Nonresponse Bias and Turnout Misreporting in a Refugee Survey: Evidence from Randomized Participation Triggers

Survey nonresponse is a major obstacle to understanding refugee integration, especially when key outcomes, such as voter turnout, are also prone to misreporting. We study what we can learn about both problems by including multiple randomized participation triggers in a refugee survey linked to administrative data. We find that the two biases interact: respondents are positively selected to the survey on integration, and more integrated respondents also over-report turnout. Moreover, the participation triggers attract different types of misreporters to the sample. Higher monetary incentives attract refugees who are more prone to over-report turnout, increasing the bias in self-reported turnout, while an extra reminder attracts refugees who misreport less. Using the timing of responses, we show that the reminder result reflects genuine changes in who responds rather than in what respondents report. Finally, we find that methods relying on the missing-at-random assumption to adjust for nonresponse bias are ineffective, while those exploiting randomized incentives seem to require impractically large effects on response rates to be informative. We recommend that surveys of hard-to-reach populations use a diverse set of randomized triggers to detect bias, and link to administrative records wherever possible.