Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
the Kansas City area in 2004. Volume 2 includes a paper on this topic.2 The following points were covered in her presentation. ⢠GPS has been used in at least 12 regional travel sur- veys over the past 10 years. The primary use of GPS in these surveys has been to audit trip reporting, to assess the level of trip underreporting, and to develop correc- tion factors for the data. Most applications have used in- vehicle devices to obtain data on both the driver and passengers. A few surveys, such as the Portland, Oregon, survey, focused only on the driver. Thus, GPS has been used mostly for passive data collection, although per- sonal digital assistants (PDAs) were used with surveys in Los Angeles and Ohio. Different processes have been used for detecting missed trips. Rates for missed trips have ranged from 5% in Reno, Nevada, to 81% in Laredo, Texas. Also, time thresholds and other consider- ations have varied. ⢠Key findings can be identified from the use of GPS in the 12 regional travel surveys. First, GPS participants are a select group of respondents. GPS participants tend to have higher incomes and higher home ownership rates. Second, different methods have been used to process GPS data. These methods appear to influence the rates of trip underreporting. The instructions provided to participants also appear to influence underreporting. In most cases, participants are instructed not to report trips out of a specifically defined area and trips for com- mercial purposes. ⢠A research project was conducted to examine the use of GPS in the 2004 Kansas City regional household survey. The project had four research objectives. The first objective was to identify the likelihood and the magni- tude of trip underreporting at the person level. The sec- ond objective was to develop a joint model that recognizes the implicit relationship between the likeli- hood and the level of underreporting. The third objective focused on examining a comprehensive set of variables related to driver demographics, driver travel characteris- tics, and driver adherence to survey protocol. The final objective was to identify methodological improvements to reduce underreporting in future travel surveys based on the research results. ⢠The Kansas City regional travel survey was con- ducted in 2004 under the sponsorship of the Mid- America Regional Council and the Kansas and Missouri Departments of Transportation. The survey included 3,049 randomly sampled households, 7,570 persons, and 32,011 trips. The GPS component involved equip- ping the vehicles of 294 households with GPS devices to record all vehicle travel during the survey period. Both computer- assisted telephone interviewing (CATI) and GPS data are available for 228 of the 294 households. The analysis focused on the 228 households and the cor- responding 377 drivers and 2,359 vehicles trips. The GPS participants were more likely to own more vehicles, to have higher incomes, and to live in single- family dwellings than non- GPS participants. ⢠Of the 377 drivers participating in the GPS com- ponent, 71% accurately reported all travel in their CATI survey, while 29% had at least one nonreported trip. Of those drivers who underreported trips, 49% missed one trip, 20% missed two trips, 10% missed three trips, and 20% missed more than four trips. ⢠The hypothesis underlying the empirical analysis was that trip underreporting is influenced by three major factors: the demographic characteristics of the driver, the characteristics of the trip, and the level of adherence to the survey protocol. Examples of driver demographics include age, type of household, employment status, and the number of vehicles in the household. Trip character- istics included the total number of trips, the average trip distance, and the level of trip chaining. Elements associ- ated with following the survey protocol included use of the travel survey to record all trips and talking directly with the interviewer. ⢠A joint binary choice- ordered response discrete model of underreporting and an ordered- response model for level of underreporting among underreporting indi- viduals were developed by adopting a systematic proce- dure of eliminating statistically insignificant variables. The exogenous inputs to the model are classified accord- ing to the three areas of influence. The binary choice- ordered response discrete model includes two equationsâ one addressing likelihood and one addressing magnitudeâ and accounts for correlation in error terms. ⢠The results of the modeling effort help identify underreporting tendencies with household travel sur- veys. The underlying mechanism that represents whether an individual underreports is different from the mecha- nism that determines the level of underreporting. There are factors that influence both the underreporting propensity and the propensity associated with the level of underreporting. The effect of driver demographics indicates that adults under the age of 30, men, individu- als with less than a high school education, unemployed individuals, individuals working in clerical and manu- facturing positions, and individuals working at residen- tial, industrial, and medical facilities are more likely to underreport trips than other respondents. Driver travel characteristics that influence underreporting include higher trip rates on the survey day, traveling long dis- tances per trip, and high levels of trip chaining. Also, drivers who do not use their travel diary to record their travel are more likely to miss trips than those who use the travel diary. Based on elasticities, the most important 35SURVEY METHODS 2 See Bricka, S., and C. Bhat. Using Global Positioning System Data to Inform Travel Survey Methods. Volume 2, pp. 89â93.
