Freight Trip Generation and Land Use (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.

21 This chapter brings together a succinct characterization of the freight transportation system, its relations to the land use system and to a study of FTG. The chapter discusses three main topics: • The freight system. This section contains a comprehensive description of the freight system and its various compo- nents. A unique aspect of the discussion is that it defines and covers the different dimensions needed to fully char- acterize freight activity (e.g., function performed, modes used, geography, and the nature of the connections among the participating agents). • The relationship between the freight system and land use. This section discusses, using both empirical evidence and theory, the relationship between freight activity and land use. It is important to consider these interactions because both systems influence each other. • The differences between passenger trip generation mod- els and FTG models. This section describes and examines the similarities and differences between passenger and FTG. It also discusses the unique aspects of FTG from the perspective of economic theory and supply chain prin- ciples. These disciplines are the ones best positioned to explain the complex dynamics that determine FTG. It is important to define and explain some key concepts and technical terms that will be used throughout the report. Their consistent and proper use is of paramount importance to studying and understanding FTG. These concepts are: • Generation: In accordance to transportation planning practice (Ortúzar and Willumsen 2001), this refers to the processes that determine productions and attractions of both demand and trips. • Demand generation: These are the processes associated with the needs of passengers and freight to be transported to/from different locations. In the case of passenger trans- portation, this is measured in units of passenger trips; while in freight, typically, units of weight are used. • Vehicle-trip generation: This refers to the number of vehi- cle trips required to transport a given amount of demand. It depends on the corresponding modal split, and is typically measured in units of passenger-car-trips, and truck trips. It follows that FG is the tonnage (or volume) of freight to be transported, while FTG is the number of freight vehicles needed to transport freight. As explained later in this chapter, there are good reasons to be rigorous and consistent when using these terms. It is also important to define precisely the main focus of the analyses in this document, in terms of modes and vehicle types. This report assumes a primary focus on trucking, as this is the most important mode in terms of economic contributions and market share, and it also has the largest impact on con- gestion and environmental pollution. A second assumption is that all vehicle-types designed for and primarily used for freight purposes must be considered. This includes all trucks from small vehicles, e.g., pick-up trucks, to the largest tractor trailer truck combinations. The reason for including such a wide range of vehicles is that, while one thinks of a “freight vehicle” as a semi-trailer or a large rigid truck, large trucks are the dominant vehicle only in interstate operations. In urban and suburban operations, the vast majority of freight traffic is associated with small vehicles such as pick-up trucks, delivery vans, etc. For example, in New York City small trucks and delivery vans comprise about 90% of the urban freight traffic. There are also commercial passenger vehicles that are used to transport freight. In Denver, a survey found that 36% of passenger vehicles (autos and sport utility vehicles) registered with commercial license plates reported transporting freight (Holguín-Veras and Patil 2005). The predominance of small vehicles in urban freight clearly suggests the need to consider them when estimating and analyzing FTG; not doing so obvi- ously translates into large estimation errors. C h a p t e r 3 The Freight System, Its Purposes, and Relations to Land Use

22 For that reason, unless explicitly stated, the term “truck” refers to all vehicles designed for and primarily used for freight purposes, irrespective of their size. This definition includes dual use vehicles such as pick-up trucks that are also used for passenger travel, but it leaves out cases such as the passenger vehicles that are used to transport freight, as these represent a small portion of the total. The Freight System This section provides a comprehensive description of the freight system. Characterizing the freight system is challeng- ing because of its multifaceted and highly heterogeneous nature. In fact, it is hard to think of any other component of the transportation system that is more varied, exhibits so many fundamentally different behaviors, involves more interacting agents, is so pervasive in modern life, and is so rarely studied than freight. Because of this complexity, it is best to describe the freight system in a systematic fashion by defining the relevant dimensions that could be used to char- acterize it, and then discussing each of these dimensions in some detail. The multi-dimensional nature of the freight sys- tem poses a major challenge to simple land use classification systems because it may not be possible to characterize such complexities by a single metric. A formal characterization requires defining the following: • Interacting agents: This includes shippers, carriers, re- ceivers, warehouses, and end-users. • Links between participants: This includes independent companies and integrated companies. • Functions: This includes long-haul transportation, deliv- ery service, and parcel service. In this context, a specific operation could be character- ized by identifying where it belongs in each of the previ- ously described dimensions. Simply identifying a company as a “for-hire carrier,” for instance, does not provide enough information to characterize its operations or to understand its behavior. The following sections discuss the relevant dimensions. Multiplicity of Economic Agents Involved The first level of complexity is related to the many agents that influence the generation of freight. This is an obvious consequence of modern economies that translates into com- plex logistics and freight systems. As a result, it is useful to envision the freight system as the physical manifestation of the economy, as in most cases, monetary transactions are accompanied by a commodity flow in the opposite direction. In essence, freight activity is the economic activity in motion. As a result, to understand the generation of freight, one must have a basic understanding of the connections among various economic agents. To decompose the process and facilitate understanding, the concept of a production-consumption (PC) link is useful. A PC link represents the transaction that connects a producer of cargo with the next consumer (which could be the end user or an intermediate one that uses the cargo as an input to another PC link). In essence, a typical supply chain is com- prised of many PC links where an economic agent produces/ ships freight that other agents process/transform and store, and ultimately deliver to the end/intermediate consumers. Obviously, if the agents are not collocated, transportation has to take place. This, in turn, is what produces the vehicle trips that transportation planners and engineers capture as trip origins (O) and trip destinations (D). In simple supply chains, e.g., a farmer who sells produce to the local market, the corresponding PC pattern is straightforward. In complex supply chains, e.g., in the automobile industry, there could be hundreds of PC links corresponding to the various stages of the production process. The multiplicity of possibilities is overwhelming. A schematic of some of the possibilities is outlined in Figure 1, together with the corresponding trip origins and destinations. Therefore, to understand freight demand, one needs to study the underlying supply chains that satisfy the needs of the PC links that comprise a production and distribution process. This is because the transportation flows generated as part of these PC connections materialize into freight traf- fic, e.g., truck trips. The main focus of this research is on locations where the cargo is produced, transformed, stored, or consumed, i.e., the nodes in the transportation network. Understanding the underlying process that determines how much freight is produced or attracted at each land use is the key objective of this project. As a result, the study of the FG and FTG must consider: (1) production sites/shippers; (2) intermediate processing points, including storage; and (3) consumer sites—both end and intermediate. There are important practical reasons to be comprehensive in the study of FTG. While it is easy to identify production sites, warehouses, trucking companies, and ports as genera- tors of freight and truck trips, the role of consumer oriented businesses as generators of truck trips is frequently over- looked. The need to study FTG by service and retail busi- nesses has long been recognized as a key priority (Fischer and Han 2001). Quite frequently, and particularly in urban areas, small establishments—when taken together—produce more truck trips than any single large generator. As an example, calculations made by the team indicate that the 6,600 res- taurants and bars in Manhattan produce more truck traffic than the Port Authority of New York and New Jersey termi- nals combined.

23 There are a number of agents relevant to the study of freight transportation: shippers, carriers, receivers, third party logis- tics, freight forwarders, and warehouses/distribution centers, among others. Their roles are briefly discussed here. As pre- viously described, freight has its origin at supply points (e.g., raw material production sites or areas and manufacturing, distribution, or assembling companies, among others), and the agents who produce and ship freight are typically referred to as shippers. These shippers need to send their cargoes to their respective destinations, which requires transportation services that the shippers provide with their own assets, or with the assets of other companies hired by them for that purpose. The companies that transport the goods are known as carriers. Carriers are classified as either for-hire carriers, those that provide services to the open market, or private carriers, those that provide transportation services to a par- ent or a related company. For transportation of the freight, the shippers may contact the carrier companies directly, or they can use the services of intermediary companies, namely the third party logistics (3PL) providers, which are companies that provide logistics services for part or all of the shippers’ supply chain needs. Typically, 3PLs provide services for integrated operations, including not only transportation but also warehousing and management of the supply chain. Alternatively, the shippers may contact freight forwarders. These are a form of third party logistics providers that make use of asset-based carriers for the dispatch of shipments, either by water, ground or air, typically for international ship- ments. Freight to be transported may have as its destination: a distribution center or warehouse, retailers, wholesale traders, the end consumer, or intermediate consumers. The destina- tion agents act as receivers of the cargo. Because of the nature of business relations, the receivers typically set constraints in terms of: delivery times, technology used, and others. Distribution centers are a special case since they can serve as both receivers and shippers of cargo; at these locations, cargo received may be stored, consolidated or split up, or even post-processed or assembled. These processes can impact shipment size, which, in turn, may affect the trans- portation mode used when shipping to the next destina- tion. Other agents worthy of mention are: wholesale retailers which in some occasions may act as distributors (shippers) of the cargo; intermediate consumers which may process or conduct transformations to the cargo received and then ship it to the next destination; and finally, the end consum- ers. It is important to note that delivering cargoes to the end consumers may require additional logistical consider- ations because part of the cargo received, when consumed, may turn into waste that may require additional processing. This is what led to the development of the emerging field of Reverse Logistics. In addition to the agents just described, one can find intermodal centers where the transfers between freight modes take place. This includes airports, ports, inter- modal rail terminals, and the like, which tend to generate a substantial amount of FTG. Links Between Participants An important and frequently overlooked aspect of FTG is the nature of the links between the various agents involved in freight activity. In general terms, the participating agents could be independent companies, or they could be integrated, i.e., part of the same company, and there are other modalities. Figure 2 shows the possibilities for a case involving shippers, Production Transportation Processing, storage Consumption Shipper, producer Carrier Distribution centers, warehouses Intermediate consumer End consumer To another PC link O 1 O 2 O 3 O 5 D 2 D 3 D 5 D 4 Note: The arrows represent transportation flows, O is a trip origin, D is trip destination, and the numbers represent the physical location of the agents in a trip end. Figure 1. Production-consumption (PC) link.

24 carriers, and receivers; obviously when the number of agents increases, the number of possibilities increases exponentially. The nature of the connection between the agents is impor- tant because it determines, among other things, the propen- sity of the agents to engage in cooperative behavior. In the case of integrated operations, the parent company internal- izes benefits and costs accrued by the participants. This leads to a decision-making environment in which cooperation and accommodation take place, if it leads to better overall perfor- mance. In independent operations—where a company hires another to outsource part of the production and transpor- tation process—the propensity to cooperate is much less as each company tries to maximize its own profits, with little regard to what happens to the other. In this context, a partner in such a transaction is not inclined to cooperate with the other if doing so adversely impacts its profits (regardless of how beneficial the cooperation may be for the other partner). The data on freight behavior confirms these assertions. For instance, private carriers were found to have delivery time windows that were almost double those of for-hire carriers (Holguín-Veras et al. 2005; Holguín-Veras et al. 2006). This suggests that the perceived differences in the behavior of com- mon and private carriers reflect the different constraints that they face, and not because they follow different behavioral rules (the research conducted has failed to find evidence of behavioral differences between common and private carriers). Partial Views of the Freight System As a consequence of the many agents involved, no single agent provides a complete picture of FG. Assembling a coher- ent description of the whole process requires assembling the views provided by the composite parts, i.e., the different agents who may be aware only of those aspects that concern their operation. A summary of the information that each agent is typically aware of is shown in Table 13. Table 13 shows that producers and shippers of cargo are typ- ically aware of the characteristics of the cargo that they receive and/or ship out. However, they do not know much about what happens once the freight vehicles leave their facilities. Carriers know the details of their operations—including the loaded and empty trips produced—though, quite frequently, they are not aware of the attributes of the cargo they transport. They know who they deliver to, though they do not necessarily know who Independent Integrated Independent Integrated Receiver Carrier Shipper Independent Integrated Figure 2. Potential links for a case with a shipper, a carrier, and a receiver. Table 13. Partial views of the freight system. Notes: (1) Only of the cargo that they handle; (2) For all the cargo they receive. Freight generation: Shippers / Producers Carriers Distribution centers / Warehouses Consumers of cargo (receivers) Transportation agencies Amount of cargo Yes (1) Yes (1) Yes (1) Yes (2) No Number of loaded vehicle-trips Yes (1) Yes (1) Yes (1) Not always Number of empty vehicle-trips No Yes (1) No No Number, frequency, of deliveries Yes (1) Yes (1) Yes (1) Yes (2) No Commodity type Yes (1) Not always Yes (1) Yes (2) Only at some ports of entry Shipment size Yes (1) Yes (1) Yes (1) Yes (2) No Cargo value Yes (1) Not always Not always Yes (2) Only at some ports of entry Land use patterns Yes (1) Yes (1) Yes (1) Yes (1) All At key links (no distinction between loaded and empty)

25 The locations of the receivers are indicated by the triangles next to the circles. As shown, at stop 1 the carrier makes a delivery to a single customer, at stop 2 the carrier delivers to two custom- ers, and so forth. Upon completion of the deliveries, the carrier returns empty to the home base. Obviously, the establishments that receive these deliveries are likely to receive cargo from other vendors, who are not shown to avoid complicating the figure. Figure 3 illustrates a number of key points: (1) the origins and destinations of the individual vehicle trips (Base-1, 1-2, 2-3, 3-4, 4-5, and 5-Base) rarely match the direction of the PC relations that link the Base to each of the consumers; they are marked by dashed arrows; (2) a typical receiver does not necessarily know how many vehicle trips are generated, much less how many empty vehicle trips (this is known only to the carrier); and (3) the flow of empty vehicle trips typically runs counter to the commodity flows. The empty trips are impor- tant to consider because they could represent sizable portions of the total freight traffic. The data show that, as a percentage of total truck traffic, empty vehicle trips typically represent 20% in urban areas, 30–40% in interstate freight, and 50% of the directional truck traffic in some corridors. In terms of vehicle-miles, the numbers are equally significant; about 57% of the miles traveled by rigid trucks, and 33% of the miles traveled by semi-trailers are empty (U.S. Census Bureau 2004b). As a result of their importance, not explicitly model- ing empty trips leads to significant estimation errors, perhaps as high as 83%, contrasted with errors of 57% considering empty trip models (Holguín-Veras and Thorson 2003a). Modes and Vehicles Used The mode refers to the type of transportation technol- ogy used to transport cargo, which in turn determines the infrastructure and operational needs. The types of modes include: animal-powered transport, human-powered trans- port, air transport, water transport, rail transport, and road transport, among others. These types of modes are charac- else is delivering to a particular customer. The consumers of the cargo, i.e., the receivers, know the details of the cargo they receive/ship out, though they do not always know how many vehicle trips have been generated because many of them only observe the number of deliveries (a truck trip could be used to make multiple deliveries). Transportation agencies have an idea about truck traffic in the network and land use patterns. However, in most cases, they know very little about the freight flows in their jurisdictions. In summary, none of the agents involved in freight have sufficient information to fully describe what happens in the system as a whole. This has important implications for data collection efforts, as most surveys rely on the information gathered from the participants in the freight activity. The fun- damental challenge is how to put that information together into a comprehensive picture of FG that is relatively accurate, practical, and conceptually correct. However, from the stand- point of FG and FTG, there should be no doubt that the agents best positioned to provide the most complete view are the consumers and producers of the cargo. This is because they are the ones that know the details of the cargo they receive and/ or ship out, and the corresponding delivery frequencies, and shipment sizes. This also implies that establishment data— and the models estimated with them—are the most accurate. Multiplicity of Metrics to Define and Measure Freight Freight—and by extension its generation—can be mea- sured by many metrics. These include the value of the cargo, the amount of cargo transported, the vehicle trips produced, and the number of stops and deliveries made. Figure 3 depicts a producer that is sending cargo to nine different customers from its home base (the black circle). In transporting the cargo, the producer makes a tour comprised of six individual trips. The physical origins and destination of the trips are marked by the circles with a number, as well as the one labeled “Base.” Base 1 2 3 4 5 Loaded vehicle-trip Commodity flow Notation: Consumer of cargo (receiver) Empty vehicle-trip Figure 3. Vehicle trips, commodity flows, and delivery tours.

26 The complexity of freight transportation is increased by the many commodity types transported around the country. To group the hundreds of thousands of individual products that are transported by the freight system, commodity clas- sification codes are used. The 2007 Commodity Flow Survey (CFS) uses the Standard Classification of Transported Goods (SCTG) codes. The SCTG system has a hierarchical structure and comprises four levels (i.e., 2 to 5-digits) that aggregate the Harmonized System (HS) four or six-digit classes used worldwide for international trade. Each level of the SCTG covers the universe of transported goods, and each category, in each level, is mutually exclusive. In total, about 500 five- digit codes are considered. Table 16 lists the most important commodities; in total they represented 81.9% in value in 2007 (Bureau of Transportation Statistics 2009). The industries with higher utilization of truck-tractors with single trailers were agricultural, forestry, fishing or hunt- ing, construction and waste management, landscaping, or administrative/support services. The industry with the high- est utilization of truck-tractors with single and double trailers corresponds to the for-hire transportation and warehousing businesses (see Tables 17 and 18) (U.S. Census Bureau 2004a). Level of Geography Involved and Functions Performed The different functions performed by the various compo- nents of the freight transportation system are closely inter- twined with the level of geography. In a rather simplified terized by the infrastructure and initial investment required, vehicles and containers, type of propulsion systems, opera- tional costs, and capacities and speeds, among other factors. In the United States, according to the 2007 Commodity Flow Survey, 93% of the tonnage (representing 81.6% of the value) is transported by a single mode (i.e., truck, rail, water, air, and pipeline). Figure 4 shows the breakdown of freight tonnage by truck, rail, water, air, and pipeline for these modes. The remainder is transported using multimodal combinations. As shown in Figure 4, truck is the mode with the larg- est market share; 75% of commodities were transported by trucks in 2007. There are emerging trends in terms of resur- gence of rail freight (including extensive use of trailer-on-flat cars and containerization) and the use of ports and rail-truck intermodal centers as major good transfer points. However, the focus of this report is on truck transportation. Table 14 describes the breakdown between for-hire (common) trucks and private trucks. It should be noted that while private truck delivers a larger share of the tonnage, for-hire trucks trans- port a larger share of the value. This suggests that private trucks are transporting cargoes with values that are lower than those transported by for-hire carriers. In addition to the single modes, there are multimodal freight transportation alternatives that use combinations, such as those presented in Table 15. Among those, the data show that more than 80% of the value transported uses parcel (e.g., packages), United States Postal Service (USPS) or courier, while this group rep- resents only 6% of the tonnage transported by multimodal modes (Bureau of Transportation Statistics 2009). Source: Adapted from Bureau of Transportation Statistics (2009). Figure 4. Breakdown of shipments by mode used. Source: Adapted from Bureau of Transportation Statistics (2009). Value (million $) Value % Tons (thousands) Tons % For-hire truck 4,955,700 59% 4,075,136 46% Private truck 3,380,090 41% 4,703,576 54% Total 8,335,790 8,778,712 Table 14. Distribution of trucking company ownership structure. Source: Adapted from Bureau of Transportation Statistics (2009). Multi modal Value Tons Parcel, USPS or courier 84% 6% Truck and rail 10% 39% Truck and water 3% 25% Rail and water 1% 10% Other multiple modes 2% 20% Table 15. Multiple modes of transportation.

27 Source: Adapted from Bureau of Transportation Statistics (2009). Commodity Type Value($mil) Value % Tons (thous) Tons % Mixed freight 7,303,091 8.48% 2,384,804 2.57% Electronic & other electrical equip & office equip 7,248,316 8.42% 350,589 0.38% Motorized and other vehicles (including parts) 6,530,511 7.58% 987,405 1.07% Pharmaceutical products 5,602,658 6.51% 134,107 0.14% Gasoline and aviation turbine fuel 4,800,101 5.58% 6,900,173 7.44% Machinery 4,675,732 5.43% 514,479 0.55% Other prepared foodstuffs and fats and oils 3,756,333 4.36% 3,643,446 3.93% Base metal in prim. or semifin. forms & in finished shapes 3,711,035 4.31% 2,757,242 2.97% Plastics and rubber 3,675,205 4.27% 1,366,540 1.47% Miscellaneous manufactured products 3,481,277 4.04% 705,856 0.76% Textiles, leather, and articles of textiles or leather 3,451,697 4.01% 354,814 0.38% Articles of base metal 2,935,810 3.41% 1,024,423 1.11% Fuel oils 2,715,113 3.15% 4,656,411 5.02% Chemical products and preparations, nec 2,536,182 2.95% 949,627 1.02% Meat, fish, seafood, and their preparations 2,192,327 2.55% 775,578 0.84% Precision instruments and apparatus 2,029,925 2.36% 36,657 0.04% Coal and petroleum products, nec 1,918,659 2.23% 4,185,478 4.51% Basic chemicals 1,909,600 2.22% 2,917,137 3.15% Sub-Total 70,473,572 81.85% 34,644,766 37.37% Table 16. Largest commodity groups. Source: Adapted from U.S. Census Bureau (2004a). Vehicular and operational characteristics 2002 trucks (thousands) Trucks (%) 2002 truck miles (millions) 2002 average miles / truck (thousands) BUSINESS For-hire transportation or warehousing 1,280.2 1.5 72,272.8 56.5 Vehicle leasing or rental 859.2 1.0 20,024.6 23.3 Agricultural, forestry, fishing, or hunting 2,239.9 2.6 24,120.0 10.8 Mining 177.6 0.2 3,411.5 19.2 Utilities 679.3 0.8 10,244.7 15.1 Construction 4,541.5 5.3 75,906.2 16.7 Manufacturing 782.9 0.9 15,384.5 19.7 Wholesale trade 735.9 0.9 16,963.5 23.1 Retail trade 1,530.5 1.8 27,470.5 17.9 Information services 376.6 0.4 5,622.0 14.9 Waste management, landscaping, admin/support 743.2 0.9 10,709.3 14.4 Arts, entertainment, or recreation services 187.1 0.2 1,784.1 9.5 Accomodation or food services 284.3 0.3 5,816.3 20.5 Other services 2,127.3 2.5 35,776.2 16.8 Personal transportation 65,343.0 76.7 766,639.8 11.7 Not reported 1,308.2 1.5 20,820.7 15.9 Not applicable 1,978.1 2.3 1,761.3 0.9 Total 85,174.8 100.0 1,114,728.0 Light 79,759.6 93.60 969,104.3 12.2 Medium 1,914.0 2.20 26,255.6 13.7 Light-heavy 910.3 1.10 11,765.7 12.9 Heavy-heavy 2,590.9 3.00 107,602.4 41.5 Total 85,174.8 100.0 1,114,728.0 Pickup, minivan, other light vans, and sport utility 79,638.4 93.5 Flatbed, stake, platform, and low boy 1,192.4 1.4 Van 1,703.5 2 Service, utility 255.5 0.3 Van, step, walk-in or multistop 425.9 0.5 Dump 851.7 1 Tank for liquids, gases, or dry bulk 340.7 0.4 Other and not applicable 766.6 0.9 Total 85,174.8 100.0 BODY TYPE VEHICLE SIZE Table 17. Truck, truck miles, and average miles per truck (2002 VIUS).

28 The key combinations of geography and functions per- formed are shown in Table 19. As shown, some functions (e.g., urban deliveries, service) are predominantly urban endeavors, while others (e.g., long-haul, parcel service) touch all levels of geography. On a conceptual basis, the freight traf- fic in urban areas includes the four major vehicle types shown in the table. Urban deliveries, parcel service and USPS, and way, one could identify the following functions and levels of geography: • Functions: urban deliveries; long-haul transportation (trucking, air, maritime); parcel service; and USPS. • Levels of geography: urban; regional; interstate/national; and international. Source: Adapted from U.S. Census Bureau (2004a). Vehicular and operational characteristics 2002 trucks (thousands) Single-unit trucks and truck-tractors without trailer Single- unit truck with trailer Truck- tractor with single trailer Truck- tractor with double trailer BUSINESS For-hire transportation or warehousing 1,280.2 574.1 21.7 637.9 45.1 Vehicle leasing or rental 859.2 781.8 0.5 76.7 Agricultural, forestry, fishing, hunting 2,239.9 1,857.5 217.6 156.3 8.5 Mining 177.6 145.9 7.4 23.2 1.1 Utilities 679.3 614.8 56.8 7.5 Construction 4,541.5 4,208.6 222.9 104.4 5.5 Manufacturing 782.9 706.3 15.6 60.1 0.9 Wholesale trade 735.9 654.8 12.6 66.7 1.8 Retail trade 1,530.5 1,411.8 55.7 62.3 0.7 Information services 376.6 375.1 0.8 0.7 Waste manag., landscaping, admin/support 743.2 565.0 159 18.8 0.4 Arts, entertainment, or recreation services 187.1 166.6 16.9 3.5 Accomodation or food services 284.3 261.7 1.1 20.6 0.8 Other services 2,127.3 2,094.4 23.3 9.6 Personal transportation 65,343.0 64,497.2 845.1 0.7 Not reported 1,308.2 1,190.4 45.4 71.1 1.3 Not applicable 1,978.1 1,978.0 Total 85,174.8 82,084.0 1702.4 1320.1 66.1 Table 18. Type of trucks used for industries (2002 VIUS). Level of geography Urban deliveries Service related Parcel service and USPS Long haul (trucking, rail, air, maritime) Urban Predominantly small trucks Predominantly small trucks Predominantly small trucks Transport large volumes of cargo to intermodal sites, distribution centers, or large businesses Regional Not applicable Predominantly small trucks Predominantly midsize and large trucks Predominantly midsize and large trucks Interstate / National Not applicable Rarely done Predominantly large trucks Predominantly large trucks International Not applicable Rarely done Predominantly large trucks, air Predominantly large trucks, rail, air, and maritime Functions performed Table 19. Key combinations of levels of geography and functions.

29 freight operations. A summary of the key features is shown in Table 20. In general, urban freight operations differ from freight oper- ations in states, regions, or nations. For the urban case, freight movement is performed almost completely by road, since other modes have shown to be inefficient in urban areas (Ogden 1992). Urban deliveries are composed of short-distance move- ments and multiple stops made on one tour, which normally starts and ends at the warehouse. In most cases, small trucks are the ones used for urban deliveries to consumer oriented establishments (e.g., retail, food) which typically have major constraints on storage space. As a result of the large traffic of small trucks, freight activity produces a significant amount of congestion. There is also a sizable amount of cargo, e.g., bread, that is locally produced and transported using small trucks. In terms of tour lengths, there are obvious differences. In urban areas, long tours are the norm; in New York City and Denver, the average number of delivery stops is about 5.5 per tour (Holguín-Veras and Patil 2005; Holguín-Veras 2006). In regional, interstate, and international travel, there typically is one delivery stop per tour or two at most. The percent of empty trips generated is also very different. In urban areas, because of the long tours and better chances of getting cargo for the return trip, the percent of empty trips is typically about 20% (Strauss-Wieder et al. 1989; Holguín- Veras and Thorson 2003a; Holguín-Veras and Thorson 2003b). In regional, interstate, and international travel—where the imbalances of trade tend to be more pronounced—the empty service trips are typically made using small trucks. In con- trast, the long-haul flows arriving to the area by air, water, rail, or trucking, tend to use large capacity freight technol- ogy. In cases where these flows arrive at intermodal sites (e.g., airports, rail terminals, ports), the cargo is usually picked up (or delivered in the case of outbound shipments) using large trucks that typically transport the cargo to distribution cen- ters or warehouses in the vicinity of the urban area. A subject deserving further research is the quantification of service trips, particularly in service areas, and the amount of freight that they transport. This is an increasingly large por- tion of the commercial vehicle market as modern economies are predominantly based on the service sector, and includes activities such as repair of photocopiers, maintenance of office equipment and computers, and the like. At the other end of the spectrum, one finds long-haul transportation that typically connects distant manufactur- ing and consumer sites, distribution centers, and warehouses where the cargo is stored and re-processed, and major inter- modal sites where the cargo is transferred to another mode for domestic and international shipping. These operations are very different from urban deliveries as their emphasis is on transporting large volumes of cargo to a relatively small set of destinations. Shipment size increases with shipment distance (Holguín-Veras 2002), which translates into urban areas having smaller shipment sizes than regional, inter- state/national, and international transport. The level of geography has a direct impact on the characteristics of the Level of geogra- phy Predominant vehicle/mode Shipment size Congestion impacts Empty trips Number of deliveries per tour Nature of clients Urban Small trucks Small, frequent deliveries High Typically about 20% 5-6 Predominantly consumer oriented goods Regional Midsize and large trucks Larger shipments Typically, only an issue at specific bottlenecks Typically about 30-40% 2-3 Mix of manufacturing and consumer oriented goods Interstate / National Midsize/large trucks, rail, and air (high priority goods and parcel) Large shipments Typically, only an issue at specific bottlenecks Typically about 30-40% 1-2 Mix of manufacturing and consumer oriented goods Interna- tional Large trucks, rail, air (high priority goods and parcel), and maritime Large shipments Typically, only an issue at specific bottlenecks Typically about 30-40% 1-2 Mix of manufacturing and consumer oriented goods Functions performed Table 20. Key features of freight activity by levels of geography.

30 it is important to consider a broader graphical scale which could be at the national or even international level. Because of the involved geographical scale, most national and interna- tional freight movements use several modes, especially when origins and destinations are far apart, and also the inclusion of different types of stakeholders. Differences Between Passenger, Freight Generation, and Freight Trip Generation To fully appreciate the differences between passenger and FTG, it is important to make a clear distinction between the generation of demand (e.g., passenger trips, tons) and the generation of traffic (e.g., car trips, truck trips). While most analysts agree that this distinction is of minor importance in passenger transportation, there is a great difference in freight transportation. The reason is related to the degree of corre- spondence between demand generation and trip generation. In the passenger case, there is a fairly tight correspondence between the amount of trips produced and the associated number of vehicle trips—particularly in areas where tran- sit’s share is small—because car occupancies are relatively low, hovering around 1.1–1.2 passengers/car. In contrast, in freight transportation, many businesses could dramati- cally change their shipment sizes (in some cases by several orders of magnitude) to minimize their total logistic costs. As a result, the tight correspondence that exists in passenger transportation between demand and traffic disappears in the case of freight. Accordingly, one cannot assume propor- tionality between FG and FTG. In terms of the underlying factors that determine the generation of demand, there are differences as well. Also, passenger trips are produced mainly at the household level—determined essentially by the socio-economic charac- teristics of the individual and the household. Freight trips are produced at the farms, factories, and transshipment points, and delivered (attracted) to shops, offices, business areas, etc. They are determined by the establishment characteristics, and by such dynamics as inventory policy and total logistic costs. The key similarities and differences can be summarized as in Table 21. Table 21 implies that it is imperative to treat the genera- tion of freight demand and the generation of freight vehicle trips as two different concepts. Because businesses have the power to significantly change the sizes of the shipments that they ship out or receive, the FTG is not directly proportional to the FG. As a result, large businesses generate proportion- ally less FTG than do small businesses, as they handle larger shipment sizes. A second complication is that because of the indivisibility of the vehicle-trip, small businesses that receive small amounts of cargo may generate a disproportionally trips could fall between 30% and 40%, and in some corridors up to 50% (Holguín-Veras and Thorson 2003a; Holguín- Veras and Thorson 2003b). The geographic scales also impact the nature of the plan- ning process. When analyzing the movement of goods or peo- ple it is important to characterize the relationships between the spatial constraints and attributes with the origin, desti- nation, quantity, nature, and purpose of the movements. In the same way, when analyzing transportation, interrelations between networks, nodes, and demands, it is also important to consider the factors that affect these components accord- ing to their spatial function. An important factor of these relations is their scale, that is, if the transportation system is established over urban/suburban, regional, national, or global geographies (Rodrigue et al. 2006). Similarly, for planning considerations of geographical scales, differences between the scopes and limitations of each decision appear. According to the Metropolitan Washington Council of Governments and their Transportation Planning Board (National Capital Region Transportation Planning Board 2010), at the regional scale, decisions are made about where and how the city will grow in the long-term. Trans- portation priorities set at this scale shape future decisions at the city, corridor, and site scale. Regional planning decisions bring together a wide range of stakeholders, from local gov- ernments to grassroots advocacy groups. Regional resources for connecting transportation and land use include long- range plans, but also technical resources on addressing issues of concern for the whole region, such as affordable housing. At the city or corridor scale, freight and transit corridors are planned at the scale of a corridor, which may involve multiple local jurisdictions. Cities and counties make decisions about where new development should occur and which areas are in need of revitalization as they prepare land use and transpor- tation plans to guide long- and short-term growth. Trans- portation and land use decisions at this level involve many stakeholders, from local jurisdictions and transit agencies to neighborhood groups and individual citizens. Detailed plans for land use and transportation are often made at the neighborhood scale. Station area plans, sector plans, and streetscape plans are all implemented at the neigh- borhood scale. Decisions about the intensity of new develop- ment or the character of key streets impact the neighborhood as a whole. Neighborhood-scale planning can also incorpo- rate planning for community benefits, such as affordable housing. Transportation and land use connections are ulti- mately implemented at the scale of individual sites. Devel- opment connects with surrounding streets, transit stations connect with public spaces and surrounding buildings, and streets create the framework for development. In addition, and considering that about half of global trade takes place between locations of more than 2,000 miles apart,

31 floor area, and non-office employment (Bartlett and New- ton, 1982) can also be factors. Some studies have produced trip rates for: different types of land use and/or vehicle types (Zavattero and Weseman 1981; Middleton et al. 1986; Tadi and Balbach 1994); special facilities, more specifically ports (Guha and Walton 1993; Wegmann et al. 1995; Al-Deek et al. 2000; Al-Deek 2001; Holguín-Veras et al. 2002; Wagner 2010); and warehouse trip productions (DeVries and Der- misi 2008; Orsini et al. 2009). Table 22 contains a summary of the key results. In addition, different methodologies have been used to esti- mate truck trip rates. Garrido (2000) used time series data to develop estimates of productions and attractions. Input- Output (IO) coefficients, data from the 1993 CFS, and average load factors were used by Sorratini (2000) to estimate truck flows for the state of Wisconsin. However, as shown in Table 22, only a subset of these factors has been empirically tested to assess their statistical significance. The latter is important because it is the only scientific way to determine if an attribute could be considered a valid explanatory variable of FTG. The Role of Shipment Size The decision about cargo shipment size is without any question one of the most important in freight transporta- tion because it directly impacts both FTG and mode/vehicle choice. Both aspects are discussed in this section. Impacts on Mode/Vehicle Choice The process of mode/vehicle choice is one of the most complex aspects of transportation modeling. The discussion here considers the process of vehicle type choice—as well as mode choice—because of their importance to freight trans- portation planning. The reason is that, since freight vehicles high amount of FTG (e.g., delivering one small box requires a truck trip, which is the same needed to transport five boxes of the same product). These elements lead to a situation in which the FTG depends on business size, with some large companies producing proportionally less FTG than small ones. Furthermore, the available data for some industry seg- ments show that there is only a weak connection between the number of vendors that deliver goods to establishments and the establishment size. In essence, the average number of vendors is about the same irrespective of size (see section on input required for business operations and Appendix B for some empirical results). As a result, the trip rates for small businesses are typically much larger than the ones for large businesses; trip rates are six times larger in the case of estab- lishments in the Wholesale Trade: Durable Goods industry (see Appendix C for estimates). Yet no such discordance hap- pens in passenger trip generation. This has major implica- tions for modeling because the use of constant trip genera- tion rates is bound to lead to huge errors in the estimation of FTG for certain activities. These issues, together with Table 21, are further discussed in subsequent sections. Attributes That Influence Freight Trip Generation As mentioned before, FTG is influenced by factors and attributes somewhat different from those that affect pas- senger trip generation, and freight generation. The factors and attributes mentioned in the literature include land use (Brogan, 1980; Jack Faucett Associates, 1999), and economic activity at the study area (Cambridge Systematics Inc., 1996). Combinations of company attributes, such as: employment and business area (Iding et al., 2002); industry segment, commodity type transported and employment (Bastida and Holguín-Veras, 2009); total employment, site area, gross Characteristic Passenger Freight Demand generated Passenger trips Tons produced or consumed at a given location Traffic generated Car trips, bus trips, etc. Truck-trips, van-trips, etc. Influencing variables Income, land use, family structure, car ownership, activity concentration Economic activity performed, line of business, business size, land use Correspondence between demand and traffic generated Very tight, almost one to one in areas where transit share is low Very loose due to: (1) the role played by the shipment size that leads to a situation where large businesses, while generating large amount of cargo, produce proportionally less traffic because of their large shipment sizes; and (2) the indivisibility of freight-trips, that translates into small businesses generating proportionally large freight trip generation in relation to the demand generated. Table 21. Passenger vs. freight trip generation.

32 One of the sources of complexity is that mode/vehicle choice is impacted by the interactions between shippers and carriers, and carriers and receivers, and the decisions they make concerning shipment size and frequency. Although it is frequently assumed that freight mode/vehicle choice is a decision solely made by the carrier—which seems rea- sonable as it parallels the behavior observed in the pas- senger case—the reality is that the agent that decides on the shipment size—whether shipper or receiver—also has a major impact on the choice of best mode/vehicle. This has been clearly established by the independent work of two Nobel Prize winners in Economics (Samuelson 1977; McFadden et al. 1986), and confirmed with the assistance of econometric models (Abdelwahab and Sargious 1991; are very heterogeneous, using an average vehicle, e.g., the average truck, leads to major distortions in the analyses. This is because such a generic vehicle unit cannot adequately rep- resent the range of capacities and operational factors of all of those between the smallest and the largest vehicles in the category. At one end, one finds pick-up trucks with typically one ton capacity and at the other end, truck combinations with load capacities that could exceed 50 tons. Obviously, it is not possible to represent such a dissimilar group of vehicles using an “average” class. This is no trivial matter, as the traffic of small freight vehicles is a significant portion of the freight- related traffic, particularly in urban areas. For that reason, the mode/vehicle choice process is one of great import to urban transportation planning, and sustainability efforts. Tested Signi- ficant Employment Yes Yes Brogan, 1980 Building area Yes Yes Tadi and Baldach, 1994 Establishments No Iding et al., 2002 Trip Purpose Employment Yes Yes Brogan, 1980 Employment Yes Yes Zavattero and Weseman, 1981; Brogan, 1980 Land use Yes Yes Zavattero and Weseman, 1981 Vehicle Type No Middleton et al., 1986 Employment Yes Yes Iding et al., 2002; Bartlett and Newton, 1982 Business area Yes Yes Iding et al., 2002; Bartlett and Newton, 1982 Establishments No Iding et al., 2002 Gross floor area Yes Yes Bartlett and Newton, 1982 Non-office employ. Yes Yes Bartlett and Newton, 1982 Commodity type Yes Yes Bastida and Holguín-Veras, 2009 Type of Business Yes Yes Bastida and Holguín-Veras, 2009 Industry Segment Yes Yes Bastida and Holguín-Veras, 2009 Industry Segment Yes Yes Bastida and Holguín-Veras, 2009 Type of Business Yes Yes Bastida and Holguín-Veras, 2009 Fleet size Company SIC Yes Yes Bastida and Holguín-Veras, 2009 Area Yes Yes Holguín-Veras et al., 2002 TEUS Yes Yes Holguín-Veras et al., 2002 Boxes Yes Yes Holguín-Veras et al., 2002 Daily total vessels Yes No Al-Deek et al., 2000 Gross tons Yes No Al-Deek et al., 2000 Gantry crane activity Yes Yes Al-Deek et al., 2000 Day of the week Yes No Al-Deek et al., 2000 Imported freight units Yes Yes Al-Deek et al., 2000; Al-Deek, 2001 Exported freight units Yes Yes Al-Deek et al., 2000; Al-Deek, 2001 Exported commodity type Yes Yes Al-Deek et al., 2001 Imported commodity type Yes Yes Al-Deek, 2001 Exported com. tonnage Yes Yes Al-Deek, 2001 Imported com. tonnage Yes Yes Al-Deek, 2001 Employment Yes Yes Bastida and Holguín-Veras, 2009 Commodity type Yes Yes Bastida and Holguín-Veras, 2009 Sales Yes Yes Bastida and Holguín-Veras, 2009 Industry Segment Yes Yes Bastida and Holguín-Veras, 2009 Vehicle Type Land Use Number of Truck Drivers Carrier vs. Receiver, Attraction vs. Production Ports Industry Sector Employment Stratification Factor Source Statistically Table 22. Summary of findings concerning attributes influencing FTG.

33 From T*, one could find the optimal frequency f* as: f T hD K Inventory Cost Demand per unit * * = = = ( ) 1 2 time Setup Cost ( ) ( )2 3( ) The relationships between these variables is summarized in Table 23. The table provides key insights on the elements that an ideal FTG model should capture, though at an appropri- ate level of detail. The following observations are important: • The type of economic activity performed by an establish- ment is very important as it determines the order costs, the amount of cargo (FG, or demand) to be transported, and inventory costs (storage + opportunity cost). • Businesses with large order costs (setup + transportation), in equality of conditions, tend to receive larger quantities of goods more spaced in time than other businesses. • Firms handling large volumes of cargo are likely to require, in equality of conditions, larger orders more often than other establishments. • Establishments with large inventory costs (e.g., limited storage space, handling perishable goods) are going to require frequent shipments of relative small orders. These variables provide conceptual support for the devel- opment of FTG models based on data collected at the estab- lishment level because they are the ones with direct knowl- edge of the freight deliveries they get/produce. Equations 1, 2, and 3 clearly show the difference between FG and FTG. While the freight generation is represented by D, the FTG is the product of the number of vehicles to trans- port Q* (in most cases, only one vehicle), times the delivery frequency f*. If the business is receiving goods from multiple vendors, the total FTG would be equal to the summation of the vehicle trips generated by all vendors. The results also imply that an increase in FG would be accompanied by a less than proportional increase in FTG Holguín-Veras 2002), analytical formulations (Baumol and Vinoud 1970; Hall 1985) and economic experiments (Holguín-Veras et al. 2009). In the words of Samuelson: “. . . the relevant transportation choice . . . is not simply a choice between modes, but a joint choice of mode and shipment size. In most cases, the shipment size is practically mode determining. . . . Hence, it follows that in freight demand modeling, shipment size and mode choice should always be modeled jointly.” (Samuelson 1977, 118–119). In other words, the carrier’s decision about mode is conditioned by the decision of shipment size. Adding to the complexity, the decision about the shipment size could be made by either the shipper or the receiver, depending on which one has more market power. Freight mode/vehicle choice also depends on other factors— more in line with the passenger mode choice practice—such as the economic attributes of the cargo (e.g., commodity type, cargo value, degree of perishability, size), as well as the char- acteristics of the competing modes/vehicles (e.g., cost, travel time, reliability, probability of cargo damage). Taken together, freight mode choice is an extremely complex subject that is currently poorly understood. Impacts on Freight Trip Generation It is rather obvious that, in equality of conditions, business size influences the amount of freight generated as the larger the business, the larger the amount of cargo it is expected to handle. In other words, the larger the business the larger its FG. However, it turns out that the FTG, i.e., the number of vehicle trips generated, is not entirely determined by the amount of freight transported, as FTG also depends on other key aspects such as inventory policy and logistic costs. Understanding the factors that determine FTG requires the use of inventory theory. A good way to start is the Eco- nomic Order Quantity (EOQ) model (Harris, 1915), as it is relatively simple and provides a conceptually solid depiction of the problem. The EOQ model considers a business that needs a certain amount of cargo, and wants to determine the combination of shipment size and delivery frequency that minimizes the total logistic cost (transportation plus inven- tory costs). Under simple assumptions (Simchi-Levi et al., 2005) the optimal order quantity (shipment size), Q*, and time between orders, T*, can be obtained as: Q KD h Setup Cost Demand per unit time * = = ( )( )2 2 Inventory Cost ( )1 T K h Setup Cost Inventory Cost Demand * = = ( ) ( ) 2 2 per unit time( ) ( )2 Effect: Optimal order quantity (shipment size) Q* Delivery frequency f* Order cost K increase Increases Decreases Demand D increase Increases Increases Inventory cost h increase Decreases Increases Table 23. Relationships among key variables.

34 ibility of the vehicle-trip (which forces a minimum number of trips regardless of the amount of cargo), and the effect of increasing shipment sizes that proportionally reduce the FTG for larger establishments. At some point, however, the increases in shipment size lead to a situation in which a change of vehicle/mode is warranted because either the smaller vehicle cannot handle it, or because it is more eco- nomical to use a larger vehicle. This produces the pattern illustrated in Figure 5 where the increase in shipment size leads to a vehicle/mode change that produces a drop in the freight traffic. Number of Inputs Required in Business Operations, Indivisibility of Truck Trips An important factor reflects the relationship between the number of inputs required by a production process, its rela- tionship to business size, and the indivisibility of vehicle trips. As is widely known, businesses need different inputs to con- duct their economic activities. The delivery of these inputs, together with the shipments produced by the firm, deter- mines the FTG. However, the (limited) data available suggest that firms of different sizes in the same line of business tend to require about the same number of inputs, though large establishments are likely to need larger amount of cargo, and thus larger shipments. In a context of economic specializa- tion—where vendors/suppliers specialize in specific seg- ments and are not prone to consolidate shipments with other vendors for fear of losing their customers—most vendors end up sending their vehicles to deliver even the smallest of shipments. As a result, the small amounts of cargo delivered to small businesses require proportionally larger amounts (which is confirmed by the empirical evidence). As an example, consider the case of a business with K = $1/order, h = $1/ item, and D = 1 item/hour. Applying equations 1 and 3 leads to an optimal shipment size of Q* = 2 items/shipment, and delivery frequency f*=1/ 2 = 0.707. Consider now a similar business with exactly the same characteristics (with K = $1/ order, h = $1/item) but with a demand that is four times larger (D =4 items/hour). In this case, the optimal shipment size is Q* = 8 = 2 2 items/shipment, and the optimal delivery frequency becomes f* = 2 = 1.41. These results show that, contrary to intuition, the larger business does not generate four times the FTG as the small one. Instead, the increase in the FG is handled by smaller increases in both shipment size and delivery frequency. As a result of this, a four-fold increase in FG leads to only double the FTG. This means that as shipment size increases as the busi- ness increases, it may prompt a change in the freight vehicle/ modes used towards those with larger capacity. This, in turn, could produce a drop in the FTG. In essence, FTG is inter- twined with the process of vehicle/mode choice, in the same way that in passenger transportation, the generation of traffic is determined by the corresponding mode choice process. Conceptually, the relationship between FG, FTG, and size (in this case measured by employment) generally follows the pattern shown in Figure 5. As shown, larger businesses are expected to generate more cargo than smaller ones. Furthermore, since they are likely to be more productive, large businesses may generate more cargo per employee than small establishments. The FTG is a different matter altogether. In the case of FTG, in proportion to their size, small businesses are expected to generate proportionally more vehicle trips than large ones. This reflects the indivis- Small trucks are used Large trucks are used Increase in shipment size lead to a change in vehicle type Figure 5. Conceptual relation between freight generation, freight trip generation, and size.

35 2. Small businesses tend to produce proportionally more FTG than large ones. To see why this is the case, consider the case of the inflow of cargo to a business (a relatively similar analysis applies to the outflow). A small business typically needs about the same number of different inputs as a large establishment in the same line of business. The fundamental difference is that the small business needs much smaller amounts of cargo. However, since the cargo is likely to be provided by different vendors, and the truck trip is indivisible, small shipments require about the same number of truck trips as larger ones, e.g., transporting one box of a product requires the same number of trips as does five boxes. As a result, in proportion to its size, small busi- nesses produce more FTG than large businesses. In the case of the SIC 50 Wholesale Trade: Durable Goods Indus- try (see Appendix C), the empirical evidence suggests that FTG rates for small businesses are about six times larger than the ones for large establishments. 3. Though FG increases with size, FTG often does so at a slower rate. In real life, businesses schedule deliver- ies to minimize their total logistic costs. Therefore, they select the combination of shipment sizes and delivery frequencies that minimizes the summation of inventory plus transportation costs. This means that, the transport of more cargo might be accomplished by a dual increase in shipment size, and the corresponding delivery fre- quency. As a result of the increase in the shipment size, the increase in the delivery frequency—which is what increases the FTG—is typically smaller than what would have been expected if the shipment size had remained constant. This explains why large establishments produce proportionally less FTG than smaller ones. The empiri- cal evidence from the literature confirms this assertion. Furthermore, since large shipment sizes can use larger vehicles or more efficient freight modes, it is also likely that the FTG could be smaller than the one for a small establishment (though the vehicles would be larger). of FTG than those for large establishments that typically use large shipment sizes. Figure 6 shows a plot of the number of vendors vs. num- ber of employees for a sample of establishments in New York City. As shown, there is no discernible pattern between these two variables. In essence, the number of vendors that deliver products to these establishments is statistically constant (sta- tistical analyses confirmed this result). Additional results for other industry sectors, and the corresponding statistical anal- yses, are shown in Appendix B. The ability of large businesses to absorb large shipment sizes along with the relative constancy of the number of inputs required by establishments of different sizes translates into small establishments proportionally generating more FTG than large ones. As a result, it follows that using constant FTG rates as a function of size variables, e.g., employment, may lead to significant errors in the estimation of FTG. Summary This section provides a summary of the insights from the analyses reported in this chapter concerning FG and FTG. The analyses lead to the following findings: 1. The FG at a particular business increases with business size. To a great extent, this correspondence in growth is to be expected from a competitive market that weeds out inefficient businesses. Under basic conditions of effi- ciency, the larger the input the larger the output. Further- more, large businesses are expected to produce propor- tionally more cargo per unit input than small businesses because, under normal conditions, they should be able to take advantage of scale economies and the like. However, in most economic processes the amount of land available may act as a constraint, not an input, thus limiting the ability of land use variables to explain FTG. 1 10 V en do rs 100 0 10 20 30 40 50 60 70 80 90 100 Employment Figure 6. Number of vendors vs. employment (retail trade).

36 industry sectors exhibit this kind of behavior for freight trip production, and another 11 out of 21 do so for freight trip attraction. The remaining industry sectors exhibit increasing FTG with business size. Among those that exhibit variable FTG, the following variables have been found to play a statistically significant role: industry segment, employment, sales, commodity type, and square footage. It should be mentioned that industry segment has only been tested in Europe, as no studies in the United States have tested its significance. However, since employment and industry segment may be correlated, one could expect that in those industry sectors where employment was found to play a role, the industry segment could work as well. 4. Both FG and FTG rates depend on business size. As a consequence of the findings described herein, FG and FTG rates depend on business size. Moreover, when nor- malized by a business size variable, the FTG rates for small businesses are significantly larger than those for large businesses. As a result of this pattern, if a constant FTG rate is used, the FTG of small businesses would likely be underestimated, and the one for large busi- nesses overestimated. 5. Variables that influence FTG. The review of the literature confirms most of the assertions of Findings 1 through 4. The analyses of the sparse data available indicate some industry sectors exhibit FTG that does not increase with business size. In the case of New York City, 7 out of 12