Anyone who’s hosted friends in London will be familiar with the etiquette advice that’s given to new arrivals: “don’t stand on the right side of the escalator”, “don’t stop moving in the middle of the street”, or the all-but-dead “don’t have a conversation on the tube”.
Increasingly, we’re starting to add safety measures like: “don’t make eye contact with anyone that seems crazy”, “a wallet isn’t worth dying for”, or “don’t get your phone out of your pocket in public” (not that you’d get any signal anyway).
As always, the first questions should be: Who are they asking? What questions are they asking them?
It’s maddening that the people giving this advice are often the same people making fun of non-urbanites for being overly scared of crime. The latest evolution in gaslighting is posting graphs showing that crime is dropping and that actually, our rapidly declining cities are the safest they’ve ever been.
As always, the first questions should be: Who are they asking? What questions are they asking them? This particular survey samples residents from designated statistical areas known as Middle Layer Super Output Areas (MSOAs) and Lower Level Super Output Areas (LSOAs). These areas don’t map cleanly onto cities, towns or police jurisdictions but have roughly equal population sizes to normalise statistical rates — an approach that’s useful for health indicators such as heart attack incidence, but less appropriate for capturing the complexity of crime.
Crime, unlike many other health or social phenomena, is highly concentrated geographically. Crime data often follows power law distributions, where 80 per cent of crimes are committed by a small number of repeat offenders and concentrated in just a handful of hotspots within any given area, typically busy streets, pubs, bus stations, and train stations. A survey that evenly samples across areas that differ drastically in crime intensity will inevitably mask these concentrations.
The survey also samples exclusively single-occupancy dwellings, deliberately excluding Houses in Multiple Occupation (HMOs). This significantly undercounts crimes in densely populated, often economically disadvantaged areas, and disproportionately oversamples wealthier, single-family homes.
The survey focusing on crimes against individuals also doesn’t capture crimes against businesses, including rampant shoplifting, vandalism, or other property damage. Nor does it accurately represent repeated victimisation, which used to be capped at 5 reports and was replaced by an opaque “crime-specific imputation method based on the 98th percentile” in 2019. This new method doesn’t solve the problem. If 98 per cent of respondents report 7 or fewer burglaries, the cap might be set at 7 for burglary, so someone reporting 20 burglaries is now recorded as 7 instead of 5.
Nor does it accurately represent repeated victimisation. Previously, the survey artificially limited or “capped” responses at 5 incidents per crime type, meaning that if someone reported being burglarised 15 times, it would only count as 5 incidents. In 2019, this was replaced by an opaque “crime-specific imputation method based on the 98th percentile”, but this new method doesn’t solve the underlying problem. If 98 per cent of respondents report 7 or fewer burglaries, the cap gets set at 7 for burglary. So someone reporting 20 burglaries would still be recorded as only 7 incidents instead of the previous cap of 5.
Historically, the demographic most likely to commit crimes is males aged between 15 and 35. Given that the UK’s median age has now reached 40, combined with significantly lower alcohol consumption among younger generations (a key factor associated with violent criminal behaviour), we would expect a substantial reduction in crime rates. It’s hard to reconcile this with imprisonment rates reaching historic highs, to the extent that the government is releasing individuals convicted of minor violent offences early to accommodate more serious offenders, with thousands more trapped in court backlogs.
In reality, definitive conclusions about whether crime rates are rising, stable, or declining are challenging to draw from this or any other crime survey in aggregate. A key question is: Which crimes? It might be the case that crime rates have dropped substantially, but people are capable of noticing that some demographics that contain a disproportionate number of criminals make up an increasing proportion of our population, and that along with this the types of crimes are changing. Mass stabbings or violent events like acid attacks that once would have been national news stories for weeks have become background noise.
Beyond this, what costs have been accepted in the name of crime avoidance? They might not be entirely aware of it, but people change their behaviour to avoid being victimised. In South Africa, Brazil, and other notoriously criminal countries, people live in secure compounds behind barbed wire fences, with CCTV cameras that alert friends if they get signal jammed. Their already notorious victimisation rates would likely be hundreds of times higher if they tried to live as we do in the West. What people are lamenting when they say crime is “out of control” might not be their personal stories of victimisation but the increasing cognitive load and behavioural adjustments required to stay safe that weren’t necessary even 5 years ago. Given that people change their behaviour, how people feel is really the only thing that matters. As Jeff Bezos said, “When the data and anecdotes disagree, the anecdotes are usually right. It’s not usually not that the data is being miscollected, it’s that you’re not measuring the right thing”.
