‘Many’, ‘Half’ or ‘One out of Two’?: Assessing the Effectiveness of Counter-Biasing Techniques in Reducing Self-Reported Turnout

When asked in surveys about their voting record, many respondents tend to report having gone to the polls, although they have not. The consequence is a large and significant turnout overreporting, with negative consequences upon the validity of many investigations in the field of voting behaviour. An individual’s tendency to exaggerate his/her voting record is usually the combined effect of several factors: social desirability bias, limitations of human memory, and acquiescence response bias. This article assesses the effectiveness of counter-biasing techniques in compensating for the social desirability bias and, consequently, reducing self-reported turnout. It compares the effects of three alternatives of ‘loaded’ turnout questions, providing variable information on the occurrence of non-voting behaviour (many people/around half of the people/one out of two people did not vote) and invariant face-saving response options. Self-reported turnout is compared against externally validated turnout. We employ a split-ballot survey experiment included in the 2009 Romanian Presidential Election panel study. Our analyses show that the ‘half’ counter-biasing formula is systematically more effective in reducing the social desirability bias than the alternatives, which perform rather similarly. When compared to validated turnout, it appears that, after applying the counter-biasing techniques, turnout overreporting due to social desirability remains at 5.7 to 8.6 percent.