Household travel surveys are typically associated with a wide range of biases resulting from various factors including undercoverage, nonresponse, data collection method, and surveying methodology. To develop accurate estimates of population parameters and, more generally, to draw inferences about the population that was sampled, analytic weights are needed. Without the use of analytic weights, population estimates are subject to biases ofunknown, possibly large, magnitude. Consequently, analytic weights are crucial to obtain survey estimates with minimal bias. While most agree on the need for weights to adjust for the various biases inherent in imperfect data collection, the actual process and variables to be considered varies based on the type of survey and the planned application for the data. Thisis particularly true when we take into account what is considered as required elements in weighting by statisticians who primarily work on social policy studies versus those considered by travel demand modelers. In thispaper, we test and evaluate alternative weighting methodologies in order to identify the most appropriate methodology for household travel surveys.Specifically, we compare two weighting methodologies—a comprehensive, all-inclusive method (preferred by social policy statisticians) and a simplified method (preferred by travel demand modelers), to assess the best method for application to household travel surveys. We find that both approaches yield similar results. Thus, depending on the planned application of the data, the simplified method is more cost-effective and takes less time,streamlining the time between data collection and application.
Abstract