ESTIMATING THE CANONICAL DISEQUILIBRIUM MODEL - ASYMPTOTIC THEORY ANDFINITE-SAMPLE PROPERTIES

The purpose of this paper is to review the various available estimatio n techniques for the canonical disequilibrium model: full information maximum likelihood (FIML), pseudo maximum likelihood methods, and thei r simulated counterparts. The use of the latter was advocated by Laroq ue-Salame (1989) when the model is analytically untractable. We provid e some new asymptotic results for all of these methods and compare the ir behavior on finite samples using Monte Carlo simulations. It appear s that FIML performs only slightly better than the second-order PML me thod, which in turn performs only slightly better than its simulated c ounterpart. Given the ease of implementation of the latter technique, these results strongly militate for its use.