Abstract
The objective of the study is to assess the nature and magnitude of reporting errors and attrition biases in the first two waves of a household panel survey data set that consists of weekly trip diaries. This paper investigates the relationship between reporting errors in a first-wave diary and the subsequent decision to participate in a second-wave survey and the relationship between the same first-wave errors and the errors in the second-wave survey for those households which chose to participate. An econometric procedure is developed which accounts for attrition bias and relates reporting errors from the two waves. The model system offers a mechanism by which relationships among mobility, trip reporting errors, and attrition behavior in a panel survey can be statistically evaluated. It consists of three elements: wave-one trip equation, attrition probability model, and wave-two trip equation. The conditional distribution of the error terms of these three components is evaluated assuming chronological dependency among the errors. The result is used to develop a correction term for attrition bias and a term which accounts for the dependency between the wave-one error and wave-two error. Results from empirical application of the model system are reported together with the use of the attrition model to develop an efficient sample weighting factor that fully utilizes the information in the survey results.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have