Do Surveys Overestimate or Underestimate Socioeconomic Differences in Voter Turnout? Evidence from Administrative Registers

Surveys generally overestimate the overall level of voter turnout in elections due to both the misreporting of voting and nonresponse. It is sometimes argued that socioeconomic differences in turnout are exaggerated in surveys because social desirability has a more pronounced effect on eligible voters in more advantaged socioeconomic positions. However, the contribution of nonresponse bias has not been taken into consideration in these assessments. Using a register-linked survey with information on the education, occupational social class, income, and voting in the 2015 Finnish parliamentary elections of both respondents and nonrespondents, this study shows that nonresponse bias leads to not only a larger overestimation of the overall level of turnout than social desirability, but also an underestimation of educational, social class, and income-related differences in the propensity to vote. Socioeconomic differences in the probability of voting in register-based data were at least two-thirds larger than differences obtained when using standard survey techniques. This finding implies that socioeconomic inequality in electoral participation is a more pressing social problem than previous evidence might indicate.