Testing for Model Selection in Predicting Aggregate Variables

This paper focuses on the choice between aggregate and disaggregate models, consisting of both univariate and multivariate specifications, in predicting aggregate variables. A formal hypothesis testing procedure for in-sample model selection is suggested. The empirical size and power of the test are investigated via the use of Monte Carlo simulations. Empirical results show that the test has good performance not only when the competitive models are non-nested specifications, but also when considering nested competitors.