Simultaneous‐Equation Systems Involving Binary Choice Variables

Simultaneous‐equation systems of continuous variables are well known in the social sciences. Unfortunately, these methods cannot be used when dealing with discrete endogenous variables. Although for discrete types of data a broad range of methods and techniques have been developed, their applicability is restricted to single‐equation systems (for example, logit or probit formulations) or to association‐type models (log‐linear models), mainly because of the lack of suitable estimation methods and computer programs. The basic problem is the multivariate error structure of the endogenous variables. This paper presents a simultaneous‐equation system of binary endogenous variables. The model is applied empirically to longitudinal data on mode choice and is compared with the well‐known log‐linear model. 1990 The Ohio State University