Identification and Evaluation of Freight Demand Factors (2012)

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.

 Appendix A - Peer Exchange Synthesis The following pages present a synthesis of feedback on an earlier version of the work presented in this report from a Peer Exchange held May 18, 2010, at the Beckman Center in Irvine, CA. Participants invited to the Peer Exchange represented potential users of the project findings, including private-sector supply chain companies and trade associations (rail, trucking, and others), ports, U.S. and Canadian government officials at the local, state, and federal levels, academics, and consultants. The purpose of the one-day meeting was to discuss the preliminary research findings; obtain feedback; and facilitate a dialogue on freight demand factors, defining those that are cost- effective as early warning indicators of change in freight demand trends. A list of Peer Exchange attendees follows the synthesis. Synthesis of Feedback from Peer Exchange NCFRP #11 – Identification and Evaluation of Freight Demand Factors May 18, 2010 – Beckman Center, Irvine Main themes: • Relationship between this work (based on national-level demand factors) and forecasting needs at subnational (regional/multi-state, corridor, metro or local area) level. o Are any of the identified demand factors transferable (i.e., significant at the subnational level)? o Are data sources available? o Is the methodology transferable? o Do these answers depend on the geographic scale of the question? (e.g., small or large metro area, vs. multi-state corridor) o What is the relationship between this work and local/regional/state travel demand models? o Detailed private-sector data would be needed for more geographically localized models. (also, see Data availability) o Procedures for obtaining/using data at a subnational level, derived from the national methods used in this project, would be helpful for local/metro areas. • Statistical issues and model specification o The model is not intended to provide a forecast (within the scope of NCFRP 11). o Should the model mix indices with actual values? o Need to examine lagging/leading variables. [Addressed as part of early warning indicator analysis] o Could the use of ratios (rates) in the model lead to spurious correlation? o Reconsider inclusion of closely correlated (collinear) variables in Principal Components Analysis (PCA). o Need to clarify how PCA was used in backcasting. • Data analysis issues   A-1

 o Many of the factors used in our model are also used by the private sector (e.g., railroads, trucking industry) in forecasting demand. o What about the effect of specific commodities on freight transportation demand? A related factor is domestic vs. import/export shares. ƒ Bulk agricultural products were evaluated in our work (though peer input indicated this system is in a state of change). ƒ Coal transportation (see Baltic dry index; data sources include EIA and railroads) ƒ Salt (as road de-icer) ƒ Soda ash ƒ Sand & gravel ƒ Phosphate/fertilizer ƒ High-fructose corn syrup (changing levels of demand due to public awareness, government policy?) ƒ Both panel and peer exchange suggested removing some commodities (esp. coal and grain) from the analysis to see if model is still valid. o How does the total tons moved by mode (as per our model) compare with national totals? o How might it be possible to fold in air, water freight modes? o How does the model account for industry/economic trends? ƒ Increase in offshore manufacturing ƒ Increase in transloading o The RFP lists a wide range of factors, but many are not addressed in the model. Why? o Could the model also be used to examine demand per capita and per unit of GDP? o How does the model account for the impact of technology on freight demand throughout the supply chain? Productivity measures may be appropriate for this purpose [these may be captured in some of the identified demand factors]. o Can any measures of freight value (by truck or rail) be incorporated in the model? o Could the development and use of supply chain data (as opposed to commodity- based data) offer a more robust means of model development for demand forecasting purposes? Supply chain archetypes could be developed as a framework for this approach. o Could the model address factors related to transportation service levels or characteristics? o What about the geographic focus (on U.S. only)? Could the model be used to forecast or backcast for Canadian, Mexican rail or trucking?   A-2

 o There was agreement that variability in data is increasing sharply, with recent examples (past ~18 months) from all sectors (e.g., fuel price, trans-border freight flows). o Need to be able to distinguish long-term trends from seasonal variations in freight demand. o How far out in time is it reasonable to forecast? Different government/private sectors may use widely varying time horizons (from one year to 50). o Which factors might prove to be leading indicators that are reliable more than 2- 3 months into the future? o Early models of freight transportation separated competitive from non- competitive goods movement; this work does not. • Data availability o Could the model be improved by considering data sources that are not available without charge (e.g., higher-frequency data for some indicators, private-sector data)? Could guidelines be developed based on the value of the data for decision making? o Private-sector data is not available except on a very limited basis, due to privacy/confidentiality concerns, potential release of competitive information, risk of use in enforcement or trigger of legal liability. These barriers can be overcome by being clear on purpose/use of data; use of written agreements to protect data privacy; de-identifying data sets (e.g. trucking company bid histories, truck probe data); identifying the benefits of data sharing for the private sector; developing a trust relationship over time as to how the data will be used; possibly paying for data. o Trade associations may offer some data, which may or may not be representative. o Private companies may share data via investor relations function. o The U.S. has a data gap in urban truck movements. Canada now monitors 25% of daily vehicle km, uses information to reduce congestion. Data could also be used in forecasting demand or the need for road maintenance or construction. Canadian data is more related to supply chain than commodities. Includes freeways, local roads, border crossings; engine type, fuel consumption, braking information, chained trips. Data is compared with volume- and speed-based performance measures. o The goal in this project was to develop a model that did not rely on the use of proprietary data sources or those that might be short-lived, or that specified measures where no data could be gathered. Not all government data is free, and not all public data is permanent or long-lived. • Underlying (or overarching) trends & context o Long-term shift from consumption-oriented society? Would shift to consumption of services vs. goods affect the model? o Global climate change impacts on fuel/transportation modes and costs?   A-3

 o “Peak oil” effects on fuel cost. o Government/regulatory policy with regard to climate change (e.g., in California). o Consumption trends in other countries (e.g., China, India). o Resolution of U.S. trade deficit. o Growing trend towards internet sales. o First-time trend towards increasing U.S. household size. o National resources available for freight infrastructure construction. o Growing interest in local sourcing of food (may be a small fraction of market). o Trend towards denser, urban-infill development patterns. o Roadway congestion in urban areas. • Value of this modeling effort o At the regional/local/gateway level, as a tool to help communicate to policy makers the role and value of freight movement in the overall economy. o At the national level, as a guide to infrastructure investment needs. o Can help academics teach students about freight demand trends. o Can help state/metro agency define the need for an infrastructure project, design funding strategy, and obtain funding.   A-4

 List of Peer Exchange Attendees (excluding Project Team members) Name Affiliation Beningo, Steve BTS/Research and Innovative Technology Admin. Bingham, Paul IHS Global Insight Brodin, Doug California Department of Transportation (Caltrans) Casgar, Tina San Diego Association of Governments Deitz, Ron Bureau of Transportation Statistics Drumm, Scott Port of Portland Fuller, John University of Iowa Goodchild, Anne University of Washington Hancock, Kathleen Virginia Tech Holguin-Veras, José Rensselaer Polytechnic Institute Ivanov, Barb Washington State DOT Kirkpatrick, Mark Canadian Pacific Railroad Lahsene, Susie Port of Portland Lepofsky, Mark Visual Risk Technologies Logan, Bob Consultant Ludlow, Donald Cambridge Systematics McCormack, Ed Washington State DOT Mickle, Michael Alpha Decision Núñez, Juan Burlington Northern Santa Fe Railway Palmerlee, Tom TRB Petzold, Roger FHWA, Borders & Corridors Pisarski, Alan Consultant Regan, Amelia University of California, Irvine Resor, Randy U.S. DOT Rhodes, Suzann Wilbur Smith Rogers, Bill TRB Shalia, Rakesh Federal Express Shi, Huajing Port Authority of NY/NJ Short, Jeffrey American Transportation Research Institute (ATRI) Tardif, Robert Transport Canada Zmud, Johanna NuStats   A-5

Literature Review to Investigate Factors Affecting Transportation Demand The Research Team used a “case” approach to review and analyze a variety of recen t and relevant studies on what factor affect the demand for transportation. The literature analysis was developed in two steps. The in itial step included a review of a significant number of documents based on their relevance to freight dem and estimation. NCFRP-01, the “Review and Analysis of Freight Transportation Markets and Relationships,” was scanned to identify relevant docum ents. A thorough review of the papers/docum ents was conducted and yielded a representative collection of freight demand models that have been used in the past. Du ring the second step of the literature review process, the Research Team performed an in-depth analysis of the various models that are featured in the following documents: In reviewing the literature for the purpose of identifying factor affecting freight demand, it was important to distinguish between general demand factors and the specific datasets actually utilized to represent these factors. In some cases, datasets employed to represent certain factors may actually be outputs of other models or summarizations (e.g., the output from an external regional economic growth model). For each of the complex and varied datasets in the model survey literature, the Research Team endeavored to tease out the underlying factors. The models in the literature are diverse. They include both commodity and vehicle-based models over a variety of different geographic contexts. For each model summary, a rough distinction is also made between factors used to help estimate pure or economic demand and factors that are used to help estimate network demand. With pure demand, the primary concern is generally the demand generation stage – which factors, if any, are used to determine the quantities and flow patterns associated with freight? They are often similar to those that are typically used in econometric demand modeling, such as employment, commodity characteristics and business activity. For the network demand, the factors tend to be associated with the distribution of freight traffic. Of course, there is some overlap between factors that can be used to inform about network demand. For instance, business activity and concentration of land use can help determine how much of a certain commodity is demanded. At the same time, such factors can help determine where and how the commodities are shipped. B-1 Appendix B - Literature Review

Economic Indices for the Transportation Sector Kajal Lahiri &Vincent Wenxiong Yao, Transportation Research Part A 40 (2006) 872-887 Indicators are developed for the transportation services sector to identify its current state and predict its future. The work tries to identify the relationship of the transportation sector to the aggregate economy as well as develop statistical procedures that can capture changes (ups and downs) in the transportation sector. A set of leading indicators is also developed using rigorous statistical processes and is found to be a useful forecasting tool. Relevance to NCFRP 11: Indicators identified help understand how the transportation service sector is changing both currently as well as in the future. While these indicators do show how the freight industry and broadly the transportation services sector is performing as well as how it might perform in the future, the nature of freight demand itself and its drivers is less well explored. Nevertheless, leading factors are identified that could be used as inputs to freight demand analysis. Freight Demand Factors Cited: Factors are identified that are representative and allow development of a composite leading indicator (CLI). The following list are not considered to be freight demand factors following the definition set forth for this project; however, these indicators may allow for identification and detection of trends and the transportation’s conditions as well as serve as predictors for future freight performance. Below some of the factors that form the CLI are shown below which include the Dow Jones Transportation Average, PMI – Inventory diffusion index, New Orders (Transportation Equipment), Shipments (Transportation Equipment), Industrial production (Transportation Equipment), Payrolls (Transportation Equipment) and the consumer sentiment index. US transportation leading factors Factors (up to 12/2003) DJTA (20 Stocks) 0.098 PMI - Inventory diffusion index (PMI- inventory) 0.091 NO (TE) 0.058 Shipments (TE) 0.140 IP (TE) 0.256 Payrolls (TE) 0.220 Consumer sentiment index (CSI) 0.137 B-2

Nationwide Freight Generation Models: A Spatial Regression Approach David C. Novak, Christopher Hodgdon, Feng Guo, and Lisa Aultman-Hall Networks and Spatial Economics This paper investigates the application of linear regression models and modeling techniques in predicting freight generation at the national level within the U.S. Specifically, the paper seeks to improve the performance and fit of linear regression models of freight generation. The paper provides insight into different variable transformation techniques, evaluation of the use of spatial regression variables and the application of a spatial regression modeling methodology to correct for spatial autocorrelation. The paper concludes that the spatial regression model is the preferred specification for freight generation at the national level. Relevance to NCFRP 11: The paper adopts a freight generation modeling methodology at the nationwide level by using proprietary data sources such as TransSearch as using public data sources such as the Commodity Flow Survey (CFS). The dependent variables include county-level freight origin – destination for two commodities (paper and machinery excluding electronics). They also adopt spatial approaches to correct for inherent spatial autocorrelation, which introduces model more sophisticated interactions between different zones across the US (i.e. they attempt to model “spill-over” effects from one region to another). Freight Demand Factors Cited: Ž Total Employment and Employment in different sectors of the origin and destinations Ž Distance between origins and destinations Ž Port tonnage for particular zone Ž Highway length for a particular zone B-3

B-4 Forecasting freight demand using economic indices Jonathon T. Fite, G. Don Taylor and John S Usher, John R. English, and John N. Roberts International Journal of Physical Distribution & Logistics Management 32 (2002): 299-308 The paper describes the results of an effort to predict freight volume in the truckload (TL) trucking industry with an objective is to develop a method and identify relevant independent economic drivers that may provide a reasonable basis for forecasting demand. With a combined dataset that covers 31 months of nationwide volumes provided by J.B. Hunt Transport and a series of 107 economic indices, the authors pre-screen each variable to identify correlations and applicable lead times for national, regional and industrial segments. Then through a series of stepwise multiple linear regression models, the authors estimate the association between TL freight volume with independent demand indicators. Relevance to NCFRP 11: This paper conducts a search for independent economic variables that may potential indicators for forecasting. The authors make a specific effort to look into lead time structures rather than lag time structures. A number of the 107 variables utilized in the study are similar to the variables used in the freight demand factor research. Unlike the NCFRP research, Fite et al. (2002) findings on the regional and industrial segments, finding that smaller segment/cell models are more sensitive to event-specific effects Freight Demand Factors Cited: National level factors: Ž Producer commodities price index of construction materials and equipment (3-mo lead)** Ž Retail store sales of automotive dealers Ž S&P 500 index Ž Producer Commodities price index for household furniture Ž US exports Ž Dow Jones industrial stock index Ž Producer commodities price index for commercial furniture Regional factors: Ž Retail inventories for all retail stores (a top pick) Ž Producer commodities price index for processed poultry and meats** The following is a complete list of factors that were considered for the analysis: Truck tonnage index (American trucking association) Purchasing managers’ index Dow Jones utilities index S&P 500 stock index Unemployment claims Manufacturers’ new orders of consumer goods and materials Consumer expectation index US exports Total manufacturing prodtn index Iron and steel prodtn index Electric machinery prodtn index Aircraft and parts prodtn index Furniture and fixtures prodtn index Paper products prodtn index Canned and frozen foods prodtn index Foods prodtn index Consumer goods prodtn index Materials prodtn index Total manufacturing sales

Manufacturing sales of nondurable goods Manufacturing finished goods inventories Manufactures’ total unfilled orders Manufacturer’s’ unfilled orders of nondurable goods Manufacturers’ new order of nondurable goods Total retail store sales Retail store sales of nondurable goods Retail store sales of furniture stores Retail store sales of apparel and accessory stores Retail store sales of general merchandise stores Retail inventories for durable goods stores Retail inventories for apparel stores Retail inventories for food stores Retail inventories for general merchandise stores Total wholesale inventories Wholesale inventories of nondurable goods stores Producer commods. price index of all food Producer commods. price index of livestock Producer commods. price index of fluid milk Producer commods. price index of fish Producer commods. price index of processed poultry Producer commods. price index of metals and metal products Producer commods. price index of nonferrous metals Producer commods. price index of machinery and motive products Producer commods. price index of construction materials and equipment Producer commods. price index of construction materials Producer commods. price index of prepared paint Producer commodity price index of lumber Producer commodity price index of household furniture Producer commodity price index of floor coverings Retail premium unleaded gasoline prices Interest rates for conventional mortgages Interest rates for prime six month commercial paper Interest rates for three month treasury bills Composite long term government securities Truck tonnage index three month CMA (American trucking association) Dow Jones transportation index Dow Jones composite index Composite index of leading indicators Contract/order plant equipment Metal mining prodtn index Nonelectric machinery prodtn index Motor vehicle and parts prodtn index Lumber and lumber products prodtn index Rubber and plastic products prodtn index Tobacco products prodtn index Final products prodtn index Equipment prodtn index Manufacturing and trade inventories Manufacturing sales of durable goods Manufacturing materials/supplies inventories Manufacturing work in process inventories Manufacturers’ unfilled orders of durable goods Manufacturers’ new orders of durable goods Wage and salary disbursements Retail store sales of durable goods Retail store sales of building supply stores Retail store sales of automobile dealers Retail store sales of food stores Retail inventories for all retail stores Retail inventories for building supply stores Retail inventories for nondurable goods stores Retail inventories for furniture and appliance stores Automotive retail inventories Wholesale inventories of durable goods stores Producer price index of all commodities Producer commods. price index of grains Producer commods. price index of eggs Producer commods. price index of meat Producer commods. price index of dairy products Producer commods. price index of sugar and confectionary Producer commods. price index of iron and steel Purchasing power of the 1982 US dollar Producer commods. price index of agricultural machinery Producer commods. price index of electric machinery and equipment Producer commods. price index of concrete ingredients Producer commodity price index of flat glass Producer commodity price index of plywood Producer commodity price index of commercial furniture Retail unleaded gasoline prices Gasoline production Interest rates for three month CD Interest rates for prime 90 days Interest rates for six month treasury bills Gross domestic products Housing starts US imports B-5

Truck Volume Estimation via Linear Regression under Limited Data Maria Boilé and Michail Golias The authors utilize a series of linear regression algorithms to ‘train models’ when training data is limited. Four algorithms are developed, including Ridge Regression (RR), Lasso Regression (LR), Partial Least Squares Regression (PLSR), and Constrained Linear Least Squares Optimization (CR). Focusing on 14 highways in New Jersey, the authors test the models on data segmented by roadway class. Demand drivers were primarily socioeconomic indicators specific to each section, and the geographic area included was determined using a search for the optimal buffer zone from each roadway class. The robustness and optimality of a modeling exercise using limited data can be maximized by systematically testing multiple training models as well as incorporating the use of precision tools such as GIS. Relevance to NCFRP 11: The iterative process employed tested multiple types of models on data with varying coverage area. Similar to the NCFRP-11 process, testing multiple models may shed light on optimal utilization of data. The successful inclusion of local socioeconomic indicators provides evidence that regional indicators may be utilized in estimating truck traffic. Future work that considers regional modeling may benefit from GIS-based approaches to determine appropriate regional influences and augment precision of study estimates. Freight Demand Factors Cited: The paper considered 3 socioeconomic variables for 11 SIC industries were considered in addition to population. Socioeconomic Factors Ž Number of employees Ž Sales Volume Ž Number of Establishments Ž Population Industries for which socioeconomic factors (excluding population) are considered Ž Mining Ž Agriculture Ž Manufacturing Ž Construction Ž Transportation Ž Utilities Ž Retail Trade Ž Wholesale Trade Ž Real Estate Ž Finance/Insurance Ž Services B-6

Forecasting Truck VMT Growth at the County and Statewide Levels Feng Liu and Robert G. Kaiser In this article, the authors develop statistical models to forecast truck VMT growth of four facility categories at the county and statewide levels. The models incorporate both socioeconomic and transportation system supply variables, and various specifications were evaluated for statistical validity. The results indicate that local socioeconomic variables explain a considerable amount of the truck VMT variance, particularly for urban interstate and non-interstate facilities. Adding external driving forces such as truck corridor or contributing state gross sate product variables increase the models’ explanatory power, particularly for rural interstate facilities. The authors find that the modeling provides reliable results across geographies and a cost effective solution to developing a statewide travel demand model. Relevance to NCFRP 11: Liu and Kaiser utilize simple OLS and Fixed Effects OLS models to construct forecasts of truck VMT. This paper largely follows similar estimation exercises by segmenting the data sample by class and incorporating socioeconomic factors to drive truck demand. The model is generalizable to broad road classes, which give some level of insight into the expected changes across a state. Freight Demand Factors Cited: Socioeconomic Factors Ž Number of households Ž Population and population density Ž Employment and employment density Ž Employment by sectors Ž Per capita income and household income Ž Population by age Ž Retail sales Freight/Transportation Data Ž Commodity Flow Survey (CFS) Ž Gross State Product (GSP) Ž Lane Miles B-7

Future Freight Transportation Demand (National Urban Freight Conference 2006) Paul Bingham The presentation focuses on underlying factors that drive freight demand. Trade growth is rapidly increasing with widespread pressure and demand for adequate levels of transportation and freight logistics services. Trade growth is influenced by factors beyond the common underlying demand for consumption of goods. The author makes particular note that global logistics sourcing, global trading blocks, harmonization of trade and regulatory policy, security standards, and increased freight and traffic congestion have dissimilar effects and thus leading to uneven growth trends across regions. A disproportionate share of freight growth is concentrating at major hubs, crossing and gateways in urban regions. Provided increased pressure, transport infrastructure is reaching capacity, and there is now increased demand for transport labor and optimal land-use. Relevance to NCFRP 11: The presentation brings into light some critical areas that are applicable to this research. While it may not be possible to explicitly model policy considerations that are not explicitly tied to demand, these qualitative concerns may aid in guiding the validity and logical assessment of forecast approaches. Provided the expansion in demand for transport, metrics concerning land-use and socioeconomics (e.g. employment) may be useful indicators for understanding the disproportionate changes in trade growth. Whereas national models may paint the overall picture, knowing that small patches of high growth are present at a regional level may suggest that regional models may be more appropriate in the future forecast environment. B-8

Monthly Output Index for the U.S. Transportation Sector Kajal Lahiri, Herman Stekler, Wenxiong Yao, and Peg Young In this article, the authors develop a monthly output index for the U.S. transportation sector from January 1980 through April 2002, covering air, rail, water, truck, transit, and pipeline activities. Additional indices are developed for each freight and passenger demand segments. A Total Transportation Output Index is constructed using five freight component series and three passenger transportation series. It is found that the total transportation output index’s performance is closely predicts the output indices produced by other federal agencies. Furthermore, the index appears to be suitable to capture growth slowdowns in the national economy. Generally, the index’s historical performance tends to lead NBER growth cycle turning points by six months for peaks and five month for troughs. Relevance to NCFRP 11: The indices constructed by the authors present a method of estimating output of the transportation sector reflective on a subset of various mode performance indicators. The weighting method does not utilize a formal estimation method. A comparison between the constructed index and existing indices finds similar prediction accuracy. Compared with a formal econometric demand model, index construction approaches estimation of economic activity from a higher level using products of demand (e.g. air revenue miles) as opposed to demand factors (e.g. wages, population), and the index method is a suitable approach for understanding larger geographies where as formal modeling approaches may be more sensitivity to micro-level interactions. Freight Demand Factors Cited: The factors used to construct the index in this article are products of demand rather than the drivers themselves. A number of these factors are considered in the NCFRP-11 modeling exercise. Freight Components: Ž Trucking tonnage Ž Air revenue ton-miles Ž Rail revenue ton-miles Ž Waterway tonnage indicators Ž Pipeline movements of petroleum products Passenger Components: Ž Air revenue passenger miles Ž Rail revenue passenger miles Ž National transit ridership B-9

Final Report on Contract Number NCTIP97-21: Development of a Freight Forecasting Model to Forecast Truck Flow Between NJ Counties Themselves and Between NJ Counties and Other States Kenneth Lawrence and Gary Kleinman The aims of the project are to develop a model with the capability of predicting commodity flow information via trucking. In the context of New Jersey, the authors develop a series of database tools to allow for decision makers to easily view the available data. A forecast analysis system is developed to inform agencies on the need for new roadways. The forecast modeling approach follows a gravity flow model for freight tonnage flows in which the best model is assumed to be one that generates the most accurate backcast. Drivers of the model included populations of O-D states, distance between O-D state, person incomes, wages, and total employment. Relevance to NCFRP 11: By explicitly incorporating physical or economic distance, the gravity model estimates the propensity to trade or move goods between two locations. Essentially, the propensity to move goods between points A and B can be a function of socioeconomic demand factors, which then allow researchers and policy analysts to understand traffic on certain links. The type of model has the ability to answer questions that are central to long term planning of transport links whereas national aggregate models for NCFRP-11 provide answers to issues with macro-scale implications. Freight Demand Factors Cited: Socioeconomic › Total employment › Population › Earnings › Total Personal Income Commodity flow survey data (1993 CFS) › Weight of shipment (or tons) › Ton-miles › Distance between origin and destination B-10

Title: A Survey of the Freight Transportation Demand Literature and a Comparison of Elasticity Estimates Chris Clark, Helen Tammela Naughton, Bruce Proulx, Paul Thoma Prepared for Institute for Water Resources, U.S. Army Corps of Engineers This study reviews various aggregate and disaggregate choice methodologies employed to estimate freight demand. Methods from notable contributions to demand estimation are discussed with close attention paid to functional form and level of aggregation. From comparing elasticity estimates across modes and methods, the authors find that generally aggregate and disaggregate models tend to produce noticeably different elasticity results noting that a number of contextual factors may also contribute to differences. Based upon the contrasting differences, the survey concludes with a discussion of considerations for improving demand estimation and research. Relevance to NCFRP 11: The authors review various modeling approaches and bring attention to a number of weaknesses that are common in model specification: 1) the lack of focus on between mode competition in demand modeling, 2) the imprecision of aggregate data in price elasticities, 3) a need to consider both short and long term estimation, and 4) a more solidified consideration for underlying motives of functional forms. An interesting insight is that the Various models are examined: Aggregate Demand Models: modal split models and neoclassical aggregate demand models Disaggregate Demand Models: inventory and behavioral models Freight Demand Factors Cited: Factors considered in both types of models are similar; however, the main difference is the level of aggregation and the type of interaction. Disaggregate models may take the perspective from a shipping manager where as an aggregate model may be somewhat removed from theory at the individual decision level. The following highlight the inputs from a selection of example models: Inventory-based demand (Disaggregate model) Some authors may include: shipping cost per unit, mean shipping time, variance of shipping time and carrying cost per unit of time while in transit. In order to determine how a shipper chooses between modes, the shipper’s indifference curve is specified Neoclassical Aggregate Demand Models The data may used may be commodity group-specific and consist of the distance of links, total tons moved, average freight rate, transit time and its variability by mode on each link. B-11

Title: Commodity Flow Modeling William R. Black Transportation Research Circular (1999): 136-54 This research focused on the primary objective of creating a database of commodity flows of counties in Indiana and to allocate commodity tragic to the state’s transportation network. The approach undertaken in the study utilized multiple quantitative and technical tools, including TransCAD (GIS), multivariate regression, entropy-based gravity model algorithms along with other database tools. A transportation planning framework is adopted to identify networks, to estimate production and attract of commodity flows, to determine traffic distribution from O-D pairs using a gravity model, to determine mode splits, and to assign traffic to links. Nineteen commodity groups were considered in this analysis and forecasts traffic for 2005 to 2015. The aims and implication of the modeling system is to aid decision- makers in determining and assaying alternative options for investment in transport infrastructure. Relevance to NCFRP 11: The commodity flow model utilizes an interdisciplinary approach to build a sophisticated representation of Indiana’s transportation network and its commodity flow patterns. The author uses TransCAD to create the transportation network, update data and traffic assignment. The particular significance of using geographically precise data with micro-level ability diverges from the larger regional or national aggregate models that have been reviewed in this literature review. Freight Demand Factors Cited: Traffic generation drivers › Population › Employment by sector B-12

Container Demand in North American Markets: A Spatial Autocorrelation Analysis Wilson, William W and Camilo Sarmiento. 2007. Researchers utilized a Cross-Sectional Spatial Autocorrelation Analysis to identify how and if demographic characteristics, primarily personal income and population were related to container demand. There was a significant positive relationship with income and population. As the study was performed for various North American markets, there was the opportunity to analyze infrastructure and spatial characteristics of the markets in relation to one another. The number of terminating railroads in an area as well as the number of interstate highway miles in a focus area had significant positive relationships with container demand in that area. Location near a port facility also had a significant positive relationship with container demand, while container demand is overall, larger on the West Coast largely due to the relative proximity of those port facilities being in a favorable position to serve Asian trading markets. Some areas in the South also have relatively high demand. Interestingly, locations adjacent to locations with high container demand had lower-than- expected demand for containers. While this is significant on a regional transportation planning basis (e.g., Nashville’s demand for containerized freight will tend to be subordinated by Memphis’ greater gravitational attraction due to its status as a rail hub), it has less relevance to a national planning study. Relevance to NCFRP 11: While the study examined and compared selected areas in North America to identify regional factors that influenced the relative demand for containers, key conclusions from the analysis indicate that both population and personal income have a significant positive relationship with the demand for containerized freight. Freight Demand Factors Cited: Socioeconomic › Total population › Total Personal Income B-13

Freight Travel Demand Modeling – Synthesi s of Approaches and Development of a Framework, Pendyala, Ram, V. Shankar and R. McCullough, 2000 William W and Camilo Sarmiento. 2007. A study of commodity flow for the State of Indiana, the model included rail and truck traffic. Trip generation was performed using linear regressions with employment and population as independent variables. The model forecast commodity productions and attractions. Distribution used a constrained gravity model. For the State of Wisconsin, Pendyala et al. developed a model Relevance to NCFRP 11: While both independent variables of employment and population were utilized and helpful in explaining commodity flow, the Study Team at Halcrow and TTI noted the high colinearity between these two variables. Freight Demand Factors Cited: Socioeconomic › Total population › Total employment B-14

Container Demand in North American Markets: A Spatial Autocorrelation Analysis Wilson, William W and Camilo Sarmiento. 2007. Researchers utilized a Cross-Sectional Spatial Autocorrelation Analysis to identify how and if demographic characteristics, primarily personal income and population were related to container demand. There was a significant positive relationship with income and population. As the study was performed for various North American markets, there was the opportunity to analyze infrastructure and spatial characteristics of the markets in relation to one another. The number of terminating railroads in an area as well as the number of interstate highway miles in a focus area had significant positive relationships with container demand in that area. Location near a port facility also had a significant positive relationship with container demand, while container demand is overall, larger on the West Coast largely due to the relative proximity of those port facilities being in a favorable position to serve Asian trading markets. Some areas in the South also have relatively high demand. Interestingly, locations adjacent to locations with high container demand had lower-than- expected demand for containers. While this is significant on a regional transportation planning basis (e.g., Nashville’s demand for containerized freight will tend to be subordinated by Memphis’ greater gravitational attraction due to its status as a rail hub), it has less relevance to a national planning study. Relevance to NCFRP 11: While the study examined and compared selected areas in North America to identify regional factors that influenced the relative demand for containers, key conclusions from the analysis indicate that both population and personal income have a significant positive relationship with the demand for containerized freight. Freight Demand Factors Cited: Socioeconomic › Total population › Total Personal Income B-15

B-16

1. The Freight Story: A National Perspective on Enhancing Freight Transportation, Federal Highway Administration, 2008 Examines the nature of freight movem ent, identifies challenges to improving freight productivity and security, and presents st rategies to increase freight productivity. Intended to be a point of departure for furt her examination of policies, program s, and initiatives that m ight be undertaken by stakeh olders at all levels of governm ent. This effort also involved the developm ent of an in tegrated freight data and analytical system, called the Freight Analysis Fram ework (FAF), which is intended to enable decision- makers to identify areas in need of capacity improvements. Relevance to NCRFP 11: Moderate relevance to NC FRP-11. Some freight dem and factors are discussed, but the majority of the docum ent focuses on decision-m aking related to key challenges facing the US freight network. Freight Demand Factors Cited: - Composition of the Local/State/National Economy (as th is has shifted, the distance goods has travel increased), i.e., shift for m anufacturing to a servi ce economy in the US (Economic) - Customer demand changes (need for more flexible, reliable, timely service: - Growth in traffic of smaller shipments - Traditional, high-volume demand will be a smaller portion of overall demand - Deregulation (Public Policy) - Rail: as a result of Staggers Act, increase in rail shipments - Trucking: as a result of truck dereg, increase in interstate carriers, decrease in in- house trucking, decrease in empty back-hauls - Air: growth in air freight, improved efficiency - Shift from “push” logistics to “pull” logistics (Economic) - Greater multi-modal integration/coordinated logistics - Globalization – increase in merchandize trade (from 11% in 1970 to 25% in 1997) (Economic/Public Policy) - increase in ocean shipping, container facility demand, intermodal demand, pressure on international gateways/feeder corridors - NAFTA – increase in North-South freight movement 2. Freight Demand Modeling – Tools for Public Sector Decision-Making, Transportation Research Board, Conference Proceedings 40 Proceedings of a conference d esigned to complement the Federal High way Administration’s work on the Freight Mode l Improvement Program and focused on (1) modeling methodologies, (2) applications of existing models at the national and local levels (including international examples), and (3) related data needed to support modeling B-17

efforts. As with othe r reports, there is the general sense that freight dem and modeling is often not a priority and too of ten influenced by inadequate tools originally designed for passenger demand modeling. Relevance to NCFRP 11: Moderate relevance. The series of papers presents a diverse overview of freight modeling issues designed for a policy conference setting, so relevant sections tend to be scattered throughout. Overall, there is more of a focus on modeling processes than on the fundamental demand drivers. Freight Demand Factors Cited: - Transportation/inventory costs (which, in aggregate, equal total logistics costs): The relative costs of each (the balance of total logistics costs) have an impact on freight decision-making - Relative Regional Purchasing Power: increases in this will positively impact freight trip attraction - Network capacity/ease of movement: typically influenced by policy decisions/ technology implementation and, in turn, has a direct impact on transportation costs - Types of commodities produced/demanded and where - “Real GDP growth in the long run is correlated well with truck transportation, but on a quarter-to-quarter basis it may not be” (p. 83) – Thus, the time period being considered may impact the relevance of certain factors. *** 3. Review of Freight Modelling http://www.dft.gov.uk/pgr/economics/rdg/rfm/ In September 2001, the UK Departm ent for Transport commissioned a short research study to review the options for m odeling and forecasting freight including those used in other countries in order to assess which techniques would be most suited in Great Britain. The review covered road freight and other modes as well as modeling light goods vehicle (LGV) movements. The study focused on three main areas: A) What is needed? - Market structures and issues in freight B) What is available? - Review of models and data C) The way ahead - specification of models and data requirements There’s a lot of material on the project website and we haven’t reviewed everything. It was determined that “Report A1: Issues in Logistics and the Freight Industry” and “Review of Data Sources” would be the most relevant to NCFRP11. Relevance to NCFRP 11: Since the scope of the study is relatively broad, only a portion of the work product is directly relevant to the NCFRP 11 work. Specifically, the section that presents a “Review of Data Sources” includes a large number of factors that are likely to impact freight demand. However, while a number of factors are listed, it is not always clear how these factors might impact freight demand (direction and magnitude of change). Freight Demand Factors Cited (From the “Review of Data Sources” Section): B-18

1) From subsection “Current Freight Models and Their Use of Data” - National Road Traffic Forecasts (NRTF): docum ented relationships established f or each industrial sector o ver the p ast 21 year s. Also used economic growth forecasts by sector (linkage between econom y and dem and for freight tonnes). Generated separate forecasts for both LGV (light ) and HGV (heavy) vehicles (b ased on simple relationship between the two). - EUNET Trans-Pennine Model: com bination of national input-out put tables, regional accounts and em ployment data by econom ic sector/zone to estim ate the inter industry trade linkages, ultimately generating the freight O/D movements…forecasts are based on forecasted growth by zone in final consumption for each economic sector - Portland (Oregon) Model: similar to EUNET, but also incorporates shipment sizes by commodity/vehicle class, carrier characteristics by commodity type, operator behavior relationships, location of main transshipment centers, detailed land used data for allocation of shipments to detailed locations 2) From sub-section “Description of Data Source” - Economic/financial data: How do operating costs/tariffs affect demand? How does the transport system react to a policy initiative that changes these variables? - Aggregate Statistics/Trends: o value and volume of imports/exports by economic sector o regional value added and overall production by sector as part of the Regional accounts o National Input-Output Tables of the linkages in value units from producers to consumers in each economic sector for both domestic and international trade o Time series of past trends in economic variables o Problem: the bulk goods that are important in transport volume are often of lesser importance in terms of total economic value o Problem: Manufactured goods are often transported as mixed loads, so any miscellaneous loads are difficult to trace back to economic sectors of production - Trends from Surveys / road/rail/port/airport traffic count series - Land use data o Population trends (Census Numbers) ƒ number of households (possible generators for freight trips); ƒ number of employees by SIC/Journey to Work tables (useful indicator, but can be misleading as to actual activity at a site) o Land Use/Changes in Land Use – may form the basis of other data gathering o Business Registries – can be used to examine spatial changes in business activity - Network characteristics/quality of connecting infrastructure/congestion measures (From the “Issues in Logistics and the Freight Industry” Report) B-19

- There is a relationship between GDP and freight movement, but the increasing importance of sustainability could result in a degree of uncoupling - Increased international trade – materials/products are more likely to be transported over greater distances - Changes in supply chain structures/ supply chain integration – longer distance freight transport movements/increase in volumes on key routes/reliance on a smaller number of supply chain partners - Reductions in lead time/Just in Time – usually more frequent deliveries of smaller quantities, but there have been some JIT strategies that effectively consolidate flows/minimize transport intensity - Developments in e-commerce – changes patterns, but impact on freight demand not entirely clear - The increasing role of logistics service providers *** 4. Truck Trip Generation Data, NCHRP Synthesis 298, 2001M This Synthesis Report sets out to identify and evaluate the validity of available truck trip generation data and provides an assessm ent of the current state of the prac tice. It examines the key considerati ons in the development of truck trip generation data and provides a detailed review of available data sources. The report highlights the distinction between vehicle-based models and comm odity-based models, each of which requires different inputs and results in different conclusions about trip generation trends. When assessing the current state of the practice, the rep ort focuses on statewide/metropolitan modeling, transportation engineeri ng applications and general organizational willingness to share data. As with similar freight studies and reports, there is recognition that the st ate of the practice in truck trip generation da ta is less developed relative to practices used in assessing passenger vehicle movements. Relevance to NCFRP-11: Moderate to high relevance. This synthesis focuses on truck trip generation data, which is one component of the NCFRP- 11 topic of freight demand factors. Some underlying demand factors emerge from the summaries of truck trip studies included in the “Review of Available Data Sources” section. Some of the m ost relevant points focu s on the difference between Vehicle-based Generation Models vs. Commodity-based Gene ration Models (and th e applicability of each when modeling short-haul and long-haul activity): - “Most of the metropolitan truck trip generation models developed to date have been vehicle-based models that adhere to methodologies similar to those used in 4-step passenger models….For studies that have more of a focus on freight transportation needs in long haul corridors, commodity-based approaches to trip generation are becoming more popular” (p. 34) Explanation of this trend: - The agencies that tend to focus on the Metropolitan/Short-Haul context tend to be most interested in the impact of regional truck activity. These regional agencies are usually B-20

constrained to accept methodologies due to regulatory requirements (federal transport planning and air quality regulations). So, vehicle models often make more sense in this context. Also, commodity-based models do not include explicit consideration of all reload/tour activities, so they may not be as useful for the metropolitan/short-haul context (they would likely under-estimate urban travel activity). On the other hand, commodity- based models have proven to be more appealing/ revealing for those conducting long- haul/statewide models. However, commodity-based trip rates are rarely published and are hard to derive from available data. Freight Demand Factors Cited: The factors cited are all related to truck trip generation, including: - Land Use – Variable examples include acreage of land used, square feet of building floor area, use designation (light industrial park, office, etc) - Employment by (major) industry – Variable examples include manufacturing, construction, agriculture, etc. – This factor is often used in generation models, not because it provides the most accurate results, but because it is often the only data measuring economic activity that are feasibly available to public agencies. Modeling inaccuracies can easily result if changes in labor productivity can impact the actual trip rates per employee. (p. 35) - Economic output – Some variable examples include annual sales, revenue, value of shipments, etc. This may be a better measure of industrial activity than employment, but this type of information is not as easily obtainable. (p 35) - Non-highway modal activity at intermodal terminals (special trip generators) – Some variable examples include number of import/export container moves, TEUs, etc. - Split between “garage-based trips” (single round trips) and “linked trips” (multi- stop trips): This is of particular relevance for metropolitan/short-haul demand studies. The way in which trips are typically executed will impact travel generation/demand. *** 5. State-of-the-Practice in Freight Data: A Review of Available Freight Data in the US Mani, A. and Prozzi, J., Center for Transportation Research, University of Texas at Austin, 2004 This study provides a comprehensive overview and evaluation of available US freight data sources (including both public and commercial databases). A short description of each database is provided, the methodology is explained, the featured freight demand characteristics are listed and the important limitations of the data source are outlined. Relevance to NCFRP-11: Low/Moderate Relevance to NCFRP-11 – While the featured databases provide information that is useful in demand forecasting, there is little discussion of the factors that potentially influence future demand. While historical demand is one determinant of future activity, there are also many other factors that come into play. B-21

Freight Demand Factors Cited: Historical Demand /Freight Flow Patterns – Trends from past years can be projected in the future for reasonable estimates, but the impact of other external factors should also always be considered. In practice, the historical/current demand is often used as a starting point for calculating future demand. *** 6. Characteristics and Changes in Freight Transportation Demand: A Guidebook for Planners and Policy Analysts NCHRP Project 8-30, Cambridge Systematics, 1995. http://ntl.bts.gov/lib/4000/4300/4318/ccf.html The document was designed to be a guidebook to assist transportation planner with all stages of freight demand analyses. It examines the changing character and composition of US multi-modal freight transportation demand and outlines processes for effectively forecasting future demand. Demand forecasting for both existing and new facilities is explained. A structured approach for freight demand policy analysis is also presented. Relevance to NCFRP-11: The document provides a good general overview of freight demand and ways in which it can be analyzed. Highly relevant to NCFRP-11 is Appendix A: Factors Influencing Freight Demand. This section presents a very comprehensive overview of factors, both those with direct impacts and those with indirect impacts on freight demand. Freight Demand Factors Cited: Factors with Direct Impact on Freight Demand - The goods-production component of GDP/GNP, which has tended to grow more slowly than overall GDP/GNP and tends to fluctuate more with business cycles - Industrial location patterns / spatial distribution of economic activity - Globalization of business – global distribution patterns and how they fluctuate over time (very dynamic) - International trade agreements - Just-in-Time Inventory Practices – could increase the frequency of in-bound shipments; decrease lead-time/size for shipments; increase importance of receiving shipments on time - Centralized Warehousing – results in increased transport demand and sometimes a shift between modes (from truck to air, for instance) - Packaging Material – reduction in average density of shipments for some products (with a lot of packaging) – higher volume shipments with less actual weight - Recycling Patterns – increasing the use of recycled materials affects the O/D patterns, lengths of haul, and modal usage for certain commodities. Location of recycling plant vs. location of processing plant. Factors that Affect Demand Through Influence on Costs/Services - Economic Regulation and Deregulation: Significant impact on Air & Road (creating more competitive markets) and Rail (improve mode profitability) B-22

- International Transportation Agreements - Intermodal Operating Agreements - Single-Source Delivery of International LTL Shipments - Carrier-Shipper Alliances - Fuel Prices - Publicly Provided Infrastructure - User Charges - Other Taxes - Government Subsidization of Carriers - Environmental Policies and Restrictions – emissions regulations, phase-out of single hull tankers, fuel quality requirements, air noise restrictions - Safety Policies and Restrictions – speed limits, route restrictions for hazmat - Effects of changes in Truck Size and Weight Limits - Congestion - Technological Advances *** 7. Modeling Freight Demand and Shipper Behavior: State of the Art, Future Directions Regan, A. and Garrido, Rodrigo., 2002 Freight transportation dem and and shipper beha vior have historically been evaluated separately. However, global supply chain trends have made it more important that freight demand models explicit incorporate shipper/ carrier behavior. This paper presents a thorough review and synthesis of research in both fields. Relevance to NCFRP-11: Moderate/high relevance to NCFRP-11 –useful report that identifies the various approaches to freight demand modeling (categorized as either modeling the International, Intercity or Urban contexts) and outlines the variables typically employed in each. Freight Demand Factors Cited: - International Context (Goods Movement Between Countries) - Ricardian theory of international trade: Wage rates, capital stock and prices - Gravity: freight volumes, network “impedance” and spatial attraction - Industrial organization: wage rates, commodity prices, costs functions and their inputs - Input-Output Analysis: technical coefficients, trade coefficients, commodity prices - Spatial and temporal interactions: series of flow in space and time - Intercity Context - Abstract mode Aggregate: level of service for each mode, sociodemographic characteristics - Aggregate logit: market shares, freight rates, level of service for each mode - Neoclassical theory of the firm: freight rates, firm’s short-run cost function and their Inputs, firms expenditures - Time series: level of service for each mode, transportation cost functions - Disaggregate-inventory based: freight rates, commodity values, transit times plus all B-23

the usual EOQ model inputs - Disaggregate-utility maximization: level of service and commodities attributes - Spatio-temporal interaction: spatial and temporal freight flows, sociodemographic data - Urban Freight Scope - Gravity: total productions and attractions in each zone, impedance (usually distance or travel time) between zones - Input Output: technical coefficients, survey and non survey based data. General Factors - Impact of improved logistics technologies/increase in logistics service providers that broker information to shippers, carriers, warehousers, etc. *** 8. The Decoupling of Road Freight Transport and Economic Growth Trends in the UK: An Exploratory Analysis Alan C. McKinnon, October 2006 (was slated to be published in Transport Review in early 2007) While previous decades have shown a very clear relationship between the growth in road freight movement and economic growth in the UK, this relationship has not been as strong in recent years. It has been observed that GDP increased significantly during the period 1997-2004, but the volume of road freight movement has remained relatively stable. This report reviews previous research on this decoupling issue and examines possible causes for this trend. Relevance to NCFRP-11: High relevance, especially if we are going to cite economic growth (defined as GDP growth) as a significant factor for NCFRP-11, this paper helps to determine the actual magnitude of impact. It will be important to determine how many of these findings are applicable in the US context. Freight Demand Factors Cited: The primary demand factor is GDP, which is observed to have a decreasing impact on freight demand. The paper determines that around 2/3 of this decoupling is due to three quantifiable factors: - increased penetration of the UK road haulage market by foreign operators – this has an impact on reporting figures (not really a factor in the US) - decline in road transport’s share of freight market (increase in rail and water shares) - real increases in road freight rates And several factors that are more difficult to quantify - relative growth of service sector - diminishing rate of geographic centralization - off-shoring of manufacturing B-24

*** 9. Upper Midwest Freight Corridor Study—Phase I and II Midwest Regional University Transportation Center, Adams, T., et al University of Wisconsin-Madison. 2007, see http://www.uppermidwestfreight.org/files/final2007.pdf for full collection of documents. The Phase II Final Report is currently saved in our files. Drawn From NCRFP 1 Bibliography/Literature Review This publication is based on eleven white pape rs that were written on factors r elating to freight movement and public policy throughout the Upper Midwest region of the United States. These papers provide information on f actors that influence performance of freight transportation in the regional context. Since the papers covering one or another aspect of our scope, we have outlined in some detail the relevant topics/headings and the associated references (for empirical data or other lit reviews that may be of relevance) Relevance to NCFRP 11: Limited Relevance to NCFRP 11 since it primarily deals with freight planning policies. However, certa in demand factors do em erge from the discussion. Freight Demand Factors Cited: The initiatives/factors covered include: • Federal Funding mechanisms • Highway Technology: • Key items covered: short term gains, technology, policy • Influence of Toll Roads-can be used to extrapolate the impact the cost of tolls to regional transportation. • Alternative Freight Transport: Water based transport • Intermodal Freight Facilities Regional Freight Agenda for the Upper Midwest Defines the priorities that would enhance re gional freight corridor through 12 initiatives. Information sources are useful for technology, economics and monitoring. Conceptual Regional Technology Plan—set s the outline for a technology efficient tracking and the benefits of deployment. Transportation and the Economy Covers the effect of transportation on the economy with China and India as examples. *** 10. Freight Transportation Infrastructure Survey: Causes and Solutions to the Current Capacity Crisis MIT Center for Transportation and Logistics, Cambridge, MA, 2006 Drawn From NCRFP 1 Bibliography/Literature Review B-25

This survey report investigates the pe rception gap am ong freight transportation stakeholders in order to discover the root causes of congestion and the capacity crisis, the resulting effects on business and the solutions that are or should be utilized to assuage these impacts. Following a discussion of the findings, recommendations for short and long term strategies to enhance communication among stakeholders are offered. Questions asked: 1. What were the root causes of the freight transportation congestion and capacity crisis? 2. What were the impacts on business due to the crisis? 3. What actions are being, or should be, taken to alleviate these impacts in the future? Relevance to NCFRP-11: Moderate relevance. While we often look at tangible “factors” contributing to freight dem and and affecting freight patterns, it is all too often forgotten that stakeholder perception and the effec tiveness of communication between various stakeholders are also a very real factors. Specifically, there appears to be a significant disconnect between governm ent stakeholders and shipper/carrier stakeholders w hen identifying root causes of freight transport pr oblems. This is obviously very difficult to model, but it still im portant to tak e this into consideration when trying to deter mine future freight demand. For instance, how w ill public inves tment and policy initiatives really affect shipper/carrier b ehavior? (probably not in the exact way that was initially intended) Drawn From NCRFP 1 Bibliography/Literature Review Freight Demand Factors Cited: Root cause identified as Highway Congestion • Operators concerned with operational inefficiencies • Interfacing with larger population • Inputs (drivers, fuel cost volatility, hours of service rule) Top action items include: • Meet more frequently with carriers • Work on establishing contingency plans to avoid supply chain disruptions • Request solution proposals from carriers • Collaborate with carriers on transportation forecasts - Carriers have most interaction with State, Local govt levels and minimum interaction at the Federal level. Key Findings: Collaboration between shippers and carriers is increasing. Collaboration between shippers/carriers and government agencies is non-existent. Perceptions in the causes and remedies of the congestion crisis differ between the government and the private sectors – but not much between shippers and carriers. On a Scale of 1-4 the following are the rated responses: 1-5 ranked Root causes: B-26

New Hours of Service agreem ent (HOS) not a major factor for carriers & govt. Shippers feel it is of medium importance. Overall transport use increasing-whatever the mode-which indicates an increase in trade volume. Ranking of ports usage: Why do govt ranking of east coat ports differ some much from the shippers ranking of east coast ports? Possible solutions: B-27

Key investments: • Improving existing highways • Building new highways as (Govt ranks neither of these in the top 5) • airport or seaport investment not considered that important. • only the Government respondents ranked building logistics hubs and parks as a top 5 investment. • All stakeholders agreed on the importance of expanding existing intermodal yards and double tracking selected rail lines. • . majority of respondents advocate mixed funding (PPP) strategy should be taken.. • All stakeholders agreed that the government should fund all highway infrastructure – to include both improvements and new Detailed survey results important to have a glance through. *** 11. Freight Capacity for the 21st Century TRB Special Report 271, National Research Council. Morris, J. 2003 This TRB report identifies constraints in the freight planning process that have lim ited the efficiency and productivity of the tr ansportation system. The r eport goes on to recommend changes in governm ent policies th at will contribute to better planning through more rational investments which will ultimately improve the efficiency of freight transportation. Drawn From NCRFP 1 Bibliography/Literature Review Relevance to NCFRP 11: Moderate relevance to N CFRP 11. The paper is primarily concerned with US fr eight system capacity constraints and potential solutions for B-28

alleviating these. Any shortcomings in the transport infrastructure network will certainly impact freight operations and these factors should be considered when modeling demand. Freight Demand Factors Cited: Trends that affect • growing congestion on important highway segments and • slowing of the rate of addition of highway capacity, • rail infrastructure downsizing and • service disturbances, • congestion at terminals and border crossings, • lengthening lead times and • rising costs of infrastructure projects, and • freight–passenger conflicts in cities. Issues: • Increasing population density • urbanization, and wealth ensure that conflicts between freight and passenger traffic; • conflicts between freight transportation and residential, • recreational, and other competing land uses; and • Requirements to control pollution will increase. • These forces will increase the cost of expanding capacity and add to th e risk of investment. Possible solutions: • Policy & increased spending on infrastructure projects. Policy that will have long lasting effect: B-29

*** 12. Principles for a US Public Freight Agenda in a Global Economy Robins, M., Strauss-Wieder, A., 2006, Principl es for a U.S. Public Freight Agenda in a Global Economy. Transportation R eform Series, The Brookings Institute, W ashington, D.C., http://www.uppermidwestfreight.org/files/Brookings_freightsystems.pdf. The objective of this research is to summ arize the key is sues and trends affecting the nation’s increasingly strained freight system, provide examples of efforts to address these strains and the land uses invol ved, and identify the curren t roles played by government agencies. It is determined that a nationwide systems-based and m ultimodal agenda is necessary to maintain America’s competitiveness and economic well-being. Drawn From NCRFP 1 Bibliography/Literature Review Relevance to NCFRP-11: Moderate Relevance to NCFRP-11. Freight Demand Factors Cited: - Changes in the Global Econom y: sourcing/production of goods at the lowest cost location (often off-shore) - US evolved in to an “import economy”. Results in an increase in goods traveling over long distances, which this trend expected to continue into the future (for all modes) - Longer, more complex and leaner supply chains/Just-in-Time Practices B-30

- Changes in the type of transport equi pment: longer truck tr ailers, doublestack containers and m assive ocean ves sels – this m ay not be a m ajor issue for commodity-based models, but vehicle-based demand models will be impacted - Port and Route diversification: logistic s professionals are finding ways to get around deficiencies in the US transport infrastructure - will reorient supply chain to minimize the impact of disruptions (even though it may not be the optimal use of infrastructure) This will shift demand from one part of the system to another, possibly without apparent cause - Rise of “value-added” warehousing: shifts part of the “m anufacture” of products to retail distribution centers - Increasing community awareness/action: opposition to freight movement through communities can im pact public transport policy, thus inf luencing the freight network and indirectly impacting freight flows 13. Forecasting Freight Demand Using Economic Indices Fite, Jonathon T., et al. Inte rnational Journal of Physical Distribution & Logistics Management, Vol 32, No. 4,2002, pp 299-308. The purpose of this paper is to develop a si mple model to forecast freight dem and in the trucking industry. The authors use simple multiple regression tools to es tablish correlations between major economic indices and freight trucking data. One of the data sources that the autho rs use is the monthl y US Departm ent of Commerce release of economic indicators including the Composite Indexes of Leading, Coincident and Lagging Indicators. Relevance to NCFRP 11: Moderate relevance to NCFRP 11. Prim arily because the freight data used was of priv ate, and the analysis was conduc ted over the relatively short time horizon of three years. Freight Demand Factors Cited: Freight Dem and factors are only cited in-d irectly in terms of those economic indices that work be st within the framework of an econometric model (and correlate highly with the freight data). On the national level this is the producer commodity price index of construc tion materials and equipm ent. They also develop regional and commodity based models and describe their correlations. B-31

*** 14. Modeling the Demand for Freight Transport: A New Approach Abdelwahab, Walid and Michel Sargious. J ournal of Transport Econom ics and Policy, Vol. 26, Issue 1, pp 49-70, 1992. The authors develop a sim ultaneous equation econometric model to determine decisions between mode choice as well as factors th at derive demand for a particular m ode. They identify two types of modeling approaches, aggregate and disaggregate. This study uses the disaggregate approach. They use data pr imarily from the US Bureau of Census, Commodity Transportation Survey. B-32

Relevance to NCFRP 11: Moderate relevance to NCFRP 11 since the paper focuses more on the m ode choice com ponent than dr ivers of freight dem and at the aggregate level. Freight Demand Factors Cited: Some of the variables us ed can be divided into commodity attributes (value and density of th e shipped cargo), modal attributes (such as freight charges, reliability and transit tim e by each mode) as well as the other attribu tes such as volume of commodity moved on each link by each mode. *** 15. NCHRP Report 606 – Forecasting Statewide Freight Toolkit The report deals with developing appropriate methodologies for forecasting freight at the state-wide level and providing regional planning organizations with modeling approaches that they can use to m odify their planning needs. The m ain focus of this study is to document methodologies to forecast prim ary freight (i.e. goods moved over long distances and between cities as well as goods by local truck that are in their initial stages of long distance m ovement). It therefore excludes practices for forecasting secondary freight ie. Freight that moves from DC to DC. The paper discusses in depth dif ferent modeling approaches including traditional 4 step models (trip generation, trip d istribution, mode split, n etwork assignment). They also look at 10 case studies and develop a template frame work for freight forecasting. Relevance to NCFRP 11: Moderate relevance at NCFRP 11. The paper dwells in length on different forecasting methodologies that can be used to m ake freight forecasts. Many of the methodologies focus on forecasting traffic by commodity type. Freight Demand Factors Cited: The study identifies various socio-economic factors that can be used to adjust or feed-into the forecasting procedures id entified above. M ost of them relate to employm ent and population forecasts. The study considers various case studies which consider the practices of different State and federal bodies. They look at some of the socio-economic drivers behind the forecasts these include: - Industrial employment projections - Labor projections for different counties - Employment in Retail, Indus trial, Public and Office secto rs (was used to forecast trip rates in the New Jersey State wide model) - Population is also used a driver of trip generation - Employment by commodity sector - Use of a spatial input / output model th at identifies economic relationships between origin and destinations (MEPLAN model coefficients). - county business pattern data from the census bureau B-33

Sources of Freight Data: US Bureau of Census, Commodity Transportation Survey Surface Transportation Boards’ Railroad Waybill Sample US Bureau of Census, Annual survey of Manufacturers US Bureau of Census, VIUS Survey FAF State to State commodity flow database *** 16. Freight Demand Characteristics and Mode Choice: An Analysis of the Results of Modeling with Disaggregate Revealed Preference Data Jiang, Fei, Paul Johnson and Christian Calzad a. Journal of Transpor tation and Statistics, December 1999, pp 149-158. The purpose of this paper is to analyze how freight demand characteristics relate to and influence shipper’s mode choice using a nested logit model as an analytical tool. A large scale national disaggregate revealed preference database is used for shippers in France in 1988. Relevance to NCFRP 11: Moderate relevance to NCFRP 11. The a uthor’s focus on disaggregate data with a focus on understand ing relationships betw een shipper’s mode choice as well as their decision to pursue priv ate (their own) versus public (sold in the market place) transport options. Freight Demand Factors Cited: Freight Demand characteristics are divided into three type s: a firm ’s (shipper’s or receiver’s) characteristics, goods physical attributes, and th e spatial and flow characteristics of shipments. Another way to look at these same factors is in terms of time horizon; Long term factors include the firm’s nature, size, location, information system, structure and trucks owned by the firm (private i.e. a firms logistical capabilities reserved for its own use, versus publicly available transportation systems). Short term factors include: physical attributes of goods, physical flow attributes (frequency, distance, origin and destination) They find that transpo rtation distance, the shipper’s accessibility to transpor tation infrastructure, the shipper’s own transportation facilities and shipment packaging (pallets and parcel) are the critical determinants of the demand for rail and c ombined (rail + truck) transportation. *** 17. What Determines Demand for Freight Transport? Bennathan, E., Julie Fraser and Louis S. Thompson, Infrastructure and Urban Development Department, The World Bank, 1992. This study looks at a cross section of count ries (both high incom e and developing) and aims to derive long run freight dem and characteristics. The authors use single equation B-34

econometric studies to identify key determ inants of demand across the cross section of countries. Relevance to NCFRP 11: Moderate relevance to NCFRP 11. Primarily since the authors looks at a cross section of c ountries, some high income and others developing that it is hard to make the conclusion that same factors might be relevant in the United States. Freight Demand Factors Cited: The authors try to establish a relationship between ton-km of freight and GDP and area of a country. They look at a cross section of developing and developed economies. *** 18. Freight Travel Demand Modeling – Synthesis of Approaches and Development of a Framework Pendyala, Ram, Venky Shankar and Robert McCullough. Transportation Research Record 1725, Paper No. 00-0200. pp. 9-16. The authors try and synthesize approaches to freight forecasting by looking at current practices of different State a nd DOT to forecast freight as well as a brief exam ination of the various forecasting methods currently in use. Relevance to NCFRP 11: Moderate relevance to NCFRP 11. The authors identify direct and indirect factors that drive transportation demand. Freight Demand Factors Cited: Figure 1 shows de mand factors broken by direct and indirect factor s. Direct factors contribute directly to the dem and of goods and services. Obvious candidates for such factors are demographic trends (literally th e number of pe ople), the disposable incom e and wealth of the population, proportion the people spend on non-durable / durable goods and proportion of incom e consumed versus saved. Other m acro variables that m ight influence are trade and how important trade is to the economy. B-35

*** 19. Logistical Restructuring and Road F reight Traffic Grow th – An Empirical Assessment McKinnon, Alan and Allan Woodburn. Transportation, May 1996 pp 141-161. Forecasting of freight traffic has relied heavily on the close correlation between GDP and road tonne-Kms and not been rooted in the causes of freight traffic growth. This paper looks at these causes by focusing on the produc tion, consumption and movement of food and drink products in the UK and by conduc ting a survey of the c hanging freight transport requirements of 88 large British-based manufacturers. Relevance to NCFRP 11: Moderately relevant to NCFR P 11. This study under-takes a detailed examination of the drivers and cau ses of freight traffi c growth. Although it focuses on a particular sector of the econom y, the principles estab lished are nonetheless of general import. Freight Demand Factors Cited: Four key factors are identified (albeit with regards to the food industry in the UK) are i) value density, ii) handling factor (the ratio of tons lifted to actual we ight of products consumed or exported), iii) average length of haul iv) and consignment size. Value density, typically increases w ith time i.e. end products are m ore valuable but weigh less i.e. they are more highly processed and more “value added”. With regard to handling factor as suppl y chains becom e more complex and the number of intermediate processes incre ase, this factor becom es inflated and the amount of freight handled is bound to increase. Changes in this factor can be traced to increased weight loss in the production process, increase in the number of separate links in the supply chain and changes in the amount of packaging. Another factor that has contribu ted to the r ise in freight in this sector is the average length of haul. Average length of haul has in creased significantly and that has growth has been driven by i) the geographical conc entration of production and inventory ii) the expansion of market areas. Lastly consignment size has grown significantly. This has been driven in turn by two factors, the increase in the m aximum weights and dim ensions of lorries; and the consolidation of food and drink into bul k loads and for further channeling at distribution centers. B-36

Additionally a survey was conduc ted of a sa mple of firms in the UK and results of that survey produced what the main drivers of freight were a tightening of custom ers requirements driven by desire to cut th eir inventory levels and the second m ost important factor was the firms own desire to cut its inventory levels. *** 20. Spatial Price Competition and the Demand for Freight Transportation Inaba, Fred S. and Nancy E. W allace. The Review of Economics and Statistics, Vol. 71, No. 4, November 1989, pp. 614-625. Two important issues in econom etric demand analysis are addr essed: i) the sim ultaneity between quantity shipped and mode/destination choices and ii) th e effect of spatial price competition on the demand for the transportation factor. The paper develops a theoretical model as well as a pra ctical application that is applied to wheat shipments in the Pacific Northwest. Relevance to NCFRP 11: Broadly relevant to NCFRP1 1. Evidence suggests that shippers choice of mode and shipment size are not independent of each other rath er are made simultaneously. A theoretical model is developed wherein spatial price competition determines the firm’s market area and thereby determines its sales and shipment size. Freight Demand Factors Cited: Shipment size Firm’s market area and choice of mode Other factors that jointly effect m ode destination (as illus trated by the decisions of country elevators that ship wheat in the pacific northwest ) are, waiting cost (having negative effects) and the size of the market (increases as the size of the market increases). *** 21. NCHRP Report 388: A gui debook for Forecasting Freight Transportation Demand Cambridge Systematics Freight demand factors are di vided into two categories. Factors that effect dem and directly and those that effect demand indirectly. The former consist of obvious candidates like population while the latter consist of those factors that effect the costs of one more transport modes and in som e cases on the serv ices offered; these factors affect dem and indirectly as a result of changes in transport costs and rates and in the services offered. B-37

Relevance to NCFRP 11: Direct relevance to NCFRP 11. Freight Demand Factors Cited: Direct Factors: i) Level of economic activity: Level of economic activity can be measured with GDP, particularly goods production co mponent of GDP. Goods production and particularly durables goods producti on tends to be more cyclical and correlates well with the growth in freight demand (as measured by ton-miles) ii) Industrial Location Factors: The spatial distribution of manufacturing centers has a great influence on the demand for freight when measured in ton-miles or the distance that freight is transpor ted. Spatial distri bution of econom ic activity also has a major influence on the modes that are used. Plants located near specific modes tend to prefer that mode, for eg: factories located near rail tend to use rail m ore where as those located near major highways tend to use trucks more. iii) Globalization of Business: Globalization has resulte d in long supply chains that are spread all over the world. In general the longer the supply chain, the greater the demand for freight. This “supply chain sprawl” varies significantly from industry to industr y. Additionally these s upply chains are dynam ic constantly adapting to the changing market and price conditions. iv) International Trade Ag reements: International Trade Ag reements reduce barriers to trade such as reduction of tr ade barriers, bring about consistency in government regulations and generally pr omote trade and cross border freight traffic. v) Just in Time Inventory Practices: Just-in-Time (JIT) systems focus on keeping inventories at minimum levels through a coordination of input deliveries with production schedules. Effects on freight demand are to increase the number of individual shipments, decrease their length of haul, and most im portantly increase the importance of on-time delivery. B-38

vi) Central Warehousing: As transportation systems have become more efficient there has been a trend towards using fewer warehouses. Third party logistics operators that specialize in distribution ha ve helped this trend. This trend has increased transport demand and associated costs in order to save on inventory costs. The trend toward centralized wa rehousing results in increased transport demand (measured in ton-m iles, shipment miles or value of service) and in some instances a shift from truck to air delivery. Appropriate measures, might be the number of firm s using one or two warehouses and the value or tons shipped. vii) Recycling: Recycling has an important effect on O-D patterns and lengths of haul and modal usage of several commodities. Recycling plants are located near urban centers or near the markets for the products while processing plants that use virgin m aterials are usually lo cated near a m ajor source of supply of these materials and they commonly ship their products long distances to their markets. Indirect Factors: Indirect factors include: i) Deregulation: The different deregulation acts (Airline Deregulation Act, Motor Carrier Act, Stagger’s Rail Act and the Shipping Act) have all had important effects by allowing private ope rators to op erate freight links and allowed them to operate efficient and high quality service. ii) International Transportation Agreem ents: There are m any international agreements that limit or have recently allowed more competition to take place. For example in the air freight secto r, the so-called “open skies” agreements, where carriers could provide service to airports other than the m ajor gateway airports of New Yor k, Miami, and Los Angeles. (Data on air traffic freigh t movements are available by airport by the North Am erican Airport T raffic Report, published by the Airports Counc il International). For trucks for example, NAFTA allows carriers to fl ow freely between Canada and the US but not between Mexico and the US. iii) Intermodal Agreements: Intermodal Operating Agreements like that of APL that used double stack trains to serve most inland origins and destinations. All major containership companies were us ing such services and m any trucking companies were adapting their trucks to handle containers. Measures for these include the number of such agreem ents and the volum e of intermodal traffic (these are published by the Association of Am erican Railroads on a weekly and annual basis for movements on class I railroads). iv) Fuel Prices: Fuel constitutes a m oderately significant and r elatively volatile component of cost for all freigh t modes. Fuel accounts for 7.1% of total operating expenses for Class I railroads: feul, oil, lubr icants and coo lants accounts for about 13.5% of operating e xpenses for truck-load carriers and about 6% of operating expenses for LTL carriers and about 30-40% f or air carriers. Sources of Data in clude the Department of Energy that publishes rates for various end use categories B-39

v) Publicly Provided Infrastructure: Air, water, and tru ck carriers are all dependent on publicly provi ded infrastructure. FAA is responsible for US airports, US Army Corps of Engineers maintains the waterway channels and the FHWA (Federal Highway Adm inistration) and all three sys tems of transportation tend to be expanded m ore slowly than users would like resulting in congestion that increases travel tim es and operating costs. This leads to increased unreliability which is a particular problem for Just-in-Tim e inventory approaches. The quality of local inf rastructure and the deg ree of congestion also effect carrier choices of ports and airports. Measures of public infrastructure include physical characteristics such lane-miles of road, channel depths, lengths of runw ays etc or fina ncial characteristics such as cap ital, maintenance and operating expenditures. vi) User Charges, Taxes and Subsidies: User Charges are used to fund m ost publicly provided infrastructure in th e US including airports, ports and highway. The major exception is inland waterways which in some cases give subsidies to barge operators. Other taxes exist in the form of fuel taxes such as the federal tax on gasoline and highway diesel fuel. All th ese taxes ad d to costs of transportation and som e cases make some modes more attractive relative to the others particularly barge transport is subsidized relative to rail. vii) Environmental Policies and Restrictions: Environmental policies place particular restrictions on ports and waterways. Port expansion is limited by the Clean Water Act or th e “no net loss” of wetlands. The O il pollution act requires oil tankers that have single hulls to b e replaced with double hull tankers and tankers w ith single hulls are required to meet other criteria. With trucks, emissions controls and clean fuel req uirements have also ef fected costs. Air transport is subject to noise reduction acts. All these factors increase costs of transportation and do af fect the demand for freight. In add ition environmental policies influence the lo cations at which raw m aterials are produced and those at which industrial plants are located. viii) Safety Policies and Regulations: Safety policies increase some costs f or carriers but also reduce other liab ility costs and their effect tends n ot be significant in increasing costs. ix) Effects of Changes in Truc k Size and W eight Limits: Changes in truck size and weight lim its can significantly aff ect the cost of goods m ovement by truck. Changes in truck size can re sult in sh ifts in f reight to or f rom other modes most importantly rail. The T RB’s Truck Weight Study estimated that eliminating the 80,000 pound lim it on gros s weight would result in a 2.2 percent reduction in rail traffic. To the extent that changes in Truck size and weight limits cause increase in th e demand for all f reight will depend on whether the savings in c osts that result from these changes w ill be passed on to consumers. Additionally, changes in weight limits might cause total amount of freight shipped to increase (decreas e) to the extent that it en courages (discourages) the centralization of production facilities. x) Congestion: In many urban areas, increasing highway congestion is affecting the efficiency of reliable truck transport and the reliability required by just-in- time shipping. B-40

xi) Technological Advances: The use of com puters and telecom munication equipment has had an important effect on the freight industry. Air carriers and many leading companies have implemented sophisticated systems for tracking shipments, computers for sorting shipments and optimizing the use of aircraft, and freight trucks. All these factors have improved efficiency and im proved the level of service and reliability and hence indirectly promoted modern Just- in-Time delivery methods to operate. *** 23. Transport & Logistics in Regional Travel Demand Models – State of the Practice Halcrow Consulting Inc. This report investigates the state of the practice in m odeling the dem and for freight transportation, with a sp ecific focus on regi onal transport models. It e xamines current approaches to regional freight m odels and highlights the limitations in current practice. The paper proceeds to examine case studies of metropolitan areas that have tried to move closer to commodity-based m odeling on the regional level (including the techniques & data sources that they employ.) Also, it discus ses the specific difficulties associated with modeling in and around Marine Gateways, which is highly dynam ic and unpredictable (thus extremely difficult to m odel – very di fferent factors involved). In general, it is determined that m ore complex models (with a lot of different factors) m ay not be the most effective at representing the freight industry’s ability to change quickly. Relevance to NCFRP-11: Highly relevant. In addition to exploring a universe of key factors (see below) it also provides some useful points regarding the current limitations of regional freight models, particularly in terms of future forecasting. For instance: “The main limitation identified is the ability and efficacy of using current freight and truck models for future forecasting, particularly into the longer-term. It is considered that the relationships that are generated between truck trips and land-use and other economic data make sense in representing the current industry and business practices, but these relationships are highly unlikely to be applicable into the future, given the rapid changes over the recent past. Hence, the concern expressed is that the future year forecasts may follow the procedures properly, and would calibrate well with a projected version of the current situation, but the results are not credible as this is unlikely to reflect the future economy and freight practices. This is basically because the modelling practices typically employed simply cannot account in forecasts for the sometimes rapid and significant changes in the economic environment as well as freight transport technologies.….The general thrust is that the eponymous four-stage process employed for personal travel is not appropriate to capture the full nature of freight, particularly in forecasting. …There is therefore a general feeling within US MPOs that commodity-based modelling approach tied to economic activity is ultimately the way forward.” B-41

(Note: This type of modeling is currently more common at the State-level and above (long-haul), where relevant data is more likely to be available and disaggregated details less significant.) Freight Demand Factors Cited: In addition to listing key demand factors, this report - categorizes them as either related to (1) production, cons umption & trade (2) logistics & associated transportation service factors or (3) network & routing factors - describes the way in which they are typically incorporated in US transport models (as shown in the listing below) - outlines the potential modeling issues/challenges associated with each of the three broad categories (see table below) Production, consumption & Trade Factors 1) Economic activity • Only included in employment numbers, if these are used to generate truck trips 2) Commodity movements and trade patterns • Very limited in regional models. Some jurisdictions aiming to do so in future. Super-regional or statewide models more likely to include commodity modeling. 3) Changes to sources of raw materials and goods • May be incorporated into commodity modeling (where relevant) Logistics & Associated Transportation Services Factors 4) Origins and destinations of freight • Internal truck matrices likely to be developed with limited reference to actual freight origins and destinations, based on features of the origin/destination (such as employment). External movements often incorporate origin/destination information based on surveys or other models 5) Distribution systems • The subtlety of distribution systems and the way that locations of distribution centers and use of modes and tour planning are coordinated is not considered. Hence the scope for these to change is largely ignored 6) Logistics facility locations • Some attempts to include locations as special generators or freight activity nodes 7) Modes available and modes used • Regional modelling usually concentrates on trucks. Mode split is then not relevant. Is generally included in commodity modelling and wider area models 8) Operational practices • Empty moves and positioning tacitly included in internal trip generation, but not specifically allocated to the key facilities that generate the need for such practices 9) Ownership of transport provider • Relates to operational practices and distribution systems and how decisions are made for forecasting purposes 10) Costs of freight transport to the users • Incorporated into assignment through generalized costs for trucks. Does not encapsulate the overall cost of goods movement, especially in so much as this is part of a production methodology that may span the globe B-42

Network & Routing Factors 11) Network availability (road, railway and waterway) • Generally not included unless multi-modal assignment modeling is done. May be included in commodity modeling 12) Access to the networks – intermodal • Intermodal freight interchanges are specifically identified if they are considered important freight generators. This is unlikely to include direct reference to transhipment. 13) Network restrictions • Assignment process include restrictions for vehicle classes 14) Network instability • Linkage to overall traffic modeling results in congestion being modelled in assignment process Tour planning • Inferred in the generation of internal trips. Similarly to operational practices, there is no specific reference to origins and destinations and hence potential trip-chaining 15) Actual routes used by Trucks • See network restrictions B-43

B-44

B-45 To be included in bibliography (not in lit review) Behavioral insights in to the mod eling of freight transportation and distribution systems Hensher and Figliozzi, Transportation Research, 2007 Assessing Rail Freight Solutions to Roadway Congestion: Final Report Bryan, Joseph , Glen Weisbrod, and Carl Martland. 2006

 Table C-1 - Correlations of Candidate Demand Factors Absolute Correlation Matrix Rail Tons Rail Ton- Miles Rail Train- Miles Rail Car- Miles Rail Rev Ton-Miles Annual Truck Ton- Miles Truck VMT Water Tons Water Ton- Miles Real GDP 0.939 0.983 0.955 0.976 0.982 0.978 0.978 -0.010 -0.979 Real GDP per Capita 0.928 0.981 0.954 0.977 0.979 0.981 0.981 0.037 -0.971 Real Personal Consumption 0.938 0.977 0.943 0.966 0.976 0.969 0.969 -0.046 -0.980 Real Income Per Capita 0.925 0.971 0.934 0.961 0.969 0.971 0.971 -0.023 -0.975 Total Housing Starts 0.466 0.414 0.839 0.827 0.422 0.428 0.428 -0.307 -0.481 Industrial Production Index 0.950 0.988 0.965 0.984 0.983 0.977 0.977 0.014 -0.980 Industrial Manufacturing Index 0.951 0.987 0.966 0.985 0.983 0.975 0.975 0.003 -0.981 Purchasing Managers' Index 0.166 0.199 0.321 0.260 0.219 0.252 0.252 0.091 -0.137 Trade Wt. Broad Cur. Index 0.824 0.925 0.833 0.834 0.915 0.966 0.966 0.066 -0.902 Trade Wt. Major Cur. Index -0.498 -0.579 -0.001 0.014 -0.586 -0.555 -0.555 -0.519 0.425 Total Employment 0.910 0.984 0.966 0.984 0.979 0.993 0.993 0.119 -0.954 Employment in Wholesale Sector 0.862 0.944 0.873 0.899 0.936 0.961 0.961 0.210 -0.909 Exports in Real $ 0.922 0.962 0.895 0.899 0.962 0.925 0.925 0.135 -0.912 Imports in Real $ 0.958 0.956 0.936 0.955 0.960 0.919 0.919 -0.083 -0.967 Total Capacity Utilization 0.006 0.077 -0.302 -0.356 0.089 0.083 0.083 0.844 0.108 Chained Inv. Sales Ratio (BEA) -0.866 -0.890 -0.955 -0.937 -0.892 -0.890 -0.890 -0.033 0.871 Inv. Sales Ratio (Census) -0.963 -0.951 -0.957 -0.955 -0.956 -0.900 -0.900 -0.048 0.911 Urban Gas Price in Real $ 0.028 -0.197 0.516 0.549 -0.180 -0.354 -0.354 -0.544 0.092 Retail Sales in Real $ 0.945 0.981 0.968 0.986 0.979 0.972 0.972 -0.003 -0.977 Lagged Inland Waterway Trust Fund Tax/Gallon NA NA NA NA NA NA NA 0.117 -0.857 Grain Tonnage NA NA NA NA NA NA NA -0.075 -0.700 Coal + Grain Tonnage NA NA NA NA NA NA NA 0.149 -0.890 Coal Production (Tonnage) NA NA NA NA NA NA NA 0.262 -0.890 The five independent variables with the highest absolute correlations were highlighted. NA indicates correlations were not determined for rail or truck demand for the independent variables added especially to model waterborne traffic. Developed by the Research Team C‐1  Appendix C - Correlation Analysis

 C‐2  TABLE C-2 Filtered Correlations by Mode (Rail) in Actual, 1980 - 2007 Variables with which Correlations are > 0.75 Demand Measures (Dependent Variables) Income Production Foreign Exchange Employment Trade Other Variables with lower Correlations Rail Tons Real GDP, Real GDP/Cap., Real Pers. Cons., Real Inc./Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Emp., Emp. Wholesale Real Imports, Real Exports Inv.-Sales Ratio BEA, Inv.- Sales Ratio Census, Real Retail Sales Trade Wt. Major Index, Total House Starts, Purch. Mgr. Index, Real Gas Price, Total Cap. Util. Rail Ton-Miles Real GDP, Real GDP/Cap., Real Pers. Cons., Real Inc./Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Emp., Emp. Wholesale Real Imports, Real Exports Inv.-Sales Ratio BEA, Inv.- Sales Ratio Census, Real Retail Sales Trade Wt. Major Index, Total House Starts, Purch. Mgr. Index, Real Gas Price, Total Cap. Util. Rail Train-Miles Real GDP, Real GDP/Cap., Real Pers. Cons., Real Inc./Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Emp., Emp. Wholesale Real Imports, Real Exports Inv.-Sales Ratio BEA, Inv.- Sales Ratio Census, Total Housing Starts, Real Retail Sales Trade Wt. Major Index, Purch. Mgr. Index, Real Gas Price, Total Cap. Util. Rail Car-Miles Real GDP, Real GDP/Cap., Real Pers. Cons., Real Inc./Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Emp., Emp. Wholesale Real Imports, Real Exports Inv.-Sales Ratio BEA, Inv.- Sales Ratio Census, Total Housing Starts, Real Retail Sales Trade Wt. Major Index, Purch. Mgr. Index, Real Gas Price, Total Cap. Util. Rail Rev Ton Miles (Annual) Real GDP, Real GDP/Cap., Real Pers. Cons., Real Inc./Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Emp., Employment Wholesale Real Imports, Real Exports Inv.-Sales Ratio BEA, Inv.- Sales Ratio Census, Real Retail Sales Trade Wt. Major Index, Total Housing Starts, Purch. Mgr. Index, Real Gas Price, Total Cap. Util. SOURCE: Developed by the Research Team

 C‐3  TABLE C-3 Filtered Correlations by Mode (Truck) in Actual, 1980–2007 Variables with which Correlations are > 0.75 Demand Measures (Dependent Variables) Income Production Foreign Exchange Employment Trade Other Variables with which Correlations are < 0.75 Truck Ton Miles Real GDP, Real GDP/Capita, Real Pers. Cons., Real Income/Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Employment, Employment Wholesale Real Exports, Real Imports Inv.- Sales Ratio BEA, Inv.- Sales Ratio Census, Real Retail Sales Total Housing Starts, Purch. Mgr. Index, Trade Wt. Major Index, Total Cap. Util., Real Gas Price Truck VMT Real GDP, Real GDP/Capita, Real Pers. Cons., Real Income/Capita Indus. Prod. Index, Indus. Mfg. Index Trade Wt. Brd. Index Total Employment, Emp. Wholesale Real Exports, Real Imports Inv.- Sales Ratio BEA, Inv.- Sales Ratio Census, Real Retail Sales Total Housing Starts, Purch. Mgr. Index, Trade Wt. Major Index, Total Cap. Util., Real Gas Price SOURCE: Developed by the Research Team

 C‐4  TABLE C-4 Rail Tonnage Sub-Sample Correlation Ranks (in actual) Rank Full Sample Rank Short- Term Sample 1 Rank Short- Term Sample 2 Rank Short- Term Sample 3 Rail Tons (Correlation Ranks in Actual) 1980- 2007 t-stat 1980- 1985 t-stat 1986- 1996 t-stat 1997 - 2007 t-stat Real GDP 6 11.3 13 -0.3 9 5.4 3 23.7 Real GDP per Capita 8 11.0 17 -0.1 7 6.1 2 25.8 Real Personal Consumption 7 11.2 8 -0.7 10 5.3 4 17.8 Real Income per Capita 9 10.9 9 -0.6 8 5.7 5 14.7 Total Housing Starts 16 2.7 14 -0.2 18 -0.3 15 1.1 Industrial Production Index 3 13.3 10 0.5 3 9.0 8 10.6 Industrial Manufacturing Index 4 12.9 11 0.4 4 8.9 6 11.8 Purchasing Managers' Index 17 1.0 18 0.1 19 -0.2 19 0.4 Trade Wt. Broad Cur. Index 14 6.1 6 -1.2 13 3.7 18 -0.7 Trade Wt. Major Cur. Index 15 -3.0 5 -1.5 14 -3.1 12 -2.9 Total Employment 11 10.0 19 0.0 6 6.7 9 8.2 Employment in Wholesale Sector 13 8.2 16 -0.1 5 7.2 16 1.1 Exports in Real $ 10 10.7 3 2.4 11 4.9 13 2.4 Imports in Real $ 2 16.8 4 1.6 1 9.7 7 11.8 Total Capacity Utilization 19 0.0 2 3.4 15 3.1 17 -1.0 Chained Inv.-Sales Ratio (BEA) 12 -9.1 12 -0.4 16 -1.8 14 -2.3 Inv.-Sales Ratio (Census) 1 -18.8 1 -3.8 12 -4.2 10 -6.6 Urban Gas Price in Real $ 18 0.1 7 1.2 17 -1.4 11 5.9 Retail Sales in Real $ 5 12.6 15 0.1 2 9.4 1 27.7 SOURCE: Developed by the Research Team

 C‐5  TABLE C-5 Rail Ton-Miles Sub-sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Short-Term Sample 3 Rail Ton-Miles (Correlation Ranks in Actual) 1980– 2007 t-stat 1980– 1985 t-stat 1986– 1996 t-stat 1997– 2007 t-stat Real GDP 1 17.4 18 0.1 6 18.2 1 17.4 Real GDP per Capita 2 16.7 13 0.4 4 18.9 2 16.7 Real Personal Consumption 4 15.2 15 -0.3 5 18.2 4 15.2 Real Income per Capita 5 13.6 17 -0.2 7 16.9 5 13.6 Total Housing Starts 17 0.9 19 0.1 18 -0.8 17 0.9 Industrial Production Index 9 8.5 6 1.0 1 51.3 9 8.5 Industrial Manufacturing Index 7 9.2 8 0.8 2 44.7 7 9.2 Purchasing Managers' Index 19 0.2 16 0.3 19 -0.3 19 0.2 Trade Wt. Broad Cur. Index 18 -0.7 9 -0.8 11 7.9 18 -0.7 Trade Wt. Major Cur. Index 12 -2.7 5 -1.0 14 -4.0 12 -2.7 Total Employment 8 8.6 12 0.5 3 19.1 8 8.6 Employment in Wholesale Sector 15 1.1 14 0.3 13 5.3 15 1.1 Exports in Real $ 13 2.6 4 1.7 10 9.8 13 2.6 Imports in Real $ 6 10.7 3 2.1 8 13.1 6 10.7 Total Capacity Utilization 16 -1.0 2 5.5 15 2.3 16 -1.0 Chained Inv.-Sales Ratio (BEA) 14 -2.1 10 -0.7 17 -1.4 14 -2.1 Inv.-Sales Ratio (Census) 11 -5.7 1 -5.9 12 -5.8 11 -5.7 Urban Gas Price in Real $ 10 6.5 7 0.8 16 -1.7 10 6.5 Retail Sales in Real $ 3 16.2 11 0.5 9 10.8 3 16.2 SOURCE: Developed by the Research Team

 C‐6  TABLE C-7 Rail Train-Miles Sub-Sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Rail Train Miles (Correlation Ranks in Actual) 1990– 2007 t-stat 1990– 1996 t-stat 1997– 2007 t-stat Real GDP 6 15.2 2 15.9 4 10.3 Real GDP per Capita 7 13.9 6 13.6 3 10.4 Real Personal Consumption 8 13.5 4 15.0 5 9.2 Real Income per Capita 11 11.3 9 10.3 6 8.2 Total Housing Starts 15 6.6 14 4.2 15 1.4 Industrial Production Index 4 16.1 5 14.0 10 6.7 Industrial Manufacturing Index 2 16.7 3 15.2 9 7.3 Purchasing Managers' Index 17 1.4 19 1.0 18 0.8 Trade Wt. Broad Cur. Index 14 6.7 8 10.6 16 -1.1 Trade Wt. Major Cur. Index 19 -0.1 18 -1.6 12 -3.6 Total Employment 5 15.6 10 10.2 11 5.4 Employment in Wholesale Sector 13 7.3 17 2.7 17 0.9 Exports in Real $ 12 9.5 11 8.3 14 2.6 Imports in Real $ 3 16.7 1 17.6 1 12.3 Total Capacity Utilization 18 -1.3 15 3.3 19 -0.7 Chained Inv.-Sales Ratio (BEA) 9 -12.8 12 -6.0 13 -2.9 Inv.-Sales Ratio (Census) 10 -12.2 13 -5.6 8 -7.8 Urban Gas Price in Real $ 16 2.3 16 -2.8 7 7.9 Retail Sales in Real $ 1 17.8 7 10.8 2 11.3 SOURCE: Developed by the Research Team

 C‐7  TABLE C-8 Rail Car-Miles Sub-Sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Rail Car Miles (Correlation Ranks in Actual) 1990– 2007 t-stat 1990– 1996 t-stat 1997– 2007 t-stat Real GDP 5 22.4 9 9.5 9 16.5 Real GDP per Capita 7 20.7 5 12.9 5 17.5 Real Personal Consumption 8 18.7 10 8.4 10 13.3 Real Income per Capita 9 15.5 7 11.7 7 12.0 Total Housing Starts 15 6.2 15 3.2 15 1.2 Industrial Production Index 3 22.8 2 23.4 2 9.2 Industrial Manufacturing Index 2 24.0 3 19.5 3 10.2 Purchasing Managers' Index 18 1.1 19 0.4 19 0.6 Trade Wt. Broad Cur. Index 14 6.5 11 5.8 11 -0.9 Trade Wt. Major Cur. Index 19 0.0 18 -1.6 18 -3.3 Total Employment 6 22.3 1 32.2 1 6.8 Employment in Wholesale Sector 13 8.3 12 4.1 12 1.0 Exports in Real $ 12 9.5 6 11.7 6 2.5 Imports in Real $ 4 22.8 4 16.4 4 13.6 Total Capacity Utilization 17 -1.5 16 3.2 16 -0.8 Chained Inv.-Sales Ratio (BEA) 11 -10.9 13 -3.7 13 -2.6 Inv.-Sales Ratio (Census) 10 -12.2 14 -3.5 14 -7.5 Urban Gas Price in Real $ 16 2.5 17 -1.9 17 6.5 Retail Sales in Real $ 1 29.7 8 11.5 8 18.9 SOURCE: Developed by the Research Team

 C‐8  TABLE C-9 Truck Vehicle-Miles Sub-Sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Short-Term Sample 3 Truck VMT (Correlation Ranks in Actual) 1980– 2007 t-stat 1980– 1985 t-stat 1986– 1996 t-stat 1997– 2007 t-stat Real GDP 2 39.6 5 6.1 2 30.0 3 15.7 Real GDP per Capita 6 32.4 7 4.3 5 14.9 4 13.5 Real Personal Consumption 5 33.5 1 12.0 1 38.3 1 20.9 Real Income per Capita 8 26.0 2 11.3 3 15.9 2 20.1 Total Housing Starts 16 2.5 9 3.3 18 -1.0 16 1.3 Industrial Production Index 4 33.7 14 2.4 6 14.6 8 6.3 Industrial Manufacturing Index 3 35.2 13 3.0 7 14.5 6 6.5 Purchasing Managers' Index 18 1.4 16 1.4 19 -0.3 19 0.3 Trade Wt. Broad Cur. Index 11 17.7 4 8.9 8 13.1 18 -0.3 Trade Wt. Major Cur. Index 15 -3.4 6 5.2 14 -3.9 12 -2.3 Total Employment 1 43.7 12 3.0 4 15.0 10 5.8 Employment in Wholesale Sector 10 17.8 11 3.0 13 4.1 17 0.4 Exports in Real $ 12 12.6 10 -3.2 9 9.8 15 1.6 Imports in Real $ 9 19.0 17 1.1 10 8.6 7 6.3 Total Capacity Utilization 19 0.4 19 0.2 15 1.7 14 -1.8 Chained Inv.-Sales Ratio (BEA) 14 -10.1 15 -1.9 17 -1.2 13 -2.0 Inv.-Sales Ratio (Census) 13 -10.4 18 0.4 12 -5.7 9 -6.0 Urban Gas Price in Real $ 17 -1.9 3 -9.9 16 -1.7 11 4.3 Retail Sales in U.S. $ 7 27.3 8 3.7 11 7.4 5 12.7 SOURCE: Developed by the Research Team

 C‐9  TABLE C-10 Truck Ton-Miles Sub-Sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Short-Term Sample 3 Truck Ton Miles (Correlation Ranks in Actual) 1980– 2007 t-stat 1980– 1985 t-stat 1986– 1996 t-stat 1997– 2007 t-stat Real GDP 2 39.7 5 6.1 2 30.0 3 15.6 Real GDP per Capita 6 32.5 7 4.3 5 14.9 4 13.5 Real Personal Consumption 5 33.5 1 12.0 1 38.2 1 20.7 Real Income per Capita 8 26.0 2 11.3 3 15.9 2 20.1 Total Housing Starts 16 2.5 9 3.3 18 -1.0 16 1.3 Industrial Production Index 4 33.7 14 2.4 6 14.6 8 6.3 Industrial Manufacturing Index 3 35.2 13 3.0 7 14.5 6 6.5 Purchasing Managers' Index 18 1.4 16 1.4 19 -0.3 18 0.3 Trade Wt. Broad Cur. Index 11 17.7 4 8.9 8 13.1 19 -0.3 Trade Wt. Major Cur. Index 15 -3.4 6 5.2 14 -3.9 12 -2.3 Total Employment 1 43.7 12 3.0 4 15.0 10 5.8 Employment in Wholesale Sector 10 17.8 11 3.0 13 4.1 17 0.4 Exports in Real $ 12 12.6 10 -3.2 9 9.8 15 1.6 Imports in Real $ 9 19.0 17 1.1 10 8.6 7 6.3 Total Capacity Utilization 19 0.4 19 0.2 15 1.7 14 -1.8 Chained Inv.-Sales Ratio (BEA) 14 -10.1 15 -1.9 17 -1.2 13 -2.0 Inv.-Sales Ratio (Census) 13 -10.4 18 0.4 12 -5.7 9 -6.0 Urban Gas Price in Real $ 17 -1.9 3 -9.9 16 -1.7 11 4.3 Retail Sales in Real $ 7 27.3 8 3.7 11 7.4 5 12.7 SOURCE: Developed by the Research Team

 C‐10  TABLE C-11 Waterborne Tonnage Sub-Sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Short-Term Sample 3 Truck VMT (Correlation Ranks in Actual) 1980– 2007 t-stat 1980– 1985 t-stat 1986– 1996 t-stat 1997– 2007 t-stat Real GDP 15 0.36 22 ‐0.02 11 1.34  3  ‐5.03 Real GDP per Capita 12 0.49 19 0.26 5 1.72  5  ‐4.86 Real Personal Consumption 19 0.20 14 ‐0.40 14 1.25  2  ‐5.60 Real Income per Capita 20 0.14 17 ‐0.32 7 1.54  1  ‐5.63 Total Housing Starts 4 ‐1.41 20 ‐0.17 15 ‐1.23  13  ‐1.21 Industrial Production Index 14 0.42 9 0.87 8 1.53  8  ‐3.66 Industrial Manufacturing Index 13 0.42 10 0.68 9 1.46  7  ‐3.67 Purchasing Managers' Index 10 0.63 21 ‐0.15 20 ‐0.28  20  0.39 Trade Wt. Broad Cur. Index 11 0.59 8 ‐0.89 17 0.63  18  ‐0.56 Trade Wt. Major Cur. Index 3 ‐2.93 6 ‐1.18 2 ‐2.76  15  1.03 Total Employment 8 0.84 13 0.45 6 1.71  6  ‐3.99 Employment in Wholesale Sector 5 1.22 16 0.33 3 2.49  21  ‐0.11 Exports in Real $ 6 1.15 4 2.04 4 1.89  17  ‐0.61 Imports in Real $ 18 0.21 3 2.16 12 1.31  10  ‐3.04 Total Capacity Utilization 1 8.00 1 5.15 1 3.61  9  3.08 Chained Inv.-Sales Ratio (BEA) 21 ‐0.11 18 ‐0.31 22 ‐0.09  14  1.10 Inv.-Sales Ratio (Census) 22 ‐0.08 2 ‐3.35 16 ‐0.93  11  3.01 Urban Gas Price in Real $ 2 ‐3.50 7 0.93 19 0.42  12  ‐2.27 Retail Sales in U.S. $ 16 0.30 15 0.40 10 1.46  4  ‐4.87 Grain & Coal Tonnage 7 0.94 11 0.61 13 1.26  16  ‐0.84 (Lagged) Inland Waterway Trust Fund Fuel Tax Rate 9 0.70 5 ‐1.29 18 0.53  22  0.00 Grain Tonnage 17 ‐0.25 12 0.54 21 ‐0.13  19  ‐0.46 SOURCE: Developed by the Research Team

 C‐11  TABLE C-12 Waterborne Ton-Miles Sub-Sample Correlation Ranks (in actual) Full Sample Short-Term Sample 1 Short-Term Sample 2 Short-Term Sample 3 Truck Ton Miles (Correlation Ranks in Actual) 1980– 2007 t-stat 1980– 1985 t-stat 1986– 1996 t-stat 1997– 2007 t-stat Real GDP 8 ‐17.41 10 ‐1.18 2 ‐4.94  3 ‐11.44 Real GDP per Capita 9 ‐16.27 14 ‐0.98 1 ‐5.09  5 ‐10.63 Real Personal Consumption 5 ‐18.09 8 ‐1.40 5 ‐4.84  2 ‐11.78 Real Income per Capita 6 ‐18.07 5 ‐1.68 6 ‐4.84  4 ‐11.22 Total Housing Starts 18 ‐2.76 18 ‐0.52 22 0.69  19 ‐0.59 Industrial Production Index 2 ‐18.90 17 ‐0.69 7 ‐4.34  9 ‐7.25 Industrial Manufacturing Index 3 ‐18.54 15 ‐0.73 9 ‐4.30  8 ‐7.52 Purchasing Managers' Index 20 ‐0.79 21 0.28 19 1.12  21 0.25 Trade Wt. Broad Cur. Index 15 ‐8.80 4 ‐1.81 3 ‐4.93  20 0.48 Trade Wt. Major Cur. Index 19 2.45 1 ‐1.99 16 1.71  13 2.24 Total Employment 10 ‐13.75 13 ‐1.03 4 ‐4.89  7 ‐8.31 Employment in Wholesale Sector 12 ‐10.51 9 ‐1.26 13 ‐3.31  18 ‐1.11 Exports in Real $ 13 ‐9.77 6 1.62 11 ‐4.11  14 ‐2.23 Imports in Real $ 1 ‐19.64 20 ‐0.29 12 ‐3.90  10 ‐6.69 Total Capacity Utilization 21 0.55 19 0.43 17 ‐1.30  17 1.32 Chained Inv.-Sales Ratio (BEA) 14 9.34 22 ‐0.07 21 0.72  15 1.62 Inv.-Sales Ratio (Census) 11 11.79 3 ‐1.85 15 2.63  12 4.46 Urban Gas Price in Real $ 22 0.46 2 1.91 20 0.84  11 ‐4.83 Retail Sales in Real $ 4 ‐18.30 16 ‐0.72 10 ‐4.29  6 ‐9.79 Rail Ton-Miles (substitute service) 7 ‐17.59 11 1.18 8 ‐4.31  1 ‐12.02 (Lagged) Inland Waterway Trust Fund Fuel Tax Rate 16 ‐7.20 7 ‐1.44 14 ‐3.11  NA  NA  Grain Tonnage 17 ‐4.58 12 ‐1.14 18 ‐1.23  16 ‐1.45 SOURCE: Developed by the Research Team

TABLE D-1 Rail Ton Log-Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 5 6 7 Period of Estimation 1981-07 1981-07 1981-07 1981-07 1981-07 1981-07 1981-07 R - Squared (adjusted) 0.97 0.97 0.97 0.97 0.97 0.98 0.96 S.E. of Regression 0.02 0.03 0.02 0.02 0.02 0.02 0.03 Durbin Watson Statistic 2.38 2.01 2.16 2.16 2.11 2.11 2.36 Model Coefficients Constant 11.20 8.29 8.36 5.38 9.99 8.20 2.23 t-stat 28.3 23.2 24.5 4.2 15.4 7.8 1.4 Candidate Demand Factor Weights GDP in Real $ 0.55 t-stat 10.0 Real Personal Consumption 0.67 t-stat 7.4 Industrial Production Index 0.84 t-stat 7.7 Total Capacity Utilization 0.67 t-stat 2.8 Total Trade in Real $ 0.36 0.27 t-stat 12.0 6.2 Exports in Real $ 0.15 t-stat 1.5 Lagged Housing Total 0.10 0.12 0.1 t-stat 2.2 1.8 1.7 Real Income Per Capita 0.58 t-stat 5.2 Inv. Sales Ratio (Census) -0.45 -0.71 t-stat -3.2 -4.5 Retail Sales in Real $ 0.66 t-stat 4.3 Lagged Purchasing Managers' Index 0.1 t-stat 1.7 Trade Wt. Broad Cur. Index -0.14 t-stat -1.7 Exogenous Impact Controls Lagged NAFTA Impact 0.04 0.03 0.04 0.04 t-stat 2.1 2.6 5.6 3.9 Rail Deregulation -0.07 -0.05 -0.07 -0.09 t-stat -2.9 -2.0 -6.2 -3.6 Statistical Error Corrections AR (1) 0.78 0.43 0.49 0.82 0.49 0.61 0.81 t-stat 7.2 3.5 4.2 10.7 3.6 4.6 7.6 SOURCE: Developed by the Research Team D‐1  Appendix D - Regression Analysis & Diagnostics

TABLE D-2 Rail Ton Log-Actual Diagnostics Rail Ton Diagnostics Tests 1 2 3 4 5 6 7 Number of Observations 28 27 27 28 27 27 28 AIC (Information Criteria) -116.79 -104.26 -87.90 -91.10 -109.72 -100.32 -99.35 BIC (Information Criteria) -110.13 -100.37 -82.72 -84.44 -104.53 -93.84 -92.69 Structural Model R - Sq 0.86 0.92 0.74 0.80 0.93 0.93 0.84 DW stat (AR corrected) 2.38 2.01 2.22 2.15 2.11 2.04 1.72 DW stat (original) 1.37 0.87 1.53 1.53 0.89 1.10 1.91 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.08 0.04 0.00 0.00 0.11 0.10 0.00 White Test (p-value) 0.44 0.21 0.12 0.37 0.40 0.19 0.11 Variance Inflation Factors (Regressors) GDP in Real $ 1.4 Real Personal Consumption 1.8 Real Income Per Capita 9.0 Industrial Production Index 9.0 Total Capacity Utilization 1.5 Total Trade in Real $ 1.4 11.4 Exports in Real $ 7.2 Inv. Sales Ratio (Census) 12.8 8.8 Lagged House Total 1.38 1.4 1.6 Lagged Purchasing Managers' Index 1.1 Retail Sales in Real $ 9.3 Trade Wt. Broad Cur. Index 12.2 Lagged NAFTA Impact 1.0 1.0 1.0 1.2 Rail Deregulation 2.3 2.2 2.0 D‐2 

TABLE D-3 Rail Ton–Miles Log-Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 5 Period of Estimation 1981-07 1981-07 1981-07 1981-07 1981-07 R - Squared (adjusted) 0.97 0.99 0.99 0.99 0.99 S.E. of Regression 0.03 0.02 0.03 0.03 0.02 Durbin Watson Statistic 2.19 1.99 2.33 2.09 1.98 Model Coefficients Constant 3.81 9.98 8.84 -4.53 -4.34 t-stat 6.2 29.0 4.4 -1.5 -3.4 Candidate Demand Factor Weights GDP in Real $ 1.12 t-stat 16.9 Industrial Production Index 0.96 t-stat 13.8 Total Employment 1.31 t-stat 3.4 Total Trade in Real $ 0.38 t-stat 3.2 Imports in Real $ 0.24 t-stat 2.1 Exports in Real $ 0.28 t-stat 3.5 Inv. Sales Ratio (Census) -0.47 -0.52 t-stat -3.6 -3.2 Retail Sales in Real $ 0.97 t-stat 6.8 Exogenous Impact Controls Lagged NAFTA Impact 0.05 0.06 t-stat 4.1 7.2 Statistical Error Corrections AR (1) 0.64 0.56 0.94 0.65 0.55 t-stat 4.3 3.7 13.0 4.4 2.9 SOURCE: Developed by the Research Team D‐3 

TABLE D-4 Rail Ton-Miles Log-Actual Diagnostics Rail Ton Miles Log – Actual Diagnostics Model 1 2 3 4 5 Number of Observations 28 28 28 28 28 AIC (Information Criteria) -84.68 -118.38 -109.90 -97.42 -116.17 BIC (Information Criteria) -80.68 -113.06 -105.90 -93.43 -112.18 Structural Model R - Sq 0.95 0.97 0.70 0.94 0.97 DW stat (AR corrected) 2.19 1.99 2.33 2.09 1.98 DW stat (original) 0.45 0.98 1.08 0.63 0.92 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.00 0.19 0.00 0.04 0.92 White Test (p-value) 0.03 0.40 0.14 0.04 0.47 Variance Inflation Factors (Regressors) GDP in Real $ 1.0 Total Employment 16.7 Industrial Production Index 7.0 Total Trade in Real $ 10.0 Imports in Real $ 16.7 Exports in Real $ 5.6 Inv. Sales Ratio (Census) 6.9 10.0 Retail Sales in Real $ 5.6 Lagged NAFTA Impact 1.0 1.0 SOURCE: Developed by the Research Team D‐4 

TABLE D-5 Rail Train Miles—Log Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 5 Period of Estimation 1990-07 1990-07 1990-07 1990-07 1990-07 R - Squared (adjusted) 0.97 0.98 0.98 0.99 0.97 S.E. of Regression 0.02 0.02 0.02 0.01 0.02 Durbin Watson Statistic 2.22 1.27 1.90 2.08 1.56 Model Coefficients Constant 0.37 2.48 1.80 1.80 -5.15 t-stat 0.2 1.8 10.4 6.7 -3.3 Candidate Demand Factors GDP in Real $ 0.57 t-stat 7.1 Industrial Production Index 0.56 t-stat 9.9 Total Capacity Utilization 0.87 t-stat 7.2 Total Trade in Real $ 0.25 t-stat 5.5 Lagged Real BLS Gas 0.10 0.27 t-stat 5.5 12.1 House Total 0.18 0.17 0.09 t-stat 4.4 4.8 4.0 Chained Inv. Sales Ratio (BEA) -0.79 t-stat -2.5 Lagged Trade Wt. Broad Cur. Index 0.13 0.11 0.50 t-stat 2.4 1.7 14.1 Lagged Purchasing Managers' Index 0.18 0.11 0.08 t-stat 3.4 1.7 2.1 Retail Sales in Real $ 0.78 t-stat 17.4 Exogenous Impact Controls Lagged NAFTA Impact 0.05 0.06 0.03 0.05 t-stat 3.2 6.4 2.3 5.5 SOURCE: Developed by the Research Team D‐5 

TABLE D-6 Rail Train-Miles Log-Actual Diagnostics Rail Train Miles Log – Actual Diagnostics Model 1 2 3 4 5 Number of Observations 18 18 18 18 18 AIC (Information Criteria) -82.35 -89.55 -88.07 -97.55 -85.47 BIC (Information Criteria) -77.89 -85.10 -83.62 -92.21 -81.02 Structural Model R - Sq 0.98 0.95 0.98 0.99 0.98 DW stat (AR corrected) 2.22 1.72 1.90 2.09 1.56 DW stat (original) 2.22 1.27 1.90 2.09 1.56 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.41 0.91 0.14 0.34 0.09 White Test (p-value) 0.49 0.25 0.39 0.43 0.95 Variance Inflation Factors (Regressors) GDP in Real $ 6.95 Industrial Production Index 4.03 Lagged Purchasing Managers' Index 1.32 1.41 1.27 Total Capacity Utilization 1.65 Lagged House Total 3.65 3.82 3.18 Total Trade in Real $ 6.87 Lagged Real BLS Gas 1.37 1.08 Inv. Sales Ratio (Census) 8.20 Retail Sales in Real $ 3.24 Lagged Trade Wt. Broad Cur. Index 6.04 5.35 4.51 Lagged NAFTA Impact 6.95 1.08 1.22 1.13 SOURCE: Developed by the Research Team D‐6 

TABLE D-7 Rail Car Miles—Log Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 5 Period of Estimation 1990-07 1990-07 1990-07 1981-07 1981-07 R - Squared (adjusted) 0.99 0.98 0.99 0.99 0.98 S.E. of Regression 0.01 0.02 0.01 0.01 0.02 Durbin Watson Statistic 2.35 2.21 1.98 1.74 1.19 Model Coefficients Constant 3.47 4.55 7.75 -4.03 2.47 t-stat 5.1 3.0 35.3 -4.8 0.7 Candidate Demand Factors Weights GDP in Real $ 0.66 t-stat 6.6 Real Personal Consumption 0.33 t-stat 1.6 Industrial Manufact. Index 0.47 t-stat 18.5 Total Trade in Real $ 0.14 t-stat 1.9 Lagged House Total 0.1 0.1 0.1 t-stat 2.1 1.7 2.9 Real Income Per Capita 0.79 t-stat 2.5 Inv. Sales Ratio (Census) -0.49 -0.48 t-stat -6.4 -1.4 Retail Sales in Real $ 0.95 t-stat 17.1 Exogenous Impact Controls Lagged NAFTA Impact 0.04 0.04 0.03 0.05 t-stat 4.3 8.5 4.7 3.3 Statistical Error Corrections AR (1) 0.66 0.71 -0.44 0.54 0.81 t-stat 3.3 2.0 -1.4 2.2 3.2 SOURCE: Developed by the Research Team D‐7 

TABLE D-8 Rail Car Miles—Log Actual Diagnostics Rail Car Miles Log – Actual Diagnostics Model 1 2 3 4 5 Number of Observations 18 18 18 18 18 AIC (Information Criteria) -95.97 -80.18 -116.86 -103.49 -77.43 BIC (Information Criteria) -92.41 -76.62 -112.41 -100.82 -73.87 Structural Model R - Sq 0.95 0.86 1.00 0.98 0.76 DW stat (AR corrected) 2.34 2.21 1.96 1.74 1.20 DW stat (original) 0.89 0.70 2.76 1.06 0.59 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.94 0.55 0.95 0.91 0.80 White Test (p-value) 0.83 0.33 0.51 0.55 0.29 Variance Inflation Factors (Regressors) GDP in Real $ 6.70 Real Personal Consumption 17.88 Real Income Per Capita 11.53 Industrial Production Index 8.93 Total Trade in Real $ 19.13 Lagged Total House 6.49 6.89 6.52 Inv. Sales Ratio (Census) 7.95 11.16 Retail Sales in Real $ 1.04 Lagged NAFTA Impact 1.08 1.06 1.04 1.11 SOURCE: Developed by the Research Team D‐8 

TABLE D-9 Rail Revenue Ton–Miles (Annual)—Log Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 5 6 Period of Estimation 1981-07 1981-07 1981-07 1981-07 1981-07 1981-07 R - Squared (adjusted) 0.99 0.99 0.99 0.99 0.99 0.99 S.E. of Regression 0.03 0.02 0.03 0.03 0.03 0.03 Durbin Watson Statistic 2.22 2.11 1.86 1.96 1.96 1.89 Model Coefficients Constant 11.23 17.39 11.27 9.57 3.26 -4.55 t-stat 20.3 41.6 20.7 6.1 2.8 -1.9 Candidate Demand Factors GDP in Real $ 1.06 t-stat 17.8 Real Personal Consumption 0.28 t-stat 1.7 Industrial Production Index 0.86 t-stat 10.4 Total Employment 2.12 t-stat 10.6 Total Trade in Real $ 0.50 t-stat 3.9 Exports in Real $ 0.27 t-stat 3.2 Real Income Per Capita 1.15 t-stat 7.8 Inv. Sales Ratio (Census) -0.64 -0.91 t-stat -3.9 -4.3 Retail Sales in Real $ 0.93 t-stat 6.7 Purchasing Manager Index 0.18 t-stat 4.5 Exogenous Impact Controls Lagged NAFTA Impact 0.06 0.07 0.07 0.04 0.07 t-stat 5.3 9.8 3.6 7.0 7.8 Statistical Error Corrections AR (1) 0.59 0.58 0.44 0.67 0.53 0.78 t-stat 3.6 3.7 2.4 4.7 2.9 8.1 SOURCE: Developed by the Research Team D‐9 

TABLE D-10 Rail Revenue Ton–Miles (Annual)—Log Actual Diagnostics Rail Revenue Ton Miles Log - Actual*Diagnostics Model 1 2 3 4 5 6 Number of Observations 28 28 28 28 28 28 AIC (Information Criteria) -85.95 -123.79 -115.00 -101.20 -109.84 -71.78 BIC (Information Criteria) -81.95 -118.46 -111.00 -95.88 -104.51 -66.45 Structural Model R - Sq 0.94 0.97 0.97 0.93 0.96 0.89 DW stat (AR corrected) 2.22 2.11 1.86 1.97 1.96 1.89 DW stat (original) 0.56 1.28 1.16 0.85 0.88 0.34 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.00 0.08 0.42 0.23 0.29 0.15 White Test (p-value) 0.03 0.20 0.83 0.72 0.78 0.03 Variance Inflation Factors (Regressors) GDP in Real $ 1.0 Real Personal Consumption 18.6 Real Income Per Capita 5.9 Total Employment 1.1 Industrial Production Index 7.0 Purchasing Managers Index 1.1 Total Trade in Real $ 18.6 Exports in Real $ 6.1 Inv. Sales Ratio (Census) 6.9 5.9 Retail Sales in Real $ 5.9 Lagged NAFTA Impact 1.0 1.0 1.0 1.1 1.0 SOURCE: Developed by the Research Team D‐10 

TABLE D-11 Truck Ton–Miles—Log Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 Period of Estimation 1981-07 1981-07 1981-07 1981-07 R - Squared (adjusted) 0.997 0.997 0.997 0.998 S.E. of Regression 0.01 0.01 0.01 0.01 Durbin Watson Statistic 1.65 1.77 1.57 1.63 Model Coefficients Constant 13.82 10.34 11.19 14.21 t-stat 17.3 6.6 5.5 10.7 Candidate Demand Factors Industrial Manufacturing Index 0.23 t-stat 4.4 Total Trade in Real $ 0.10 t-stat 2.3 Exports in Real $ 0.05 t-stat 1.5 Total Employment 0.37 t-stat 2.6 Chained Inv. Sales Ratio (BEA) -0.17 -0.22 t-stat -2.0 -2.7 Retail Sales in Real $ 0.26 t-stat 2.9 Gas Price in Real $ -0.05 -0.05 -0.03 -0.04 t-stat -1.3 -1.4 -1.1 -1.4 Exogenous Impact Control NAFTA Impact 0.01 0.02 t-stat 1.5 1.5 Statistical Error Corrections AR (1) 0.98 0.99 0.98 t-stat 74.8 57.4 55.3 SOURCE: Developed by the Research Team D‐11 

TABLE D-12 Truck Ton–Miles—Log Actual Diagnostics Truck Ton Miles Log - Actual* Diagnostics Model 1 2 3 4 Number of Observations 28 28 28 28 AIC (Information Criteria) -95.93 -111.13 -114.88 -114.35 BIC (Information Criteria) -90.60 -105.80 -108.22 -109.02 Structural Model R - Sq 0.34 0.36 0.42 0.44 DW stat (AR corrected) 1.65 1.77 1.57 1.62 DW stat (original) 0.89 0.71 0.83 0.89 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.09 0.05 0.28 0.67 White Test (p-value) 0.27 0.12 0.41 0.39 Variance Inflation Factors (Regressors) Total Employment 4.9 Industrial Manufacturing Index 1.1 Total Trade in Real $ 4.6 Exports in Real $ 5.7 Inv. Sales Ratio (BEA) 4.8 4.6 Retail Sales in Real $ 5.9 Real Gas Price 1.1 1.1 1.3 1.2 NAFTA Impact 1.1 SOURCE: Developed by the Research Team D‐12 

TABLE D-13 Truck Vehicle Miles—Log Actual Regressions Candidate Demand Factor Weights Model 1 2 3 4 Period of Estimation 1981-07 1981-07 1981-07 1981-07 R - Squared (adjusted) 0.997 0.997 0.997 0.998 S.E. of Regression 0.01 0.01 0.01 0.01 Durbin Watson Statistic 1.66 1.78 1.58 1.64 Model Coefficients Constant 12.05 8.59 9.43 12.44 t-stat 15.1 5.5 4.7 9.4 Candidate Demand Factors Industrial Manufacturing Index 0.23 t-stat 4.3 Total Trade in Real $ 0.10 t-stat 2.3 Exports in Real $ 0.06 t-stat 1.5 Total Employment 0.37 t-stat 2.6 Chained Inv. Sales Ratio (BEA) -0.17 -0.22 t-stat -2.0 -2.7 Retail Sales in Real $ 0.26 t-stat 2.9 Gas Price in Real $ -0.05 -0.05 -0.03 -0.04 t-stat -1.3 -1.4 -1.1 -1.3 Exogenous Impact Control NAFTA Impact 0.01 0.02 t-stat 1.5 1.5 Statistical Error Corrections AR (1) 0.98 0.98 0.99 0.98 t-stat 74.6 66.2 57.2 54.9 SOURCE: Developed by the Research Team D‐13 

TABLE D-14 Truck Vehicle Miles—Log Actual Diagnostics Truck Vehicle Miles Travelled Log - Actual*Diagnostics Model 1 2 3 4 Number of Observations 28 28 28 28 AIC (Information Criteria) -95.96 -111.12 -114.87 -114.33 BIC (Information Criteria) -90.63 -105.79 -108.21 -109.01 Structural Model R - Sq 0.34 0.35 0.42 0.44 DW stat (AR corrected) 1.66 1.78 1.58 1.62 DW stat (original) 0.89 0.72 0.83 0.89 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.09 0.05 0.27 0.67 White Test (p-value) 0.27 0.12 0.41 0.39 Variance Inflation Factors (Regressors) Total Employment 4.9 Industrial Manufacturing Index 1.1 Total Trade in Real $ 4.6 Exports in Real $ 5.7 Inv. Sales Ratio (Census) 4.8 4.6 Retail Sales in Real $ 5.9 Real Gas Price 1.1 1.1 1.3 1.2 NAFTA Impact 1.1 1.1 SOURCE: Developed by the Research Team D‐14 

TABLE D-15 Water Tons—Log Actual Regressions Candidate Demand Factors Model 1 2 3 Period of Estimation 1981- 2007 1981- 2007 1981- 2007 R-Squared Adjusted 0.76 0.75 0.66 S.E. of Regression 0.02 0.02 0.02 Durbin Watson Statistic 1.75 1.91 1.72 Model Coefficients Constant 15.53 17.61 16.39 t-stat 23.24 26.89 21.42 Candidate Demand Factors Weights Total Capacity Utilization 0.86 0.82 0.88 t-stat 6.63 5.73 5.07 Grain and Coal Tons 0.09 t-stat 2.68 Grain Tons 0.03 t-stat 3.00 Real BLS Gas -0.10 -0.08 t-stat -3.01 -3.14 Lagged Inland Waterway Trust Fund Tax/Gallon -0.03 t-stat -1.03 Exogenous Impact Controls Rail Deregulation -0.01 t-stat -0.70 Statistical Error Corrections AR(1) 0.66 0.53 0.86 t-stat 4.32 3.31 5.22 D‐15 

TABLE D-16 Water Tons—Log Actual Diagnostics Model Coefficients Number of Observations 28 28 28 AIC (Information Criteria) -148.56 -145.51 -139.02 BIC (Information Criteria) -143.23 -140.18 -133.69 Structural Model R-Squared 0.79 0.78 0.70 DW Stat (AR Corrected) 1.75 1.91 1.72 DW Stat (Original) 0.98 1.04 0.74 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.14 0.27 0.02 White Test (p-value) 0.09 0.12 0.13 Variance Inflation Factors (Regressors) Total Capacity Utilization 1.19 1.25 1.05 Grain and Coal Tons 1.02 Grain Tons 1.63 Real BLS Gas 1.20 1.60 Lagged Inland Waterway Trust Fund Tax/Gallon 1.65 Rail Deregulation 1.61 D‐16 

TABLE D-17 Water Ton Miles—Log Actual Regressions Candidate Demand Factors Model 1 2 3 Period of Estimation 1981- 2007 1981- 2007 1981- 2007 R-Squared Adjusted 0.95 0.87 0.96 S.E. of Regression 0.04 0.03 0.03 Durbin Watson Statistic 1.82 1.95 1.62 Model Coefficients Constant 23.23 26.13 24.65 t-stat 11.01 22.93 26.83 Candidate Demand Factors Weights Rail Ton-Miles -0.44 -0.65 -0.68 t-stat -4.13 -11.95 -26.29 Real BLS Gas -0.15 t-stat -2.79 Total Capacity Utilization 0.69 t-stat 3.88 Lagged Inland Waterway Trust Fund Tax/Gallon -0.13 t-stat -2.30 Purchasing Managers Index 0.25 t-stat 2.70 Exogenous Impact Controls Lagged NAFTA Impact 0.14 0.11 0.09 t-stat 4.54 8.25 3.59 Statistical Error Corrections AR(1) 0.51 t-stat 3.76 D‐17 

D‐18  TABLE D-18 Water Ton Miles—Log Actual Diagnostics Model Coefficients Number of Observations 27.00 27.00 28.00 AIC (Information Criteria) -95.91 -103.48 -107.08 BIC (Information Criteria) -89.43 -98.30 -101.75 Structural Model R-Squared 0.96 0.88 0.97 DW Stat (AR Corrected) NA 1.95 DW Stat (Original) 1.82 1.20 1.62 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.48 0.03 0.71 White Test (p-value) 0.72 0.20 0.97 Variance Inflation Factors (Regressors) Rail Ton-Miles 13.04 1.09 1.02 Real BLS Gas 2.55 Total Capacity Utilization 1.11 Lagged Inland Waterway Trust Fund Tax/Gallon 16.03 Purchasing Managers Index 1.07 Lagged NAFTA Impact 1.16 1.03 1.12

TABLE E-1 Principal Component Analysis Groups Demand Factor Commodity Consumption Foreign Exchange Production Purchasing Manager Index & Capacity Utilization Group Employment 1 Urban Gas Price in Real $ Real Personal Consumption 2005 Chained $ Trade Weighted Foreign Exchange Index (Broad Trading Partners) Real GDP 2005 Chained $ Purchasing Managers Index Total Employment 2 Increase in Coal Producer Price Index Total Housing Starts Trade Weighted Foreign Exchange Index (Major Trading Partners) Real Income/Capita, Chained 2005 $ Change in Capacity Utilization Employment Wholesale 3 Urban Gas Price in Real $ 1 yr Lag Inv.-Sales Ratio (Census) Trade Weighted Foreign Exchange Index (Broad Trading Partners) 1 Yr. Lag Industrial Production Index Purchasing Managers Index 1 yr Lag Total Employment, 1 Yr. Lag 4 Chained Inv.- Sales Ratio (BEA) Trade Weighted Foreign Exchange Index (Major Trading Partners) 1 Yr. Lag Industrial Manufacturing Index 5 Retail Sales in Real $ Real Exports in Goods (in $) 6 Real Imports in Goods (in $) Real GDP 2005 Chained $, 1 Yr. Lag 7 Real Personal Consumption 2005 Chained $, 1 Yr. Lag Real Income Per Capita 2005 Chained $, 1 Yr. Lag 8 Total Housing Starts, 1 Yr. Lag 9 Inv.-Sales Ratio (Census),1 Yr. Lag 10 Chained Inv.- Sales Ratio (BEA),1 Yr. Lag 11 Retail Sales in Real $ 1 Yr. Lag SOURCE: Developed by the Research Team E‐1  Appendix E - Primary Component Analysis Tables

TABLE E-2 Principal Component Results for the Commodity Group Commodity Group Results Principal Components in Actual Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 2.3521 1.8060 0.7840 2.3521 0.7840 2 0.5461 0.4444 0.1820 2.8983 0.9661 3 0.1017 --- 0.0339 3.0000 1.0000 SOURCE: Developed by the Research Team TABLE E-3 Principal Component Results for the Consumption Group Consumption Group Results Principal Components in Actual Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 9.1971 8.0693 0.8361 9.1971 0.8361 2 1.1279 0.7319 0.1025 10.3250 0.9386 3 0.3960 0.2716 0.0360 10.7210 0.9746 4 0.1244 0.0527 0.0113 10.8453 0.9859 5 0.0716 0.0076 0.0065 10.9170 0.9925 6 0.0641 0.0550 0.0058 10.9810 0.9983 7 0.0091 0.0036 0.0008 10.9901 0.9991 8 0.0055 0.0020 0.0005 10.9956 0.9996 9 0.0035 0.0026 0.0003 10.9991 0.9999 10 0.0008 0.0007 0.0001 10.9999 1.0000 11 0.0001 --- 0.0000 11.0000 1.0000 SOURCE: Developed by the Research Team TABLE E-4 Principal Component Results for the Foreign Exchange Group Foreign Exchange Results Principal Components in Actual Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 2.7825 1.7142 0.6956 2.7825 0.6956 2 1.0683 0.9215 0.2671 3.8508 0.9627 3 0.1468 0.1444 0.0367 3.9976 0.9994 4 0.0024 --- 0.0006 4.0000 1.0000 SOURCE: Developed by the Research Team E‐2 

TABLE E-5 Principal Component Results for the Production Group Production Group Results Principal Components in Actual Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 6.8376 6.7103 0.9768 6.8376 0.9768 2 0.1273 0.0969 0.0182 6.9648 0.9950 3 0.0303 0.0274 0.0043 6.9951 0.9993 4 0.0029 0.0019 0.0004 6.9981 0.9997 5 0.0010 0.0002 0.0001 6.9991 0.9999 6 0.0008 0.0008 0.0001 7.0000 1.0000 7 0.0000 --- 0.0000 7.0000 1.0000 SOURCE: Developed by the Research Team TABLE E-6 Principal Component Results for Manager Index/Capacity Utilization Group Purchasing Manager Index & Capacity Utilization Group Results Principal Components in Actual Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 1.9898 1.0921 0.6633 1.9898 0.6633 2 0.8977 0.7852 0.2992 2.8875 0.9625 3 0.1125 --- 0.0375 3.0000 1.0000 SOURCE: Developed by the Research Team TABLE E-7 Principal Component Results for Employment Group Employment Group Results Principal Components in Actual Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 2.9597 2.9224 0.9866 2.9597 0.9866 2 0.0373 0.0342 0.0124 2.9969 0.9990 3 0.0031 --- 0.0010 3.0000 1.0000 SOURCE: Developed by the Research Team E‐3 

TABLE E-8 Principal Component Candidate Demand Factor Weights: Commodity Group Principal Component Candidate Demand Factor Weights: Commodity Group Actual Candidate Demand Factors 1 2 3 Urban Gas Price in Real $ 0.62 -0.26 -0.74 Increase in Coal Producer Price Index 0.51 0.85 0.12 Urban Gas Price in Real $ 1 yr Lag 0.60 -0.45 0.66 SOURCE: Developed by the Research Team TABLE 1 Principal Component Candidate Demand Factor Weights: Consumption Group Principal Component Candidate Demand Factor Weights: Consumption Group Actual Candidate Demand Factors 1 2 3 4 5 Real Personal Consumption 2005 Chained $ 0.32 -0.16 0.14 0.28 0.03 Total Housing Starts 0.19 0.72 0.45 -0.01 0.45 Inv.-Sales Ratio (Census) -0.32 0.07 -0.15 0.53 0.41 Chained Inv.-Sales Ratio (BEA) -0.31 -0.16 -0.33 0.24 0.38 Retail Sales in Real $ 0.33 -0.11 0.10 0.26 -0.02 Real Imports in Goods (in $) 0.33 -0.12 0.01 0.01 -0.04 Real Personal Consumption 2005 Chained $, 1 Yr. Lag 0.32 -0.19 0.11 0.28 0.01 Total Housing Starts, 1 Yr. Lag 0.23 0.55 -0.64 0.14 -0.38 Inv.-Sales Ratio (Census),1 Yr. Lag -0.31 0.19 0.31 0.57 -0.49 Chained Inv.-Sales Ratio (BEA),1 Yr. Lag -0.32 -0.05 0.34 -0.06 -0.31 Retail Sales in Real $ 1 Yr. Lag 0.32 -0.18 0.00 0.30 0.02 SOURCE: Developed by the Research Team TABLE E-10 Principal Component Candidate Demand Factor Weights: Foreign Exchange Group Principal Component Candidate Demand Factor Weights: Foreign Exchange Group Actual Candidate Demand Factors 1 2 3 4 Trade Weighted Foreign Exchange Index (Broad Trading Partners) 0.52 0.48 0.22 -0.67 Trade Weighted Foreign Exchange Index (Major Trading Partners) -0.50 0.47 0.70 0.19 Trade Weighted Foreign Exchange Index (Broad Trading Partners) 1 Yr. Lag 0.52 0.47 -0.14 0.70 Trade Weighted Foreign Exchange Index (Major Trading Partners) 1 Yr. Lag -0.46 0.56 -0.66 -0.17 SOURCE: Developed by the Research Team E‐4 

TABLE E-11 Principal Component Candidate Demand Factor Weights: Production Group Principal Component Candidate Demand Factor Weights: Production Group Actual Candidate Demand Factors 1 2 3 4 5 Real GDP 2005 Chained $ 0.38 -0.13 0.04 0.14 -0.87 Real Income/Capita, Chained 2005 $ 0.38 -0.31 0.12 0.76 0.35 Industrial Production Index 0.38 0.03 -0.58 -0.16 0.27 Industrial Manufacturing Index 0.38 0.00 -0.56 -0.07 -0.07 Real Exports in Goods (in $) 0.36 0.88 0.26 0.13 0.06 Real GDP 2005 Chained $, 1 Yr. Lag 0.38 -0.12 0.28 -0.53 0.15 Real Income Per Capita 2005 Chained $, 1 Yr. Lag 0.38 -0.31 0.44 -0.27 0.12 SOURCE: Developed by the Research Team TABLE E-12 Principal Component Candidate Demand Factor Weights Manager Index/Capacity Utilization Group Principal Component Candidate Demand Factor Weights: Purch. Mangr. Index & Capacity Utilization Group Actual Candidate Demand Factors 1 2 3 Purchasing Managers Index 0.64 -0.38 0.67 Change in Capacity Utilization 0.68 -0.11 -0.72 Purchasing Managers Index 1 yr Lag 0.35 0.92 0.19 SOURCE: Developed by the Research Team TABLE E-13 Principal Component Candidate Demand Factor Weights: Employment Group Principal Component Candidate Demand Factor Weights: Employment Group Actual Candidate Demand Factors 1 2 3 Total Employment 0.58 -0.20 -0.79 Employment Wholesale 0.57 0.79 0.23 Total Employment, 1 Yr. Lag 0.58 -0.59 0.57 SOURCE: Developed by the Research Team E‐5 

TABLE E-14 Rail Tons PCA—Log Actual Regressions Rail Tons Log – Actual* Model 1 2 2* 3 Period of Estimation 1981-2007 1982-2007 1982-2007 1981-2007 R - Squared (adjusted) 0.94 0.98 0.99 0.95 S.E. of Regression 0.03 0.02 0.01 0.03 Durbin Watson Statistic 1.21 1.70 2.38 1.19 Model Coefficients Constant 14.25 14.25 14.24 14.25 t-stat 2143.5 3983.9 4401.7 2151.3 Principal Component Regression Weights Production Principal Component 0.05 t-stat 21.12 Employment Principal Component 0.08 t-stat 21.36 Employment Comp Lag 0.02 0.03 t-stat 2.19 6.59 Consumption Principal Component 0.03 0.03 t-stat 6.94 12.74 Commodity Principal Component 0.03 0.01 0.00 0.02 t-stat 6.20 4.82 2.19 3.64 Purch. Mngr. & Cap. Util Component 0.01 0.01 t-stat 4.23 2.07 Exogenous Impact Control Lagged NAFTA Impact 0.03 0.03 0.03 0.03 t-stat 2.40 3.74 4.50 2.31 SOURCE: Developed by the Research Team E‐6 

TABLE 2 Rail Tons PCA—Log Actual Diagnostics Rail Tons Log - Actual* Model 1 2 2* 3 Number of Observations 27 26 23 27 AIC (Information Criteria) -104.26 -117.92 -132.99 -104.08 BIC (Information Criteria) -97.78 -111.63 -127.31 -97.61 Adjusted Model R - Sq 0.94 0.97 0.99 0.94 DW stat (original) . 1.70 2.38 1.19 Serial Corr. LM Test (1-lag) 0.16 0.44 0.31 0.12 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.05 0.06 0.81 0.05 White Test (p-value) 0.59 0.01 0.02 0.20 Variance Inflation Factors (Regressors) Production Principal Component 1.0 Employment Principal Component 6.17 Employment Comp Lag 7.82 6.17 Consumption Principal Component 7.35 7.71 7.35 Commodity Principal Component 1.58 1.27 1.58 1.15 Purch. Mngr. & Cap. Util Component 1.09 Lagged NAFTA Impact 1.06 1.06 1.06 1.07 SOURCE: Developed by the Research Team E‐7 

TABLE E-16 Rail Ton Miles PCA—Log Actual Regressions Rail Ton Miles Log - Actual* Model 1 2 3 Period of Estimation 1981-2007 1981-2007 1982-2007 R - Squared (adjusted) 0.99 0.99 0.99 S.E. of Regression 0.03 0.02 0.03 Durbin Watson Statistic 1.36 1.81 1.76 Model Coefficients Constant 14.02 14.04 14.48 t-stat 2456.6 1667.5 12.7 Principal Component Weights Production Principal Component 0.10 t-stat 41.5 Employment Lag Principal Component 0.16 t-stat 35.6 Commodity Principal Component 0.01 0.02 0.01 t-stat 1.8 3.8 2.0 Consumption Principal Component 0.05 t-stat 2.8 Purch. Mngr. & Cap. Util Component 0.03 t-stat 6.5 Exogenous Impact Control Lagged NAFTA Impact 0.07 0.08 0.05 t-stat 9.1 9.1 10.7 Statistical Error Corrections AR (1) 0.32 0.98 t-stat 2.3 17.0 SOURCE: Developed by the Research Team E‐8 

TABLE E-17 Rail Tons Miles PCA—Log Actual Diagnostics Rail Ton Miles Log - Actual* Model 1 2 3 Number of Observations 27 25 26 AIC (Information Criteria) -111.42 -111.25 -100.60 BIC (Information Criteria) -106.23 -105.15 -94.31 Adjusted Model R - Sq 0.99 0.98 0.56 DW stat (original) 1.36 1.18 0.43 DW stat (corrected) - - - Serial Correlation LM Test 0.55 0.05 0.00 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.11 0.24 0.18 White Test (p-value) 0.04 0.52 0.05 Variance Inflation Factors (Regressors) Production Principal Component 1.0 Employment Lag Principal Component 1.02 Consumption Principal Component 1.01 Commodity Principal Component 1.09 1.12 1.07 Purch. Mngr. & Cap. Util Component 1.0 Lagged NAFTA Impact 1.06 1.1 1.1 SOURCE: Developed by the Research Team E‐9 

TABLE E-18 Rail Revenue Ton Miles PCA—Log Actual Regressions Rail Revenue Ton Miles Log - Actual* Model 1 2 3 Period of Estimation 1981-2007 1982-2007 1982-2007 R - Squared (adjusted) 0.99 0.99 0.99 S.E. of Regression 0.03 0.03 0.03 Durbin Watson Statistic 1.20 1.43 1.34 Model Coefficients Constant 20.90 20.92 20.89 t-stat 4050.1 3856.0 1043.9 Principal Component Weights Production Principal Component 0.09 t-stat 45.4 Employment Lag Principal Component 0.14 t-stat 53.5 Commodity Principal Component 0.01 0.03 0.01 t-stat 3.2 7.8 1.3 Consumption Principal Component 0.05 t-stat 9.0 Purch. Mngr. & Cap. Util Component 0.01 0.04 t-stat 3.8 8.4 Foreign Exchange Principal Component 0.06 t-stat 3.8 Exogenous Impact Control Lagged NAFTA Impact 0.09 0.12 0.04 t-stat 12.9 11.4 5.0 SOURCE: Developed by the Research Team E‐10 

TABLE E-19 Rail Revenue Ton Miles PCA—Log Actual Diagnostics Rail Revenue Ton Miles Log - Actual* Model 1 2 3 Number of Observations 27 26 26 AIC (Information Criteria) -116.65 -111.48 -113.94 BIC (Information Criteria) -110.17 -105.19 -107.65 Adjusted Model R - Sq 0.99 0.99 0.94 DW stat (original) 1.20 1.43 0.50 DW stat (corrected) Serial Correlation LM Test 0.25 0.21 0.00 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.14 0.43 0.08 White Test (p-value) 0.04 0.16 0.83 Variance Inflation Factors (Regressors) Production Principal Component 1.04 Employment Lag Principal Component 1.02 Consumption Principal Component 3.93 Commodity Principal Component 1.15 1.12 1.29 Purch. Mngr. & Cap. Util Component 1.09 1.04 Foreign Ex. Principal Component 4.48 Lagged NAFTA Impact 1.07 1.07 1.18 E‐11 

TABLE E-20 Rail Train Miles PCA—Log Actual Regressions Rail Train Miles Log - Actual* Model 1 2 3 Period of Estimation 1990-2007 1990-2007 1990-2007 R - Squared (adjusted) 0.97 0.97 0.98 S.E. of Regression 0.02 0.02 0.02 Durbin Watson Statistic 1.24 1.53 1.22 Model Coefficients Constant 6.05 6.05 6.08 t-stat 707.7 745.2 823.0 Principal Component Weights Production Principal Component 0.07 t-stat 18.0 Employment Principal Component 0.11 t-stat 17.2 Commodity Principal Component 0.01 t-stat 2.0 Consumption Principal Component 0.05 t-stat 23.7 Purch. Mngr. & Cap. Util Component 0.03 0.01 t-stat 4.0 2.2 Exogenous Impact Control Lagged NAFTA Impact 0.05 0.06 0.05 t-stat 6.6 8.1 5.8 E‐12 

TABLE E-21 Rail Train Miles PCA—Log Actual Diagnostics Rail Train Miles Log - Actual* Model 1 2 3 Number of Observations 18 18 18 AIC (Information Criteria) -80.17 -85.17 -88.18 BIC (Information Criteria) -75.72 -81.61 -85.51 Adjusted Model R - Sq 0.97 0.97 0.98 DW stat (original) 1.24 1.53 1.22 DW stat (corrected) Serial Correlation LM Test 0.18 0.63 0.32 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.23 0.24 0.42 White Test (p-value) 0.76 0.98 0.08 Variance Inflation Factors (Regressors) Production Principal Component 1.11 Employment Principal Component 1.25 Consumption Principal Component 1.04 Commodity Principal Component 1.31 Purch. Mngr. & Cap. Util Component 1.02 1.05 Lagged NAFTA Impact 1.08 1.05 1.04 E‐13 

TABLE E-22 Rail Car Miles PCA—Log Actual Regressions Rail Car Miles Log - Actual* Model 1 2 3 Period of Estimation 1990-2007 1991-2007 1990 2007 R - Squared (adjusted) 0.99 0.99 0.99 S.E. of Regression 0.01 0.01 0.02 Durbin Watson Statistic 1.42 2.07 2.20 Model Coefficients Constant 10.26 10.31 10.27 t-stat 2226.9 1161.9 1565.3 Principal Component Weights Production Principal Component 0.08 t-stat 30.4 Employment Lag Principal Component 0.12 t-stat 26.1 Commodity Principal Component 0.01 t-stat 2.5 Consumption Principal Component 0.05 t-stat 18.1 Purch. Mngr. & Cap. Util Component 0.01 0.03 t-stat 3.1 11.9 Exogenous Impact Control Lagged NAFTA Impact 0.05 0.04 0.08 t-stat 6.2 3.4 10.5 Statistical Error Corrections AR (1) 0.50 t-stat 2.5 E‐14 

TABLE E-23 Rail Car Miles PCA—Log Actual Diagnostics Rail Car Miles Log - Actual* Model 1 2 3 Number of Observations 18 17 18 AIC (Information Criteria) -99.41 -97.50 -92.81 BIC (Information Criteria) -95.85 -95.00 -88.36 Adjusted Model R - Sq 0.99 0.97 0.99 DW stat (original) 1.42 1.11 2.20 DW stat (corrected) - - - Serial Correlation LM Test 0.72 0.06 0.48 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.17 0.93 0.73 White Test (p-value) 0.28 0.79 0.67 Variance Inflation Factors (Regressors) Production Principal Component 1.11 Employment Lag Principal Component 1.32 Consumption Principal Component 1.04 Commodity Principal Component 1.34 Purch. Mngr. & Cap. Util Component 1.05 1.02 Lagged NAFTA Impact 1.05 1.04 1.10 E‐15 

TABLE E-24 Truck Ton Miles PCA—Log Actual Regressions Truck Ton Miles Log - Actual* Model 1 2 3 Period of Estimation 1982-2007 1983-2007 1984-2007 R - Squared (adjusted) 0.99 0.99 0.99 S.E. of Regression 0.01 0.01 0.02 Durbin Watson Statistic 1.36 1.91 2.13 Model Coefficients Constant 14.43 14.56 13.80 t-stat 52.1 26.7 1289.8 Principal Component Weights Production Principal Component 0.02 t-stat 3.1 Employment Lag Principal Component 0.14 t-stat 15.8 Commodity Principal Component 0.00 0.00 t-stat -1.9 -1.0 Consumption Principal Component 0.00 t-stat -1.0 Purch. Mngr. & Cap. Util Component 0.01 t-stat 3.5 Exogenous Impact Control NAFTA Impact 0.97 0.01 0.02 t-stat 80.4 2.0 2.4 Statistical Error Corrections AR (1) 0.97 1.42 1.25 t-stat 80.4 6.6 6.5 AR (2) -0.44 -0.64 t-stat -2.1 -3.6 E‐16 

TABLE E-25 Truck Ton Miles PCA—Log Actual Diagnostics Model 1 2 3 Number of Observations 26 25 24 AIC (Information Criteria) -153.31 -156.20 -127.40 BIC (Information Criteria) -5.65 -5.72 -4.61 Adjusted Model R - Sq 1.00 0.65 0.63 DW stat (original) 0.00 0.00 0.00 DW stat (corrected) Serial Correlation LM Test 0.87 0.04 0.07 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.78 0.06 0.22 White Test (p-value) 0.00 0.00 0.00 Variance Inflation Factors (Regressors) Production Principal Component 1.03 Employment Lag Principal Component 1.00 Consumption Principal Component 1.01 Commodity Principal Component 1.09 1.07 Purch. Mngr. & Cap. Util Component 1.03 NAFTA Impact 1.06 1.06 1.03 E‐17 

TABLE E-26 Truck VMT PCA—Log Actual Regressions Truck VMT Log - Actual* Model 1 2 3 Period of Estimation 1982-2007 1983-2007 1984-2007 R - Squared (adjusted) 0.99 0.99 0.99 S.E. of Regression 0.01 0.01 0.02 Durbin Watson Statistic 1.38 1.92 2.14 Model Coefficients Constant 12.67 12.80 12.04 t-stat 45.8 23.6 1120.7 Principal Component Weights Production Principal Component 0.02 t-stat 3.1 Employment Lag Principal Component 0.14 t-stat 15.7 Commodity Principal Component 0.00 0.00 t-stat -1.8 -1.0 Consumption Principal Component 0.01 t-stat 2.1 Purch. Mngr. & Cap. Util Component 0.01 t-stat 3.5 Exogenous Impact Control NAFTA Impact 0.02 0.01 0.02 t-stat 1.3 1.9 2.4 Statistical Error Corrections AR (1) 0.97 1.41 1.25 t-stat 80.0 6.5 6.5 AR (2) -0.43 -0.63 t-stat -2.0 -3.6 E‐18 

E‐19  TABLE E-27 Truck VMT PCA—Log Actual Diagnostics Truck VMT Log - Actual* Model 1 2 3 Number of Observations 26 25 24 AIC (Information Criteria) -153.04 -149.71 -117.58 BIC (Information Criteria) -5.64 -5.70 -4.60 Adjusted Model R - Sq 0.99 0.99 0.99 DW stat (original) 0.83 0.65 0.63 DW stat (corrected) Serial Correlation LM Test 0.00 0.00 0.00 Heteroskedasticity Tests (Null Hypothesis: Constant Variance) Breusch-Pagan Test (p-value) 0.87 0.04 0.07 White Test (p-value) 0.78 0.06 0.22 Variance Inflation Factors (Regressors) Production Principal Component 1.0 Employment Lag Principal Component 1.00 Consumption Principal Component 1.01 Commodity Principal Component 1.09 1.07 Purch. Mngr. & Cap. Util Component 1.0 NAFTA Impact 1.06 1.1 1.0