Crime statistics have a credibility problem. In many countries the official numbers fall year after year, while polls keep finding that people think crime is rising.
The usual explanations are media panic, politics or poor numeracy. A new working paper by economists Elisa Facchetti, Imran Rasul and Joost Sijthoff offers a plainer one: people judge crime by what they see, and the statistics mostly count something else.
The study counted witnesses as well as victims
Official crime figures are built around victims: who was burgled, robbed or assaulted.
The authors used the Crime Survey for England and Wales from 2001 to 2018. Besides asking whether people were victims, the survey asks whether they saw a crime happen. That let the authors track two numbers side by side for 17 years:
- how many people were victims of crime
- how many people witnessed crime
Victims fell by a third while witnesses held steady
- Victimisation fell by 32%. By the end of the period, roughly one in three fewer people were victims than at the start.
- The share of people witnessing crime stayed roughly flat.
- Witnessing crime was about twice as common as being a victim of it.
So while the headline figure improved, the part of crime people actually encounter did not. For a typical person, the chance of seeing a crime stayed about the same over the whole period, even as the chance of suffering one fell.
The statistics measured what happened to people. People were describing what happened in front of them.
Witnessed violence rose by 28%
Violent crime made up a growing share of all crime over the period. And violence is unusually public: each violent incident had roughly ten times as many witnesses as other crimes.
To picture it (our illustration, not the paper’s): a burglary at night may leave one victim and nobody who saw it. A fight outside a pub can be seen by a dozen passers-by.
As a result, the share of people who witnessed violent crime rose by 28%, even as victimisation overall fell.
Seeing crime and suffering it were linked to different worries
Both experiences went with more negative views about crime, but not the same ones:
- Witnessing was more strongly linked to lower confidence in the justice system and to worse views of the local neighbourhood.
- Being a victim was more strongly linked to feeling less safe personally.
These are associations in survey data, not proof that witnessing crime changes people’s views.
Counting witnesses left the cost of crime roughly flat
The authors also estimated the social cost of crime by putting a value on the harm it causes:
- Counting victims only, the total cost of crime fell by 17%.
- The costs borne by witnesses rose by 36%.
- Counting both, the total stayed roughly unchanged.
In other words, the improvement on the victims’ side was roughly cancelled out by a growing burden on bystanders.
What gets counted decides whether crime looks better
The gap between crime statistics and public opinion is usually treated as the public getting it wrong. This paper suggests both sides may be describing something real.
Victim statistics answer one question: how many people were harmed? People’s sense of crime answers another: how much crime do I see? When violence becomes a larger share of crime, the two answers can move in opposite directions.
Every statistic is a choice about what to count. If witnesses are left out, official figures will understate what crime costs society and keep clashing with what people experience. A reasonable inference, not something the paper tested, is that reducing visible public violence could do more for public confidence than the overall crime rate suggests.
What the study doesn’t show
- It’s a working paper (October 2026) and hasn’t been peer reviewed yet.
- The data cover England and Wales from 2001 to 2018. They don’t include the pandemic years or anything since.
- The links between witnessing and attitudes are correlations. People in areas with more visible crime may differ in other ways.
- The witness cost estimates depend on how the harm of seeing a crime is valued, which is harder to measure than a stolen car or a hospital visit.
- Like any survey, it relies on what people remember and what they consider a crime.
Falling crime and rising worry can both be true
When crime figures fall but people feel less safe, the public isn’t necessarily wrong. They may simply be counting what they see.
Before dismissing fear of crime, ask what people are witnessing.