FINAL REPORT OF THE URBANA TRAFFIC STOP DATA TASK FORCE VOLUME II: STATISTICAL APPENDIX. October 31, 2015

Transcription

1 FINAL REPORT OF THE URBANA TRAFFIC STOP DATA TK FORCE VOLUME II: STATISTICAL APPENDIX October 31, 215

2 Recommended citation: City of Urbana Traffic Stop Data Task Force Final Report of the Urbana Traffic Stop Data Task Force, Volume II: Statistical Appendix. Urbana, Illinois: Mayor s Office. Published 215

3 TABLE OF CONTENTS Contents INTRODUCTION... 1 MEMBERS OF THE TK FORCE... 2 CHAIR... 2 MEMBERS... 2 ACKNOWLEDGMENTS... 3 VOLUME II: STATISTICAL APPENDIX... 4 VOLUME II: STATISTICAL APPENDIX... 4 VOLUME I: MAIN REPORT... 4 CONTACT INFORMATION... 78

4 INTRODUCTION Introduction In January 214, the Urbana City Council established a Traffic Stop Data Task Force to examine data regarding racial disparities in traffic stops by the Urbana Police Department. The data we were tasked with examining was collected by the Police Department, in part to provide to the Illinois Department of Transportation for their study of traffic stops. In June 214, the Task Force met to begin its work. The Task Force divided its work into four major areas of study: A survey of wider literature regarding traffic stops and racial disparities An analysis of the collected statistics regarding traffic stops in order to look for racial disparities and possible causes of any such disparities A study of the impact to the community of racial disparities in traffic stops, regardless of the causes of the disparities A review of current police procedures and how the police engage with the community This report is a compilation of the results of those four areas of study over the past year, along with the Task Force s conclusions and recommendations. The Task Force considers its work as the beginning, rather than the end, of this endeavor. While we have been able to do a significant review of the statistics, community impact, and police procedures and public engagement, the most we could do in the very short amount of time we were given was to identify areas of further exploration and give recommendations for future action. There is a great deal of work ahead to address the issues we have identified in this report. Page 1

5 MEMBERS OF THE TK FORCE Members of the Task Force CHAIR Peter Resnick MEMBERS Dr. Nicole Anderson-Cobb Patricia Avery Sgt. Andrew Charles Dr. Shinjinee Chattopadhyay Alejandra Coronel Dr. Eric Jakobsson Will Kyles Shandra Summerville Paul Testa Page 2

6 ACKNOWLEDGMENTS Acknowledgments The Task Force gratefully acknowledges the many people who contributed to this report. In particular, we would thank the many members of the public who attended our Town Hall meeting to give their input into the Community Impact section of this report, with special gratitude to Mr. Sam Smith who facilitated the discussion. Also, our thanks to all of the members of the public for their contributions during the public input section of our meetings and during the public comment period for our preliminary report, with a special note of thanks to Mr. Durl Kruse who not only provided valuable feedback during meetings but also contributed a great deal of research and information throughout our work. We are also grateful to the entire staff of the Urbana Human Relations Office for all of their support and to the staff of Urbana Public Television for their assistance with all of our meetings. Our thanks to the members of the Urbana Police Department staff who collected the statistical data that went into this report, and to Chief Patrick Connolly for his support of this process and willingness to engage with the Task Force. Finally, we would like to thank the Urbana City Council and Mayor Laurel Prussing for their courage and confidence in creating the Task Force and giving us the opportunity to address this important issue. Page 3

7 VOLUME II: STATISTICAL APPENDIX Volume II: Statistical Appendix VOLUME II: STATISTICAL APPENDIX The present publication, Volume II: Statistical Appendix, is a companion to Volume I: Main Report of the Final Report of the Urbana Traffic Stop Task Force, published in 215. VOLUME I: MAIN REPORT You may download Volume I: Main Report of the Final Report of the Urbana Traffic Stop Task Force at Page 4

8 Statistical Appendix Overview This appendix contains the analyses reported in the Urbana IDOT Tra The appendix is organized as follows: c Stop Data Task Forces final report. Section 1: IDOT Disparities presents the yearly disparity ratios from the IDOT report, as well as disparties for each racial group (Whites, African Americans, Hispanics, and Asians). Both the total and race-specific figures are calculated by comparing the proportion of stops that involve a minority driver (or specific racial group) to the estimated proportion of the driving population in Urbana that are minorities or from a specific racial group. Section 2: Demographic and Socio-economic Di erences explores demographic and socio-economic di erences that may factor into the observed disparities in tra c stops. Specifically, this section examines di erences in the driver age, vehicle age, and gender of drivers stopped. It also provides a description of driver residency. Section 3: Tra c Stops and Patterns of Policing examines the relationship between calls for service, tra c stops, and the racial composition of neighborhoods in Urbana. The analysis is limited to -213 (the years for which data on calls for service are available). The primary unit of analysis here is the Urbana Police Department s geocode. Urbana is divided into five police beats. Each beat is divided into smaller regions called geocodes, which are used to report the locations of both stops and calls for service There are around a 14 unique geocodes in the data depending on the year. Geocodes vary in size. In residential neighborhoods, they generally correspond to several city blocks, and are somewhat larger in more commercial areas or sparsely populated sections of Urbana. Estimates for the minority population of each geocode were obtained from the U.S. Census. The data for the race of residents in Urbana are avaialbe at the Census block level. Estimates of the racial compositin of each geocode were obtained by taking a weighted average of corresponding census blocks contained within that geocode. The section also explores whether, conditional on the number of calls for service, the precent of minoirities living in a geocode also predicts the number of tra c stops, through regression analyses, some of which control for the possibility of spatial dependence in the data. This section also provides local estimates of the disparity in tra c stops for each geocode. As with the measures reported in Section 1, for each geocode, we compare the proportion of stops involving a minority driver to the estimated minority population living in that area. Finally, the section also explores disparities in the Urbana Police Department s Selective Tra c Enforcement Program (STEP), a project designed to address high levels of accidents and other community concerns through concentrated policing. Section 4: Testing for Racial Profiling Using the Veil of Darkness presents the results from a series of tests designed for racial profiling using a procedure called the Veil of Darkness. 1. The logic of this test is outlined in the main body of the report. The first pair of figures show the set of stops that occur during the inter-twilight period that are used in the analysis. The three tables correspond to set of logistic regressions with three di erent outcomes: Whether the driver stopped was a minority (1 if minority, if white) Whether the driver stopped was African American (1 if African American, if not) Whether the driver stopped was African American or White (1 if African American, if white, Asian and Hispanic drivers are excluded from these models) The first column in each table presents the simplest model, testing whether whether drivers stopped when it is dark out are more or less likely to be minority or African American. A negative coe cient here would 1 See Grogger, Je rey, and Greg Ridgeway. Testing for racial profiling in tra c stops from behind a veil of darkness. Journal of the American Statistical Association 11.4 (26):

9 suggest evidence of profiling since when it is dark out, it should be harder to determine the driver s race. The next model adds a control for time of day, since the driving population at 5 pm may di er from the driving population at 8 pm. The third model, also this e ect to vary non-linearly through a cubic spline. The fourt model, then alows the e ects of darkness to vary conditionally on the time of day. The final model then allows these conditional e ects to vary by year as well. The figures associated each table are produced from the estimates of the fifth model. The solid line shows the predicted e ect of darkness on the log-odds that a driver is a minority or African American at di erent times of day. The dotted lines provide a 95 percent confidence interval for these estimates. When the prediction (solid line) and its confidence interval (dotted lines) are below zero (dashed line) this provides evidence that is consistent with the presence of racial profiling. Section 5 Disparities in Financial Impact examines the average fines and types of fines associated with tra c stops for each racial group. Section 6: Additional Analysis contains a number of other descriptive summaries of the data, breaking down the types, rates of citation, searches, contraband and duration of stops by racial group. Please feel free to contact Paul Testa (ptesta2@illinois.edu), the chair of the Task Force s Statistics Subcommittee, with any questions, comments, or concerns. Contents 1 IDOT Disparities 8 IDOT Disparity Ratios by... 8 ly Disparities by Demographic and Socio-economic Di erences 1 Driver Residency... 1 Driver Age... 1 Vehicle Age Gender Tra c Stops and Patterns of Policing 13 Stops and Calls for Service Calls for service Correlation between Calls for Service and Tra c Stops OLS Regressions of Stops on CFS and Minortiy population Results controlling for Spatial Dependence Disparities by Geographic Region Population estimates by Geocode Total Stops by Geocode... 2 Disparity Ratio Recent s: Only statistically significant disparties Recent s: (Only statistically significant disparties) Stops and the STEPS program

10 4 Testing for Racial Profiling Using the Veil of Darkness 27 Models ly Estimates of Racial Profiling of Minorities with Log-Odds ly Estimates of Racial Profiling of African Americans with Log-Odds ly Estimates of Racial Profiling of African Americans with Log-Odds (Excluding Other Minoriities from Analysis) Disparities in Financial Impact 37 Merging IDOT Data with Court Data Average Fine by Types of Charges by Number of Charges by Average Fine by Violation and Additional Analyses 42 Type of Stop Total Stops Percent of Total Stops Type of Stop by Citations Total Number of Citations Percent of Total Citaitons Rates of Citation Searches Total Number of Searches Propotion of Total Searches Rates of Searches Contraband Number of Stops with Contraband Found Percent of Total Contraband Found... 6 Percent of Stops with Contraband Found Duration of Stops Hitrates of Searches for Contraband 66 8 Tra c Stops and Cannabis 68 Aggregate Incidents Cannabis and Tra c Stops... 7 Incident Counts Incident Codes

11 1 IDOT Disparities IDOT Disparity Ratios by The State of Illinois requires that police departments collect information on tra c stops for the purpose of assessing racial bias, disparities and profiling in policing. One approach to measuring racial disparities with these data is to compare the proportion of minorities who are stopped to the estimated proportion of minority drivers in the population. The disparity measured by this ratio for Urbana, IL, from to 213 ranges between a high of 1.7 in and a low of 1.7 in. Table 1: ly IDOT Disparity Ratios # White Stops # Minority Stops % Stops White % Stops Minority Min % of Driv Pop Disparity ly Disparities by The observed disparity among minorities as a whole is due almost entirely to disparities in the rates at which African Americans are stopped, which ranges from a low of 1.71 in and 213 to a high of 2.18 in. 8

12 Table 2: ly Disparities by African Americans Stops Total Stops % Total Est % Population Disparity Hispanics Stops Total Stops % Total Est % Population Disparity Asians Stops Total Stops % Total Est % Population Disparity Whites Stops Total Stops % Total Est % Population Disparity Note: In 29 stops the drivers identified themselves as Native American. These cases are not included in the analysis above. 9

13 2 Demographic and Socio-economic Di erences Driver Residency Table 3: Tra c Stops and Driver Residency Driver From: # Stops % Total Urbana Urbana-Champaign Local Within 5 Miles Chicago 55.1 Illinois Just over half of the drivers stopped from -213 had addresses in Urbana, IL. Three-quarters lived in Urbana-Champaign (Local includes Savoy and St Jospeh),about 85 percent lived within 5 miles, and close to 98 percent lived in-state. Driver Age 1 Driver Age s 5 Comments Figure 1: Distribution of Driver s Age by There s greater variation in the age of white drivers, who also on average, tend to be slightly older than minority drivers. 1

14 Vehicle Age 1 Vehicle Age s 5 Comments Figure 2: Distribution of Vehicle Age by African Americans and Hispanics tend to drive slighltly older cars than Whites and Asians. 11

15 Gender Driver Gender Proportion Female Figure 3: Proportion of Stopped Drivers who are Female Comments The figure shows the proportion of drivers stopped who are female for each racial group each year. For the most part, men are more likely to be stopped than women, particularly for Asians and Hispanics. 12

16 3 Tra c Stops and Patterns of Policing Stops and Calls for Service Calls for service CFS Figure 4: Total Calls for Service -213 Correlation between Calls for Service and Tra c Stops 13

17 Table 4: Correlations between CFS and Tra c Stops Correlation Table 5: Correlations between CFS and Minority Percent of Population Correlation Table 6: Correlations between CFS and Minority Percent of Population Correlation

18 OLS Regressions of Stops on CFS and Minortiy population The models below present the results from a series of regression analyses, examing how the total number of tra c stops in a police geocode varies accordding to the number of calls for service and the percentage of minorities that live in that geocode. The first set of models ignore the possibility for spatial dependence in the data which can bias the models estimates (i.e. that regions high or low values of our variables may cluster together). Statistical tests suggests there is spatial dependence in the data, and seem to a favor an autoregressive lag model. 2 Without controlling for spatial dependence, the minority population in the geocode is a larger positive predictor of the number of tra c stops in a region, when holding constant the number of calls for service. However, when the spatial dependence of the data is taken into account, the percent of minoirities living in an area is no longer a siginficant predictor of tra c stops. Table 7 TotCFS crime crime211 crime crime213 Dependent variable: TotStops TotStops1 TotStops11 TotStops12 TotStops13 (1) (2) (3) (4) (5). úúú (.56).26 úúú (.49).194 úúú (.38).224 úúú (.66).348 úúú (.79) Min.p úú úú ú úú (4.84) (9.151) (7.532) (12.673) (13.68) pop (.35) (.8) (.7) (.11) (.12) Constant úú (16.855) (3.793) (3.98) (5.265) (5.68) Observations R Adjusted R Residual Std. Error (df = 134) F Statistic (df = 3; 134) úúú úúú úúú úúú úúú Note: ú p<.1; úú p<.5; úúú p<.1 2 We also estimated autorgressive error models, and used a n-nearest neighbors weighting matrix. The results are substantively the same to those reported above. 15

19 Results controlling for Spatial Dependence Neighbor Matrix 16

20 Table 8 TotCFS crime crime211 crime crime213 Dependent variable: TotStops TotStops1 TotStops11 TotStops12 TotStops13 (1) (2) (3) (4) (5).196 úúú (.47).1 úúú (.41).15 úúú (.32).169 úúú (.58).278 úúú (.69) Min.p (35.2) (7.842) (6.397) (11.3) (12.99) pop (.3) (.7) (.6) (.1) (.1) Constant (15.18) (3.32) (2.8) (4.877) (5.16) Observations Log Likelihood , Akaike Inf. Crit. 1, , , , , Wald Test (df = 1) úúú úúú úúú úúú úúú LR Test (df = 1) úúú úúú úúú úúú úúú Note: ú p<.1; úú p<.5; úúú p<.1 17

21 Table 9 TotCFS crime crime211 crime crime213 Dependent variable: TotStops TotStops1 TotStops11 TotStops12 TotStops13 (1) (2) (3) (4) (5).217 úúú (.49).19 úúú (.44).164 úúú (.34).194 úúú (.6).298 úúú (.7) Min.p (36.582) (8.387) (6.774) (11.74) (12.236) pop (.31) (.7) (.6) (.1) (.11) Constant (16.33) (3.64) (3.19) (5.155) (5.298) Observations Log Likelihood , ,8.693 Akaike Inf. Crit. 1, , , , , Wald Test (df = 1) úúú úúú úúú úúú úúú LR Test (df = 1) úúú 2.88 úúú úúú úúú.15 úúú Note: ú p<.1; úú p<.5; úúú p<.1 18

22 Disparities by Geographic Region Working with data from the census, we ve produced population estimates weighted by the census block for the racial composition of the 13+ police geocodes. 3 Population estimates by Geocode 213 Est Min Pop Figure 5: Estimated Minority Population 3 Specifically, we overlayed the police geocode map onto the census block maps and then weighted populations for each block by the proportion of the blocks total area within the geocode. Consider a block with 1 people. If that block falls entirely within ageocode,all1arecountedtowardtheestimatedpopulationofthegeocode.ifonlyhalfoftheblockfallswithinageocode, that block would add 5 people to the estimate of the total population of that geocode. 19

23 Total Stops by Geocode 213 # Stops Figure 6: Estimated Minority Population We can use information from the figures above to produce geocode-level measures of the IDOT disparity or relative risk of a minority being stopped based on the estimated minority population in each geocode. Spefically, for each geocode, i we calculate i, a ratio of two proportions: i = Minority Stops Total Stops Minority Population Total Population The figures below shows these estimates for each geocode, with blue being values below 1 (lower than expected risk of being stopped based on relative the proportion of minorities in the geocode s population), white being values close to 1 and red being values above 1 (more than expected risk). The same caveats about the IDOT measures apply to these, and note that when there few stops and/or small population in a geocode these estimates can be quite volatile. Disparity Ratio

24 213 Disparity Figure 7: Disparty Ratio by Geocode Recent s:

25 211 Disparity Figure 8: 211 Disparty Ratio by Geocode Only statistically significant disparties To capture this volatility, we also constructed confidence intervals for the point estimates, that reflect the uncertainty of estimates where their are relatively few stops or small populations. The figures below shows the geocodes with >1 (i.e. more than expected risk) whose 95-percent confidence intervals do not include 1. 22

26 Disparity Figure 9: Disparty Ratio by Geocode 213 Disparity Figure 1: 213 Disparty Ratio by Geocode 23

27 213 Disparity (95% ci > 1) 4 2 Figure 11: Statistically Significant Disparties by Geocode Recent s: (Only statistically significant disparties) 24

28 211 Disparity (95% ci > 1) 4 2 Figure 12: 211 Disparty Ratio by Geocode Disparity (95% ci > 1) 4 2 Figure 13: Disparty Ratio by Geocode

29 213 Disparity (95% ci > 1) 4 2 Figure 14: 213 Disparty Ratio by Geocode Stops and the STEPS program Disparities are lower for STEP-stops relative to non-step stops Table 1: Comparing Disparities in Steps vs Non-Steps Stops Est Pop % STEPS % STEPS Disp Non-STEPS % Non-STEPS Disp White Black Hispanic Asian Minority Total

30 4 Testing for Racial Profiling Using the Veil of Darkness All Stops Intertwilight Stops Time of Day : 1: 2: Time of Day 16:27 18:26 2:26 28 Day of 28 Day of Figure 15: Tra c Stops by Time of Day: Grey dots show stops that occured during the day and black dots show stops that occurred at night. Blue lines show dawn,sunrise,sunset,dusk. Red lines (left panel) denote the intertwilight period (right panel) used in the veil of darkness analysis Models 27

31 No Time of Day Linear E ect Cubic Spline Interaction FE Dark Out.12 ú (.6) (.7) (.7) (.51) (.51) Time of Day. úúú (.) Spline(Time of Day) (.21) (.) (.) Spline(Time of Day) 2.74 ú (.34) (.45) (.45) Spline(Time of Day) 3.88 úúú 1.12 úúú 1.12 úúú (.22) (.31) (.31) Spline(Time of Day) 4.78 úúú (.18) (.34) (.34) Spline(Time of Day) úú (.4) (.51) (.51) Spline(Time of Day) 6.54 úú (.17) (.48) (.49) Dark Out X Spline(Time of Day) (.53) (.53) Dark Out X Spline(Time of Day) (.81) (.81) Dark Out X Spline(Time of Day) 3.3. (.58) (.58) Dark Out X Spline(Time of Day) ú.98 ú (.5) (.5) Dark Out X Spline(Time of Day) (1.16) (1.16) Dark Out X Spline(Time of Day) (.53) (.54) AIC BIC Log Likelihood Deviance Num. obs úúú p<.1, úú p<.1, ú p<.5 Table 11: Testing for Racial Profiling of Minorities ly Estimates of Racial Profiling of Minorities with Log-Odds 28

32 No Time of Day Linear E ect Cubic Spline Interaction FE Dark Out.15 ú (.6) (.7) (.7) (.56) (.56) Time of Day. úúú (.) Spline(Time of Day) (.22) (.27) (.27) Spline(Time of Day) 2.72 ú (.36) (.47) (.48) Spline(Time of Day) 3.83 úúú 1.15 úúú 1.18 úúú (.23) (.31) (.32) Spline(Time of Day) 4.62 úú.6.6 (.19) (.36) (.36) Spline(Time of Day) 5.94 ú (.43) (.54) (.54) Spline(Time of Day) 6.52 úú (.18) (.5) (.5) Dark Out X Spline(Time of Day) (.58) (.58) Dark Out X Spline(Time of Day) (.87) (.87) Dark Out X Spline(Time of Day) (.63) (.63) Dark Out X Spline(Time of Day) ú 1.11 ú (.53) (.53) Dark Out X Spline(Time of Day) (1.28) (1.27) Dark Out X Spline(Time of Day) (.55) (.56) AIC BIC Log Likelihood Deviance Num. obs úúú p<.1, úú p<.1, ú p<.5 Table 12: Testing for Racial Profiling of African Americans Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day 29

33 213 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day 211 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds 5 16:57 17:49 18:41 19:33 2:26 Time of Day Time of Day Figure 16: ly Estimates of Racial Profiling of Minorities (2-13) Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day Figure 17: ly Estimates of Racial Profiling of Minorities (26-9) 3

34 No Time of Day Linear E ect Cubic Spline Interaction FE Dark Out.15 ú (.7) (.8) (.8) (.56) (.56) Time of Day. úúú (.) Spline(Time of Day) (.23) (.28) (.28) Spline(Time of Day) 2.81 ú (.37) (.49) (.49) Spline(Time of Day) 3.92 úúú 1.23 úúú 1.27 úúú (.24) (.33) (.33) Spline(Time of Day) 4.76 úúú (.2) (.37) (.37) Spline(Time of Day) úú.7.66 (.44) (.56) (.56) Spline(Time of Day) 6.58 úú (.18) (.52) (.52) Dark Out X Spline(Time of Day) (.59) (.59) Dark Out X Spline(Time of Day) (.89) (.89) Dark Out X Spline(Time of Day) (.64) (.64) Dark Out X Spline(Time of Day) ú 1.14 ú (.54) (.54) Dark Out X Spline(Time of Day) (1.29) (1.29) Dark Out X Spline(Time of Day) (.57) (.57) AIC BIC Log Likelihood Deviance Num. obs úúú p<.1, úú p<.1, ú p<.5 Table 13: Testing for Racial Profiling of African Americans (Other Minorities Excluded) ly Estimates of Racial Profiling of African Americans with Log-Odds 31

35 213 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day 211 Effect on Log Odds 1 16:57 17:49 18:41 19:33 2:26 Effect on Log Odds 3 16:57 17:49 18:41 19:33 2:26 Time of Day Time of Day Figure 18: ly Estimates of Racial Profiling of African Americans(2-13) Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day 32

36 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day Figure 19: ly Estimates of Racial Profiling of African Americans (-6) 33

37 ly Estimates of Racial Profiling of African Americans with Log-Odds (Excluding Other Minoriities from Analysis) 213 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day 211 Effect on Log Odds 1 16:57 17:49 18:41 19:33 2:26 Effect on Log Odds 3 16:57 17:49 18:41 19:33 2:26 Time of Day Time of Day Figure 2: ly Estimates of Racial Profiling of African Americans(2-13) 34

38 29 28 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day 35

39 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Effect on Log Odds :57 17:49 18:41 19:33 2:26 Time of Day Time of Day Figure 21: ly Estimates of Racial Profiling of African Americans (-6) 36

40 5 Disparities in Financial Impact Merging IDOT Data with Court Data To obtain estimates of the financial impact of tra c stops, we merged data on driver s race from the IDOT data with the Champaign County Court data on tra c citations from to 214 using the driver s first and last names.there are a total of 4,868 charges, with 26,389 unique defendants, with some defendants receiving multiple charges. Overall, we were able to match 77 percent of the court records with the IDOT data. In a given year, we are able to match between 15 and 2 percent of the cases, while in 214, 58 percent of the cases are unknown (labeled UK below). Since there are only 13 respondents who identify as Native American or Alaskan, they are excluded from subsquent analyis. Table 14: Defendants by (-213) NA UK Count Proportion Average Fine by In the sample, the average fine paid by a person in given case, (for which there may be multiple charges) is about $ The median fine is $77 dollars. The distribution of fines is very skewed. About 22 percent of the sample pay no fine, while 6 percent of the sample pay over $6 in fines. Looking at the distribution of fines by race, we see that African Americans and Hispancis, on average, are ordered to pay more fines than Whites and Asians. There are several possible reasons for this disparity, each of which we explore in more detail below. Table 15: Average Fines by (-214) Average Fine Stnd Dev 5th percentile percentile Maximum UK Types of Charges by First, the distribution of charges may vary across racial groups. African Americans and Hispanics, may be more likely to be charged with o enses that carry a higher fine. The table below provides some evidence of this. Driving without insurance or on a revoked license carry higher average fines than moving violations, and are more common among African Americans and Hispancis, than Whites and Asians. 37

41 Table 16: Top 1 Charges by (-214) White Count Mean Fine Driving 15-2 Mph Above Limit 2411 $18. Operate Uninsured Mtr Vehicle 21 $1. Driving Mph Above Limit 23 $14.17 Disregard Stop Sign 172 $11.5 Seat Belt Required/driver 717 $52.92 Disreg Tra c Control Light 636 $1.5 Fail To Reduce Speed 578 $112.3 Driving On Suspended License 464 $ Driving 1-1 Mph Above Limit 423 $12.64 Drvg Under Influ Of Alcohol 421 $79.87 African American Count Mean Fine Operate Uninsured Mtr Vehicle 3 $ Driving On Suspended License 1121 $29.82 Driving 15-2 Mph Above Limit 969 $92.81 Unlicensed 893 $ Driving Mph Above Limit 828 $92.8 Disregard Stop Sign 76 $82.17 Op Veh W/loud System > Ft 452 $65.89 Driving On Revoked License 426 $9.45 Seat Belt Required/driver 333 $44.6 Fail To Reduce Speed 297 $83.53 Hispanic Count Mean Fine Unlicensed 4 $172.2 Operate Uninsured Mtr Vehicle 394 $ Driving 15-2 Mph Above Limit 136 $14.56 Disregard Stop Sign 118 $86.19 Driving Mph Above Limit 13 $9.55 Driving On Suspended License 94 $ Drvg Under Influ Of Alcohol 61 $ Improper Tra c Lane Usage 43 $73.86 Disreg Tra c Control Light 41 $88.5 Drvg Under Influ/bac.8 41 $ Asian Count Mean Fine Driving 15-2 Mph Above Limit 528 $17.6 Driving Mph Above Limit 317 $15.95 Operate Uninsured Mtr Vehicle 33 $7.73 Disregard Stop Sign 298 $1.96 Disreg Tra c Control Light 16 $17.3 Unlicensed 97 $49.93 Unsafe Equipment/1st and 2nd 9 $ Fail To Reduce Speed 84 $ Driving 21- Mph Above Limit 7 $12.86 Improper Tra c Lane Usage 55 $

42 Number of Charges by Second, members of di erent racial groups may be more or less likely to be charged with multiple o enses (e.g. speeding and driving without insurance), which would raise the average fine per person in these groups. Again, the data support this view. Fourty-two percent of African Americans and 49 percent of Hispanics are charged with more than one violation, compared to percent of Asians and 26 percent of Whites. Individuals with one charge, pay between $1 and $13 dollars in fines. Those charged with more than one fine pay about $3 to $4 dollars more Table 17: Number of Charges by (-214) One Two Three Four Five UK Table 18: Proportion of Multipe Charges by (-214) One Two Three Four Five UK

43 .6 Proportion.4 UK Number of Charges per Defendant 4

44 Average Fine by Violation and Finally, it is possible, that for the same o ense, di erent minorty groups recieve di erent fines. The evidence here is mixed. African Americans and Hispancis are significantly more likely to pay higher fines for driving without insurance and being unlicensed. Whites pay more for moving violations and DUIs compared to African Americans and Hispanics, but not Asians. Asians are fined more for tra c lane violations Table 19: Di erences in Average Fines for Selected Charges by (-214) White-African American Mean Fine Mean Fine Di erence Driving 15-2 Mph Above Limit * Driving Mph Above Limit * Seat Belt Required/driver * Disregard Stop Sign * Improper Tra c Lane Usage Operate Uninsured Mtr Vehicle * Unlicensed * Driving On Suspended License Driving On Revoked License Drvg Under Influ Of Alcohol * White-Hispanic Mean Fine Mean Fine Di erence Driving 15-2 Mph Above Limit Driving Mph Above Limit * Seat Belt Required/driver * Disregard Stop Sign * Improper Tra c Lane Usage Operate Uninsured Mtr Vehicle * Unlicensed * Driving On Suspended License Driving On Revoked License Drvg Under Influ Of Alcohol * White-Asian Mean Fine Mean Fine Di erence Driving 15-2 Mph Above Limit Driving Mph Above Limit Seat Belt Required/driver Disregard Stop Sign Improper Tra c Lane Usage * Operate Uninsured Mtr Vehicle Unlicensed * Driving On Suspended License Driving On Revoked License Drvg Under Influ Of Alcohol Note:*p <.5 41

45 6 Additional Analyses Complete Summary of Stops, Citations, Searches, and Contraband by 42

46 Type of Stop Total Stops Stops All Stops 28 Stops Moving Violation Stops Equipment 28 Stops License/Registraion Figure 22: Total Number of Stops by and The figure shows the total number of stops by year and type of stop for each racial group. Comments Moving violations are the most common reason for stop, followed by equipment violations, and stops for License plates/registration (L/R) Increase in total stops peaks at 29, driven by rises in the number of equipment and L/R stops. Increase from reflects increase across all type of stops. White and African American drivers make up the majority of stops. 43

47 Percent of Total Stops % of Stops All Stops 28 % of MV Stops 1 5 Moving Violation % of Eq Stops Equipment 28 % of L/R Stops 1 5 Figure 23: Proportion of ly Stops by License/Registraion The figure shows for a given year and type of stop, what proportion of the stops are from what racial group. - The proportion of total stops by race is relatively constant over the years. - Whites and African Americans account for generally over 9 percent of all stops - Whites make up the majority of moving violations - African Americans account for the plurality of Equipment and L/R stops Type of Stop by The figure shows the proportion of each racial group s total stops that are for moving violations, equipment, and L/R. Comments Moving violations are the most common type of stop for all races Equipment and L/R stops tend to be more common among African Americans and Hispanics 44

48 % of 's Stops Moving 28 % of 's Stops Equipment 28 % of 's Stops 1 5 License/Registration Figure 24: Type of Stop by and Table 2: Tra c Stops by Total # % # % # % # %

49 Table 21: Moving Violations by Total # % # % # % # % Table 22: License and Registration Violations By Total # % # % # % # % Table 23: Equipment Violations by Total # % # % # % # %

50 Citations Total Number of Citations Citations All Stops 28 Citations Moving Violation Citations Equipment 28 Citations License/Registration Figure : Total Number of Citations by,, and Type of Stop The figure shows total number of citations issued in a given year to drivers of a certain race. 47

51 Percent of Total Citaitons % Citations All Stops 28 % MV Citations 1 5 Moving Violation % Eq Citations 1 5 Equipment % L/R Citations 1 5 License/Registration Figure 26: Proportion of Total Citations by,, and Type of Stop The figure shows the proportion of total citations in a year issued to each racial group for all stops, and then separately for moving, equipment and L/R violations. Comments Gaps between Whites and African American Drivers in terms of citations for Equipment and L/R stops 48

52 Rates of Citation % Stops w/ Citation All Stops 28 % MV w/ Citaiton 1 5 Moving Violation % Eq w/ Citation Equipment 28 % L/R w/ Citation 1 5 License/Registration Figure 27: Rates of Citations by,, and Type of Stop The figure shows the rates of stops which result in citations for each racial group. Comments Hispanics are far more likely to get a citation, particularly for L/R stops. 49

53 Table 24: Citations by Total # % # % # % # % Table : Moving Violation Citations by Total # % # % # % # % Table 26: Lic/Reg Citations by Total # % # % # % # %

54 Table 27: Equipment Citations by Total # % # % # % # % Table 28: Percent of Stops with Citations by Stops # % Stops # % Stops # % Stops # % Table 29: Percent of Stops with Citations for Moving Violations by Stops # % Stops # % Stops # % Stops # %

55 Table 3: Percent of Stops with Citations for Lic/Reg Violations by Stops # % Stops # % Stops # % Stops # % Table 31: Percent of Stops with Citations for Equipment Violations by Stops # % Stops # % Stops # % Stops # %

56 Searches Total Number of Searches Searches All Stops 28 Searches Moving Violation Searches Equipment 28 Searches License/Registraion Figure 28: Total Number of Searches by,, and Type of Stop The figure shows the overall number of stops in year by racial group. Comments Overall, it seems the number of searches has been declining. The format for reporting searches are reported in the data frequently changed over -. 53

57 Propotion of Total Searches % of Total Searches All Stops 28 % of MV Searches 1 5 Moving Violation % of Eq Searches 1 5 Equipment % of L/R Searches 1 5 License/Registration Figure 29: Proportion of Total Searches by,, and Type of Stop The figure shows for each year what proportion of the years searches were conducted on drivers from each racial group Comments African Americans consistently make up the majority of drivers searched. 54

58 Rates of Searches % Stops w/ Search All Stops 28 % MV w/ Search 1 5 Moving Violation % Eq w/ Search Equipment 28 % L/R w/ Search 1 5 License/Registration Figure 3: Rates of Searches by,, and Type of Stop The figure shows a given racial group, what proportion of their stops result in a search Comments Hispanic and African American drivers are consistently more likely to be searched during a stop 55

59 Table 32: Total Searches by Total # % # % # % # % Table 33: Searches for Moving Violations by Total # % # % # % # % Table 34: Searches for Lic/Reg by Total # % # % # % # %

60 Table 35: Searches for Equipment Violations by Total # % # % # % # % Table 36: Percent of Stops with Searches by Stops # % Stops # % Stops # % Stops # % Table 37: Percent of Stops with Searches for Moving Violations by Stops # % Stops # % Stops # % Stops # %

61 Table 38: Percent of Stops with Searches for Lic/Reg Violations by Stops # % Stops # % Stops # % Stops # % Table 39: Percent of Stops with Searches for Equipment Violations by Stops # % Stops # % Stops # % Stops # %

62 Contraband Number of Stops with Contraband Found Contraband All Stops 28 Contraband 1 5 Moving Violation Contraband Equipment 28 Contraband 1 5 License/Registraion Figure 31: Amount of Contraband by,, and Type of Stop The figure shows the total number of stops that resulted in contraband (drugs, paraphernalia,alcohol,weapons) being found. ** Comments** The data start in 26. Finding contraband is a relatively rare experience Decline mirrors decline in total number of searches A back of the envelop calculation suggests a third of searches produce contraband (will follow up,more formally) 59

63 Percent of Total Contraband Found % All Contraband All Stops 28 % MV Contraband 1 5 Moving Violation % Eq Contraband Equipment 28 % L/R Contraband 1 5 License/Registration Figure 32: Porportion of Contraband by,, and Type of Stop The figure shows the porportion of contraband found by driver s race. ** Comments** Majority of contraband found from stops involving African Americans and Whites 6

64 Percent of Stops with Contraband Found % Stops w/ Contr All Stops 28 % MV w/ Contr 1 5 Moving Violation % Eq w/ Contr 1 5 Equipment % L/R w/ Contr 1 5 License/Registration Figure 33: Porportion of Stops with Contraband by,, and Type of Stop The figure shows the proportion of the stops which result in contraband being found for each racial group. Comments A relatively small proportion of stops result in contraband being found. 61