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.
40 S t e p 5 5.1 Goal The goal of this step is to develop appropriate forecasts of freight activity and transport costs for the build and no-build alternatives. It is important to understand how the stage of project develop- ment can influence the individual tasks in this step. The purpose of the study (in effect, the project stage) determines the level of confidence required, which then determines the data requirements and the level of analytical rigor. Each progressive stage of development has an increasing demand for certainty and accuracy that drives the approach. This step focuses on determining the appropri- ate level of effort for BCA and how multimodal aspects can or should be considered. 5.2 Tasks Resolve Issues Related to Transport Forecasts for Project Alternatives The analytical goals of this task are to obtain modal freight flows on the corridor and transport cost changes to allow for incremental analysis of the build and no-build (or do-minimum) sce- narios over the analysis period. This task includes the development of behavioral forecasts of the project(s) over the selected analysis period and the selected build and do-minimum scenarios. Behavioral effects typically fall within three categories: â¢ Routes (projects that impact routes taken). â¢ Mode choice. â¢ Departure time. These complex effects are not captured in typical travel demand models, except for dynamic traffic assignment or microsimulation models. BCA requires that benefits be captured for existing users, induced users (including diverted users), and generated users. If the project has the ability to create network effects, use the demand model zonal framework (network approach) to allow approximation of generated benefits using the well-known rule of half (Appendix A). This typically comes into play for multiregional and multistate ventures for all line-haul networks. The approach leads to two sets of forecast flow and impedance matricesâone for the build and the other for the no-buildâfor the use of the rule of half. This is important for ensuring that the consumer and producer surplus measures are correctly approximated. It also facilitates distributional analysis in later steps because of the zonal coverage for the area. This is an effort and time-intensive approach; hence, it is recom- mended only for feasibility studies. Models may also not be available to capture all nuances of the investment. The project scale and type for analyzing behavioral effects should be matched to data that can be used for forecasts. Develop Forecasts
Develop Forecasts 41 Modal flows refer to truck freight volumes, rail freight volumes, air cargo volumes, and barge and marine cargo volumes reported as trips or movements of motive power. In addition, some- times volumes may refer to tonnage flows. A commodity forecast breakdown, if available, can enrich the analysis for subsequent stages. The freight flow forecast data outputs (flows and impedances/times and costs) required to sat- isfy these analytical goals for line-haul modes are often provided by travel demand models or commodity flow models based on geographic scale that state departments of transportation or metropolitan planning organizations may have developed for their freight planning needs. In addition, mode choice equations that are part of the freight demand models can also be used in developing a response to the build and no-build alternatives. Apply Appropriate Models and Data and Develop Forecasts A BCA is guided by three factors: (1) the geographic scope of the project; (2) the level of analy- sis, and (3) the availability of suitable network models for developing forecasts. For each mode of transportation, network models analyze the demand for use of routes and terminals, the func- tional capacity of those routes and terminals, and resulting changes in their performance. Their value lies in their capability to simulate how improvements to network or terminal facilities can improve direct performance (as measured by time, distance, and speedsâfactors influencing transport costs). The following discussion suggests how to develop forecasts and the outputs that will be needed. The proportionality rule plays an important role in model selection by taking into account staffing and monetary resources and the required level of analysis. Type of Analysis-Line Haul Network Improvement Conceptual Analysis: A post processed or sketch plan approximation of volumes on the entire corridor (vehicle miles traveled [VMT] and vehicle hours of travel [VHT]) in the build and reasonable no-build scenarios can be used to analyze trip alternatives for truck freight. Modal connections such as gateway terminal capacity changes or additions of landside terminals can be included in the preliminary information used to obtain VMT and vehicle hours of travel, subjected to sensitivity tests. Freight volumes and ton-miles should be used for other modes. The inputs for this type of analysis forecast can be developed using baseline freight volumes from travel models suited for the context, aggregate databases, and elasticities. Sketch plan BCA spreadsheet tools can be used or developed for the BCA. The data for conceptual analysis can be obtained from aggregate freight flows. While these are less precise, the BCA precision can be enhanced by appropriate sensitivity or risk analysis. Feasibility and Investment Analysis: The inputs for this type of forecast are best developed using baseline freight volumes from network travel models specifically suited for the context. They are more time and cost intensive, but recommended for large projects. For instance, this means relying on statewide, regional, or custom models as they are consistent with planning needs for the geographic scope of planning for projects of such scales. This will also be important for ensuring greater precision in results. In addition, they can also be supported by special studies to examine specific considerations such as mode shifts, value of reliability, or other factors. â¢ Catalog behavioral aspects of projects from Step 3: â If the project motivations are to induce route shifts, then the forecast for the build alterna- tive must include those effects. For smaller scale projects, a fixed demand forecast for the terminal period may be acceptable and is simpler to develop. The travel and cost outputs for both scenarios can be post-processed for use in BCA. â If the project is expected to induce modal shift, then the forecasts must include mode choice factors that will capture cost change responses in the forecasts.
42 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments â¢ Ensure that baseline and forecast conditions with the project are reflected in the model and networks for both the build and no-build alternatives. â¢ Develop the forecast over the analysis period. If the analysis period of the project exceeds the horizon period in the travel model, suitable extrapolations must be made. â¢ If network effects from route or mode diversion are expected over time and the analysis uses a fixed demand forecast, be prepared to adjust the forecasts to be more reflective of the induced demand using elasticities or risk analysis methods. A fixed demand forecast or (fixed trip matrix) is one that ignores induced demand between origins and destinations since the trip matrix is simply reassigned to alternate routes or modes for the base period and forward. For large-scale projects, this is a source of error since volumes and costs and first-order direct ben- efits will be biased. The assumptions of tolls and fuel taxes as transfers (as discussed in Step 8) are violated in this approach. â¢ Network post processor BCA tools can be used if available or developed to support BCA. Working Without a Travel Model (O-D Matrices and Network Analysis)â Simplest Approach If a commodity freight model or travel model is not available to the analyst (or does not exist at all), the analyst will have to construct his or her own estimation of travel costs based on values derived from network analysis. Network analysis relies on the use of geographic information sys- tem route networks. Carefully document all assumptions and the basis for the estimation of travel costs and forecasts. Another approach is to develop an O-D matrix from surveys. This approach can be used for new options, but it is a costly process. Sketch plan BCA spreadsheet tools can be used or developed for the BCA. Network Post Processing Models by Scale and Geography Models are characterized by project scale and geography. A caveat arises in that network models do not exist for all scales and geography for all modes. Some state and regional models allow dis- aggregate zonal analysis of truck freight, which can be useful for feasibility and investment analysis of freight options as well for examining distributional effects. â¢ Metropolitan/Regional Scale Models: Most metropolitan regions typically do not get very involved in freight movements, and freight choices (mode and routes) do not go beyond truck freight corridors. In some cases (e.g., multimodal truck-rail freight projects), regional models may have to be supported by statewide freight models. â¢ Statewide Models: States and metropolitan regions vary with respect to types of models they maintain or have developed for freight. States may have many types of demand modelsâtypical highway travel demand models, heavy-truck models, activity models, integrated economic activity models, logistics type models, or integrated transport land use models with a focus on highways that sometimes also include transit. They may also have freight models such as commodity flow models. These are more suited for rail corridors and truck-rail freight diver- sion projects. In some projects, both travel demand and commodity flow models can be used simultaneously (especially if two modes are being compared). â¢ Custom Models for Multistate, Megaregion Truck Freight: Methods discussed in NCHRP Report 606 (26) using O-D matrices can be used in large-scale, multistate, or state projects when states have no demand models. The actual O-D may be obtained from public-domain (FAF) or private-domain (e.g., Transearch) sources or otherwise established from resource-intensive sur- veys. At multistate and megaregion scales, custom models used for planning purposes can sup- port BCA. Examples of such developments are those for the Chesapeake megaregion model and the Mid-America Freight Model (27). The Appalachian Regional Commission requested that a custom model be developed for the economic evaluation of the Appalachian Development Highway System (28). The Appalachian Regional Commission TransCAD model is a passenger and truck model developed from FAF1 data covering 697 counties. The Interstate 95 Coalition recently developed demand modeling capabilities called the Integrated Corridor Analysis Tool
Develop Forecasts 43 (ICAT) which was used for an economic study. These are all resource intensive (in terms of development and calibration) and are used to support broader corridor planning needs that span multiple geographies; they may potentially have great value for BCA of multijurisdictional corridor projects. If such models already exist, they should be used for BCA with attendant assumptions in corridor multimodal BCA. â¢ National-Scale Multimodal Freight Networks: FAF is a national network multimodal tool. Conceptual studies could use the FAF, supplemented by other regional or state models as needed. There are limitations in using public domain FAF; however, it can serve as a prelimi- nary basis for BCA as long as it is supported by suitable risk analysis and assumptions. Private sector railroads often rely on commercial network models such as the Rail Traffic Control- ler (RTC), developed by Berkeley Simulation Software. It is used extensively by the U.S. rail industry for modeling capacity. While coverage is for the entire nation, it works best for smaller areas so as to provide travel performance metrics for use in BCA. It also works best for handling projects where grade crossings are part of the alternatives compared. Projects with Nodal Involvement (Terminals at Ports and Airports and Intermodal Terminals) Terminal area models, if available, can be used if the investment involves the terminal signifi- cantly. In some terminal project contexts, terminal area forecasts are sufficient to address the problem at hand. Additional Considerations These additional points are worth noting: â¢ Rail network data are private information, but statewide commodity flow models allow exam- ination of rail corridors, as does the FAF. States that have commodity-flow-based models may also have mode choice elements that can allow the effects of diversions to be considered. â¢ Waterways can be addressed using the FAF to a very limited extent (for non-USACE projects). Many movements that use water are included in the multimodal flow category, resulting in underreporting of water traffic in some cases. Private data sources may fill this gap partly. USACE maintains its own network model for the inland waterways, ship channels, and harbors. When modal diversion is a possibility (either to or from the water), USACE typically relies on stated-preference surveys to determine the possible extent of such diversions, while still using its core network model. Desired Outputs from Models for BCA Ultimately, the key outputs that are needed for BCA for the build and no-build include: â¢ At the Corridor Level: â Freight volumes (daily, annual volumes) for the route examined and for all segments in the corridor (line haul) for all user classes (e.g., medium, heavy trucks for truck corridors). They can be broken down by time of day if examining congestion effects. It is more time consum- ing to have O-D profiles but best if available. They should be broken down by volumes into, out of, and through the region if possible. The forecast should cover the analysis period. â Freight volumes by mode in other cases and by commodity. â Travel costs (as measured by time, distance changes) changes (mode/route costs). â Assumptions on time costs and other costs used in assignments (route and mode choice), if used. Ultimately, the BCA should also rely on using these same values in valuation as defaults instead of those provided by the U.S. DOT or other agencies. â¢ Network BCA Area AnalysisâAlternatively, for later stage studies, these outputs may be obtained as matrices which will allow a better approximation of the rule of half: â Annual freight/commodity volumes, by commodity type, for O-D pairsâthink of this as a matrix, or several matrices.
44 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments â Mode/route choice for each cell in the above matrices. â Mode/route costs: b Travel time for each mode for each O-D pair. b Other transportation costs by mode by O-D pair. b Externalities (safety, emissions) by mode O-D pair. Steps to Obtain Desired Outputs 1. Begin by forecasting A-freight volume forecast for the alternatives. For a conceptual analysis, these commodity flows may be the same for all alternatives. There are several resources discussed in this section that support this step (discussed under network models by scale and geography). These also include public domain data obtained through FAF, Waybill, and other sources like these. These data bases provide data by O-D pair that can be used for conceptual analysis. 2. In more advanced studies, the forecasts may be developed from an assignment model based on current and future transportation costs for each mode/route in the future, and to assign com- modity volumes to each mode/route. This most typically occurs in the case of truck freight. 3. Develop the forecast for the analysis period: If the data are only for the base year of the analy- sis, develop a forward forecast for the rest of the duration. Data sources like travel models and freight models develop forecasts over the period of the model. Aggregate data sources like FAF on the other hand provide forecasts through 2040 (FAF4). If the forecast is provided or can be developed only for a few points over the analysis period, then the interim data points have to be interpolated. If the analysis is longer than the forecast period, then extrapolate the forecasts. 4. Use the volumes and transport costs to develop estimates of TEE benefits. 5. The same volumes can be used to estimate externalities and wider benefits. Quality of Inputs and BCA Credibility Since data inputs are critical to the process and the credibility of the BCA itself, risk and uncer- tainty in the treatment of costs, volumes, and benefits are important in all cases. The process of identification of risky variables on the volume side is important; it involves isolating the factors that influence volume forecasts for the build and no-build alternatives. In principle, best use must be made of all existing models and tools to conduct the BCA since the level of confidence is largely driven by data inputs. Table 6 provides a summary of the matching data requirements based on proportionality principles. Consider Methodological Assumptions in Using FAF, Waybill, and Transearch for Line-Haul NetworksâRail, Waterways, and Pipelines The use of public-domain aggregate data such as FAF, Waybill, or private-sector data such as Transearch is certainly a way forward in the absence of models. However, a few issues must be recognized in the use of such for any level of analysis because they can contribute to uncertainty on the benefit side: â¢ FAF forecasts assume modal shares will be constant over the years. For line-haul investments such as rail, waterways, and marine investments, the future flows would not reflect these modal shift changes in the future forecast years. However, individual year estimates of mode shares can be adjusted using procedures and tools noted in Table 7. FAF and other aggregate annual flows mask seasonality aspects of freight flows that are important in some types of freight movements. â¢ FAF networks are fixed for the entire forecast period, so induced demand effects must be addressed exogenously. â¢ The use of the FAF is also contrary to any BCA where assignment models can be used to model the effects of specific investments on network flows. â¢ The use of macroscopic O-D flows must be accompanied by assumptions that must be used to apportion macro-flows to be useful for analysis. The FAF, for instance, is often used for truck
Develop Forecasts 45 Model/Forecast Approach Level of Analysis Resource Requirements Sketch Plan, Network Analysis, Aggregate Models (FAF, Waybill etc.), Origin-Destination Matrices ConceptualâLine Haul Budget: Low Staff Expertise: Low-Medium Time Requirements: 2 months Network Existing Post Processing Models (Statewide, Metropolitan) Feasibility/Alternatives, Investment Analysis for state and regional geographiesâLine Haul Budget: Medium Staff Expertise: Medium Time Requirements: 4â7 months. This will increase based on modal interactions and need to rely on terminal forecasts. Network Custom Post Processing Models (Statewide, Metropolitan) Feasibility/Alternatives, Investment Analysis for multi-state geographiesâ Line Haul Budget: High Staff Expertise: Advanced Time Requirements: 8 months- 1 year. Modal interactions and need to rely on terminal forecasts or interface with other models will add to complexity. Notes: 1. Network analysis and network models also provide impedance or travel costs forecasts for line-haul contexts. Terminals and smaller regional geographies for rail, air, and ports can benefit from simulation models. 2. Confidential Waybill, Transearch databases will add to budget and time costs. 3. Models need to be run for both build and no-build contexts. Assumptions must be documented for costs, investments as part of networks; demand projections. A trained economist working with demand modelers and planners is helpful for interpreting outputs. Table 6. Summary of tools and models by scale and geography matched to resource requirements for freight volume forecasts. and auto volume assignments by a disaggregation to the county level. Similarly, when the FAF is used for rail and waterways, it must be accompanied by apportioning assumptions. This is an area of continuing research and advanced modeling for assigning freight flows. Absent that, apportioning assumptions for rail could include (but are not limited to): â Carrier availability (number of carriers) serving the O-D pair and/or terminals or market share. â Total tonnage flow between the origin and destination. â Total flow or tonnage by commodity between the O-D pair. â Other rules for using specific routes/tracks. Similar issues arise with the use of the Waybill where flows are organized by origin and desti- nation, Bureau of Economic Analysis (BEA) zones, and freight rate territory. Methodological Approaches to Modeling Modal Diversion Table 7 and the supporting discussion in this section provide different methodologies by which modal diversion can be determined. Estimating diversion from one mode to another requires assumptions about the baseline excess demand relative to modal capacity. Several methods have been used in the literature, but the methods have rarely ever been well discussed. Table 7 provides a listing of five types of models (as denoted by the column headers) that have been observed/ reported when modal diversion is an intended goal or a specific feature of a project that is a com- ponent driving benefits. Examples are included in the table and in the Examples section to show specific instances of use or applications in freight BCA. Of the five methods listed, most conceptual and feasibility analyses can rely on the first four methods. Demand models may be used if states and regions have diversion considerations and mode choice models built directly into their demand models. Of the four methods, the most resource-intensive yet contextually appropriate and robust method is the choice method (mode or route) developed for the specific mode or route in question.
46 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments Market Segmentation by Commodity Type and Rules of Thumb for Major O-D Market Pairs. This method relies on six steps applied for defining the O-D markets (O-D matrix of com- modity flows) for the corridor and modes in consideration. It involves defining a âfromâ mode and a âtoâ mode. An O-D pair comparative assessment implies that flows by both modes (from and to) are established for the corridor. Comparative analysis is most useful if capacity considerations are also examined so that estimated diversions can be accommodated on the segments and terminals. The six steps are as follows: 1. Develop a base set of commodity flows of the âfromâ and âtoâ modes (the FAF and Waybill), along with their units by O-D pair. This step is aimed at understanding base modal shares and flows by mode between the O-D pair. 2. Demonstrate the suitability of the target market or O-D pair and equipment for diversion since modal diversion for freight rail or waterways may occur over the full route. Trail-rail competi- tion is strongest in distances of 400 to 600 miles. Therefore, long-distance volumes may provide the appropriate market segments for truck-rail or rail-truck and truck-waterway diversion. Tool Type Attributes Market Segmentation Rules of Thumb, Truck to Rail Modal Elasticity Mode Choice Models (New Modes) Built into Travel Models Examples Mid-Atlantic Rail Operations (MAROps) Comparable Markets Method for Assigning Rail SharesâPhase 2 Study (29) NCHRP Report 8-42 (30) and mode shift parameters from within freight models FHWAâs Intermodal Transportation and Inventory Costing Model (ITIC), AARâs Intermodal Competition Model, and Global Insightâs Comparative Cross- Modal Economics Model Discrete choice North Carolina Department of Transportation Demand Model Geographic scope Local to nationwide, based on data availability Regional, statewide, and national developed for corridor analysis State-nationwide truck-rail diversion developed for corridor analysis Regional, statewide, or national based on context Statewide Data sources Commodity flows, network and modal characteristics (data by O-D pair). FAF3 or Transearch zones FAF or private- domain Transearch. Truck flows segmented by geography, in, out, and through. Commodity flows, network and modal characteristics, and cost or pricing information Commodity flows, network and modal characteristics, shipper/carrier preference surveys, and employment data Truck flows and FAF Use Identifies markets for modal shift and diversion potentialâ modal shift by commodity First-order approximation Estimates potential modal shift in response to price or level-of-service changes Estimates modal shares based on transportation and other factors. Flexibility for scenarios. Truck rail diversion Strengths Least data intensiveâ Implemented on a national scale. Simple to use. Not data intensive. Is also a market segmentation approach. Simple to use. Relies on goods movement data to account for infrastructure- related cost changes. Can provide context appropriate mode choice and share parameters. Relies on goods movement data to account for infrastructure- related cost changes. Point-to-point intermodal diversion between rail and truck in the demand model Weaknesses Limited by quality of data, unable to account for changing characteristics Limited by quality of data Results can be highly case-specific. Some resources are proprietary. Resource intensive/ costly/surveys Not known Table 7. Accepted mode diversion models/tools.
Develop Forecasts 47 3. Develop commodity category filters to identify divertible cargo (see Step 5 worksheets on filters and modal elasticities in Appendix M). 4. Apply filters to the base mode and percentage for diversion. 5. Apply capacity constraints along the corridor (and at terminals) when allocating diversion percentages (if applicable). 6. Establish truck-rail equivalence or conversion of resultant divertible flows to equivalent flows in the âtoâ mode with assumptions on phasing in the diversion over time. It is complicated to provide a unique set of conversion factors because of commodity type, train type, and routes. Start out with tons per truck and/or FAF payloads based on the context, and improve this with more context-specific information, as available. Modal Switch ElasticitiesâRail-Truck/Truck-Rail. This approach relies on the use of cross elasticities established from earlier studies (Friedlander and Spady , Abdelwahab , Clark et al. , and McCullough ). These estimates are summarized in Appendix D. The more recent studies indicate that substitution opportunities exist between truck and rail and that the markets can be quite competitive. These commodity-specific elasticities are the results of mode choice studies developed for varying policy contexts. They should be used along with market segments (or O-D pairs) and estimates of cost reduction in those market segments to define the extent of diversion. The following additional points are worth noting when using this approach for conceptual analysis: â¢ The elasticities represent only a suggested maximum diversion opportunity for any commod- ity category. A conservative approach is to apply moderate diversion elasticities around 0.5 to divertible commodities. â¢ Stakeholder inputs into this diversion potential are very useful in refining estimates for feasibility studies. â¢ Consider a phase-in time frame for applying the diversion since not all diversion occurs instantaneously. â¢ The parameter can be subjected to sensitivity testing. ITIC Model (National and State Versions). This is the only formal model that has been devel- oped for examining diversions and based on mode shares. Therefore, it is discussed here. FHWAâs ITIC is an open-source tool that can be used for state and national scales of analysis. It has a state-level version called ITIC-ST (developed in 2006), which was intended to be used along with the Highway Economic Requirement SystemâState Version [HERS-ST]. It estimates diversion generated by a change in the transportation levels of service or price caused by improvements in infrastructure, operations, or policy. The model includes a significant number of decision vari- ables determining mode choice, supplier, and shipment size. ITIC is an inventory costing model assuming mode choice (and implied modal elasticity) as a factor in the inventory/warehousing decision. It is based on a translogarithmic shipper cost function. Documentation is provided along with the tool, but no definitions and derivations for param- eters are provided. The model calculates the modal share based on an internal set of assumptions and parameters. The user can change parameters and some assumptions to run scenarios. ITIC uses an all-or-nothing assignment rule in determining diversion. It includes obsolescence fac- tors ranging from 1 to 10% for 50 Standard Commodity Classification Codes (1â50). It assumes a speed of 23 mph for trailer-on-flat-car traffic on rail line hauls and 60 mph for trucks. Dray speed is set at 35 mph. The modal shares in each scenario can be compared in order to determine modal shift. Characteristics of ITIC include the following: â¢ Scope: movements by O-D pair for truck, rail, and intermodal. â¢ Inputs: relies on data sources that are publicly available such as the FAF and O-D, truck and rail rates, payload factors, and other parameters such as wage rates and fees.
48 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments â¢ Outputs: modal shares, including truck, rail, and intermodal. Comparison between scenarios can be used to determine modal shifts. ITICâs advantages include its open-source Excel spreadsheet with reliance on public domain (or private) data and ease of use. The model allows the user to adjust parameters to run scenario analysis so that modal shifts due to change in a variety of parameters can be examined. ITIC can be used both for examining truck-to-rail and rail-to-truck diversions (e.g., semi-trailers to carloads, rail intermodal, and vice versa). ITICâs disadvantages include its fairly detailed input requirements. ITIC is not updated, and the 2006 user interface is incompatible with some newer operating systems. The Excel worksheets accompanying this guidebook (Appendix M) provide diversion filters for use in BCA for truck-to-rail or rail-to-truck diversions. A hybrid analysis includes market segmentation and use of modal elasticities to determine diversion potential. The conceptual analysis of truck-rail or rail-truck diversion modeling should be guided by market segmenta- tion by O-D pairs for rail and truck competitive movements. The hybrid method or FHWA ITIC model may be used in such analysis. New choice models should be reserved for later-stage analysis and for projects of state and national significance since the data input requirements are more onerous. Scenario analysis (e.g., high-growth scenario, medium-growth scenario, and low- or no-growth scenario) may be used in all contextsâconceptual or later stage. These choice models may allow for more precise assessment of consumer surplus. Diversion and Uncertainty Models of diversion are often significant sources of uncertainty and so must be scrutinized with sensitivity, scenario, and/or risk analysis. Factors such as reliance on aggregate data also contribute to risk. A Monte Carlo simulation approach is useful for this context as an alterna- tive to sensitivity and scenario analyses. This process assumes that each of the variables that contribute to diversion can be assumed to be a random variable described by a known prob- ability distributionâincluding freight volumes, growth in trade, capacity outlook forecasts at terminals, and transport cost changes. A very common distribution used for quantifiable factors is the triangle distribution. The triangle distribution has three parameters: a maximum value, a minimum value, and a most likely value (the mode of the distribution) for base and forecast years. Other distributions can be specified for specific variables. Draws from assumed distributions are then used in simulations of diversion potential. This process is very useful for all stages of an analysis, but it requires sig- nificant stakeholder inputs to refine the input values and expertise in modeling, analyzing, and reporting the results. The development of commercial tools makes it easier to apply Monte Carlo simulation methods. This approach should be used more often; the present reliance on point estimates ignores uncertainty, leading to risk of inaccurate forecasts of project benefits and costs and of suboptimal decision making. Steps to Account for Uncertainty. Regardless of the approach adopted, the following steps are necessary: 1. Obtain corridor flows and segment the corridor markets by O-D pair, direction, and com- modity type. If examining truck-to-rail or barge diversion, development of an O-D matrix (trip table) is assumed to be part of this segmentation. 2. Obtain the change in total transport cost and base mode shares. 3. Use suitable payload factors to convert the diverted freight flows to the number of move- ments. For instance, if truck tonnage is diverted, then truck payload factors can be used to estimate trucks diverted. The FAF provides default payload factors, which can be supplanted by better, more context-specific information if applicable.
Develop Forecasts 49 Analysts are required to: â¢ Maintain transparency in assumptions used. Precision in diversion can be improved by assign- ment modeling. Developing assignment for rail flows is quite complex and requires rules and routable networks. The Oakridge National Laboratory provides such a network. Due to a lack of full information on logistical choices8 for routing and modal competition and high-level data such as Transearch, the FAF, or the Waybill, solutions will likely be approximate in nature. â¢ Use sensitivity and uncertainty analyses since diversion can be influenced by assumptions, and economic and exogenous factors. Route competition by carriers can effectively limit the size of actual substitution possibilities as suggested by cross-price elasticities. Diversion is impacted by freight competition in the corridor (rail-rail, rail-truck, truck-water, rail-water, and rail-pipeline). â¢ Ensure that diversion estimation is accompanied by a realistic projection of freight flows over time. â¢ Recognize that estimates of the number of trucks diverted are used to establish the public benefits from diverted trucks based on estimated changes in safety, environmental emissions, and noise pollution as part of Step 6. â¢ Recognize that the new demand generated (âtoâ mode) is the basis for the rule of half. For example, shipping cost reductions for existing users are fully included, but only half of the benefit is attributed to the new demand. Many of the BCAs evaluated and the approaches examined as part of this project did not meet this criterion. â¢ Consider domestic demand and international-trade-related demand (flows associated with export and imports) separately. Rail-Truck-Waterway (Barge) Mode Switch. Since inland waterways and marine transport also require landside access to and from the water, rail and barge modes sometimes compete with one another. This is an area of significant uncertainty because very few studies examine new barge demand or barge traffic diverted to trucks. A multimodal regional routing model to increase annual wheat flows from the Pacific Northwest region through the Columbia-Snake river system demonstrated that even a 20% increase in wheat production only produced a 1% increase in barge traffic (35). In such cases, a cautious approach should be used, guided by the same philosophy as truck-rail rail-truck diversions, but with a much more conservative approach to projecting future freight traffic as well as to projecting the magnitude of diversion. Stakeholder input may be critical in driving these mode shift parameters. Determine Forecasting Volumes and Behavioral Effects across Scenarios and Induced Demand The aim of this part is obtaining freight volume forecasts for incremental analysis of all alter- natives compared. The data are used to develop measures of consumer surplus and producer surplus. For multimodal BCAs, reliance on consumer surplus and the rule of half for approxi- mating user benefits for both existing and induced or generated users is recommended in all conceptual analysis.9 The consumer surplus user benefit measure implies an association between changes in transport costs and times and forecasted demand on an incremental basis across the alternatives. The induced user benefits can be very significant if the project scale is very large, and existing data may not do justice to the potential for future demand. The following are some 8For instance, tools such as the Geospatial Intermodal Freight Transportation (GIFT) model developed by the Rochester Institute of Technology discusses emerging motivations for routing. 9Existing traffic is also referred to as normal traffic. Generated traffic refers to new traffic induced due to cost reduction. New or diverted freight volumes or users refer to freight diverted from other modes. The methods adopted should be able to address all three categories of freight users.
50 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments issues that must be kept in mind about forecasts and induced demand (or new user forecasting) when using different types of models or network analyses: â¢ Travel or commodity freight models with fixed trip tables: In a majority of demand models, trip tables are fixed, and mode choice redistributes existing trips by mode for the same O-Ds. This is the case with most travel models. The outputs of such models are often used for link and corridor analysis. This option is the most commonly used but is limited in its ability to capture induced demand beyond route shift. â¢ Travel or commodity freight models with variable trip tables: Advanced capabilities include the use of models with mode choice and destination choice components. Logsums can be used. Advanced methods can model new modes and/or destination shifts. This option con- siders induced demand for route and mode or destination, thereby allowing a better approxi- mation of the full consumer surplus from new users. The volumes and travel time outputs are reflective of all network effects and induced demand. This is important because ignoring induced demand and network effects in the context of large-scale improvements leads to bias. â¢ Network analysis: A conceptual analysis relying on network analysis tools to obtain travel costs does not consider induced demand. The logsum provides an alternative measure of surplus gains and losses associated with the approximation of induced demand and dealing with large-scale or non-marginal investments, but the rule of half is, and will remain, the most used and widespread tool due to its simplicity. The use of logsums is explained in Appendix D. In some situations, the logsum is recommended as a better way to approximate TEE user benefits. The caveat is that use of logsums requires that choice models be calibrated or have the capability of being calibrated for the context. They are resource intensive in terms of time, budget, and requisite analytical ability for analysis and interpretation of discrete choice models. Consider using logsum in these situations: â¢ A completely new modal option: Since the option does not exist in the reference alternative (base case), there are no existing users to refer the variation of costs to, and the rule of half cannot be applied. An estimation of the absolute value of the generalized cost is needed, which requires stated-preference surveys as future freight is difficult to predict. â¢ A very high change in the transportation cost with respect to its absolute value: This can happen when major bottlenecks or missing links are evaluated (e.g., connectors between two places that were only accessible via very long detours). The literature often suggests a thresh- old value of greater than 30% relative to the baseline. â¢ Both additive and restrictive projects as part of a bundled alternative: This leads to a situation where benefits are bestowed on some users and costs on others. In this case, it is impossible to know how much a simulated mode shift depends on the new costs for the origin mode or the benefits for the destination mode. â¢ Comparison of non-dominant alternatives (e.g., comparing a transport option that is not the best one for any of the user groups). â¢ Feasibility and investment grade studies: The development of logsums is time intensive due to data collection and detailed analysis. We recommend the use of this option if these model options are already built into travel models (as some regions and states do in United States) or use in advanced analysis. 5.3 Inputs: Recommended Tools and Data Sources A number of tools can assist the analyst with developing purpose, project alternatives, analysis time periods, and forecasts: â¢ Regional network models (travel or commodity) for regional corridor studies appropriately adjusted to account for long-haul traffic or private-domain models.
Develop Forecasts 51 â¢ Statewide network models (travel or commodity) for regional corridor studies appropriately adjusted to account for long-haul traffic (public- or private-domain models) and shipper surveys, if available. â¢ For multistate BCA, the FAF, custom-built models, O-D flows (public or private domain), and shipper surveys. Similarly, Global Insightâs Transearch may also be used although there are considerable differences between the FAF and Transearch. The Car Waybill Data (public use) version can provide base year O-D flows and container movements for freight rail. The confidential datasets may not be available without official permission. These base year flows can be forecasted. â¢ FHWA resources on Life Cycle Cost Analysis: https://www.fhwa.dot.gov/infrastructure/ asstmgmt/lcca.cfm. â¢ California Department of Transportation Life Cycle Cost Analysis Resources Using REALCOST: https://www.fhwa.dot.gov/infrastructure/asstmgmt/lcca.cfm. â¢ Modal diversion tools Table 7. â¢ Network analysis in Esri ArcView, TransCAD, or other similar tools. â¢ NCHRP Report 606: Forecasting Statewide Freight Toolkit. â¢ Oakridge National Laboratories Multimodal Networks. â¢ Cross-price elasticities (truck-rail, rail-truck, truck-water, water-truck, and others as available). â¢ Payload factors from databases such as the FAF. â¢ See the Excel worksheets in Appendix M (Step 5: DiversionâNew Users; Step 5: Diversion 1âSegmentation, and Step 5: Diversion 2âCross Price Elasticities). 5.4 Best Practices and Examples Best practices for Step 5: â¢ Clearly present the impact area used for analysis in relation to the corridor along with the appropriate network. â¢ Use suitable existing models for conceptual analysis. Custom models and private-sector mod- els may be justified for multistate corridor analysis or later stage analysis. â¢ Clearly state the assumptions included in the forecasts concerning timeline, cargo/commodity types, base year, how the alternatives are considered, and limitations if any (e.g., use of aggre- gate public or private-domain data for corridor applications such as the FAF, the Waybill, and Transearch). â¢ If a network model or commodity flow model is used, clearly discuss assumptions in the fore- cast of the effects of build and the do-minimum scenarios. This includes networks included, modes included and their treatment, assumptions and parameters included or used in route assignment, and mode choice. â¢ Base the induced demand forecast on O-D pair flows by mode for modal shift opportunities. â¢ Use assignment models to support modal analysis where possible and clearly present assump- tions on methods, values, and networks used. Examples of Forecasting Freight Volumes, Behavioral Effects, and/or Costs Example 1: The SR 167 and I-5/SR 509 Corridor projects in the state of Washington were evaluated using the Puget Sound Regional Council regional travel demand models (36). Example 2: The I-5 Portland Freight Corridor is a multijurisdictional project. It is a key north- south route for both freight and passenger movement through the Portland, OregonâVancouver, Washington, metropolitan region. The BCA used the full Portland metro regional travel model to evaluate five alternative freight strategies for the truck corridor. The Phase 1 study was
52 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments conducted for consideration of federal funding (https://casesimportal.newark.rutgers.edu/ program-evaluation) (37). Example 3: The MAROps Initial Benefit Assessment Study for the I-95 Corridor Coalition (29) defines an impact area by the scope of freight flows spanning CSXâs north-south cover- age of MAROps states in five census regions. Three states in the Southwest South Central, East South Central, and South Atlantic, and two states in the Northeast-Middle Atlantic and New England were included. Freight flows that touch (e.g., move into, out of, within, or through) the MAROps states and have an origin or destination in any one of the five regions were considered. Freight flows were used for the base year and forecast year 2025 based on Transearch. Most of the analysis was driven by conventional network analysis. Example 4: National Gateway Corridorâs Phase 1 uses truck-to-rail diversions based on a concep- tual study using a market segmentation method with point estimates of modal elasticity of 0.67 for truck-to-rail along with a rail cost reduction estimate. A similar analysis was also used in the Corpus Christi Rail Capacity Expansion project for the BNSF corridor (38). The analysis used a three-year phase-in period. Baseline corridor operations were developed from Port of Corpus Christi and Port of Brownsville statistics. The following notable assumptions were used in both studies: â¢ 17 tons per truck. â¢ Three trucks per railcar for conversion. â¢ Rail shipping cost savings of 5% due to heavier loads per railcar (after improvements). â¢ A baseline period beginning in 2010 and a first year of diversion in 2013 when the bridge improvement construction is completed. Example 5: The Virginia Department of Transportation (VDOT) commissioned a feasibil- ity study (New UsersâMode Shift) as part an environmental impact statement for a 325-mile stretch of the I-81 corridor (39). This case study is the only example that demonstrates the use of FHWAâs and the FRAâs ITIC model for diversion forecasting. The study used the following: â¢ Data sources: commodity flows from Transearch, line-haul rail variable cost data from Uni- form Rail Costing System Plan 1.0, and Truck Trip Analyzer developed by Jack Faucett Asso- ciates, which forecast flows (similar to the FAF). â¢ Inputs: heavy-truck volumes (annual average daily volumes) at various count stations (base- line and 2035 forecasts). â¢ Factors determining diversion: deterioration in truck service due to increased congestion on I-81; and the availability of improved rail service speeds, reliability, and cost reductions that result from four initial improvement concepts and improved intermodal service. â¢ Scenarios evaluated: high growth medium and long term at year 2020, and estimated values of 10.4% for traffic diverted medium term and 5.2% for traffic diverted long term. â¢ Alternatives examined: four railcar build scenarios combining rail infrastructure and roll- ing stock with costs ranging from $11 million to $3.5 billion; and no-build or do-minimum/ normal capital improvements inside and outside Virginia for railroads. Assumptions: developed in consultation with FHWA, Surface Transportation Board (STB), the Virginia Department of Rail and Public Transportation, Norfolk Southern Railroad, and others. See Appendix D for details. The corridor location and study area are shown in Figure 8 and Figure 9. â¢ Methodology, process, and assumptions: The methodology relies on primary data collection via surveys and extensive proprietary data in combination with VDOT models: â Corridor intercept surveys on truck usage within and through the 325-mile corridor. â A shipper-carrier survey aimed at identifying commodities moving along the corridor, the nature of operation at the facility, the carrier profile, truck usage, and rail usage. Surveys
Develop Forecasts 53 were targeted at major employers in the corridor, motor carriers using I-81, and freight stakeholders in Roanoke. â 2035 forecasts using traffic counts along I-81, regional economic model forecasts, 1997 Vehicle Inventory and Use Survey, Jack Faucett Associatesâ Truck Trip Analyzer, Virginia Statewide Travel Demand Model, National Transportation Atlas Database, and the 1998 VDOT Transearch database. The Transearch database included O-D flows for nine counties in West Virginia, 14 BEA zones, 11 states, and eight multistate regions. â Rail traffic from the STB Waybill sample. Figure 8. The virginia I-81 corridor study area. (vDOT) Destination Origin 37 Corridor Counties Rest of Virginia Rest of U.S. 37 corridor counties Internal truck flows (truck loads) by commodity Internal- external U.S. Rest of Virginia External to corridor Rest of U.S. U.S. external- internal External to Virginia Figure 9. The virginia I-81 truck flows O-D matrix for forecasting.
54 Guide for Conducting Benefit-Cost Analyses of Multimodal, Multijurisdictional Freight Corridor Investments â Growth factors for VDOT counts developed from truck flows by O-D and direction (truck load and less than truck load) using the Transearch proprietary database Motor Carrier Data Exchange and Freight Locator Database. Northbound flows were separated from south- bound flows, converted to truckloads using load factors, forecast to 2035, and disaggregated to the counties in the corridor. Empty trucks were handled adjusting directional forecasts. â A no-build truck O-D matrix was developed from forecasts as shown in Figure 9 (green shaded boxes showing origin location and destination locations). These were calibrated with truck counts. â For diversion, inputs from Norfolk Southern for use with ITIC (Appendix Table D4 pro- vides a discussion of ITIC use in this specific context) and selected potentially divertible commodities from Transearch. The inputs used in this study as shown in Appendix D may not be applicable to other contexts and would have to be updated. ITICâs internal parameters, however, cannot be updated. Example 6: Analysts on the Trans-Tennessee Railroad Diversion project (35) evaluated the modal diversion opportunities by developing independent modal cost functions to examine cost differences: TSC RTM TM RH H EVD THC ED (2)= ( ) + ( ) + + + where TSC = total shipment cost; RTM = transportation rate per ton-mile; TM = number of shipment ton-miles; RH = inventory value per hour; H = total transit time in hours; EVD = expected value of delay outside of performance guarantee; THC = total handling (terminal) costs; and ED = equipment differential. Three scenarios for diversion were examined assuming: â¢ A static growth scenario; â¢ Highway traffic growth; and â¢ Fuel price increases or rail network congestion. The project examined truck-rail and rail-rail container/trailer diversions for four alternatives using the following assumptions: â¢ Only locations that currently have mechanized intermodal terminals are considered. â¢ Only shipments to/from origins and destinations in the terminal county or contiguous coun- ties are considered. â¢ Movements with shipment distances of less than 600 miles are excluded. â¢ Dry-bulk commodities (coal, ore, nonmetallic minerals, etc.) are excluded. â¢ Private truck movements in shipper-owned equipment are excluded. â¢ Less-than-truckload movements are included. â¢ Intermodal movements are costed for 53-foot domestic container movements in a double- stack configuration. â¢ Assumes service improvements (transit times and reliability) sufficient to secure diversions. â¢ Assumes no additional incremental service costs. â¢ Assumes containers are already double-stacked. No further details are provided. The analysis presents six benefits included in the BCA and five non-quantifiable benefits.
Develop Forecasts 55 5.5 Common Mistakes Common mistakes occur when the project team: â¢ Uses a methodology and parameters for estimation of current and future demand that are not explicitly presented or justified. â¢ Assumes overoptimistic usersâ growth rates throughout the entire reference period of the project. Where uncertainty exists, it is wise to assume a stabilization of demand after the first few years of operation. â¢ Fails to clearly establish the link between the demand analysis and design capacity of the project (supply). (The design capacity of the project should always refer to the year in which demand is highest.) â¢ Does not discuss biases associated with induced demand and does not apply the rule of half to induced demand, leading to overestimation of user benefits. â¢ Does not clearly present forecasting assumptions and models. â¢ Does not maintain transparency in the use of diversion (induced demand) modeling. â¢ Fails to recognize that existing public-domain data such as the Waybill and FAF are not cali- brated for corridor use. â¢ Counts route diversion demand twice.