Quantifying key errors in household travel surveys

Identifying and minimizing potential errors in household travel surveys can facilitate collecting more representative and accurate data. Through a comparison of two recent travel surveys with census data, this paper presents how sampling, noncoverage, nonresponse, and measurement errors work their way into surveys. The 2009 National Household Travel Survey (NHTS) Add-On in Virginia was implemented with a comprehensive survey instrument and random-digit-dial (RDD) sampling. The 2008 National Capital Region Household Travel Survey collected behavioral data with a concise instrument, while adopting address-based sampling (ADD). Focusing on a common area of Northern Virginia, this study examined differences in sociodemographics and travel behavior of the extracted samples (N = 597 and N = 3,581, respectively). Results show that the ADD survey collected data on more single-person households, younger individuals, and Hispanics and Mexicans, which are generally identified as hard-to-reach groups. A comparison of the two samples with the census data shows that the ADD sample was more representative of the population and area, partly because of the inclusion of mobile phone-only households (28%), which were not fully covered in RDD. To quantify a measurement error, this study estimated rigorous statistical models in regard to reported trip frequency. Results show that the NHTS captured 10% more trips, partly as a result of diary instructions and the presence of walking and biking questions in the instrument. Details of other errors and implications for reducing key survey errors are discussed.