Asking income and consumption questions in the same survey: what are the risks?

Sample surveys providing high quality information on both total household expenditure (consumption) and income are not commonly available. Nevertheless, surveys focusing on income usually do collect some information on expenditure data. A main drawback of this practice is that it could let some researchers think that both sets of information have similar accuracy, as they are derived from the same survey. This paper provides an empirical investigation of the consequences of such an assumption. We draw on the Survey of Household Income and Wealth (SHIW, thereafter) as a case study, since it collects information on both income and consumption. We combine this survey with the information coming from other surveys that are assumed to be more reliable than the SHIW for specific items. On average, we find that the underestimation of household income is lower than the one relating to consumption. As a consequence, in the survey saving rates are likely to be overestimated. We also find evidence that measurement error in income data is proportionally higher for high incomes. This does not appear to be the case for consumption data. Household saving is likely to be overestimated, especially for households in the low income classes. Finally, we find evidence that measurement error may bias the relationship between household savings and its determinants.