A Bayesian Extension of the J-Test for Non-Nested Hypotheses

It is a common practice to use the Davidson and MacKinnon's J-test in empirical applications to test non-nested model specifications. However, when the alternate specifications fit the data well the J- test may fail to distinguish between the true and false models: the J-test will either reject, or fail to reject both specifications. We show that it is possible to use the information generated by the J-test and combine it with the Bayesian posterior odds approach that would yield an unequivocal and acceptable solution for non-nested hypotheses. We show that the approximations of Schwarz and Bayesian Information Criterion based on classical estimates for the J- test yield the Bayesian posterior odds without any need for the specification of the prior distributions and the onerous Bayesian computations.