Using a Multinomial Logit Specification to Model Two Interdependent Processes with an Empirical Application

Theories in sociology and other social sciences often postulate reciprocal causal relationships. Yamaguchi advanced a model for the interdependence of two discrete-time, discrete-state endogenous processes. The Yamaguchi model is introduced with a discussion of its advantages over conventional methods and a comparison with recently developed relevant models. To overcome the obstacle that existing statistical software cannot directly estimate the Yamaguchi model, the author has developed a method that converts estimated parameters from standard multinomial logit estimation into parameters of the Yamaguchi model. This method makes it possible and easier for researchers to estimate the Yamaguchi model using standard statistical software and a simple programming of linear transformation. The method is simple and straightforward and thus merits application to an analysis of interdependence with panel data. This article also provides a detailed empirical example to illustrate an application of the method.