ABSTRACT
As a student, I had the great pleasure of being taught by the statistician A.R. Jonckheere, originator of the eponymous trend test. I held him in awe at the time because I was told he could play Scrabble well in seven languages. His insistence on prediction was remorseless. He described a visit he had made to the laboratory of the eminent ethologist Nikolas Tinbergen. Jonckheere said the two men stood in front of a tank of sticklebacks, and he challenged Tinbergen to predict the behaviour of one fish in five-second units. Jonckheere said that Tinbergen came out reliably better than chance according to the sign test which he conducted in his head. Criminology has not consistently engaged with the future. Visions of
future crime trends are sparse in comparison with analyses of past trends. Such early work on prediction in criminology as was undertaken was more concerned with the prediction of future offending by individuals, rather than prediction of crime trends. Even that limited emphasis on prediction was lost from the criminology scene in the 1960s and 1970s. Since then, massive changes in the discipline have meant that the focus of attention has moved to the crime event from the criminal disposition. However, a renewed preoc cupation with the future has not attended the change. Most analyses of trends in crime offered by criminologists in routine activity and rational choice perspectives are persuasive but retrospective. Both first used data from historical trends (in violent crime and suicide rates, respectively). Recent work in the traditions is subject to the same observation. Taking examples of quality work more or less at random, Ross Homel’s (1993)
superb analysis of drink-driving, the Homel and Clark modelling of vio lence in pubs and clubs (1994) and La Vigne’s (1994) study of gasoline drive-offs all develop models on the basis of extant data. The analysis of ransom kidnapping in Sardinia (Marongiu and Clarke, 1993) is likewise wonderfully provocative but post hoc. This is not to criticize these ground breaking studies, but to suggest that they were designed to describe, not to predict. For instance, if a predictive intent had been foremost, the sample could have been divided (to use one of the cruder alternatives) into construc tion and validation samples, to get a more appropriate idea of the predictive power of the relationships observed. This matters little in so far as the past is a simple extrapolation of the future, but it is not. The last century has seen the disappearance of theft of large animals and (almost completely) of theft from employers, to be replaced by motor theft and plastic fraud (see Bottomley and Pease, 1986). Felson’s Crime and Everyday Life (1994) likewise is not predictive in the
sense that the deciding data do not exist at the point of study design. Further, his concluding chapter, ‘Into the Future’, does not directly talk about crime. One can sympathize with not giving hostages to fortune in that book. People who predict the future tend to look silly when overtaken by events. How ever, the point stands that, even when talking about the future, the future of crime is veiled. The lack of future orientation in studies of the crime event contrasts markedly with the resurgent emphasis on prediction in the study of criminality.
