Abstract
Many people are concerned about what citizens know about politics. Surveys provide widely cited
measures of such knowledge. Do unusual aspects of survey interviews affect these measures? An
experiment on a nationally representative sample of over 1200 Americans reveals an answer. Respondents
are randomly assigned to one of four groups. The control group answers questions in a typical survey
context. Other respondents are given extra time to answer knowledge questions, a small monetary
incentive for answering these questions correctly, or both. These variations produce significant
performance increases for almost every knowledge question we asked. Overall, average knowledge scores
in the treatment groups are 10-20% higher than in the control group. The treatments also cause significant
reductions in the magnitude of error entailed in incorrect responses. Our results imply that more reliable
measures of political knowledge require improved elicitation strategies.
2