An Evaluation of HRS Sample Weights
21 Pages Posted: 24 Aug 2013
Date Written: August 23, 2013
The Health and Retirement Study (HRS) documentation describes HRS sample weights as post-stratified by age, race/ethnicity, marital status and gender (Ofstedal, et al., 2011). Yet, we find systematic differences between distributions of these variables in the HRS and the Current Population Survey (CPS). In particular, in the HRS, minorities totals are under-estimated in older cohorts and over-estimated in younger cohorts in 2004, 2006 and 2008 survey waves. We hypothesize that these discrepancies result from of a simplified post-stratification approach, with post-stratification by race/ethnicity not being done jointly with all other factors. Given the HRS study design, however, a joint post-stratification is called for.
Keywords: Health and Retirement Study, sample weights, post-stratification, race
JEL Classification: C82, C83
Suggested Citation: Suggested Citation