How Worried Should We Be? The Implications of Fabricated Survey Data for Political Science

Abstract

Abstract Surveys are ubiquitous in the study of politics, making enumerator fabrication a critical issue. A prevailing view is that faked interviews affect inferences drawn from compromised datasets. Researchers have generated theories about how fabrication might affect inferences. Yet, speculation has outpaced systematic testing. We leverage a rare dataset to address this gap: a national face-to-face survey in Venezuela in which a uniquely high volume of falsified interviews was detected, canceled, and replaced. Comparing the verified and fraudulent datasets, we find that descriptive inference is sometimes affected, but correlational results hold, even in a dataset with an unusually high-fabrication rate. Enumerators largely fabricate plausible data. Though still egregious, enumerator fabrication may not constitute a grave threat to political science research.