Proactive Policing and Robbery Rates across Large U.S. Cities: Assessing Robustness Charis E. Kubrin George Washington University Steven F. Messner Glenn Deane Kelly McGeever State University of New York, Albany Thomas D. Stucky Indiana University-Purdue University Indianapolis
Aims of Current Study To replicate Sampson and Cohen (1988) To expand their model specification To explore the possible implications of endogeneity
Explanations for Discrepant Findings on Policing and Deterrence Police work is not devoted to crime reduction Police practices do not affect arrest certainty Displacement of offenders Methodological issues: –Limitations with arrest certainty measures – Nature of causal relationship between police strength and crime rates
Proactive Policing and Crime Indirect effect of proactive policing on crime through arrest risk –Increasing arrest/offense ratio Proactive policing may directly affect crime rate by influencing community perceptions regarding the probabilities of apprehension for illegal behavior –Public disorder
Specifying a More Complete Model Index of concentrated disadvantage –Poverty, family disruption, joblessness Role of local politics –Wilson (1968) Varieties of Police Behavior –Policing styles: watchman, legalistic, service –Elected mayors, partisan elections, district based council representation
Data and Methods Sample: U.S. cities with pop. of 100,000+ with at least 1,000 blacks in 2000 (n=181) 5 data sources: (1) counts of robberies known to police and city pop. totals; (2) yearly arrest counts for DUI and disorderly conduct; (3) police employee data; (4) demographic data from 2000 census; (5) two databases on political system characteristics of city governments
Data and Methods Contd. Dep. vble= robbery offenses known for all cities that were available in UCR for 4-yr. period: –Smoothed data Key Indep. vble= proactive policing –Sum of # arrests for DUI and disorderly conduct / # sworn police officers –Lagged measure of proactive policing using data for 4-year period ( ) immediately preceding period of interest Indep. vble= robbery arrest/offense ratio –Lagged measure
Data and Methods Contd. Controls: city pop size (logged), median family income, % divorced, % non-Hisp. Black, racial income inequality, dummy vble. for West location Model extension: –Resource deprivation: % poverty, % non-Hisp. Black, % unemployed, % high school grad, % female- headed households, median family income –Residential instability, % young males –City political system characteristics 3 elements: (1) mayor-council forms of government, (2) council members represent specific geographic areas, and (3) city elections are partisan
Table 1. Regressions of Certainty of Arrest and Robbery Rates. Certainty of Arrest (log) Robbery Rate Model I (log) Robbery Rate Model II b b b Intercept.587*-3.773*-1.732*- (log) Proactive Policing * *-.135 (log) Population-.026* * *.198 Percent Divorced *.127 Western Location * Racial Inequality * *.150 Median Income (in $1000s) a.002* * Percent Non-Hispanic Black a -.002* * Resource Deprivation Index *.690 Traditional Government Index Percent Young Males Percent Moved R-Square *Statistically Significant for a Two-Tailed Test at the.05 Level a Incorporated in the "Resource Deprivation Index" for Model II of Robbery Rates
Table 2. Non-Recursive Models of the Police Measures and Robbery Rates. (log) Robbery Rate Model 1 (log) Robbery Rate Model 2 b b Intercept * *- (log) Proactive Policing -.129* *-.098 (log) Population.201* *.137 Percent Divorced.048* West Racial Inequality.294* *.160 Resource Deprivation Index.541* *.595 City Politics Index Percent Young Males Percent Moved Certainty of Arrest *-.274 * Statistically Significant for a Two-Tailed Test at the.05 Level Model 1 = 2SLS with lagged Proactive Policing as instrument Model 2 = 2SLS with lagged Proactive Policing and lagged Certainty of Arrest as Instruments