Migrant Sanctions and Recidivism in Border Apprehensions (Appendix)

Appendix A: Figures and Tables Figure A1: Number of Border Patrol Ocers along Southwestern Border and Nationwide 25000 No. of officers 20000 15000 10000 5000 0 1992 1996 2000 2004 Year 2008 Southwestern States Note : Data from the U.S. Customs and Border Protection, 2012 2016 Nationwide CBP Border Security Report, FY2017. Figure A2: Apprehensions by the U.S. Border Patrol at the Southwestern Border 1200 Apprehensions (000s) 1000 800 600 400 200 0 2004 2006 2008 2010 2012 2014 Total apprehensions Apprehensions of Mexican nationals Note : Data from the U.S. Department of Homeland Security, various years. Yearbook of Immigration Statistics, 80 60 40 20 0 Intending to cross the border (%) 100 Figure A3: Apprehended Migrants Intending to Cross the Border within next 3 months Note : 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Data from EMIF-Norte Surveys (Surveys of Migration in the Northern Border of Mexico) 2005 to 2015 and Roberts (2017). Table A1: Summary Statistics Characteristics of migrant Fraction Age 16-17 18-20 21-24 25-28 29-33 34-40 41-50 0.049 0.159 0.189 0.175 0.178 0.163 0.087 Birth region in Mexico Border North Center North Center Center South South 0.115 0.125 0.180 0.198 0.314 0.067 Number of prior apprehensions 1 2 3 4 5 6 0.458 0.254 0.140 0.077 0.044 0.027 Re-apprehended Within 3 mos. Within 6 mos. Within 12 mos. Within 18 mos. 0.206 0.226 0.250 0.264 Administrative consequences Removal order Reinstatement order Total 0.571 0.429 344,974 Programmatic consequences ATEP MIRP Total 0.862 0.138 189,532 Criminal consequences Streamline Standard Prosecution Total 0.835 0.165 85,683 Location, time of apprehension Fraction Border Patrol San Diego sector El Centro Yuma Tucson El Paso Big Bend Del Rio Laredo Rio Grande Valley 0.173 0.059 0.013 0.531 0.035 0.007 0.039 0.043 0.104 Fiscal year 2008 2009 2010 2011 2012 0.282 0.233 0.198 0.145 0.143 Month January February March April May June July August September October November December 0.083 0.102 0.143 0.130 0.100 0.077 0.064 0.064 0.058 0.075 0.059 0.045 Day of week Sunday Monday Tuesday Wednesday Thursday Friday Saturday 12am-7am 7am-12pm 12pm-6pm 6pm-12am 0.134 0.142 0.149 0.149 0.149 0.142 0.135 0.258 0.222 0.297 0.223 Time of day Note: ThisThis tabletable provides summary statistics onon ourour sample of of apprehensions of of male Mexican Notes: provides summary statistics sample apprehensions male Mexicannationals, nationals,ages 16 toages 50, 16 with six of fewer previous apprehensions, where the apprehension in question occurred between ports to 50, with six of fewer previous apprehensions, where the apprehension in question occurred of entry along the Southwester border between 2008 and 2012. The re-apprehension statistics are cumulative between ports of entry along the Southwester border between 2008 and 2012. The re-apprehension statistics rather mutually exclusive. For those apprehended in 2005, we track whether they had been apprehended arethan cumulative rather than mutually exclusive. during the 18 months back into 2003; for those apprehended in 2012, we track whether they were apprehended in the 18 months out into 2014. The full data cover 2,824,776 apprehensions of Mexican nationals between 2005 and 2012. Restricting the sample to men drops 437,618 apprehensions of women, to ages 16 to 50 drops 71,519 apprehensions of younger and older males, and to those with fewer than seven previous apprehensions drops another 102,704 apprehensions. The nal sample contains 973,171 apprehensions. Table A2: Details on CDS Rollout Consequence Type 2008 2009 2010 2011 2012 2008-12 Administrative 0.154 0.261 0.330 0.550 0.739 0.354 Programmatic 0.148 0.167 0.393 0.492 0.195 Criminal 0.083 0.086 0.136 0.226 0.088 Programmatic or Criminal 0.229 0.249 0.510 0.680 0.273 Administrative & Programmatic 0.004 0.043 0.242 0.385 0.100 Administrative & Criminal 0.072 0.081 0.132 0.218 0.083 Administrative & Programmatic/Criminal 0.076 0.120 0.355 0.567 0.174 0.414 0.458 0.705 0.852 0.454 Any 0.154 Notes: Fraction of sample apprehended migrants (male Mexican nationals, ages 16-50, with 6 or fewer prior apprehensions) subject to given consequence programs during rollout period for CDS. Table A3: Impact of Administrative Consequences on Probability of Re-Apprehension (1) (2) (3) (4) (5) (6) Panel A: Pr(Re-Apprehension within 3 months) Administrative Consequences -0.064*** -0.063*** -0.065*** -0.063*** -0.064*** -0.066*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) [0.010] [0.014] [0.022] [0.015] [0.011] [0.005] Oster |δ| Statistic Relative to Column … 972,754 0.074 0.061 83.7 ∞ 482 2 2 5 713,528 0.326 0.079 512,727 0.401 0.099 Number of Observations R-squared Adjusted R-squared 973,171 0.060 0.059 972,721 0.076 0.061 495,668 0.409 0.103 Administrative Consequences -0.055*** -0.054*** -0.055*** -0.054*** -0.055*** -0.058*** (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) [0.019] [0.020] [0.048] [0.036] [0.020] [0.013] Panel B: Pr(Re-Apprehension within 6 months) Oster |δ| Statistic Relative to Column … 972,754 0.068 0.054 144.4 ∞ 2 2 713,528 0.320 0.072 512,727 0.396 0.092 308 5 Number of Observations R-squared Adjusted R-squared 973,171 0.053 0.053 972,721 0.070 0.055 495,668 0.405 0.096 Administrative Consequences -0.047*** -0.046*** -0.047*** -0.047*** -0.047*** -0.050*** (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) [0.025] [0.033] [0.083] [0.063] [0.032] [0.046] Panel C: Pr(Re-Apprehension within 12 months) Oster |δ| Statistic Relative to Column … 972,754 0.062 0.048 125.8 2 131.9 2 713,528 0.315 0.065 512,727 0.392 0.086 291.7 5 Number of Observations R-squared Adjusted R-squared 973,171 0.047 0.047 972,721 0.064 0.049 495,668 0.400 0.089 Administrative Consequences -0.042*** -0.041*** -0.042*** -0.043*** -0.042*** -0.046*** (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) [0.035] [0.042] [0.098] [0.068] [0.039] [0.054] Panel D: Pr(Re-Apprehension within 18 months) Oster |δ| Statistic Relative to Column … Number of Observations R-squared Adjusted R-squared Interactive Fixed Effects Sector x Fiscal Year x Month … x Day of Week x Time of Day … x Age Category x Birth State … x Number of Prior Apprehensions 113.2 2 60.7 2 973,171 0.046 0.045 972,754 0.060 0.047 713,528 0.313 0.063 512,727 0.391 0.084 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Sector x Calendar Date … x Age Category x Birth State x Prior Apprehensions 208.4 5 972,721 0.062 0.047 495,668 0.399 0.086 ✓ ✓ ✓ This table reports estimates of equation (1) for the eect of administrative consequences on the probability of re-apprehension within 3, 6, 12, and 18 months after the initial apprehension. Coecients and standard errors are those shown in Figure 5. Sample sizes decline with the inclusion of additional interactive xed eects because we omit the singleton cells for which there is one observation. Standard errors (clustered by sector-year-month) are in parentheses. Stars indicate signicance at 10% *, 5% **, and 1% *** level for those standard errors. Within brackets, we report the p-values based on a wild bootstrap procedure clustering at the sector level (of which there are 9). Notes: Notes: This table reports estimates of equation (1) for the effect of administrative consequences on the probability of re-apprehension within 3, 6, 12, or 18 months after the initial apprehension. Coefficients and standard errors are shown in Figure 6. Sample sizes decline with the inclusion of additional interactive fixed effects because we omit the singleton cells for which there is a single observation. Standard errors (clustered by sector-year-month) are in parentheses. Stars indicate significance at 10% *, 5% **, and 1% *** level for those standard errors. Within brackets, we report the p-values based on a wild bootstrap procedure clustering at the sector level (of which there are 9). Table A4: Impact of Any CDS Sanction on Probability of Re-Apprehension (1) (2) (3) (4) (5) (6) Panel A: Pr(Re-Apprehension within 3 months) Any Consequences -0.071*** -0.071*** -0.080*** -0.079*** -0.071*** -0.081*** (0.004) (0.003) (0.004) (0.004) (0.004) (0.004) [0.009] [0.028] [0.028] [0.009] [0.034] [0.009] Oster |δ| Statistic Relative to Column … 972,754 0.075 0.062 22.8 2 26.8 2 713,528 0.327 0.081 512,727 0.402 0.101 116.8 5 Number of Observations R-squared Adjusted R-squared 973,171 0.061 0.060 972,721 0.077 0.062 495,668 0.410 0.104 Any Consequences -0.064*** -0.064*** -0.072*** -0.071*** -0.064*** -0.074*** (0.003) (0.003) (0.004) (0.004) (0.003) (0.004) [0.006] [0.023] [0.024] [0.007] [0.034] [0.008] Panel B: Pr(Re-Apprehension within 6 months) Oster |δ| Statistic Relative to Column … Number of Observations R-squared Adjusted R-squared 973,171 0.054 0.054 972,754 0.069 0.055 23.5 2 2.79 2 713,528 0.321 0.074 512,727 0.397 0.094 118.4 5 972,721 0.070 0.056 495,668 0.406 0.098 Panel C: Pr(Re-Apprehension within 12 months) Any Consequences -0.056*** -0.056*** -0.064*** -0.064*** -0.055*** -0.066*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) [0.005] [0.005] [0.014] [0.014] [0.004] [0.032] Oster |δ| Statistic Relative to Column … Number of Observations R-squared Adjusted R-squared 973,171 0.048 0.048 972,754 0.063 0.049 21.4 2 22.4 2 713,528 0.316 0.066 512,727 0.393 0.088 103.4 5 972,721 0.065 0.050 495,668 0.401 0.090 Panel D: Pr(Re-Apprehension within 18 months) Any Consequences -0.052*** -0.051*** -0.059*** -0.059*** -0.051*** -0.061*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) [0.005] [0.004] [0.012] [0.019] [0.004] [0.028] Oster |δ| Statistic Relative to Column … Number of Observations R-squared Adjusted R-squared Interactive Fixed Effects Sector x Fiscal Year x Month … x Day of Week x Time of Day … x Age Category x Birth State … x Number of Prior Apprehensions 19.8 2 20.7 2 973,171 0.046 0.046 972,754 0.061 0.047 713,528 0.314 0.064 512,727 0.391 0.085 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Sector x Calendar Date … x Age Category x Birth State x Prior Apprehensions 110.5 5 972,721 0.062 0.048 495,668 0.399 0.087 ✓ ✓ ✓ This table replaces administrative consequences with any consequences (administrative, programmatic, and (or) or criminal) and re-estimates the specications in Table A3. Coecients and standard errors are those shown in Figure 5. Standard errors (clustered by sector-year-month) are in parentheses. Stars indicate signicance at 10% *, 5% **, and 1% *** level for those standard errors. Within brackets, we report the p-values based on a wild bootstrap procedure clustering at the sector level (of which there are 9). Notes: Notes: This table replace administrative with any consequences (administrative, programmatic, or criminal) and re-estimates the same set of specifications as in Table A.XX. The coefficients and standard errors are the input to Figure XX. Stars indicate significance at 10% *, 5% **, and 1% *** level for those standard errors. Within brackets, we report the p-values based on a wild bootstrap procedure clustering at the sector level (of which there are 9). Table A5: Comparing Administrative and Other Consequences, 3-Month Horizon Dep. Var.: Pr(Re-Apprehension within 3 months) (2) (3) (4) (5) (6) (1) Panel (A) Admin. Conseq. (AC) -0.060*** -0.060*** -0.065*** -0.060*** -0.063*** -0.066*** (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) Program or Crim. Conseq. (PC/CC) -0.043*** -0.042*** -0.059*** -0.059*** -0.042*** -0.060*** (0.006) (0.006) (0.007) (0.007) (0.006) (0.006) Number of Observations Adjusted R-squared 973,171 0.060 972,754 0.062 713,528 0.081 512,727 0.101 972,721 0.062 495,668 0.105 Panel (B) Admin. Conseq. (AC) -0.070*** -0.070*** -0.075*** -0.075*** -0.070*** -0.076*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) Program or Crim. Conseq. (PC/CC) -0.059*** -0.060*** -0.076*** -0.077*** -0.058*** -0.077*** (0.006) (0.006) (0.007) (0.007) (0.006) (0.006) AC x PC/CC 0.034*** (0.005) 0.036*** (0.005) 0.039*** (0.005) 0.043*** (0.006) 0.033*** (0.005) 0.039*** (0.006) Number of Observations Adjusted R-squared 973,171 0.061 972,754 0.062 713,528 0.081 512,727 0.101 972,721 0.063 495,668 0.105 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Interactive Fixed Effects Sector x Fiscal Year x Month ✓ … x Day of Week x Time of Day … x Age Category x Birth State … x Number of Prior Apprehensions Sector x Calendar Date … x Age Category x Birth State x Prior Apprehensions This table reports estimates of equation (1) for the probability of re-apprehension within 3 months after the This tableallowing reportsadministrative estimates of equation (1) for the probability of re-apprehension within 3 on initialNotes: apprehension, and programmatic/criminal consequences to have dierent eects recidivism in after apprehensions. A enters the two consequences separately; panel(programmatic/criminal) B allows for their interaction. months the initial Panel apprehension, allowing administrative and other Standard errors are clustered by sector-year-month. Stars indicate signicance at 10% *, 5% **, and 1% *** level. consequences to have different effects on recidivism. Panel A enters the two consequence Notes: treatments separately, and Panel B allows for their interaction. Standard errors are clustered at the sector-year-month level. Stars indicate significance at 10% *, 5% **, and 1% *** level. Table A6: Comparing Administrative and Other Consequences, 18-Month Horizon Dep. Var.: Pr(Re-Apprehension within 18 months) (1) (2) (3) (4) (5) (6) Admin. Conseq. (AC) Panel (A) -0.039*** -0.039*** -0.042*** -0.043*** -0.039*** -0.046*** (0.003) (0.003) (0.003) (0.004) (0.003) (0.004) Program or Crim. Conseq. (PC/CC) -0.037*** -0.036*** -0.049*** -0.048*** -0.036*** -0.050*** (0.005) (0.005) (0.006) (0.006) (0.005) (0.006) Number of Observations Adjusted R-squared 973,171 0.046 Admin. Conseq. (AC) Panel (B) -0.047*** -0.047*** -0.052*** -0.053*** -0.047*** -0.055*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) Program or Crim. Conseq. (PC/CC) -0.050*** -0.050*** -0.065*** -0.065*** -0.049*** -0.065*** (0.006) (0.006) (0.006) (0.006) (0.006) (0.006) AC x PC/CC 0.028*** (0.005) 0.030*** (0.005) 0.036*** (0.006) 0.038*** (0.006) 0.027*** (0.005) 0.033*** (0.006) Number of Observations Adjusted R-squared 973,171 0.046 972,754 0.048 713,528 0.064 512,727 0.085 972,721 0.048 495,668 0.088 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Interactive Fixed Effects Sector x Fiscal Year x Month ✓ … x Day of Week x Time of Day … x Age Category x Birth State … x Number of Prior Apprehensions 972,754 0.047 Sector x Calendar Date … x Age Category x Birth State x Prior Apprehensions 713,528 0.064 512,727 0.085 972,721 0.048 495,668 0.088 This table reports estimates of equation (1) for the probability of re-apprehension within 18 months after the initial apprehension, allowing administrative and programmatic/criminal consequences to have dierent eects on recidivism in apprehensions. Panel A enters the two consequences separately; panel B allows for their interaction. Standard errors are clustered by sector-year-month. Stars indicate signicance at 10% *, 5% **, and 1% *** level. Notes: Table A7: Heterogeneous Impacts of Consequence Programs by No. of Previous Apprehensions (A) Administrative Consequences Pr(Re-apprehension within … months) Administrative Consequences Administrative Consequences x 2 Prior Apprehensions Administrative Consequences x 3 Prior Apprehensions Administrative Consequences x 4-6 Prior Apprehensions Number of Observations R-squared Adjusted R-squared (1) (2) (3) (4) 3 6 12 18 -0.058*** (0.003) -0.025*** (0.005) -0.029*** (0.010) -0.011 (0.017) -0.049*** (0.003) -0.024*** (0.005) -0.027*** (0.010) -0.007 (0.018) -0.043*** (0.003) -0.022*** (0.005) -0.015 (0.011) -0.000 (0.018) -0.039*** (0.003) -0.021*** (0.005) -0.015 (0.011) 0.009 (0.020) 512,727 0.401 0.099 512,727 0.396 0.092 512,727 0.392 0.086 512,727 0.391 0.084 (1) (2) (3) (4) 3 6 12 18 -0.074*** (0.004) -0.024*** (0.005) -0.019** (0.009) -0.004 (0.018) -0.067*** (0.004) -0.022*** (0.006) -0.014 (0.009) -0.003 (0.019) -0.060*** (0.004) -0.019*** (0.006) -0.002 (0.010) 0.003 (0.019) -0.056*** (0.004) -0.019*** (0.006) -0.001 (0.011) 0.011 (0.020) 512,727 0.402 0.101 512,727 0.397 0.094 512,727 0.393 0.088 512,727 0.391 0.085 ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ (B) Any Consequences Pr(Re-apprehension within … months) Any Consequences Any Consequences x 2 Prior Apprehensions Any Consequences x 3 Prior Apprehensions Any Consequences x 4-6 Prior Apprehensions Number of Observations R-squared Adjusted R-squared Interactive Fixed Effects Sector x Fiscal Year x Month … x Day of Week x Time of Day … x Age Category x Birth State … x Number of Prior Apprehensions This table reports estimates of the regressions in column 4 of Table A3, shown in panel A, and in column 4 of Table A4, shown in panel B, in which we allow the impact of consequence programs on the probability of reapprehension to vary with the number of previous apprehensions for an individual. Standard errors are clustered by sector-year-month. Stars indicate signicance at 10% *, 5% **, and 1% *** level. Notes: Appendix B: Estimating Capacity Constraints in Figure 3 In Section 2.3, we discuss the results in Figure 3 demonstrating the stang constraints in moving from voluntary return to administrative consequences under the CDS. For each Border Patrol sector s, we compute the share of ocer time that would be absorbed by applying AC to all apprehended migrants on a given day d from 2008 to 2012 based on the equation:  agent timesd (1.5 − 0.25) × (0.92 × apprehensionssd ) = 100 × (agentssd − (0.8 × agentss,2007 )) × 8 × 0.51 × 0.916  (4) where (1.5-0.25) captures the increase in agent man-hours to go from processing one VR to processing one AC; 0.92 is the share of apprehensions that were not already subject to AC as of 2008 (i.e, 92% of migrants received VR in 2007); (agentssd − (0.8 × agentss,2007 )) is agent time available after subtracting the fraction needed for essential operations (e.g., patrolling the border, making apprehensions) which is set to 80% of the level of 2007 agent activities; 8 is the number of potential hours available per agent per day; 0.51 is the fraction of each hour that agents work in operations after accounting for reported time not on duty, on breaks, in training, or performing administrative tasks; and 0.916 is the fraction of operations time not spent on trac checkpoints (which occur relatively far from the border itself, impeding agents who man check points from performing other duties). These parameter values in equation (4) are based on an in-depth analysis by the U.S. General Accounting Oce of time use by Border Patrol agents along the U.S.-Mexico border in the early 2010s (GAO, 2017b). Figure 3 plots the resulting variation using sectorday observations on the total number of apprehensions, sectoryear observations on the total number of agents. Note that the number of apprehensions used in equation (4) is based on our sample and hence likely understates demands on agent time, as it excludes minors, serious criminals, and non-Mexican nationals, which account account for 15% of total apprehensions over 2008-2012.