Measuring the disutility of imprisonment to offenders

Recently, quite a lot of work has been done on deducing parameters from observed decisions using dynamic programming models, and Wolpin (1996) provides a good introduction to this literature. While a dynamic programming framework to this problem has not been utilised in this paper, the concept of bringing together actual decisions and a constructed model to explain these to infer an unobservable parameter is the same. For more discussion of the dynamic programming methodology and related issues, the reader is referred to Wolpin’s paper. Our alternative framework is to undertake a calibration exercise based on defendants’ optimal plea decisions for three different offences. For representative ex ante combinations of plausible parameter values for the cost of a guilty plea and expected cost of a trial we generate a very large sample of values of the discount rate at which a defendant would be indifferent between either plea. The median r value of our distributions is interpreted as the defendant’s willingness to pay to delay the cost or expected cost of imprisonment, and hence a measure of the disutility of imprisonment. Alternatively, our estimates can be interpreted as implicit premiums over and above the unskilled wage rate.