Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Since 1999, the Panel Study of Income Dynamics (PSID) has asked respondents to report their health insurance coverage in the two years since the previous interview. These questions are somewhat unusual among major health insurance surveys in their use of a relatively long recall period. In order to investigate the quality of the resulting data, I benchmark PSID health insurance data from 1999 through 2005 to data from the Medical Expenditure Panel Study, which are treated as "truth" for purposes of the benchmarking exercise. I find that respondent reports of private coverage are very accurate. PSID estimates of the fraction of the population with both private and public coverage during a two-year reference period are low; respondents who had both types of coverage apparently fail to mention having had public coverage and instead report having had only private coverage. This problem is concentrated among children and the elderly (ages 65 and older) – the groups with the highest true rates of having both types of coverage. As a result of this under-reporting, PSID estimates of the fraction with any public coverage during a two-year reference period are low. On the other hand, PSID estimates of the fraction with public coverage only during a two-year reference period are quite accurate. As a result, PSID estimates of the uninsured are very accurate, as long as "uninsured" is defined as individuals without insurance for the entire reference period rather than individuals without insurance at some point during the reference period. These results should provide useful background for analysts working with the new health insurance questions in the PSID. I also discuss briefly the health insurance questions in the earliest years of the PSID (1968 – 72) and the more recent health insurance data in the Child Development Supplement to the PSID.