Making Trucks Count: Innovative Strategies for Obtaining Comprehensive Truck Activity Data (2014)

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.

9 C H A P T E R 2 This chapter evaluates the current state of truck-activity data. Truck-activity data encompass many datasets and sta- tistical reports. Ideally, such data should provide complete information on the movement of trucks on U.S. road net- works, including the number of miles and the roads traveled, as well as goods and weight hauled. As this assessment of pri- mary data sources reveals, there are many limitations to the current state of the practice. The researchers began the study by reviewing statistical databases and reports on truck activity. They developed pro- files of each, summarizing their definition of truck, data items and their source and method, responsible agency, and years covered in each source. The researcher also noted the back- ground and purpose, coverage, sources, reliability, time series, and availability for each. The team derived nearly all informa- tion from published sources. Appendix A provides a descrip- tive profile of each, as well as a matrix that systematically arrays the information from each source. The study team also examined the truck-activity data for a subset of these sources to determine what issues and limi- tations exist that could be overcome through innovative strategies for data gathering or reporting. Where possible, the team examined data elements to better understand the following: • How the data was generated (e.g., traditional survey, roadside intercept survey, sensors, GPS); • What auditing procedures were used (e.g., manual or auto- mated review); • How the data are maintained (e.g., single data collection project or ongoing program); • How the data can be accessed (e.g., only with summaries on paper, microdata); • How the data are archived (e.g., central or decentralized system, on CD only); • Whether metadata availability exists (e.g., electronic listings, paper details); and • What were previous uses of the data (e.g., listings of reports or analyses produced from a particular dataset). The sources are profiled below. Appendix B contains a detailed evaluation of each of these sources in terms of its methodology, quality control, stewardship, accessibility, archival practices, and metadata availability. The text that follows presents a high-level summary of the detailed infor- mation in Appendix B. 2.1 U.S. Census Bureau Data The quinquennial economic census is the standard used by federal statistical agencies in classifying business establish- ments for collecting, analyzing, and publishing statistical data economic data. Among its elements are the Commodity Flow Survey (CFS), conducted with the Bureau of Transportation Statistics, and the Vehicle Inventory and Use Survey (VIUS), conducted from 1963 to 2002. Assessment of Primary Data Sources for Truck-Activity Data CHAPTER 2 KEY TAKEAWAYS • A weakness for many current data sources is difficulty of use. • Transparency in data sources and auditing procedures is a problem across nearly all sources. • New innovative sources would need to focus on statistical rigor, quality assurance, and accessibility. Improvements could either expand existing sources or offer a method for integrating and synthesizing data.

10 2.1.1 Commodity Flow Survey The CFS is the most comprehensive public tool for under- standing U.S. freight flows. It has evolved continually since its introduction in 1993. In its latest iteration, the CFS provides origin-destination flows among and within state portions of major metropolitan areas and balances of states for economic census years ending in 2 and 7. While CFS data are collected from over 100,000 establishments and are based on 4.9 million shipments, the CFS has coverage limitations, such as imports and farm-based shipments, and it is not timely. The FAF is built on the CFS and uses additional data and models to fill in gaps in coverage and provide annual estimates. The FAF has less commodity detail than the CFS and does not include several CFS shipment attributes such as shipment size. Local data collection is necessary to provide greater geographic detail for either the CFS or the FAF as indicated in NCFRP Report 26. 2.1.2 Services Annual Survey The U.S. Census Bureau also conducts the Services Annual Survey (SAS), a detailed high-level report of activity in selected industries, including Transportation and Warehousing, which are North American Industry Classification System (NAICS) codes 48 and 49 (Trucking and Warehousing are only two of the dozen sectors covered in the survey). SAS has its roots in the Transportation Annual Survey, which originated in 1985. These datasets provide information that is self-reported directly by firms and not directly available from any other source. NAICS coverage includes all carriers, both employer establishments and non-employers (owner-operators). The survey covers all for-hire (TL, LTL), heavy and tractor-trailer, light or delivery services identified by NAICS code—private carriage is not included. Data include motor carrier revenue by commodity classes, end of year fleet size by type, fuel expen- ditures, payroll, and purchased freight transportation. Although these data are excellent for understanding general features of the transportation industry, and especially for-hire trucking, its summaries can only serve as a baseline to understand trends because SAS measures revenue and expenses and not trucking activity. Among the advances of these data are up-to-date high-level information on trucking industry operations, and the ability they offers to calibrate older data with current metrics. 2.1.3 Vehicle Inventory and Use Survey (VIUS) Discontinued VIUS provided data on the physical and operational char- acteristics, such as VMT, on trucks. Conducted every 5 years as part of the economic census, the primary goal of VIUS is to produce national and state-level estimates of the total number of trucks. Its truck inventory data were used by several state and federal agencies for developing transportation plans, analyzing highway safety issues and environmental impacts of emissions, and creating studies on vehicle performance, fuel demands, and fuel conservation practices. Some have suggested restarting VIUS for monitoring data on heavy-duty trucks, which will soon have new fuel-economy standards, and for differentiating commercial and personal use of light- duty trucks. VIUS would also be the only source of data on difficult-to-locate operators of vehicle fleets (Transportation Research Board, 2011). Strengths of the data include its detail on vehicles that could be used to understand U.S. freight movement, includ- ing configuration and commodities hauled. In addition to its discontinuation in 2002, challenges include its quinquennial collection and separate imputation processes for missing data on length, average weight, and annual mileage for trucks. 2.2 Amalgamated Datasets Amalgamated datasets use advanced post-processing tech- niques on a number of different sources to depict U.S. highway freight movements. Among these are the FAF, freely available from the Federal Highway Administration (FHWA) office of Operations, and IHS Transearch data, commercially available from IHS Global Insight. Both datasets cover the U.S. high- way system. Although Transearch has more granular O/D and is updated more frequently, its closed and proprietary nature makes it difficult to ascertain its quality and reliability. 2.2.1 The FAF The FAF estimates tonnage, value, and ton-miles of all goods shipped to, from, and within the United States by ori- gin, destination, commodity type, and mode. The FAF is based primarily on the CFS, and uses the CFS geography of state portions of major metropolitan areas and balances of states plus eight foreign regions. The FAF has less commodity detail and fewer shipment characteristics than the CFS, but has complete coverage of goods movement and includes annual updates between CFS years. In addition to origin-destination data, the FAF assigns truck tonnages to individual routes. The FAF and CFS divide commodity movements by truck among three modal categories: truck only, multiple modes and mail, and other and unknown. For national and state totals, the CFS distinguishes types of multiple modes such as truck and rail. Estimates of truck activity based on the truck- only category miss trucking activity that can be significant in selected areas. Neither the FAF nor CFS provides local origin-destination flows such as for pairs of counties. Several techniques exist to disaggregate FAF region-to-region flows to the county-to-

11 county level, but the statistical reliability of these techniques is highly suspect and remains to be thoroughly tested. Most, if not all, are based to some extent on County Business Patterns that, like the FAF, is a composite estimate of economic activity rather than direct observation. As described in NCFRP Report 26, supplemental data collection is required to get local detail. 2.2.2 IHS Global Insight Transearch Data This dataset originated when Reebie Associates invited truck- ing companies to “volunteer” their shipping data in exchange for all other data Reebie collected. IHS Global Insight later bought Reebie and integrated its transportation data with Global Insight’s wealth of economic data. Many of its data elements are similar to those of the FAF. The amalgamated data is generally available at the county level rather than the broad economic regions of the FAF and CFS. Transearch reports to have a sample size of more than 70,000 shipments. Its data, however, are limited to those firms that contribute data, and therefore may not be representative like CFS. The restriction of the data also can make it expensive— and maybe even inappropriate—for public planning. With no opportunity for oversight or verification, there could be the appearance, or potential, for biases in the analysis due to undis- covered mistakes, or purposeful targeting of projects, places, individuals, or modes. 2.3 Roadway Operations Data Sources Roadway operations data are data that are measured directly from road or vehicle operations. There are two categories of these data: traffic count data and GPS data. Traffic count data can be subdivided into short and continuous categories, and again into weigh in motion (WIM), classification, and speed and volume data. 2.3.1 ATR Classification Count Data There are more than 6,000 automatic traffic recorders (ATRs) in the United States; only about 2,000 can differenti- ate among FHWA vehicle classes. These stations provide an insight into fleet mix on all different road classifications in each state. Although the data from class count stations are not as detailed as WIM data, the greater number of stations can make the dataset very useful in understanding highway truck flows. The continuous count classification data are collected monthly from each state by the FHWA Office of Highway Policy Information as a part of their Traffic Monitoring Anal- ysis System (TMAS). These data are be used to calibrate data obtained from the FAF, especially when FAF data may have become unreliable as a result of economic or other changes since last published. There are some challenges to their use, because they are not centrally available to the public. This, in turn, makes it difficult to determine how complete or uni- form the data are, as well as to develop a nationally represen- tative sample of the data. 2.3.2 Weigh-in-Motion (WIM) Data WIM devices are designed to capture and record axle weights and gross vehicle weights as vehicles drive over a measurement site. Unlike static scales, WIM systems do not require the vehicle to stop. Every state department of transportation has a WIM pro- gram responsible for the upkeep of WIM stations and archiving and analyzing the collected data. All states are required to sub- mit their WIM data annually to FHWA, which uses it as a mea- sure in determining Truck VMT as well as for use in creating Vehicle Travel Information System reports. Virtual WIM stations are now being added to WIM Pro- grams. Such stations use cameras to capture images of each passing truck and software to digitize information such as license plate numbers. This additional information can be used to link data from WIM stations to other datasets, such as truck size and weight permits, or even with WIM data from other stations. Although these data have a centralized stan- dard, they do not have a centralized repository, which makes it difficult to conduct analyses using multiple years of data. Data quality procedures can also vary by state. 2.3.3 Highway Performance Monitoring System (HPMS) The Highway Performance Monitoring System (HPMS) provides data on the extent, condition, performance, use, and operating characteristics of the nation’s highways. It was developed in 1978 as a national highway transportation sys- tem database. It includes limited data on all public roads, more detailed data for a sample of arterial and collector func- tional systems, and certain statewide summary information. HPMS replaced numerous uncoordinated annual state data reports as well as biennial special studies conducted by each state. These special studies had been conducted to support a 1965 congressional requirement that a report on the condi- tion of the nation’s highway be submitted to Congress every 2 years. HPMS supports development and evaluation of the Admin- istration’s legislative, program, and budget options. HPMS pro- vides the rationale for Federal-Aid Highway Program funding level requests, and is used for apportioning federal-aid funds back to the states under TEA-21. HPMS includes TMAS, which has all continuous count traffic data, including class counts and WIM data, from across the country since the 1990s. It also

12 supports Traffic Volume Trend Reports, a monthly report of vehicle miles traveled for each state as deduced from the base- line HPMS data. HPMS uses ATRs to estimate percentages of single-unit and combination trucks; these estimates, in turn, can be used to develop VMT for these types of trucks (Beagan et al., 2007). However, HPMS is not available in its published form from a centralized database. In addition, the distribution of sensors was originally designed to estimate passenger vehicles using volume counts. Increasing the accuracy of truck counts would require using more classification count and weigh-in-motion equipment. The lack of microdata makes it difficult to assess data completeness. Quality procedures may vary by state. 2.4 Vehicle-Based Operations Data The rapid progress of GPS and cellular technology has led to rapidly increasing use of GPS data for fleet management. Although this data is used mostly in real time for trucking industry operations, such data, if archived, could be a power- ful tool for understanding truck freight behaviors. GPS data can provide speed, latitude, longitude, and heading of trucks in operation. Gathering detailed O/D information requires special data handling techniques to deal with privacy con- cerns. Although many users of GPS do not archive their data, truck logistics companies have many uses for it and have been collecting it for years. Companies do not share the data, though many firms are now gathering GPS data from public and private parties. Studies conducted by the Washington State Department of Transportation using truck GPS data have included research aimed at understanding freight bottlenecks and the impact of construction on freight flows, as well as developing good quality checking tests for these data as well. Truck GPS data could provide a robust form of truck-activity data, with suf- ficient sample size and advances in collection and processing. 2.4.1 Freight Performance Measure Program The Freight Performance Measure Program (FPM) is an ini- tiative to develop freight-specific performance measures and to promote the practice of freight performance measurement to help identify needed transportation improvements and monitor their effectiveness. As part of FPM, FHWA is advanc- ing tools and research for multimodal freight performance measurement, especially in the use of truck probe data to ana- lyze freight performance on highways. The most recent tool FHWA provides is the newly acquired National Performance Management Research Data Set (NPMRDS) consisting of a passenger and freight average travel time probe reported every 5 minutes on the national highway system that is available for use by states and MPOs. This dataset is provided through a contract with HERE, formerly Nokia. HERE acquires the freight data from probe data collected by ATRI, the research arm of the American Trucking Associations. ATRI’s ties to the trucking industry have helped it collect detailed GPS data over a number of years that other agencies have not been able to collect. Information on the quality of this data is available to users and has been assessed as part of this report. Additionally, FHWA contracts with ATRI directly to assist FHWA with the development of tools and analysis based on the probe data they provide. Two tools have been in use to under- stand performance using the probe data, Freightperformance. org or FPM Web and the National Corridors Analysis and Speed Tool (N-CAST). FPM Web allows the user to sub- mit a query, specifying a highway or group of highways, and the time interval. It returns the average speed on that high- way in 3-mile segments. This data can be downloaded as a CSV file. The site also includes instructions for visualizing the data with an available corresponding ESRI shapefile for ArcMap 9.3 or later. N-CAST is a second-generation prod- uct with greater coverage of the national highway system and 1-mile segmentation. In its current state, it does not allow users to query by time and only shows the average speed for A.M., midday, and P.M. peaks without indicating when the data was gathered. Because these datasets have not been publicly assessed, the research team was not able to evalu- ate their quality procedures. The team also was not able to judge how complete the data are. Some information related to this data can be derived from the quality reports associ- ated with the NPMRDS. FHWA is currently assessing these tools and determining what next-generation format would be most useful for users. Through ATRI’s data and tools, FHWA is also able to analyze the probe data to understand corridor performance, aggregated origin and destination information, locations of significant freight activity, freight node perfor- mance, and incident/weather impacts. At the time of this report, FHWA is developing a significant body of freight performance research, some of which is evalu- ating probe use and developing approaches and resources, as well as identifying new tools and data to consider. 2.5 Administrative Data Sources For a truck to operate anywhere in the United States it must be registered with at least one state Department of Motor Vehicles. A truck operating in more than one state (or Canada) may register with the International Registration Plan (IRP) and report the number of miles it plans to travel in each state. Trucks hauling goods over legal size or weight limits are required to get a permit from each state to be traveled. Each year, a massive amount of data is collected regarding trucking operations for administrative datasets, yet these are

13 little used in planning or research. This lack of use could be attributed to the lack of access to the data, although truck registration data have been provided upon request (Lawson et al., 2002). The Federal Motor Carrier Safety Administration (FMCSA) requires companies operating commercial vehicles transporting passengers or hauling cargo in interstate com- merce to obtain a U.S.DOT Number. Commercial intrastate hazardous-materials carriers who haul quantities requiring a safety permit must also register for a U.S.DOT Number that serves as a unique identifier when collecting and monitoring a company’s safety information acquired during audits, compli- ance reviews, crash investigations, and inspections (see http:// www.fmcsa.dot.gov/registration-licensing/registration- U.S.DOT.htm). 2.5.1 Federal Motor Carrier Management Information SYSTEM (MCMIS) FMCSA manages their vast information resource on the safety and fitness of commercial motor carriers and hazardous material shippers through MCMIS (see http://mcmiscatalog. fmcsa.dot.gov/). The MCMIS system has its roots in the Motor Carrier Safety Act of 1984 and the Commercial Motor Vehicle Safety Act of 1986. There are several types of reports available from MCMIS, as follows: • Crash File Extracts: This data extract describes commer- cial vehicle crashes reported to FMCSA. • Crash Count Report: Shows the number of federally record- able crashes by user-specified categories. A “recordable” crash is one that has a fatality, an injury requiring medical attention away from the scene of the crash, or a vehicle towed away. • Inspection File Extract: The FMCSA Inspection File con- tains data from state and federal inspection actions involving motor carriers, shippers of hazardous materials, and trans- porters of hazardous materials operating in the United States. Most inspections were conducted roadside by state personnel under the Motor Carrier Safety Assistance Program. • Inspection Count Report: Shows the number of inspec- tions by user-specified categories. • Company Safety Profiles: Are the most comprehensive summary of a specific carrier’s national safety performance. • Census File Extracts: Provide descriptive information on every active company in the MCMIS Census File. The file from which the extract is generated is updated biweekly and currently contains more than 1.5 million interstate carriers and hazardous materials shippers. It also has many broader uses for transportation planning, containing records for the roughly 1.7 million companies or entities subject to the Federal Motor Carrier Safety Regulations or Hazardous Materials Regulations. Census file extracts offer more than 130 data elements, including physical address information, the number and type or trucks owned and leased, the types of commodities hauled by the entity, the number of drivers hired, whether the entity ships less or more than 100 miles, and whether it ships by inter- or intra-state routes. It also includes annual VMT by entity. These data are mainly collected through mandatory biennial self-report forms, although safety and crash data also are aggregated for the census file and other MCMIS reports. Many entries in the census file have incomplete fields and have not been updated recently. In addition, data quality pro- cedures for these files are not documented or publicly available. 2.5.2 International Registration Plan (IRP) Data The International Registration Plan was initially developed in the 1960s and early 1970s as a means of replacing the sys- tem of registration reciprocity, which then prevailed and was rapidly becoming inadequate to meet the needs of expanding interstate and international commerce. With the related Inter- national Fuel Tax Agreement, the IRP is unique in that it is an inter-jurisdictional agreement administered and managed by the states and provinces that are its members without any significant federal involvement (International Registration Plan, 2011). All apportioned vehicles must be registered under IRP. The plan defines an apportionable vehicle as one that is, among other things, used in two or more member jurisdictions for transportation of persons for hire or of property and exceeds 26,000 pounds or has at least three axles (excepting buses and recreational vehicles). IRP has estimates of the number of miles traveled by jurisdiction for each truck registered. Unfortunately, its data are not currently available to the public for research. 2.5.3 Truck Oversize – Overweight Permitting Data According to federal law, no truck can operate with a gross vehicle weight greater than 80,000 pounds, have a single-axle weight greater than 20,000 pounds, or have tandem axles with weight greater than 34,000 pounds. In addition to the regula- tions on weight, there are numerous regulations on the size of trucks that can operate on certain roads. Trucking compa- nies may apply for permits to transport oversize loads (such as manufactured homes or transformers) that, if divided, would be unsuitable for their intended purpose or require more than 8 hours to dismantle using appropriate equipment. States may also grant permits for oversize divisible loads such as fuel, logs, milk, or trash. The resulting permit data includes details on oversize and overweight truck shipments in each state, but

14 these data are not currently available for public research. Such data also are collected by each state transportation depart- ment, with availability varying by year. Information generated in response to these questions, along with other relevant factors, was assembled into tables and graphs to help illustrate the relationships across truck-activity datasets, with an opportunity to look for patterns of strengths and weaknesses within and between datasets. This process pro- vided additional insight into what is currently available. Table 2-1 summarizes the datasets reviewed, including their frequency of publication and elements. It illustrates the patchwork nature of the truck-activity data infrastructure, indicating that users may need to piece together information from many sources to answer critical policy questions. At the same time that there are critical gaps, there is also overlap among sources in the data that are available. Fig- ure 2-1 illustrates overlapping data elements in the data- sets. The researchers realize that these datasets are not really Dataset Avail- ability VMT Tons Ton- Miles Commod- ities Speed Weight O/D Volume Counts CFS 5 years --- X X X --- --- X --- FAF 5 yearsa --- X X X --- --- X X VIUS Dis- continued --- X X X --- X --- X Transearchb Annual --- --- --- --- --- --- --- --- HPMS Monthly --- --- --- --- --- --- --- X WIM > Monthly --- --- --- --- --- X --- X Class Counts > Monthly --- --- --- --- --- --- --- X FreightPerformance.org > Monthly --- --- --- --- X --- --- --- GPS > Monthly --- --- --- --- X --- X --- IRP** Monthly --- --- --- --- --- --- --- --- Oversize/Weight > Monthly --- --- --- --- --- X X --- SAS Annual X X --- X --- --- --- --- MCMIS Annual X X X X --- --- --- --- a FAF availability is 5-year benchmarks and annual estimates. b The researchers were not able to obtain any samples of Transearch data so they used secondary sources to write about it. Table 2-1. Data availability and elements. Figure 2-1. Diagram of data areas and datasets.

15 comparable and have different objectives, but can be used to provide an overview of the overlap. The MCMIS dataset and the two amalgamated datasets, FAF and Transearch, contain commodity, tons, and routing and volume. MCMIS is administrative data (an updated census), while both of the other sources are amalgamated sources. The methods of the FAF are available but can only be applied to very large geographies. Transearch data integration is not known and most likely is not based on a random sample. Many statistical techniques can address bias of non-random samples, but are difficult to apply without knowing the under- lying methods of Transearch data. Furthermore, Transearch provides county-level estimates, while certain types of trans- portation planning efforts for freight need corridor- or subcounty-level data. Drilling down into potential quality issues, Table 2-2 sum- marizes the dataset according to other data quality character- istics. The quality assessments were judgments of the research team. The value definitions are presented following Table 2-2. The datasets available to freight planners often are difficult to compare in their potential use because each one is designed to address different needs in different communities of planners Table 2-2. Data quality characteristics. Dataset Centralized Data Availability Years Available Data Completeness Price Frequency of Collection Quality Pro- cedures Metadata Avail- ability CFS 5 5 4 5 3 5 4 SAS 4 5 5 5 3 5 4 VIUS 5 2 5 5 0 5 5 FAF 4 5 5 5 3 4 5 Transearch 5 5 5 1 5 n/a 2 ATR 2 2 2 5 5 3 5 WIM 2 2 2 5 5 2 5 HPMS 3 3 2 5 4 3 4 Truck GPS 1 1 n/a 3 5 4 2 ATRI 5 4 n/a 5 n/a n/a 5 MCMIS 4 4 3 4 5 3 4 IRP 2 2 4 n/a 5 n/a 1 Oversize 2 2 5 n/a 5 n/a 1 Definitions for Values Centralized Data Availability: 1—neither summaries nor microdata publicly available; 2— microdata collected by many agencies but not uniformly available or uniformly available only as summary data; 3—data collected by many agencies and is uniformly available but not in a centralized location; 4—data collected or reported to central location but quality and format type varies due to lack of centralized procedures; 5—data collected or reported to central location, with central quality control and easily available microdata. Years Available: Historical data is 1—not available, 2—available in some cases but not uniformly, 3—uniformly available for 10 years, 4—uniformly available for 20 years, 5— uniformly available for 30 years. Data Completeness: 1—data unusable, 2—significant problems to use, 3—data issues occur but are documented and do not pose a significant problem, 4—data can be shown to be complete, 5—data completeness is considered in creating data and is measured. Price: 1—more than $10,000; 2—$1,000 to $10,000; 3—$1 to $1,000; 4—free but processing fees; 5—free and publicly available. Frequency of Collection: 1—discontinued, 2—every 10 years, 3—every 5 years, 4—annually, 5—monthly or more frequently. Quality Procedures: 1—none; 2—limited, insufficient, or not publicly available; 3— sufficient and publicly available; 4—publicly available with data in pre- and post-processing forms; 5—specific data quality procedures uniformly applied. Metadata: 1—not available, 2—exists but not publicly available, 3—publicly available, 4— publicly available with documentation, 5—publicly available with documentation and examples.

16 and engineers. Nevertheless, two characteristics of use offer some basis of comparison: completeness and timeliness. Fig­ ure 2­2 ranks each dataset by these attributes based on the scores in Table 2­2. These data are not publicly available for analysis of a number of important characteristics (e.g., auditing tech­ niques) because they are produced by a private­sector company. In some cases, it might be possible to correlate Transearch data with FAF and CFS data. A correlation between Tran­ search and CFS/FAF would identify which portions of Tran­ search used CFS/FAF numbers to fill in missing coverage. Providing standard practices for such correlations could be useful for cross validation. Figure 2­2 also indicates that while other datasets are scored lower on the comprehensiveness criterion, they were collected and processed much more fre­ quently than other datasets. It may be possible to unify these more frequent datasets using a data architecture approach (such as the Freight Data Architecture program described in NCFRP Report 9), making it possible to increase the com­ pleteness and timeliness of publicly available datasets. Among those datasets for which all features are known, SAS ranks highest overall on statistical characteristics, qual­ ity assurance, and stewardship, followed by VIUS and CFS. Statistical characteristics are primarily based on the use of a random sample if data are not a complete census of opera­ tions. Administrative data and that on roadway and vehicle operations could all be available as a complete census going forward. Data products produced through the economic census follow strict procedures to ensure generalizability. Quality assurance is highest for SAS and CFS. Establishing a trans parent, documented set of procedures is critical for any future data to ensure all users understand data conditions. Dimensions of stewardship vary by dataset. Roadway opera­ tions, vehicle operations, and administrative data (except MCMIS) score low on centralized data availability but can be formatted well and have strong quality procedures, making it possible to produce a decentralized system with analytical value. At this time, Transearch and some GPS data have asso­ ciated costs, while some “free” data are difficult to obtain due to institutional issues. Key findings from the review of the limitations and chal­ lenges associated with current data sources follow. • CFS is the basis for understanding freight in America. For collecting freight activity in the metrics of tons and ton­ miles of goods transported, all current data sources depend on CFS. CFS is conducted only every 5 years, making it difficult to accurately model intervening years. The FAF is directly based on CFS, and Transearch is likely calibrated by it. Improving CFS by increasing its reach, granularity, and frequency would lead to downstream improvements for all other datasets. • MCMIS provides tons and ton­miles in their adminis­ trative data program. Its continual updating provides an opportunity (as well as a challenge) for analysis. • A large issue for many current data sources is usability. Many require a great deal of technical knowledge to use and visualize the data. For example, operations data (e.g., roadway and vehicle based) contain a great deal more Figure 2-2. Timeliness and comprehensiveness of data sources.

17 information than are currently “harvested” and used for statistical estimation. There is a great opportunity to help planners at all levels by creating tools that could make using these data more intuitive. • A widespread weakness is that there is no method for verifying modeled truck-activity data, such as the FAF or Transearch. A method or tool that could be used to verify these data would be of great value to users. It also might be possible to use the MCMIS on some specific factors as a source for verification and validation. • Transparency in data sources and auditing procedures is a problem across nearly all the reviewed data sources. This review of current primary data sources for truck- activity data provided direction to the research team in terms of the types of new innovative strategies to consider. An improved, comprehensive source would either be (1) an expansion to an existing source so that it would con- tain more of the needed measures than are currently avail- able or (2) a method for integrating and synthesizing data, such as the FAF, that provides more of the needed measures. A new innovative source would need to focus not only on critical data gaps, but also on data elements for addressing the important areas of statistical rigor, quality assurance, and accessibility.