Estimating underreporting of consumer expenditures using Markov latent class analysis

Abstract

This paper examines reporting in specific consumer item categories (or commodities) and estimates expenditure underreporting due to survey respondents who erroneously report no expenditure in a category. Our approach for estimating underreporting errors is a two-step process. In the first step, a Markov latent class analysis is performed to estimate the proportion of consumers in various subpopulations who fail to report their actual expenditure in a particular commodity. Once this proportion is estimated, the dollar value of the missing expenditure is estimated using the mean expenditure of those in that subpopulation that did report an expenditure. Finally, the estimates are evaluated and discussed in light of external data on expenditure underreporting.