Stochastic Trends, Demographics and Demand Systems

Techniques for determining the number of stochastic trends generating a set of non-stationary panel data are applied to budget shares for a number of commodity groups from the Family Expenditure Survey (FES) for the UK for the years 1973-2001. It is argued that some stochastic trends in macro data are generated by the aggregation of fixed demographic effects in the micro data. From cross section data, fixed effect coefficients are estimated which incorporate both age and income distribution effects. The estimated coefficients are combined with age proportion variables to form a set of I(1) indices for broad commodity groups which are then incorporated into a system of aggregate demand equations. The equations are estimated and tested in a non-stationary time series setting.