Assessing earnings and income data from a short web survey

Abstract

In short web-surveys, or in surveys that prioritize other content domains, earnings and income are often elicited with single questions or small question sets. This contrasts with the detailed question sets recommended for surveys that focus on earnings and income. We evaluate earnings and income data collected with a short question set in a series of recent web surveys: the Understanding Society COVID-19 Study. The fact that many COVID-19 Study respondents also contemporaneously answered the main annual Understanding Society survey provides individual- and household-level validation data. We find that measures of earnings and income in the COVID-19 Study are noisier than those from the main annual Understanding Society survey, and that there is evidence of systematic under-reporting for household totals.However, we find no evidence that measurement errors in the COVID-19 Study are substantively correlated with true values. We conclude that the COVID-19 Study collected useful data on earnings and income, and therefore, that simple collection of useful earnings or income datais feasible.