Despite meta-analytic evidence on the effectiveness of offender treatment programs, little is known about how therapeutic changes to dynamic risk factors contribute to this effect. The present study explores the relationship between prison climate, changes in dynamic risk factors, and recidivism in a sample of 185 male violent and sexual offenders. Participants completed psychometric tests on dynamic risk factors (procriminal attitudes, antisocial personality patterns, empathy, anxiety/neuroticism) and perceived prison climate before and after correctional treatment (length: M = 32 months) in a social-therapeutic facility. Recidivism data were available for 92 participants with a follow-up of M = 4 years. Medium-sized prosocial changes to the dynamic risk factors of procriminal attitudes and anxiety/neuroticism in all offenders were found, while antisocial personality patterns only decreased among violent offenders. Positive ratings of different aspects of prison climate significantly correlated with prosocial changes in all dynamic risk factors except empathy. Cox regressions showed that prosocial changes in dynamic risk factors were not predictive of general or sexual/violent recidivism. The reasons that could account for the missing link between therapeutic change and reduced recidivism are discussed. These range from problematic methodological issues like the repeated measurement of self-reports to the omission of relevant constructs from desistance theory and protective factors. It is concluded that more attention should be given to creating a prison climate that is conducive to therapeutic change.