Analysis of trip item non-response bias in the Houston-Galveston area council household travel surveys

Abstract

Information obtained through the collection and analysis of origin-destination travel surveys is used to develop trip rate information for input to travel demand models. Three major sources of systematic error can exist relative to typical sample survey data sets: survey non-response, item non-response, and inaccurate reporting. This thesis presents the results of a study examining the effect of item non-response on trip rates for a household survey conducted by the Houston-Galveston Area Council from October 1994 to June 1995. The variables of age cohort, household size, household income, and auto availability were used to stratify the trip rates. These four variables were used individually and in combination to examine internal vehicular person trip rates for home based work, home based other, non-home based, and all three trip purposes combined and for internal auto driver trip rates for all three trip purposes combined. The results were evaluated to determine the bias introduced into the trip rates by item non-response with regard to the quality of the collected data. No attempt is made to quantify the bias for the purpose of establishing or recommending correction factors or adjusted trip rates for the survey data. Statistical evidence in the form of significantly different trip rates in the removed questionable households indicate that item non-response bias may have resulted in a change in mean trip rates. Empirical evidence was also found to indicate that bias due to trip item non-response caused a change in total trips and total vehicle miles traveled were observed after the removal of the questionable samples. A reduction in the variation of the sample data within stratification cells also resulted from the removal of the questionable samples.