Appendix 3 - Addendum Report: Additional Benchmarking Of TPS Use Of Force And Charge Data

 

Dr. Scot Wortley and Dr. Ayobami Laniyonu
Centre for Criminology and Sociolegal Studies
University of Toronto

 

Prepared for the Ontario Human Rights Commission
November 2022

 

Introduction

This report is designed as an addendum to A Disparate Impact, published by the Ontario Human Rights Commissions as part of their inquiry into anti-Black racism within the Toronto Police Service.  This addendum incorporates data that was not available during the preparation of A Disparate Impact.

• Part A of the addendum benchmarks police use of force incidents against race-based arrest and street check data;
• Part B presents a multivariate analysis, predicting use of force incidents, that incorporates racial differences in arrests and police contact;
• Part C of the report benchmarks “out-of-sight” traffic charges with estimates of Toronto’s driving population;
• Part D of the report benchmarks failure to comply charges against TPS arrest statistics;
• Finally, we provide a methodological appendix that further explains the multivariate analyses conducted in the original use of force report. [1]

 

PART A: Additional Use of Force Benchmarking

Our original use of force report (Wortley, Laniyonu, and Laming 2020) documented that, compared to their presence in the general population, Black people are grossly over-represented in Toronto Police Service (TPS) use of force cases.  Indeed, we found that Black people were over-represented in both use of force cases that resulted in civilian death or serious injury (as documented by SIU investigations), and lower-level use of force cases that did not result in an injury that would warrant SIU attention (as measured by an analysis of TPS Injury Reports and General Occurrence data).[2] 

General population benchmarking captures the overall impact of police use of force on racialized communities.  Proponents maintain that general population benchmarking reveals the likelihood that people from different racial backgrounds will experience police contact and/or  a police use of force incident.  A growing number of researchers recognize that census benchmarking is a valuable first step in the research process and that it serves to effectively document the extent to which different racial groups experience different types of police contact.  For example, a recent Home Office study concluded that: “When they are based on a wide enough geographical area, statistics based on resident populations still give us an important indication of how often members of different ethnic communities are actually stopped and searched in that area” (MVA and Miller 2000: 84).  Similarly, Riley and his colleagues (2009: 26-27) conclude that “comparisons based on the residential population remain important because they illustrate the experience of different ethnic groups irrespective of the reasons that may explain any disparities.  Disproportionality is a critical issue for the police service because evidence shows that negative police practices can damage public confidence and because being stopped and searched has been linked with lower satisfaction levels with the police.”  Miller (2010) has also argued that census benchmarking is likely the best method for documenting racial disparities over time (see Miller 2010).  The argument in favour of census benchmarking is also articulated by Benjamin Bowling and Coretta Phillips (2007).  Following their review of different benchmarking strategies used within racial profiling research, these prominent British scholars concluded that:

It is our view that the most robust measure of disproportionality in the use of police stop and search powers, and which relies on the fewest assumptions, is the per capita stop/search rate….The issue of availability provides no defence against the charge that routine practices are having a disproportionate impact on people from minority groups; thus prompting the Lawrence Inquiry label of ‘institutional racism.” The most important point is that the per capita rate provides, by definition, an estimate of the population group experience.  Thus, in a large geographical context such as the London Metropolitan Police Area or England and Wales as a whole, statistics based on resident populations provide an important indicator of how often members of different ethnic communities are actually stopped and searched within that area.  As Home Office researchers bluntly put it, per capita stop/search rates show clearly that being Black means that you are going to be stopped more often (Bowling and Phillips 2007: 952-953).

We strongly believe that the logic used to justify census benchmarking with respect to police stop and search activities can be applied to studies of police use of force.  However, we also acknowledge that, while general population benchmarking may highlight the over-representation or under-representation of racialized people in use of force statistics, these statistics may not completely explain racial disparities. In other words, general population benchmarking is not the only method that can be used to capture the “population at risk” of experiencing police use of force.  It could be argued, for example, that racial groups with high levels of contact with the police are at greater risk of experiencing police violence than those with lower levels of contact.  It could also be argued that those who have broken the law – and targeted for arrest – are at especially high risk of police use of force. Furthermore, it has been argued that violent offenders (i.e., those involved in arrests for violent crime) are more likely to demonstrate “resistance” to the police and are thus particularly vulnerable to police use of force incidents (see Tregle, Nix and Alpert 2019).  With these arguments in mind, in this section we augment our original general population benchmarking with benchmarks that document racial differences with respect to both police contact (street checks) and arrests.

To the best of our knowledge, during the study period (January 1^st, 2013 to June 30^th, 2017), the TPS did not collect or disseminate data documenting racial differences in police contact.  For example, the TPS did not release data documenting racial differences with respect to traffic stops, pedestrian stops, or calls for service. Thus, we decided to use TPS street check data, collected between 2008 and 2013, to estimate racial differences in police contact.[3]

During the development of our initial report, we did not have information on whether the TPS compiled or would be willing to release statistics on race and crime. However, following the release of A Disparate Impact the OHRC requested and received TPS data on the race of accused persons, arrested for various violent and non-violent offences, between 2014 and 2017.

The data provided in Table A1 reveal that Black people are significantly over-represented in TPS street checks and arrests.  Although they represent only 8.8% of Toronto’s general population (according to the 2016 Census), Black people were involved in 22.8% of all street checks, 24.8% of all arrests, 23.3% of arrests for property crime, 27.6% of arrests for violent crime, 38.6% of arrests for aggravated assault, 44.5% of homicide arrests, 42.3% of arrests for attempted homicide, and 51.6% of arrests for firearms-related offences.

It should be noted that arrests for serious violence are quite rare.  For example, between 2014 and 2017, the TPS made 110,218 arrests. However, only 164 of these arrests (0.1%) were for homicide.  Similarly, only 0.2% of all arrests were for attempted murder, 0.8% were for aggravated assault, and 2.2% were for firearms-related offences.

It is not the purpose of this addendum report to provide an in-depth explanation for the over-representation of Black people in arrest statistics.  However, as discussed in our earlier report, most criminologists agree that it is likely a combination of both racial bias within the criminal justice system and higher rates of “street-level” offending (see Wortley and Jung 2020).  Racial bias contributes to racial disparities in arrest statistics in several ways.  To begin with, Black people often come under higher levels of police surveillance than White people.  For example, numerous studies reveal that Black people are grossly over-represented in police stop and search activities.  Biased police surveillance practices entail that Black and other racialized people are more likely to be caught for breaking the law – and subsequently arrested – than White people who engage in exactly the same behaviour.  Research also indicates that, when illegal activity is identified, Black people are more likely to be charged with a crime than cautioned by the police or offered diversion programs.   As highlighted by Goff and his colleagues in their report entitled The Science of Justice:

Unfortunately, there is no way to take a true measure of criminality within a population, and the nearest approximation is problematic. Arrest data, which provide the closest estimate of criminal activity within a population (short of direct observation), are compromised by the very nature of who makes arrests. That is, because police arrest people and our concern is with the possibility that police behave in a biased manner when applying force, there is the strong likelihood that arrest data would be biased in the same manner as use of force data. Benchmarking use of force data to arrest data likely underestimates the level of bias that may exist in police use of force (Goff et al. 2016: 5).    

Nonetheless, we can’t discount the possibility that some of the racial disparity with respect to use of force is related to racial differences in offending behaviour.  As documented by Ontario’s Roots of Youth Violence Inquiry (McMurtry and Curling 2008), higher rates of offending among Black and Indigenous peoples in Canada can be traced back to colonialism and the institution of slavery.  These historical processes resulted in systemic racism, multi-generational trauma, and contemporary racial inequality.  As a result, Black Canadians are more likely to live in disadvantaged communities and suffer from unemployment, poverty, limited social capital, social alienation, and hopelessness.  A large volume of criminological research reveals that these factors are significantly related to criminal offending.  It is also important to note that, although Black people may be statistically over-represented in some TPS crime categories – including gun violence – the vast majority of Black people are law abiding.  Despite facing the perils of racism and inequality – most Black people are resilient and never break the law.  This majority does not deserve to be profiled because of the actions of a small number of Black offenders.  We will return to this issue at the conclusion of this section. 

 

Measuring Black Representation in TPS Use of Force Incidents

Consistent with the strategy used in our previous reports, Odds Ratios were calculated, using different population benchmarks, to determine the representation of Black people in TPS use of force cases. Odds ratios were calculated by dividing the percentage of all use of force cases involving Black people by their percent representation within each benchmark. An Odds Ratio approaching 1.00 indicates that Black people are neither over-represented nor under-represented in use of force cases. An odds ratio less than 1.00 indicates that Black people are under-represented in use of force incidents. An odds ratio greater than 1.00 indicates that Black people are over-represented in use of force cases. For example, an Odds Ratio of 2.00 would indicate that Black people are twice as involved in TPS use of force cases as they are in the population benchmark under consideration. By contrast, an Odds Ratio of 0.50 would indicates that Black people are 50% less represented in use of force cases than their proportion of the benchmark population would predict.

As discussed in our earlier report, there is no set standard for determining when racial disproportionality (i.e., the over- or under-representation of a particular racial group with respect to a specific social outcome) is cause for concern. However, for the purposes of this study, we have used a relatively high threshold of 50%. In other words, for the purposes of the present analysis, an Odds Ratio of 1.50 or higher will be used to determine whether the over-representation of Black people in TPS use of force cases is noteworthy or not. At times we will discuss the notion of “gross” racial disparity. For the purposes of this report, a gross racial disparity exists when the level of over-representation is 200% or greater (i.e., as indicated by an odds ratio of 3.00 or higher).

 

Findings

As reported in our earlier reports, the data presented in Table A2 demonstrate that, compared to their presence in the general Toronto population, Black people are highly over-represented in TPS use of force cases documented between January 1^st, 2013 and June 30^th, 2017.[4]  For example, compared to their presence in the general population, Black people are 3.27 times more likely to be involved in an SIU use of force investigation, 4.09 times more likely to be involved in an SIU shooting investigation, 4.42 times more likely to be involved in a lower-level use of force incident, 6.99 times more likely to be involved in a TPS use of force incident that resulted in civilian death, and 7.95 times more likely to be involved in a TPS shooting-related death (see Table A2 below).

We next benchmarked use of force incidents against street checks conducted by the TPS between 2008 and 2013.  The results indicate that, using this alternative benchmarking method, Black people remain over-represented in TPS use of force statistics.  Black people are, in fact, significantly over-represented in lower-level use of force incidents (Odds Ratio=1.71), SIU shooting investigations (Odds Ratio=1.58), and TPS use of force incidents that resulted in civilian death (Odds Ratio=2.69).  Furthermore, using street checks as a benchmark, Black people are still grossly over-represented in TPS shooting deaths (Odds Ratio=3.07).  It is important to note, however, that street check benchmarks produced lower Odds Ratios than general population benchmarks.  This finding suggests that higher rates of police contact may help explain the over-representation of Black people in TPS use of force statistics.  These results are also consistent with other report findings which suggest that, compared to cases involving White people, use of force incidents involving Black people are more likely to involve proactive policing practices (i.e., traffic stops).  Overall, these findings are consistent with the argument that racial profiling contributes to the over-representation of Black people in use of force incidents by increasing the number of negative, involuntary contacts between the police and Black residents.  The higher the number of negative, involuntary contacts, the greater the likelihood that some cases will devolve into an incident involving police use of force.

TPS use of force incidents were next benchmarked against overall TPS arrest statistics.  As with street checks, this benchmarking method produces lower odds ratios than general population benchmarks.  However, even when the arrest benchmark is used, Black people are still significantly over-represented in lower-level use of force incidents (Odds Ratio=1.57), SIU Death investigations (Odds Ratio=2.48), and TPS shooting deaths (Odds Ratio=2.82).  Black people are also over-represented with respect to TPS shootings (Odds Ratio=1.45) and SIU use of force investigations (Odds Ratio=1.16) – but the Odds Ratio fall below the 1.50 significance threshold used in the current study.  Overall, the fact that Black people are over-represented in TPS arrest statistics cannot explain the overrepresentation of Black people in TPS use of force incidents.

TPS use of force incidents were next benchmarked against the TPS arrests for property crime.  The results are very similar to those produced by the total arrest benchmark.  Using the property crime benchmark, Black people remain significantly over-represented in lower-level use of force incidents (Odds Ratio=1.67), SIU shooting investigations (Odds Ratio=1.54), SIU death investigations (Odds Ratio=2.64), and TPS shooting deaths (Odds Ratio=3.00).  However, the over-representation of Black people in SIU use of force investigations falls below the 1.50 level of significance established by this study.

Following the recent example set by American researchers (Tregle et al. 2019), we next benchmarked TPS use of force incidents against TPS arrests for violent crime.  It should be noted that when Tregle and his colleagues (2019) benchmarked American fatal officer-involved shootings with American violent crime arrests – they found that Black citizens were less likely to be fatally shot by the police than their White counterparts.  This is not the case with the TPS.  Indeed, using TPS arrests for violent crime as a benchmark, Black people are still 2.23 times more likely to be involved in a TPS fatal use of force incident and 2.54 times more likely to be involved in a fatal, officer-involved shooting.[5]  Furthermore, using arrests for violent crime as a benchmark, Black people are also over-represented in TPS lower-level use of force incidents (Odds Ratio=1.41) and SIU shooting investigations (Odds Ratio=1.30).  However, these odds ratios do not meet the 1.50 significance threshold established for this study.

Even when we use arrests for “serious” violence as the benchmark – Black people remain significantly over-represented in TPS fatal shootings.  For example, when we use arrests for aggravated assault as the benchmark, Black people are still 1.81 times more likely to be involved in a fatal officer-involved shooting.  When we use attempted homicide arrests as the benchmark, Black people are still 1.65 times more likely to be fatally shot by a TPS officer.  Finally, when we use homicide arrests as the benchmark, Black people are still 1.57 times more likely to become the victim of a fatal police shooting. 

The only benchmark that renders Black over-representation insignificant is arrests for firearms offences (see Table A2).  Using firearms-related arrests as the benchmark, Black people are only 1.36 times more likely to be involved in a fatal TPS shooting.  This odds ratio is below the 1.50 significance threshold established for this study.  Furthermore, using the firearms arrest benchmark, Black people become under-represented in both lower-level use of force incidents (Odds Ratio=0.75) and SIU use of force investigations (Odds Ratio=0.55).

It must be stressed that, due to very small numbers, the use of “serious violence” to benchmark use of force incidents may be statistically problematic.  For example, between 2014 and 2017, the TPS made only 164 arrests for homicide (41 per year), 281 arrests for attempted homicide (70 per year), 911 arrests for aggravated assault (228 per year), and 2,469 arrests for firearms offences (617 per year).  By contrast, during this same period, the TPS conducted 110,218 arrests in total (27,554 per year) and 43,245 arrests for violent crime (10,811 per year).  Based on these numbers, the overall arrest and violent arrest benchmarks are likely far more stable than the benchmarks for “serious violence” (i.e., homicide, attempted homicide, aggravated assault, and firearms violations).

 

Context and Caution  

The analysis presented above reveals that both street check and arrest benchmarking practices reduce – but do not eliminate – the over-representation of Black people in TPS use of force incidents.  In other words, even when we consider the proportion of arrests that involve Black suspects, Black people remain significantly over-represented in TPS use of force incidents -- including police shootings and shooting deaths.  These findings, in our opinion, provide further evidence that racial bias contributes to racial disparities in TPS use of force.  As stated by Goff and his colleagues:

If, however, a department were to demonstrate racial disparities in the application of force even controlling for arrest rates, this would provide reason for pause. If that pattern held for a plurality of departments, it would also cast doubt on the prospect that disparities in criminal behavior explain disparities in force. In this light, benchmarking police use of force to arrest rates may prove a usefully conservative (prone to false negatives, if anything) test of departmental bias despite the problem of endogeneity.

Nonetheless, the results also reveal that the more serious the arrest category – the less significant the over-representation of Black people.  Some may interpret these findings as “evidence” that it is “serious criminal behaviour,” not race, that explains why Black people are more likely to be involved in TPS use of force incidents. Such an interpretation of the data should only be considered with great caution. Indeed, aggregate level associations between arrest statistics and use of force statistics diverge significantly from the information provided in individual case files.

For example, the fact that Black people are over-represented in TPS arrest statistics may be misinterpreted as evidence that the Black individuals involved in police use of force incidents have lengthy criminal records involving violent offences and are thus”known to be dangerous” during police encounters.  However, between 2013 and 2017, 55.6% of the Black people involved in SIU use of force investigations had no previous criminal record.  Furthermore, the fact that Black people are over-represented in firearms arrests may give the impression that the Black individuals involved in TPS use of force cases were usually armed with a gun at the time of the incident.  This is not the case.  The data indicate that, between 2013 and 2017, two-thirds of the Black individuals involved in SIU investigations were unarmed during the use of force incident. Only 8.3% were in possession of a firearm.  Further analysis reveals that very few of the TPS use of force incidents documented by this study – including lower-level use of force cases – involved an attempt to arrest a suspect for a serious violent offence like homicide, attempted homicide, aggravated assault, or firearms possession.

How can we reconcile the fact that, while Black people are over-represented in TPS arrests for violent crime, most Black people involved in TPS use of force incidents were unarmed at the time of the incident and did not have a criminal record?  One possibility is that, although serious violence remains quite rare in Toronto, police officers are aware that Black males are over-represented in such cases. This awareness may stem from exposure to race-based arrest statistics, negative media depictions of Black males, orr through informal narratives shared within the police subculture. Officer awareness of the over-representation of Black males in violent crime may lead some officers to stereotype all Black people as potentially dangerous.[6]  As a result, police officers may become more fearful or hyper-vigilant when dealing with Black people in the community.  This fear or hyper-vigilance may cause some officers to interpret incidents involving Black people as “more dangerous” and thus deserving of use of force.  In sum, the data support the argument that racial stereotyping and can help explain the over-representation of Black people in use of force incidents.  This argument is further supported by the Toronto Police Service’s own analysis of 2020 use if force data.  This analysis further documents that Black people are over-represented in TPS use of force incidents and that this over-representation cannot be explained by other factors including age, gender, nature of police contact, arrest statistics, or the presence of weapons.  For example, consistent with the racialized fear or stereotype argument, the TPS analysis reveals that, in 2020, TPS officers were 2.3 times more likely to point a firearm at an unarmed Black person than an unarmed White person (Toronto Police Service 2022). 

In the next section of this addendum report we continue our examination of this issue by providing a multivariate analysis that further documents the association between patrol zone characteristics, race-based arrest statistics, street checks, and use of force incidents.

TABLE A1: The Representation of Black People in TPS Street Checks (2008-2013) and Arrest Statistics (2014-2017), by Crime Type

 

Type of Benchmark	Total Number	Number Involving Black People	Percent Involving Black People	Odds Ratio
Street Checks	2,026,258	461,468	22.8%	2.59
Total Arrests	110,218	27,314	24.8%	2.82
Arrests for Property Crime	50,093	11,664	23.3%	2.65
Arrests for Violent Crime	43,245	11,940	27.6%	3.14
Arrests for Aggravated Assault	911	352	38.6%	4.39
Arrests for Homicide	164	73	44.5%	5.06
Arrests for Attempted Homicide	281	119	42.3%	4.81
Arrests for Firearms Offences	2,469	1,275	51.6%	5.87

 

TABLE A2: The Representation of Black People in 2013 to 2017 TPS Use of Force Cases,

By Type of Benchmark

 

Benchmark	Lower- Level Use of Force	SIU Use of Force Investigations	SIU Shooting Investigations	SIU Death Investigations	SIU Shooting Death Investigations
General Population	4.42	3.27	4.09	6.99	7.95
Street Checks (2008-2013)	1.71	1.26	1.58	2.69	3.07
Total Arrests	1.57	1.16	1.45	2.48	2.82
Arrests for Property Crime	1.67	1.24	1.54	2.64	3.00
Arrests for Violent Crime	1.41	1.04	1.30	2.23	2.54
Arrests for Aggravated Assault	0.99	0.66	0.93	1.59	1.81
Arrests for Homicide	0.87	0.57	0.81	1.38	1.57
Arrests for Attempted Homicide	0.92	0.68	0.85	1.45	1.65
Arrests for Firearms Offences	0.75	0.55	0.70	1.19	1.36

 

PART B: Multivariate Analyses -- Additional Benchmarks

The results of the multivariate analysis presented in the main report benchmarked use of force incidents against the population prevalence of each racial group in TPS patrol zones. As discussed in the main report, the principal purpose of these models was to test whether the observed racial disparities in the odds of experiencing police use of force persist after controlling for the independent effects of aggregate patrol zone characteristics. Again, patrol zone-level characteristics like violent crime rates or poverty rates may affect the odds that an individual will experience force (see section B of the main report). As we explained in Part A of this Addendum, these analyses relied on population benchmarks, which capture the overall impact of police use of force on racialized communities.

However, general population benchmarks, as discussed above, may not perfectly measure the population at risk of police use of force. Consider the following scenario. Suppose that as a matter of police policy all young persons under the age of 12 years—regardless of race—are at zero risk of experiencing force. If young persons under 12 constitute a larger share of the Black population than the White population, then population-based disparity measures for Black persons will be downwardly biased or will underestimate the risk of force for Black persons. This is because the Black population at risk of force is smaller than that assumed by the benchmark and force is more concentrated among those who are actually at risk. Similarly, if Black persons engage in illegal behaviors at higher rates  than Whites, and thus legitimately draw the attention of the police, then population-based disparity benchmarks may be upwardly biased or overstate racial disparities.

It is impossible to determine the actual population at risk of experiencing police use of force using administrative police data (see Knox, Lowe and Mummolo 2019; Knox and Mummolo 2020). As a result, some recommend estimating disparities in use of force across multiple potential benchmarks (see discussion in Part A). Again, previous research focuses on three methodologies: population-based benchmarks, police contact-based benchmarks, and arrest-based benchmarks. Contact-based measures benchmark police use of force against rates at which different populations come into contact with the police, often estimated through general population surveys, police contact (street check) reports, or calls for service. Crime-based measures benchmark rates of force against rates at which different groups are arrested for criminal behaviour.

We note at this juncture that both contact-based and arrest-based benchmarks may underestimate racial disparities in police use of force (Knox, Lowe and Mummolo 2019). As discussed above, if police officers discriminate in stop, search, or arrest decisions—for example, by disproportionately stopping, investigating or charging Black citizens—this will increase the size of the Black population deemed at risk of police use of force and thereby decrease the estimated risk of force for Black people. In other words, racial bias will increase the denominator in an estimated risk ratio, downwardly biasing the overall estimated risk.  A sizable literature, summarized in the main report, suggests that police officers in Canada engage in racially biased stop and arrest practices, meaning that contact and arrest-based measures of racial disparities will be biased downward.  As we noted above, Black persons in Toronto are grossly over-represented in TPS street checks and arrests. This overrepresentation is almost certainly a consequence, at least in part, of racial bias among TPS officers. Overall, this means that benchmarking on street checks and arrests should be considered conservative estimates of racial disparities.

That said, below we present the results of a series of negative binomial models which: 1) benchmark use of force against rates for Black, White, and Other Torontonians against the arrest rates for each group; and 2) benchmark use of force rates against rates of police contact for each group as measured by field information reports (street checks).^[7] In other words, we estimate whether or not use of force rates are out of proportion with the rates at which members of each group are arrested or come into contact by the police. These models also control for patrol zone characteristics which may affect the odds that an individual will experience force (see section B of the Main report). We control for the patrol-zone violent crime rates, median household income, and the share of single mother headed households in each patrol zone, to account for the potential effects that officer perception of danger in the patrol zone, economic marginalization, and neighborhood disadvantage/social disorder may have on the odds of experiencing force.

Overall, the results from this analysis are consistent with those presented in main report, with some important caveats. We find that Black Torontonians are still far more likely to experience force relative to White Torontonians, and that other racialized minorities are less likely to experience force than White Torontonians, even when race-specific rates of contact with the police and race-specific arrests are set as benchmarks. The relative risk that Black Torontonians will experience force estimated in these models, however, is smaller than the relative risk estimated in the main report. Also, we do not estimate a statistically significant difference between risk of force between Black and White Torontonians when arrests for violent crimes are set as the population at risk for force.

We also analyze disparities in use of force events that result in SIU investigations and lower-level use of force separately. Here we find that Black Torontonians are at greater risk of low-level force when race specific rates of contact and arrests are set as benchmarks. That the over-representation of Black people in lower-level use of force incidents persists even when we employ these two, more conservative benchmarks bolsters our confidence that Black people are indeed over-represented in these types of force incidents. Results from models specifically on force events that result in SIU investigations tend to indicate disparity, but disparity that does not reach statistical significance. Overall, the results presented in this addendum report still point to the unjustified and disparate involvement of Black Torontonians in force incidents with the TPS.

 

Police Contacts

Tables B1, B2, and B3 present results where we benchmark use of force for each racial group against the total number of contacts police had with members of that racial group. Such contacts are not always criminal or lead to an arrest – but may serve as a measure for contacts with police. Our estimates of race-specific contacts with the police are generated from TPS Field Information Reports (FIRs), commonly known as street checks, ranging from 2008 to 2013. Race specific contacts after 2013 were unavailable at the time of writing, but contacts over this range may nevertheless serve as a measure for race-specific rates of contact with police from 2014 to 2017 (see discussion in Section A).

As in the main report, we report results from our final model, Model 5, with simultaneously control for several patrol zone characteristics including the violent crime rate (logged), median household income (logged), and the share of single mother households in the patrol zone. Results presented in Table B1 suggest that benchmarked against rates of police contact and controlling for these patrol zone characteristics, Black Torontonians are 1.59 times more likely to experience force resulting in an SIU investigation relative to White Torontonians when police contacts that generate a TPS Field Information Report are set as the benchmark or population at risk of force. Using the same benchmark, other racialized minorities appear to be slightly less likely to experience force, but the results are not statistically significant.

 

 Table B1: Predictors of all use of force cases in Toronto by
race and patrol zone factors, with race-specific contacts with police as benchmark (January 1^st, 2013 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.61 (1.31–1.96)	1.59 (1.29–1.91)	1.58 (1.30–1.92)	1.60 (1.30–1.95)	1.59 (1.30–1.94)
Other racial minority	0.94 (0.74–1.19)	0.94 (0.75–1.17)	0.94 (0.75–1.17)	0.94 (0.75–1.19)	0.94 (0.75–1.17)
Patrol zone factors
Violent crime rate (log)		1.78 (1.29–2.39)			1.60 (1.12–2.26)
Median household income (log)			0.42 (0.19–0.87)		0.53 (0.19–1.53)
% Single mother households				1.00 (0.97–1.03)	0.98 (0.94–1.01)
Note: Negative binomial models of low-level use of force cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1. Data for serious use of force cases ranged from January 1, 2013 to June 30th 2017. Data for lower-level use of force cases ranged from July 1st, 2016, to June 30th, 2017

 

Tables B2 and B3 estimate racial disparities in SIU and lower-level use of force respectively. In Table B2, while we estimate that Black Torontonians are more likely to experience force than White Torontonians when those who have contact with the police are set as the benchmark, the results are not statistically significant. Similarly, and using the same benchmark, we do not estimate a statistically significant difference between the risk that Other racialized minorities will experience force relative to White Torontonians.

 

Table B2: Predictors of SIU cases in Toronto by race and
patrol zone factors, with race-specific contacts with police as benchmark  (January 1^st, 2013 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.26 (0.89–1.74)	1.23 (0.88–1.72)	1.24 (0.88–1.73)	1.28 (0.90–1.75)	1.28 (0.90–1.80)
Other racial minority	0.94 (0.64–1.36)	0.94 (0.64–1.36)	0.93 (0.63–1.35)	0.94 (0.63–1.36)	0.95 (0.65–1.36)
Patrol zone factors
Violent crime rate (log)		1.30 (0.95–1.80)			1.25 (0.86–1.87)
Median household income (log)			0.86 (0.38–1.84)		0.76 (0.23–2.40)
% Single mother households				0.98 (0.95–1.02)	0.97 (0.93–1.02)
Note: Negative binomial models of SIU cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

In contrast, we estimate in Table B3 that Black Torontonians are significantly (1.71 times) more likely to experience lower-level police force when the population who have contact with the police are set as the benchmark. Although we estimate that Other racialized minorities are slightly less likely to experience lower-level force than White Torontonians using the same benchmark, the estimate is not statistically significant.

 

 

Table B3: Predictors of lower-level use of force cases in Toronto by
race and patrol zone factors, with race-specific contacts with police as benchmark (July 1^st, 2016 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.74 (1.38–2.17)	1.71 (1.37–2.14)	1.72 (1.36–2.15)	1.73 (1.37–2.16)	1.71 (1.35–2.19)
Other racial minority	0.92 (0.70–1.18)	0.91 (0.69–1.18)	0.91 (0.69–1.19)	0.91 (0.69–1.18)	0.91 (0.69–1.20)
Patrol zone factors
Violent crime rate (log)		2.08 (1.40–3.10)			1.79 (1.12–2.92)
Median household income (log)			0.28 (0.10–0.78)		0.46 (0.11–1.80)
% Single mother households				1.01 (0.97–1.06)	0.98 (0.94–1.03)
Note: Negative binomial models of low-level use of force cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

All criminal offenses

Tables B4, B5, and B6 show results when use of force is benchmarked against persons arrested for criminal offenses. For use of force cases that result in an SIU investigation, we benchmark against the total number of arrests associated with each racial group in each patrol zone from 2014 to 2017^[8]. Since lower-level use of force cases range from July 2016 to June 2017, we set the benchmark to the average number of arrests associated with each group in each patrol zone in 2016 and 2017. When analyzing disparities in all forms of force, we set the total number of arrests from 2014-2017 as the relevant benchmark.

Table B4 considers all use of force events together. Results from Model 5 suggest that Black Torontonians are about 1.27 times more likely to experience any form of force relative to White Torontonians when persons who are arrested for any criminal offense are set as the benchmark or population at risk for force. We also estimate that Other racialized Torontonians are 50% less likely to experience force than White Torontonians using the same benchmark.

 

Table B4: Predictors of all use of force cases in Toronto by
race and patrol zone factors, with arrests set as the benchmark (January 1^st, 2013 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.28 (1.05–1.54)	1.26 (1.03–1.53)	1.26 (1.04–1.53)	1.27 (1.06–1.53)	1.27 (1.04–1.53)
Other racial minority	0.51 (0.41–0.64)	0.50 (0.40–0.63)	0.50 (0.41–0.63)	0.50 (0.40–0.63)	0.50 (0.40–0.64)
Patrol zone factors
Violent crime rate (log)		1.60 (1.25–2.04)			1.42 (1.07–1.89)
Median household income (log)			0.42 (0.22–0.76)		0.54 (0.23–1.24)
% Single mother households				1.01 (0.98–1.04)	0.99 (0.96–1.02)
Note: Negative binomial models of SIU cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

Tables B5 and B6 analyze SIU and lower-level force independently. The results from Table B4 suggest that Black Torontonians are only slightly more likely than their White counter parts to experience force resulting in an SIU investigation which the arrested population is set as the benchmark, but the result is not statistically significant. We do find, however, that Other racialized minority groups are 48% less likely to experience serious force relative to White Torontonians using this benchmark.

 

Table B5: Predictors of SIU cases in Toronto by race
and patrol zone factors, with arrests set as the benchmark (January 1^st, 2013 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.02 (0.72–1.43)	1.02 (0.72–1.45)	1.02 (0.73–1.41)	1.03 (0.73–1.43)	1.04 (0.75–1.43)
Other racial minority	0.52 (0.36–0.76)	0.52 (0.36–0.76)	0.52 (0.36–0.73)	0.52 (0.36–0.76)	0.52 (0.36–0.75)
Patrol zone factors
Violent crime rate (log)		1.17 (0.87–1.61)			1.11 (0.76–1.63)
Median household income (log)			0.80 (0.36–1.69)		0.69 (0.23–2.09)
% Single mother households				0.99 (0.96–1.02)	0.98 (0.94–1.02)
Note: Negative binomial models of SIU cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

In Table B6, however, we estimate that Black Torontonians are about 1.35 times more likely to experience lower-level use of force relative to White Torontonians which the arrested population is set as the population at risk for force. These results here are statistically significant. Similarly, to the estimate presented in Table B5, we estimate that Other racialized minorities in Toronto are 46% less likely to experience force relative to White Torontonians

 

 

 

 

Table B6: Predictors of lower-level use of force cases in Toronto by

 race and patrol zone factors, with arrests are set as the benchmark

(July 1^st, 2016 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.36 (1.09–1.68)	1.35 (1.07–1.67)	1.35 (1.07–1.69)	1.35 (1.07–1.69)	1.35 (1.08–1.67)
Other racial minority	0.54 (0.41–0.70)	0.54 (0.41–0.70)	0.54 (0.41–0.70)	0.54 (0.41–0.70)	0.54 (0.41–0.69)
Patrol zone factors
Violent crime rate (log)		1.88 (1.33–2.65)			1.62 (1.08–2.45)
Median household income (log)			0.28 (0.11–0.66)		0.46 (0.14–1.51)
% Single mother households				1.02 (0.99–1.06)	1.00 (0.95–1.04)
Note: Negative binomial models of low-level use of force cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

 

Violent Crime Arrest Rates

 

Finally, Tables B7, B8, and B9 present the results of negative binomial models where violent crime arrests are set as the benchmark. As before, when analyzing SIU use of force cases, we benchmark against the total number of arrests for violent crimes associated with each racial group in each patrol zone from 2014 to 2017. When analyzing low level use of force, we use the average number of arrests for violent crimes in 2016 and 2017. When analyzing disparities in all forms of force, we set the total number of criminal offenses from 2014-2017 as the relevant benchmark.

Table B7 presents the results of analysis of disparities in any form of force with arrests for violent crime set as the benchmark. Here, while we estimate that Black Torontonians are slightly more likely to experience any form of force that White Torontonians, the results are statistically insignificant. Other racial minorities are 50% less likely to experience any form of force compared to White Torontonians when arrests for violent crime is set as the benchmark and the results are statistically significant. 

 

 

Table B1: Predictors of SIU cases in Toronto by race

 and patrol zone factors, with arrests for violent crime set as the benchmark

(January 1^st, 2013 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.14 (0.94–1.39)	1.13 (0.93–1.36)	1.13 (0.93–1.37)	1.14 (0.94–1.36)	1.13 (0.93–1.38)
Other racial minority	0.50 (0.40–0.63)	0.50 (0.39–0.62)	0.50 (0.40–0.61)	0.50 (0.40–0.62)	0.50 (0.39–0.61)
Patrol zone factors
Violent crime rate (log)		1.89 (1.44–2.48)			1.68 (1.22–2.31)
Median household income (log)			0.35 (0.18–0.70)		0.50 (0.20–1.36)
% Single mother households				1.01 (0.98–1.04)	0.98 (0.95–1.01)
Note: Negative binomial models of SIU cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

Table B2 presents results specifically for SIU cases and suggests that Black Torontonians are slightly less likely to experience force relative to their White counterparts when arrests for violent offenses are set as the benchmark, but the results are not statistically distinguishable from zero. We do estimate, however, that Other racialized minority groups are about 49% less likely to experience force resulting in an SIU investigation when those force incidents are benchmarked against rates of arrest for violent crime.

 

 

 

 

Table B2: Predictors of SIU cases in Toronto by race

and patrol zone factors, with arrests for violent crime set as the benchmark

(January 1^st, 2013 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	0.91 (0.65–1.27)	0.90 (0.64–1.25)	0.89 (0.64–1.25)	0.92 (0.66–1.28)	0.92 (0.65–1.29)
Other racial minority	0.52 (0.35–0.74)	0.51 (0.35–0.74)	0.51 (0.35–0.74)	0.51 (0.35–0.73)	0.51 (0.35–0.75)
Patrol zone factors
Violent crime rate (log)		1.39 (1.03–1.93)			1.32 (0.91–1.92)
Median household income (log)			0.68 (0.31–1.50)		0.69 (0.22–2.14)
% Single mother households				0.99 (0.95–1.02)	0.97 (0.93–1.02)
Note: Negative binomial models of SIU cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

Finally, Table B3 presents results that analyze low-level force incidents. Again, while we estimate that benchmarked against arrests for violent crimes and controlling for various patrol zone characteristics, Black Torontonians are slightly more likely to experience lower-level use of force relative to White Torontonians, the results of this analysis are not statistically significant. We do estimate, however, that other racialized minorities are about 50% less likely to experience low-level force relative to White Torontonians when such force events are benchmarked against participation in violent crime arrests.

 

 

 

 

 

Table B9: Predictors of low-level use of force cases in Toronto by race and patrol zone factors, with arrests for violent crime set as the benchmark

(July 1^st, 2016 – June 30^th, 2017)

 

	Model 1	Model 2	Model 3	Model 4	Model 5
Race (White set as reference group)
Black	1.24 (0.90–1.71)	1.20 (0.88–1.62)	1.21 (0.89–1.67)	1.23 (0.90–1.72)	1.21 (0.90–1.63)
Other racial minority	0.50 (0.35–0.73)	0.50 (0.36–0.70)	0.50 (0.35–0.70)	0.50 (0.34–0.71)	0.50 (0.35–0.70)
Patrol zone factors
Violent crime rate (log)		2.23 (1.71–2.95)			1.93 (1.41–2.63)
Median household income (log)			0.27 (0.13–0.54)		0.42 (0.17–1.05)
% Single mother households				1.00 (0.98–1.03)	0.98 (0.95–1.01)
Note: Negative binomial models of low-level use of force cases in Toronto patrol zones. 95% credible intervals are given in parentheses.  Effect of race is relative to White reference group. Cell values give effect of a unit change on odds of force. Values in bold are those where 95% credible intervals do not overlap with 1.

 

Conclusion

Our results show that when contacts with the police and arrests are used to estimate the population at risk of police use of force, Black Torontonians remin at an elevated risk of force compared to White Torontonians, while other racialized minorities are at lower risk. We do not, however, estimate any significant racial differences when we narrowly benchmark force incidents against arrests for violent offenses, or when looking specifically at incidents of force that result in SIU investigations.

As anticipated, the size of the Black-White disparity we estimate when we benchmark use of force incidents against arrests and police contact are smaller than when general population is used as the benchmark. Again, this is likely a consequence, at least in part, of racially biased stop and arresting practices by TPS officers. Racial bias in stop and arrest decisions inflates the denominator used to calculate risk ratios and thereby downwardly biases the estimated risk that Black persons will experience force. It may also be a consequence of differential involvement in behaviors and activities that result in police use of force. As in the main report, we also find that use of force is generally more likely where violent crime is higher. Despite controlling for violent crime and other patrol zone factors, however, our overall finding is that racial disparities persist and remain troubling. That is, they point to the unjustified and disparate involvement of Black Torontonians in force incidents that can erode mental and physical wellbeing, police legitimacy, success in school for children, and trust in government.

 

PART C:  Benchmarking “Out-of-Sight” Traffic Offences

Our previous report (Wortley and Jung 2020) documented that, compared to their presence in the general Toronto population, Black people are grossly over-represented in “out-of-sight” traffic offences.  These offences include charges for driving without a license, driving while suspended, driving without insurance, and driving without proper vehicle registration.  These offenses are often labelled as “out-of-sight” offences because, unlike violations for speeding or other illegal driving practices, officers cannot observe these violations from the street or their patrol vehicles.  These violations are only identified once a traffic stop has been initiated.  Scholars maintain that the over-representation of Black people in out-of-sight traffic charges provides additional evidence of racial bias or racial profiling with respect to who the police decide to stop, question, and investigate.  In other words, racial differences in “out-of-sight” driving offences – especially those that do not involve another visible offence (like speeding) – reflect police discretion with respect to surveillance and proactive investigation (Harris 2003; Wortley and Tanner 2003). Consistent with the racial disparities observed in TPS street check data, Black people may be over-represented in “random” traffic stops compared to White people (see Foster and Jacobs 2018). Ultimately, greater exposure to “random” traffic stops is a form of racial bias that increases the likelihood of Black people being identified for an “out-ofsight” driving offence.  Since they are less likely to be stopped by the police to begin with, White drivers are also less likely to be caught for an “out of sight” traffic violation than their Black counterparts.  

Table C1 documents the representation of Black people in “out-of-sight” traffic violations using general population benchmarks.  This data was presented in our earlier report.  The results indicate that, although they represent only 8.8% of Toronto’s population, Black people were identified as the accused in 35.2% of TPS out-of-sight traffic offences documented between 2013 and 2017.  In other words, Black people are four times more likely to be involved in an out-of-sight traffic offence than their presence in the general population would predict.  Furthermore, the out-of-sight charge rate for Black people (1,194 per 100,000) is 4.9 times greater than the rate for White people (244 per 100,000) and 6.9 times greater than the rate for other racial minorities (174 per 100,000).

Scholars,community advocates and police officials have all identified that, when it comes to benchmarking driving activity, general population estimates have limitations and should be supplemented with estimates of the actual driving population.  Thus, to address these concerns, we draw upon data from the 2016 Canadian Census that captures the number of Toronto residents who drive to work using a car, truck, or other personal motor vehicle.  Commute to work estimates may be considered superior to population benchmarks because they better capture the driving population (i.e., those who are of the legal driving age and have access to a motor vehicle).  Commute to work benchmarks may also capture people who drive frequently and are thus at greater risk of police-initiated traffic stops.  However, commute to work benchmarks are not without their limitations.  These figures do not capture, for instance, people who walk or use public transit to commute to work -- but use a car frequently for leisure purposes.  These estimates also do not capture people who have access to motor vehicles but are not currently employed – including retired people, the unemployed, and homemakers.  Commute to work estimates also do not capture young people who may drive to get to high school, college, or university or those who drive often for leisure purposes. Unfortunately, we could not find any alternative benchmarks of Toronto’s driving population that disaggregate by the driver’s racial background.

Table C2 benchmarks “out-of-sight” traffic offences against the Toronto population that commutes to work by motor vehicle.  The results indicate that, using this driving benchmark, Black people become even more over-represented in out-of-sight traffic offences.  For example, although they represent only 6.9% of Torontonians who drive to work, Black people were involved in 35.2% of all out-of-sight traffic offences documented by the TPS between 2013 and 2017.  In other words, Black people are now 5.1 times more likely to be involved in an out-of-sight traffic offence than their presence in the driving population would predict.  Thus, the Odds Ratio documenting Black representation in out-of-sight charges climbs from 4.00 using the general population benchmark to 5.10 using the driving benchmark.  Furthermore, the Black out-of-sight offence rate (7,182 per 100,000) is now 6.8 times greater than the White rate (1,054 per 100,000) and 8.1 times greater than the rate for other racial minorities (889 per 100,000). 

 

TABLE C1: Total Charges for “Out-of-Sight” Driving Offences, by Race of Civilian,

Toronto Police Service, November 5, 2013, to July 31, 2017

(2016 General Population Benchmark)

 

Racial Group	Population Estimate	Percent of Population	Number of Charges	Percent of Charges	Odds Ratio	Charge Rate (per 100,000)
White	1,322,656	48.4	3,230	39.7	0.82	244.2
Black	239,850	8.8	2,864	35.2	4.00	1,194.1
Other Minority	1,169,065	42.8	2,035	25.0	0.58	174.1
TOTAL	2,731,571	100.0	8,129	100.0	1.00	297.6

 

 

TABLE C2: Total Charges for “Out-of-Sight” Driving Offences, by Race of Civilian,

Toronto Police Service, November 5, 2013, to July 31, 2017

(2016 Census Benchmark of Toronto Population that Commutes to Work by Motor Vehicle)

 

Racial Group	Population that Drives to Work	Percent of Population	Number of Charges	Percent of Charges	Odds Ratio	Charge Rate (per 100,000)
White	306,380	53.3	3,230	39.7	0.74	1,054.2
Black	39,875	6.9	2,864	35.2	5.10	7,182.4
Other Minority	229,005	39.8	2,035	25.0	0.63	888.6
TOTAL	575,260	100.0	8,129	100.0	1.00	1,413.1

 

 

Tables C3 benchmarks out-of-sight traffic offences against general population estimates broken down by both race and gender.  Using general population benchmarking, Black males emerge as massively over-represented in out-of-sight traffic offences.  Although they represent only 4.0% of Toronto’s population, they were involved in 30.1% of all out-of-sight traffic offences captured by the TPS between 2013 and 2017.  In other words, Black males are 7.5 times more likely to be involved in an out-of-sight traffic offence than their presence in the general population would predict.  By contrast, using general population estimates, Black women are neither over-represented nor under-represented in out-of-sight traffic charges.  Black women represent 4.8% of Toronto’s population and 5.2% of those charged with an out-of-sight traffic offence (Odds Ratio=1.08).

Table C4 benchmarks out-of-sight traffic offences against Census estimates of Toronto’s driving population broken down by race and gender.  The results indicate that Black males remain grossly over-represented in out-of-sight traffic offences regardless of the benchmarking method used.  Black males are still 7.5 times more likely to be involved in an out-of-sight traffic offence than their presence in the general driving population.  Furthermore, the Black male charge rate (10,596 per 100,000) remains 6.8 times greater than the rate for White males (1,547 per 100,000) and 8.7 times greater than the rate for males from other racial minority groups (1,214 per 100,000).

While the use of the driving benchmark does not change the representation of Black males in out-of-sight traffic offences – it does change the situation for Black women.  As discussed above, when we use the general population benchmark, Black women are not over-represented in these types of offences.  However, when we use the driving benchmark, Black women become significantly over-represented (see Table C4).  Although they represent only 2.9% of Toronto’s driving population, Black women were involved in 5.2% of all out-of-sight traffic offences documented by the TPS between 2013 and 2017.  In other words, Black women are 1.8 times more likely to be involved in an out-of-sight traffic violation than their presence in the general driving population would predict.  Furthermore, the out-of-sight charge rate for Black women (2,498 per 100,000) is now 6.9 times higher than the rate for White women (361 per 100,000) and 8.1 times greater than the rate for women from other racial minority groups (309 per 100,000).  In fact, using the driving population benchmark, the out-of-sight charge rate for Black women (2,498 per 100,000) is now 1.6 times greater than the rate for White males (1,547 per 100,000) and 2.1 times greater than the rate for other minority males (1,214 per 100,000).

In sum, when we use an estimate of Toronto’s driving population as our benchmark, Black people remain grossly over-represented in TPS out-of-sight traffic charges.  In fact, the over-representation of Black people – particularly Black women – increases when we use the driving benchmark as opposed to the general population benchmark.  These findings are consistent with both police statistics and survey data that suggest that Black people are much more likely to be stopped and questioned by TPS officers than people from other racial backgrounds.  Together, these findings strongly support the argument that the TPS has engaged in racial profiling. 

 

TABLE C3: Total Charges for “Out-of-Sight” Driving Offences, by Race and Gender of Civilian,

Toronto Police Service, November 5, 2013, to July 31, 2017

(2016 General Population Benchmark)

 

Racial Group	Population Estimate	Percent of Population	Number of Charges	Percent of Charges	Odds Ratio	Charge Rate (per 100,000)
White male	645,960	23.6	2,766	34.0	1.44	428.2
White female	676,690	24.8	461	5.7	0.23	68.1
Black male	109,870	4.0	2,444	30.1	7.53	2,224.4
Black female	129,980	4.8	420	5.2	1.08	323.1
Other minority male	557,760	20.4	1,781	21.9	1.07	319.3
Other minority female	611,315	22.4	254	3.1	0.14	41.5
TOTAL	2,731,571	100.0	8,126	100.0	1.00	297.5

 

 

TABLE C3: Total Charges for “Out-of-Sight” Driving Offences, by Race and Gender of Civilian,

Toronto Police Service, November 5, 2013, to July 31, 2017

(2016 Census Benchmark of Toronto Population that Commutes to Work by Motor Vehicle)

 

 

Racial Group	Population that Drives to Work	Percent of Population	Number of Charges	Percent of Charges	Odds Ratio	Charge Rate (per 100,000)
White male	178,500	31.0	2,766	34.0	1.10	1,549.6
White female	127,880	22.2	461	5.7	0.26	360.5
Black male	23,065	4.0	2,444	30.1	7.53	10,596.1
Black female	16,810	2.9	420	5.2	1.79	2,498.5
Other minority male	146,705	25.5	1,781	21.9	0.86	1,214.0
Other minority female	82,300	14.3	254	3.1	0.22	308.6
TOTAL	575,260	100.0	8,126	100.0	1.00	1,412.6

 

 

 

 

 

PART D: Benchmarking Failure to Comply Charges

In our earlier report (Wortley and Jung 2020) we explored the representation of Black people in TPS failure to comply charges. Using Toronto’s resident population as a benchmark, we found that Black people were grossly over-represented in failure to comply charges (see Table D1). Although they represent only 8.8% of Toronto’s population, Black people represent 32.7% of those involved in the failure to comply charges documented by the TPS between 2013 and 2017. In other words, Black people are 3.7 times more likely to be charged with a failure to comply offence than their representation in the general population would predict. By contrast, White people and people from other racial minority groups are under-represented.  The failure to comply charge rate for Black people (2,013 per 100,000) is 4.1 times greater than the White rate (493 per 100,000) and 6.9 times greater than the rate for people from other racial minority groups (292 per 100,000).

 

 

Table D1: Total charges for failure to comply offences, by race of civilian,

Toronto Police Service, November 5, 2013, to July 31, 2017

(2016 General Population Benchmark)

 

Racial group	Population estimate	Percent of population	Number of charges	Percent of charges	Odds ratio	Charge rate (per 100,000)
White	1,322,656	48.4	6,514	44.1	0.91	492.5
Black	239,850	8.8	4,828	32.7	3.71	2,012.9
Other minority	1,169,065	42.8	3,417	23.2	0.54	292.3
Total	2,731,571	100.0	14,759	100.0	1.00	540.3

 

 

Although general population benchmarking documents the impact that these types of charges have on the Black community in general, critics may argue that this benchmarking method does not capture the population “at risk” of facing failure to comply offences.  A superior benchmark may be the population that has experienced an arrest during the study period.  After all, one must be arrested before release conditions can be applied.[9]  Thus, in Table D2, we benchmark failure to comply charges against the population experiencing a TPS arrest between 2014 and 2017.  The data suggest that using arrest as opposed to general population benchmarking greatly reduces the over-representation of Black people in failure to comply charges.  Black people represent 24.8% of those arrested by the TPS between 2014 and 2017.  They also represent 32.7% of those charged with a failure to comply offence during this time period (Odds Ratio=1.32).  Thus, using the general population benchmark, Black people are 272% more likely to experience a failure to comply charge.  However, using the arrest benchmark, Black people are only 32% more likely to be charged with this type of offence.  Similarly, using the general population benchmark, the Black failure to comply charge rate is 4.1 times greater than the White rate.  However, when we use the arrest benchmark, the Black charge rate is only 1.3 times greater than the White rate.

In sum, using the arrest benchmark, rather than the general population benchmark, significantly reduces the over-representation of Black people in failure to comply charges.  In fact, using the arrest benchmark, the Odds Ratio for Black people drops below the 1.50 threshold established by this inquiry.  It must be stressed, however, that this reduction in Black over-representation does not eliminate evidence of racial bias.  Indeed, the data still reveal that Black arrestees are 32% more likely to experience a failure to comply charge than their presence within the arrested population.  Furthermore, many steps are involved in the application and enforcement of release conditions.  To begin with, an individual must first be arrested by the police.  Thus, as documented by previous research, if racial bias exists with respect to police surveillance and arrest decisions, this bias will directly contribute to the over-representation of Black people in failure to comply charges (see discussion in Wortley and Jung 2020; Goff et al 2016).  Furthermore, after arrest, the police must decide whether to release an accused person or hold them for a show-cause hearing.  Those held for show-cause hearings are at increased risk of having conditions applied to their release.  Thus, as previous research indicates, if police are more likely to hold Black accused for show-cause hearings, this bias would further contribute to the over-representation of Black people in failure to comply charges (see Kellough and Wortley 2002). Next, during a show cause hearing, accused persons can either be detained in custody or released with or without conditions.  Thus, as previous research exists, if Black accused are more likely to be released with a high number of conditions, this bias would further increase their risk of facing a failure to comply charge (see Kellough and Wortley 2004).  Finally, as demonstrated by previous research, Black accused released to the community are subjected to higher levels of police surveillance than accused from other racial backgrounds.  This type of racial profiling will, once again, contribute to the over-representation of Black people in failure to comply charges.  Clearly, these findings underscore the need to further study – through the collection of race-based data – how racial bias may contribute to decision-making at various stages of the justice system.      

 

 

Table D2: Total charges for failure to comply offences, by race of civilian,

Toronto Police Service, November 5, 2013, to July 31, 2017

(2014 to 2017 TPS Arrest Benchmark)

 

Racial group	Total Arrests	Percent of Arrests	Number of Failure to Comply charges	Percent of Failure to Comply charges	Odds Ratio	Charge rate (per 100,000)
White	46,067	41.8	6,514	44.1	1.06	14,140.3
Black	27,314	24.8	4,828	32.7	1.32	17,675.9
Other minority	36,837	33.4	3,417	23.2	0.69	9,276.0
Total	110,218	100.0	14,759	100.0	1.00	13,390.7

 

 

References

 

Bowling, Ben and Coretta Phillips.  2007.  “Disproportionate and Discriminatory: Reviewing the Evidence on Police Stop and Search.  Modern Law Review 70 (6): 936-961.

Bürkner, Paul-Christian. 2017. “Brms: An R Package for Bayesian Multilevel Models Using Stan.” Journal of Statistical Software 80(1):1–28.

Edwards, Frank, Michael H. Esposito, and Hedwig Lee. 2018. “Risk of Police-Involved Death by Race/Ethnicity and Place, United States, 2012–2018.” American Journal of Public Health 108(9):1241–48.

Foster, Lorne and Jacobs, Les. 2018. Racial Profiling and Human Rights Policy in Canada: The New Legal Landscape. Toronto: Irwin Press.

Geller, Amanda, and Jeffrey Fagan. 2010. “Pot as Pretext: Marijuana, Race, and the New Disorder in New York City Street Policing.” Journal of Empirical Legal Studies 7(4):591–633.

Gelman, Andrew, Jeffrey Fagan, and Alex Kiss. 2007. “An Analysis of the New York City Police Department’s ‘Stop-and-Frisk’ Policy in the Context of Claims of Racial Bias.” Journal of the American Statistical Association 102(479):813–23.

Gelman, Andrew, and Jennifer Hill. 2006. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge: Cambridge University Press.

Goff, Phillip Atiba, Tracey Lloyd, Steven Rapheal, and Jack Glaser. 2016. The Science of Justice: Race, Arrests and Police Use of Force. Los Angeles, CA: Centre for Policing Equity, University of California.

Harris, D. 2003. Profiles in injustice: Why racial profiling cannot work. New York, NY: The New Press.

Jiang, Shuang, Guanghua Xiao, Andrew Y. Koh, Jiwoong Kim, Qiwei Li, and Xiaowei Zhan. 2019. “A Bayesian Zero-Inflated Negative Binomial Regression Model for the Integrative Analysis of Microbiome Data.” Biostatistics (kxz050). doi: 10.1093/biostatistics/kxz050.

Kellough, Gail and Scot Wortley.  2002.  "Remand for Plea: The Impact of Race, Pre-trial Detention and Over-Charging on Plea Bargaining Decisions."  British Journal of Criminology 42 (1): 186-210.

Miller, Joel.  2010. “Stop and Search in England: A Reformed Tactic or Business as Usual?” British Journal of Criminology 50 (5): 954-974.

MVA and J. Miller.  2000.  Profiling Populations Available for Stops and Searches: Police Research Series Paper 131. London, England: Home Office.

Pew, Timo, Richard L. Warr, Grant G. Schultz, and Matthew Heaton. 2020. “Justification for Considering Zero-Inflated Models in Crash Frequency Analysis.” Transportation Research Interdisciplinary Perspectives 8:100249. doi: 10.1016/j.trip.2020.100249.

Riley, James, Davnet Cassidy and Jane Becker.  2009.  Statistics on Race and the Criminal Justice System.  London, England: Ministry of Justice.Ross, Cody T. 2015. “A Multi-Level Bayesian Analysis of Racial Bias in Police Shootings at the County-Level in the United States, 2011–2014.” PloS One 10(11):e0141854.

Toronto Police Service. 2022. Race and Identity-Based Data Collection Strategy: Understanding Use of Force and Strip Searches in 2020. Toronto Police Service and Toronto Police Services Board.

Tregle, Brandon, Justin Nix and Geoffrey Alpert.  2019. “Disparity does not mean bias: Making sense of observed racial disparities in fatal officer-involved shootings with multiple benchmarks.” Journal of Crime and Justice 42 (1): 18-31.Wortley, Scot ,Ayobami Laniyonu and Erick Laming. 2020. Race and the Use of Lethal and Non-Lethal Force by the Toronto Police Service: Final Report. Toronto: Ontario Human Rights Commission. http://www.ohrc.on.ca/sites/default/files/Use%20of%20force%20by%20the%20...

Wortley, Scot and Maria Jung. 2020. Documenting Racial Disparity: An Analysis of Arrest and

Charge Data from the Toronto Police Service. Toronto: Ontario Human Rights Commission. http://www.ohrc.on.ca/sites/default/files/Racial%20Disparity%20in%20Arrests%20and%20Charges%20TPS.pdf#overlay-context=en/disparate-impact-second-interim-report-inquiry-racial-profiling-and-racial-discrimination-black.

Wortley, Scot and Gail Kellough.  2004. “Racializing Risk: Police and Crown Discretion and the Over-representation of Black People in the Ontario Criminal Justice System.”  Pp. 173-205 in Anthony Harriott, Farley Brathwaite and Scot Wortley (Eds.), Crime and Criminal Justice in the Caribbean and Among Caribbean Peoples.  Kingston, Jamaica: Arawak Publications.

Wortley, S., & Tanner, J. 2003. “Data, denials and confusion: The racial profiling debate in Toronto.” Canadian Journal of Criminology and Criminal Justice, 45(3): 367 – 390.

 

Methodological Appendix A

Here we provide additional details on the multi-level Bayesian[10] models run to test for racial disparities in use of force across patrol zones while accounting for patrol zone level characteristics. We provide the following details: a) additional data cleaning that was required to perform the analysis; b) the precise specification of our models; and c) technical details on how the models were fit.

 

Data Cleaning and Manipulation

Geocoding Use of Force Incidents – GIS software was required to geolocate some lower-level use of force cases and a lack of geographic data on some SIU cases meant that those cases were dropped from the analysis. Of the 198 SIU cases in the dataset, 9 cases (4.5%) did not contain information on the patrol zone in which they occurred or did not occur in the City of Toronto and were dropped from the analysis.

Of the 578 lower-level use of force cases analyzed 3 (.3%) incidents were marked as occurring outside the City of Toronto and were dropped from the analysis. Officers did not mark the patrol zone in which incidents took place in 26 incidents. In 13 of these incidents, however, they did note the XY coordinates where the incident occurred. GIS software was used to successfully geolocate these incidents to the patrol zone where they occurred. The remaining 13 incidents were dropped from the analysis.

In S1, we characterize the racial composition of the 26 total cases that were not geolocated.

 

S1: Cases dropped due to a lack of geo-identifiers
Race	Total	% Of Cases
Minor Use of Force	578	100%
Black	1	0.2%
White	8	1.4%
Other	4	0.7%
Outside Toronto	3	0.5%
SIU Cases	198	100%
Black	4	2.0%
White	4	2.0%
Other	1	0.5%
All Use of Force	25	3.2%

 

 

Model Specification

As we describe in the main text, we fit multi-level negative binomial models to evaluate whether racial disparities in police use of force persist after accounting for precinct-level characteristics and estimate them in a Bayesian framework. We do so to simultaneously account for the overdispersion in counts of use of force cases and the grouping of use of force cases into patrol zones.[11] We estimate the total number of use of force cases for each racial group across patrol zones using the following model: 

where y_ij is the total number of force cases experienced by members of racial group i in patrol zone j,  pop_ij is the race-specific population at risk of force, a_i  is an indicator variable for each racial group, β_j  is a vector that stores patrol-zone specific variables (including the log of violent crime rate, the log of the median household income, and share of households headed by single mothers), β_p controls for unmeasured patrol zone level variation, and the parameter ϕ controls the shape of the negative binomial distribution and is estimated from the data (c.f. Gelman and Hill 2006; Jiang et al. 2019; Pew et al. 2020). Following Bürkner (2017), we set noninformative priors on u, a_i, and βj. 

 

Fitting the Model

We fitted the model using Stan  in R suing the brms package (Bürkner 2017). Hamiltonian Monte Carlo (MCMC) sampling was performed on four chains, each with 1,000 warm-up draws and 2,000 sampling draws, resulting in 8,000 draws from the posterior total. Trace plots were used to confirm model convergence.