Outcome-related nonresponse bias, not systematic misreporting, explains overestimation of vaccine uptake in COVID-19 surveys

Authorities increasingly rely on surveys to generate real-time evidence on population behavior during epidemics, yet the reliability of such evidence is debated. We combine data from a large-scale Danish survey (n = 393,375; 50.42% women; mean age 49.46) conducted during the COVID-19 pandemic with individual-level administrative registries to identify sources of error in survey-based estimates of vaccine uptake. Comparing survey responses and actual vaccine uptake, we find no evidence of systematic overreporting. We evaluate nonresponse bias by comparing actual vaccination behavior among respondents with the full pool of 1.68 million invitees. Despite weighting for sociodemographic characteristics, respondents are 5.99 percentage points more likely to initiate vaccination and 12.91 percentage points more likely to receive a booster than the invited sample. These results indicate that outcome-related nonresponse bias-not misreporting-is the dominant source of survey error, implying that even weighted estimates may overstate vaccine uptake during epidemics.