Cointegration tests in present value relationships: A Bayesian look at the bivariate properties of stock prices and dividends

Abstract

A Bayesian methodology is used to examine the bivariate time series properties of stock prices and dividends. The hypothesis that stock prices and dividends are cointegrated is emphasized. Exact small sample results are calculated using a specific parameterization for the bivariate model. Some commonly encountered problems in conventional cointegration tests based on asymptotic theory are thereby eliminated. We consider three hypotheses whose nonlinear form entails the need for multidimensional Monte Carlo integration. In addition, three data sets are used, two of which are annual U.S. sets over long time periods and one a post-war Canadian set. The intention is to demonstrate that a more accurate understanding of the value and limitations of cointegration tests for investigating present value relationships is possible with a Bayesian methodology, employed over a variety of data configurations.