An exploration of the application of PLS path modeling approach to creating a summary index of respondent burden

The potential effect on respondent burden is a major consideration in the evaluation of survey design options, so the ability to quantify the burden associated with alternative designs would be a useful evaluation tool. Furthermore, the development of such a tool could facilitate more systematic examination of the association between burden and data quality. In this study, we explore the application of Partial Least Squares path modeling to construct a burden score. Our data come from a phone-based, modified version of the Consumer Expenditure Interview Survey in which respondents were asked post-survey assessment questions on dimensions thought to be related to burden – e.g., effort, survey length, and the frequency of survey requests (Bradburn, 1978). These dimensions served as the latent constructs in our model. We discuss model development and interpretation, assess how the measured items relate to our latent constructs, and examine the extent to which the resulting burden scores covary with other survey measures of interest.