Predicting reconviction for middle class offenders. Do predictors change with length of follow-up?

Are just a few variables important for predicting reconvictions under various conditions? Or do particular variables come into play at different points? This paper analyses the contribution which various variables make to a variety of outcome measures. The resource for the study is a consecutive series of 388 offenders who were seeking white-collar employment between 1 January 1970 and 31 March 1973. Thus, their class position at this point - defined in occupational terms - is quite clear. The series is considered in criminological terms to the end of 2006 (a follow-up of 33-36 years). Variables considered include demographic characteristics, educational background, employment history, and previous criminal record. The outcome measures are the likelihood of being reconvicted for (a) any ‘standard list’ offence,(b) ‘white-collar’ offences and (c) any sex or violence offences within 1, 5, 10, 20 and 35 years after being interviewed between 1970 and 1973.