Matchmaker, Matchmaker, Make Me a Match

Effective marriage matching, i.e. forming specific couples from pools of potential marriage partners, is vital for the success of dynamic longitudinal microsimulation models. This paper identifies some problems with a deterministic algorithm previously used in CORSIM, and in its “child” models, e.g., DYNACAN and POLISIM, for matching prospective marriage partners. It describes an alternative, stochastic, algorithm that provides somewhat better results. A set of validation measures illustrates the extent of this improvement, and indicates something of the ample room remaining for additional improvement. The paper situates the choice of a matching algorithm within the broader arena of marriage markets, and identifies some other areas in which improvements are desirable.