School of Law, King's College London.This article has been prompted by that of Larry Sherman (2009) To anticipate the article's conclusions, there has been over-investment (both financially and intellectually) in a technocratic model of reducing reoffending that attaches too much importance in accredited programmes and packages, and under-investment in models that see the process of 'people changing' as a complex social skill 2 . The technocratic model seriously underestimates this complexity, and its advocates wrongly assumes that experimental research can readily identify the causal processes at work in helping people stop offending. They mistaken suggest that clinching evidence 3 about 'what works' can be accumulated when in reality this is a field where evidence is perennially tentative and where knowledge is perennially labile. This is not to deny that there is a place for experimental methods in this field. They constitute one form of evidence about what works in reducing reoffending, and in some circumstances this can be very important evidence.
2
What works? What research has told usFor as long as I have being doing criminology, there have been tensions between researchers at the 'applied' end of the spectrum and their more traditionally academic colleagues. Government research, and government-sponsored research, tends to be largely a-theoretical, or rather, it implicitly accepts the conceptual frameworks within which political and governmental debate about crime and its control are conducted 4 . This body of work tends to be narrowly focused, and addresses specific policy questions. In general it is empirical, quantitative and increasingly incorporates a cost-benefit assessment. It is overwhelmingly short-termist -designed to answer the question whether whatever is being evaluated is having an immediate impact; and it tends to be uncritical, in the sense that it does not (or cannot) question general government policy. Rather, it assesses whether the evidence favours investment in one policy tactic as opposed to another. Traditionally this sort of research has either been carried out by Home Office (and now Ministry of Justice) researchers, by the NAO or the Audit Commission, or by academics on contract to government departments. Increasingly, though, 'niche consultancies' are also carrying out this sort of work.In contrast, there is a growing body of academic research which is much more theoretically orientated, and substantially detached from policy dilemmas -even when crime policy is the focus of its attention, as indeed is often the case. Typically this work is concerned with conceptual rather than empirical analysis, and in so far as it engages with empirical work is as likely to draw on qualitative as quantitative work. Probably the best known and the most cited British criminologist at the moment is David Garland (eg Garland, 2001(eg Garland, , 2002 5 . His work is an extended commentary on government crime control policy that is nevertheless largely detached from the day-to-day questions with...