SBAM: An Algorithm for Pair Matching

This paper introduces a new algorithm for pair matching. The method is called SBAM (Sparse Biproportionate Adjustment Matching) and can be characterized as either cross-entropy minimizing or matrix balancing. This implies that we use information e\u000eciently according to the historic observations on pair matching. The advantage of the method is its e\u000ecient use of information and its reduced computational requirements. We compare the resulting matching pattern with the harmonic and ChooSiow matching functions and find that in important cases the SBAM and ChooSiow method change the couples pattern in the same way. We also compare the computational requirements of the SBAM with alternative methods used in microsimulation models. The method is demonstrated in the context of a new Danish microsimulation model that has been used for forecasting the housing demand.