Data Sources & Methodology

1. Our data principles

• Official only. Every crime statistic in StudySafeLink comes from a police force or city government open-data platform. We never estimate, model or "fill in" numbers.
• Honest about gaps. If an area or university has no suitable official data, we say so — you'll see "no official data", never an invented score.
• Traceable. Each university page shows which sources cover it and the exact data month behind every figure on this site and in the App.

2. Current sources

We currently ingest 53 official sources:

Coverage notes: UK street-level data covers England, Wales and Northern Ireland (Police Scotland publishes no comparable open data). US and Canadian sources are city-level; Australian point-level data covers Queensland cities via Queensland Police. Regions without point-level data (Sydney/Melbourne via NSW BOCSAR and VIC CSA, Canadian census metropolitan areas via Statistics Canada, other US jurisdictions via FBI UCR) are scored from area-level official counts in a separate pool. Universities outside any covered area have no local source yet.

3. From raw records to three severity levels

Each publisher uses its own category system (e.g. "Violence and sexual offences" vs "BURGLARY-RESIDENCE"). To compare countries fairly we normalise every record into one of three levels at import time, using transparent per-source rules that are also viewable inside the App:

• Violent — violence against the person, robbery, sexual offences…
• Property — burglary, theft, vehicle crime, arson, criminal damage…
• Minor — anti-social behaviour, minor disorder, other non-serious categories…

4. How safety scores are computed

1. Take all officially recorded crimes within 3 km of the university's campus location.
2. Restrict to the 12 most recent months actually published by the covering sources (a rolling window).
3. Compute a density indicator weighting violent incidents more heavily than property/minor ones.
4. For point-level data, rank all scored universities within the same country and convert to a percentile-based 0–100 score and A–D grade. These scores are relative to each country's pool, so a UK score and a US score are both "relative among peers".
5. For area-level data, score on a fixed absolute scale defined once per data source: green = below that source's typical (median) area, red = above twice it, with three shades inside each band. Because the scale is absolute, a low-crime region can stay green throughout and is never forced to contain a red area. A–D grades correspond to the four colour bands.
6. Recompute weekly when publishers release new months; pages on this website regenerate weekly from the same database the App reads.

A university is only scored when its sample is large enough to be meaningful (20+ recorded crimes and sufficient month coverage). Otherwise it stays unscored and shows raw counts instead.

Where no point-level data is published, universities are scored from area-level official counts (the whole administrative area: local government area in Australia, census metropolitan area in Canada, police agency jurisdiction in the US), ranked by weighted incidents per 100,000 residents in a separate pool that is never mixed with point-level rankings. Area scores use a fixed absolute scale per data source and are not comparable across sources or countries; areas with very small populations are excluded because their rates are statistically unreliable.

5. Known limitations

• Recorded ≠ total. Official statistics under-report crime; comparisons between very different neighbourhoods should always be read with care.
• Publisher revisions. Recent months can be revised upwards after initial publication.
• Different scopes. UK data is national but offence-level; US, Canadian and Australian point-level data is city-limited; area-level sources cover whole administrative areas on monthly, quarterly or annual cycles. We therefore rank within countries (and within data-type pools) only; area-level scores follow a fixed absolute scale per data source and are not comparable across sources or countries.
• Not real-time. This is monthly aggregate data for orientation — it cannot tell you what is happening right now.