Source: (2004) Revised paper presented at the 2003 meetings of the Australian and New Zealand Society of Criminology, 1-3 October, Sydney, Australia. Forthcoming in the Australian and New Zealand Journal of Criminology. Downloaded 19 August 2004.Restorative justice conferencing, in response to youthful offending, has grown in popularity around the world. While there is now substantial empirical evidence that shows offenders and victims are satisfied with outcomes and perceive the process as generally fair, available data on re-offending have produced mixed results. Uncertainty about how conferencing affects future offending may result from how re-offending is analysed. In this paper, I used survival analysis to reanalyse data from the Bethlehem, Pennsylvania Restorative Policing Experiment and assessed two methodological approaches: a standard comparative approach to examine differences in re-offending between offenders in conference and court and a variation analysis to examine differences in re-offending within conference and court groups. Comparative analyses showed that violent offender referred to conference were less likely to re-offend compared to violent offenders referred to court. There were no differences in re-offending for property offenders in conference and in court. Variation analyses showed that female offenders attending conferences were less likely to re-offend than male offenders in conferences. There were no differences in re-offending for males and females in court. These results suggest that there is value in comparing the effects of traditional and restorative interventions, as well as assessing how variation within interventions is related to future offending. Author's abstract.