The Validity and Reliability of Retrospective Measures of Childhood Socioeconomic Status in the Health and Retirement Study: Evidence from the 1940 U.S. Census

Abstract

Objectives: Retrospective measures of childhood socioeconomic status (SES) in cohort studies of aging that first observe people late in life—such as the Health and Retirement Study (HRS)—are widely used. However, their measurement validity and reliability are unknown. We assess the reliability and validity of the HRS’s retrospective measures of parental education and childhood family finances. Methods: We use records for 6,343 HRS sample members who were children in 1940 that have been linked to records from the complete-count 1940 U.S. Census. We assess interrater reliability by comparing (a) retrospective reports of childhood SES collected from sample members in the 1992–2018 HRS to (b) prospective measures of parallel concepts collected from HRS sample members’ parents in the 1940 Census. We assess predictive validity by comparing the results of analyses that model later-life outcomes as a function of childhood SES as measured both prospectively and retrospectively. Results: Interrater reliabilities of retrospective measures of parental education are high; however, the same is not true of the retrospective measure of childhood family finances. Both retrospective and prospective measures of childhood SES are predictive of later-life outcomes, and with similar strengths and directions of associations for most outcomes. Discussion: Researchers who rely on retrospective indicators of childhood SES from the HRS should be aware of their measurement properties. They are measured with error, and that error modestly attenuates estimates of their associations with later-life outcomes. However, prospective and retrospective measures of childhood SES have similar predictive validity. These findings should reassure researchers who rely on retrospective measures of childhood SES in the HRS and similarly designed surveys.