Conferences

Declining response rates and increasing efforts to convert reluctant respondents have led to research interest in the relationship between reluctance to respond and quality of survey answers. According to Krosnick (1991), respondents with lower motivation are more likely to engage in satisficing, i.e. skipping cognitive stages of response process, resulting in lower data quality. Previous research (e.g. Fricker, 2007) has supporting evidence of a negative relationship between the quality of answers and response propensity. The focus of the current study is on whether the relationship between respondent reluctance and satisficing remains significant after controlling for cognitive ability. The study is based on the European Social Survey data (third round) conducted in 2005 / 2006. The comparison of four countries (Belgium, the Netherlands, Norway and Sweden) is implemented. The quality measure is developed using latent class analysis with four indicators of measurement error, including “don’t know” responses, “extreme and middle” responses, straight-lining and inconsistent responses. Among two classes found, a class of satisficers is consistently more likely than the other class, optimizers, to provide answers with measurement error. A structural equation model is developed with reluctance and cognitive ability measures based on call record information, interviewer evaluation indicators and respondent’s demographic characteristics. The findings suggest that a negative relationship between reluctance and response quality is wholly mediated by level of cognitive ability. The structural invariance of the model indicates that the negative effect of cognitive ability on reluctance and the positive effect of cognitive ability on response quality have identical magnitudes across the four countries. The findings have important practical implications suggesting that data quality may be improved by decreasing the task difficulty for reluctant respondents.