Probability Feedback in a Recursive System of Logit Models: Estimation

Abstract

Some estimation procedures are considered for the recursive system with probability feedbacks introduced in Q. H. Vuong (1982b) for the case in which the probability models are logit models. The asymptotic distributions of the two-step estimator, of the estimator obtained at each iteration of a natural iterative sequential procedure, and of the estimator obtained at each iteration of an efficient iterative sequential procedure are derived. It is shown that, upon convergence, this latter procedure produces an asymptotically efficient estimator.