One of the main conclusions is that, because of the recession, sex workers are forced to bring their price down and accept more clients to earn more money, thus opening themselves to higher risk of facing violence behaviour, rapes, robbery and murder.
The report (which I have flipped through) also includes some rudimentary economic evaluation which aims at highlighting what the cost-saving could be if Westminster council were to implement some measures to increase the number of crimes reported by sex workers.
Using data from various sources (including a questionnaire that is part of the research, as well as official data from the UK Office for National Statistics), the report tries to assess the cost of implementing the intervention against the potential costs of avoiding negative events (eg a sex worker being raped).
The authors of the report are pretty upfront in stating that their analysis is partial, only considers "first-order effects" and that the figures for the return of investment for the council are "not exact". Also, I suppose that their main objective was to raise the issue, so in a sense that's fine.
But I can't help feeling that with a little extra effort, a proper economic model could have been set up and run to give even more robust findings. For example, the report suggests that:
If one rape offender was convicted (due to a sex worker reporting the incident) and the perpetrator’s potential to re-offend was prevented, an average of £73,564 would be saved in a 12 month period (this could be greater if the offender re-offended in subsequent years). Calculation:If 10 people committed rapes, this would cost £960,000 [my comment: This is taken from a ONS report]. 2.67 of these people are likely to commit 2.87 more rapes in a 12 month period [my comment: this is based on data from the UK Home Office], so (2.67 x 2.87 x £96,000 =735,638.4 /10 people) = average of £73,564 per offenderSurely all these numbers are subject to huge variability, which is not accounted for, here. Of course data are difficult to obtain for such a sensitive topic, but this seems one of those cases where the inclusion of informative priors to describe uncertainty around point estimates (which might be genuinely given with no measure of uncertainty in the available data) is extremely valuable.