The paper deals with measurement error, and its potentially distorting role, in information on industry and professional status collected by labour force surveys. The focus of our analyses is on inconsistent information on these employment characteristics resulting from yearly transition matrices for workers who were continuously employed over the year and who did not change job. As a case-study we use yearly panel data for the period from April 1993 to April 2003 collected by the Italian Quarterly Labour Force Survey. The analysis goes through four steps: (i) descriptive indicators of (dis)agreement; (ii) testing whether the consistency of repeated information significantly increases when the number of categories is collapsed; (iii) examination of the pattern of inconsistencies among response categories by means of Goodman's quasi-independence model; (iv) comparisons of alternative classifications. Results document sizable measurement error, which is only moderately reduced by more aggregated classifications. They suggest that even cross-section estimates of employment by industry and/or professional status are affected by non-random measurement error.