Why crime maps disagree

Look up the same neighborhood on three different crime websites and you will often get three different answers. Not slightly different. One will call it safer than average, another will call it dangerous, and the third may have no data at all.

They are not all wrong. They are mostly measuring different things, on top of a national dataset with large holes in it that almost nobody mentions.

Reporting crime data is voluntary

This is the root of it, and it surprises most people.

There is no national crime database that police departments are required to fill in. The FBI collects and publishes crime data, but departments submit it voluntarily. There are around 18,000 law enforcement agencies in the United States and they participate to wildly differing degrees.

Most take part. Some never have. Others send partial counts, or report for a few years and then stop, or change systems and produce a gap while they migrate. None of this is visible in the finished numbers unless someone goes looking.

A recent change made the gaps worse

The FBI spent years moving from a simple summary count to a more detailed incident-level system. The new system is better. It records far more about each incident.

It also costs money and effort to adopt, and when the older system was retired, a large number of agencies had not made the switch. Some big-city departments were missing from national figures entirely for a period. Coverage has been recovering since, but the dip is still sitting in the historical data, and anything that averages across those years inherits it.

Why this produces a map that lies quietly

Here is the failure that matters for anyone house hunting.

If a city reports little or nothing, and a site draws it anyway, the city comes out looking calm. Low reported crime and low actual crime produce the same clean map. The map cannot tell you which one you are looking at.

That is worse than an empty map. An empty map makes you go and ask. A confident wrong map makes you stop asking.

The other reason two maps disagree

Even with identical data, sites make different choices, and the choices move the answer a lot.

• What counts as crime. Violent only? Property too? Include minor offences and the ranking reshuffles.
• What you divide by. Residents, or people actually present? A business district with few residents and a large daytime population looks alarming per resident and unremarkable per person on the street.
• How big an area. City-wide averages hide everything interesting. Cities routinely contain both the safest and the least safe areas in their region.
• Which years. One year is noisy in a small area. Several years is stable but slower to notice a real change.

None of these is the single correct choice. They are trade-offs. The problem is that most sites do not tell you which ones they made.

The fair counter-argument

You could reasonably say: incomplete data is still better than nothing, and leaving areas blank just makes the product less useful.

There is something to that. Someone comparing two neighborhoods wants an answer, and “we don’t know” is an unsatisfying one. A partial count does carry some signal.

The reason we come down the other way is what the mistake costs. Choosing where to live is expensive, slow, and hard to reverse. If someone rules out a good area, or moves to one they misjudged, they carry that for years. In that situation a confident wrong answer does more damage than an honest gap, because an honest gap prompts a phone call and a wrong answer ends the conversation.

What we do about it

We compare what each agency reports against the FBI’s published figures for the same area. Where the difference is too large for the numbers to carry meaning, we leave that area blank rather than drawing it.

Blank means we do not know. It does not mean nothing happens there. We would rather show you a hole than a comfortable fiction.

The full detail of what we measure and what it cannot tell you is on our crime methodology page.

How to read any crime map, including ours

• Check whether it says where its data came from. Many do not.
• Check whether it distinguishes “low crime” from “no data.”
• Be suspicious of a single letter grade for a whole city.
• Compare areas within one metro rather than across the country. Reporting is more consistent locally.
• For anywhere you are serious about, ask someone who lives there. Data is a filter, not a verdict.

You can see the areas we do and do not have solid data for on the crime map. The gaps are visible on purpose.