Abstract

Deviant interviewer behavior and interviewer falsification in face-to-face surveys can have substantial effects on survey results, including biased survey estimates and increased variance of survey estimates. Given that survey results are widely used in research and policy consulting, identifying deviant interviewers before the release of the data is therefore of highest importance. Statistical methods developed to identify such interviewers usually rely on indicators generated from survey data or paradata obtained during the survey. However, constructing these indicators requires extensive effort, which prohibits efficient data control procedures during or after the field period. Thus, the goal of this project is to develop methods to automate the construction of indicators of deviant interviewer behavior and falsification. This will be tested and implemented on multiple surveys conducted by the IAB.