As paradata gains prominence in survey research as an effective tool for addressing nonresponse bias, researchers are exploring different types of paradata to evaluate their potential for nonresponse adjustment or directing fieldwork. One type of paradata sometimes collected in interviewer administered surveys is an interviewer evaluation of the case’s willingness to participate in the survey. Such “willingness ratings” were collected on each call attempt with contact on two surveys, one done over the telephone and the other done face-to-face. These data allow us to investigate how interviewers use these ratings and if they are effective at predicting the outcomes for the cases, thereby displaying potential for addessing nonresponse.

The analysis first investigates how interviewers use these ratings by closely examining the distribution and trends in assignment of the ratings. The analysis will test various interviewer characteristics to explain different applications of the rating scale. We then use multinomial models to explore whether the willingness ratings are more effective than other information contained in the call history data, such as number and outcome of calls, at predicting cases’ final outcomes. Both steps of the analysis will account for the clustering of ratings within cases, which are within interviewers.

If the ratings can replace the call outcomes or be used in conjunction with them to more effectively predict case outcomes, they would be useful to survey practitioners when making fieldwork decisions such as case prioritization or introducing protocol changes for responsive design.