Estimating the Cost of Crime

Governments decide how much to spend on policing, courts, prisons and prevention programmes with finite budgets, and cost-of-crime estimates give that decision a common currency. Because official crime statistics record only offences that reach the police, they cannot on their own say how much a crime problem actually costs the country running it, which is the gap cost-of-crime research fills.

If a prevention scheme costs a fixed sum to run each year and can be shown to prevent a measurable number of burglaries or assaults, comparing that cost against the estimated cost of each prevented crime lets analysts judge whether the programme is worth funding relative to competing uses of the same money.

The same figures surface in sentencing debate. When a court, a law commission or a legislature weighs whether a harsher penalty for a given offence is proportionate, cost-of-crime research supplies one input into that judgment: an estimate of the harm the offence causes to victims and to public funds, set against the cost of the additional punishment.

This does not settle sentencing policy on its own, since proportionality also rests on culpability and moral seriousness rather than price alone, but it gives legislators a harder number than intuition.

Insurers, compensation schemes and civil damages awards depend on similar reasoning. A state-run victim compensation fund has to decide how much to pay a victim of assault or robbery, and cost-of-crime research, particularly the intangible-cost component drawn from jury-award and willingness-to-pay studies, gives compensation boards a benchmark rather than an arbitrary figure.

Finally, aggregate cost-of-crime totals, once added across offence types and multiplied by estimated volumes, are used to argue for or against entire policy programmes at national level, from stricter alcohol licensing to funding for youth diversion schemes. Because so much rides on the number, how it is built matters as much as the number itself.

The oldest and most straightforward approach is cost-of-illness accounting, borrowed directly from health economics. It adds up every item with a market price attached to a crime: medical treatment, property replacement, lost earnings, and the public cost of police, courts and corrections attributable to that offence.

Because each line item is drawn from an invoice, a wage record or a government budget, cost-of-illness totals are relatively easy to defend, but they systematically understate the true burden of crime because they say nothing about pain, fear or lost quality of life.

The second approach, jury-award research, tries to close that gap by treating civil damages awards as a market price for pain and suffering. Economist Mark Cohen's 1988 study analysed jury awards in personal injury and wrongful death cases with injury profiles similar to those suffered by crime victims, then used the non-economic portion of those awards as a stand-in for the intangible cost of a comparable crime.

The method is useful because juries are, in effect, already pricing suffering when they set a damages figure, but it inherits every bias present in the civil litigation system, including which cases reach a jury at all.

The third approach, willingness-to-pay research using contingent valuation, asks a representative sample of the public directly how much they would pay in additional taxation for a programme that reduces a stated crime by a stated amount. This method has the advantage of asking ordinary citizens rather than juries to reveal how much they value safety, though contingent valuation surveys are sensitive to how the question is framed and to respondents overstating what they would actually pay.

No single method is treated as definitive. Serious cost-of-crime estimates typically report a tangible total from cost-of-illness accounting and a separate intangible estimate from jury-award or willingness-to-pay research, so readers can see which figure rests on which method rather than being handed one blended number.

Tangible costs are the components of a cost-of-crime estimate that have a direct market price and can, in principle, be checked against a receipt, a wage record or a public budget line. For an individual victim this starts with medical costs: emergency treatment, follow-up care, and in serious cases long-term rehabilitation.

Property crime adds the value of items stolen or destroyed, priced at replacement or repair cost rather than sentimental value, and violent crime adds lost earnings for time away from work during recovery, which can be substantial for injuries that keep a victim off work for weeks.

A second tier of tangible cost sits with the state rather than the individual victim: the public money spent on the police response, forensic examination, prosecution, court time and, where a conviction follows, the cost of a custodial or community sentence.

These figures are drawn from government budgets divided by caseload, so they carry their own measurement problems, chiefly that overhead costs common to many cases have to be allocated somehow to a single offence, but they are still grounded in real expenditure rather than judgment calls about suffering.

A third category, sometimes called second-party or precautionary costs, covers spending that would not exist without the crime problem at all: household security spending such as locks and alarms, private security guards, insurance administration, and the value of time people lose taking precautions against victimisation, such as avoiding certain routes or areas after dark. These costs are real but harder to attribute to any single incident, which is why some published estimates include them and others do not.

Because every tangible line item rests on an observable price, tangible-cost totals travel well across studies and are the part of a cost-of-crime figure least likely to be disputed. Their limitation is not accuracy but completeness: on their own they capture only the financial shadow of a crime, not the harm felt by the person it happened to.

Intangible costs cover the burden of crime that never generates an invoice: physical pain, psychological trauma, fear, disrupted sleep, loss of trust and a general reduction in quality of life following victimisation. There is no market transaction that prices a night of interrupted sleep after a burglary or the anxiety that follows a violent assault, which is precisely why economists turned to jury-award and willingness-to-pay research rather than accounting records to estimate this component.

Across the studies that separate the two categories, intangible costs consistently come out larger than tangible costs for crimes against the person, sometimes by a wide margin, because the human cost of violence and its psychological aftermath dwarfs the medical bill and lost wages attached to the same incident.

This is the single most important reason cost-of-illness accounting alone understates the true cost of crime: leaving out intangible harm does not just shave a little off the total, it typically removes the majority of it for violent offences.

Including intangible costs is also the more contested part of any cost-of-crime study, since critics point out that pricing suffering necessarily involves a modelling choice rather than an observed fact, and that jury awards and contingent valuation responses can be inconsistent across studies, jurisdictions and even the specific wording of a survey question.

Researchers respond that leaving intangible costs out entirely is itself a modelling choice, one that implicitly assigns pain and fear a value of zero, which is harder to defend than an imperfect estimate.

For policy purposes the practical lesson is to read whether a published figure includes an intangible component at all before comparing it against another study or another country's number, since a tangible-only estimate and a full estimate for the same offence can differ by a large multiple without either being wrong on its own terms.

A tangible-only figure is not incorrect, it is simply answering a narrower question, and treating it as the whole cost of the crime is the mistake, not the number itself.

Once tangible and intangible components are priced for a single incident of an offence, that figure becomes a unit cost, and researchers build a national total by multiplying each offence type's unit cost by an estimate of how many such offences occurred, then summing across offence types.

Two research programmes illustrate the approach at national scale: economists Kathryn McCollister, Michael French and Hai Fang published crime-specific unit costs for the United States in 2010, and the UK Home Office has run a parallel programme since the early 2000s, most recently updated by Matthew Heeks and colleagues in 2018, both combining cost-of-illness accounting with intangible estimates drawn from the jury-award and willingness-to-pay literature covered earlier in this topic.

Where the volume figure comes from changes the total more than any choice of costing method. Full guidance on comparing police-recorded counts, victimisation-survey estimates and other administrative sources sits in sources of crime and criminal justice data. The short version for a cost-of-crime estimate is that a volume figure drawn only from police records inherits every offence that was never reported, so multiplying an undercounted volume by an accurate unit cost still produces an undercounted national total.

India illustrates the problem starkly. The National Crime Records Bureau publishes an annual police-recorded volume in its Crime in India report, but no equivalent large-scale national victimisation survey exists to correct for underreporting the way the CSEW does for England and Wales or the NCVS does for the United States.

A cost-of-crime estimate built on the NCRB volume alone therefore compounds two separate undercounts, the offences victims never reported and the offences the resulting total never priced, and any such figure should be presented as a floor rather than a complete national cost.

Comparing totals across countries adds a further layer: unit costs are denominated in local currency and local wage and price levels, so a raw comparison between a UK and a US total has to convert one currency into the other and, ideally, adjust for purchasing power rather than the market exchange rate, or the comparison will mostly reflect exchange-rate movements rather than any real difference in the cost of crime.

Not every payment connected to a crime belongs in its cost. Victim compensation, insurance payouts and the value of stolen property that is later recovered are transfer payments: money or goods move from one party to another, but no resource is destroyed in the process.

Economists exclude transfer payments from the net social cost of crime, because counting both the loss to the victim and the compensation paid to cover it would double count the same underlying harm. What insurance and compensation schemes do add to the total is their own administrative cost, the staff time and overhead needed to run the scheme, which is a genuine resource cost even though the payout itself is not.

Published cost-of-crime totals also understate the true picture because they are strongest for the offence types with the best volume and unit-cost data, chiefly violent crime, burglary and vehicle theft, and weakest for corporate, financial and cyber offences, where victims are often organisations rather than individuals, unit costs are harder to standardise, and a large share of incidents never surface in either police records or a household victimisation survey. A national total built from the well-measured offence types alone is not wrong, but it answers a narrower question than its headline figure suggests.

The main practical use of a unit cost, rather than the national aggregate, is benefit-cost analysis of a specific prevention programme or policy change. If a diversion scheme or a policing initiative can be shown, through a credible evaluation, to have prevented a measurable number of offences of a known type, multiplying that number by the unit cost of the offence gives an estimate of the harm avoided, which can then be set against what the programme cost to run.

This is the calculation introduced briefly earlier in this topic, and it is the reason unit costs, not just the national headline total, are the figure most policy evaluators actually reach for.

That use also marks the limit of what a cost-of-crime figure can do. It weighs harm already measured against money already spent, but it cannot value culpability, cannot substitute for a proportionality judgment in sentencing, and depends entirely on the quality of the evaluation that produced the prevented-crime count in the first place.

Treated as one input among several rather than a verdict on its own, a well-documented unit cost remains the most defensible number a policy debate about crime is likely to get.