posted on 2014-12-15, 10:36authored byJustine A. Freeman
The reputed deterrent effect exerted by capital punishment, that executing convicted murderers reduces the number of homicides, is frequently cited as a justification for this penalty's employment. Considerable attention has been devoted to the deterrence hypothesis, with economists producing an impressive amount of research although with inconclusive results. However, the majority of studies focus on the USA with a major void in research relating to Britain. This thesis addresses the issue of deterrence utilising data from England and Wales. We test the robustness of an earlier study's results, reproducing Wolpin's (1987) data set and following a comparable regression analysis. As Wolpin's investigation only includes three years post-abolition data, it may have failed to capture the true deterrent effect should policy changes only modify behaviour in the long term. Consequently, this thesis extends the analysis to include the full impact of data from the post-abolition period. Additionally, we examine the series for stationarity as it has recently been demonstrated that regression analysis using non-stationary series produces inconsistent coefficient estimates, biased estimates of coefficient standard errors and invalid standard statistical inference procedures. Ultimately, we consider a more appropriate modelling technique.;Contrary to previous studies the general conclusion reached in this thesis is that there is limited evidence that the death sentence exerted a significant deterrent effect on murders and manslaughters in England and Wales. Although a robust deterrent effect did not emerge using many of the models, it must be stressed that this hardly qualifies as unequivocal confirmation that capital punishment does not exert a deterrent effect, for absence of proof is not at all the same thing as proof of absence. Indeed, while a statistically significant deterrent effect could only be detected using a 'stripped down' model, the fact that such a result emanated from an extended data set and the preferred model specification does convey this finding extra weight.