0.61% 6.80% 1.23% 9.58% 20 Nassau-Suffolk, NY 0.33% 1.87% 14.16% 2.59% 18.10% 21 New York, NY 2.05% 5.25% 18.06% 6.67% 20.95% 23 Newark, NJ 0.74% 2.69% 22.24% 4.02% 20.64% 24 Norfolk-Virginia Beach-Newport News, VA 0.09% 0.76% 5.57% 1.02% 7.59% 25 Oakland, CA 0.60% 1.71% 6.40% 1.46% 6.87% 26 Orlando, FL 0.31% 0.88% 16.80% 2.01% 12.47% 27 Philadelphia, PA-NJ 0.06% 1.09% 5.40% 1.33% 5.84% 28 Phoenix-Mesa, AZ 0.68% 0.93% 8.28% 1.20% 13.51% 29 Pittsburgh, PA 0.00% 0.57% 6.00% 0.55% 6.84% 30 Portland-Vancouver, OR-WA 0.10% 1.07% 8.90% 1.76% 14.62% 31 Riverside-San Bernardino, CA 1.49% 1.56% 11.24% 2.89% 22.81% 32 Sacramento, CA 0.66% 1.37% 11.06% 3.24% 13.96% 33 St. Louis, MO-IL 0.02% 0.39% 5.64% 0.44% 7.43% 34 San Antonio, TX 2.31% 0.62% 6.57% 1.14% 7.54% 35 San Diego, CA 2.29% 1.78% 20.73% 3.36% 26.89% 36 San Francisco, CA 1.50% 3.44% 17.15% 5.69% 24.69% 37 San Jose, CA 1.25% 3.28% 13.27% 8.06% 20.00% 38 Seattle-Bellevue-Everett, WA 0.11% 1.87% 5.70% 2.73% 10.46% 39 Tampa-St. Petersburg-Clearwater, FL 0.61% 1.07% 13.78% 1.66% 15.18% 40 Washington, DC-MD-VA-WV 0.49% 2.00% 24.58% 3.48% 29.46% 41 West Palm Beach-Boca Raton, FL 0.67% 1.46% 22.92% 3.01% 25.02% Total 3.68% 2.06% 20.31% 3.27% 25.77%

Reporting and Victimization DiD Reporting Model predicts no effects for non-hispanics (use as control group). Do (undocumented) Hispanics show an increase in reporting rates right after the IRCA compared to non-hispanics? misclassification issue Victimization Both Hispanics and non-hispanics are affected. Exploit comparative statics results: do Hispanics in places where there is a large fraction of IRCA eligible Hispanics show larger reductions in victimization?

Difference-in-differences in reporting rates between Hispanics and non-hispanics in MSAs with the top third of undocumented Hispanics -.15 -.1 -.05 0.05.1.15 Reporting Rates in MSAs with many Illegal Hispanics Difference between Hispanics and Non-Hisp. 1981 1984 1987 1990 1993 year Reporting effects 5/95 CI

Interpretation of the Effects Reporting rates for Hispanics and non-hispanics differ by about 5.5 percentage points (33.2 versus 38.7 percent) Assuming legal Hispanics have the same reporting rate as natives, one can use γ = 0.20 to compute the unobserved reporting rate of illegal Hispanics. The overall Hispanic reporting rate 0.332 = 0.20R illegals + 0.80 0.387.

Reporting Regressions R it = β P OST H i P OST t + β P RE H i P RE t + β H H i + ξ X i + ɛ i. H i = 1 if hispanic; P RE=1981-1986; P OST =1989-1994 X i contains year and MSA fixed effects, and in some specifications crime-type fixed effects, as well as MSA specific time trends.

Reporting Regressions (5) (6) MSAs with many illegal Hisp. Pre and Hispanic -0.035* -0.043** (0.019) (0.017) Post and Hispanic -0.044** -0.036** (0.019) (0.017) Hispanic -0.015-0.007 (0.017) (0.015) MSA fixed effects Year fixed effects Socioeconomic characteristics Crime-type fixed effects MSA specific time trends Observations 44149 44149 R-squared 0.007 0.121 Mean dep. var 0.383 0.383 Illegal hispanics 0.163 0.163 Illegal non-hispanics 0.0176 0.0176

Misclassification Aim: legalized individuals change their reporting behavior? Just observe the reporting behavior of H=1 (and H=0) before/during/after IRCA

Misclassification Aim: legalized individuals change their reporting behavior? Just observe the reporting behavior of H=1 (and H=0) before/during/after IRCA Misclassification not all hispanics were eligible (some were already legal or were not fulfilling the eligibility criteria): P (A = 1 H = 1) 0.2 = γ some non-hispanics were eligible: P (A = 1 H = 0) 0.02 = 1 q... formally the misclassification...

Inflated Effects E (R H = 1, T = t) = γe (R A = 1, T = t) + (1 γ) E (R A = 0, T = t) E (R H = 0, T = t) = qe (R A = 0, T = t) + (1 q) E (R A = 1, T = t). Taking a difference, and rearranging we get that E (R A = 1, t) E (R A = 0, t) = E (R H = 1, t) E (R H = 0, t) γ (1 q) Taking a difference of this difference across different T s we get that the difference-in-difference needs to be inflated by 1 γ (1 q). E(R A = 1) E(R A = 0) = E(R H = 1) E(R H = 0). γ (1 q) The DiD is approx. 4%, inflated by 5 it becomes 20%, meaning that the amnesty eliminates 3/4th of the reporting gap

Differences in differences in victimization rates of Hispanics in MSAs with the top third and bottom two-thirds fraction of undocumented Hispanics -.08 -.06 -.04 -.02 0.02.04.06.08 Hispanic Victimization Rates Difference between MSAs w/ Large and Small Fraction of Illegal Immigrants 1981 1984 1987 1990 1993 year Victimization effects 5/95 CI

Differences in differences in victimization rates of non-hispanics in MSAs with the top third and bottom two-thirds fraction of undocumented Hispanics -.08 -.06 -.04 -.02 0.02.04.06.08 Non-Hisp. Victimization Rates Difference between MSAs w/ Large and Small Fraction of Illegal Immigrants 1981 1984 1987 1990 1993 year Victimization effects 5/95 CI

Victimization Regressions V it = β P OST T OP (γ) i P OST t + β P RE T OP (γ) i P RE t + ξ X i + ɛ i T OP (γ) i = 1 if individual i resides in a MSA with a large fraction of IRCA eligible undocumented immigrants of H origin over the entire population (top tercile) separate regressions for Hispanics and for non-hispanics

Victimization Regressions (1) (2) (3) (4) (5) (6) Dependent Variable: Has the Respondent been victimized? (0/1) Hispanics Non-Hispanics Pre 1987, MSAs w/ large fraction of ill. Hisp. 0.029*** 0.029*** 0.013-0.001-0.001 0.000 (0.010) (0.010) (0.014) (0.004) (0.004) (0.006) Post 1987, MSAs w/ large fraction of ill. Hisp. 0.031*** 0.029*** 0.049*** 0.010** 0.009** 0.007 (0.010) (0.010) (0.016) (0.004) (0.004) (0.006) MSA fixed effects Year fixed effects Socioeconomic characteristics MSA specific time trends Observations 92182 92182 92182 822077 822077 822077 Mean dep. var 0.111 0.111 0.111 0.109 0.109 0.109 R-squared 0.007 0.017 0.018 0.008 0.030 0.030

Victimization for different T-C splits based on γs Pre/Post-Amnesty Difference in Victimization -.02 0.02.04.06 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Number of MSAs in the Treatment Group Notes: Initially the control cities are the bottom 2/3 or 28 in terms of γ. MSAs are added to the treatment group starting from top in terms of γ. Once the 13th MSA is added to the treatment group previous control MSAs switch to the treatment MSAs.

Conclusions We provide evidence of undocumented immigrants underreporting (11 percent against 38 percent). 11.3 million undocumented immigrants are unable to safeguard their fundamental right to protect their property and their human right to security! Effect of legalization substantially increases the reporting rate of legalized immigrants accounting for misclassification issues increases by 20 percentage points getting close to that of legal citizens affects the likelihood of becoming target of criminals