Do retrospectively-collected employment measures match prospective reports?

• In Event: Measurements and Methods for Understanding Dynamics in Occupations, Work, and Organizations

Abstract

Demographers and social scientists have long been interested in the accuracy of reports made about employment histories and the conditions people experience on the job. Retrospective reports of start and end dates of employment spells and retirement, and of the nature of employment conditions in a particular job, are frequently used because collecting prospective survey data over decades is time consuming and expensive. This study examines the consistency of retrospective and prospective reports about employment status and job characteristics across midlife. We use novel data from an adaptation of the SHARELIFE retrospective life history (RLH) interview – the ACLLIFE interview – that was collected in 2021 from respondents of the long running Americans’ Changing Lives (ACL) longitudinal study (1986-2019). Preliminary results show that about 81-87% of respondents consistently report that they were working (both reports yes) or not working (both reports no) in a given survey year when comparing retrospective and prospective reports. Few report being currently unemployed and consistency ranges from 92-99%, increasing as prospective survey waves grow closer to the retrospective life history interview. Consistency in reporting being retired falls sharply from 98 to 80% over survey waves from 1986 to 2019, as more individuals enter retirement ages, and the fraction making false negative (said yes prospectively but no retrospectively) or false positive (said no prospectively but yes retrospectively) reports rises. Keeping house is reported consistently by from 78 to 91% of respondents, depending on survey year, with false negatives higher than for any other activity (8-19% of respondents). We find evidence that consistency varies by gender for all activities, especially drastically for reports of keeping house as an activity, which is almost entirely reported by women and for whom false negative reports make up 15-34% of all types of consistency or inconsistency. Planned next analytic steps are discussed.

Authors

• Sarah Burgard, University of Michigan-Ann Arbor

• Janet Wang, University of Michigan

• Lucie Kalousova, Vanderbilt University

• David Yang, University of Michigan