Safety, with the working shown.

Why per-capita, and why person-crime counts double

Every scored listing's band is weighted incidents per 1,000 residents nearby — not a raw count. A dense area racks up more total incidents than a quiet one simply by having more people and more potential targets present; dividing by the population living in the same 0.8 km radius is the only way two areas of different density are comparable at all. And a purse-snatching and an aggravated assault aren't the same thing: crimes against a person count 2× toward the score, property crimes once. The raw and person-crime counts are both kept and shown on every listing, so the weighting is auditable, not a black box.

Neither choice is an invention here — it's the same shape of method used by the crime-data products people already check before signing a lease. Researched and cross-checked against six of them before building this one:

• NeighborhoodScout — crime index built on crime rate per 1,000 population
• CrimeGrade.org — FBI UCR/NIBRS data, ML-weighted by impact, census-block-group granularity
• SpotCrime — incident-level mapping with severity-aware summaries
• DoorProfit — rental-focused neighborhood safety scoring
• AreaVibes — crime rates normalized per 100,000 residents, violent crime weighted higher
• Niche.com — FBI UCR data on a standardized x/100,000-residents scale

All six converge on the same shape: normalize by population, weight by severity. None of them score off a raw incident count, and neither does this site. Those products are themselves built on the same kind of local-agency open data used here — the fix was adopting their calculation, not their data.

The population denominator

Resident counts come from the U.S. Census Bureau's TIGERweb Census-2020 block layer — population carried directly on the block geometry, no API key, no estimation step. A listing whose radius holds fewer than 300 residents (a park, an industrial strip, a stretch of coastline) shows “No data” rather than a wildly unstable rate built on a tiny denominator.

Why LAPD listings show no trend arrow

A trend needs two comparable 12-month windows. LAPD switched its records system to NIBRS on March 7, 2024, and its open datasets show a multi-month reporting ramp as the switch was backfilled — comparing any two windows that touch that ramp would read as a citywide crime spike that is really just paperwork catching up. So LAPD listings get one clean trailing-12-month snapshot and no trend, until the feed has a full trailing year entirely past its own ramp.

Why San Diego ships without a trend

Same discipline, different reason: SDPD moved to NIBRS in 2021 and states that its incident records may change or be deleted as investigations proceed. San Diego listings are scored exactly like LAPD's — one trailing-12-month snapshot, same radius, same weighting, same quartile cuts — and the trend stays blank rather than resting on records the department itself calls provisional. Only geocode-verified incident rows are used.

Why Orange County has no block tier at all

The OC Sheriff publishes no open incident-level data — only a proprietary third-party map embed with no API. Orange County listings therefore carry no block-level band; where their city has clean CA DOJ counts they show the citywide tier, clearly labeled, and otherwise they show no safety number at all. No number is scraped out of a widget, and no gap is filled with a guess — that rule holds everywhere on this site.