Can you trust survey responses? Evidence using objective health measures

- We investigate measurement error in objective survey variables.
- Major medical operations are used as a case study.
- Cross-checking self-reports with administrative data we find large reporting errors.
- These errors are not explained by observed individual characteristics.
- Hence coefficients are not biased when the survey variables are used as outcomes.

We investigate the common assumption in applied research that reporting errors are negligible in variables where there is no clear incentive for misreporting. Using major medical operations, we find high misreporting rates, but the coefficients of their predictors remain unbiased.