Using POLIMOD to evaluate alternative methods of expenditure imputation

Micro- level information on both expenditure and income is useful for a wide range of purposes: to conduct investigations into standards of living, for example. It is necessary for the analysis of the combined effects of direct and indirect personal taxes. However, it is unusual to have one data source with high quality information on both, particularly since the information often needs to be at a detailed level of disaggregation (income by source and recipient; and expenditure by category of goods and services). Our particular problem is to create a synthetic micro-dataset containing information on household incomes and expenditures by statistically matching independent data sources.\n We are able to examine the success of two matching techniques and a number of variants to one of the methods with the Family Expenditure Survey (FES). The UK is quite unusual in having a household survey with high-quality data on both incomes and expenditures. In many other countries Household Budget Surveys contain very limited income information of relatively low quality, and income surveys do not contain comprehensive expenditure information. Thus the UK data offer us the opportunity to experiment with expenditure imputation and to evaluate our results against actual expenditure information. For experimental purposes we confine ourselves to one set of 27 expenditure.