We examine the effects of a major pension reform in 2015 on the joint retirement decisions of working couples in the Netherlands. The reform abolished the partner allowance, a state pension supplement for a nonworking partner below the state pension age, and introduced a law that made actuarially generous early retirement arrangements much less attractive. Using rich administrative data, we estimate a multivariate mixed proportional hazards model that distinguishes between two sources of joint retirement: causal mechanisms that make retirement of one spouse more likely when the other spouse retires (“state dependence”, e.g. due to complementarities in leisure or social norms) and correlated preferences (observed and unobserved heterogeneity). We find that the reform weakened state dependence in joint retirement.