Bayesian Inference in a Simultaneous Equation Model with Limited Dependent Variables

Abstract:     
This paper develops new inferential procedures for conducting a finite sample likelihood-based analysis of a simultaneous equation model with limited dependent variables (SLDV). By employing the combination of Gibbs sampling and data augmentation, we can draw from the exact posteriors of these SLDV models and avoid direct evaluation of the nontrivial likelihood functions. The merits of the proposed estimation/model comparison procedures are illustrated through an application to default data of original issue high yield bonds. Recognizing the endogenous nature of a financially distressed firm's Chapter 11 filing decision, we are able to estimate both the length of time in default and the likelihood of a formal reorganization within the SLDV framework. The computed Bayes factor for the single-equation formulation against the simultaneous-equations formulation does not provide any overwhelming evidence in support of either model. We believe the pooled Bayesian inferential framework adopted here provides richer insight on the default experiences of original issue high yield bonds.

Note: This is a description of the paper and not the actual abstract.

JEL Classification: C11, C15, C34, C35, G33

Accepted Paper Series