This paper studies the labour force participation dynamics of older couples in the United States. Longitudinal data from the five available waves of the Health and Retirement Study (HRS) is used to investigate if the dynamics introduced by considering both spouses’ behavior provide additional information in trying to fit observed participation sequences. The paper uses a bivariate dynamic binary choice model with unobserved heterogeneity and serial correlation to disentangle the many sources of dynamics and correlation in a couple’s decision making. First, strong true state-dependence is found and results in a bunching of participation and non-participation sequences. Cross-spouse state-dependence is also found which points to indirect effects of social security and pension incentives through complementarity in leisure. Second, the Spouse Allowance program is found to have predicted effects on participation of the couple and these effects are statistically significant. A simulation exercise presents evidence that the elimination of the spouse allowance can raise participation of wives at age 62 by more than the decrease in participation of husbands at age 65.