Toronto break-in atlas
Six analyses of every break-in reported to Toronto Police in 2024 and 2025: which neighbourhoods stand out once size and chance are accounted for, where hot spots formed and faded, whether one break-in predicts the next, and how each area has changed since 2014. Everything is recalculated in your browser from the police service's open data each time the page loads.
Calculated when the data arrives.
Which neighbourhoods really stand out
Each dot is a neighbourhood: its 2025 home break-in rate against how many households it has. Small neighbourhoods swing a lot by chance, so the funnel lines show how far a rate can stray from the city average before it is unlikely to be luck. Dots outside the funnel are the ones that genuinely differ.
A fairer ranking for small neighbourhoods
A neighbourhood with 2,000 households can top a raw ranking because of a handful of break-ins. Stable rates pull each neighbourhood's rate toward the city average in proportion to how little data it has: large neighbourhoods barely move, small ones move more. Public health agencies rank small areas this way.
Calculating…
| Rank | Neighbourhood | Households | Break-ins | Raw rate | Raw → stable | Stable rate | Move |
|---|
Hot spots, new and fading
The city is cut into 250 m squares. A square is a hot spot when it and its neighbours had far more break-ins than the city pattern would produce by chance (Getis-Ord Gi*, with the false discovery rate held at 5%). Comparing 2024 with 2025 shows which hot spots persisted, which are new and which faded.
Does one break-in lead to another nearby?
A Knox test compares how many pairs of home break-ins happened close together in both place and time with how many you would expect if the dates were shuffled at random. A ratio of 1.5 means 50% more pairs than chance.
Highlighted cells are significant at the 5% level. Locations are offset to the nearest intersection, so “same location” means the same intersection, not the same home. Because the reshuffles are random, ratios can move slightly each time the page loads.
How each neighbourhood has changed since 2014
A twelve-year trend for every neighbourhood, fitted as a yearly percentage change and tested for whether it is more than year-to-year noise (allowing for extra variation, with the false discovery rate held at 5% across all 158).
| Neighbourhood | 2014–2025 | 2014 | 2025 | Per year | Verdict |
|---|
Trend counts are police records of home break-ins. One incident can produce more than one record, so these can be slightly higher than the incident counts on neighbourhood pages.
When break-ins happen, and how long until they are reported
2024 and 2025 combined. Brighter squares are busier hours. Times are often estimates, because people find out when they get home.
Time from break-in to police report
Download the results
Both files are generated from this page's live calculation and include the source and retrieval date. Grid squares carry counts only, no addresses.
Method and limits
- Source. Toronto Police Service, Major Crime Indicators open data, Open Government Licence – Ontario. Break and enter covers forced and attempted entry and being unlawfully in a dwelling. Each incident is counted once, by the date it occurred. Homes are houses and apartments; businesses are commercial premises.
- Rates. Households are from the 2021 Census. The city average is incidents with a known neighbourhood divided by all households. Rate intervals use Byar's approximation; funnel limits use exact Poisson percentiles.
- Stable rates. Global empirical Bayes smoothing (method of moments, Marshall 1991): each rate is a weighted average of the neighbourhood's own rate and the city rate, with more weight on its own rate the more households it has.
- Hot spots. Getis-Ord Gi* on 250 m squares with a 3 × 3 neighbourhood, over squares within one square of any recorded incident, with Benjamini–Hochberg correction.
- Near repeats. Knox test on home break-ins with a known location, 2024–2025, against 99 random permutations of dates.
- Trends. Poisson log-linear regression on yearly records, 2014–2025, with a quasi-Poisson correction for overdispersion and Benjamini–Hochberg correction across neighbourhoods.
- Limits. Only reported crime is counted. Locations are offset to the nearest intersection by police for privacy, and the data is preliminary and can be revised, so results can shift slightly over time. Police advise against comparing these figures with other crime data. Patterns describe places, not people, and are not a prediction for any home or street.
All calculations are by MkaeSense. MkaeSense is not affiliated with Toronto Police Service.