How bad is crime?

A first pass

There is a temptation to produce only ‘the finished word’ on some topic. Obviously I think that finished word articles are good, and I both consume and produce them. But I think that this tendency can make us self-censor less well-finished thoughts, which is probably bad for the same reasons I think excessive pre-publication peer review thresholds are bad, the same reasons trade secrets have costs, and in analogy to how many drugs’ best uses are only found after they are approved for something else.

So here’s the starting point of some thoughts about crime. I’m hoping that putting thoughts out there as I develop them will mean aggressive policing of my factual mistakes and errors of reasoning, and people will suggest lots of useful ideas to consider and things to read.

I think crime is underrated as a topic.

Crime is important because being victimised is extremely bad. People hate it. Even on top of the obvious costs, say if you lose a $1,000 phone, it feels humiliating and scary, shaking your faith in strangers and the world. This means that when crime does happen, it imposes large costs on people’s welfare: not just literally having stuff stolen, and not even just sustaining injuries, losing days at work, but deeper psychological harm. It means that people go to great lengths to avoid it, often in ways that have negative impacts on them and society at large.

Of course, crime is also pretty much inevitable, and a feature of all societies. It’s going to be impossible, with the best education, social programmes, and criminal justice system, to reduce it to zero. But it’s worth having a gauge of how costly crime is, so we can think about how important it is, and therefore what programmes it’s worth funding to reduce it, and how much effort we should spend on thinking about crime policy.

One way of trying to work this out is studying people’s stated willingness to pay for crime prevention efforts. One experiment asked people if they were willing to personally pay a given amount (randomised between people) to reduce one in every ten of a given crime (randomised between people) in their community. So they might be asked ‘Last year a crime programme reduced murders 10 percent in your area – would you be willing to pay $200 to continue this.’ It asked each individual in a representative sample of the US population a bunch of questions like this, varying the situations and amounts.

Obviously studies like this are imperfect. The situation is hypothetical, and there are all sorts of issues that might be playing into the question. For example, when the asking price gets sufficiently high they might think that it would be better to just move, or to invest in security for themselves rather than the programme benefiting the whole group.

You can add this up across the country. If an individual is willing to pay $80 to reduce murders 10 percent, and there are 25,000 murders, then they implicitly value one tenth of murders (2,500) at $80, and thus each murder at just over three cents. Multipled by the hundreds of millions of Americans, that would be a total value of just under $10m per murder. In the paper, whose calculations were done in 2006, Americans were willing to pay $25,000 to avert a burglary across their society, $70,000 to avoid a serious assault, and nearly $10m to avoid a murder.

A more practical situation comes when juries award money to ‘make people whole’ for physical injury, pain, suffering, mental anguish, shock, and discomfort that they have experienced due to some illegal action. For example, one 68-year old lady was shot through the spine in a drive-by shooting, and left paraplegic – a jury gave her $2.7m in addition to her medical costs.

If you combine these awards, in a large sample, with separate ‘physician impairment ratings’ – basically how bad doctors think the injury is compared to death – then this is another method of estimating the statistical value of a life, something we have hundreds of estimates for, which typically comes out somewhere above $5m, depending on the wealth of the country and the methodology.

Some of these studies use extremely ingenious methodologies. For example one study looked at Sierra Leoneans travelling between the airport, and the capital Freetown, choosing between death trap Soviet-era helicopters, intermittently-available hovercraft, ferries, and water taxis.

I suspect, however, that people would generally prefer a risk of a natural death to a risk of murder, especially if it’s not happening to them but a friend or loved one, so it might not be unsurprising if people’s willingness to pay to avoid murder was quite a bit higher than their willingness to pay to avoid death. 

A third way of looking at the cost of individual crimes is listing all the possible costs, both monetary and non-monetary, and both individual and society-wide, and then totting them up. One study digs through existing estimates of mental health costs, productivity losses to society, use of public services, the courts, quality of life, and so on, to come up with an estimate. A major downside of this study is that it takes the criminal justice response as a given – whereas in this post I am treating our response to crime as the variable we should control, and the cost of crime as a reason or driver for our response. That caveat aside, this study finds roughly similar costs of crime – and the response to each crime is generally under 10 percent of the cost. This methodology finds murders cause $7.8m of harm.

A fourth method is looking at house prices. Crime tends to be extremely local. The murder rate of East Pilsen in Chicago is perhaps sixty times higher than Pilsen, immediately bordering it, according to official data from the police department. Walking from the centre of one area to the other would take perhaps 10 minutes – though it would involve going under a major underpass. A random, unscientific approach shows very similar properties on Zillow worth $100,000 more just for walking two minutes past the underpass.

Turning that property differential into a value per murder is difficult. One study looks at property in Sydney, Australia: they find an individual murder causes a roughly four percent drop in property prices within a 0.2 mile radius. Calculating roughly, I reckon there are about 1,000 properties in one of those circles on average across Sydney, where properties are worth an average of A$1.1m, or $750,000. That would imply a capitalised value of $30m.

I suspect this $30m estimate of the social cost of a murder, as measured in house price terms, both understates and overstates the ‘true’ value. If an unexpected murder happens in a high murder location, that probably doesn’t shift house prices much, since it’s a small percentage of the overall murder rate, and is hard to distinguish from noise. This doesn’t, however, imply that murders in that location aren’t weighing heavily on property value there. So that element would bias the estimate down from its true value. On the other hand, property prices in Sydney are correlated inversely with density, so my rough and ready model of dwellings per square mile is putting too many near the expensive properties in low crime areas that are driving the changes, and too few in the cheaper areas in higher crime areas where marginal murders are not making as much of an impact.

So far I have just been talking about individual crimes. But let’s multiply this up for a whole country. One of the studies I tackle earlier has done this for the USA, adding up their estimated social cost for every crime, both recorded and reported. They use cost estimates that are about in the middle of the range for individual crimes. Their central estimate is that crime costs America $2.6 trillion annually, mostly coming from violent crime. This is about 12 percent of US GDP. By this metric, it would be, in GDP terms, one of the US’s biggest problems, on par with housing. For a country like the UK with a murder rate about five times lower, the problem is probably about five times smaller.

I actually think the American problem is considerably bigger than this estimate, because this study only includes the costs of crimes that actually get committed. However, people try their damnedest to avoid being the victims of crime. This leads to many extremely socially costly behaviours. 

Some of these have really obvious ‘macro’ effects. People who fear new neighbours have a substantial risk of committing crimes tend to oppose new development nearby, and to live in extremely spread out ‘sprawl’ suburbs, where sheer walking distance between places makes crime more difficult. What’s more, people in high-crime areas prefer to travel with metal shields around them at all times – that is by car – causing dramatically higher carbon emissions. By contrast ultra-low-crime Japan is tolerant of high density development throughout its cities, and rates of cycling, walking, and transit use are all extremely high, while carbon emissions are much lower.

You can aggregate some of these points by looking at the links between economic success and social trust, which is closely associated with crime.

But other effects are more ‘micro’. In a society with extremely low crime, people don’t bother locking up their bicycles when they go around town. Car rental companies need not impose laborious and strict checks and rules to prevent their cars from being used in crimes – a huge number of annoying regulations cover off rare edge cases. Landlords need not ask for large deposits. Women can walk about freely at night, including through poorly-lit parks. 

So crime is a big problem, I think most people underrate it, and I think we should think more about it.