SELECTION OF NONNESTED DEMAND MODELS

This paper applies the likelihood dominance criterion (LDC), a new met hod of model selection developed by Pollak and Wales (1991), to four p opular demand equation systems, using a set of seasonally adjusted qua rterly Canadian data series. The results indicate that the almost idea l demand system (AIDS) dominates various alternative demand systems. T he LDC method is useful in assisting demand analysts in selecting the proper demand specification. The paper also includes tests for unit ro ots and cointegration and, based on these tests, an error corrections model is estimated and tested against the popular but not theoreticall y plausible linear approximate almost ideal demand system.