Comment: 40 pagesConsider a university administrator who must assign students to both courses and dorms, and runs (random) serial dictatorship independently in each market. While serial dictatorship is efficient for each individual matching problem holding the other fixed, Pareto improvements can be found via students jointly trading their allocated course and dorm. To resolve this problem, we introduce paired serial dictatorship where students report relative preferences between courses and dorms that then influence their priority in either market. By aligning allocation priorities with students' relative preferences, paired serial dictatorship improves welfare compared to independently running serial dictatorship in each market. Any deterministic allocation that arises in equilibrium is furthermore Pareto efficient, pointing towards randomization as the key barrier to optimality.