Aggregation, Distribution and Dynamics in the Linear and Quadratic Expenditure Systems

Using Canadian data (1965-86) we confirm and extend Stoker's (1986) results on the rule of distributional effects in demand systems. The confirmation consists of evidence from the LES model showing that distributional effects are statistically significant and can displace AR(1) dynamics in the disturbances. The extension is made to the QES model and an argument is advanced that standard habit formation dynamics may reflect omitted distributional effects. The evidence supports this conjecture. This suggests that we may have been drawing the wrong conclusions from expenditure studies. Rather than inferring dynamic behaviour we should have been concluding that these models are misspecified.