4.0 FREIGHT MOVEMENT BY RAIL

4.1 Introduction

The most efficient method for updating the railroad freight flows depicted within the FAF² would employ the confidential version of the Surface Transportation Board's annual Carload Waybill Sample (CWS). However, the scope of the current project explicitly states that data sources used in the updates must be publicly available. The public use CWS is of limited use for flow analysis given that origin and/or destination information is removed in many records to preserve confidentiality. Accordingly, an alternative method was developed that makes heavy use of the public use CWS, but which also relies on the existing FAF² flows, annual rail traffic data from the Association of American Railroads (AAR), the Surface Transborder Freight Dataset (STFD), and data available from the US Census Bureau.

The railroad freight flows depicted within the FAF² have two dimensions – commodity and geographic. It is essential to reflect variations across both dimensions. The proposed process begins by using the AAR annual carloading data to obtain traffic growth rates for the 19 AAR commodities. Next, the 19 AAR commodity groups were mapped into the 43 FAF² commodity groups, so that each FAF² commodity flow will have an associated overall commodity-based growth rate. These rates may be used to develop control flow totals for each commodity.

Next, the growth rates for each FAF² flow element were modified to reflect any volume variations that are attributable to origin and/or destination location. Adjustments may be based on any combination of three factors – demographic variations, variations in industry-specific economic activity, and variations in general economic conditions.⁹

The final methodological step combines the origin and destination specific growth rates with existing FAF² flow data in an application of a traffic growth factor model such as that developed by Fratar.¹⁰ These models have a number of attributes and limitations. Most notably, however, they do not require the incorporation of impedances, and so are relatively simple to implement.¹¹

The methodology for estimating international flows employed the annual rail records of the STFD, which are complete for rail movements. These shipments were disaggregated to the FAF² region level. For the purpose of the FAF² update, railroad carload and rail/truck intermodal shipments were treated as separate modes, even though many calculations were integrated across the two.

There are a number of challenges inherent in the processes described above. These include, but are likely not limited to:

• Difficulty in mapping the 19 AAR commodity groups into the 43 FAF² commodity groups.
• The AAR data are presented for each Class I carrier. Thus, it is possible for a particular shipment to be duplicated if it is interchanged between two carriers.
• There is no immediately available source for establishing changes in commodity values for domestic shipments.
• AAR data are expressed in carloads; FAF² flows are expressed in tons. Thus, changes in car loading weights could distort the AAR-based growth rates.
• What about the time lagging?

The proposed response to each of these challenges is provided in the descriptions of the methodologies presented in the following sections.

4.2 Principal Data Sources

Weekly Railroad Traffic – This weekly publication contains information on carload and intermodal traffic for the U.S. Class I railroads, the two large Canadian railroads, a major Mexican railroad, and selected U.S. non-Class-I railroads. It includes carload information for 19 commodity groups and intermodal traffic, which is reported for trailers and containers. It can be accessed at:
http://www.aar.org/catalogandpublications/PublicationDetails.asp?ContentType_ID=22

Carload Waybill Sample (CWS) – A stratified sample of carload waybills for terminated shipments by railroad carriers. This waybill data is used to create a movement specific Confidential Waybill File and a less detailed Public Use Waybill File. This is published by the Surface Transportation Board (STB) and can be accessed at:
http://www.stb.dot.gov/stb/industry/econ_waybill.html

Surface Transborder Freight Database – Published by the Bureau of Transportation Statistics (BTS) and contains data on North American merchandise trade by commodity, surface mode (rail, truck, pipeline, mail, and other), and by port of entry and geographic detail for the U.S. trade to and from Canada and Mexico. This source provides the dollar value of both imports and exports, and tonnage of imports. The data are published with a three-month time lag. It can be accessed at:
http://www.bts.gov/programs/international/transborder/

County Business Pattern Database – Published by the U.S. Census Bureau on an annual basis and provides national, state, and county level data on payroll, employment, and number of establishments by detailed NAICS industries. The series provides subnational economic data by industry and excludes data on self-employed individuals, employees of private households, railroad employees, agricultural production employees, and most government employees. The report is released with a two-year time lag. It can be accessed at:
http://www.census.gov/epcd/cbp/view/cbpview.html

Producer Price Index – Measures the average change over time in the prices received by domestic producers of goods and services. This measures price changes from the point of view of the producer. The data are reported by detailed industry and detailed type of commodities. The Bureau of Labor Statistics (BLS) publishes these data on a monthly basis with a time lag of one month. It can be accessed at:
http://www.bls.gov/ppi/

4.3 Domestic Rail Flows

The primary data source for the FAF² domestic rail flow update was the Surface Transportation Board (STB) Public Use version of the Carload Waybill Sample (CWS). The waybill sample is a stratified sample of the population of rail movements that originate or terminate within the United States. The actual degree of sampling depends on the variability of shipment characteristics across commodities. In most cases, the sample represents between one and two percent of the overall shipment population.

The initial step involved gleaning tonnage variations between 2002 and 2005. CWS commodity definitions (based on Standard Transportation Commodity Codes) were bridged to the Standard Transportation Commodity Group (STCG) definitions employed within the FAF². Once this was accomplished, national traffic growth factors were calculated based on variations in tonnage between 2002 and 2005.

Unfortunately, the O-D information in the Public Use CWS is left incomplete in order to protect the confidentiality of both shippers and rail carriers. Consequently, it was impossible to use the CWS to identify geographic variations in traffic volumes that might be relevant to intertemporal variations in the FAF² flows. To remedy this problem, industry-specific employment values, derived from County Business Pattern data were used to build indexes reflecting FAF region employment, population, and income trends for both origin and destination regions. The hypothesized relationship between employment, demographic values and rail flows are summarized in Table 4.1. The formal derivation of the indexes is explained in the following equation. Once established, the indexes were applied to the national growth factors to simulate geographic variations in flows. Finally, the adjusted growth factors were applied to the 2002 FAF² values to yield 2005 FAF² estimates.¹²

.

where OVAL and TVAL represent the appropriate origin and destination employment or population variables.

Revising the FAF² data also required estimating changes in the value of commodity flows. Changes in values are a function of both changed flow volumes and per-unit commodity value variations. Flow changes were based on tonnage volume flows as described above. Per-unit values were based on commodity-specific variations where possible, as captured by changes in the components of the Producer Price Index (PPI). The bridge between FAF commodity definitions and PPI values is provided in Table 4.2.

4.3.1 Rail Flows to Deep-Draft Ports

Somewhat inexplicably, the FAF² rail flows over U.S. deep-draft ports entirely neglect import flows and capture export flows only. Consequently, export flows alone were updated from 2002 values to reflect changed economic conditions in 2005. The basis for this update was export data obtained from the US Department of Commerce. As in the case of domestic rail flows, the initial step involved reconciling export data commodity definitions with the commodity definitions used within the FAF².

Next, because the Department of Commerce export volumes are expressed in dollar values only (as opposed to tons) it was necessary to account for intertemporal changes in commodity values between 2002 and 2005. As in the case of domestic flows, the PPI was used for this purpose. Once price variations were accounted for, export data were used to scale 2002 FAF flows to reflect 2005 FAF² deep-draft port flows. Given that no geographic variation is reflected in the Department of Commerce data, the revised FAF² values assume that the distribution of rail export flows across US ports is unchanged between 2002 and 2005.¹³

The final task in the adjustment to the FAF data involved again using price index data in order to inflate the value field in the 2002 data. Where possible, industry or product specific values were used. In the absence of such data, the value corresponding to “Industrial Commodities Less Fuels” was used.

4.4 Methodology for International Freight

4.4.1 Methodology for International Freight

FAF contains international rail freight shipments of two types: (1) all-rail shipments to/from Canada and Mexico and (2) shipments to/from countries outside of North America that use rail for the domestic portion of the movement. Different methodologies are used for addressing the two categories.

Transborder Rail Freight to and from Canada and Mexico by U.S. State and Port of Entry or Exit
The approach for estimating rail freight flows between the U.S. and Canada, and between the U.S. and Mexico, is as follows:

1. Determine state-level transborder rail freight to and from Canada and Mexico for the current year using information from BTS' Transborder Freight Dataset (TFD);
2. Disaggregate state level transborder rail freight flow to FAF region-level based on FAF patterns from the base year; and
3. Allocate FAF region level flows to and from Canada/Mexico to ports of entry/exit (actually border crossing points) based upon FAF² patterns from the base year or data on port use from the current year TFD.

4.4.2 Determine State-Level Transborder Rail Freight to and from Canada and Mexico

BTS's Transborder Freight Dataset provides freight data on tons and value of exports and imports from Canada and Mexico to the United States by rail. The data are reported by origin/destination state, country, and type of commodity. A separate group of files provide data on total weight and value through ports of entry/exit by O-D state and county, without regard to specific commodities. The rail records are extracted from all of the annual TFD datasets for the target year.

The TFD uses the Harmonized Schedule (HS) commodity classification, rather than the SCTG employed by FAF. Using a translation table matching HS and SCTG codes, the TFD records are processed to add the appropriate SCTG. The port of entry/port of departure (POE/POD) in the TFD is described using Customs Port Codes. A translate table maps these codes to FAF regions and international gateways.

The following sections describe processing steps to handle import and export data.

4.4.2.1 Imports

The following TFD files contain import data at the commodity level:

• 09yyyy–imports from Mexico with state of destination and 2-digit commodity detail, where yyyy is the year of release, e.g. 2005; and
• 10yyyy–imports from Canada with state of destination and 2-digit commodity detail.

The import records are processed to tally the total weight and value by HS commodity. Separate totals are kept for imports from Mexico and Canada. This information is used in processing the export records as described in the next section.

For compatibility with FAF, the weights and values in each record are respectively converted from kilograms to kilotons and dollars to millions of dollars. The origin field is set to the appropriate FAF code for Mexico or Canada.

4.4.2.2 Exports

Export data at the commodity level is contained in the following TFD files:

• 3ayyyy–exports to Mexico with state of origin and 2-digit commodity detail; and
• 4ayyyy–exports to Canada with state of origin and 2-digit commodity detail.

Unlike the import records, the export records lack weight. Accordingly, the weight for each flow must be imputed. To estimate weight, value is multiplied by a weight/value ratio for the commodity with different ratios used for Canada and Mexico. These ratios are derived using the commodity tallies collected from the import records. To minimize variance, the tally is based at the HS level, a lower level of aggregation than the SCTG.

As with the import records, the weights and values in each record are respectively converted from kilograms to kilotons and dollars to millions of dollars. The destination field is set to the appropriate FAF code for Mexico or Canada.

4.4.2.3 Initial Aggregation

Following initial processing, the import records for Mexico and Canada are combined into a single file representing all import rail traffic for the target year. The separate export record files are also combined. As a result of the conversion from HS to SCTG, each combined file may contain several records for a given origin, destination, and SCTG. The files are processed to combine these records into a single record containing the totals for the origin, destination, and SCTG.

4.4.3 Disaggregate Estimates of Transborder Rail Freight by State to FAF-Level Estimates

The state-level transborder rail freight tonnage and value are disaggregated to FAF-level estimates using the existing patterns from the original FAF base year.

4.4.3.1 Imports

The target year estimate of rail freight import tonnage W_i,c,r,t of commodity i from country c to FAF region r in state s for year t, is calculated as:

W_i,c,r,t = W_i,c,s,t*P_i,c,s,r,t-1

where:
W = Rail freight tonnage of imports,
P = share variable,
i = commodity,
c = country of origin (Canada or Mexico),
s = destination state,
r = destination FAF region (in state s),
t = target year, and
t-1 = base year.

The share variable P is based upon the weight of i destined from c to r in the base year as a portion of the total weight destined from c to s. Domestic FAF regions lie entirely within state boundaries; a crosswalk table allows state totals to be derived from FAF totals. The share variable is formally calculated as:

The value of the import tonnage for the commodity is calculated in a similar fashion using the same share variable, P:

V_i,c,r,t = V_i,c,s,t*P_i,c,s,r,t-1

where V is the value and all subscripts have the same meaning as previously.

Where commodity flows did not exist in the base year, the flow is allocated in equal portions to each FAF region in the state. Future refinements may use county level indicators from the Census County Business Patterns Database (CBP) to disaggregate the flows. However, more study is needed to establish the appropriate metrics to use for this purpose. The correlated variable for imports is likely different from exports. Also, 2004 is currently the latest year for which CBP data are available.

4.4.3.2 Exports

Disaggregation of exports from the state to the FAF region level follows a similar approach to that of imports. The target year estimate of rail freight export tonnage of commodity i from FAF region r in state s to country c for year t, is calculated as:

W_i,r,c,t = W_i,s,c,t*P_i,s,r,c,t-1

where all variables and subscripts remain as previously defined. The share portion for r is calculated as follows:

This share is also used to apportion value for the flow.

4.4.4 Allocate FAF-Level Flows to Ports

The final processing step is to allocate FAF region level imports and exports across ports. FAF defines a number of international gateways, and these correspond well to the limited number of international rail border crossings in North America. The term port is used here to generally include rail border crossings. However, a significant portion of the FAF rail transborder records do not presently use the designated gateways, instead having ports identified as regular FAF regions. No effort was made to improve allocation of shipments to the defined rail border crossings.

4.4.4.1 Imports

For flows occurring in both the base year and the target year, import tonnage of a given commodity from a foreign source to a FAF region is allocated among ports of entry using the following formula:

W_i,c,p,r,t = W_i,c,r,t*P_i,c,p,r,t-1

where:
W = Rail freight tonnage of imports,
P = share variable,
i = commodity,
c = country of origin (Canada or Mexico),
p = port,
r = destination FAF region (in state s),
t = target year, and
t-1 = base year.

The value of the import tonnage for the commodity is calculated in a similar fashion using the same share variable, P:

V_i,c,p,r,t = V_i,c,r,t*P_i,c,p,r,t-1

where V is the value and all subscripts have the same meaning as previously.

The port share is again based upon the weight of i destined from c to r via p in the base year as a portion of the total weight of i destined from c to r via all ports involved in the trade. Mathematically, this share is expressed as:

The set X contains all ports handling commodity i between country s and region r during the base year.

In the case where a commodity was not handled by the FAF region in the base year, the allocation is based on the each port's share of total trade, by value, from the origin country to the FAF region's state during the target year. This information is found in the port level TFD, which provides total weight and value for all freight between a country, and state via a port.

4.4.4.2 Exports

Allocation of export flows between a FAF region and foreign country to ports follows a similar approach to that of imports. The target year estimate of rail freight export tonnage of commodity i from FAF region r in state s to country c for year t, is calculated as:

W_i,r,p,c,t = W_i,r,c,t*P_i,r,p,c,t-1

with all variables and subscripts as previously defined. The share portion for r becomes:

with X being the set of ports handling commodity i between region r and country s during the base year. Again, the share is also used to apportion value for the flow. Flows not existing during the base year are apportioned to ports based upon the target year TFD port level export data for the state of origin.

4.4.5 Combine Import and Export Data

The previous steps result in two output files for transborder rail freight:

• import flows (weight and value) by commodity, country, port, and destination domestic FAF region; and
• export flows (weight and value) by commodity, domestic origin FAF region, port, and country.

The final processing step is to combine the two files and eliminate any records for which the weight is less than 0.01 kilotons and the value less than 0.01 million dollars. The current FAF limits these values to two decimal points, so values less than these will appear as zeros. In practice, a single rail carload carries 0.07 to 0.10 kilotons, so this step will have minimal impact on overall flows.

4.4.6 Future Directions

The current process appears to give reasonable results when checked against known rail freight flows. However, following are TFD issues that may affect the results:

• The TFD field for value includes more than the actual commodity value for exports, and there is no straightforward way to account for this;
• TFD state of origin/destination refers to the location of the exporter/importer, rather than the ultimate origin or destination of the goods; and
• Some HS codes can map to more than one SCTG.

The disaggregation of state level data to FAF regions should ideally have a theoretical underpinning based on economic variables associated with each commodity. Allocation of origin-destination-commodity flows across ports should also be based upon some current year metric, rather than past patterns. Rail flows have a limited number of crossing points, however, and, given the structure of the industry, changes in allocation patterns are likely of less effect than for maritime or air movements. Of greater interest, perhaps, is addressing the large number of rail transborder records in FAF that do not use the defined set of rail border crossing points. Perhaps the TFD port data can be used more effectively to evaluate these issues, despite its lack of commodity specificity.