In 2004, the Justice Research Consortium (JRC) completed a randomised controlled trial testing the crime reduction effects of restorative justice conferences. Burglary and robbery offenders and victims who agreed to participate in the trial were randomly assigned to either a face to face conference (experimental) in addition to normal court processing or to a control condition where their case proceeded as usual through London's Crown Courts. Restorative justice has been shown to be effective at reducing reoffending in some contexts. More research is needed, however, to determine why some individuals taking part in restorative justice desist from offending whilst others reoffend quickly. Research suggests that individual characteristics such as gender, race and age, criminal history features such as frequency, seriousness, and custody, and ancillary factors such as drug use and victimization can have an impact on offending, and this thesis investigates the extent to which these variables predict time to reoffending. Using survival analysis, this thesis also explores whether, in a sample of serious adult offenders, the effects of restorative justice on time to reoffending are different at different values of these variables. Key findings include: o In the burglary experiment, female offenders reoffended significantly faster than males, and females in the control condition reoffended twice as fast as females in the experimental condition. o The higher the frequency of arrests or convictions, and/or the greater the seriousness of offending prior to random assignment, the faster project offenders reoffended. This effect was significant in the burglary experiment. o Highly frequent and highly serious offenders (based on criminal history prior to random assignment) took significantly longer to reoffend in the experimental condition than the control condition. o Offenders who were identified as using crack, heroin, or crack and heroin, reoffended at a faster rate if they were in the experimental conference group. This effect was significant in the robbery experiment. Those who were not drug users offended more slowly in the experimental condition. o The number of times offenders had been a victim of crime in the five years pre-random assignment was significantly related to survival time in the burglary experiment.