This paper explores the presence of politically sourced sensitivity bias, which leads respondents to report alternative answers to their true beliefs, in Afrobarometer’s survey results. To this end, we run a set of regressions on Afrobarometer Round 7 data to explore mechanisms through which the perceived sponsor of the survey is associated with response patterns related to trust in political parties. Questions regarding individuals’ trust levels in political parties are prone to sensitivity bias insofar as political concerns can lead respondents to favour a certain response for strategic reasons. Moreover, if respondents believe that answers will be shared with the perceived sponsor of the interview, they are likely to tailor their answers to what they judge is the preferred answer of their perceived survey sponsor. We find a number of statistically significant results that suggest some sensitivity bias is present in the survey responses. In particular, we show that there are statistically significant differences in responses for self-reported trust in both opposition and ruling parties when the respondent perceives the government to be the survey sponsor compared to when they perceive the sponsor to be Afrobarometer. We also explore how these responses vary by country- specific characteristics, such as regime type. Reducing respondents’ variation in perception of the survey sponsor might alleviate biases throughout Afrobarometer’s interviews, and a set of methodological changes might mitigate the effects of this sensitivity bias in future surveys. While other studies have focused on potential bias from interviewer effects, our research suggests a new avenue for research, in Africa and beyond, on respondent perceptions of sponsorship.
Working paper