Crime & perception ·
Is crime actually rising?
What the fear says, what the figures say — and where they honestly part.
Written with AI from the books and data named at the end. Both cases below are made at full strength — the honest version means you leave knowing the strongest form of the side you disagree with. How we use AI
In twenty-nine of the last thirty-two years, most Americans have told Gallup that national crime is rising. Over the same three decades, the surveys that ask people what actually happened to them recorded violence falling by about seven-tenths on both sides of the Atlantic. Somebody must be wrong — except that, on inspection, each answer is telling the truth about something different. That is what makes this a real question, and why it deserves the honest version.
The crime moved, and the old rulers don't measure it
More than half of measured crime against people in England & Wales is now fraud and computer misuse — roughly 5.3 million of 9.6 million incidents — categories the famous long-run comparisons did not contain before 2017. Fraud alone is 32% above its 2017 level, even as computer misuse has fallen. The visible offences have not vanished either: police recorded 507,086 shoplifting offences, only 4% below the previous year's 530,324.
And the reassuring averages have soft spots. Only 48% of violent victimisations in America were reported to police in 2024, so recorded numbers cannot hold the whole picture. The 2020 murder spike — up 29% — showed how fast "solved" problems return. The US victim survey itself has turned: 23.3 violent victimisations per 1,000 in 2024, up roughly two-fifths from 16.5 in 2021. The fear is not innumeracy; it is tracking real things the headline series miss.
The latest CSEW still puts violence 78% below its 1995 peak, while England and Wales recorded 499 homicides — the lowest count under current recording practices.
By the best measures we have, this is a much safer age
The victim surveys — which don't care what gets reported — are unambiguous about the long run. Violence in England & Wales: 4.5 million incidents in 1995, 989,000 now — down 78%. Theft: 11.6 million to 2.6 million — down 77%. Homicide, the one crime that can't hide, tells the same story: 499 in the latest year, the lowest count under current recording practices, with knife homicides at their lowest since 1991. America's reported murder count fell 14.9% in 2024 and a preliminary 18.1% in 2025; the first half of 2026 was lower again in the cities reporting promptly.
Meanwhile the gap between fear and experience has a measured shape: Gallup's national-versus-local seriousness question has averaged a 43-point spread for decades — last autumn 49% called national crime extremely or very serious, while 12% said the same of their own area. Most Britons say national crime rose over the decade; asked about their own street, they split. We remember the stories. We live in the denominators.
The NCVS violent-victimisation rate rose from 16.5 per 1,000 in 2021 to 23.3 in 2024 — a 41% increase from the survey's recent floor.
The long fall
Focus, tap or hover a year to read the source value.
Choose a year to inspect its value and source.
What both sides accept
The long fall is real: violence down ~78% (England & Wales) and ~71% (US) from 1990s peaks, on victim surveys.
Neither side seriously disputes the series — the argument is about what it leaves out.
Receipt 1 of 4Fraud and computer misuse are now ~55% of measured crime in England & Wales — and absent from every pre-2017 comparison.
"Lower than the 1990s" and "more than the 1990s measured" can both be true: different rulers.
The perception gap is stable in two Gallup series: national "increasing" majorities in 29 of 32 readings, and an average 43-point national–local gap on crime seriousness.
Whatever crime does, the fear has its own engine — and it points away from home.
The official rulers themselves are contested: UK police-recorded figures lost their statistical accreditation in 2014; the US victim survey is mid-redesign.
Which is why single-year headlines are the least trustworthy number in the whole argument.
Receipt 4 of 4What would change our mind
Towards "rising"
BJS National Crime Victimization Survey violent-victimisation rate above 25 per 1,000 people aged 12 or older for three consecutive annual NCVS releases.
Source:BJS · NCVS 2024CSEW fraud incidents above 5.0 million incidents for two consecutive annual CSEW releases.
Source:ONS · YE March 2026ONS police-recorded homicide count above 550 homicides for two consecutive annual ONS releases.
Source:ONS · YE March 2026
Towards "falling further"
FBI national murder rate at or below 4.0 per 100,000 people for two consecutive final annual UCR releases.
Source:FBI · 2024 finalONS police-recorded shoplifting count below 450,000 offences for two consecutive annual ONS releases.
Source:ONS · YE March 2026Gallup national-versus-local crime-seriousness gap below 30 percentage points for two consecutive annual Gallup readings.
Source:Gallup · 2025
When the evidence moves, this page moves — that is the promise of the honest version.
The arguments · non-fiction
Steven Pinker · 2011
The long decline of violence — the case at civilisational scale.
Tom Gash · 2016
The myths on both sides of the crime argument, tested one by one.
Jill Leovy · 2015
Where the violence concentrates — and who the averages hide.