Mainstreaming Methodology for Estimating Costs of Crime
In the early stages of developing cost of crime estimates for a country it can be very helpful, where data are missing or incomplete, to make use of estimates made for other countries. Thought and care are needed when doing this. We use three examples to illustrate.
1. Burglary. Suppose there is an estimate that 5% of households in country X are burgled each year but nothing is known about the value of items stolen. Estimates of the latter are however available for some other country Y. The estimates for Y (average loss is €2,000 while average income is €20,000) can be adapted for use in X by reference to the income level there. This adjustment can be a simple one, that assumes the average loss is the same proportion of average annual income (10%) in the two countries. Or it can be derived from a more elaborate model of the relationship between the average value of stolen items. If there are data on the average loss and average income for many countries then a regression model can be used to derive a formula for the link between the two. This formula can then be used to make an estimate of the average loss in a country with a given income level.
2. In the valuation of the emotional and physical impacts of harm (e.g. suffered in assaults) transferability of evidence and data raises issues under two principal headings:
The similarity of the physical and emotional impacts of crimes. Evidence on the physical impacts of a particular crime in one country might be transferable to other countries if the manner in which those crimes are committed is similar between the two countries. However, where there is a disparity between countries in the modus operandi for a particular crime – for instance, in the use of firearms in robberies (which would be expected to result in a greater incidence of gunshot wounds) – then average crime severity might well differ between countries, and evidence obtained in one country will be less useful for application in the other.
The similarity of individual preferences over emotional and physical health states associated with victimisation. If individual preferences towards the sorts of impacts which victims of crime can experience do not vary significantly across countries, the evidence from one country relating to the value of preventing crimes of different severities might be suitable for use in other countries. One way in which economic preferences do vary significantly across countries is in terms of personal income – income per head in the poorest countries in the EU can be significantly lower than that in the richest, although the disparity is reduced when figures are compared on a Purchasing Power Parity basis (e.g. Romanian GDP per head was estimated by Eurostat to be 35 per cent of the EU27 average in 2005, compared with Sweden at 124 per cent of the average). Conventional practice when transferring values in these situations is to apply the following simple adjustment:
where V(TC,SC) is the value to be transferred to the target country or from the source country, I(TC, SC) is the income per head of the target or source country, and eI is the income elasticity of demand for health. The value of this last parameter is usually assumed to be equal to one, which means that the adjustment effectively reduces to multiplying the transferred value by the ratio of incomes in each country.
3. The proportion of police time going on a particular offence type is not known for most countries. But there are one or two countries (Sweden and England & Wales, for example) where this matter has been carefully studied by analysis of police diaries or activity data. Since the policing costs associated with, say, burglary have to be estimated one possible strategy is to apply the time proportions estimated in Sweden or England. Of course, this procedure may be seriously flawed if there are large variations across countries in policing priorities or time allocation. But the alternatives, such as estimating the time allocation on the basis of the proportion of total crimes that are burglaries, may be even less reliable.
The question of transerability of estimates is considered in more general terms in the latter part of a slide show based on a presentation by Dubourg & Gordon (2007).
