How wrong were we? Dependent interviewing, self-reports and measurement error in occupational mobility in panel surveys

Abstract

Occupation is a central concept in sociology and economics and individual change in occupation is of major importance to literatures on wage determination, human capital, careers and social mobility. The collection of occupational data in surveys, particularly panel surveys, is challenging due to measurement error, and observed rates of occupational mobility are argued to be overestimated. We use a methodological discontinuity in the collection of occupational data from independent interviewing (respondents are asked to describe their occupation each year) to dependent interviewing (respondents are shown their previous response and only asked to describe their occupation if this has changed) and information on self-reported occupational changes in two panel surveys to estimate the degree of error in occupational mobility in panel data. We also test whether observed patterns differ by the level of aggregation of occupational classifications and examine the external validity of different measures of occupational mobility through their predicted impacts on selected labour market outcomes. Results indicate that occupational mobility is dramatically lower under dependent than independent interviewing (particularly for highly disaggregated occupational classifications) and that there is an evident mismatch between respondents’ self-reports of occupational switches and mobility measures inferred from changes in occupational codes. The impacts of occupational changes on earnings and job satisfaction are more consistent with theoretical predictions under dependent than independent interviewing and when occupational mobility is inferred from respondents’ self-reports. These findings have important implications for survey design, question the validity of existing studies on occupational change and call for further research.