How We Travel: A Sustainable National Program for Travel Data (2011)

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.

3 New Approaches for Meeting Travel Data Needs T his chapter explores new approaches for meeting travel data needs. It begins with a summary of key barriers to survey data collection. Then, opportunities for addressing these challenges are discussed. These opportunities range from greater use of technology for more accurate and timely data capture, to alternative methods of data collection that have the potential to yield improved understanding of travel behavior and more stable cost and staffing requirements than are obtained through traditional large-scale periodic surveys. The discussion includes the pros and cons of these approaches, drawing on examples of their use. The chapter ends with a series of findings regarding implications for travel data programs. Barriers to Survey Data Collection Travel data are collected using a wide range of means, from surveys, to administrative records (e.g., the rail Carload Waybill Sample), to automated data collection (e.g., use of Global Positioning System [GPS] tracking). This section focuses on survey data because the flagship passenger and freight travel surveys—the National Household Transportation Survey (NHTS) and Commodity Flow Survey (CFS), respectively—are the primary sources of multimodal travel data. In examining barriers to the collection of travel data with surveys, it is important to distinguish between the different types of respondents. Households and individuals are the   45  

46  How We Travel: A Sustainable National Program for Travel Data units surveyed to obtain data on personal travel, whereas businesses (e.g., establishments, shippers, carriers) are surveyed to obtain data on freight movement. Each target group poses different challenges. Personal Travel Data The past several decades have seen a general decline in the willingness of the public to respond to surveys; at best, response rates have remained constant (Zmud 2010a). In the telephone survey area, particularly relevant to the NHTS, response rates have fallen steadily over time to very low levels (see Curtin et al. 2005). Response rates for other survey modes have also either declined or remained relatively constant, but at much greater cost. A recent and visible example of steady response rates at greatly increased cost is the 2010 U.S. census. The mail portion of the census achieved a 74 percent response rate, matching the response rate of the 2000 census. Both the 2000 and 2010 censuses had substantially higher response rates for the mail portion than the 1990 low of 65 percent (Billitteri 2010; Zmud 2010a). However, achieving this response rate came at a significant cost. Overall, the mail and subsequent face-to-face follow-up cost was $13 billion, representing the most expensive census ever conducted (GAO 2010).1 A large share of this cost was allocated to efforts to boost response rates, including an extensive media campaign emphasizing the importance of the census to local communities, use of the Internet to publicize the importance of the census for the entire country, and a significant simplification of the census instrument itself to a brief 10-question form to reduce respondent burden (Billitteri 2010).2 Travel surveys have much less visibility and far fewer resources than the census. The typical cost of a local travel survey for a large metropolitan area, for example, is about $2–4 million, or about $150 per surveyed household, and typical response rates are generally in the range of 30–40 percent (Zmud 2010a).3 The response to the initial recruitment 1. The cost was twice the $6.5 billion cost of the 2000 census, or 1.57 times the 2000 cost in inflation-adjusted dollars (GAO 2001; GAO 2010). 2. The simplification was possible because the “long form” Census questionnaire, administered to approximately one of every six households in the previous censuses, was replaced with a separate con- tinuous survey—the American Community Survey (ACS). 3. The total cost figures were reported by Ronald Kirby, Transportation Director of the Washington Area Council of Governments, in a briefing to the committee at its second meeting (February 18, 2010) for recent household travel surveys conducted for the Washington and Baltimore metropolitan areas. The estimated cost of the 2010 Travel Behavior Inventory, a local travel survey conducted by the Minneapolis-St. Paul metropolitan area every 10 years in conjunction with the decennial census, is $4 million (information provided by committee member Timothy Henkel, Aug. 2010). 

New Approaches for Meeting Travel Data Needs  47  for the 2009 NHTS was only 23 percent, nearly 60 percent less than the 56 percent response rate for the 2001 survey.4,5 Of those households that did agree to participate, however, 80 percent completed the survey, some 10 percentage points higher than the 70 percent completion rate for the 2001 survey, reflecting in part the increased training and effort involved in ensuring that initial recruits would actually complete the survey.6 What accounts for the decline in willingness to participate in surveys? The decline has been attributed to a wide range of societal factors and technological changes. Less discretionary time has reduced the moti- vation of respondents to cooperate and limited opportunities for contact, particularly at home (Lepkowski 2010a; Zmud 2010a). Norms of civic duty and cooperation for the common good are less powerful motivators than in the past, affecting participation in publicly sponsored surveys in particular. Declining participation has also been the overall result of declining levels of trust in government (Pew Research Center 2010), greater concerns about privacy, the rise of telemarketing and the corresponding introduction of no-call registers, and the ability to screen out calls (Stopher 2009). A random telephone survey of U.S. residents, for example, conducted since 1982 by the Council for Marketing and Opinion Research, a nonprofit organization working on behalf of the survey research industry to improve respondent cooperation, found that the percentage of those who had “refused to participate in a survey in the past year” had risen from 15 percent in 1982, to 31 percent in 1992, to 45 percent in 2001 (Zmud 2010a). Finally, the population’s increased mobility and location in large metropolitan areas has made it more difficult both to find and to contact respondents. The most difficult populations to reach are males; young people; the less well educated; nonwhites; and the nonemployed, including students (Princeton Survey Research Associates 2008). Tech- nological changes have played a role as well, particularly the use of cellular phones and the Internet, which have increased the difficulty of reaching younger, minority, and lower-income groups through traditional survey methods. A growing number of households, for example, no longer use landline telephones, still the primary method for conducting the NHTS household interviews. 4. This is the response rate reported to the Office of Management and Budget. 5. T. Tang, Federal Highway Administration (FHWA), personal communication, June 11, 2010. 6. T. Tang, FHWA, personal communication, June 11, 2010. 

48  How We Travel: A Sustainable National Program for Travel Data Freight Travel Data Collecting freight travel data typically involves the private sector and a different set of challenges for data collection managers. It is difficult to generalize about response rates because some surveys, such as the CFS, are mandatory.7 The collection and reporting of other administrative data, such as data on rail carload waybills and on waterborne commerce, are required by federal regulation or statute for railroads and domestic vessel operators, respectively. Nevertheless, as shown by the experience with the most recent 2007 CFS—with a response rate of 83.1 percent8—nonresponse can be an issue. Respondent burden in filling out the traditional mail-out, mail-back survey is part of the explanation. The accuracy of survey responses is also a problem; for example, only 58.7 percent of the total number of establishments sampled in the 2007 CFS provided complete and usable responses.9 More generally, data providers in the private sector are most concerned about protection of proprietary data.10 In the context of growing interest in detailed travel data by transportation planners and modelers, companies are worried about the risk of revealing such data to competitors. Many businesses also are skeptical of data collection by the federal government, particularly for open-ended purposes. The fear is that the data will be used to regulate the industry or in legal action against it. This is a key concern, for example, with the use of electronic data recorders, which many trucking companies have adopted to track the locations of drivers and shipments (Murray 2010). In the event of a crash, the recorder data could be sub- poenaed to determine culpability. Many companies also are in the business of selling data, not giving them away for free. Thus, they are looking for some exchange of value or incentive to share data with the public sector, with the exception, of course, of data that must be provided by law or regulation. Some federal agencies are already purchasing private data (e.g., the U.S. Army Corps of Engineers purchases data on foreign water- borne commerce from the Port Import Export Reporting Service [PIERS]), and, as discussed subsequently, new data ownership and licensing arrange- ments are emerging. Finally, the burden of lengthy surveys or those  7. However, enforcement measures, such as civil penalties, to coerce firms to participate have not been used.  8. This is the official rate reported to the Office of Management and Budget.  9. R. Duych, BTS, personal communication, April 14, 2010. 10. This discussion draws heavily on briefings to the committee by committee members Joseph Bryan and Daniel Murray at the committee’s third meeting (May 6, 2010) and Thom Pronk, CR England, who participated in a roundtable at the committee’s second meeting (February 18, 2010). 

New Approaches for Meeting Travel Data Needs  49  conducted over an extended period is an issue for busy company staff and may encourage ignoring the request or handing the survey off to less knowledgeable staff unless it is perceived to be of value to the company. Implications of These Barriers The increasing difficulty of collecting travel data, particularly through surveys, has important implications for data providers and users. First, the cost of data collection is increasing, often just to keep response rates constant. Second, declining response rates may introduce bias, calling into question the representativeness of survey results.11 For household surveys, the difficult-to-reach nonrespondents are a key problem. A pilot test of a sample of cellular telephone–only users conducted for the 2009 NHTS, for example, found different travel patterns for this group (Contrino 2010). To what extent do other nonrespondent groups have different travel patterns? The link between response rates and bias is not well understood, and existing research on the topic may offer guidance to the transportation community. For freight surveys, particularly the mandated CFS, the issue is less nonresponse to the survey than the completeness and accuracy of the data. Third-party logistics companies, for example, which handle shipments for many large firms and carriers, are not surveyed in the CFS. As a result, those who do fill out the establishment-based survey may not have the detailed knowledge about freight shipments that they once did when transport and logistics typically were handled in house. Another explana- tion may lie in the fact that respondents do not see the value of the data or understand the purpose for which they will be used.12 Both factors under- score the importance of establishing close ties with data providers and users, involving them in helping to structure data collection instruments. Overcoming the Barriers Strategies for overcoming the barriers discussed above fall into two broad categories: capitalizing on technology and other techniques to improve data collection, and employing alternative methods of data collection for surveys. 11. The issue here is nonresponse bias that is introduced when some members of the population are more likely to be included than others, and their responses differ from those of nonrespondents. 12. This discussion draws heavily on briefings to the committee by committee members Joseph Bryan and Daniel Murray at the committee’s third meeting (May 6, 2010). 

50  How We Travel: A Sustainable National Program for Travel Data Capitalizing on Technology and Other Techniques to Improve Data Collection A range of techniques are being used to help overcome many of the barriers described in the previous section, especially to improve survey response rates. In particular, greater use of technology has the potential to improve the timeliness, efficiency, and accuracy of current travel data collection efforts by substituting automated methods for manual processes. New data collection methods reduce some barriers but do not solve all problems. On the contrary, new issues arise, such as extensive post-processing of data, technical difficulties resulting in missing information, and difficulties collecting socio-demographic information about mode of transport, trip- purpose, and vehicle occupancy (Stopher et al. 2010). Moreover, none of these techniques is likely to reduce the cost of data collection in the short term. Improving Response Rates of Existing Travel Surveys For household surveys, data collectors are using a variety of approaches to improve response rates, ranging from media campaigns to use of incentives (e.g., compensating survey respondents) (see Box 3-1). The use of incentives Box 3-1 Approaches to Overcoming Barriers to the Collection of Passenger Travel Data Most approaches to overcoming barriers to the collection of passenger travel data are focused on boosting response rates to household travel surveys. These approaches include • Media campaigns, • “Rest and recycle” (staged telephone callbacks) for telephone interviews, • Data gathering at a convenient time for the respondent and not necessarily by telephone (e.g., scheduled personal interview), • Special targeting of difficult-to-access socioeconomic groups, and • Use of incentives. 

New Approaches for Meeting Travel Data Needs  51  has become a routine part of many survey research efforts, and survey researchers are generally convinced that incentives should be used to obtain respondent cooperation and ensure proper sample representation (Berry et al. 2008). Nevertheless, their use raises many complex issues. Incentives improve cooperation but do they reduce bias in the estimates produced? Is the use of incentives a reflection of changes in societal norms away from a more altruistic view of survey participation and toward an economic information exchange model? While the use of differential incentives to different groups may prove cost-effective, is the practice “fair”? These questions are beyond the scope of this study, and they represent important questions to be addressed within the recommended travel data program going forward. Many advances in household travel surveys, including greater use of technology, especially GPS tracking, have become commonplace within the United States over the past decade (Zmud 2010b). Until 2006, vehicle-based studies were dominant due to technology limitations of wearable GPS devices. With the relatively recent “explosion” of small, battery-powered, commercially available GPS data loggers, these GPS augments have switched almost entirely to a person-based approach, given the desire to capture detailed data on all modes of travel. A split tech- nology design (in-vehicle or wearable) allows for the collection of many days of highly accurate vehicle-based GPS data with minimal respondent burden. Passive data collection of travel with GPS equipment has many proven benefits, including trip-making rate correction due to under- reporting, improved accuracy of travel times and trip destinations, and detailed travel paths. In addition, multiday data collection enables the evaluation of day-to-day variability of travel along with weekend travel patterns, which can be useful in designing policies to affect choice of time or route of travel (Wolf 2009). A concern for the environment (specifically air quality and emissions regulations), coupled with the modeling community’s desire for more robust data, has led to an increase in the use of on-board diagnostic (OBD) sensors in air quality studies. These sensors monitor vehicle engine performance and store engine operating parameters useful for evaluating the environmental impacts of personal travel and activity patterns. By coupling GPS-based location details with OBD-provided vehicle operations data, engine and vehicle activity can be mapped to the transportation network. In the California Statewide Travel Survey, the California Energy Commission and the California Air Resources Board are planning to fund 

52  How We Travel: A Sustainable National Program for Travel Data an additional in-vehicle GPS/OBD sample focused on alternative fuel, flex fuel, and hybrid vehicle owners.13 With the rapid introduction and use of smart phones, their use to track travel is the next horizon beyond GPS (Schuman 2010). For example, mobile text surveys, completed in real time on hand-held devices, are being increasingly used to collect travel data and beam location or GPS information. Data can be collected on origin–destination flows, travel times, and speeds (The Economist 2007). This technology application may be a mechanism for reducing nonresponse, particularly among hard-to-survey population groups, such as young adults. This is a relatively new use for the transportation field, however, and there has not been a great deal of study on how taking surveys on a mobile device may change the survey process or results. A number of problems must be addressed. For example, the signals are recorded in the cellular phone network, and thus the data belong to the service provider and require provider cooperation for release. Moreover, subscriber cooperation and identification are needed so that the traveler can be contacted and the reasons for the travel added to the flow data—all of which are currently major limitations to gathering survey data with smart phones. And the distribution of smart phones is not universal. Economic disparities related to smart-phone penetration may lead to biased estimation when persons with lower socioeconomic status are under covered. Nevertheless, California is exploring the use of smart phones for data collection for a portion of its next statewide house- hold travel survey, a $12 million project (Zmud 2010b). Greater use of the Internet to gather survey data has the potential to increase the efficiency and timeliness of data collection and may also reduce respondent burden. Travel surveys using paper travel diaries can take a long time to complete and process.14 Web-based diaries not only can “remember” and automatically populate repetitive information, but also are typically linked to interactive maps (such as Google Maps) that allow easy identification of exact locations. Automatic error checking can be built into these web-based diaries as well, making the information provided by respondents more accurate than that recorded in paper diaries. Electronic processing and cleaning of the travel diary data is also more efficient and less prone to errors. At present, however, travel diary 13. Personal communication with J. Wolf, GeoStats, Feb. 11, 2011. 14. The diaries capture information on the total number of trips as well as their characteristics, including purpose, time of travel, transportation mode, and location (i.e., origins and destinations), among other information. 

New Approaches for Meeting Travel Data Needs  53  surveys generally are not being conducted online. Rather, the Internet is being used to advertise the survey, recruit respondents, and display survey results (Zmud 2010b). Greater use of the Internet is limited by household access to high-speed connections, although such access has been growing.15 Another difficulty is obtaining a representative sample; no list of households with Internet access and e-mail addresses currently exists from which a sample can be drawn. Opt-in respondents are the hallmark of many web surveys but are not a suitable sample for travel surveys because of self-selection bias, among other issues. To date, use of technologies that are becoming state-of-the-practice for data collection in local travel surveys is limited for the flagship NHTS. FHWA has recognized the problem and is undertaking a $1.6 million project to explore a wide range of methods (e.g., different sampling frames, different response options) for conducting the next NHTS to boost response rates.16,17 A broader-based research initiative is needed, however, focused on the CFS as well. Some technology innovations were introduced for the most recent CFS but did not directly affect how the survey was conducted.18 Staff acknowledged the need to do much more electronically to move away from the traditional mail-out, mail-back survey approach and help reduce respondent burden (Fowler, 2009). More generally, numerous approaches for overcoming barriers to the collection of freight travel data are being explored and implemented (see Box 3-2). Most apply to data collected from the private sector that are not required by statute or regulation. The focus is less on technology than on arrangements for data sharing and protection of proprietary data. Nevertheless, technology is playing a role. As more source documents become electronic (e.g., rail carload waybills, automated customs data on imports and exports used by PIERS), respondent burden is reduced or eliminated entirely, the speed of data collection is enhanced, and the cost may be reduced. As the PIERS 15. In the 2007 Internet and Computer Use Supplement to the Current Population Survey, the Census Bureau found that 62 percent of households reported having Internet access in the home in 2007, an increase from 18 percent in 1997, the first year the bureau collected such data (U.S. Census Bureau 2009). 16. T. Tang, FHWA, personal communication, June 11, 2010. 17. The project is funded by FHWA ($1 million) and the Office of the Secretary ($600,000). To date, no funds have been provided by the Research and Innovative Technology Administration, but its Bureau of Transportation Statistics is part of the study team. 18. For example, a geographic information system (GIS) postprocessing routing tool was developed to compute mileage for origin–destination data reported on freight shipments to improve accuracy (Duych 2009). 

54  How We Travel: A Sustainable National Program for Travel Data Box 3-2 Approaches for Overcoming Barriers to the Collection of Freight Travel Data A broad range of approaches, focused mainly on arrangements for data sharing with the private sector and protection of proprietary data, are being considered and implemented to overcome barriers to the collection of freight travel data. These approaches include • New data ownership arrangements, with the data being pur- chased or leased from the private sector for public use; • More cooperative public–private arrangements and data sharing to increase value to private data providers; • Greater clarity about the use of the data, increasingly specified in licensing agreements; • Sanitizing of the data to substantially alleviate disclosure con- cerns, either by the Census Bureau (for the CFS) or through cooperative agreements with third-party providers; • Fusion of disparate data sources for the purpose of obscuring competitive information; • Greater use of modeling in cases where the data are particularly sensitive; and • Use of incentives. example discussed in Appendix E illustrates, however, considerable funds still must be spent on data quality control. In summary, a wide range of methods are being explored, including greater use of technology, to reduce respondent burden and improve survey response rates and increase the accuracy and efficiency of both passenger and freight travel data collection. However, use of these methods, particu- larly technology, requires the resolution of numerous issues, which often involves further research and testing before the effectiveness of the methods can be confirmed and they can be widely adopted. Nor will their use necessarily reduce the cost of travel data collection. 

New Approaches for Meeting Travel Data Needs  55  Gathering New Kinds of Travel Data Some of the most innovative uses of technology for gathering travel data are occurring in the private sector, where the focus has been less on conducting surveys than on capturing raw data, often in real time, an approach made possible only recently with the widespread introduction and adoption of new smart technologies and applications. To date, the usefulness of both passenger and freight travel data has been hampered by the lack of timeliness and inadequate detail of the data, particularly for metropolitan and smaller geographic areas. Using technology, the private sector is offering solutions to both of these problems. Two examples are provided here to illustrate the type of automated travel data being collected by the private sector, its public applications, and the implications for data ownership and use. To date, the major focus has been on new ways of tracking vehicle movements. INRIX, a leading provider of traffic and navigation services in North America, aggregates traffic data from more than 2 million GPS-enabled vehicles and cellular probes in its Smart Driver Network, along with other traffic-related data sources, to provide real-time traffic information to both private- and public-sector clients (INRIX 2010a) (see Box 3-3).19 Coverage includes about 100,000 miles of arterials, city streets, and secondary roads, as well as nearly all limited-access highways in the United States (INRIX 2010b). INRIX provides its data to the public sector through licensing agreements with public agencies.20 The data can be used at various levels of aggregation and road coverage for operational purposes, such as dynamic message signs, weather safety alerts, and statewide 511 services (INRIX 2008).21 The data also can be used for congestion analysis on major corridors in 19. The 2 million drivers of the vehicles currently in the INRIX Smart Driver Network report their loca- tion, heading, and speed from vehicles with embedded GPS systems, portable navigation devices, and smartphones. The data are combined with traditional road sensor information, and real-time and predictive traffic speeds are sent to INRIX commercial customers and drivers in the INRIX Smart Driver Network (Schuman 2010). INRIX pays some drivers to provide the needed data where the location information is critical to support its traffic data services. For others, INRIX provides the data free or at a reduced price in exchange for drivers passively reporting their location and speed (personal communication with R. Schuman, INRIX, June 10, 2010). 20. INRIX provides clients with ready access to data through a simple application programming inter- face, a web-based monitoring site, and traffic tile map overlays (INRIX 2008). 21. The telephone number 511 is designated by the Federal Trade Commission for traveler information. Established in 1999, 511 information services vary widely both by provider (ranging from state departments of transportation [DOTs] to local transportation and transit agencies) and by information provided (from traffic delays and weather, to transit and tourism information) (description provided by the 511 Deployment Coalition at http://www.deploy511.org/whatis511.html). 

Box 3-3 INRIX and Private-Sector Travel Data Collection INRIX, a privately held corporation founded by former Microsoft executives in 2004, aggregates and enhances traffic-related data from its own unique and growing Smart Driver Network, along with data obtained from traditional sources such as road sensors. The result is a critical mass of real-time data on vehicle speeds on a broad road network. Once received, the traffic data are fused and processed, using advanced algorithms, to produce information for both individual private-sector clients (INRIX was chosen to be Ford Motor Company’s in-vehicle traffic advisory service, for example) and public-sector clients ranging from individual state departments of transportation to the multistate I-95 Corridor Coalition. Applica- tions of the data range from real-time, in-vehicle traffic information and advisories for drivers, to more aggregated data on traffic flows combined with information on incidents and weather alerts, used by transportation agencies for daily operational purposes. Data also are archived for future retrieval, for example, by public agencies wishing to measure traffic flows and bottlenecks for safety and emergency planning and evacuation purposes and for investment analyses for new capacity. The data are limited to traffic movements and speed, but INRIX announced early in 2010 that it was partnering with the Texas Transportation Institute (TTI), which has data on traffic volumes from the Highway Performance Management System (see the description in Appendix E). For TTI, the partnership will provide a timelier, more accurate, and more complete picture of traffic volumes and congestion in 100 cities, covering virtually every major U.S. metropolitan area. For INRIX, the partnership will increase its visibility as an important contributor to TTI’s highly publicized annual Urban Mobility Reports. In theory, INRIX could also provide origin–destination data (Schuman 2010). Because every vehicle INRIX tracks has an iden- tification number, the company could create an application that would capture vehicle trip traces on the network. At present, how- ever, these data are rapidly discarded for confidentiality reasons. INRIX would have to obtain authorization to keep the vehicle iden- tification data for the purpose of creating origin–destination data. 

New Approaches for Meeting Travel Data Needs  57  metropolitan areas, as well as for traffic management on multistate cor- ridors (e.g., the I-95 corridor on the East Coast). INRIX provides a free annual National Traffic Scorecard, which reports on nationwide conges- tion trends over time, identifies and ranks the worst traffic bottlenecks, and provides regional traffic comparisons on the nation’s major highways (INRIX 2010c).22 Currently, INRIX travel data are limited to vehicle speeds. The main issue for transportation planners and modelers is the lack of infor- mation on trip origins and destinations, who is traveling (e.g., socio- economic characteristics), and the purpose of their travel—behavioral data that are essential to building policy-sensitive predictive models. A partnership between FHWA and the American Transportation Research Institute (ATRI), the independent research arm of the American Trucking Associations,23 offers another model for providing current data on traffic movements (see Box 3-4). In 2002, FHWA launched the Freight Performance Measurement (FPM) initiative to fill a gap in information on how congestion and delay affect goods movement by trucking companies (Mallet et al. 2006). Because data on freight movements generally reside in the private sector, FHWA partnered with ATRI to collect intercity travel data from motor carriers on significant freight corridors and international land-border crossings. Working with trucking companies and third-party vendors protected by contractual arrangements and nondisclosure agreements that main- tain the confidentiality of the data, ATRI currently collects data from approximately 600,000 GPS-instrumented trucks throughout North America (Jones and Murray 2010).24 The core FPM initiative centers on data on travel speeds and reliability for some 25 significant Interstate freight corridors, border crossing times and reliability for 15 major U.S. international land-border crossings, truck origins–destinations, and truck parking activities.25 In both 2008 and 2009, FHWA and ATRI released an 22. Most recently, the company launched INRIXTraffic.us—a free web service providing state, regional, and municipal transportation agencies with information on real-time traffic flows on all major high- ways, Interstates, arterials, and secondary roads in major cities and rural areas and across state lines (INRIX 2010a). 23. Legally, ATRI is a 501(c)3 not-for-profit organization. 24. INRIX also collects data on commercial vehicle movements as a subsample of its network data, but its sample does not include as large a number of heavy-duty, over-the-road trucks as ATRI’s (Schuman 2010). 25. The data on Interstate corridors are supplemented with average annual daily traffic data from FHWA’s Highway Performance Measurement System, described in Appendix E (Mallet et al. 2006). Much of the data on travel time and delay at U.S.–Mexico border crossings is provided by vehicle-mounted radio frequency identification tags (Jones 2010). 

58  How We Travel: A Sustainable National Program for Travel Data Box 3-4 Federal Highway Administration–American Transportation Research Institute Partnership to Collect Truck Travel Data for Freight Performance Management The collaboration between the Federal Highway Administration (FHWA) and the American Transportation Research Institute (ATRI) represents a new public–private approach to travel data collection. With its close association to the trucking industry, ATRI has assumed primary responsibility for recruiting a nationwide volunteer sample of heavy-duty, over-the-road, GPS-instrumented trucks to provide the data. FHWA set general guidelines for the intended uses of the data (Jones and Murray 2010). The data are collected from participating vehicles using anonymous, randomly generated identification numbers to main- tain the confidentiality of the truckers and trucking companies (Mallet et al. 2006). Data on the position of the truck (latitude and longitude), spot speeds, and time and date are received at pre- determined intervals. The data are then matched to latitude and longitude coordinates of the Interstate corridors of interest. The processed data can be incorporated into data models, dashboards, or visualization software. Beyond spot-speed data, the recording of location and time along a route enables the calculation of average “processed” speed for each truck on a specific road segment; speeds of multiple trucks are then aggregated to determine average speed on a road segment or network (Mallet et al. 2006). Finally, average truck speeds in miles per hour are calculated for the entire length of a corridor. Either spot-speed or processed speed data can be used to calculate travel time reliability measures such as buffer time indices, variability measures, and simple standard deviations from mean speeds. Since the Freight Performance Measurement (FPM) initiative began, FHWA has continually revised the program by increasing the sample size, the geographic coverage, and the representativeness of (continued) 

New Approaches for Meeting Travel Data Needs  59  the truck fleets. The expansion of the number of trucking firms and trucks involved in the project reflects in part the negotiated disclosure agreements with multiple data sources; these agreements create binding stipulations on the purpose of the data collection and what data can be collected, provisions to ensure anonymity (including legal and financial consequences for violations), and sunset provisions (Jones and Murray 2010). Private parties receive some remuneration for the provision of the data. To ensure that the core mission of the program is not endangered by inappropriate uses, both FHWA and ATRI monitor requests for data and deter- mine cooperatively who should have access and at what level of aggregation. Currently, public distribution of the travel data is limited to travel speed and reliability. Data on average speed can also potentially be coupled with ideal speed (posted speed limits), as well as with data on truck volumes and percentage of truck traffic from FHWA’s Highway Performance Management System, to derive measures of delay. Other potential measures for future use include expansion of border crossing monitoring, truck parking applications, and weather and work zone impact analyses (Mallet et al. 2006). annual report on the top freight-critical nodes and bottlenecks in the United States (see Short et al. 2009 for the most recent report).26,27 The main incentive for ATRI and the trucking industry to participate in the FPM initiative is to educate U.S. DOT leadership about the critical effect of congestion on trucking operations and the economic costs and productivity losses that accrue from those delays. By specifically identifying the worst bottlenecks on roadways and at border crossings, the trucking 26. Using truck position and speed data, ATRI identified and analyzed the bottlenecks. They were ranked on the basis of severity through an analysis of speed data for 24 1-hour time blocks, which involved comparing actual speeds with a free-flow criterion—55 mph—and computing the miles per hour below free-flow on an hour-by-hour basis for the 24-hour period (Short et al. 2009). 27. Most recently, FHWA, in partnership with ATRI, launched a free web tool—FPMWeb—that enables state and local transportation agencies, as well as businesses and freight companies, to access data on where and when trucks are moving at slower than free-flow speeds, to visualize the results in a web-based GIS environment, and to probe the data more deeply through a customized query option (FPMweb undated; AASHTO Journal 2010). 

60  How We Travel: A Sustainable National Program for Travel Data industry hopes to garner support for greater investment in transportation infrastructure projects. The main limitations of the publicly available data are twofold: (a) like the INRIX data, the ATRI data are restricted to travel speeds, and (b) the coverage properties of the data are unknown (i.e., what kinds of trucks are not covered and the extent to which uncovered trips differ from those covered). INRIX and the FHWA–ATRI partnership are only two examples from a growing field of private-sector providers of travel data, whose data collection approaches range from Bluetooth-enhanced traffic surveillance equipment to airborne traffic data collection.28 Implications and Assessment INRIX and ATRI provide good examples of new ways of thinking about data collection, ownership, and access in a postregulatory environment. The data are not collected through a traditional sample survey approach but through automatic collection of millions of bits of raw data, which are then fused with other data, aggregated, and archived for different applications. In this context, no one entity owns all the data; a common arrangement, if the public sector is interested in accessing the data, is to lease them from private data aggregators. With this arrangement, the government can use the data for well-defined purposes, but the private sector is protected from Freedom of Information Act disclosure. The benefits to the public sector are that the data are timely, detailed, and scaleable. The main drawbacks are the lack of control over the data, the lack of transparency with respect to their collection and quality (such as coverage), and the need in most cases to purchase the data from the private sector. Finally, much of the data is focused on vehicle movements and speeds but not connected to information on traffic volumes, trips, people, or vehicle characteristics; travel behaviors; or the condition of the infrastructure on which the travel occurs.29 28. Traffax Inc., for example, uses its proprietary Blufax traffic surveillance units, together with in-vehicle Bluetooth technology, to provide state and local governments with continuous real-time measure- ment of travel times (between data collection stations) for vehicle and pedestrian applications on freeways, arterials, and pedestrian environments. Skycomp contracts with transportation agencies and engineering firms to collect traffic data using time-lapse aerial photography across large regions and built-up urban areas and at small sites with complex vehicle movements. 29. Wrap-up commentary by committee member Lance Grenzeback at Session 2, on Capitalizing on New Technologies to the Committee on Strategies for Improved Passenger and Freight Travel Data, May 6, 2010. 

New Approaches for Meeting Travel Data Needs  61  Employing Alternative Methods of Data Collection for Surveys The flagship federal travel surveys, which are key elements of current travel data programs, are conducted as periodic cross-sectional surveys. They are expensive to conduct, their results are dated by the time the data are released, and the data provide an incomplete picture of travel patterns and issues. Alternative data collection methods for surveys hold potential for providing data that are less expensive to obtain, more timely, and more appropriate for answering today’s transportation questions. These alternative methods include (a) continuous surveys with responsive design, (b) panel surveys, and (c) hybrid approaches. Both continuous and panel surveys collect data over long time periods. Continuous surveys provide repeated cross-sectional “snapshots” of a population using a new sample of the population each time. These sur- veys are constantly sampling and including new groups, enabling direct measurement of changes in the overall population. Panel surveys, on the other hand, track the behavior of a fixed sample of subjects over relatively long time periods (i.e., years). It may be necessary to replenish panel samples because of attrition and the introduction of new groups into the population. Both types of surveys can provide data on a more timely basis since the data collected can be processed and released even as data collection continues for new cross-section samples in a continuous survey or reinterviews with panel members. Continuous Surveys Continuous surveys may require smaller staff than periodic cross-sectional surveys and thus may cost less per unit of data collected. This benefit is realized through more efficient utilization of management and supervision in data collection operations and less need for staff specialization, as staff perform multiple duties throughout the data collection. Periodic cross- sectional surveys, in comparison, require the development of a large staff dedicated to one operation in a limited time period, a relatively expensive way to collect data. Continuous surveys enable planners and decision makers to monitor travel behavior over time so as to understand changes at different stages in an economic cycle or during periods of high or low fuel prices (Raimond 2009). Periodic surveys, such as the NHTS, however, may be conducted during atypical travel periods and are less well configured to 

62  How We Travel: A Sustainable National Program for Travel Data measure change except by comparison with previous periodic surveys. For example, the 2001 NHTS was conducted at the time of the terrorist attacks of September 11, which sharply depressed travel, and the 2009 NHTS was conducted during a deep recession, which also depressed travel.30 The absence of intermediate measurements leaves the transportation community without adequate data with which to assess the severity of the travel reduction during these periods. Continuous surveys also reduce the pig-in-a-python effect of funding for large periodic surveys by spreading out the costs more evenly over several years and can offer greater flexibility (e.g., new topics and questions can be added without having to wait until the next survey) (Raimond 2009; Stopher 2009; Zumkeller and Ottmann 2009). Continuous surveys also offer the opportunity to monitor the performance of the data collection system more carefully, identify and measure indicators of data quality, and intervene to improve data quality as the survey is conducted. These monitoring and intervention methods are referred to as responsive or adaptive design, and they offer the potential for continuing improvement in some properties of quality (Groves and Heeringa 2006). Examples are beginning to appear in the survey literature (see, for example, Lepkowski et al. 2010), indicating how responsive designs can be implemented in continuous surveys and what impact responsive design techniques can have on cost and data quality. Significant savings have been demonstrated, for example, by moving from periodic cross-section surveys to continuous data collection. But the gains are not guaranteed and must be coupled with recent advances in responsive survey designs. These emerging techniques deserve the careful attention of the transpor- tation survey community. Continuous surveys require a different way of analyzing and interpreting the data. Because the data are collected continuously, they are received in smaller increments over extended periods of time compared with those collected by periodic one-time efforts. Any one year of data in a continuous survey would have larger sampling variances than a single cross-sectional survey for the same year. Continuous data must often be aggregated over time to obtain the same sample sizes for small groups that would be obtained from periodic surveys conducted at one point in time. Continuous surveys thus require pooling data over several years to increase confidence 30. NHTS program managers note, however, that the travel effects were somewhat mitigated because data collection took place over the period of a year. 

New Approaches for Meeting Travel Data Needs  63  in the estimates derived from the data (Raimond 2009).31 This is a major issue for small-area data, an issue discussed below with respect to the Journey to Work data in the ACS. The aggregation forces analysts to use such techniques as moving averages over time rather than single-point- in-time estimates generated from periodic surveys (Lepkowski, 2010b). Analysts also face challenges in interpreting trend data from continuous surveys. For example, if multiple years are to be compared, care must be taken to avoid overlap in years across aggregated data time periods (Plewes 2010). Other data quality concerns arise with continuous surveys. Some suggest that team fatigue and waning motivation with continuous surveying lead to declining response rates, while limited evidence suggests that response rates that have been declining could actually be stabilized by more consis- tent staffing and methods (Lepkowski et al. 2010). Finally, continuous surveys even out funding requirements, a particularly valuable feature in stable funding environments. If budgets remain flat over time while per unit costs increase, however, sample sizes must be reduced, and overall confidence in the estimates diminishes (Lepkowski 2010a). The ACS is already an example of how flat budgets over extended time periods reduce sample size as a survey progresses.32 To date, the transportation community’s experience with the ACS has not been entirely satisfactory, particularly with respect to small-area data (Christopher 2009; Murakami 2009). Part of the problem is transitional. It has taken 5 years after the start of the ACS to pool sufficient data for small-area analysis; each year after that, new data based on a moving 5-year average will be made available. But even with such accumulations, the variability of the data and disclosure issues for small areas are likely to remain (Plewes 2010). Although small-area estimates derived from the long form of the census were less variable, however, they were also less timely, the data being collected only once every 10 years. A project now under way (NCHRP forthcoming) is exploring possible solutions, ranging from combining small geographic units (e.g., traffic analysis zones) to using 31. Of course, smaller annual sample sizes can also be viewed as a benefit from the perspective of cost and burden for staff and respondents. 32. The sample size of the ACS is about 3 million households per year, but respondents number fewer than 2 million annually, and the 5-year cumulated sample is less than the 2000 long-form sample (Plewes 2010). The Census Bureau’s fiscal year 2011 budget includes funding to boost the annual ACS household sample size by 500,000 to achieve original precision and sample-size goals and preserve the reliability of small-area estimates (Plewes 2010). 

64  How We Travel: A Sustainable National Program for Travel Data synthetic data and modeling.33 Data managers viewing the ACS experience are concerned that a shift to continuous survey data collection for the NHTS will pose similar challenges and trade-offs with respect to small- area estimates. Continuous surveys have long been successful in other fields, such as health, where they have generally proved less expensive than periodic surveys and provided better value, largely through smaller, better-trained, and more experienced staff (Lepkowski 2010a). Continuous surveys also are used in other countries.34 For example, Great Britain has successfully used continuous surveying since 1988 for its National Travel Survey (see Box 3-5). That survey provides regular, up-to-date data on personal travel, including long-distance travel (i.e., greater than 50 miles) within Great Britain, which enables monitoring of changes in travel behavior and helps inform the development of policy (Anderson et al. 2009). The smallest geographic units for which the data are generally published are the nine Government Office Regions.35 Panel Surveys Panel surveys are another way of collecting data that can be particularly useful in understanding the dynamics of travel behavior, although experi- ence with these surveys in transportation research, particularly in the United States, is limited. In comparison with periodic and continuous surveys, which rely on cross-sectional designs, longitudinal panels enable analysts not only to study changes in travel behavior over time, but also to understand the reasons for shifts in behavior or attitudes because the same group (panel) of respondents is queried in each survey wave (Zmud 2009).36 33. Synthetic data replace underlying microdata with values derived from a model-dependent imputation approach (e.g., using regression models), data swapping, or an additive noise technique. A random component is used in the generation of synthetic data, and thus “noise” is added to the data as a means of disclosure control. For example, in a particular locality where revealing household identity could be an issue, the characteristics of one household could be swapped with those of another to protect the identity of persons in the households. The goal of the approach is to retain household characteristics and travel patterns at an aggregate level, capture the error component due to the masking procedure, and retain multivariate associations between household characteristics (T. Krenzke, Westat, personal communication, Aug. 17, 2010). 34. Committee member Johanna Zmud briefed the committee on international practices, particularly the use of panel surveys, at the third committee meeting in Session 3: Alternative Data Collection Methods to Support Future Data Programs. She also directed the committee to a book, summarizing the results of the 8th International Conference on Survey Methods in Transport at Annecy, France, in 2008 (Bonnel et al. 2009a), which provided many examples of international practice. 35. Analyses at finer geographic levels (e.g., urban, rural) are possible if sample sizes are large enough. 36. In a panel survey, a wave is the interviewing period during which the entire panel is surveyed and asked the same questions. A panel survey consists of multiple waves. 

New Approaches for Meeting Travel Data Needs  65  Box 3-5 The National Travel Survey of Great Britain An Example of a Continuous Survey The National Travel Survey (NTS) of Great Britain, sponsored by the Department for Transport (DfT), provides continuous data on personal travel within Great Britain. The sample frame is postal addresses in Great Britain, and data are collected continuously during every month of the year on the basis of a stratified sample of 40 regions (relating roughly to counties or groups of counties in England and groups of unitary authorities or council areas in Scotland and Wales), with oversampling in London. The results are weighted to help reduce the effect of nonresponse bias. The process of recruiting and interviewing households includes an advance recruitment letter, followed by a face-to-face interview with all household members (or proxies). During the interview, point data on household characteristics and vehicle ownership are collected, and a £5 gift voucher is offered if all household members complete every section of the survey. Households are informed of their travel week and left with a 7-day travel diary in which they record each trip, including origin–destination details, purpose, mode used, distance traveled, trip time, and number traveling. Within 6 days of the end of the travel week, a pick-up interview is conducted, and the travel diaries are collected. The data are coded and entered into a data system, and quality checks are performed. Response rates are high—around 60 percent overall, but lower in inner and outer London (46 percent and 49 percent, respectively, in 2008) (Anderson et al. 2009). The data are analyzed at various levels (e.g., by household, individual, vehicle, day, trip), but the smallest geographic unit typically published is at the Government Office Region level; nine such regions exist in Great Britain. Long-distance trips (more than 50 miles) within Great Britain are also recorded, with respondents being asked to note any such journeys during their travel week and during an additional week. Finally, questions may be added periodically to gather information on a particular policy (continued on next page) 

66  How We Travel: A Sustainable National Program for Travel Data Box 3-5 (continued) The National Travel Survey of Great Britain An Example of a Continuous Survey or question. Key results are published annually in a statistical bulletin available on the DfT website. Technical reports and additional analyses, including a set of factsheets, are also available on the web. Finally a nondisclosure version of the NTS data set is deposited at the UK Data Archive at the University of Essex. DfT funds the NTS, which is currently carried out under contract by the National Centre for Social Research, an independent social research institute. The contractor is responsible for questionnaire development, sample selection, data collection and editing, and data file production (Anderson et al. 2009). DfT, supported by a staff of five full-time equivalents (FTEs), is responsible for the building of the database, data analysis, publication, archiving, and research on future survey methods. The total cost of the sur- vey (contractor and DfT staff costs) is currently about £2.8 million (about $4.18 million) annually, about two-thirds of which is for basic fieldwork and incentives (L. Avery, Department for Transport, UK, personal communication, June 24, 2010). Thus, panel surveys provide a more sophisticated understanding of travel behavior than can be derived from cross-sectional analyses, and the data can be used in travel demand models to better predict travel behavior (Zmud 2009). Questions can readily be added to the survey to explore traveler responses to a particular policy or transportation investment (e.g., expanded transit services). Panel surveys also provide timely infor- mation and require smaller sample sizes than periodic or continuous surveys and thus have lower recruitment and staff costs, at least in the early years of a panel (Zmud 2009). Panel surveys pose challenges, not the least of which are initial recruit- ment in the face of the continuing nature of the survey, imposing a heavier respondent burden; natural attrition of the panel and declining response rates over time; and panel fatigue and poorer quality of responses in later 

New Approaches for Meeting Travel Data Needs  67  survey waves (Zmud 2009).37 These problems can be addressed through such measures as refreshing the panel by replacing members who have left and providing incentives to panel members, but these measures complicate longitudinal analyses. Finally, taking advantage of the data provided by panel surveys requires knowledgeable staff and sophisticated models. In fact, one of the reasons given for the lack of more panel surveys in the United States is the absence of dynamic models, such as activity- based models, which can make use of the results (Stopher 2009).38 The primary example of a panel survey in the United Sates is the Puget Sound Transportation Panel, which ran nearly annually from 1989 to 2002. Data on one day of travel activity were collected from about 1,700 respon- dents in 10 annual survey waves (Zmud 2009). In 2002, the panel was discontinued and replaced with a typical cross-sectional local travel survey. The main reasons for its termination were the time-consuming nature of maintaining the panel, the resulting cost, and the lack of sophisticated dynamic models for using the data captured from the panel.39 The cost, for example, increased by more than 2.5 times, from $75,000 in 1989 to $200,000 in 2002, or about 1.8 times in inflation-adjusted dollars (Howard 2010). The cost of the cross-sectional household travel survey, which was conducted in 2006 and replaced the panel, was $1 million; it is planned to be repeated no later than 2015.40 One way to reduce the initial costs of establishing a panel and anticipate the challenges of response bias, panel maintenance, and panel attrition is to use an existing panel source. There are private firms that specialize in running or establishing customized longitudinal panels for both public and private clients.41 Special care must be taken to ensure that the selected panel meets rigorous standards of accuracy and reliability through probability-based, statistically valid (not opt-in) sampling, and that panel 37. Panel attrition is not a trivial problem. The Puget Sound Transportation Panel experienced about a 20 percent attrition rate between the first two survey waves, the German Mobility Panel a 43 percent attrition rate, and the Dutch National Mobility Panel a 44 percent attrition rate (Zmud 2009, 3). 38. Activity-based models capture the dynamic interaction between the activities of households and individuals and their travel decisions. They are based on a more comprehensive understanding of the trade-offs that affect decisions about whether to make a trip, what time to make it, what destination to visit, what mode to use, and what path to take (TRB 2007). 39. At the committee’s third meeting, on May 6, 2010, Elaine Murakami (FHWA) noted that one of the reasons for the decision not to continue the Puget Sound Transportation Panel was the lack of modeling capacity to take advantage of the survey-generated data. 40. N. Kilgren, Puget Sound Regional Council, personal communication, July 1 and July 6, 2010. 41. For example, D. K. Shifflet & Associates, which collects tourism-oriented travel data (described in Appendix E), uses a panel company to recruit nationally representative panels of households, which have agreed in advance to participate in periodic surveys. 

68  How We Travel: A Sustainable National Program for Travel Data recruits have similar access to technology (e.g., a computer and free Internet service), particularly when online participation is desired. The German Mobility Panels are an example of long-standing use of panel surveys in transportation research. This panel has been conducted nationally each year since 1994, with a sample of about 1,000 households reporting on travel activity in a 7-day diary (Zmud 2009).42 A rotating panel approach is used, whereby respondents participate for three con- secutive years, replaced by new panel respondents, so as to ensure reliable and motivated participants (Zumkeller et al. 2008).43 Provision is also made for stratified recruitment of new cohorts to balance any dropout bias (Zumkeller 2007). The national panel survey is complemented by several similarly designed regional panels to obtain more detailed data on travel in major regions of the country and to increase the opportunities for pooling data (Zumkeller et al. 2008).44,45 The panel surveys are part of a family of personal travel surveys, described in the following subsection. A Hybrid Approach In view of the pros and cons of the different survey methods, the most efficacious strategy may be to combine several different types of sur- veys to meet a range of needs that motivate the surveys (Bonnel et al. 2009b). Germany provides an excellent example of this approach for household surveys. It conducts periodic national cross-sectional surveys with large samples every 5 to 10 years. These surveys are supplemented by two longitudinal panel surveys at the national level—the annual German Mobility Panel focused on everyday travel (previously discussed) and the INVERMO panel survey of long-distance travel (i.e., distances greater than 100 kilometers)—as well as selected regional panel surveys (also previously discussed) (Zumkeller 2007).46 42. The diary survey of travel activity is conducted during September through November of each year. A 3-month odometer survey with a focus on fuel consumption is administered during April through June (Zumkeller 2007). 43. Response rates are relatively low—about 20 percent of the original sample recruited by telephone. 44. Panel participants at the national level are not required to geocode their trips, easing respondent burden. However, these data are collected in the regional panels because they are needed for planning and modeling purposes (Zumkeller et al. 2008). 45. In the early years of a regional panel, household data from the national survey for a specific region are pooled with the regional data, so that the regional authorities have immediate results. Over time, the national sample data are phased out. 46. The INVERMO survey was last conducted between 1999 and 2002. Using a combination of a screen- ing telephone interview and a postal survey, panel members reported their long-distance travel for a 2-month period over four reporting time frames (Zumkeller 2007). 

New Approaches for Meeting Travel Data Needs  69  The primary sponsor of the German Mobility Panel is the German Ministry of Transport.47 The cross-sectional surveys are cosponsored by regional and state authorities, whose funding enlarges sample sizes for their geographic areas. The INVERMO panel is funded by the German Federal Ministry for Education and Research and includes several private partners.48 Together these surveys provide a broad picture of personal travel behavior in Germany and have enabled in-depth analyses of such topics as the stability and variability of weekly travel behavior, fuel price elasticities, coordination of travel among different household members, and car dependency and multimodal travel behavior (Zumkeller 2007). As discussed in the following subsection, a similar approach could be adopted in the United States. Implications and Assessment The different approaches to data collection just reviewed suggest that there is no one best method. Each approach has its pros and cons, and each serves a particular purpose. The United States should consider adopting an approach similar to the German model—using a portfolio of surveys at the core of comprehensive data programs to meet future travel data needs, both passenger and freight. This approach should include • Consideration of continuous surveys to replace or supplement the federal flagship surveys to provide more timely travel data or, at a minimum, a regular cycle of periodic surveys with updates in interim years using a smaller sample; • Establishment of a national panel survey to improve understanding of the dynamics of household travel behavior and to track national travel trends over time, which could be supplemented by periodic surveys targeting traveler response to particular policies and investments; • Partnerships with state and local governments to expand national surveys to collect more state- and regional-level data and to work toward more common formats for state and local travel surveys so as to encourage pooling of data, or substitution of modeled data, particularly for use across small metropolitan areas with common characteristics; and 47. Technical support is provided by the University of Karlsruhe, and the fieldwork is conducted by several market research companies. 48. Among these are the private German Rail system (Deutsche Bahn AG), Lufthansa German Airlines, and the German arm of the global market research company TNS Infratest (Zumkeller 2007). 

70  How We Travel: A Sustainable National Program for Travel Data • Partnerships with the private sector to acquire more fine-grained data on the travel patterns of individuals and private firms, using digital methods of data capture and methods to protect sensitive competitive data, and integrating and aggregating the automated data for analysis and decision making. Unlike the German top-down model, however, the portfolio approach envisioned for the United States would be a more decentralized data collection system. It would be built on a strong, federally supported core of surveys and data collection activities to enable the gathering and dis- semination of essential travel data, but well integrated with travel data collected by the states, MPOs, transit agencies and other local authorities, as well as the private sector. This concept is described in greater detail in the next chapter. Findings The transportation community needs to change the way it collects travel data to address many significant barriers to data collection. Traditional methods of collecting essential national travel data through large-scale, periodic surveys should be adapted to address issues of public willingness to provide data and should take advantage of evolving technologies and data collection approaches. Fortunately, alternative methods of data collection are available, but each involves trade-offs compared with large- scale, periodic surveys. Use of continuous cross-sectional surveys and panel surveys can help spread out the costs of data collection, maintain a well- trained core staff, and provide more timely results. Experience with such approaches is limited in the transportation sector, however, and the learning curve for properly collecting, analyzing, and using the data is likely to be steep. In addition, more evidence is needed on whether these methods will improve or stabilize response rates compared with periodic surveys. Greater use of automated data sources (e.g., passive probes) and technology (e.g., web surveys, GPS) may reduce respondent burden and improve response accuracy, but most of these methods are unlikely to reduce the costs of data collection. Furthermore, much of the data collected with these methods is focused on vehicle movements and speeds and trip origins and destinations, without being linked to information about who is traveling and for what purpose—behavioral information critical for modeling and analysis to support policy making. 

New Approaches for Meeting Travel Data Needs  71  A program of methods research is needed to examine a wide range of approaches to data collection. Such research would help determine the optimal frequency for surveys and updates, involve pilot testing of new techniques before they are adopted more widely, and identify opportuni- ties for purchasing commercial data or contracting with private vendors for data collection. References Abbreviations AASHTO American Association of State Transportation and Highway Officials FHWA Federal Highway Administration GAO U.S. Government Accountability Office NCHRP National Cooperative Highway Research Program TRB Transportation Research Board AASHTO Journal. 2010. FHWA Launches New Technology Tool to Pinpoint Freight Congestion, May 28. Anderson, T., O. Christophersen, K. Pickering, H. Southwood, and S. Tipping. 2009. National Travel Survey 2008 Technical Report, No. P2820. Prepared for the Department for Transport, National Centre for Social Research, London, England, July. Berry, S. H., J. S. Pevar, and M. Zander-Cotugno. 2008. Use of Incentives in Surveys Supported by Federal Grants. Rand Corporation, Santa Monica, Calif. Paper presented at the seminar of the Council of Professional Associations on Federal Statistics on Survey Respondent Incentives: Research and Practice, Washington, D.C., March 10. Billitteri, T. J. 2010. Census Controversy. CQ Researchers, May 14, pp. 433–455. Bonnel, P., M. Lee-Gosselin, J. Zmud, and J. L. Madre (eds). 2009a. Transport Survey Methods: Keeping Up with a Changing World, Emerald Group Publishing Limited, Bingley, United Kingdom. Bonnel, P., M. Lee-Gosselin, J. L. Madre, and J. Zmud. 2009b. Keeping up with a Changing World: Challenges in the Design of Transport Survey Methods. In Transport Survey Methods: Keeping Up with a Changing World (Bonnel et al. eds.), Emerald Group Publishing Limited, Bingley, United Kingdom, pp. 4–14. Christopher, E. 2009. Census Data for Transportation Planning. Federal Highway Administration, Matteson, Ill. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., Dec. 10. Contrino, H. 2010. The National Household Travel Survey. Federal Highway Administration, Washington, D.C. Presentation to a stakeholder meeting at the American Automobile Association, Washington, D.C., Feb. 25. 

72  How We Travel: A Sustainable National Program for Travel Data Curtin, R., S. Presser, and E. Singer. 2005. Changes in Telephone Survey Nonresponse over the Past Quarter Century. Public Opinion Quarterly, Vol. 69, pp. 87–98. Duych, R. 2009. Bureau of Transportation Statistics, Washington, D.C. Briefing on the Commodity Flow Survey presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., Dec. 10. Fowler, J. 2009. The U.S. Bureau of the Census, Washington, D.C. Briefing on the Commodity Flow Survey presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., Dec. 10. FPMweb. Undated. Accessing Freight Performance Measures through the Inter- net. Federal Highway Administration, U.S. Department of Transportation, Washington, D.C. GAO. 2010. Data Collection Operations Were Generally Completed as Planned, but Long-standing Challenges Suggest Need for Fundamental Reforms. GAO-11-193. Washington, D.C., December. GAO. 2001. Significant Increase in Cost per Housing Unit Compared to 1990 Census. GAO-02-31. Washington, D.C., December. Groves, R. M., and S. Heeringa. 2006. Responsive Design for Household Surveys: Tools for Actively Controlling Survey Errors and Costs. Journal of the Royal Statistical Society, Series A, Vol. 169, No. 3, pp. 439–457, July. Howard, C. 2010. Alternative Data Collection Methods. Session 3. Puget Sound Regional Council, Seattle, WA. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6. INRIX. 2010a. INRIX Launches Free Traffic Operations Service for Transportation Agencies Nationwide. INRIX News Alert, June 3. INRIX. 2010b. INRIX Technology Breakthrough Significantly Improves Accuracy of Real-Time Traffic Information for Navigation on Arterials, City Streets and Secondary Roads. Press Release. http://www.inrix.com/pressrelease.asp?ID=88. Accessed March 3, 2010. INRIX. 2010c. INRIX National Traffic Scorecard. 2009 Annual Report, Synopsis. http://scorecard.inrix.com. Accessed April 30, 2010. INRIX. 2008. INRIX, the Public Sector’s Leading Source for Private Traffic Data. Kirkland, Wash. Jones, C. 2010. Border Data Initiatives. Federal Highway Administration, Wash- ington, D.C. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., Feb. 19. Jones, C., and D. Murray. 2010. Freight Performance Measurement—FPM. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., Feb. 19. Lepkowski, J. M. 2010a. Handout accompanying briefing on Barriers to Data Collection, and How We Might Overcome Them. Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6. Lepkowski, J. M. 2010b. Alternative Data Collection Methods. Session 3. Uni- versity of Michigan, Ann Arbor. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6. 

New Approaches for Meeting Travel Data Needs  73  Lepkowski, J. M., W. A. Mosher, K. A. Davis, R. M Groves, and J. van Hoewyk. 2010. Continuous National Survey of Family Growth: New Concepts for Sample Design and Analysis. National Center for Health Statistics, Hyattsville, Md. Mallet, W., C. Jones, J. Sedoc, and J. Short. 2006. Freight Performance Measurement: Travel Time in Freight-Significant Corridors. FHWA-HOP-07-071. FHWA, U.S. Department of Transportation, Washington, D.C., December. Murakami, E. 2009. Census Transportation Planning Products. FHWA, Seattle, WA. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., Dec. 10. Murray, D. 2010. Barriers to Data Collection. Session 1. American Transportation Research Institute, St. Paul, Minn. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6. Pew Research Center. 2010. Distrust, Discontent, Anger and Partisan Rancor. Wash- ington, D.C., April 18. http://pewresearch.org/pubs/1569/trust-in-government- distrust-discontent-anger-partisan-rancor. Accessed April 30, 2010. Plewes, T. 2010. Using the American Community Survey: Benefits and Challenges. Committee on National Statistics, Washington, D.C. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6. Princeton Survey Research Associates. 2008. How Different are People Who Don’t Respond to Pollsters? Pew Research Center Publications, April 21. Raimond, T. 2009. Moving Towards Continuous Collection of Large-Scale Mobility Surveys: Are There Compelling Reasons? A Discussant Response. In Transport Survey Methods: Keeping Up with a Changing World (Bonnel et al. eds.), Emerald Group Publishing Limited, Bingley, United Kingdom, pp. 541–548. Schuman, R. 2010. Capitalizing on New Technologies: INRIX’s Perspective. Session 2. INRIX, Kirkland, WA. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6. Short, J., R. Pickett, and J. Christianson. 2009. Freight Performance Measures Analysis of 30 Freight Bottlenecks. The American Transportation Research Insti- tute, Arlington, Va., March. Stopher, P. R., C. Prasad, and J. Zhang, 2010. Can GPS Replace Conventional Travel Surveys? Some Findings. Proceedings of the 33rd Australasian Transport Research Forum, Canberra, Australia, Sept. 29–Oct. 1. Stopher, P. R. 2009. The Travel Survey Toolkit: Where to From Here? In Transport Survey Methods: Keeping Up with a Changing World (Bonnel et al. eds.), Emerald Group Publishing Limited, Bingley, United Kingdom, pp. 15–46. The Economist. 2007. Visualisation: Go with the Flow. Economist Intelligence Unit, March 8. http://www.ebusinessforum.com/index.asp?layout=rich_story&doc_ id=10276&title=Visualisation%3A+Go+with+the+flow&categoryid=1&channel id=3. Accessed June 11, 2010. TRB. Forthcoming. Producing Data Products from the American Community Survey that Comply with Disclosure Rules. NCHRP Project No. 08-79. Transportation Research Board, Washington, D.C. 

74  How We Travel: A Sustainable National Program for Travel Data TRB. 2007. Special Report 288: Metropolitan Travel Forecasting: Current Practice and Future Direction. Transportation Research Board of the National Academies, Washington, D.C. U.S. Census Bureau. 2009. Internet Use Triples in a Decade. Press release, June 3. http://www.census.gov/newsroom/releases/archives/communication_ industries/cb09-84.html. Accessed June 11, 2010. Wolf, J. 2009. Mobile Technologies: Synthesis of a Workshop. In Transport Survey Methods: Keeping Up with a Changing World (Bonnel et al. eds)., Emerald Group Publishing Limited, Bingley, United Kingdom, pp. 393–402. Zmud, J. 2010a. Barriers to Data Collection. Session 1. NuStats, LLC, Austin, TX. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6, 2010. Zmud, J. 2010b. Capitalizing on New Technologies. Session 2. NuStats, LLC, Austin, TX. Briefing presented to the Committee on Strategies for Improved Passenger and Freight Travel Data, Washington, D.C., May 6, 2010. Zmud, J. 2009. Draft Technical Memorandum on Panel Design Options. Task 2.1.1. NYMTC Regional Household Travel Survey, prepared for the New York Metropolitan Transportation Council, Dec. Zumkeller, C. 2007. Mobility Panel Surveys: The German Experience. Universität Karlsruhe Research University, Germany, PowerPoint presentation in Paris, Oct. 8. Zumkeller, D., B. Chlond, and M. Kagerbauer. 2008. Regional Panels against the Background of the German Mobility Panel. Institute for Transport Studies, Uni- versity of Karlsruhe, Germany, prepared for the 8th International Conference on Survey Methods in Transport, Annecy, France, May 25–31. Zumkeller, D., and P. Ottmann. 2009. Moving from Cross-Sectional to Continuous Surveying: Synthesis of a Workshop. In Transport Survey Methods: Keeping Up with a Changing World (Bonnel et al., eds.), Emerald Group Publishing Limited, Bingley, United Kingdom, pp. 533–539.