We assess to what extent the associations between marriage and offending differ for high-risk men marrying in two distinct periods: 1930-70 and 1971-2006. Between these two periods, power relations between the sexes, laws governing marriage and in general the role and expectations attached to marriage differed. Based on these differences, we argue that - following two explanations for the 'marriage effect', that is the control and the social capital explanation - a different effect of marriage is expected for the two marriage cohorts. Our results confirm these expectations and thus provide support for both explanations. © The Author(s) 2012.