Robust Bayesian analysis of a multivariate dynamic model

Abstract

Recently, a Bayesian approach has been widely used to infer parameters of complex models from finite sample data in business, finance, social and physical sciences. Bayesian applications in finance literature are growing due to their exposure to prior beliefs that macroeconomic variables influence the financial behavior of customers, investors and firm’s performance. This paper employs a robust Bayesian analysis model to capture a complex multi-dimensional process and estimate the required parameters of interest. It considers a class of prior distributions for the parameters that is a mixture of natural conjugate base priors and a contamination class of prior distributions. The conditional type-II maximum likelihood posterior densities and posterior means for the coefficients vector and autoregressive parameter are derived for the coefficients vector and autoregressive parameter of interest. The Markov Chain Monte Carlo (MCMC) method has been used for obtaining the marginal posterior densities and Bayes estimators for various combinations of the parameters. The usefulness of the model is demonstrated on Indian financial firms’ balance sheets data between 1992 to 2015 to observe the sensitivity of the proposed model. The result shows most of the corporate performance variables are significantly affecting the firm’s profitability with the exception of interest coverage ratio.