OpenData: Analyzing Motor Vehicle Theft Trends in Toronto
The Toronto Police Service is committed to the ongoing release of open data for public safety, awareness, greater openness, and transparency. The Service’s Open Data Program strives to release valuable open data and provide continuous support for public understanding, use, and application of police information.
Background
Motor vehicle theft (MVT) is the criminal act of stealing or attempting to steal a motor vehicle. According to Stats Canada, in 2021 more than eighty-three thousand incidents were reported in Canada, of which 11.4% were cleared with charges laid or otherwise.
This dataset is published to OpenData Toronto by the Toronto Police Service (TPS) and contains more than a decade of reported incidents. To learn more about the dataset click this link. All tables and visualizations were created using MS Excel (16.67) and Tableau Public (2022.3.0).
The goal of this analysis is to describe and visualize the current landscape of the data. Specifically, to identify trends that are informative to citizens that own or lease a vehicle.
Data Cleaning
Data cleaning steps were employed as described in the summary below:
The dataset appears to have been collected systematically and in a manner that resulted in standardized data collection and formatting practices. As a relatively clean dataset, minimal efforts were necessary to prepare the data for analysis.
76,195 unique records are contained within the dataset, of which 7,112 (9.3%) were determined as missing or incomplete. The majority of blank values belong to fields named occuranceHour and reportingHour, 83% and 16%, respectively. All reported events (rows) with complete or partially complete data were included in the analysis.
Upon reviewing the dataset shows minimal reports between the years 2000-2013 and 2021. Therefore, the aforementioned years are excluded from between-year comparisons.
Exploratory Analysis
MVT is 2.5x more likely to occur on weekdays vs. weekends. Weekends are defined as dates that fall on Saturday and Sunday.
| Day | Total | % |
|---|---|---|
| Weekday | 46,822 | 71.4 |
| Weekend | 18,713 | 28.6 |
Vehicles parked outside were most frequently stolen followed by house, and other. Each premise category is defined under the data field location type. Click here to learn more about the definitions applied to premise categories listed in the table below.
The Toronto Police recommend reporting MVT immediately. Most MVT are reported within 48 hrs however a significant proportion are submitted well after the incident occurred. Note, for the histogram below, the y-axis is log transformed to improve visibility of the long thin tail of the distribution.
MVT is most frequent in the evening and/or throughout nocturnal sleeping hours as compared to events occurring during daytime or daylight hours. A sharp rise in reporting is evident within a few hours of waking. Note to reduce the risk of recall bias the results were filtered to exclude incidents reported >1 day after the event was believed to have occurred.
Divisional Trends:
There are 17 divisions in the TPS each is responsible for providing services to communities within a defined geographical region.
The charts below show conditionally formatted to highlight the percent change in MVT relative to the prior year within and across divisions. 2014-2015 and 2015-2016 saw nearly all divisions with YoY decline in MVT. Since then, MVT has increased YoY (as evidenced by the sea of red cells left-to-right), however, the rate of rise in YoY change trended lower each year thereafter.
Seasonality
Month-of-year effects were observed between 2014 and 2020. The red dotted line, grey-shaded area, and grey dotted line indicate the average, average with ±1 STDDEV, and a linear trend model applied to each data pane, respectively.
MVT was the least frequent in 2016 and peaked in 2020. 5 of 7 years show steep increases in MVT throughout the calendar year with incidents/month reaching yearly highs in either Oct or Nov followed by a significant drop-off throughout the colder months. The month with the fewest MVT trend towards Feb.
Ranking Neighbourhoods Most Impacted By MVT
In 2020 Tableau introduced "Bar Chart Races" to visualize data over time in a ranked format. Click here to view an animation of the top 15 neighborhoods most impacted by MVT.
Summary and Recommendations for Further Analysis
Although it's difficult to tease out actionable insights or practical recommendations, from my point of view, a key takeaway is that MVT is far more common than I thought. The data indicates that no neighborhood is safe however some areas are harder hit than others. Where possible exercise the use of precautionary measures such as the tips provided by the as listed on the website.
Recommendations and questions for those interested to expand beyond the scope of this analysis:
Are vehicles parked closer to the downtown core more or less likely to be stolen than those parked further away? Calculate the distance between geocodes relative to a fixed point (e.g. City Hall).
How likely are successive MVT events to take place within 500m of the prior incident occurring in a short period of time (e.g. 1-2 wks)?
What types of weather conditions (e.g. snow, rain, or fog) deter or encourage unlawful activities such as MVT?
If external insurance data can be sourced, analyze trends in fire and theft insurance premiums over time to observe a correlation and predict future rates.
To help make this data more personally relevant, the TPS could include in the reported data: the make, model, color, year, and even price of the vehicle (new). On a related note, CP24 (news outlet) reported Ontario's most stolen vehicles in Ontario. Having access to more detailed reporting may help inform prospective buyers which vehicle type and model to buy