Improved Estimates of Ton-Miles
Improved Estimates of Ton-Miles
SCOTT M. DENNIS *
This paper describes recent improvements in measuring ton-miles for the air, truck, rail, water, and pipeline modes. Each modal estimate contains a discussion of the data sources used and methodology employed, presents a comparison with well-known existing estimates for reference purposes, and discusses the limitations of the data. The resulting estimates provide more comprehensive coverage of transportation activity than do existing estimates, especially with respect to trucking and natural gas pipelines.
KEYWORDS: Transportation measurement, ton-miles.
The Bureau of Transportation Statistics (BTS) is improving some of its basic estimates of transportation activity. This paper describes proposed ton-mile estimates for the air, truck, rail, water, and pipeline modes. Each modal estimate contains a discussion of the data sources used and methodology employed, presents a comparison with well-known existing estimates for reference purposes, and discusses the limitations of the data. This paper should be viewed as part of a continuing series of steps forward. Additional planned work will allow BTS to further improve its basic estimates of transportation activity.
Ton-miles is the primary physical measure of freight transportation output. A ton-mile is defined as one ton of freight shipped one mile, and therefore reflects both the volume shipped (tons) and the distance shipped (miles). Ton-miles provides the best single measure of the overall demand for freight transportation services, which in turn reflects the overall level of industrial activity in the economy. In addition, a ton-mile estimate is necessary in order to construct other estimates of transportation system performance, such as energy efficiency and accident, injury, and fatality rates.
Domestic ton-mile estimates are usually developed by aggregating data for individual freight transportation modes. Data for air freight, railroad, and water transportation are readily available as a result of government provision of infrastructure or residual economic regulation. Comprehensive pipeline data are difficult to obtain, because a significant percentage of pipeline traffic is "in-house" transportation for companies that produce and refine oil. Ton-mile data for the trucking sector are even more problematic due to the large number of shippers, receivers, and trucking firms, as well as the substantial percentage of in-house trucking traffic. All data sources suffer, at least to some degree, from gaps in the desired scope of coverage.
The Eno Transportation Foundation has published historical ton-mile estimates for many years (Eno 2002, p. 42), but no longer does so. In more recent years, BTS has provided alternative estimates in National Transportation Statistics (NTS) (USDOT BTS 2003). But due to the problems described above, these well-known sources do not appear to provide complete, reliable estimates of this basic transportation measure. BTS, therefore, undertook a research program to address these shortcomings. This paper presents the results of this research.
BTS developed its improved estimates of domestic ton-miles (traffic within and between the 50 states, the District of Columbia, Puerto Rico, and the U.S. Virgin Islands) to maintain compatibility with other U.S. Department of Transportation Strategic Plan data. These annual ton-mile estimates illustrate long-term trends. Comprehensive coverage is achieved by combining reported data from established sources, estimates from surveys, and calculations based on certain assumptions. Table 1 briefly compares the scope of the improved BTS ton-mile estimates with the NTS and Eno estimates, while figure 1 presents all three estimates for all modes (also see appendix table A1). The NTS and Eno estimates are not available for the most recent years.
Figure 2 shows air freight ton-mile data from the three datasets (also see appendix table A2). The improved BTS data are compiled in Air Carrier Traffic Statistics Monthly (USDOT BTS 1990-2003), which presents the results of the T-100 reporting system, supplemented by special tabulations of data on domestic all-cargo operators from the Federal Aviation Administration (FAA).
The T-100 data represent the population of all domestic freight traffic for Section 401 air carriers, which operate planes with a passenger seating capacity of more than 60 seats or a maximum payload capacity of more than 18,000 pounds. These data include the vast majority of all domestic air freight traffic. As a result of a BTS rulemaking, data for smaller carriers have been included in this source starting with the fourth quarter of 2003. The inclusion of smaller carriers does not substantially affect the value of the data series. Domestic all-cargo operators (Section 418 carriers) have been gradually integrated into Air Carrier Traffic Statistics Monthly. The FAA data captured those carriers who had not yet reported in Air Carrier Traffic Statistics Monthly, thus allowing representation of the full population of domestic all-cargo operators.
BTS's proposed estimates of air freight ton-miles are essentially the same as the Eno estimates. Neither estimate includes private carriage of air freight or air freight forwarders who do not use T-100 reporting carriers. These exceptions account for well under 5% of all air freight traffic. The substantial difference between the two data series in 2001 is due apparently to Eno's use of preliminary data.
Oak Ridge National Laboratory produced estimates of truck ton-miles based on the 1993 and 1997 Commodity Flow Surveys (CFS) (USDOT and USDOC 1993 and 1997), supplemented with data on farm-based shipments and imports arriving by truck from Canada and Mexico. Transportation Statistics Annual Report provides this 1997 estimate of truck ton-miles (USDOT BTS 2000, p. 124). To produce the improved BTS estimate, the 1993 and 1997 estimates were updated and backdated using intercity and intracity vehicle miles-traveled (VMT) for single-unit and combination trucks, as reported in Highway Statistics (USDOT FHWA 1990-2003). Figure 3 presents the resulting estimates (also see appendix table A3). The trend in both series is the same, because the same VMT data were used to update each series. After making these adjustments for different time periods and population coverage, the difference between the 1993 and 1997 estimates is less than 2%.1
The CFS captures export movements, as well as movements of imports once they reach their first domestic destination, such as a warehouse. In order to provide a more complete estimate of truck traffic, the data in figure 3 were further adjusted to reflect truck ton-miles from maritime movements prior to reaching their first domestic destination. The number of loaded 20-foot equivalent unit (TEU) containers shipped through U.S. ports is reported in U.S. Waterborne Container Traffic by Port (U.S. Army Corp of Engineers WCSC 2003). These figures were then divided by 2.4 to convert them to an equivalent number of 48-foot trucks. Estimates of the percentage of import traffic, the truck share of import traffic, miles to the first domestic destination, and tons per truck for East, Gulf, and West Coast ports were obtained through interviews with port personnel in New York, Houston, and Los Angeles, respectively. The resulting estimates added between 7 billion and 12 billion truck ton-miles each year. This represents approximately 1% of all truck ton-miles currently estimated.
Figure 4 shows trucking ton-mile estimates (also see appendix table A4). The improved BTS estimates are based on the Oak Ridge National Laboratory supplement to the 1997 study, which is the more recent of the two studies. The improved BTS estimate is about 10% higher than the NTS and Eno estimates, each of which reflects only intercity truck traffic. Therefore, the improved BTS estimate provides a more comprehensive estimate of truck traffic.
The CFS data used to construct the improved trucking ton-miles estimate exclude shipments by households and retail, service, utility, and government establishments (including the U.S. Postal Service); and certain noncommercial freight shipments, such as construction traffic and municipal solid waste. The existing NTS and Eno estimates do not include intracity traffic. Therefore, it appears that a significant percentage of truck VMT and a somewhat smaller percentage of truck ton-miles are not included in any of these estimates. Clearly more work is needed in this area.
BTS developed its improved railroad ton-miles estimates using data from the Carload Waybill Sample (USDOT STB 1990-2003). The population estimate in this source is based on a 500,000 record sample of all traffic terminating on all railroads in the United States. The sample implicitly includes traffic originating on U.S. railroads and terminating on Mexican railroads, because almost all such traffic is rebilled to U.S. border crossings.2
Population data on the tonnage of railroad shipments originating in the United States and terminating in Canada come from Transportation in Canada (Transport Canada 1990-2002) for years prior to 2003. The average length of haul for U.S. railroad shipments was applied to this tonnage to obtain an estimate of U.S. railroad ton-miles for shipments terminating in Canada. This assumption seems reasonable, because even though much of this traffic originates in states bordering Canada, more distant states such as California, Texas, and Georgia are also among the 10 largest originating states. The Carload Waybill Sample's improved coverage of Canadian terminations of rail shipments originating in the United States allows estimates of all railroad ton-miles for 2003 and subsequent years.
Figure 5 illustrates the railroad ton-mile estimates (also see appendix table A5). From 1998 to 2001 (the most recent years for which all three estimates are available), the improved BTS estimates are about 5% greater than the NTS estimates, which include only Class I railroads; about 1% greater than the Waybill estimates, which do not include Canadian terminations; and almost identical to the Eno estimates, which include both non-Class I railroads and Canadian terminations. The increase in the improved BTS estimates relative to the other estimates is probably due to better coverage of the rapidly growing railroad shipments originating in the United States and terminating in Canada. Further, while Eno's ton-mile estimates for non-Class I railroads are based on financial survey data, the improved BTS estimates are based on actual ton-mile data and should be considered more reliable.
Finally, railroad ton-mile data may not include shipments originating in Mexico and terminating in Canada. Based on data from Transport Canada, it appears that these shipments account for less than one tenth of one percent of all U.S. railroad traffic.
Domestic waterborne ton-mile estimates are presented in figure 6 (also see appendix table A6). The current NTS estimates of annual water transportation ton-miles were taken from Waterborne Commerce of the United States (U.S. Army Corps of Engineers 2003). Data in this source are developed from lock data and individual trip reports that must be filed with the U.S. Coast Guard. Therefore, this source represents the entire population of all domestic water traffic, including inland waterways, coastwise, Great Lakes, and intraport traffic, along with traffic to and from Alaska, Hawaii, and Puerto Rico. The NTS and Eno estimates differ substantially, because NTS includes coastwise (domestic ocean) traffic and Eno does not. Thus, the current NTS data, which are proposed for use here, are more comprehensive than Eno's estimates.
Figure 7 shows pipeline ton-miles (also see appendix tables A7(a) and A7(b)). Annual oil and oil products pipeline ton-miles were obtained from Shifts in Petroleum Transportation (Association of Oil Pipelines 2003). These data represent the entire population of crude petroleum and petroleum products carried in domestic transportation by both federally regulated and nonfederally regulated pipelines. Both NTS and Eno currently use these data, which we also propose for use here.
Natural gas pipeline ton-miles are also presented in figure 7. These new estimates are based on natural gas deliveries reported in the Annual Energy Review (USDOE 2003a). BTS first converted the gas deliveries, measured in cubic feet, to metric tons and then to tons using a standard conversion factor of 48,700 cubic feet per metric ton as reported in the International Energy Annual (USDOE 2001).
There are no data available on the length of haul for natural gas shipments, because natural gas is drawn from a common pipeline rather than shipped to a specific consignee. Origination and termination data indicate that natural gas has a distribution pattern similar to oil and oil products (USDOE 2003b and 2003c). The oil and natural gas pipeline networks are also very similar. Therefore, the length of haul for oil and oil products was applied to the tonnage of natural gas to estimate natural gas ton-miles in transmission lines.3
Natural gas ton-miles in distribution lines (i.e., local utilities) were estimated using 5% of transmission length of haul, which is approximately half the diameter of a major metropolitan area. Natural gas ton-miles in gathering lines (i.e., from well to processing plant) were estimated using the same length of haul as in distribution lines. The ton-miles for gathering, transmission, and distribution lines were then summed to provide an estimate of total natural gas ton-miles. Natural gas ton-miles, which have not to our knowledge been previously estimated, represent nearly as much traffic as that carried on the inland waterway system. These new estimates fill a substantial gap in the existing ton-mile data.
The natural gas pipeline data do not include gas used to repressurize gas fields or power the pipeline itself, because these uses do not represent gas carried in revenue transportation. The pipeline data also exclude coal slurry, ammonia, and other types of pipelines. There are only a few such pipelines, which tend to have either short haul or low volume, and appear to account for well under 1% of all pipeline ton-miles. BTS will investigate the recent decline in the oil pipeline ton-mile data and the resulting reduction in the estimate of natural gas ton-miles in order to improve the most recent estimates.
The improved ton-mile estimates for the air, truck, rail, water, and pipeline modes described in this paper are both more comprehensive and more reliable than well-known existing estimates. The improvements are most noticeable with respect to trucking and natural gas pipelines. Additional work will allow BTS to further improve these basic estimates of transportation activity.
BTS has already incorporated these improved estimates of domestic freight ton-miles into the Transportation Statistics Annual Report (USDOT BTS 2004, p. 213).4 BTS plans to extend the improved estimates back to 1980 in the fall 2005 update to National Transportation Statistics.5 Future research will be conducted to further extend the improved estimates back to 1960 where data are available.
Association of Oil Pipelines. 2003. Shifts in Petroleum Transportation. Washington, DC. Table 1.
Eno Transportation Foundation. 2002. Transportation in America. Washington, DC.
Transport Canada. 1990-2002. Transportation in Canada. Ottawa, Ontario, Canada. Addendum, table A6-10.
U.S. Army Corps of Engineers. 2003. Waterborne Commerce of the United States. Washington, DC. Part V, Section 1, Table 1-4, Total Waterborne Commerce.
U.S. Army Corps of Engineers, Waterborne Commerce Statistics Center (WCSC). 2003. U.S. Waterborne Container Traffic by Port. Washington, DC.
U.S. Department of Energy (USDOE), Energy Information Administration. 2001. International Energy Annual. Washington, DC. Table C-1.
_____. 2003a. Annual Energy Review. Washington, DC. Table 6.5.
_____. 2003b. Natural Gas Annual. Washington, DC. Table 12.
_____. 2003c. Petroleum Supply Annual, Volume 1. Washington, DC. Table 33.
U.S. Department of Transportation (USDOT), Bureau of Transportation Statistics (BTS). 2000. Transportation Statistics Annual Report. Washington, DC.
_____. 2003. National Transportation Statistics. Washington, DC. Table 1-44.
_____. 2004. Transportation Statistics Annual Report. Washington, DC.
U.S. Department of Transportation (USDOT), Bureau of Transportation Statistics (BTS), Office of Airline Information. 1990-2003. Air Carrier Traffic Statistics Monthly. Washington, DC. Freight, Express, and Mail Revenue Ton-Miles table, p. 2, line 3.
U.S. Department of Transportation (USDOT), Bureau of Transportation Statistics and U.S. Department of Commerce (USDOC), Economics and Statistics Administration, U.S. Census Bureau. 1993. Commodity Flow Survey. Washington, DC.
______. 1997. Commodity Flow Survey. Washington, DC.
U.S. Department of Transportation (USDOT), Federal Highway Administration (FHWA). 1990-2003. Highway Statistics. Washington, DC. Table VM-1.
U.S. Department of Transportation (USDOT), Surface Transportation Board (STB). 1990-2003. Carload Waybill Sample. Washington, DC.
1. The VMT data include a substantial amount of truck traffic that is outside the scope of the CFS, e.g., shipments by households and retail, service, utility, and government establishments (including the U.S. Postal Service); and certain noncommercial freight shipments, e.g., construction traffic and municipal solid waste. The VMT data can, therefore, be taken to provide a reasonable estimate of the trend in truck ton-miles, but not the level, and should not be used to make inferences about operational parameters such as empty mileage or average load per truck.
2. Traffic originating on U.S. railroads and terminating on Mexican railroads is treated for accounting purposes as if it terminated at the U.S. border crossing, and is therefore included in the Carload Waybill Sample. This practice is known as rebilling.
5. Available on the BTS website at http://www.bts.gov/.
ADDRESS FOR CORRESPONDENCE
* S. Dennis, Bureau of Transportation Statistics, Research and Innovative Technology Administration, U.S. Department of Transportation, 400 7th St. SW, Room 3430, Washington, DC 20590. E-mail: email@example.com