Publication Detail

Weighting Methods for Choice-Based Panels with Correlated Attrition and Initial Choice



Download PDF
This paper develops a weighting method that can be applied to choice-based panel samples. The need to study infrequent travel choices has motivated the use of choice-based sampling procedures where sample entities are chosen based on endogenous choice variables. As a choice-based sample is not representative of the population, unbiased inferences can be drawn only after applying weights to the sample. The issue is further complicated if a choice-based sampling technique is employed in a panel survey, where the same behavioral units are observed over time. A choice-based panel sample would need additional treatment for selective attrition, the non-random leaving of panel survey participants. While past research has developed weights to treat for choice-based sampling and attrition separately, this study is the first attempt to account for both issues simultaneously. In this study, weights are developed for a choice-based panel sample from the Puget Sound region to obtain unbiased population estimates of transitions in mode choice. This is accomplished by estimating a bivariate probit simultaneous equation system of mode choice and attrition.
Published in Transportation and Traffic Theory, ed. C.F. Daganzo, Elsevier Science Publishers B. V.