Blog article: Are Diverse Neighbourhoods Safer? We Checked the Data
Article text
This blog post is the result of the 2025 York University Blogathon, in which York University students were invited to explore their city and community in more detail using data from Toronto’s Open Data Portal.
We are two international students who came to Canada to study Computer and Data Science and quickly fell in love with Toronto. We love seeing people from all over the world connect and we love being part of these connections. Soon after we arrived, we started thinking more deeply about why we feel comfortable and safe interacting across cultures and what safety means to us.
Our reflections tell us that safety is multidimensional. Safety is not only the absence of physical harm but also a sense of psychological ease, social belonging, and trust in institutions. Safety for us means being able to exist in public spaces without shrinking our identity or worrying about how we will be perceived or if we will be targeted. We have felt most safe in two situations: when surrounded by people from our communities from our home countries and in neighbourhoods where many cultures overlap and share the same space. In these environments, diversity reduces visibility as a point of difference, making us feel less likely to be singled out.
But people do get targeted in Toronto for their identities, which is reflected in recent crime and incident statistics (see Figure 1). Since the COVID-19 pandemic, Toronto’s visible minority community has reported greater feelings of insecurity due to crime than the rest of the population (Statistics Canada, 2020). Hate crimes in the city, in particular, are on the rise (Royal Canadian Mounted Police, 2025). According to the Toronto Police Service (TPS), these crimes are offences committed against a person or property that are motivated in whole or in part by bias, prejudice or hate. The TPS codes these crimes by motivation; this can be related to race, national or ethnic origin, language, colour, religion, sex, age, mental or physical disability, sexual orientation, or gender. Hate crimes can include acts of vandalism, harassment and even assault.
Figure 1.Police-reported hate crime incidents in Canada, 2009 to 2024 (RCMP, 2025). Note that an incident report may or may not result in a criminal charge.
Given this data, we have wondered whether the sense of safety we feel in diverse Toronto neighbourhoods is subjective or whether our feelings are reflected in empirical evidence. Theorists have, in fact, posited a relationship between diversity and hate crime. One theory is called the power-differential hypothesis; this suggests that as minority groups in a neighbourhood grow, they become more able to resist hate-motivated targeting (Piatkowska et al, 2019) . The idea behind the theory is that as minority group sizes increase, members of dominant groups become less likely to act on bias, which results in less hate-motivated crime (Green, Strolovitch, & Wong, 1998).
We thought it would be interesting to test this hypothesis using data about hate crimes and diversity from Toronto. So, we looked at publicly available datasets from the city of Toronto’s Open Data Portal and set out to examine if our perception of safety aligns with measurable patterns in this data.
Our research question: Does Toronto’s neighbourhood diversity inform the rate of race-based hate crimes?
1. Data: Measuring Diversity and Hate Crime
We examined two different datasets published on Toronto’s Open Data Portal: the Hate Crimes Dataset and Toronto’s Neighbourhood Profiles. We combined these two datasets to help us relate diversity and the rate of hate crime for each of Toronto’s 158 neighbourhoods. The details of our datasets and the measures we used follow.
The Hate Crime Dataset
To measure the incidence of hate crimes in Toronto, we explored the Toronto Police Hate Crimes Dataset. This contains information about 1805 reported hate crimes that took place in Toronto between 2014 and 2024. Each row represents a single hate crime incident, including when it occurred, in which Toronto neighbourhood, what type of offence was committed, and what bias motive was involved. The bias motives attached to each report were coded by police and include age, mental or physical disability, race, ethnicity, language, religion, sexual orientation, gender and other. A graph that illustrates counts of some of these crimes in recent years can be found in Figure 2.
The Neighbourhood Profiles Dataset
To measure the diversity of Toronto neighbourhoods, we explored the Neighbourhood Profiles Dataset. This dataset provides a detailed demographic snapshot of Toronto’s 158 officially defined neighbourhoods and is based on data from the Census of Population, which is conducted every five years across Canada by Statistics Canada. The data we explored was drawn from the 2021 census, in particular. Conveniently enough, the 158 neighbourhood identifiers in this dataset are identical to the neighbourhood identifiers in the Hate Crime Dataset, which made it possible for us to relate the two datasets to one another.
Each Neighbourhood Profile includes aggregated information about the ethnicity, religion, language, education, income, and housing status of neighbourhood residents. For our project, we used the following features from the 2021 census to define a diversity measure:
- The Total Population: which is a count of all individuals in a given neighbourhood.
- The Visible Minority Population: which is a count of the number of individuals who identified as a belonging to a visible minority group as defined by the Employment Equity Act. This includes everyone who responded on the census that they were a member of one or more of the following population groups: South Asian, Chinese, Black, Filipino, Latin American, Arab, Southeast Asian, West Asian, Korean or Japanese.
2. Method of Data Analysis
Our analysis related measures of diversity to the rate of racially motivated hate crimes in each Toronto neighbourhood.
We defined the diversity of each neighbourhood as the visible minority count in that neighbourhood divided by the total neighbourhood population.
We calculated the rate of racially motivated hate crimes in each neighbourhood as the aggregate number of racially motivated hate crimes in that neighbourhood between 2014 and 2024 divided by the neighbourhood population in the neighbourhood profile and expressed per 10,000 residents.This aligns with a similar analysis from Piatkowska and colleagues, whose work revealed a relationship between race-based hate crimes and measures of diversity across different counties in the United States (Piatkowska et al. 2019).
Outlier Detecting
To ensure that our analysis was not biased by extreme values, we applied the Cook’s Distance method to identify and remove outlier neighbourhoods based on the hate crime rate per 10,000 residents. We focused on this variable because it is the dependent variable in our regression.
Cook’s Distance is a measure used in regression analysis to estimate the influence of a data point when performing a least squares regression analysis. Data points with a large Cook’s Distance value are considered to have a high influence on the regression model. A common rule of thumb for identifying influential points is when Cook’s Distance is greater than 4/N, where N is the number of observations (which is the same as number of Toronto neighborhoods in our case). We applied this rule to identify outliers, and the process resulted in the removal of the following 4 of Toronto’s 158 neighbourhoods prior to subsequent analysis: Yonge–Bay Corridor, Moss Park, and Downtown Yonge East, and Danforth. These neighbourhoods exhibited higher concentrations of racial hate crimes relative to other areas, which likely reflects both their exceptionally high population density and their relative concentration of major institutional and commercial services (Statistics Canada, 2022).
Racially motivated hate crime counts were aggregated across the full 2014–2024 period to produce a single value for the hate crime rate of each of the remaining 154 neighbourhoods that were not considered to be outliers.
Relating Crime Rate to Diversity
To examine the relationship between neighbourhood diversity and rate of hate crimes in each neighbourhood, we conducted an Ordinary Least Squares (OLS) Regression using our measure of the rate of hate crimes (i.e. the aggregated racially motivated hate crime rate per 10,000 residents) as the dependent variable and our combined diversity measure (i.e. the average of the visible minority share in each neighbourhood) as the independent variable.
3. Results
Figure 3. At left, a heat map illustrating our diversity measure – the visible minority share in Toronto neighbourhoods, based on the 2021 census. At right, a heat map illustrating the racially motivated hate crime rate per 10,000 residents in Toronto neighbourhoods between 2014 and 2024, calculated from the Toronto Hate Crime Dataset and adjusted for neighbourhood population.
Figure 4. The hate crime rate (crimes per 10,000 residents) by the percentage of visible minorities across the 154 Toronto neighbourhoods retained after outlier removal.
Our findings reveal a significant negative relationship between neighbourhood diversity and racially motivated hate crime rates. As shown in Figure 4, neighbourhoods with a higher visible minority share tended to experience fewer reported race-based hate crimes per 10,000 residents.
- The estimated coefficient relating percent of visible minority population to race-related hate crime is −0.3186, indicating that, on average, a 10‑percentage‑point increase in a neighbourhood’s visible minority population share is associated with a decrease of approximately 3.2 race‑based hate crime incidents per 10,000 residents, holding other factors constant.
- The F‑statistic (p = 0.0232) indicates that the regression model is statistically significant at the 5% level.
- The model explains a modest proportion of the overall variation in hate crime rates (R² = 0.033).
Overall, the regression model indicates a meaningful and non‑random negative relationship between visible minority concentration and race-motivated hate crime incidence. In other words, neighbourhoods in Toronto with higher levels of visible minorities tended to exhibit lower rates of hate crimes that were motivated by race.
4. Discussion and Conclusion
Our analysis suggests that diversity plays a role in neighbourhood safety, with more diverse areas associated with a lower rate of racially motivated hate crime. This relationship may mean residents of diverse neighbourhoods feel safer because they are, in fact, safer from hate crime.
Our findings are echoed by others who research hate crimes as well. For example, Piatkowska and colleagues found that the size of the Black population in any given US county has a negative and significant effect on anti-Black crime in that area. These same authors found that hate crimes occur more frequently not only in less racially diverse counties, but in younger and more affluent ones. This gestures to the idea that diversity, like safety, is multidimensional; socio-economic and generational diversity, in addition to racial or ethnic diversity, may help communities resist bias-motivated crime.
Hate crimes in Toronto are, however, still rising, and their impacts are serious, shaping how people experience their daily lives and highlighting ongoing inequities. Support and reporting options remain available through Toronto Police Service, Ontario Human Rights Commission, and community-based organizations such as the Canadian Race Relations Foundation. Overall, while diversity can strengthen both safety and the feeling of safety, continued attention and community resources are needed to address the increase in hate-motivated incidents.
Limitations
This analysis has several limitations that should be considered when interpreting the results. First, hate crimes are widely underreported, with estimates suggesting that 60% to 90% of incidents are never reported to police (RCMP, 2025). This is especially relevant for migrants and racialized communities, who may be less likely to report due to fear or mistrust. As Brianna Garneau (2024) notes, “political, media and public discourses have fused images of migrants with those of criminals, destining the racialized non-citizen to be perpetually identified as a source of potential risk.” The TPS reports of hate crime may therefore under-represent true rates of hate crime for this reason.
In addition, police data may include inconsistencies or subjective interpretations. For example, as Timothy Bryan reported in his doctoral dissertation, police may rely on pre-existing narratives about what a hate crime looks like to code hate related incidents and/or they may lack training to correctly code hate crimes that are subtle. In addition, because reporting hate crime incidents require administrative work above and beyond reporting of other crimes, there are organizational dis-incentives to reporting crimes in this way. Police may also have a pre-existing bias to see crimes as related to ignorant individuals rather than social structures, again meaning that some hate crimes may not receive the “hate crime” label. All of these factors also contribute to the potential under-reporting of hate crimes in the TPS database.
There is also a mismatch in time, as our crime data spans 2014–2024 while our diversity data is based on the 2021 census. Aggregating data at the neighbourhood level may also have hidden some differences within smaller areas or groups. Importantly, our findings are correlational and do not imply causation, as other factors like income, education, policing, and population density may influence the results. We also assumed a linear relationship between diversity and hate crime, which may be overly simple. Finally, even after removing outliers, our regression model explains only about 10% of the variation in racially motivated hate crime, meaning many other factors may contribute to neighbourhood safety from these kind of crimes.
Finally, because we removed neighbourhoods with unusually high hate crime rates, our results describe the relationship between diversity and hate crime in most but not all Toronto neighbourhoods. The downtown and commercial-hub neighbourhoods that were excluded are also places where many Torontonians live, work, and visit, and the hate crime experiences of those who live and work in these areas warrant a separate analysis.
5. Acknowledgements
We owe a tremendous debt of gratitude to Professor Sonya Allin of York University’s Lassonde School of Engineering, who guided this work from its first tentative steps to its completion. Her expertise, generosity of time and genuine care for our development shaped every part of what you are reading. We are better thinkers and collaborators because of her.
We are also grateful to Librarian Alexandra Wong, and members of the York Data Science Club for organizing the 2025 York University Blogathon event and helping to develop this submission into a blog post.
We would like to acknowledge Brianna Garneau and Bahareh Banaei from York University’s Collaborative For Racial Justice (CRJ) for discussing their work on racialization, state violence, and the experiences of marginalized communities with us; this helped shape how we think about safety beyond just crime statistics.
In addition, we would like to recognize Timothy Bryan’s research on Race, Diversity, and the Politics of Hate Crime, which examines how police respond to racially motivated hate crimes in the Greater Toronto Area; this provides important context for understanding how hate crimes are reported and addressed in our city.
We also want to recognize the contributions of Manas Tripathi who was involved in the early stages of this research and who encouraged us to explore safety in Toronto neighbourhoods, specifically.
Finally, a huge acknowledgment to Toronto’s policymakers, educators, and community advocates for their ongoing commitment to address our city’s hate crimes. Their work to strengthen awareness, improve reporting systems, and advance inclusive education will hopefully mean more people can come enjoy, feel welcome and safe in Toronto, much as we have.