Imputation of Pension Accruals and Investment Income in Survey Data

This paper explores the problem of augmenting the data in the UK's Living Costs and Food Survey in order to address two issues. First we are concerned with broadening the definition of income to include accrual of pension rights and secondly we aim to address the point that investment incomes are materially underrecorded. We draw on the Wealth and Assets Survey to address the first point and the Survey of Personal Incomes for the second. We present an approach to stochastic imputation which largely replicates the distributional properties of the source data and show how it can be adapted to address the issue of covariance between the variables imputed. Our initial results suggest that imputation of pension accruals raises both the Gini coefficient and the geometric mean of equivalised household income materially, while the effects of imputing investment income are more marked on the Gini coefficient than on the geometric mean of household income.