Accounting for Trends in the Almost Ideal Demand System

Abstract

Flexible functional forms of indirect utility and expenditure functions are frequently used in approximating the behavior of utility maximizing consumers to arrive at demand systems that can be easily estimated. A common finding in time series estimations of the Almost Ideal Demand System is strong persistence in the estimated residuals. This paper suggests two explanations for this result. First, the functions used to approximate total expenditure does not allow for the possibility of economic growth. Hence when the data on expenditure have trends, the inadequacy of the approximation results in residuals that are serially correlated. Second, when the economy grows and/or prices trend at different rates, Stone's price index provides a poor approximation to the theoretically appropriate price variable. The consequence is also reflected in the error term. Simulations are used to illustrate these arguments and cointegration is proposed as a guide to model specification.