Imputation of missing data: testing a model-based approach

One of the most pressing issues surrounding media and marketing research is the decline in respondent cooperation. Despite years of experimentation and implementation of unique techniques intended to cajole prospective respondents to participate in surveys, and maximize their level of completion, as an industry we are at best holding our own. With increased respondent confusion in the U.S. over the differences between legitimate survey research versus telemarketing, direct marketing, and fund-raising, researchers are encountering a suspicious, uncooperative, and often angry population. As an industry, we have compounded the issue by continuing to increase the amount of information and the level of detail we ask respondents to reveal. Readership researchers learned long ago that we needed to separate the critical element of print measurement from the onerous task of completing detailed product purchase behavior and other measures in order to achieve the highest achievable levels of cooperation. Even with this separation, we clearly have seen evidence of respondent bias or laziness when faced with the burden of overly complicated or lengthy readership questions.

Download Paper (PDF)