Corrupted Estimates? Response Bias in Citizen Surveys on Corruption

Abstract: There is now a near consensus among researchers about the destructive consequences of corruption. In the light of this, measuring corruption has become a global industry. An important and commonly used data source are several large-scale multi-country projects that survey citizens directly about their perceptions and experiences of corruption. However, we still know little about the quality of many of these measures. This paper deploys a large survey experiment to investigate two potential sources of bias in indicators based on citizens' perceptions and experiences of corruption, stemming from political bias and sensitivity bias. First, I draw upon research on economic perceptions and argue that respondents are likely to respond in a political manner when asked how they perceive the level of corruption in their country. I test this argument by experimentally priming respondents' political affiliations before asking for their perception of corruption. Second, I argue that standard questions probing peoples' corruption experiences are likely to be subject to sensitivity bias. I test this second argument using by constructing a list experiment. Overall, the results show strong and predictable sources of response bias.