In recent years, there has been an increasing interest in unethical work behavior. Several types of survey instruments to collect information about unethical work behavior are available. Nevertheless, to date little attention has been paid to design issues of those surveys. There are, however, several important problems that may influence reliability and validity of questionnaire data on the topic, such as social desirability bias. This paper addresses two important issues in the design of online surveys on unethical work behavior: the response scale for questions regarding the frequency of certain types of unethical work behavior and the location of the background questions in an online survey. We present the results of an analysis of a double split-ballot experiment in a large sample (n = 3,386) on governmental integrity. We found that, when comparing response scales that have labels for all categories with response scales that only have anchors at the end, the latter provided answers with higher validity. The study did not provide support for the conventional practice of asking background questions at the end. © 2013 Springer Science+Business Media Dordrecht.