In this paper, we examine the effect of interview length on wave nonresponse in a longitudinal survey, controlling for respondent-specific characteristics known to affect survey response. We use the National Longitudinal Survey of Youth (NLSY), a sample of over 10,000 individuals who were 14-22 years old when first interviewed in 1979. These individuals have been interviewed annually every year since then, providing 16 years of data. The interviews have been conducted in person in all years except one. Unlike the CPS or SIPP, the NLSY does not allow proxy responses. The NLSY attempts to interview virtually all living respondents each year. Over the years, the length of the interview has varied. It also varies substantially across individuals in the sample within years.
A transition probability model is estimated using hazard equations. Holding constant personal, demographic, and environmental factors known to influence survey response as well as several measures of respondent attitude and cooperation, we find that longer interview length is associated with sample retention. Hypothesizing that interview length may proxy for some uncontrolled dimension of respondent cooperation, an alternative measure to interview length, namely the number of questions asked, was constructed. Reestimating the hazards with this variable generates similar findings.
We conjecture that survey length, whether measured in minutes or number of questions asked, measures the saliency or applicability of the survey to the respondent. Those respondents who possess the characteristics most important to the content of the survey have the longest interviews but are also the most interested. The policy prescription we propose is to design survey instruments which include sets of questions applicable to all respondents, focusing less on the average length of the interview and more on the range of potential interview lengths.