Crime and the Legalization of Recreational Marijuana Davide Dragone Paolo Vanin Giovanni Prarolo Giulio Zanella University of Bologna IZA Discussion Paper N. 10522, 2017 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 1 / 32
Research question and contribution Research question Does recreational marijuana increase crime? To answer Quasi-experimental evidence from staggered legalization in US Washington (2012) and Oregon (2014) Difference in discontinuity design to identify causal effects We find No significant increase in crime Significant reduction in rapes (30%) and thefts (20%) Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 2 / 32
Hot topic The prohibitionist propaganda links cannabis to violence and crime: Film poster 1937 Film poster 1942 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 3 / 32
Hot topic The prohibitionist propaganda links cannabis to violence and crime: Film poster 1936 Film poster 1936 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 4 / 32
Hot topic The prohibitionist propaganda links cannabis to violence and crime: Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 5 / 32
Hot topic According to anti-prohibitionists, such as Becker and Murphy (2013): De-criminalizing and legalizing reduces the costs of the crime associated with drug trafficking Just as gangsters were largely driven out of the alcohol market after the end of prohibition...... violent drug gangs would be driven out of a decriminalized drug market. Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 6 / 32
Literature Positive association between violence and drug use (Goldstein 1985 J Drug Issues): causal? Evidence on medical marijuana laws and crime in the US: No association (Keppler et al. 2012 J Stud Alcohol and Drugs; Shepard, Blackley 2016 J Drug Issues; Morris et al. 2014 PloS One; Braakman, Jones 2014 Soc Sci & Med; Freisthler et al. 2016 Addiction) Less homicides (Ingino 2016 DP) and violent crime in general (homicides, assaults, robberies) due to reduced activity by drug trafficking organizations (Gavrilova et al. 2014 DP) Depenalization of possession in London reduced crime due to reallocation of police effort (Adda et al 2014 JPE) No causal evidence on recreational marijuana and crime Partial exception: more possession arrests in neighboring counties (Hao, Cowan 2017 NBER WP) Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 7 / 32
Legalization in the US Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 8 / 32
Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 9 / 32
Bird s-eye view Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 10 / 32
A quasi-experiment WA and OR: neighboring and similar states, with staggered legalization Washington legalized in 2012 Initiative 502: approved Nov 2012, enforced since Dec 2012 Legal possession since December 2012 Regulations for producers, processors and sellers approved in 2013 Retail sales of recreational cannabis since July 2014 Details Allows producing, processing, and selling cannabis, subject to licensing and regulation by the Liquor Control Board Allows limited possession by persons aged 21 and over (but not home cultivation) Taxes sales Oregon legalized in 2014 Measure 80: rejected Nov 2012 Measure 91: approved Nov 2014, enforced since July 2015 Similar to WA, more permissive for home cultivation Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 11 / 32
Data 1. US Uniform Crime Report County-level violent and property crime counts, 2010-2014 Reported by sheriff or police department counties 25K inhab. cities 10K inhab. Unbalanced panel of 335 observations 75 counties: 36 in OR and 39 in WA 2. 2010 US Census County-level crime rates per 100K inhab. 3. National Survey on Drug Use and Health Averages over 2010-2012, 2012-2014 Substate regions Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 12 / 32
Crime rates in all counties Year Murder Rape Assault Robbery Burglary Theft All WA counties (N = 39) 2010 0.76 10.96 46.66 12.17 265.79 458.97 2011 0.85 9.65 40.84 10.30 265.08 440.87 2012 1.03 9.16 42.70 9.99 287.77 432.55 2013 0.80 9.07 41.23 9.21 258.73 419.59 2014 0.73 9.70 41.21 10.47 246.90 399.60 All OR counties (N = 36) 2010 0.80 7.22 34.31 6.82 132.96 393.71 2011 0.66 7.26 32.02 6.26 142.14 387.37 2012 0.84 7.51 29.31 6.75 150.93 412.93 2013 0.88 5.69 22.48 5.40 146.14 433.22 2014 0.66 7.22 30.21 4.72 115.17 335.12 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 13 / 32
Crime rates in border counties Year Murder Rape Assault Robbery Burglary Theft Border WA counties (N = 11) 2010 0.35 15.37 33.69 8.51 224.00 529.80 2011 0.48 13.56 33.55 9.69 212.19 491.00 2012 0.75 12.80 42.00 7.58 223.30 445.11 2013 0.59 10.28 40.78 6.15 210.41 407.93 2014 0.71 10.52 39.48 6.97 184.76 357.10 Border OR counties (N = 10) 2010 0.34 1.58 13.40 3.04 41.88 163.57 2011 0.44 2.51 11.22 1.31 49.15 158.78 2012 0.31 2.59 10.76 1.14 56.88 176.11 2013 0.10 1.77 11.67 1.67 41.04 144.27 2014 0.11 0.91 14.89 2.39 40.91 128.08 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 14 / 32
Substance consumption rates in all counties Year Marijuana Other drugs Alcohol Binge alcohol All WA counties (N = 39) 2010-2012 0.102 0.044 0.560 0.222 2012-2014 0.127 0.039 0.542 0.206 All OR counties with consumption data (N = 34) 2010-2012 0.112 0.042 0.596 0.214 2012-2014 0.122 0.040 0.579 0.213 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 15 / 32
Substance consumption rates in border counties Year Marijuana Other drugs Alcohol Binge alcohol Border WA counties (N = 11) 2010-2012 0.093 0.042 0.535 0.223 2012-2014 0.101 0.034 0.486 0.199 Border OR counties (N = 10) 2010-2012 0.145 0.050 0.630 0.238 2012-2014 0.130 0.043 0.600 0.233 Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 16 / 32
Empirical strategy Difference in discontinuity Combine diff-in-diff (DID) and spatial regression discontinuity (SRD) Counties at the border offer a better comparison Max similarity between the two states Non-parametric diff-in-disco Plot pre-post variation in crime rates as a function of distance from OR-WA border (in 100 Km) Observe the jump at the border Parametric diff-in-disco estimates Control for distance from the border Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 17 / 32
Pre-post variation in crime and border distance -1 0 1 2 murders -4-3 -2-1 0 1 2 3 4-6 -4-2 0 2 rapes -4-3 -2-1 0 1 2 3 4-20 -10 0 10 20 assaults -4-3 -2-1 0 1 2 3 4-3 -2-1 0 1 2 robberies -4-3 -2-1 0 1 2 3 4-40 -20 0 20 40 60 burglaries -4-3 -2-1 0 1 2 3 4-200 -100 0 100 200 thefts -4-3 -2-1 0 1 2 3 4 Note: distance > 0 in WA; < 0 in OR. Smoothed county-level differences from local linear regressions, weighted by county population, triangular kernel, bandwidth of 100 Km Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 18 / 32
Parametric DID-SRD estimates Difference in dicontinuity (DID-SRD) c it = k + αp t + βw i p t + f (d i )p t + g(d i )w i p t + θ i + ξ it (1) c it is the crime rate in county i and year t p t = 1 if t > 2012 (post), p t = 1 if t 2012 (pre) w i = 1 if i WA (treatment), w i = 0 if i OR (control) f (.) and g(.) are polynomials of the same order in distance from the border θ i are county fixed effects k is a constant and ξ it an error term β is the DID-SRD effect of the policy By how much liberalizing recreational cannabis in WA changed the difference in crime rates right across the WA-OR border Estimates: OLS; quadratic polynomials in distance (Gelman, Imbens 2014 NBER WP); robust s.e., clustered at county level if larger; counties weighted by pop. Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 19 / 32
DID-SRD estimates of the effect of recr. cannabis Murder Rape Assault Robbery Burglary Theft Estimated β 0.23 4.21*** 1.30 1.26 36.32 105.62** (0.45) (1.26) (8.79) (1.92) (22.20) (40.21) Observations 335 335 335 335 335 335 No evidence of crime increase Negative point estimates for rape, assault, robbery, burglary and theft Positive for murders, but not significant Rapes and thefts: significant and relevant reduction Rapes: 4.2 offenses per 100K inhab., about 30% of the 2010-2012 rate Thefts: 105.6 offenses per 100K inhab., about 20% of the 2010-2012 rate Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 20 / 32
Robustness Two specific threats to identification Concomitant reforms affecting crime in WA Non-complying counties Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 21 / 32
Crime-related reforms in WA in 2013-2014 Several crime-related minor reforms in WA in 2013-2014 Reorganization of 911 Reforms related to health services, regulation of wine and beer, and drug courts Changes in the statute of limitations for child molestation, incest (victim under age eighteen), and rape (victim under age eighteen) New norms concerning commercial sale of sex and commercial sexual abuse, sexually violent predators, and sexual violence at school These changes were too marginal to exert a first-order effect on crime Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 22 / 32
Excluding non-complying counties Re-estimate the DID-SRD model after excluding 5 WA counties where cannabis business is prohibited Columbia, Franklin, Garfield, Wahkiakum, and Walla Walla County, all bordering Oregon except Franklin County Murder Rape Assault Robbery Burglary Theft Estimated β 0.20 3.77*** 0.36 1.19 41.84 117.51*** (0.49) (1.49) (9.14) (2.04) (25.40) (39.67) Observations 310 310 310 310 310 310 Results confirmed in sign and magnitude, and improved in significance Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 23 / 32
Channels? Some candidates (Goldstein, 1985) Psychotropic effect Drug wars Microcriminality But also Changes in market structure Substitutability/complementarity Reallocation of police effort Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 24 / 32
Channels? Some candidates (Goldstein, 1985) Psychotropic effect Drug wars Microcriminality But also Changes in market structure Substitutability/complementarity Reallocation of police effort Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 25 / 32
Psychotropic effect Source: Green et al (2003) Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 26 / 32
Pre-post variation in consumption and border distance -.02 0.02.04.06 marijuana consumption -4-3 -2-1 0 1 2 3 4 -.01 -.008-.006-.004-.002 0 other drugs consumption -4-3 -2-1 0 1 2 3 4 -.06 -.04 -.02 0 alcohol consumption -4-3 -2-1 0 1 2 3 4 -.03 -.02 -.01 0.01 binge alchool -4-3 -2-1 0 1 2 3 4 Note: distance > 0 in WA; < 0 in OR. Smoothed county-level differences from local linear regressions, weighted by county population, triangular kernel, bandwidth of 100 Km Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 27 / 32
Substitution effects Marijuana Other drugs Alcohol Binge alcohol Estimated β 0.025*** 0.005*** 0.023* 0.020*** (0.009) (0.001) (0.014) (0.007) [0.016] [0.002] [0.016] [0.010] Observations 135 135 135 135 Substitution of marijuana for other drugs and alcohol Marijuana: 2.5 pp increase, about 25% of the 2010-2012 consumption rate Other drugs: 0.5 pp reduction, about 10% of the 2010-2012 rate Alcohol: 2.3 pp reduction, about 5% of the 2010-2012 rate Binge alcohol: 2 pp reduction, about 10% of the 2010-2012 rate Data: National Survey on Drug Use and Health (Population percentages and substate regions) Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 28 / 32
Microcriminality and police attitude Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 29 / 32
Changes in police attitude and objectives Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 30 / 32
Conclusion First-pass evidence on legal cannabis on crime Drop in crime at the OR-WA border Possible channels Psychoactive effects: relaxation and euphoria reduced likelihood of violence Substitute for violence-inducing substances such as alcohol, cocaine and amphetamines Less role for drug-trafficking criminal organizations Reallocation of police effort Further research needed to disentangle the relative importance of different channels Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 31 / 32
Thanks! Paolo Vanin (University of Bologna) Marijuana and Crime IZA Discussion Paper N. 10522, 2017 32 / 32