Forecasting Statewide Freight Toolkit (2008)

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.

16 In order to forecast statewide freight flows, data are needed to develop and validate the models and methods used as inputs. Quality and precision are the keys to freight modeling, with the accuracy of the freight flow forecast dependent on the accuracy of the database. If the underlying database is not complete and correct, then the estimated freight flow will be inaccurate. This section of the Toolkit identifies data sources unique to freight forecasting or applied to freight forecasting in a unique way. General data sources used in transportation forecasting should be famil- iar to users of the Toolkit and will be mentioned briefly. Freight-specific data sources for important databases also will be briefly summarized, while sources used in the case studies will be described in more detail. 5.1 Model Development As described in Sections 4.0 and 6.0, statewide freight fore- casting methods employ a variety of techniques, models, and formulas for processing data. The nature and form of the equations and the values for their coefficients and parameters are determined through a model development process famil- iar to those that have developed passenger forecasting models. The sources of this data for freight are described below. Local Surveys The construction of a passenger transportation forecasting model often begins with a travel survey. A travel survey gath- ers information about the number of trips, the purpose of these trips, the time the trips were taken, the cost, the distance traveled, the mode choice, and information about the trav- eler. A travel survey thus provides the behavioral data needed to establish the trip generation, trip distribution, mode split, and assignment relationships specific to a study area. The survey size must be designed to provide a statistically valid sample of all potential travelers. When conducting a survey of freight movements, one encounters a basic problem: determining the size of the mar- ket that should be surveyed. Conducting a cordon survey around an entire state boundary is generally impractical, and matching vehicles passing through a statewide cordon can be extremely difficult. Cordon surveys do not usually provide information about the contents of vehicles, such as, com- modity information, which makes it impossible to tie the freight flows back to economic development data. Generally, shipper and carrier surveys prove more man- ageable. In a shipper survey, a major shipper is asked to fill out a form detailing each shipment dispatched in a given time period. The information collected might include the type of commodity, the place of origin, the destination, the transport mode or modes, the dollar value and physical volume of the shipment, and other general information. In a carrier survey, a major carrier is asked to detail all the shipments carried and possibly also the route chosen. With the consent of the car- rier’s management and staff, electronic driver information systems may be used to collect similar data. Determining a statistically valid sample of shippers and carriers for a specific statewide survey and the appropriate expansion factors is extremely difficult and expensive. However a shipper diary survey is regularly conducted by the U.S. Census Bureau for the CFS. Compilations Developers of freight forecasting models may wish to avoid the expense of conducting a behavioral survey and instead use the rates, coefficients, and relationships developed by others. While not as well developed as those for passenger planning, several publications provide values that can be used in truck generation and distribution models. These publications include the Institute of Transportation Engineers’ Trip Generation Handbook7, the National Cooperative Highway Research Program’s Truck Trip Generation Data8, and the C H A P T E R 5 Data Sources

Federal Highway Administration’s Quick Response Freight Manual3, and Accounting for Commercial Vehicles in Urban Transportation Models.9 National Surveys Shipper surveys typically require a nationwide sample. While a survey of shippers within the target market area will provide a sound picture of outbound shipments, this limited coverage will otherwise miss inbound activity. To avoid the size and complexity of conducting such a study for an indi- vidual freight forecasting project, existing surveys may be obtained. The two most common surveys, the CFS and TRANSEARCH, are described in detail as part of this Toolkit. Also described is the FAF’s Commodity Database, a publicly available database created from TRANSEARCH. Commodity Flow Survey CFS is conducted every five years as part of the U.S. Cen- sus Bureau’s Economic Census and is designed to provide data on the flow of goods and materials by commodity, ori- gin, destination, and mode of transport. Prior to 1997 the CFS reported commodity information using the STCC code. Beginning with the 1997 CFS, the SCTG codes were used, providing a more modern focus and a better link to industry classification and output measures. Due to variations in methodology, sample size, and other changes, the CFS is not particularly consistent from year to year, making it hard to build time-series data. Nonetheless, the CFS remains the only shipper survey to which a response is mandatory. As such, it is less likely to be biased than other shipper surveys. In terms of statewide forecasting, the CFS presents diffi- culties. Since the survey is predominantly designed to map national-level traffic, the small sample size within each state means that data must be aggregated to preserve shipper con- fidentiality. Consequently, origin and destination data are publicly released on CD-ROM at two aggregation levels: state-to-state and between 86 of the largest metropolitan areas (portions within the primary state boundary only). Fur- thermore, the 1997 data are available only as predefined data files through a browser, as shown in Figure 5.1. While com- modity, origin, destination, and mode information is avail- able, only three of the four characteristics are reported in any one table. Individual tables by origin, destination, and com- modity must be transformed and aggregated to produce a national database. Additional processing is necessary to esti- mate data that is aggregated or suppressed to preserve confi- dentiality. 17 Figure 5.1. Commodity flow survey.

The 2002 data has been collected and partial releases began to be made available in 2004. The 2002 survey excluded ship- ments by establishments classified in the North American Industry Classification System (NAICS) as farms, forestry, fish- ing, government agencies, construction, transportation, and most retail and service industries. The 2002 survey also excluded shipments from logging establishments, because under NAICS the classification of this industry moved from manufacturing (included in the scope of the CFS) to agriculture (out-of-scope for the CFS). The CFS is a survey of domestic establishments and measures shipments leaving an establish- ment’s facility, and it includes exports but not imports (unless the imported goods are received by an included domestic busi- ness at the port of entry and reshipped by that business). The 2002 CFS also excludes shipments of crude petroleum by the oil and gas extraction industries because of issues with how these companies record and report shipment information. The 1997 CFS is available on CD-ROM from the U.S. Census Bureau at http://www.census.gov/econ/www/cfsmain. html. The 2002 CFS will be available in February 2005. TRANSEARCH TRANSEARCH is a database of freight traffic flows avail- able from Reebie Associates. Although proprietary, it also is the most commonly used source of freight data; four of the case studies reported in this Toolkit rely on TRANSEARCH. TRANSEARCH uses several mode-specific data sources to create a picture of the nation’s freight traffic flows on an ori- gin to destination commodity basis, refining the geographic market identification to the county level. TRANSEARCH is updated annually using the following sources: 1. Annual Survey of Manufacturers by state and industry; 2. Surface Transportation Board (STB) Carload Rail Waybill Sample of market-to-market rail activity by industry; 3. Army Corps of Engineers waterborne commerce data describing market-to-market water activity by industry; 4. Federal Aviation Administration (FAA) enplanement sta- tistics and airport-to-airport cargo volumes; 5. Rail, water, and air freight flow data deducted from the Bureau of Census Annual Survey of Manufacturers (ASM)- based production data; and 6. Reebie Associates’ proprietary Motor Carrier Data Ex- change Program, which provides information on actual market-to-market trucking industry movement activity. The truckload sample covers about 6% of the market, and Reebie Associates’ less-than-truckload sample is about 40%. In total, information is received on over 75 million individual truck shipments. TRANSEARCH’s county-to-county market detail is devel- oped through the use of Reebie Associates’ Motor Carrier Data Exchange inputs and its Freight Locator database of shipping establishments, which provides information about the specific location of manufacturing facilities, measures of facility size (both in terms of employment and annual sales), and a description of the products produced. Primary coverage of truck traffic is limited for nonman- ufactured products. For manufactured products informa- tion is provided using the STCC Code, which can be aggregated from a four-digit level. Supplemental material for agricultural and mining resource extraction shipments from the source to a processing plant not ordinarily cov- ered in commodity flow surveys is available for an addi- tional charge. Traffic movements originating in warehouses or distribu- tion centers or drayage movements of intermodal rail or air freight are shown as STCC 50. These are by definition truck movements. Movements to warehousing and distribution centers may be by other STCC codes and by any mode. Details on the types of items being moved in STCC 50 are not available. The CFS defines the use of multiple modes, such as truck and rail, as a separate mode. The TRANSEARCH database, shown in Figure 5.2, is an unlinked trip table that reports the portion of a trip by each mode, and in some cases submodes, separately. This allows the volume of shipments at inter- modal transfer points to be identified, but the information on the lining of the trips is lost. As discussed, TRANSEARCH is constructed from many commercial and public sources of data, representing domes- tic and NAFTA trade flows. Economic modeling is used to adjust the surveys where data is lacking or confidential and to check elements such as spatial patterns and logic. Given the complexity of its sources and the additional analysis that is undertaken, the construction of TRANSEARCH cannot be easily summarized. This inability to completely document all elements and proprietary sources has led to some concerns by some users about the data’s inclusiveness. Despite these con- cerns, TRANSEARCH is an accepted freight database widely used for planning by the FHWA, many U.S. states and met- ropolitan planning organizations (MPOs), as well as private freight carriers and shippers. The inland or surface movement of import and export traffic volumes to locations outside of North America is included in the data but only to and from the location where the freight crosses the U.S. border. However, the flow patterns of this freight are based on the movement patterns of domestically sourced goods in the same market areas and are not the actual movements of the import/ export freight. 18

TRANSEARCH is available for purchase from Reebie Associates at http://www.reebie.com. STB Carload Rail Waybill Sample STB is the official authority of the Carload Waybill Sample. Railroads terminating over 4,500 cars per year are required to file a sample of waybills with the STB. The primary purpose of the Carload Waybill Sample is regulatory oversight. The Waybill Sample contains rail shipments data such as origin and destination points; type of commodity; number of cars, tons, revenue; length of haul; participating railroads; inter- change locations; and Uniform Rail Costing System shipment variable cost estimates. It contains confidential information and is used primarily by Federal and state agencies. While the Waybill Sample is not available for public use, a public-use version contains aggregated nonconfidential data. Move- ments are generally aggregated to the Bureau of Economic Analysis (BEA) region to BEA region level at the five-digit Standard Transportation Commodity Code level. The STB Waybill Sample is a stratified sample of carload waybills for terminated shipments by railroad carriers. Army Corps of Engineers Waterborne Commerce Data Waterborne traffic movements are reported to the Army Corps of Engineers by all vessel operators. The reports are gen- erally submitted on the basis of individual vessel movements completed. For movements with cargo, the point of loading and the point of unloading of each individual commodity must be delineated. Military cargo moved in commercial ves- sels is reported as ordinary commercial cargo; military cargo moved in Department of Defense vessels is not reported. In summarizing the domestic commerce certain movements Cargo carried on general ferries; coal and petroleum products 19 Figure 5.2. TRANSEARCH database.

loaded from shore facilities directly into bunkers of vessels for fuel; and insignificant amounts of government materials (less than 100 tons) moved on government-owned equipment in support of Corps projects. Foreign commerce data are fur- nished to the Corps of Engineers by the Bureau of the Census under a working arrangement sponsored by the Office of Management and Budget. Freight Analysis Framework Commodity Database FAF, described in Section 8.5, produced a Commodity Flow Database (CFD) that provides O-D information on commodity flows by mode for the years 1998, 2010, and 2020. These flows, given in tons, are organized by commodity and mode. The CFD is divided into domestic flows (state-to- state) and international flows. The data are available in both Microsoft Access 2000 (*.mdb) format and tab-delimited text (*.txt) format, the latter suitable for importing into a Microsoft Excel spreadsheet. A set of lookup tables of the STCC Commodity and Federal Information Processing Sys- tem (FIPS) codes for states also is provided. Separate tables are provided for domestic, international, international air, and petroleum flows. For domestic flows, the Federal Highway Administration provides a single file that contains state-to-state freight flows by commodity and mode for 1998, 2010, and 2020. Figure 5.3 illustrates the Microsoft Access data format. The first and sec- ond columns indicate a state freight flow from origin “05” to destination “06.” The FIPS reference table translates this to a flow from Arkansas to California. The next three columns indicate rail flows of 4,800 tons in 1998, 5,772 tons in 2010, and 5,930 tons in 2020. The sixth column labeled STC corre- sponds to the STCC reference table. In this case, “01” corre- sponds to farm products. Note that the O-D pair in the database is not unique. The rail freight flow of farm products is not the only movement from Arkansas to California. Rather, many records for all commodities by modes between Arkansas and California can be found later in the table. The International database file represents freight flows of international origin or destination, by commodity and mode for 1998, 2010, and 2020. It adds the international region of origin or destination (Mexico, Canada, Europe, Latin Amer- ica, Asia, and Rest of World) to the database and indicates whether the freight is exported or imported. The actual state in which the freight enters or exits the United States is reported as the origin or destination. The International Air dataset contains freight by air only, from international origins or destinations. The beginning records in the file contain records on foreign air shipments that do not have a U.S. destination or origin. These are labeled foreign in the Direction column. Both the origin and destination states are designated “00” since the flow is purely international. In addition, records show domestic shipments classified as international flows. These state-to-state flows are labeled starting with the point of entry or exit into the United States shown respectively as the origin or destination for reporting purposes. The international origin or destination also is given. The STCC13-Petro file contains international pipeline flows in tons for 1998, 2010, and 2020. These flows are not usually part of statewide freight forecasting models. A summary of the contents of the FAF Commodity Flow Databases is shown in Table 5.1. The database can be obtained from the Federal Highway Administration at http:// ops.fhwa.dot.gov/freight/freight_analysis/faf. 5.2 Flow Conversion Flow data available or forecast using the methods in this Toolkit may require conversion into other units for process- ing or analysis. Commodity flow data, reported and forecast in terms of annual tons, is typically converted into vehicles and economic value. Vehicle conversion is generally done for commodity flow by trucks, since freight trucks are assigned together with automobiles and other trucks as daily trip tables. Tons to Vehicles The assignment model component for truck freight on highways, described in Section 4.0, is most often calculated in terms of daily truck trips. For the truck model class with fore- casts in those units, this is obviously a straightforward proce- dure. For commodity models that forecast flow in annual tons per year up to and through mode split, a conversion process is required. The Indiana case study uses the Carload Waybill sample to relate tons shipped per carload to develop factors to convert from annual tons to rail carloads and then applies a factor relating the volume of a rail car to the volume of a com- bination truck trailer to develop tons per truck trailer. More commonly, the Vehicle Inventory and Use Survey (VIUS)from the Economic Census is used to develop these factors. Vehicle Inventory and Usage Survey VIUS, conducted every five years as part of the U.S. Eco- nomic Census, provides detailed information on the physical and operational characteristics of the nation’s truck popula- tion. VIUS is based on a sample of approximately 150,000 trucks, or 2,000 trucks per state. From this sample, state and national estimates are produced. Operational characteristics, which are of particular interest to forecasters, include major use, products carried, annual and lifetime miles, area of 20

operation, miles per gallon, operator classification, and haz- ardous materials transported. The sample also includes expansion factors for each record. VIUS uses product classes similar to the commodity classes used in the CFS or TRANSEARCH/FAF. It records the percentage of the miles that a truck carries certain products, equipment, materials, etc. “No Load” is treated by VIUS as a separate product cate- gory. The VIUS survey also includes buses and service trucks. Certain VIUS product categories, such as passengers carried, do not correspond to the freight model commodity classes. A correspondence between the VIUS product classes and the more common commodity classes can be easily developed based on the definition of each classification scheme. The weighted annual mileage for each VIUS product car- ried distance class can be calculated for each record in a state database. That mileage can be multiplied by the average pay- load for that record to obtain the weighted annual pound- miles by product class. The weighted annual pound-miles and the weighted annual miles can be summed over all records by product class. The average payload for each com- modity can be obtained by dividing the average annual pound-miles by the average annual miles. This payload does not include the percentage of mile that a truck travels while empty. This percentage by commodity also can be calculated from the VIUS “No Load” product class. The factor to be used to covert from annual tonnage to annual trucks could 21 Database File Name Content Domestic Flows Domestic State-to-state flows By commodity By mode By year (1998, 2010, 2020) International Flows International International flows (by state origin or destination) By commodity By mode By year (1998, 2010, 2020) International Flows International Air International flows By air By year (1998, 2010, 2020) By state origin or destination Foreign shipments only Domestic shipments only International Flows STCC13-Petro (by state origin or destination) International flows of crude petro/natural gas By pipeline (other) By year (1998, 2010, 2020) Table 5.1. Contents of commodity flow datasets. Figure 5.3. Freight analysis framework domestic flows database in Microsoft Access.

account for both the average payload and the percentage of empty trucks in each commodity. The 1997 VIUS is available on CD-ROM from the U.S Census Bureau at http://www.census.gov/svsd/www/ 97vehinv.html. Tons to Value Converting tons per year to dollars shipped is useful in eco- nomic analysis or to account for forecasting methods that consider the value of the freight being shipped. These con- version factors can be obtained from the CFS. Commodity Flow Survey The 1997 CFS reports commodities by SCTG code and con- tains both value and tonnage data for each commodity by state. This information can be used to develop conversion tables of value per ton by SCTG commodity. Values by commodity by mode can be used to account for differences in the mix of com- modities at the SCTG two-digit level by mode. This is useful when, for example, high value commodities that only can be identified at the SCTG three- or four-digit level move prefer- entially by air and distort the overall average calculations of value at the two-digit level. Table 5.2 shows the values from the 22 Note: A symbol of 1 represents zero or less than one unit of measure. A symbol of 2 represents data that does not meet publication standards due to high sampling variability or other reasons. Area California SCTG Electronic and other electrical equipment and components and office equipment Characteristic Value ($ million) Tons (000) Value per Ton ($) Item Data Symbol Data Symbol Mode All modes $ 206,731 – 5,057 – $ 40,880 Single modes $ 132,620 – 4,274 – $ 31,029 Truck $ 109,862 – 4,050 – $ 27,126 For-hire truck $ 78,259 – 2,796 – $ 27,990 Private truck $ 29,664 – 1,129 – $ 26,275 Rail $ 414 – – 2 Water $ 53 – – 2 Shallow draft – 2 – 2 Great Lakes – 1 – 1 Deep draft $ 53 – – 2 Air (includes truck and air) $ 22,291 – 148 – $ 150,615 Pipeline – 2 – 2 Multiple modes $ 57,088 – 396 – $ 144,162 Parcel, U.S. Postal Service or courier $ 56,595 – 383 – $ 147,768 Truck and rail – 2 – 2 Truck and water – 2 – 2 Rail and water – 1 – 1 Other multiple modes – 1 – 1 Other and unknown modes $ 17,023 – 387 – $ 43,987 Table 5.2. Value per ton by commodity and mode for the state of California.

CFS that can be produced for electronics and electrical equip- ment, using the table for California. The average value per ton for all modes is about $41,000, based on an average of goods moving by land with values of approximately $27,000 per ton and goods of the same commodity moving by air with a value of approximately $151,000 per ton. Table 5.2 also shows that for many modes the data are not reported because the small sample size produces unreliable results. The 1997 CFS is available on CD-ROM from the U.S. Census Bureau at http://www.census.gov/econ/www/ cfsmain.html. 5.3 Network Data Modeling truck freight movements requires the use of net- works with physical information about the highway network links. The network used in assigning freight flows must account for characteristics such as segment capacity, volume, free flow speed, and travel time. Networks exist for other modes (rail, air, water), but typically do not include infor- mation to allow the calculation of congestion and route choice in the same fashion as truck/highway networks. Many freight shipments use more than one mode in a trip, and data on the intermodal terminals where freight can change modes also are required. Modal Networks National modal networks are needed in statewide freight forecasting, particularly for non-highway modes and for highway networks for areas of the United States beyond the area covered by a statewide model. The Bureau of Trans- portation Statistics provides attribute information for water- way and railroad networks, although this information is not compatible with conventional travel demand modeling soft- ware. The Oak Ridge National Laboratory has created a mul- timodal network to determine distances and routes for the CFS. However, Oak Ridge uses special software that other agencies may find difficult to use. A comprehensive source of network data can be created by matching the network of the National Highway Planning Network (NHPN) Geographic Information Systems (GIS) shapefile with the attribute data from the Highway Performance Monitoring System (HPMS) Data collected by each state. This task already was undertaken by the FAF and is described below. Freight Analysis Framework Highway Capacity Database The FAF road network leverages existing Federal road inventories that contain, or can be linked to, HPMS data. After analyzing data availability, the Federal Highway Administration (FHWA) developed the network as a subset of the NHPN, version 3. The FAF highway network not only includes FAF truck counts, but passenger automobile counts and non-FAF truck counts. Data was obtained from traffic databases in HPMS and other state sources. After integrating the data sources, the FAF network was converted to TransCAD, a proprietary travel demand model software package. TransCAD allows the assignment of daily freight truck trips to routes using stan- dard network assignment techniques. The end result was the completed FAF highway network database containing traffic volume, capacities, speeds, locations, and travel times for each road segment. The Highway Capacity Dataset contains estimated truck volumes and system capacities for each road segment on the FAF network, obtained through freight demand analy- sis. The 1998 freight volume data are included, as well as forecasts for 2010 and 2020. Volume is provided for FAF trucks, non-FAF trucks, and general traffic. The non-FAF trucks were calculated by subtracting model-assigned trucks from observed truck counts. Both automobiles and non-FAF trucks were treated as preloaded volumes that contribute to highway congestion in the FAF route assign- ment model. Additional attributes such as volume/capacity ratio, delay, and derived speed also are included. The data files are available in either TransCAD or ESRI GIS format, with all the querying and mapping capabilities of these two programs. The Federal Highway Administration also provides a data dictionary for use in understanding abbreviated column headings in the dataset. One layer of the High-Capacity Dataset contains informa- tion on the FAF highway network, linked to information on each road segment. Figure 5.4 shows the GIS representation of U.S. highways on the FAF network. Each road segment is described using up to 17 attributes. These attributes include length in miles, state and county identifiers, signs, road name, function class, status, Na- tional Highway System (NHS) designation, and rural code. In a separate file, the Highway Capacity Dataset contains freight flow data that can be overlaid on the FAF highway network maps. Using GIS software, a user can then identify segments with specified levels of congestion, delay, or ca- pacity. Besides road characteristics, the freight volume data contained in this file includes annual average daily traffic, FAF/non-FAF trucks, speed, delay, flow, and capacity for both low and high growth estimates in 1998, 2010, and 2020. The FAF Highway Capacity Database is available from the FHWA at http://ops.fhwa.dot.gov/freight/freight_analysis/ faf. 23

Intermodal Terminals Intermodal terminals are facilities for transferring freight from one mode, such as truck, to another mode, such as rail. Knowing the location of these terminals is important when assigning a complete freight shipment from an initial origin to its ultimate destination. It also is important in forecasting the behavior of freight since freight is neither produced nor consumed at these terminals but merely transshipped. The Bureau of Transportation Statistics provides data on the location and attributes of these intermodal terminals, includ- ing the type of commodity handled. While intermodal freight is often considered freight moving in sealed containers, the intermodal terminals include all facilities where freight – including bulk shipments – changes modes. Bureau of Transportation Statistics Intermodal Terminals The Bureau of Transportation Statistics Intermodal Termi- nal Facilities data set contains geographic data for freight trans- fer facilities in the United States. Attribute data includes the modes serving the facility, the name of the railroad (if any) serving the facility, the type of cargo, and the direction of the transfer. The database provides location and attribute infor- mation for use in national and regional network analysis applications. Attribute data are extracted from a variety of rail- road and port carriers operators and associations. Data reflects conditions at facilities in 1995-1996 and is subject to frequent change. Some facilities may be dormant or permanently closed. The intermodal terminal database is available from the Bureau of Transportation Statistics Mapping Center at http://www.transtats.bts.gov/mappingcenter.asp. National Transportation Atlas Database (NTAD) The National Transportation Atlas Database (NTAD) is a collection of GIS data layers in 1:1,000,000 scale developed by the U.S. Department of Transportation and other Federal agencies. The NTAD is available from the Bureau of Trans- portation Statistics Mapping Center at http://www.transtats. bts.gov/mappingcenter.asp. 5.4 Forecasting Data Population Population data used in freight forecasting is typically used in traditional transportation forecasting. This includes both a base and a forecast horizon year or years for a variety of TAZ. For areas outside of the state study area, population data can be obtained from the U.S. Census Bureau, which typically is the basis for base year passenger transportation forecasting. Forecasts of national population only are avail- able through commercial vendors. Employment While employment data are typically used in passenger transportation forecasting, the level of industry detail is insufficient for freight forecasting. Industry information is developed from mandatory quarterly ES-220 submittals by employers to state employment security agencies and used by the U.S. Bureau of Labor to compute unemployment statis- tics. However processed, the data released to the public is aggregated to suppress confidential information. That data, available from the Census Bureau’s County Business Pat- terns, is described below. More geographic detail is available 24 Figure 5.4. Freight Analysis Framework highway network.

from private vendors such as Dun & Bradstreet, InfoUSA, Wood & Poole, and IMPLAN. These vendors also provide employment data in the more commonly used SIC system and provide forecasts not available from public agencies. County Business Patterns County Business Patterns is an annual series published by the U.S. Census Bureau that provides subnational economic data by industry. The series is useful for studying the eco- nomic activity of small areas, analyzing economic changes over time, and providing a benchmark for statistical series, surveys, and databases between economic censuses. Busi- nesses use the data for analyzing market potential, measuring the effectiveness of sales and advertising programs, setting sales quotas, and developing budgets. Government agencies use the data for administration and planning. County Business Patterns covers most economic activity in the United States. The series excludes data on self-employed individuals, employees of private households, railroad em- ployees, agricultural production employees, and most gov- ernment employees. Beginning in 1998, data was tabulated by industry as defined in the NAICS. Data for 1997 and earlier years is based on SIC codes, described in Section 5.6. As shown in Figure 5.5, the County Business Patterns data are available for all counties by three-digit NAICS code. Typ- ically freight forecasting models need only two-digit data. Additionally where industrial employment is not available for a given county, it may be possible to estimate that data from the establishments in the employment ranges. For example, as shown in Figure 5.5, the employment data for industry 113 in County 12001 is suppressed but might be estimated at 2.5 employees from the one firm with one to five employees, seven employees from the one firm with five to nine employ- ees, 14.5 employees from each of the two firms with 10 to 19 employees. This estimated total of 38.5 employees (2.5 + 7 + 14.5 + 14.5) agrees closely with the reported total of 39 em- ployees. Since the County Business Patterns data are typically applied to already available TAZ data at a more aggregated scale, the resulting percentages by detailed industry are generally suitable for forecasting. County Business Patterns data are available from the U.S Census Bureau at http://www.census.gov/epcd/cbp/view/ cbpview.html. 5.5 Validation Data In most cases validation data for freight forecasting is lim- ited to observed trucks, which include both freight and non- freight purposes. Thus, truck classification counts and weight and motion counts prepared by states can be used in validat- ing the truck portion of freight models only in combination with a multi-class assignment of all vehicles. Tolled facilities with electronic data collection mechanisms also can provide a way to validate freight forecasts, since trucks could theoretically be tracked on an individual basis and extensive data about truck movements (including entry-exit points) might be available. This type of data has been used to validate truck models on a study-by-study basis but is rarely used to validate a statewide model because tolled highways 25 Figure 5.5. County business patterns employment data.

constitute a small percentage of the statewide highway net- work. Some components of freight models are typically not vali- dated since the only data available was used to develop the model and no independent data are available for validation. For some industries it may be possible to find alternate information on production, and sometimes consumption for a particular market. Specific state agencies, such as agricul- ture, mining, or forestry departments, may maintain annual production information, particularly for sectors of the econ- omy with significant levels of activity within the State. Industry- specific trade associations also may compile this type of information. However, in many instances these agencies will disseminate statistical information that comes from Federal government sources, or the other primary sources commonly used in freight forecasting and modeling. 5.6 Classification Schemes The data sources used in freight forecasting report on ship- ments by commodity and their associated industries using different classification schemes. Understanding these schemes is necessary to properly utilize the data. Additionally, some models may require the use of multiple data sources based on different classification schemes. Understanding the relation- ships between the alternate coding systems is essential to prop- erly integrate the information. Commodity Classification There are two primary commodity classification schemes. Prior to the 1997 CFS , U.S. freight data was collected and reported using the STCC code. This classification code was developed in the early 1960s by the American Association of Railroads to analyze commodity movements by rail. It also is the reporting system that continues to be used in the STB’s Carload Waybill Sample. The United States and Canada have adopted the SCTG and this system was used in reporting the 1997 and subsequent CFSs. SCTG is similar to the Harmo- nized Schedule classification, which is the predominant prod- uct coding system currently in use worldwide. However much of the available commercial freight economic data are available using the older STCC codes. STCC codes are numerical codes that group similar prod- ucts, and the codes are arranged in a very structured, hierar- chical manner. The first digit identifies a major Economic Division, such as 2-Nondurable manufacturing. The second digit identifies an Economic Major Group, such as 20-Food and kindred spirits. The third digit identifies an Industry Group, such as 202-Dairy products. The fourth digit identi- fies a Specific Industry, such as 2024 Ice Cream and Frozen Desserts. Additional detail is provided through up to seven digits, although this level of detail is primarily used by the car- riers in setting rates and little volume information is available at this level. The STCC system also is compatible with the SIC industrial classification system discussed below, while the SCTG is not completely consistent with the SIC or the NAICS. A correspondence table between the STCC and SCTG sys- tems, at the two-digit level, is shown in Appendix A. At this hierarchical level there is considerable overlap between the two systems. Industry Classification Industry classifications are similar to commodity classifi- cations. Historically, the United States has reported economic industry data using the SIC. SIC has being replaced with the new NAICS. SIC codes are numerical codes that group companies that produce similar products or services. SIC codes are arranged in a very structured, hierarchical manner. The first digit iden- tifies a major Economic Division, such as 2-Nondurable manufacturing. The second digit identifies an Economic Major Group, such as 20-Food and kindred spirits. The third digit identifies an Industry Group, such as 202-Dairy prod- ucts. The fourth digit identifies a Specific Industry, such as 2024 Ice Cream and Frozen Desserts. The United States has started collecting data using the NAICS codes. While similar in approach, the NAICS codes cover a much wider variety of industries, technologies, prod- ucts and services, particularly new, emerging, and advanced technology industries. NAICS reorganizes industries into cat- egories that reflect the service-oriented economy. The NAICS codes are detailed up to six digits. The correspondence between the SIC and NAICS codes exists only at the four-digit SIC and six-digit NAICS level. The NAICS is a joint effort of the United States, Canada, and Mexico and will make it eas- ier to compare U.S. industrial statistics with economic data from other countries. The SIC hierarchical structure matched that of the STCC commodity classification system, making it easy to compare between industrial activity and commodity shipment. Since the NAICS was only introduced in 1997, historical data in that system is lacking. While U.S. government information is only available in the NAICS, many private users of industry and employment data, particularly private data providers still use the SIC system. A correspondence table between the SIC and NAISC systems, at primarily the two-digit level, is shown in Appendix A. The exception is in NAICS categories 31-33 Manufacturing, 44-45 Retail Trade, and 48-49 Transporta- tion and Warehousing which are distinguished only at the three-digit level. At this hierarchical level there is consider- able overlap between the two systems. 26