USF Libraries
USF Digital Collections

Development of structural equations models of statewide freight flows

MISSING IMAGE

Material Information

Title:
Development of structural equations models of statewide freight flows
Physical Description:
Book
Language:
English
Creator:
Jonnavithula, Siva S
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
sensitivity analysis
ADF-WLS estimation
employment
population
commodity groups
data requirements
Dissertations, Academic -- Civil Engineering -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: The modeling of freight travel demand has gained increasing attention in the recent past due to the importance of efficient and safe freight transportation to regional economic growth. Despite the attention paid to the modeling of freight travel demand, advances in modeling methods and the development of practical tools for forecasting freight flows have been limited. The development of freight demand models that incorporate the behavioral aspects of freight demand face significant hurdles, partially due to the data requirements, which are a consequence of the inherent complexity of the mechanisms driving freight demand. This research attempts to make a contribution in this context by proposing a relatively data simple, but behaviorally robust statewide modeling framework for the state of Florida, in the spirit of an aggregate level four-step planning process. The modeling framework that is developed in this research can be applied to the modeling of freight travel demand using data contained in readily available commercial databases such as the Reebie TRANSEARCH database and the InfoUSA employer database. The modeling methodology consists of a structural equations modeling framework that can accommodate multiple dependent variables simultaneously. This framework predicts freight flows on various modes between two zipcodes based on the socio-economic characteristics and the modal level of service characteristics. Separate models have been developed for various commodity groups. The estimated models for various commodity groups are found to offer statistically valid indications and plausible interpretations suggesting that these models may be suitable for application in freight transportation demand forecasting applications. The sensitivity analysis conducted on these models clearly added evidence to the fact that employment is the key factor influencing freight flows between two regions.
Thesis:
Thesis (M.S.C.E.)--University of South Florida, 2004.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Siva S. Jonnavithula.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 237 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001469376
oclc - 55644090
notis - AJR1130
usfldc doi - E14-SFE0000278
usfldc handle - e14.278
System ID:
SFS0024973:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Development of Structural Equations Models of Statewide Freight Flows by Siva S. Jonnavithula A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Civil Engineering Department of Civil and Environmental Engineering College of Engineering University of South Florida Major Professor: Ram M. Pendyala, Ph.D. Steven E. Polzin, Ph.D., P.E. Jian J. Lu, Ph.D., P.E. Date of Approval: March 25, 2004 Key Words: commodity groups, population, employment, ADF-WLS estimation, sensitivity analysis, data requirements Copyright 2004, Siva S. Jonnavithula

PAGE 2

Dedication I dedicate this thesis to my father, Mr. J. V. S. R. K. Prasad, who sacrificed all his happiness for me. I was really fortunate to have him as my father, and I take this opportunity to express my deep respect for him. I am also indebted to my mother, sister and brother for their constant encouragement and co-operation.

PAGE 3

Acknowledgements It has been a wonderful experien ce for me to work with grea t minds at the University of South Florida. First of all, I am grateful to Dr. Ram M. Pendyala for providing the opportunity to pursue my gradua te study at USF. I wish to express my sincere thanks & deep sense of gratitude to him for his c onstant guidance, encouragement & valuable suggestions in my research efforts. I also express my sincere thanks to Dr. Steven E. Polzin for his encouragement & support thro ughout my graduate study and for being instrumental in enabling me produce better-qua lity research. I am also thankful to Dr. Jian John Lu and Dr. Juan Pernia for serv ing on my committee and providing valuable suggestions. I also thank the Department of Civil and Environmental Engineering and the Center for Urban Transportation Research for providing with such excellent facilities and research environment. I am extremely thankful to my friends Mr. Srikanth Vaddepalli and Mr. Abdul Pinjari in helping me all through my research and for being wonderful pals. I would also like to thank my friends Mr. Ram Nehra, Mr. As hish Agarwal, Mr. Amlan Banerjee, Mr. Constantinos Tringides and Mr. Xin Ye for al l their help. Finally, I am grateful to Mr. Sashi Kumar Konda, Mr. Ravi Kiran Gorti, Mr. Uttam Bandugula, Mr Srikalyan Challa, Mr. Manohar Buggana, and Mr Vipan Reddy for making my life at USF a memorable one.

PAGE 4

i Table of Contents List of Tables iv List of Figures viii Abstract xii Chapter 1: Introduction 1 1.1 Background 1 1.2 Importance of Freight Transportation 2 1.3 Problem Definition 4 1.4 Objectives 6 1.5 Outline of Thesis 7 Chapter 2: Complexity of Freight Demand Modeling 8 2.1 Factors Affecting Fr eight Transportation 8 2.2 Freight Transpor tation Planning 10 2.2.1 Strategic Planning 11 2.2.2 Tactical Planning 11 2.2.3 Operational Planning 12 2.3 Complexity of Freight Demand Modeling 12 Chapter 3: Literature Review 16 3.1 Introduction 16 3.2 Freight Demand Modeling Based on the Dimension of Demand 17 3.2.1 Trip-Based Models 19 3.2.2 Commodity-Based Models 23 3.2.3 Trip-Based vs Commodity-Based Freight Models 24 3.3 Freight Demand Modeling Based on the Spatial Resolution 27 3.3.1 Global Freight Transport Models 27 3.3.2 Intercity Freight Transport Models 29 3.3.2.1 Aggregate Level Models 29 3.3.2.2 Disaggregate Level Models 30 3.3.3 Urban Freight Transport Models 31 Chapter 4: Model Framework 33 4.1 Introduction 33 4.2 Data Requirements for Interregional Freight Demand Models 34 4.3 Model Framework 37

PAGE 5

ii Chapter 5: Database Preparation 40 5.1 Introduction 40 5.2 Available Freight Data Sources 40 5.2.1 Commodity Flow Survey 40 5.2.2 Transborder Surf ace Freight Dataset 42 5.2.3 Transportation Annual Survey 44 5.2.4 Vehicle Inventory and Use Survey 45 5.2.5 Reebie TRANSEARCH Database 46 5.3 Preparation of the Freight Da ta from TRANSEARCH Database 47 5.4 Socio-Economic Data 50 5.5 Modal Level of Service Data 54 5.6 Final Dataset 55 Chapter 6: Database Description 57 6.1 Introduction 57 6.2 Distribution of Freight Fl ows by Commodity and Mode 57 6.3 Distribution of Freigh t Flows by Trip Length 60 6.4 Distribution of Freight Flows by Region 63 6.5 Distribution of Freight Flow s by Region and Socio-Economic Characteristics 81 6.6 Conclusions 81 Chapter 7: Modeling Methodology 83 7.1 Introduction 83 7.2 Structural Equations Modeling 84 7.3 ADF-WLS Estimation 87 Chapter 8: Model Estimation Results 90 8.1 Introduction 90 8.2 Model Estimation Results 91 8.3 Validity of the Model Estimation Results 96 Chapter 9: Sensitivity Analysis 100 9.1 Background 100 9.2 Sensitivity Analysis 101 9.3 Results of Sensitivity Analysis 102 Chapter 10: Conclusions and Further Research 106 10.1 Background 106 10.2 Conclusions 108 10.3 Role in the Overall Planning Process 112 10.4 Model Responsiveness 114 10.3 Further Research 114 References 116

PAGE 6

iii Appendices 122 Appendix A 123 Appendix B 174 Appendix C 190 Appendix D 206

PAGE 7

iv List of Tables Table 3.1 Modeling Platforms and Appr oaches Most Frequently Used 18 Table 5.1 STCC Commodity Classification Groups 49 Table 6.1 Distribution of Freight Flows by Commodity Group (Weight) 58 Table 6.2 Distribution of Freight Flows by Mode 59 Table 6.3 Distribution of Freight Flows by Commodity and Mode 60 Table 6.4 Distribution of Fr eight Flows by Trip Length 61 Table 6.5 Distribution of Freight Flows by Trip Length and Mode 62 Table 6.6 Distribution of Freight Flows by County 65 Table 6.7 Distribution of Freight Outflows by County and Mode 69 Table 6.8 Distribution of Freight Inflows by County and Mode 71 Table 6.9 Distribution of Truck Outflows and Inflows by County 73 Table 6.10 Distribution of Rail Ou tflows and Inflows by County 75 Table 6.11 Distribution of Air Ou tflows and Inflows by County 77 Table 6.12 Distribution of Water Outflows and Inflows by County 79 Table 8.1 Structural Equations Model Estimation Results for All Commodity Groups 94 Table 9.1 Percentage Increase in To tal Freight Flow (Base Case I) 102 Table 9.2 Percentage Increase in To tal Freight Flow (Base Case II) 104 Table 10.1 Model Responsiveness in Comparison to Other Methodologies 114

PAGE 8

v Table B.1 Structural Equations Mode l Estimation Results for Agriculture Commodity Group 174 Table B.2 Structural Equations Model Estimation Results for Other Minerals Commodity Group 175 Table B.3 Structural Equations Model Estimation Results for Food Commodity Group 176 Table B.4 Structural Equations Mode l Estimation Results for Non-Durable Manufacturing Commodity Group 177 Table B.5 Structural Equations Model Estimation Results for Lumber Commodity Group 178 Table B.6 Structural Equations Model Estimation Results for Paper Commodity Group 179 Table B.7 Structural Equations Mode l Estimation Results for Chemicals Commodity Group 180 Table B.8 Structural Equations Mode l Estimation Results for Petroleum Commodity Group 181 Table B.9 Structural Equations Model Estimation Results for Rubber Plastics Commodity Group 182 Table B.10 Structural Equations Model Estimation Results for Durable Manufacturing Commodity Group 183 Table B.11 Structural Equations Model Estimation Results for Clay, Concrete & Glass Commodity Group 184 Table B.12 Structural Equations Model Estimation Results for Primary Metals Commodity Group 185 Table B.13 Structural Equations Mode l Estimation Results for Fabricated Metal Products Commodity Group 186 Table B.14 Structural Equations Model Estimation Results for Transportation Equipment Commodity Group 187 Table B.15 Structural Equations Model Estimation Results for Miscellaneous Freight Commodity Group 188

PAGE 9

vi Table B.16 Structural Equations Model Estimation Results for Warehousing Commodity Group 189 Table D.1 Percentage Increase in Ag riculture Flow (Base Case I) 206 Table D.2 Percentage Increase in Agriculture Flow (Base Case II) 206 Table D.3 Percentage Increase in Other Minerals Flow (Base Case I) 207 Table D.4 Percentage Increase in Othe r Minerals Flow (Base Case II) 207 Table D.5 Percentage Increase in Food Flow (Base Case I) 208 Table D.6 Percentage Increase in Food Flow (Base Case II) 208 Table D.7 Percentage Increase in N on-Durable Manufacturing Flow (Base Case I) 209 Table D.8 Percentage Increase in N on-Durable Manufacturing Flow (Base Case II) 209 Table D.9 Percentage Increase in Lumber Flow (Base Case I) 210 Table D.10 Percentage Increase in Lumber Flow (Base Case II) 210 Table D.11 Percentage Increase in Paper Flow (Base Case I) 211 Table D.12 Percentage Increase in Paper Flow (Base Case II) 211 Table D.13 Percentage Increase in Chemicals Flow (Base Case I) 212 Table D.14 Percentage Increase in Chemicals Flow (Base Case II) 212 Table D.15 Percentage Increase in Petroleum Flow (Base Case I) 213 Table D.16 Percentage Increase in Petroleum Flow (Base Case II) 213 Table D.17 Percentage Increase in Rubbe r Plastics Flow (Base Case I) 214 Table D.18 Percentage Increase in Rubber Plastics Flow (Base Case II) 214 Table D.19 Percentage Increase in Durabl e Manufacturing Flow (Base Case I) 215 Table D.20 Percentage Increase in Durabl e Manufacturing Flow (Base Case II) 215

PAGE 10

vii Table D.21 Percentage Increase in Clay, C oncrete & Glass Flow (Base Case I) 216 Table D.22 Percentage Increase in Cla y, Concrete & Glass Flow (Base Case II) 216 Table D.23 Percentage Increase in Prim ary Metals Flow (Base Case I) 217 Table D.24 Percentage Increase in Prim ary Metals Flow (Base Case II) 217 Table D.25 Percentage Increase in Fabricated Metal Products Flow (Base Case I) 218 Table D.26 Percentage Increase in Fabricated Metal Products Flow (Base Case II) 218 Table D.27 Percentage Increase in Trans portation Equipment Flow (Base Case I) 219 Table D.28 Percentage Increase in Trans portation Equipment Flow (Base Case II) 219 Table D.29 Percentage Increase in Misce llaneous Freight Flow (Base Case I) 220 Table D.30 Percentage Increase in Misce llaneous Freight Flow (Base Case II) 220 Table D.31 Percentage Increase in Warehousing Flow (Base Case I) 221 Table D.32 Percentage Increase in Warehousing Flow (Base Case II) 221

PAGE 11

viii List of Figures Figure 3.1 Model Components of Trip-based Approaches 18 Figure 3.2 Model Components of Commodity-based Approaches 19 Figure 4.1 Conceptual Framework fo r Modeling Freight Transportation Demand 38 Figure 6.1 Distribution of Ton-Miles in Florida 63 Figure 8.1 Path Diagram for the Total Commodity Group Structural Equations Model 95 Figure 10.1 Flow Diagram for the Planning Process 113 Figure A.1 Freight Outflows in Annual Tons by Zipcode 123 Figure A.2 Freight Inflows in Annual Tons by Zipcode 124 Figure A.3 Ratio of Freight Outf lows to Inflows by Zipcode 125 Figure A.4 Truck Outflows in Annual Tons by Zipcode 126 Figure A.5 Truck Inflows in Annual Tons by Zipcode 127 Figure A.6 Ratio of Truck Outflows to Truck Inflows by Zipcode 128 Figure A.7 Rail Outflows in Annual Tons by Zipcode 129 Figure A.8 Rail Inflows in Annual Tons by Zipcode 130 Figure A.9 Ratio of Rail Outflows to Rail Inflows by Zipcode 131 Figure A.10 Water Outflows in Annual Tons by Zipcode 132 Figure A.11 Water Inflows in Annual Tons by Zipcode 133 Figure A.12 Ratio of Water Outflows to Water Inflows by Zipcode 134

PAGE 12

ix Figure A.13 Air Outflows in Annual Tons by Zipcode 135 Figure A.14 Air Inflows in Annual Tons by Zipcode 136 Figure A.15 Ratio of Air Outflows to Air Inflows by Zipcode 137 Figure A.16 Freight Outflows in Annual Tons by County 138 Figure A.17 Freight Inflows in Annual Tons by County 139 Figure A.18 Ratio of Freight Ou tflows to Inflows by County 140 Figure A.19 Truck Outflows in Annual Tons by County 141 Figure A.20 Truck Inflows in Annual Tons by County 142 Figure A.21 Ratio of Truck Outflows to Truck Inflows by County 143 Figure A.22 Rail Outflows in Annual Tons by County 144 Figure A.23 Rail Inflows in Annual Tons by County 145 Figure A.24 Ratio of Rail Outflows to Rail Inflows by County 146 Figure A.25 Water Outflows in Annual Tons by County 147 Figure A.26 Water Inflows in Annual Tons by County 148 Figure A.27 Ratio of Water Outflo ws to Water Inflows by County 149 Figure A.28 Air Outflows in Annual Tons by County 150 Figure A.29 Air Inflows in Annual Tons by County 151 Figure A.30 Ratio of Air Outflows to Air Inflows by County 152 Figure A.31 Population by Zipcode 153 Figure A.32 Annual Tons Exported per Person by Zipcode 154 Figure A.33 Annual Tons Imported per Person by Zipcode 155 Figure A.34 Population by County 156 Figure A.35 Population Density by County 157

PAGE 13

x Figure A.36 Annual Tons Exported per Person by County 158 Figure A.37 Annual Tons Imported per Person by County 159 Figure A.38 Employer Locations in Florida 160 Figure A.39 Employment by Zipcode 161 Figure A.40 Agricultural, Forestry an d Fishery Employment by Zipcode 162 Figure A.41 Mining and Construction Products Employment by Zipcode 163 Figure A.42 Light Manufactured Products Employment by Zipcode 164 Figure A.43 Heavy Manufactured Pr oducts Employment by Zipcode 165 Figure A.44 Transportation, Communi cation and Utilities Employment by Zipcode 166 Figure A.45 Wholesale and Retail Trade Employment by Zipcode 167 Figure A.46 Finance, Insurance and R eal Estate Employment by Zipcode 168 Figure A.47 Entertainment, Accommoda tion and Food Services Employment by Zipcode 169 Figure A.48 Other Services Employment by Zipcode 170 Figure A.49 Public Administra tion Employment by Zipcode 171 Figure A.50 Freight Exports in Annu al Tons per Employee by Zipcode 172 Figure A.51 Freight Imports in Annu al Tons per Employee by Zipcode 173 Figure C.1 Path Diagram for the Agri culture Commodity Group Structural Equations Model 190 Figure C.2 Path Diagram for the Other Minerals Commodity Group Structural Equations Model 191 Figure C.3 Path Diagram for the Food Co mmodity Group Structural Equations Model 192 Figure C.4 Path Diagram for the N on-Durable Manufacturing Commodity Group Structural Equations Model 193

PAGE 14

xi Figure C.5 Path Diagram for the Lu mber Commodity Group Structural Equations Model 194 Figure C.6 Path Diagram for the Paper Commodity Group Structural Equations Model 195 Figure C.7 Path Diagram for the Chemicals Commodity Group Structural Equations Model 196 Figure C.8 Path Diagram for the Pe troleum Commodity Group Structural Equations Model 197 Figure C.9 Path Diagram for the Rubbe r Plastics Commodity Group Structural Equations Model 198 Figure C.10 Path Diagram for the Du rable Manufacturing Commodity Group Structural Equations Model 199 Figure C.11 Path Diagram for the Cla y, Concrete & Glass Commodity Group Structural Equations Model 200 Figure C.12 Path Diagram for the Primar y Metals Commodity Group Structural Equations Model 201 Figure C.13 Path Diagram for the Fa bricated Metal Products Commodity Group Structural Equations Model 202 Figure C.14 Path Diagram for the Tr ansportation Equipment Commodity Group Structural Equations Model 203 Figure C.15 Path Diagram for the Mi scellaneous Freight Commodity Group Structural Equations Model 204 Figure C.16 Path Diagram for the Wa rehousing Commodity Group Structural Equations Model 205

PAGE 15

xii Development of Structural Equations Models of Statewide Freight Flows Siva S. Jonnavithula ABSTRACT The modeling of freight travel demand has gain ed increasing attention in the recent past due to the importance of efficient and safe freight transportation to regional economic growth. Despite the attention paid to the m odeling of freight travel demand, advances in modeling methods and the development of prac tical tools for foreca sting freight flows have been limited. The development of fr eight demand models that incorporate the behavioral aspects of freight demand face si gnificant hurdles, partially due to the data requirements, which are a consequence of th e inherent complexity of the mechanisms driving freight demand. This research attempts to make a contributi on in this context by proposing a relatively data simple, but behaviorally robust statewide modeling framework for the state of Florida, in the sp irit of an aggregate level four-step planning process. The modeling framework that is developed in this research can be applied to the modeling of freight travel de mand using data contained in readily available commercial databases such as the Reebie TRANSEA RCH database and the InfoUSA employer database. The modeling methodology consists of a structural equations modeling framework that can accommodate multiple dependent variables simultaneously. This

PAGE 16

xiii framework predicts freight flows on variou s modes between two zipcodes based on the socio-economic characteristics and the modal le vel of service characteristics. Separate models have been developed for various commodity groups. The estimated models for various commodity groups are found to offe r statistically valid indications and plausible interp retations suggesting that thes e models may be suitable for application in freight transportation dema nd forecasting applications. The sensitivity analysis conducted on these mode ls clearly added evidence to the fact that employment is the key factor influencing frei ght flows between two regions.

PAGE 17

1 Chapter 1 Introduction 1.1 Background Profound and revolutionary transformations in the areas of co mputer technology, communication networks, and information and production systems have characterized the latter part of the 20th century and the be ginning of the 21st century. The development and growth of the internet combin ed with the convergence of th ese trends has made possible ever deeper changes in the ways both bus inesses and consumers do their economic transactions. Business-to-bus iness and business-to-consumer s internet systems enable businesses to effectively integrate their ope rations in a seamless way, and enable both businesses and consumers to have access to vendors way beyond the traditional geographic boundaries. Such internet systems are intensely changing the nature and characteristics of production systems, commerce in genera l, and the supporting freight transportation system (Holguin-Veras et. al ., 2003). Of 50 leading shipping companies polled in 1996 at the World Express & Mail Co nference in Brussels, mo re than half cited e-commerce as the single most important fact or driving their grow th (Holguin-Veras et. al., 2003). In that year DHL Worldwide Expr ess, for example, projected 40% annual growth for its online business. Also, Finger hut Business Services Inc. in Minneapolis credited 70% of its fulfillment business to internet companies including new customers such as the Wal-Mart Stores Inc. Web site and eToys Inc. (Holgui n-Veras et. al., 2003).

PAGE 18

2 In spite of the burst of the ‘‘dotcom bubble’’, US census data indicated that ecommerce sales in the first quarter of 2001 increased 33.5% with respect to the first quarter of 2000 (Hu, 2001; Holguin-Veras et. al., 2003). There is agreement among business analysts that: (a) e-commerce will keep growing in the foreseeable future; and (b) the totals for business-to-consumers will be dwarfed by business-to-business transactions, once businesses fully integrate th eir operations with e-comm erce (Lahsene, 2001; HolguinVeras et. al., 2003). All of th is points toward an increasing role of the freight transportation system as the conveyor of goods for the e-commerce systems. 1.2 Importance of Freight Transportation In addition to e-commerce, economic globa lization, high-tech warehousing, and Just-InTime production systems are also increasing the already important role of the freight transportation system. More and more, busine sses and consumers alike are relying on the freight transportation industry for the deliv ery of goods on demand, thus reducing the need for inventory stocks (Holguin-Veras et al., 2003). Thus, freight transportation is one of today's most important activities, not on ly as measured by the yardstick of its own share of a nation's gross national product (GNP ), but also by the increasing influence that the transportation and distribution of goods have on the performance of virtually all other economic sectors (Crainic et. al ., 1997). A few figures illustrate these assertions. In 1978, Taff estimated that transportation accounts fo r approximately 10% of the United States GNP and current figures could ve ry well be significantly larger (Crainic et. al., 1997). In the United Kingdom, for example, it has been estimated that transportation represents 15% of national expenditures (Crainic et. al ., 1997). These figures are similar to those

PAGE 19

3 observed for Canada (some 16%) (Zalatan, 1993) and France (around 9% ) (Crainic et. al., 1997). Furthermore, transportation represents a si gnificant part of the cost of a product. In Canada, for example, this part may reach 13% for the primary industrial sector and 11% for the transformation and production industry (Owoc et. al., 1992; Cr ainic et. al., 1997). At the same time, there is increasing pressure from both community and environmental groups to ameliorate the negative impacts of freight activity. More and more, local communities, environmentalists, and resear chers are demanding actions to reduce the negative externalities of freight traffic. In th is context, a number of studies are looking at the environmental impacts of freight activity upon local communities (Holguin-Veras et. al., 2003). However, in spite of the negative externalities that fr eight activity produces, there is no doubt that freight transportation makes significant contributions to the vitality of the nation’s economy. In 1997, the value of the cargoes transported amounted to 6.9 trillion dollars, with a total tonnage equal to 11 billion tons, to taling 2.66 trillion tonmiles across the continental United States Trucking, the dominant mode, accounts for 70% of the total tonnage (Bureau of Transportation Statistics, 1999). Moreover, in a context of increasing econom ic globalization and in the interest of minimizing the total costs of producing a nd delivering goods, production systems are reaching out to global markets of supply and demand. The net effect of economic globalization is to extend the geographic rea lm of freight transportation systems. Once often confined by national boundaries, the tran sportation systems of today and tomorrow

PAGE 20

4 will have to operate across multiple nations at a global scale, and at some point in time they will operate as if polit ical boundaries did not exist. All of the above imply that the freight transp ortation systems of the 21st century will be expected to cover a larger geographic area, be more re sponsive to user needs and expectations, and reduce the environmental, safe ty & health externalities associated with truck traffic. Moreover, the freight transportation systems have to achieve all of this in a context in which the provision of additional freight infrastructure capacity will become more difficult and expensive. 1.3 Problem Definition Freight transportation lies at th e heart of our economic life. In industrialized countries, it accounts for significant share of the gross national product. In developing countries, it is the essential ingredient of sustainable deve lopment. With free trade zones emerging in several parts of the world and with the gl obalization of the economic system, freight transportation will in all like lihood play an even more major role in years to come. The trend towards larger, more integrated a nd more efficient transportation systems is likely to remain and should creat e the need for better planning at the strategi c, tactical and operational levels (discussed in Chapte r 2). Thus, various freight demand modeling methodologies have emerged over time to assist in freight transporta tion planning efforts. Some of the models are simple growth factor models while others are more complex and accurate autoregressive integrated moving av erage models (ARIMA), elasticity models,

PAGE 21

5 network models of logistics, direct dema nd and aggregate demand models, disaggregate demand models, and economic input-output models. Major re search efforts have been devoted to the design of models for dynamic and stochastic problems. Key developments are also taking place in the arti ficial intelligence-related area of metaheuristics such as tabu search, genetic algorithms, neural networ ks, etc (Crainic et. al., 1997). These have already given a new impetus to the whole area of global optimization and have lead to a rethinking of the entire field of heuristics. These developments, coupled with the growth of parallel methods, mean th at in the near future larg er and more complex problems should be amenable to analysis and optimi zation. In particular, significant advances should be expected in the areas of dynami c, stochastic and real-time programming, central to so many transportation systems. Chapter 3 furnishes a comprehensive literature review on the various freight demand models that have been developed till date. Despite the attention paid to the modeling of freight travel demand and advances in modeling methods, the development of practical and reliable tools for forecasting freight flows have been limited. This limitation has be en due to complexity of freight demand modeling arising from the multiple dimensions of freight demand (volume, weight and trips) under the control of a number of deci sion-makers (drivers, dispatchers, freight forwarders) who interact in a rather dynamic environment. Moreover, freight transportation data has been traditionally difficult to collect due to the proprietary nature of the data and due to the difficulty with identifying the proper entity to which a freight transportation survey needs to be admini stered. These factors contributing to the complexity of freight demand modeling ar e discussed in detail in Chapter 2.

PAGE 22

6 It is against the backdrop of such limitations that the Florida Department of Transportation, as part of its ongoing research into the development of statewide freight transportation models, desired to develop a robust and practical statewide modeling framework that can be used to estimate freight travel demand in Florida at a microscopic level using data contained in readily available commercial da tabases. Thus, the objective of this study is to propose a relatively da ta simple, but behaviorally robust statewide freight travel demand modeling framework at a microscopic level in the spirit of an aggregate level four-step planning process. 1.4 Objectives The primary objective of this dissertation is to develop a behaviorally robust and practical modeling framework that can quantify and predict freight flows by various modes between origin-destination pair s in the state of Florida. The other distinguishable objectives would be as follows: To understand the factors that contribute to the complexity of freight demand modeling; To perform a comprehensive literatu re review on frei ght demand modeling techniques and study their advantages and limitations; To develop a model concept that is data simple, but largely in line with paradigms and freight transportation de mand-supply relationships iden tified in th e literature; To identify a suitable freight data sour ce and prepare a comprehensive freight flow database merging freight data with other data, as requi red by the developed model concept;

PAGE 23

7 To perform a descriptive analysis of the developed freight flow database, in order to understand the freight fl ow patterns in Florida; To estimate the statewide freight travel demand at a microscopic level using an appropriate modeling methodology, and To analyze the potential influence of va rious factors on freight travel demand using sensitivity analysis. 1.5 Outline of Thesis The remainder of this thesis is organized as follows. The next chapter provides a good understanding of the factors that contribute to the co mplexity of freight demand modeling. This chapter is followed by a compre hensive literature review that discusses the advantages and limitations of vari ous freight demand modeling methodologies developed earlier. The fourth chapter identif ies the paradigms and freight transportation demand-supply relationships identified in the literature and leads to the model concept that will be used in this study. In the fift h chapter, various data sources available at disposal are reviewed and the database prepar ation is discussed. This chapter is followed by a description of the database used in the study. The seventh chapter discusses the identification of an appropriate modeli ng technique and its methodology. The model estimation results are provided in the eighth chapter. The sensitivity analysis performed to analyze the potential influence of various factors on freight travel demand, using the developed model systems is discussed in chapter nine. Finally, conclusions and implications of the research findings are di scussed in the tenth chapter together with future directions in freight demand modeling.

PAGE 24

8 Chapter 2 Complexity of Freight Demand Modeling 2.1 Factors Affecting Freight Transportation The freight transportation indus try, as all other economic sect ors, has to achieve high performance levels both in terms of economi c efficiency and service quality. Economic efficiency because a transportation firm ha s to make profits while at the same time competing in an increasingly open and competitive market where cost is still the major decision factor in selecting a carrier or distribution firm Yet, one also observes an increasing emphasis on the quality of the serv ice offered. Indeed, the new paradigms of production and management, such as small or no inventories associated to just-in-time procurement, production and distribution, quali ty control of the entire logistics chain driven by customer demand and requirements, etc., impose high service standards on the transportation industry. This applies, in pa rticular, to total de livery time and service reliability (Crainic et. al., 1997). The political evolution of the world also ha s an impact on the transportation sector. The emergence of free trade zones, in Europe and on the American continent in particular, has tremendous consequences for the evolution of freight transportation systems, not all of which are yet apparent or well understood. Fo r example, open borders generally mean that firms are no longer under the obligation to maintain a major di stribution center in

PAGE 25

9 each country. Then, distribution systems are re organized and this often results in fewer warehouses and transportation over longer distances (which still have to perform according to low cost-high service standards). A significant increase in road traffic is a normal consequence of this process, as ma y be observed in Europe. A study conducted for the European Parliament forecasts a 34% increase in land-based transport for the countries of the European Economic Co mmunity between 1988 and the year 2000 (Crainic et. al., 1997). Additional factors which impact on the organization, operation policies and competitiveness conditions in the transportati on industry are the in ternationalization of the economy and the opening of new markets due to political changes, mainly in central Europe and Asia, and the evolution of the regulatory environment. The first two imply larger economic spaces and transportation networks. Thus, from 1971 to 1988, the total volume of goods moved by ship has doubled, while the total number of kilometers covered by air cargo quadrupl ed (Crainic et. al ., 1997). Changes to the regulatory environment of transportation, particularly si gnificant in North Am erica and starting to gather momentum in Europe and elsewhere also has a powerful impact on the operation and competitive environment of transportation firms. The deregulation drive of the 80's has seen governments remove numerous rules a nd restrictions, especially with regard to the entry of new firms in the market and the fixing of tariffs and r outes, resulting in a more competitive industry and in change s in the number and characteristics of transportation firms. At the same time, more stringent safety regulations have been

PAGE 26

10 imposed, resulting in more complex planning and operating procedures (Crainic et. al., 1997). Moreover, there are several different types of players in the transportation field, each with its own set of objectives and means. Ho wever, the most important players are the shippers, carriers and governments (Crainic et. al., 1997). Produc ers of goods require transportation services to move raw material s and intermediate produ cts and to distribute final goods in order to meet demands. He nce, they determine the demand for transportation and are often called shippers (Other players, such as brokers, may also fall in this category). Transportation is usually performed by carriers, such as railways, shipping lines, motor carriers, etc. Thus, one may describe an intermodal container service or a port facility as a carrier. Governments constitute another important group of players. First, they regulate several aspect s of freight transporta tion. Then, they also provide a large part of the transportation in frastructure: roads and highways, and often a significant portion of the port, intern al navigation, and rail facilities. 2.2 Freight Transportation Planning As described in the previous section, tr ansportation systems are rather complex organizations which involve a great deal of human and material resources and which display intricate relationships and trad e-offs among the various decisions and management policies affecting their differe nt components. Crainic and Laporte (1997) classify these policies according to the following three planning levels.

PAGE 27

11 2.2.1 Strategic Planning Strategic (long term) planning at the firm le vel typically involves the highest level of management and requires large capital investments over long time horizons. Strategic decisions determine general development policies and broadly shape the operating strategies of the system. Prime examples of decisions at this pla nning level are the design of the physical network and its evolution (upgr ading or resizing), the location of main facilities (rail yards, multimodal platforms, etc.), resource acquisition (motive power units, rolling-stock, etc.), the definition of broad service and tariff policies, etc. Strategic planning also takes place at the internati onal, national and regional levels, where the transportation networks or se rvices of several carriers ar e simultaneously considered. State transportation departments, consultants, international shippers, etc. engage in this type of activity. 2.2.2 Tactical Planning Tactical (medium term) plan ning aims to ensure, over a medium term horizon, an efficient and rational allocation of existing re sources in order to improve the performance of the whole system. At this level, data is aggregated, policies are somewhat abstracted and decisions are sensitive only to broad variat ions in data and system parameters (such as the seasonal changes in traffic demand) without incorporating the day-to-day information. Tactical decisions need to be made mainly co ncerning the design of the service network, i.e., route choice and type of service to operate general operating rules for each terminal and work allocation among term inals, traffic routing using the available

PAGE 28

12 services and terminals, repos itioning of resources (e.g., em pty vehicles) for use in the next planning period. 2.2.3 Operational Planning Operational (short term) planni ng is performed by local ma nagement (yardmasters and dispatchers, for example) in a highly dynamic environment where the time factor plays an important role and detailed representations of vehicles, facilities and activities are essential. Scheduling of services, maintena nce activities, crews, etc., routing and dispatching of vehicles and crews, res ource allocation are important operational decisions. This classification highlights how the data flows among the decision-making levels and how policy guidelines are set. Th e strategic level sets the ge neral policies and guidelines for the decisions taken at the tactical level, which determines goals, rules and limits for the operational decision level regulating the transportation system. The data flow follows the reverse route, each level of planning s upplying information essential for the decision making process at a higher level. This hierar chical relationship prev ents the formulation of a unique model for the planning of freight transportation systems a nd calls for different model formulations addressing specific probl ems at specific levels of decision making. 2.3 Complexity of Freight Demand Modeling The hierarchical relationship discussed in th e above section puts a significant amount of pressure on metropolitan planning organizat ions (MPOs) in enha ncing their freight

PAGE 29

13 transportation planning processes. Further, fe deral legislations such as the Intermodal Surface Transportation Efficiency Act of 1991 (ISTEA) and the Transportation Equity Act for the 21st Century (TEA-21), and a host of state-level initiatives necessitate the undertaking of comprehensive freight trans portation planning and mobility strategies. However, this objective is confounded by the lack of freight-trans portation-specific demand modeling methodologies. For the most pa rt, the bulk of freight transportation modeling applications are nothing more than adaptations of transportation modeling methodologies originally designed for passe nger transportation, that tend to overlook the fundamental differences between freight move ments and passenger transportation. This is because passenger issues traditionally have been assigned the highest priorities, effectively reducing the amount of resour ces and attention allocated to freight transportation research and educat ion (Holguin-Veras et. al., 2000). Thus, the most significant hurdle to including freight transportation in the transportation modeling process is that most of the demand forecast methodologies have been developed for passenger trips, not freight trip s. This methodological void usually is filled by simplistic approaches such as assuming that freight trips follow the same behavioral mechanisms as passenger trips, which is an implicit assumption when truck traffic is estimated as a function of passenger tra ffic (Holguin-Veras et. al., 2000). Although the error induced by this assumption may not be important for small urban areas where the number of freight trips is re latively small, it can not be used in large metropolitan areas such as New York City where freight-relat ed trips are a major contributor to urban congestion, and freight-specific tran sportation policie s are warranted.

PAGE 30

14 The complexity of modeling freight demand arises from a combination of factors (Holguin-Veras et. al., 2000). First and fo remost, multiple dimensions are to be considered (Ogden, 1992; Holguin-Veras et. al., 2000). Whereas in passenger transportation there is only one unit of dema nd – that is, the passenge r, who for the most part happens to be the decision-maker – in freight transportation there are multiple dimensions (volume, weight and trips) under the control of a number of decision-makers (drivers, dispatchers, freight forwarders) who interact in a rather dynamic environment. Also, a significant portion of freight demand is discretionary in nature. In this context, a relatively small number of companies have co ntrol over a significant number of freight movements. Integrating their behavior into pl anning models is rather challenging because the dynamics of their decision-making proce ss, marked by their commercially sensitive nature, are not part of the public domain. The second major factor contri buting to the complexity in modeling freight demand is significant difference in time value, or opportunity costs, exhibited by cargoes (Cambridge Systematics, 1997). Cargo time va lue – determined by opportunity costs – exhibits a much wider range compared to the passenger’s time value ranges within the same order of magnitude. Cargoes’ opportunity costs are determined by a combination of the intrinsic cargo value (determined by ma rket value and replacement costs) and the logistic cargo value (a func tion of the importance of the cargo for the production system at a given moment in time and inventory le vels) (Holguin-Veras et al., 2000). At one end, low-priority cargoes may have intrinsic cargo values as low as $9/ton (gypsum); and at the other, high priority cargoes have in trinsic cargo values that frequently exceed

PAGE 31

15 $500,000/ton (e.g., computer chips) (Holguin-Ve ras et. al., 1999). Th ese figures would increase significantly once the logi stic cargo value is factored in. The third and the most important factor that makes the freight demand modeling complicated is its data collection. Major advances have been made in the development of freight transportation modeling methods a nd frameworks (Este, 2002; Regan et. al., 2002). However, many of these methods have not seen application in practice partially due to the lack of adequate da ta to support their estimation a nd application to forecasting. Freight transportation da ta has been traditionally difficult to collect due to the proprietary nature of the data and due to the difficulty with identifying the proper entity to which a freight transportation survey needs to be administered. The absence of freight transportation data is particularly critical at the disaggregate (spatial and temporal) level, making the estimation of disaggregate m odels of freight transportation demand a challenge that needs to be addressed. While there are aggregate level freight transportation data sets such as the commodity flow survey (CFS) data these data sets are generally insufficient to develop models that can estimate origin-destination freight flows by mode and commodity. The multitude of factors discussed in this chapter result in the complexity of freight demand modeling. In this cont ext, there is a need for de veloping practical and reliable modeling frameworks for directly estimati ng freight transportation flows by commodity and mode between origin-destination pairs. The next chapter provides an overview of the various freight demand models that have been developed till date.

PAGE 32

16 Chapter 3 Literature Review 3.1 Introduction Freight transportation is a v ital component of the economy. It supports production, trade, and consumption activities by ensuring the effi cient movement and timely availability of raw materials and finished goods. In conseque nce, freight transportation represents a significant part of the cost of a product, as well as of the nationa l expenditures of any country. Moreover, the significan ce of freight movement and activity has been increasing in terms of both its role in the economy and its potentially adverse impacts on safety and congestion on the transportation system (Pen dyala et. al., 2000; Czerniak et. al, 2000; DeWitt et. al., 2000; Regan et al., 2000). Thus, it is not surprising that freight transportation planning has attr acted the attention of resear chers since the early 1970s (Baumol et. al., 1970; Allen, 1977; Slavin et. al., 1976). Freight travel demand and supply are key elem ents of the overall freight transportation planning process that also considers th e socio-economic environment, intermodal transportation network, policy and regulator y environment, and system performance measures. Freight demand models can be used to support a host of planning applications including facility planning, corridor planning, strategi c planning, business logistics planning, and economic development. Wide rang e of demand models from simple growth

PAGE 33

17 factor models to more complex and accurate autoregressive inte grated moving average models (ARIMA), elasticity models, network models of logistics, direct demand and aggregate demand models, disaggregate demand models, and economic input-output models have been developed (Ogden, 1991; Ca mbridge Systematics, 1997; Faris et. al., 1999; Hancock et. al., 2 000; List et. al., 1995). Historically, freight transpor tation demand models have been classified based on (a) the dimensions of freight demand and (b) the spat ial levels of consideration. Holguin-Veras (undated) presented an extensive literature review on the dimension based freight demand modeling, and Regan and Garrido (2002) presented a similar one on the spatial resolution based freight demand modeling. The following sections offer a combined review of freight demand modeling based on th ese two literature reviews. 3.2 Freight Demand Modeling Based on the Dimension of Demand One of the unique features of freight transpor tation planning is that there are a number of different dimensions to be taken into account most notably: weight volume, number of vehicle trips, and value of the commodities be ing transported. Each of these dimensions represents a different way to define and measure freight trans portation demand, with important implications for freight demand m odeling. The existence of these different dimensions has resulted into two major mode ling platforms: vehicle-trip based modeling and commodity-based modeling. Various modeli ng approaches have been used on each of these platforms. The most widely used opt ions include (a) variants of the Four Step Model, (b) direct demand models, and (c) i nput-output models. Jose Holguin-Veras and

PAGE 34

18 Ellen Thorson (2000) summarize the major co mbinations as shown in Table 3.1. They also describe the processes for both the approaches as shown in Figures 3.1 and 3.2. Table 3.1 Modeling Platforms and Approaches Most Frequently Used Modeling Approach Variants of Four Steps Model Direct Demand Models Input-Output Models Commoditybased Used in both urban and regional applications Frequently used in corridor analysis Used in regional economic development studies, rarely in urban areas, though land use – transportation models are based on I-O Modeling Platform Trip-based Used in both urban and regional applications Frequently used in corridor analysis Not applicable Source: Holguin-Veras and Thorson, 2000 Figure 3.1 Model Components of Trip-based Approaches Trip generation Trip distribution Traffic assignment Approach: Trip generation rates or zonal regression models Gravity models (simply or doubly constrained) or Intervening Opportunities Standard traffic assignment techniques Step: Source: Holguin-Veras and Thorson, 2000

PAGE 35

19 Figure 3.2 Model Components of Commodity-based Approaches Commodity generation Vehicle-trip estimation Commodity mode split Commodity distribution Traffic assignment Step Approach: Commodity generation rates or zonal regression models Gravity models (simply or doubly constrained) or Intervening Opportunities Logit models based on panel data. Rarely done in urban areas. Loading rates based upon previous surveys and complementary emtpy trip models Standard traffic assignment techniques Source: Holguin-Veras and Thorson, 2000 3.2.1 Trip-Based Models As the name implies, trip-based models fo cus on modeling vehicle-trips. As shown in Figure 3.1, these models have three component s: trip generation, trip distribution and traffic assignment. Trip-based models do not need mode split or vehicle loading models, since the focus is on vehicle-tr ips, which assumes that the mode selection and the vehicle selections were already done. List and Turnquist (1995) pr esented a trip-based modeli ng method for estimating multiclass truck trip matrices from partial and fr agmentary observations. Data sets of widely varying character were combined in an efficient and effective manner so that each piece of information had a role in developing the es timated flows. O-D matrices were estimated

PAGE 36

20 from truck traffic counts, O-D synthesis. Th e technique assumed th at the links in the analysis network consisted of at least three attributes: a directional flag, a use label (e.g. truck class) and a travel time which may vary according to the time of day. In addition, the technique assumed the study area is divisible into non-ov erlapping zones, each zone having a 'centroid', where trips originate and/or terminate. Th e input data were of three types namely: link volumes or classification counts, partial O-D estimates for various zones, including time periods and truck classi fication, and origin/termination information. In essence, nine O-D matrices were estima ted for three time peri ods (6:00-10:00 A.M., 10:00 A.M.-3:00 P.M., and 3:00-8:00 P.M.) and three truck classifications (van, medium truck, and heavy truck). Another example of trip-based models is the Quick Response Freight Manual (QRFM), developed by Cambridge Systematics, Inc. (1 996). The model consists of the following steps: Obtaining data on economic activity for traffic analysis zones (including employment by type and the number of households); Applying trip generation rates to estimate the number of commercial vehicle trips for each traffic analysis zone; Estimation of commercial vehicle volumes at external stations; Estimation of the number of commercial vehicle trips between pairs of traffic analysis zones or external stations; Estimation of the mode share for each trip; Loading the O-D trip to the network; and

PAGE 37

21 Comparison of control VMT with estimated VMT. Jack Faucett Associates (1999b) illustrate the trip-based modeling process by generating truck trips as a function of di fferent land uses and trip data from trip diaries or shipper surveys. They calculate the trip rates as a function of socio-economic data (trips per employee) and land use data (trip per acre). The generated trips are distributed using some form or other of spat ial interaction models, most commonly a form of gravity model. The gravity model is typically calibra ted using trip length fr equency distributions obtained from trip diaries. Holgun-Veras (2000) proposed the “Integra tive Freight Market Simulation” (IFMS), currently being funded by the National Scie nce Foundation (NSF) E xploratory Transport Industries (ETI) program (2000). The underlying assumption of IFMS is that urban good movements are to be modeled as a market in equilibrium in which the different players, such as trucking companies maximize profit s. Thus, IFMS considers the fundamental interactions between the ke y participants (i.e., producer s, consumers and freight companies) in a game theoretical formula tion. IFMS deals with two different problems based on a bi-level formulation. The firs t problem, formulated as a Cournot-Nash equilibrium problem, entails the estimation of the transportation service (i.e., amount of loaded and empty trips that will be contri buted to the market) pr ovided by the different clusters of companies in order to maximi ze profits. The second problem involves a multivehicle routing problem that is intended to estimate the origin-destination patterns that, while consistent with the Cournot-Nash solution, also meet the other system constraints.

PAGE 38

22 In essence, the proposed framework would attempt the estimation of the trips made by freight transportation providers in the study area, such that: Freight transportation companies maximize profits while market equilibrium conditions are met (Condition I) The user requirements are met, i.e., th e commodities produced by and attracted to each TAZ are transported (Condition II) The resulting trip chains are consistent w ith trip chain patterns captured in travel diaries, or alternatively, known Trip Length Distributions (Condition III) The resulting commercial vehicle traffi c is consistent with secondary data sources, e.g., ITS traffic data (Condition IV). Houlguin-Veras (2000) term s conditions I and II as primary constraints i.e., constraints that must be met. He refers to conditions III to IV as secondary constraints which could be relaxed under certain circumstances. Cond ition I ensures proper consideration of the interactions among freight providers in the supply market. Condition II ensures consideration of user require ments. Conditions III and IV are information constraints expected to bound the solution. All the studies reviewed above treat trips as aggregated in a zonal level such as a TAZ, county, state, etc. Watson (1975), however, pres ented an approach where each firm or industrious agency is viewed as a unit that ha s attributes of employee, floor acreage, etc. The trips from these agencies can be regresse d to these attributes. This disaggregate approach is attractive due to its capability to capture the attributes that influence the trip

PAGE 39

23 generation and attraction. This approach, how ever, is not adopted in practice, probably due to its extensive data n eed to develop these models. 3.2.2 Commodity-Based Models Commodity-based modeling, on the other ha nd, focuses on the amount of commodities being transported, usually defined by its we ight. The focus on commodities enables these models to depict the fundamental processes taking place and, in doing so, to take into account the economic characteristics of cargoes. The components of the modeling process are depicted in Figure 3.2. Commodity based modeling is comprised of th e following process. The first step is the commodity generation models th at are used to estimate the total number of tons produced and attracted by each zone in the study area. The second step is the distribution phase, wherein the tonnage moving between each origin-destination pair is estimated using gravity models and other forms of spatial inte raction models. The thir d step is the mode split component, intended to estimate the number of tons moved by the various modes which is done by applying discrete choice mode ls and/or panel data from focus groups of business representatives or fr eighters (Cutler et. al., 2000) Finally, in the traffic assignment phase of commodity-based models, a combination of vehi cle loading models and complementary models that capture em pty trips, applied to origin-destination matrices by mode, are used to assign vehicle trips to the network. (Holguin-Veras et. al., 2000; Hautzinger, 1984).

PAGE 40

24 Jack Faucett Associates (1999b) summarize commodity based modeling as follows: The starting point is a known regi on-to-region table of commod ity flow tonnage based on economic output forecasts and established regi onal trade patterns. Th e region-to-region flows, depicting inbound and outbound flow patt erns, are disaggregated to the zonal level based on economic data, reflective of th e intensity of production and consuming industries. Economic data and input-output tabl es are used to estimate the quantity of each commodity produced and consumed w ithin each geographic unit. Knowing the commodity being shipped, it becomes possi ble to link producers of a good with its consumers through these economic relationships Once commodity flows are assigned to origins and destinations, they are convert ed to truck trips or vehicle trips. 3.2.3 Trip-Based vs. Commodity-Based Freight Models Trip-based models have some advantages. The most important advantage is that traffic data is relatively easy to obtain. Furtherm ore, an increasing number of Intelligent Transportation Systems (ITS) applications are able to track the movements of vehicles through the highway networks, becoming an im portant source of traffic data. The other advantage is that these types of models consider both loaded and empty trips because of their focus on vehicle trips, unlike the commod ity-based models (Holguin-Veras et. al., 2000; Holguin-Veras et. al., 1999; Cambri dge Systematics, 1997; Ogden, 1992). However, trip-based models have some significant limitations. As demonstrated by Holguin-Veras and Jara Diaz ( 1999) and McFadden et al (1986) they are unable to take into account the economic characteristics of the ca rgoes that play an im portant role in the

PAGE 41

25 vehicle selection, mode choice and routing pro cesses. The other limitation is that these models are of limited applicab ility to multimodal freight transportation systems because of their focus on the vehicle trip, which in itself is the result of a choice process that already took place (Holgu in-Veras et. al., 2003). In regard to these limitations of trip-based models, commodity-based models are attractive because of their focus on modeling th e amount of freight measured in tons, or any comparable unit of weight. Thus, they capture more accurate ly the fundamental economic mechanisms driving freight moveme nts, which are largely determined by the cargoes’ attributes (e.g., shape, specific weight, volume). Thus, this study focuses on developing statewide commodity-based freight demand models, rather than the vehicletrip based models. However, commodity based models are limited in their ability to model empty trips, that are the result of logistic decisions not di rectly explained by the commodity flows. Researchers address this limitation at the calibration stage by expanding the commodity distribution matrix so that the resulting traffic assignment resembles the calibration values (Houlguin-Veras et. al., 2000 and 2003). Th is approach is debatable, although it is widely used. First, there is no way to ensure that the resulting number of empty trips is consistent with the area wide estimates of the total numb er of empty trips. Second, expanding a commodity trip matrix to comp ensate for the missing empty trips implies that empty trips are directly correlated with the commodity flows. This assumption is

PAGE 42

26 very weak because the commodity flow betw een two zones determines the amount of loaded trips between them, not the amount of empty trips. Holguin-Veras and Thorson (2003) have ove rcome this limitation of commodity based models by developing complementary model fo rmulations to depict empty trips as a function of the routing choices that the commercial vehicl e operators make, which are based on the commodity flows in the study area. These formulations were based on probability principles and spatial interaction concepts. The models were based on the concept of order of a trip chain, defined as the number of additional stops with respect to the primary trip, and provided a statistical link between the first order and higher order trip chains. Three different destination choice probability functions were hypothesized based on different assumptions about the de stination choice pro cess. One of these formulations included a memory component th at takes into account the amount of travel already done in the destinati on choice process. An exampl e, based on data from an origin–destination study in Guatemala City, wa s included to show th e practicality of the proposed models. The numerical results indicate d a slight superiority of the formulation that takes into account the length of the previous trip. In all cases, their model outperformed the previous models which seem to be an indication of the reasonableness of its fundamental assumptions and specifically of the benefits of including a memory function. The paper also provided empirical evidence of the importance of modeling empty trips.

PAGE 43

27 3.3 Freight Demand Modeling Based on the Spatial Resolution Regan and Garrido (2002) classify freight transport models by spatial considerations into three broad categories: global (int ernational), intercity, and urban. 3.3.1 Global Freight Transport Models Global freight transport mode ls aim to model the goods movement between different countries. In the recent years, multinational firms have dispersed all over the world to take advantage of competitive prices for both materials and labour. The goods components are manufactured in different locatio ns and hence need to be transported to a certain location to be assembled and shippe d abroad again. This has resulted in an explosive growth of global freigh t transport in the last decade. Regan and Garrido (2002) identified three main approaches to model global demand for shipping. The first approach follows the sta ndard theory of intern ational trad e (Cassing, 1978). The second approach reli es on an aggregate cost f unction for a given industrial sector, from which a demand function for sh ipping is derived. The demand function is such that it minimises the cost function. This approach allows to work with an analytical expression for the demand function. Nevert heless, Regan and Garri do (2002) identified that this approach has two main drawbacks: the data requirement and the computational complexity of the solving process (a non-linear multidimensional minimization). The third approach identified by Regan and Garrido (2002) is the use of spatial interaction models to estimate tr ade flows. This approach is attractive for practical use in

PAGE 44

28 the medium and short run as it models vehi cle movements directly, instead of modeling the demand for commodities. However, they are not adequate for forecasting purposes as they are cross-section models. These models are not data demanding as they reflect mere tendencies in spatial distribu tion according to an impedanc e function. However, these models do not capture behavioural aspects of freight demand and thus are less powerful than the more data-demanding approaches. T hus, it can be observed that the freight demand models developed till da te need to compromise between the data requirements and the behavioural aspect. In this context, this study focuses on overcoming this limitation by developing a statewide behavioural freight demand model that is not data intensive. Markusen and Venables (1998) used the i ndustrial-organisation approach to model international trade. This model endogenously generates both national and multinational enterprises and goods flows. However, this mode l has not been applied in practice due to the extensive data requirement. The model re quires estimates of demand elasticity for each type of good, wages, transport cost factors, as well as utility functions to represent the consumers in each country. Garrido (2000) described a space-time autorregressive moving average model for truck flows through the Texas-Mexico border. The mode l data needs are series of international vehicle flows at different points in space, which is easily measured and available for public use. However, this model does not capture behavioural aspects of the freight movement.

PAGE 45

29 Input-output analysis has been used for both intercity and globa l freight transport modeling since 1940s. The basic in put-output model consists of a table that accounts the amount of a good involved in the production of another good, which is reflected by the purchases of an industrial sector from the re st of the industries within a given market. Leontief (1941) developed a standard approach assuming constant technical coefficients (i.e. the share of each good involved in the production of a given good), constant trade coefficients (i.e., ratio betw een the production of a good in a given location and the total production of that good), and constant modal split. These assumptions significantly lower the data requirements and the ma thematical treatment. However, these assumptions rarely hold in practice and hence th e prediction capability of this approach is rather scarce. Inamura and Srisurapanon (1998) developed a more pr actical approach by estimating a rectangular input-output model with fixed coefficients but disaggregated not only by products but also by regi on of origin and re gion of destination. The latter gives the model more flexibility than the or iginal fixed implementation by Leontief. 3.3.2 Intercity Freight Transport Models Intercity freight transport models have been widely addressed in the literature. Winston (1983) classifies these models into aggregate and disaggregate levels. 3.3.2.1 Aggregate Level Models Quandt and Baumol (1966) developed one of the first aggregate models reported in the literature which is called "abstract mode" mo del. It assumes that the freight travel demand for a mode depends on the attributes of that specific mode and the attributes of

PAGE 46

30 the available "best mode". However, even this model is data intensive that makes it impractical. Another early approach is the "a ggregate logit" modal split model (Morton, 1969; Boyer, 1977; Levin, 1978). This model is a log-linear regression whose dependent variable is the ratio between the market shar es of two modes. These models are attractive in practical applications, especially for larg e-scale problems, as the model's structure is very simple and not computationally demandi ng. However, the drawback is the lack of theoretical foundation. Oum (1979) analysed two aggregate modal split models used in practice: the "pricedifference" and "price-ratio" models. Oum showed that both specifications have weaknesses from the economic poi nt of view, arising when logit models are estimated with aggregate data. 3.3.2.2 Disaggregate Level Models Regan and Garrido (2002) iden tified two classes of disaggr egate freight demand models reported in the literature: the so-calle d "behavioural" and "inventory" models. Behavioural models focus on the mode choice decision made by either the consignee or the shipping firm, whereas inventory models analyse the freight demand from the viewpoint of an inventory manager. The behavioural models assume rational beha viour from the decision-maker and attempt to explain the freight travel demand as the result of a process of utility maximization made by the decision-maker. The drawback of this approach is that the decision maker

PAGE 47

31 must be identified before the data is gathered. This is not an easy task, especially for complex enterprises within a complex supply chain, where some decisions are simultaneously made (for instance transport mode and shipment size) and many different actors participate in the deci sion. The data needed for thes e kinds of models are the components of the level of service offered by th e different modes, such as rates, travel time, flexibility of the service, reliability, insu rance costs, etc. In addition, the choice set of each decision maker must be known to the modeller. The second type of disaggregate models are called inventory ba sed models. These models attempt to integrate the mode choi ce and the production decisions made by a firm (Baumol et. al., 1970; Das, 1974; McFadden et al., 1981). These type of disaggregate models can take the simultaneity of the deci sions into considerat ion (McFadden et. al., 1985). 3.3.3 Urban Freight Transport Models Regan and Garrido (2002) iden tified that the urban freigh t transport models are under developed with only a handful of published wo rks addressing the freight movement in the urban scope. Most of the litera ture on urban freight transpor t models deals primarily with vehicle flows, especially truck flows (He et. al., 1998; Gorys et. al., 1999). However, Harris and Liu (1998) predicted purchases and sales for different commodity categories, unlike the other studies, within and outside the city limits.

PAGE 48

32 In conclusion, the freight dema nd models that were simple and practical did not capture the behavioural aspects of freight demand. T hus, these models are less powerful than the more data-demanding behavioural approaches. These practical approaches are based on several assumptions that significantly lower the data requirements and the mathematical treatment. However, these assumptions rare ly hold in practice and hence the prediction capability of these approaches is rather limite d. The models that were able to capture the behavioural aspects of freight demand suff ered with two main drawbacks: the data requirement and the computational complexity of the model. This study addresses this problem by developing statewide freight dema nd models that are be haviourally valid and are less data intensive. Thus, this study pr oposes to use commodity-based modeling of inter-regional freight flows in Florida.

PAGE 49

33 Chapter 4 Model Framework 4.1 Introduction As can be concluded from the literature revi ew in the previous chapter, the freight demand models that captured th e behavioral aspects of frei ght demand are data intensive and impractical. Those models that are simple and practical do not capture the behavioral aspects of freight transportation in a comp rehensive manner. Thus, there has been a compromise with the behavioral aspect a nd data requirements in the freight demand models that were developed till date. As the objective of this thesis is to m odel statewide freight travel demand at a microscopic level, it is required to propose a relatively data simp le, but behaviorally robust modeling framework in the spirit of an aggregate level four-step planning process. Thus, the current research focuses on a stat ewide freight demand model that takes the structure of a commodity-based model for th e state of Florida. The model formulation and empirical analysis are specifically targ eted toward the trip generation, trip distribution and mode choice steps. Thus, th e goal is to propose a modeling framework that can quantify and predict freight flows by various modes between origin-destination pairs in the state of Florida. These origin-destination pairs may be traffic analysis zones, census tracts, zip codes, cities, counties, or ev en states depending on the particular freight

PAGE 50

34 transportation planning context of interest. But, it is desire d to conduct this study at the microscopic level of a zipcode unlike any of the earlier studies. The zip code level is considered an appropriate level of disaggrega tion where the data can be considered to be reliable avoiding large amounts of missing data. Thus, the intent of this chapter is to identif y the data requirements, paradigms and freight transportation demand-supply relationships id entified in the literature, and develop a model concept that will guide the model de velopment effort of this study. The data commonly used in the statewide freight demand modeling and the underlying paradigms are identified in section 4.2. Section 4.3 presents the development of the model framework based on the literature that is used in this study. 4.2 Data Requirements for Inter-regional Freight Demand Models The hypothesis commonly used in freight tr ansportation planning is that population is assumed to affect the attraction of freight to an area, and industry employment is assumed to affect the generation of fr eight in an area. Thus, it can be easily learnt from the literature review in the previous chapter that the data mainly required in freight demand modeling is the socio-economic information of the origin an d destination. The other data that is required for the estimation of freight demand models is the modal level of service characteristics. Data on estimates of future dwelling construc tion and other major c onstruction sites (e.g. new road or rail links, or major urban redevelopment sites) could be used to estimate the

PAGE 51

35 demand for construction related commodities, which comprise a significant proportion of all overall freight transport. Moreover, land use data may also be used for estimating the location of production for agricultural production. Garrido (undated) reviewed the data requi rements for inter-regional freight demand models and proposed that the data on modal split, fleet’s attribut es and composition, and network characteristics are required. At the ag gregate level, models typically regress the proportion of market shares (between pure mode s) against some aggregate attributes such as prices, travel time and cost, etc. Therefore, accurate data on modal split and some level of service attributes is requi red. Inter-regional models are especially sensitive to the network resolution and level of service. Routing options as we ll as the costs at each arc have a tremendous impact on the quality of mo del results. The network costs structure is especially relevant when the freight flows ar e found as a result of an equilibrium process – usually under Wardrop’s second principle (Friesz et. al., 1983). Sivakumar and Bhat (2002) also propose that the data requirements for inter-regional freight demand models are freight origin-des tination flow data, business and employment patterns, population projections, regional ec onomic information and modal level of service variables. Using these data, they developed an approach that estimates the fraction of commodity consumed at each destina tion zone that originates from alternative production zones. The resulting fractional split model for commodity flow distribution is more general in structure than the typical gravity model used for statewide freight

PAGE 52

36 planning. This empirical analys is applying the fractional spli t model was used to analyze inter-regional commodity flows in Texas. Brogan et al (2001) also proposed to use da ta such as population, employment, regional size, per-capita income, population density, da ily electric coal demand, KW capacity and Coal tons/KW. In their study, county-level commodity flow data were commercially procured to describe freight flows into, out of, within, and through Virginia. With the use of these data, they identified Virginia’s “key” commodities and the flows of these commodities were assigned to county-level orig in-destination tables. Predictive equations of freight generation and attr action relationships for each of Virginia’s key commodities were developed. A strategy for developing re gression equations was developed using a series of robust and stepwise regressions to minimize the effects of outliers. For each key commodity with a two-digit Standard Tran sportation Commodity Classification code, a set of generation and attraction equations was developed, including relationships for nonoutliers, first-order outliers, and second-order outliers. In a ddition, the authors identified several socioeconomic variable s that significantly affect fr eight generation and attraction within Virginia. Cambridge Systematics, Inc (1996) also pr oposes obtaining data on economic activity for traffic analysis zones (inc luding employment by type and the number of households) in their Quick Response Freight Manual (QRFM). Jack Faucett Associates (1999b) also propose their trip-based modeling process by generating truck trips as a function of different land uses and trip data from trip diaries or shipper surveys. They calculate the

PAGE 53

37 trip rates as a function of so cio-economic data (trips per em ployee) and land use data (trip per acre). 4.3 Model Framework Thus, in order to be in line with paradi gms and freight transportation demand-supply relationships identified in the literature, it is desired to use populat ion characteristics, employment characteristics, and the modal leve l of service characteri stics in the statewide freight demand modeling for Florida. Growth and land use data are not included in this study as inclusion of such data hinders the prac tical application of th e models that will be developed. Moreover, relevant data sources could not be identified. However, population data indirectly captures the effects of these kinds of data. Thus, only socio-economic data and the modal level of service data ar e considered for model development. Figure 4.1 shows the overall model framewor k that directed the model development effort of this study. The following assumptions are made in forming this model concept: Origin and destination employment charac teristics are assumed to influence the flow of a commodity between an origin-d estination pair and the amount of flow by each mode. Origin and destination population characte ristics are assumed to influence the total flow of a commodity between an or igin-destination pair and the amount of flow by each mode.

PAGE 54

38 Modal level of service characteristics incl uding travel distance, travel time, and travel cost influence the total flow of a commodity between an origin-destination pair and the amount of flow by each mode. The total freight flow of a commodity between an origin destination pair has an influence on flows by each mode. Figure 4.1 Conceptual Framework for Modeling Freight Transportation Demand The motivation for these assumptions is th at freight movement is fundamentally generated by the demand for consumption of commodities at the destination (or attraction) end, which is met by the flow of commodities from one or more origin (or Population Characteristics Employment Characteristics Modal Level of Service Characteristics Total Flow Flow on Mode 1 Flow on Mode 2 Flow on Mode 3 Flow on Mode ‘n’ Population Characteristics Employment Characteristics Origin Area Destination Area

PAGE 55

39 production) points (Ogden, 1978). This model framework represents a direct demand model where the movement of a commodity by a certain mode is esti mated directly from socio-economic characteristics of the origin and destination and the modal level of service variables. A link is also provided from the total flow to the modal flows to accommodate any influence that total freight flow has on individual modal flows. Thus the model framework provides a mech anism by which freight flows on various modes between an origin and destination can be estimated. The changes in freight flows can be determined in response to changes in the socio-economic characteristics of the origin or destination, and changes in modal a ttributes. The model framework is simple & practical, but behaviorally robust and can ther efore be easily estimated on a database that can be assembled by any public agency that has resources to purchase some commercial databases.

PAGE 56

40 Chapter 5 Database Preparation 5.1 Introduction From the model framework developed in the prev ious chapter, it is required to obtain the freight data containi ng commodity flows by various modes for origin-destination pairs in Florida. The other data that is required to estimate these modal flows between O-D pairs in Florida are the socio-economic characteri stics of the origin & destination and the modal level of service characteristics. A review of the various freight data s ources is conducted in order to choose an appropriate database for this study. The intent is to identify a database that supports commodity based modeling of statewide inter -regional flows at a microscopic level. 5.2 Available Freight Data Sources 5.2.1 Commodity Flow Survey http://www.census.gov/econ/www/se0700.html The Commodity Flow Survey (CFS) provides da ta on the flow of goods and materials by mode of transport. The CFS is sponsored by the Bureau of Tran sportation Statistics, United States Department of Transporta tion (USDOT), and U.S. Census Bureau,

PAGE 57

41 Department of Commerce, and performed by the U.S. Census Bureau, Department of Commerce and Oak Ridge National Laboratory. The CFS follows a series of publicly avai lable datasets from 1963 through 1997. Samples of domestic establishments engaged in mining, manufacturing, wholesale, auxiliary establishments (warehouses) of multi-estab lishment companies, and some selected activities in retail and service were used to collect the data through th e completion of a questionnaire. The current version of the CFS contains a geographic co verage of data at national level, stratified by State and Metr opolitan Area (Garrido, undated). It provides information on commodities shipped, their valu e, weight and mode of transportation, as well as the origin and destination of shipments at the national, state, and large metro-area levels. Thus, it is quite a detailed aggregate level dataset that can be used to study overall trends in commodity flows between major geographic areas. It provides a convenient mechanism to obtain control totals regarding freight movements. Although data from the Bureau of Transpor tation Statistics Commodity Flow Survey (CFS) is widely used in other types of st udies, several inherent weaknesses of the CFS make it inappropriate for use in this study. Firs t, CFS data are available only at the state level; Zipcode level data are more appropria te for use in a statewide freight planning process. If CFS data were to be used in this study, a methodology to disaggregate the statewide commodity flows to individual zipc odes would have to be developed. Such a disaggregation process, most likely based on zipcode-level employment and population, would affect the accuracy of the final (zipcode-level) commodity flow data.

PAGE 58

42 The other drawback is that CFS data is not comprehensive. Because the CFS is published by the U.S Census Bureau, it must comply with federal law governing census reports, including the prohibition of publishing data that would disclose the operations of an individual firm or establishment. As a result, much of the data are not published, severely reducing the accuracy and scope of the CFS. 5.2.2 Transborder Surface Freight Dataset http://www.bts.gov/ntda/tbscd/ Since 1993 the Bureau of Transportation Statis tics (BTS) at the United States Department of Transportation (USDOT) has contracted with Bureau of the Census (Census) at the U.S. Department of Commerce (DOC) to provide previously unpublished surface transportation data (other than air or maritime vessel) for U.S. import and export trade with Canada and Mexico. This dataset is referred to as the Transborder Surface Freight Data. Under the contract, Census provides two sets of data tables to BTS; one provides detailed transportation flows while the ot her is commodity based without as much transportation detail. After Census processes and summarizes the data, BTS receives these monthly files and makes them public ly available as soon as possible. The Transborder Surface Freight Dataset provides North American merchandise trade data by commodity type, by surface transport mo de (including pipeline) with geographic detail for U.S. exports to and imports fr om Canada and Mexic o, updated on a monthly basis. Its objective is to provide transportation informa tion on North American trade flows. The source is the official U.S. in ternational merchandise trade dataset.

PAGE 59

43 Currently, these data are being used to m onitor transborder freight flows since the beginning of the North American Free Trade Agreement (NAFTA) in 1994. Other uses of this database are: trade corridor studies, transportation infrastruct ure planning, logistics strategy analyses amongst other purposes. The dataset is compiled from the Census Fo reign Trade Statistics Program. Import and export data are collected fr om administrative records re quired by the Departments of Commerce and Treasury of the US. Most of the imports data from Canada and Mexico, are collected electronically via an Automated Broker Interface, and the Customs entry documents collected by the Customs Service and transmitted to the Census Bureau. Information on U.S. exports of goods from the U.S. to all countries (except Canada) is compiled from copies of Shipper's Export Declarations (SEDs) and data collected from shippers, forwarders or carriers. On the export side about half of the data are collected electronicall y, through a U.S./Canada Data Exchange agreement and the Automated Export Reporting Program. The official U.S. import and export stat istics provide information on shipments of merchandise between foreign countries and the U.S. Customs Territory, U.S. Foreign Trade Zones, and the U.S. Virgin Islands, w ithout regard to whether or not a commercial transaction is involved. The statistics record the physical movement of merchandise between the United States and foreign countries Thus, this dataset is inappropriate for use in the statewide freight demand modeling for Florida.

PAGE 60

44 5.2.3 Transportation Annual Survey http://www.census.gov/econ/www/se0800.html The Transportation Annual Survey is formerly known as the Motor Freight Transportation and Warehousing Survey. It is carried out annually by the US Census Bureau. It provides national estimates of reve nue, expenses, and vehi cle fleet inventories for commercial motor freight transportati on and public warehousing service industries. This survey covers companies with employ ment that provide commercial motor freight transportation and public warehous ing services. It excludes pr ivate motor-freight carriers operating as auxiliary establishments to non-transportation companies and independent owner-operators with no paid employees. The survey covers all employer firms with one or more establishments that are primarily e ngaged in providing commercial motor freight transportation or public warehousing services. Th e results of this su rvey are published in a report where statistics are summarised by kind-of-business classification based on the Standard Industrial Classificat ion (SIC) Manual issued by th e Office of Management and Budget. This survey is conducted annually since 1985. Data collection begins about 3 months after the reporting y ear and continues for about 4 months. Samples are selected every 5 years and updated annually. Even this dataset is inappropriate for use in the statewide freight demand modeling, as it prov ides only the national estimates of revenue, expenses, and vehicle fleet inventories fo r commercial motor freight transportation.

PAGE 61

45 5.2.4 Vehicle Inventory and Use Survey http://www.census.gov/econ/www/se0501.html This survey is formerly known as the Tr uck Inventory and Use Survey (TIUS). The survey name was changed to account for areas of future expansion, including the addition of automobiles and buses. The aim of this survey is to measure the physical and operational characteristics of the Nation's truck population. Th is survey covers private and commercial trucks registered (or licensed) in the United States as of July 1 of the survey year. The survey excludes vehicles ow ned by Federal, state, or local governments; ambulances; buses; motor homes; farm tractors; unpowered trailer units; and trucks reported to have been sold, junked, or wrecked prior to July 1 of the year preceding the survey. The dataset on physical characteristics include date of purchase, weight, number of axles, overall length, type of engine and body type. The operational characteristics data include type of use, l ease characteristics, operator classification, base of operation, gas mileage, annual and lifetime miles driv en, weeks operated, commodities hauled by type, and hazardous materials carried. Several private and public agencies use these data on a regular basis. Public agencies such as the Department of Transportation use the data for analysis of cost allocation, safety issues, proposed investments in ne w roads and technology, and user fees. The Environmental Protection Agency uses the data to determine per mile vehicle emission estimates, vehicle performance and fuel econom y, and fuel conservation practices of the trucking industry. The Bureau of Economic An alysis uses the data as a part of the

PAGE 62

46 framework for the national investment and personal consumption expenditures component of the Gross Domestic Product. Private agencies such as tire manufacturers use the data to calculate the longevity of products and to determine the usage, vocati on, and applications of their products. Heavy machinery manufacturers use the data to track the importance of various parts distribution and service networks. Truck ma nufacturers use the data to determine the impact of certain types of equipment on fuel efficiency. This survey is conducted every 5 years since 1963, for years ending in "2" and "7." Data collection begins in January following the cens us year and continues for approximately 9 months. Reported data are for activity during the census calendar year. This dataset is inappropriate for use in the statewide fr eight demand modeling, as it covers only the nation’s truck population. 5.2.5 Reebie TRANSEARCH Database http://www.reebie.com/images/transearch.asp The TRANSEARCH database contains origin-d estination freight movements in the US covering major modes of transport, and suits best for statewide fr eight demand modeling. It is compiled and produced on an annual basis since 1980 by the firm Reebie Associates. Records are kept for freight traffic sh ipments across geographic markets and commodities for seven modes of transport, including truckload, less than truckload (LTL), private truck, rail, in termodal, rail carload, water borne, and air. The database

PAGE 63

47 contains the freight activity of U.S. do mestic, Canada/U.S. and Mexico/U.S. This database has been used in va rious statewide freight models such as Texas and Virginia (Sivakumar and Bhat, 2002; Brogan et. al., 200 1). This dataset is the best commodity flow data available, and has been chosen for this study. The Florida Department of Transportation (FDO T), as part of its ongoing research into the development of urban and statewide fr eight transportation models, purchased the TRANSEARCH freight flow database at th e zip code level. The TRANSEARCH data may also be prepared and purchased at other levels of aggregation; however, the zip code level was considered an appropriate level of disaggregation where the data could be considered reliable and large amounts of missing data could be avoided. 5.3 Preparation of the Freight Da ta from TRANSEARCH Database Reebie Associates’ TRANSEARCH database at the zipcode level for the state of Florida was used as the primary source of commodity flow data in this study. Although it is the best commodity flow data currently availa ble, the TRANSEARCH database suffers with several limitations: The commodity flow data consists of a national database built from companyspecific data and other publicly available databases. These different data sources use different commodity classifications th at must be converted to a consistent format. These conversions can sometimes lead to some data being put in the wrong category or left unreported.

PAGE 64

48 The level of detail provided from particular companies when they report their freight shipment activities to Reebie Asso ciates can limit the accuracy of the final TRANSEARCH dataset. Specific origin-destination (O-D) inform ation is not available for overseas waterborne traffic through marine ports Overseas ports are not identified, and Reebie Associates estimates the domes tic distribution of maritime imports and exports. The TRANSEARCH 2000 database at the zi p code level, purchased by the FDOT consists of commodity flows (by ton) into, out of, and through Florida. However, in order to keep the model estimation database tractable, only those commodity flows that originated and ended at zip codes in Florid a were used for model estimation. Remaining freight flow data was availa ble only at the county and Bu reau of Economic Analysis level. This study utilizes only the zip code-z ipcode level flow data for model estimation and thus captures only about 70 percent of the in tra-state freight flow within Florida. At this point, it should be noted that the TR ANSEARCH database is not necessarily a complete and comprehensive coverage of al l freight transportati on flows. There are certain types of movements that are not captured in the TRANSEARCH database. Thus, the freight database prepared for this study contained commodity flows at the zip code to zip code level with commodities classi fied at the level of the two-digit Standard Transportation Commodity Classification (S TCC) code. In order to reduce the commodity groups to a more manageable leve l, the commodities at the two digit STCC

PAGE 65

49 level were collapsed into 17 commodity group s. The STCC codes that were collapsed into the 17 commodity groups are shown in Table 5.1. Table 5.1 STCC Commodity Classification Groups Commodity Group STCC in Commodity Group Code Name Code Name 1 Agriculture 1, 8, 9 Agriculture, Forest Products, Fisheries 2 Coal 11 Coal 3 Other Minerals 13, 14, 19 Crude Petroleum, Nonmetallic Minerals 4 Food 20 Food 5 Non-Durable Manufacturing 21, 22, 23, 25, 27 Tobacco, Textiles, Apparel, Furniture, Pr inted Goods 6 Lumber 24 Lumber 7 Paper 26 Paper 8 Chemicals 28 Chemicals 9 Petroleum 29 Petroleum 10 Rubber Plastics 30 Rubber/Plastics 11 Durable Manufacturing 31, 36, 38, 39 Leather, Electrical Equipment, Instruments, Miscellaneous Manufacturing Products 12 Clay, Concrete, Glass 32 Clay, Concrete, Glass 13 Primary Metals 33 Metal 14 Fabricated Metal Products 34 Metal Products 15 Transportation Equipment 37 Transportation Equipment 16 Miscellaneous Freight 40-48 Waste, Miscellaneous Freight Shipments, Shipping Containers, Mail, Freight Forwarder Traffic, Shipper Association Traffic, Miscellaneous Mixed Shipments, Small Packaged Freight, Hazardous Materials/Waste 17 Warehousing 5010, 5020, 5030 Secondary Traffic, Truck Intermodal, Truck Air Drayage Commodity flows were broadly assigned to four modes: truck, rail, water, and air. The truck mode was further subdivided into full tr uck load, less than truck load, and private truck load. Full truck load was defined as a “for-hire” commodity flow on a truck with greater than 10,000 pounds, and less than truck lo ad was defined as “for-hire” commodity

PAGE 66

50 flow on a truck with less than 10,000 pounds. A Private Truck was defined as any company truck that is part of a private fleet The Rail mode was also subdivided into rail intermodal and rail car load modes. Thus, th e TRANSEARCH dataset consists of freight flows between pairs of zip codes in Florid a by commodity (two-digit STCC code) and in annual weight (in tons) for each of the followi ng modes of transport: full-truck-load, lessthan-truckload, private truck, rail carl oad, rail intermodal, water, and air. 5.4 Socio-Economic Data In order to model freight flows between zip co des, three more pieces of information were required as per the model construct presente d in Figure 3. Socio-economic information represented by population and employment char acteristics was needed. All population information was derived from the 2000 Census databases. Census data was obtained at the zip code level and appropriately ma tched to the TRANSEARCH commodity flow database so that each record contained the population char acteristics of the origin zip code and the destination zip code. The fo llowing information for both origin and destination zipcodes was merged to the freight data: Overall Zipcode Population Number of Males Number of Females Number of Persons with Age between 0 and 5 years Number of Persons with Age between 6 and 10 years Number of Persons with Age between 11 and 15 years Number of Persons with Age between 16 and 21 years

PAGE 67

51 Number of Persons with Age between 22 and 24 years Number of Persons with Age between 25 and 29 years Number of Persons with Age between 30 and 34 years Number of Persons with Age between 35 and 39 years Number of Persons with Age between 40 and 44 years Number of Persons with Age between 45 and 49 years Number of Persons with Age between 50 and 54 years Number of Persons with Age between 55 and 59 years Number of Persons with Age between 60 and 64 years Number of Persons with Age between 65 and 69 years Number of Persons with Age between 70 and 79 years Number of Persons with Age between 80 and 84 years Number of Persons with Age 85 years or above Number of Households with H ousehold Incomes less than $ 20,000 Number of Households with House hold Incomes between $ 20,000 and $ 39,999 Number of Households with House hold Incomes between $ 40,000 and $ 59,999 Number of Households with House hold Incomes between $ 60,000 and $ 99,999 Number of Households with Hous ehold Incomes between $ 100,000 and $ 199,999 Number of Households with H ousehold Incomes $ 200,000 or above Number of Persons whose Occupations ar e Management, Business, or Financial Operations Number of Persons whose O ccupations are Professional

PAGE 68

52 Number of Persons whose Occupa tions are related to Service Number of Persons whose Occupations are related to Sales or Office Number of Persons whose Occupations are Farming, Fishing, or Forestry Number of Persons whose Occupati ons are Construction, Extraction, or Maintenance Number of Persons whose Occupations are Production, Transportation, or Material moving Number of Households with One Person Number of Households with Two Persons Number of Households with Three Persons Number of Households with Four or more Persons Number of Families with Zero Workers Number of Families with One Worker Number of Families with Two Workers Number of Families with Three or more Workers Number of Households with Zero Vehicles Number of Households with One Vehicle Number of Households with Two Vehicles Number of Households with Three or more Vehicles Number of Persons whose Educationa l Attainment is below 9th grade Number of Persons whose Educational Attainment is between 9th and 12th grades, no diploma, High school gra duate, Some college or no degree

PAGE 69

53 Number of Persons whose Educational A ttainment is Associate's Degree or Bachelor's Degree Number of Persons whose Educational A ttainment is Graduate or Professional Degree Employment characteristics were derived fr om the InfoUSA 2000 database which is a commercial database that contains informati on on every employer in th e state of Florida. This database was purchased by the Florida Department of Transportation for use by public agencies in developing employment characteristics for their travel demand models. This database has information about each employer, such as Company name, Contact name, Address, City, State, Zipcode, Phone SIC code, Franchise code, and Employee size. Aggregation of this employer database was performed at the zip code level and information on employment by one digit SIC wa s obtained. This data was appropriately matched to the TRANSEARCH commodity flow database so that each record contained the employment characteristics of the origin zip code and the destination zip code. Thus, the following information for both origin and destination zipcodes was merged to the freight data: Overall Zipcode Employment Agricultural, Forestry and Fishery Employment Mining and Construction Products Employment Light Manufactured Products Employment Heavy Manufactured Products Employment Transportation, Communicati on and Utilities Employment

PAGE 70

54 Wholesale and Retail Trade Employment Finance, Insurance and Real Estate Employment Entertainment, Accommodation a nd Food Services Employment Other Services Employment Public Administration Employment 5.5 Modal Level of Service Data Finally, as per the model construct, detail ed information on modal level of service variables is needed. Different modes have different service characteristics, and the commodities carried by those modes differ. The truck and air modes tend to be dominated by low-weight, high-value commod ities, such as automobile and computer parts. Conversely, th e rail and water modes tend to be dominated by high-weight, lowvalue commodities, including coal, gravel, and timber. Thus, for every zip code pair, it would be ideal to have travel time, distance, and cost information by all modes identified in the database. This effort is currently ongoing and as such all modal level of service variables have not yet been merg ed into the database. Thus, at this time, the models are estimated using simple map distance (center of zip code to center of zip code) as a measure of impedance between them. However, in the future, modal level of service variables should be included in the model specification to ma ke sure that the model is sensitive to modal level of service attributes In its current form (as presented in this thesis), the model is sensitive only to distance. The importance of the role of distance and trip length in freight transpor tation modeling is well recogni zed (Holguin-Veras et. al., 2000; Garrido et. al., 2000).

PAGE 71

55 5.6 Final Dataset The origin & destination popul ation & employment characteristics were merged to the freight flow database as sepa rate variables such as origin population, origin employment, destination population, destination employment etc. The distance between the origin and destination zipcodes was also merged to the freight database. This database was developed using SPSS, which is a statistical software package developed for use in the social sciences. Each record in this database consists of the following data: Origin Zipcode Destination Zipcode Commodity Group Total Flow (in annual tons) Total Flow by Truck (in annual tons) Total Flow by Rail (in annual tons) Commodity Flow by Air (in annual tons) Commodity Flow by Water (in annual tons) Commodity Flow by Full Truck Load (in annual tons) Commodity Flow by Less-Than-T ruck-Load (in annual tons) Commodity Flow by Privat e Truck (in annual tons) Commodity Flow by Rail Car (in annual tons) Commodity Flow by Rail Intermodal (in annual tons) Origin Population Characteristics (49 popul ation variables shown in section 4.4) Destination Population Characteristics ( 49 population variables shown in section 4.4)

PAGE 72

56 Origin Employment (11 employme nt variables in section 4.4) Destination Employment (11 empl oyment variables in section 4.4) Distance Thus, the final dataset consists of comm odity flows (by the 17 defined commodity groups) in annual tons by each mode (full truc k load, less than truck load, private truck, rail car load, rail intermodal, air, wa ter rail and truck), along with population, employment and distance information fo r all pairs of zip codes in Florida.

PAGE 73

57 Chapter 6 Database Description 6.1 Introduction As the basic objective of this study is to develop a statewide modeling framework for estimating freight flows for various commod ities on various modes for origin-destination zipcodes, it is important to understand the distribution of freight flows by commodity, mode, region and socio-economic characteristic s. Understanding these distributions can help identify and explain the fundamental re lationships between among freight flows (by commodity, mode and region) and the soci o-demographics. The following sections describe the data prepared for this study. 6.2 Distribution of Freight Flows by Commodity and Mode The distribution of total annual flow (w eight) by commodity group at the zipcodezipcode level in Florida is s hown in Table 6.1. As this databa se is focusing on intra-state movements, the warehousing commodity group is found to account for more than 50 % of the flows by weight. Other major commodity groups include other minerals (15.09 %), clay, concrete & glass (14.52 %), Chemicals (6.01 %), and Food (4.49 %). The overall mode share of total annual comm odity flow by weight at the zipcode-zipcode level in Florida is shown in Table 6.2. The truck mode accounts for the major portion of

PAGE 74

58 overall freight flow, carrying 77.6 % of the total annual commodity flows. The truck mode is a combination of two types of truc ks, for-hire and private truck. For-hire truck and private truck account for 37.3 % and 40.2 % of the total annual commodity flow respectively. Thus, for-hire truck and the pr ivate truck account for 48 % and 52 % of the total commodity flow carried by truck respectively. For-hire truck mainly constitutes of Full Truck Load which accounts for 36.1 % of the total annual comm odity flow, while the Less-than-Truck Load accounts for a small share of 1.2 % of the total annual commodity flow. Rail mode accounts for about 20 % of the total annual commodity flow. Air accounts for a very small share of commodi ty flow by weight at less than one percent while water accounts for a slig htly higher share at 2.8 %. Table 6.1 Distribution of Freight Flows by Commodity Group (Weight) Commodity Group Total Code Name Weight (tons) Percentage 1 Agriculture 71,258 0.07 2 Coal 149,729 0.15 3 Other Minerals 15,149,353 15.09 4 Food 4,505,846 4.49 5 Non-Durable Manufacturing 844,395 0.84 6 Lumber 1,815,571 1.81 7 Paper 542,107 0.54 8 Chemicals 6,035,128 6.01 9 Petroleum 2,396,885 2.39 10 Rubber Plastics 80,190 0.08 11 Durable Manufacturing 57,924 0.06 12 Clay, Concrete, Glass 14,582,320 14.52 13 Primary Metals 220,888 0.22 14 Fabricated Metal Products 455,012 0.45 15 Transportation Equipment 150,569 0.15 16 Miscellaneous Freight 1,917,070 1.91 17 Warehousing 51,427,628 51.22 Total 100,401,873 100.00

PAGE 75

59 Table 6.2 Distribution of Freight Flows by Mode Mode Share Mode Weight (tons) Percentage Truck Load (TL) 36,242,617 36.10 Less Than Truck Load (LTL) 1,217,705 1.21 Private Truck (PVT) 40,411,201 40.25 Railcar (CL) 17,630,641 17.56 Rail Intermodal (IMX) 2,012,488 2.00 Air 120,521 0.12 Water 2,765,268 2.75 Truck 77,872,937 77.56 Rail ** 19,643,147 19.56 Total*** 100,401,873 100.00 *Truck = TL + LTL + PVT **Rail = CL + IMX ***Total = Air + Water + Truck + Rail Table 6.3 presents mode shares by commod ity group at the zipcode -zipcode level in Florida. This table shows the 17 commod ity groups and the mode share for each commodity in percent by weight. Thus it can be observed that coal is carried completely by rail car while warehousing is completely mo ved by truck. It can be concluded that the truck mode is generally dominated by low-weight, high-value commodities, such as fabricated metal products and non-durable ma nufacturing. Conversely, the rail mode is dominated by high-weight, lowvalue commodities, such as coal. Differences across commodity groups with respect to modal share are quite impo rtant and noticeable. As the commodities vary with respect to density, value, and time-sensitivity, there may be fundamental differences in the relationships am ong variables that can be used to predict

PAGE 76

60 their flows. Therefore, in this study, it is proposed to estimate freight flows separately for each commodity group. Table 6.3 Distribution of Freight Flow s by Commodity and Mode Commodity Group Truck Rail Total Code Name TL LTL PVT CL IMX Air Water Truck Rail Total 1 Agriculture 48.4 0.0 17.2 0.0 27.3 7.1 0.0 65.6 27.3 100.0 2 Coal 0.0 0.0 0.0 100.0 0.0 0.0 0.0 0.0 100.0 100.0 3 Other Minerals 0.0 0.0 0.0 95.0 0.0 0.0 5.0 0.0 95.0 100.0 4 Food 32.8 0.5 63.6 1.8 1.2 0.0 0.0 96.9 3.0 100.0 5 Non-Durable Manufacturing 29.8 2.8 65.2 0.0 0.7 1.2 0.3 97.8 0.7 100.0 6 Lumber 45.9 0.8 47.7 5.6 0.0 0.0 0.0 94.4 5.6 100.0 7 Paper 30.8 7.0 54.7 4.6 2.3 0.6 0.0 92.4 6.9 100.0 8 Chemicals 55.3 0.5 0.1 43.3 0.8 0.1 0.0 55.8 44.1 100.0 9 Petroleum 5.4 0.1 16.3 0.0 0.1 0.0 78.0 21.9 0.2 100.0 10 Rubber Plastics 24.4 6.6 67.9 0.0 0.7 0.0 0.4 98.9 0.7 100.0 11 Durable Manufacturing 25.1 5.5 20.3 0.0 8.7 39.8 0.6 50.9 8.7 100.0 12 Clay, Concrete, Glass 30.2 0.1 67.3 1.2 1.3 0.0 0.0 97.5 2.5 100.0 13 Primary Metals 92.2 0.6 0.0 3.7 2.0 0.0 1.5 92.8 5.7 100.0 14 Fabricated Metal Products 42.4 2.1 50.1 0.2 0.1 4.4 0.8 94.5 0.3 100.0 15 Transportation Equipment 86.7 2.8 0.4 4.7 0.0 3.8 1.6 89.9 4.7 100.0 16 Miscellaneous Freight 0.0 0.0 0.0 3.8 87.1 2.5 6.7 0.0 90.9 100.0 17 Warehousing 48.7 2.0 49.2 0.0 0.0 0.0 0.0 100.0 0.0 100.0 6.3 Distribution of Freight Flows by Trip Length Table 6.4 provides an overall trip length dist ribution across all co mmodity groups at the zipcode-zipcode level in Florid a. This distribution should be viewed in light of the intrastate nature of the frei ght flow database. More than one-t hird (37.1 %) of the freight flow

PAGE 77

61 by weight travels less than 50 miles. Intra Zipcode flows constitute 12.5 % of the overall share of freight transport in Florida. Only about 20 % of the commodity flow by weight travels farther than 250 miles in this intra-st ate context. Obviously, this distribution will vary greatly from state to state; however, it is important to note that the trip length (distance between origin and dest ination) is likely to play a role in determining the freight flow. Therefore, the model systems devel oped in this study incl ude distance as an explanatory variable. More ideally, future model systems should contain detailed modal level of service information to ensure that the models are sensitive to modal level of service attributes. Table 6.4 Distribution of Freight Flows by Trip Length Total Trip Length Weight Percentage Intra Zipcode Flows 12,563,003 12.5 1 10 4,786,842 4.8 10 20 3,828,755 3.8 20 30 5,698,571 5.7 30 40 5,424,241 5.4 40 50 4,993,672 5.0 50 100 9,919,012 9.9 100 150 11,650,764 11.6 150 200 19,324,143 19.2 200 250 5,190,897 5.2 250 300 10,108,793 10.1 300 350 2,680,752 2.7 350 400 1,869,830 1.9 400 450 556,064 0.6 450 500 1,806,534 1.8 Total 100,401,873 100.0

PAGE 78

62 Table 6.5 Distribution of Freight Flows by Trip Length and Mode Truck Rail For-Hire Private Car Intermodal Total Trip Length FTL LTL PVT CL IMX Air Water Truck Rail Total Intra Zipcode 57.20 0.17 5.59 27.27 0.00 0.95 8.81 62.96 27.27 100.00 1 10 49.70 1.29 38.94 10.07 0.00 0.00 0.00 89.93 10.07 100.00 10 20 36.03 0.73 26.07 37.00 0.00 0.00 0.18 62.83 37.00 100.00 20 30 27.86 1.16 37.75 32.07 0.00 0.00 1.16 66.76 32.08 100.00 30 40 13.09 0.51 18.55 66.72 0.00 0.01 1.13 32.14 66.72 100.00 40 50 24.08 1.05 34.22 40.46 0.00 0.00 0.18 59.35 40.46 100.00 50 100 33.44 1.44 49.97 11.20 0.02 0.00 3.93 84.85 11.22 100.00 100 150 35.78 1.50 53.07 8.22 0.10 0.00 1.32 90.36 8.32 100.00 150 200 36.12 1.66 52.91 7.12 0.53 0.00 1.66 90.69 7.65 100.00 200 250 34.62 1.60 50.78 4.69 5.13 0.00 3.17 87.01 9.82 100.00 250 300 27.96 1.20 40.35 10.65 16.12 0.00 3.72 69.50 26.77 100.00 300 350 37.19 1.69 54.78 2.26 0.00 0.00 4.09 93.65 2.26 100.00 350 400 40.14 1.90 57.70 0.26 0.00 0.00 0.00 99.74 0.26 100.00 400 450 41.30 1.19 57.25 0.26 0.00 0.00 0.00 99.74 0.26 100.00 450 500 40.51 1.72 57.40 0.37 0.00 0.00 0.00 99.63 0.37 100.00 The distribution of flows by trip length and mo de at the zipcode-zipc ode level in Florida are presented in Table 6.5. In the intra-state na ture of the freight flow database, it is evident from the table that truck dominates all other modes. Thus, truck is the most important mode to be considered in frei ght demand modeling. Rail is a dominant mode for trips with lengths around 30–40 miles. It al so has a considerable mode share for all short trip lengths (0-50 m iles) and trip lengths from 250-300 miles. Again, these distributions should be viewed in light of the intra-state nature of the freight flow database. The distribution of ton-mile s in Florida is shown in Figure 6.1.

PAGE 79

63 Figure 6.1 Distribution of Ton-Miles in Florida 0.00 2.00 4.00 6.00 8.00 10.00 12.001 4 ,000 4,0 01 12,0 00 12,001 16,50 0 16,50 1 25,000 25 ,001 40 ,00 0 40,001 68,00 0 68,00 1 115,0 00 115,001 200,00 0 200,001 275,000 27 5,001 4 00,00 0 40 0,00 1 1,00 0,000 > 1 ,000,000Ton-MilesPercentage 6.4 Distribution of Freight Flows by Region The distribution of freight fl ows by county is presented in a tabular format in Table 6.6. The outflows and inflows at the county level are given in this table. It can be observed that Miami-Dade has the highest freight ou tflows in Florida, followed by Polk and Hillsborough counties. Miami-Dade accounts for about 22.4 % of freight exports in Florida, while Polk and Hillsborough account for 14.8 % and 8.7 % respectively. It can also be noticed that Hillsborough has the highe st freight inflows in Florida, followed by Miami-Dade and Duval. Hillsborough accounts for about 18.5 % of freight imports in Florida, while Miami-Dade and Duval accoun t for about 14.1 % and 11.7 % respectively.

PAGE 80

64 Table 6.6 also presents the export to import ra tio for counties in Florida. Hardee has the highest export to import ratio of 11.01, fo llowed by Liberty and Suwannee at 4.42 and 2.78 respectively. It can be obser ved in general that Florida imports higher than what it exports. Thus, only a few count ies with export to import ra tio greater than 1 can be identified. The export to import ration fo r the entire state of Florida is 0.78. TRANSEARCH data indicates that only 287 million annual tons of freight is exported, while 367 million annual tons is imported at th e county level in Florida. However, as mentioned earlier, this study focuses on zipcode -zipcode flows in Florida that account for about 100.5 million annual tons. Table 6.7 presents the distributi on of freight outflows by county and mode for the state of Florida. It can be observed that there are many counties such as Baker, Citrus, Dixie, Flagler etc. that export only by truck. It can also be observed that most of the counties have high shares of outflows using truck. Thus it is evident that truck is the dominant mode in Florida. Rail is a dominant mode of export in a few counties such as Bradford, Gadsden, Glades, Hamilton, Hardee, Hernando, Liberty and Polk. Water is a dominant mode of export in the Charlo tte and Wakulla counties. Air does not have a significant share in any of the counties. Similarly, Table 6.8 presents the distribution of freight inflows by county and mode for the state of Florida. It can be observed as in the case of frei ght outflows, that most of the counties have high shares of inflows using truck. Rail comes next to truck and has a significant share in many counties.

PAGE 81

65 Table 6.6 Distribution of Freight Flows by County FIPS COUNTY Outflows Percentage Inflows Percentage Export to Import Ratio 12001 ALACHUA 2,834,631 0.99 2,324,574 0.63 1.22 12003 BAKER 351,227 0.12 192,723 0.05 1.82 12005 BAY 3,356,830 1.17 4,441,292 1.21 0.76 12007 BRADFORD 535,762 0.19 256,580 0.07 2.09 12009 BREVARD 5,403,695 1.88 7,542,605 2.06 0.72 12011 BROWARD 13,587,432 4.73 17,073,597 4.65 0.80 12013 CALHOUN 91,818 0.03 189,697 0.05 0.48 12015 CHARLOTTE 704,604 0.25 754,040 0.21 0.93 12017 CITRUS 314,058 0.11 4,577,157 1.25 0.07 12019 CLAY 531,527 0.19 996,165 0.27 0.53 12021 COLLIER 1,612,798 0.56 7,367,182 2.01 0.22 12023 COLUMBIA 765,840 0.27 2,031,142 0.55 0.38 12027 DE SOTO 173,558 0.06 112,450 0.03 1.54 12029 DIXIE 169,395 0.06 521,867 0.14 0.32 12031 DUVAL 23,598,933 8.22 42,804,554 11.67 0.55 12033 ESCAMBIA 6,759,714 2.36 8,222,440 2.24 0.82 12035 FLAGLER 180,979 0.06 432,164 0.12 0.42 12037 FRANKLIN 55,794 0.02 110,658 0.03 0.50 12039 GADSDEN 3,849,211 1.34 1,429,757 0.39 2.69 12041 GILCHRIST 157,807 0.05 66,920 0.02 2.36 12043 GLADES 126,137 0.04 1,678,406 0.46 0.08 12045 GULF 806,073 0.28 1,321,435 0.36 0.61 12047 HAMILTON 3,976,920 1.39 3,970,531 1.08 1.00 12049 HARDEE 4,176,665 1.46 379,378 0.10 11.01 12051 HENDRY 767,901 0.27 428,691 0.12 1.79 12053 HERNANDO 1,041,478 0.36 2,324,267 0.63 0.45 12055 HIGHLANDS 700,822 0.24 754,780 0.21 0.93 12057 HILLSBOROUGH 25,051,909 8.73 68,016,048 18.54 0.37 12059 HOLMES 156,471 0.05 112,227 0.03 1.39 12061 INDIAN RIVER 753,147 0.26 1,084,462 0.30 0.69 12063 JACKSON 305,961 0.11 966,552 0.26 0.32

PAGE 82

66 Table 6.6 (Continued) 12065 JEFFERSON 34,274 0.01 65,022 0.02 0.53 12067 LAFAYETTE 58,639 0.02 22,134 0.01 2.65 12069 LAKE 2,545,532 0.89 2,126,154 0.58 1.20 12071 LEE 3,112, 420 1.08 3,979, 066 1.08 0.78 12073 LEON 2,414,093 0.84 1,227,746 0.33 1.97 12075 LEVY 271,838 0.09 269,211 0.07 1.01 12077 LIBERTY 336,946 0.12 76,293 0.02 4.42 12079 MADISON 632,253 0.22 277,793 0.08 2.28 12081 MANATEE 4,113,694 1.43 4,867,0 50 1.33 0.85 12083 MARION 2,709,297 0.94 3,125,620 0.85 0.87 12085 MARTIN 935,283 0.33 2,403,116 0.66 0.39 12086 MIAMI-DADE 64,225,982 22.38 51,535,773 14.05 1.25 12087 MONROE 375,079 0.13 2,109,437 0.58 0.18 12089 NASSAU 935,839 0.33 2,062,340 0.56 0.45 12091 OKALOOSA 2,454,365 0.86 1,357,923 0.37 1.81 12093 OKEECHOBEE 123,762 0.04 380,188 0.10 0.33 12095 ORANGE 13,900,842 4.84 16,841,014 4.59 0.83 12097 OSCEOLA 264,703 0.09 770,163 0.21 0.34 12099 PALM BEACH 11,039,095 3.85 19,122,300 5.21 0.58 12101 PASCO 745,349 0.26 2,454,151 0.67 0.30 12103 PINELLAS 9,563,127 3.33 8,879,033 2.42 1.08 12105 POLK 42,554,919 14.83 27,758,308 7.57 1.53 12107 PUTNAM 1,213,478 0.42 5,583,740 1.52 0.22 12109 SAINT JOHNS 748,150 0.26 1,712,638 0.47 0.44 12111 SAINT LUCIE 1,768,573 0.62 2,482,766 0.68 0.71 12113 SANTA ROSA 1,465,641 0.51 1,589,212 0.43 0.92 12115 SARASOTA 3,109,034 1.08 3,767,471 1.03 0.83 12117 SEMINOLE 1,673,479 0.58 3,080,150 0.84 0.54 12119 SUMTER 297,170 0.10 566,137 0.15 0.52 12121 SUWANNEE 1,663,375 0.58 597,346 0.16 2.78 12123 TAYLOR 1,444,818 0.50 2,792,681 0.76 0.52 12125 UNION 99,309 0.03 86,566 0.02 1.15 12127 VOLUSIA 5,627,700 1.96 7,153,129 1.95 0.79 12129 WAKULLA 1,056,669 0.37 2,342,520 0.64 0.45 12131 WALTON 511,941 0.18 744,805 0.20 0.69 12133 WASHINGTON 74,631 0.03 148,180 0.04 0.50 Total 287,000,396 100.00 366,841,517 100.00 0.78

PAGE 83

67 Table 6.9 presents the distribution of freight outflows and inflows by truck at the county level. From this table, it can be observed that Miami-Dade has the highest outflows by truck, followed by Duval and Hillsbor ough. Thus Miami-Dade accounts for about 25.57% of freight exports by truck in Florid a, while Duval and Hillsborough account for 9.2% and 7.32% respectively. It can also be noticed that these three coun ties have the highest shares of freight imports by truck in the state of Florida with the figures of 20.74%, 10.35%, and 7.94% respectively. Table 6.10 presents the distribut ion of freight outflows and inflows by rail at the county level. From this table, it can be identified that Polk has the highest outflows by rail, followed by Miami-Dade and Hamilton. Polk accounts for a very high 45.37 % of freight exports by rail in Florida, while MiamiDade and Hamilton account for 16.97 % and 5.47 % respectively. For freight imports by rail, Hi llsborough has the highest share with 27.97 %, followed by Polk at 16.38 % and Duval at 12.72 %. Table 6.11 presents the distribut ion of freight outflows and inflows by air at the county level. It can be easily seen that only a few c ounties use air as a mode of freight transport. Miami-Dade has the highest share of fr eight outflows by air accounting for 51.48 %. Broward and Orange follow Miami-Dade with shares of 15.83 % and 12.49 % respectively. For freight imports by air, Miam i-Dade has the highest share with 48.8 %, followed by Orange at 16.27 % and Hillsboro ugh at 14.72 %. Table 6.12 presents the distribution of freight out flows and inflows by wate r at the c ounty level.

PAGE 84

68 Table 6.12 presents the distribut ion of freight outflows and in flows by water at the county level. It can be easily seen that only a fe w counties use water as a mode of freight transport. Hillsborough has the highest share of freight outflows by water accounting for 44.45 %. Pinellas and Duval follow Hillsboro ugh with shares of 19.38 % and 11.2 % respectively. For freight imports by water, Hillsborough has the highe st share with 39.37 %, followed by Duval at 14.54 % and Collier at 11.16 %. Appendix A contains an extensive descriptive analys is of freight flows in Florida both at the zipcode level and the count y level. The distribution of freight outflows (exports) in annual tons at the zipcode level in Florid a is shown in Figure A.1, while Figure A.2 depicts the distribution of frei ght inflows (imports) in annual tons by zipcode in Florida. The freight outflows to inflow s ratio (export to import rati o) distribution is shown in Figure A.3. Figures A.4 A.15 illustrate the distributions of freight outflows, freight inflows and the ratio of outflow s to inflows for truck, rail, water and air respectively for zipcodes in Florida. Similar distributions at the county level are shown in Figures A.16 – A.30.

PAGE 85

69 Table 6.7 Distribution of Freight Outflows by County and Mode FIPS COUNTY Outflows Truck Rail Air Water 12001 ALACHUA 2,834,631 99.69 0.29 0.02 0.00 12003 BAKER 351,227 100.00 0.00 0.00 0.00 12005 BAY 3,356,830 75.59 12.37 0.00 12.03 12007 BRADFORD 535,762 21.39 78.60 0.00 0.00 12009 BREVARD 5,403,695 91.63 0.21 0.01 8.15 12011 BROWARD 13,587,432 97.37 2.17 0.45 0.01 12013 CALHOUN 91,818 89.94 0.00 0.00 10.06 12015 CHARLOTTE 704,604 33.83 0.00 0.19 65.97 12017 CITRUS 314,058 100.00 0.00 0.00 0.00 12019 CLAY 531,527 66.36 33.64 0.00 0.00 12021 COLLIER 1,612,798 88.31 0.00 0.00 11.69 12023 COLUMBIA 765,840 99.60 0.40 0.00 0.00 12027 DE SOTO 173,558 85.98 14.02 0.00 0.00 12029 DIXIE 169,395 100.00 0.00 0.00 0.00 12031 DUVAL 23,598,933 77.55 14.94 0.07 7.45 12033 ESCAMBIA 6,759,714 85.21 6.11 0.03 8.66 12035 FLAGLER 180,979 100.00 0.00 0.00 0.00 12037 FRANKLIN 55,794 98.47 0.00 0.00 1.53 12039 GADSDEN 3,849,211 14.83 85.17 0.00 0.00 12041 GILCHRIST 157,807 100.00 0.00 0.00 0.00 12043 GLADES 126,137 22.74 63.59 0.00 13.67 12045 GULF 806,073 49.89 50.11 0.00 0.00 12047 HAMILTON 3,976,920 1.15 98.85 0.00 0.00 12049 HARDEE 4,176,665 10.77 89.23 0.00 0.00 12051 HENDRY 767,901 69.77 30.23 0.00 0.00 12053 HERNANDO 1,041,478 21.57 78.43 0.00 0.00 12055 HIGHLANDS 700,822 99.72 0.28 0.00 0.00 12057 HILLSBOROUGH 25,051,909 58.13 13.84 0.17 27.86 12059 HOLMES 156,471 100.00 0.00 0.00 0.00 12061 INDIAN RIVER 753,147 100.00 0.00 0.00 0.00 12063 JACKSON 305,961 100.00 0.00 0.00 0.00

PAGE 86

70 Table 6.7 (Continued) 12065 JEFFERSON 34,274 100.00 0.00 0.00 0.00 12067 LAFAYETTE 58, 639 100.00 0.00 0.00 0.00 12069 LAKE 2,545,532 98.49 1.51 0.00 0.00 12071 LEE 3, 112,420 99.77 0.08 0.15 0.00 12073 LEON 2,414,093 99.91 0.00 0.09 0.00 12075 LEVY 271,838 100.00 0.00 0.00 0.00 12077 LIBERTY 336,946 3.80 96.20 0.00 0.00 12079 MADISON 632,253 83.77 16.23 0.00 0.00 12081 MANATEE 4,113,694 76.30 23.64 0.00 0.05 12083 MARION 2,709,297 90.78 9.22 0.00 0.00 12085 MARTIN 935,283 82.30 17.70 0.00 0.00 12086 MIAMI-DADE 64,225,982 79.22 19.01 0.31 1.46 12087 MONROE 375,079 100.00 0.00 0.00 0.00 12089 NASSAU 935,839 56.75 43.25 0.00 0.00 12091 OKALOOSA 2,454,365 99.99 0.00 0.01 0.00 12093 OKEECHOBEE 123,762 100.00 0.00 0.00 0.00 12095 ORANGE 13,900,842 97.10 2.55 0.35 0.00 12097 OSCEOLA 264,703 100.00 0.00 0.00 0.00 12099 PALM BEACH 11,039,095 93.86 4.89 0.07 1.17 12101 PASCO 745,349 98.99 1.00 0.00 0.01 12103 PINELLAS 9,563,127 64.38 3.80 0.00 31.81 12105 POLK 42,554,919 23.32 76.68 0.00 0.00 12107 PUTNAM 1,213,478 79.02 20.98 0.00 0.00 12109 SAINT JOHNS 748,150 90.27 4.71 0.00 5.02 12111 SAINT LUCIE 1,768,573 98.62 1.36 0.02 0.00 12113 SANTA ROSA 1,465,641 70.34 29.66 0.00 0.00 12115 SARASOTA 3,109,034 98.76 1.17 0.00 0.08 12117 SEMINOLE 1,673,479 97.21 2.79 0.00 0.00 12119 SUMTER 297,170 61.63 38.37 0.00 0.00 12121 SUWANNEE 1,663,375 100.00 0.00 0.00 0.00 12123 TAYLOR 1,444,818 60.52 39.48 0.00 0.00 12125 UNION 99,309 100.00 0.00 0.00 0.00 12127 VOLUSIA 5,627,700 85.87 14.12 0.01 0.00 12129 WAKULLA 1,056,669 34.74 0.00 0.00 65.26 12131 WALTON 511,941 97.57 0.89 0.00 1.54 12133 WASHINGTON 74,631 100.00 0.00 0.00 0.00 Total 287,000,396 69.34 25.06 0.14 5.47

PAGE 87

71 Table 6.8 Distribution of Freight Inflows by County and Mode FIPS COUNTY Inflows Truck Rail Air Water 12001 ALACHUA 2,324,574 72.23 27.74 0.03 0.00 12003 BAKER 192,723 72.24 27.76 0.00 0.00 12005 BAY 4,441,292 40.40 19.48 0.00 40.12 12007 BRADFORD 256,580 78.16 21.84 0.00 0.00 12009 BREVARD 7,542,605 57.82 38.60 0.01 3.58 12011 BROWARD 17,073,597 84.08 13.58 0.42 1.92 12013 CALHOUN 189,697 100.00 0.00 0.00 0.00 12015 CHARLOTTE 754,040 80.80 0.00 0.03 19.17 12017 CITRUS 4,577,157 20.55 79.45 0.00 0.00 12019 CLAY 996,165 93.56 6.44 0.00 0.00 12021 COLLIER 7,367,182 12.19 0.00 0.00 87.81 12023 COLUMBIA 2,031,142 58.64 41.36 0.00 0.00 12027 DE SOTO 112,450 100.00 0.00 0.00 0.00 12029 DIXIE 521,867 100.00 0.00 0.00 0.00 12031 DUVAL 42,804,554 49.61 30.60 0.09 19.70 12033 ESCAMBIA 8,222,440 52.93 16.61 0.05 30.41 12035 FLAGLER 432,164 62.31 37.59 0.00 0.10 12037 FRANKLIN 110,658 100.00 0.00 0.00 0.00 12039 GADSDEN 1,429,757 89.90 2.91 0.00 7.19 12041 GILCHRIST 66,920 100.00 0.00 0.00 0.00 12043 GLADES 1,678,406 1.11 0.00 0.00 98.89 12045 GULF 1,321,435 62.00 38.00 0.00 0.00 12047 HAMILTON 3,970,531 25.28 73.93 0.00 0.79 12049 HARDEE 379,378 94.87 5.13 0.00 0.00 12051 HENDRY 428,691 89.76 10.24 0.00 0.00 12053 HERNANDO 2,324,267 72.87 26.92 0.00 0.21 12055 HIGHLANDS 754,780 90.69 9.31 0.00 0.00 12057 HILLSBOROUGH 68,016,048 23.96 42.33 0.16 33.56 12059 HOLMES 112,227 100.00 0.00 0.00 0.00 12061 INDIAN RIVER 1,084,462 77.85 13.15 0.00 9.00 12063 JACKSON 966,552 75.03 24.97 0.00 0.00

PAGE 88

72 Table 6.8 (Continued) 12065 JEFFERSON 65,022 100.00 0.00 0.00 0.00 12067 LAFAYETTE 22,134 100. 00 0.00 0.00 0.00 12069 LAKE 2,126,154 100.00 0.00 0.00 0.00 12071 LEE 3,979,066 99 .68 0.00 0.17 0.15 12073 LEON 1,227,746 97.34 2.44 0.21 0.00 12075 LEVY 269,211 100.00 0.00 0.00 0.00 12077 LIBERTY 76,293 67.73 32.27 0.00 0.00 12079 MADISON 277,793 99.99 0.01 0.00 0.00 12081 MANATEE 4,867,050 81.24 18.74 0.00 0.02 12083 MARION 3,125,620 96.73 3.27 0.00 0.00 12085 MARTIN 2,403,116 42.67 57.29 0.00 0.04 12086 MIAMI-DADE 51,535,773 82.59 9.45 0.69 7.26 12087 MONROE 2,109,437 18.69 0.00 0.00 81.31 12089 NASSAU 2,062,340 28.84 60.94 0.00 10.22 12091 OKALOOSA 1,357,923 96.71 3.16 0.13 0.00 12093 OKEECHOBEE 380,188 47.23 52.77 0.00 0.00 12095 ORANGE 16,841,014 65.42 33.82 0.71 0.06 12097 OSCEOLA 770,163 95.38 4.62 0.00 0.00 12099 PALM BEACH 19,122,300 65.75 9.90 0.10 24.25 12101 PASCO 2,454,151 91.31 8.46 0.00 0.23 12103 PINELLAS 8,879,033 96.50 3.35 0.00 0.15 12105 POLK 27,758,308 39.21 60.75 0.00 0.04 12107 PUTNAM 5,583,740 24.84 72.60 0.00 2.56 12109 SAINT JOHNS 1,712,638 42.86 8.65 0.00 48.49 12111 SAINT LUCIE 2,482,766 54.32 45.51 0.00 0.16 12113 SANTA ROSA 1,589,212 71.77 28.23 0.00 0.00 12115 SARASOTA 3,767,471 99.97 0.00 0.03 0.00 12117 SEMINOLE 3,080,150 94.42 5.49 0.00 0.09 12119 SUMTER 566,137 54.58 45.42 0.00 0.00 12121 SUWANNEE 597,346 70.02 29.98 0.00 0.00 12123 TAYLOR 2,792,681 88.20 11.80 0.00 0.00 12125 UNION 86,566 100.00 0.00 0.00 0.00 12127 VOLUSIA 7,153,129 62.86 37.03 0.01 0.10 12129 WAKULLA 2,342,520 17.27 0.00 0.00 82.73 12131 WALTON 744,805 50.05 43.83 0.00 6.12 12133 WASHINGTON 148,180 91.43 8.57 0.00 0.00 Total 366,841,517 55.94 28.06 0.20 15.80

PAGE 89

73 Table 6.9 Distribution of Truck Outfl ows and Inflows by County FIPS COUNTY Truck Outflows Percentage Truck Inflows Percentage 12001 ALACHUA 2,825,802 1.42 1,679,092 0.82 12003 BAKER 351,227 0.18 139,231 0.07 12005 BAY 2,537,548 1.28 1,794,411 0.87 12007 BRADFORD 114,625 0.06 200,554 0.10 12009 BREVARD 4,951,310 2.49 4,360,825 2.13 12011 BROWARD 13,230,271 6.65 14,355,032 7.00 12013 CALHOUN 82,584 0.04 189,697 0.09 12015 CHARLOTTE 238,396 0.12 609,257 0.30 12017 CITRUS 314,058 0.16 940,537 0.46 12019 CLAY 352,741 0.18 932,052 0.45 12021 COLLIER 1,424,256 0.72 897,863 0.44 12023 COLUMBIA 762,751 0.38 1,191,012 0.58 12027 DE SOTO 149,233 0.07 112,450 0.05 12029 DIXIE 169,395 0.09 521,867 0.25 12031 DUVAL 18,300,454 9.20 21,236,208 10.35 12033 ESCAMBIA 5,760,023 2.89 4,352,153 2.12 12035 FLAGLER 180,979 0.09 269,278 0.13 12037 FRANKLIN 54,939 0.03 110,658 0.05 12039 GADSDEN 570,670 0.29 1,285,312 0.63 12041 GILCHRIST 157,807 0.08 66,920 0.03 12043 GLADES 28,687 0.01 18,687 0.01 12045 GULF 402,162 0.20 819,310 0.40 12047 HAMILTON 45,881 0.02 1,003,600 0.49 12049 HARDEE 449,809 0.23 359,933 0.18 12051 HENDRY 535,775 0.27 384,807 0.19 12053 HERNANDO 224,642 0.11 1,693,760 0.83 12055 HIGHLANDS 698,874 0.35 684,493 0.33 12057 HILLSBOROUGH 14,561,736 7.32 16,295,204 7.94 12059 HOLMES 156,471 0.08 112,227 0.05 12061 INDIAN RIVER 753,147 0.38 844,301 0.41 12063 JACKSON 305,961 0.15 725,197 0.35

PAGE 90

74 Table 6.9 (Continued) 12065 JEFFERSON 34,274 0.02 65,022 0.03 12067 LAFAYETTE 58, 639 0.03 22,134 0.01 12069 LAKE 2,507,188 1.26 2,126,154 1.04 12071 LEE 3,105,372 1.56 3,966, 165 1.93 12073 LEON 2,412,029 1.21 1,195,123 0.58 12075 LEVY 271,838 0.14 269,211 0.13 12077 LIBERTY 12,816 0.01 51,677 0.03 12079 MADISON 529,646 0.27 277,760 0.14 12081 MANATEE 3,138,8 60 1.58 3,95 3,983 1.93 12083 MARION 2,459,495 1.24 3,023,269 1.47 12085 MARTIN 769,764 0.39 1,025,412 0.50 12086 MIAMI-DADE 50,878,648 25.57 42,564,815 20.74 12087 MONROE 375,079 0.19 394,247 0.19 12089 NASSAU 531,072 0.27 594,776 0.29 12091 OKALOOSA 2,454,106 1.23 1,313,273 0.64 12093 OKEECHOBEE 123,762 0.06 179,549 0.09 12095 ORANGE 13,497,035 6.78 11,017,404 5.37 12097 OSCEOLA 264,703 0.13 734,553 0.36 12099 PALM BEACH 10,361,818 5.21 12,572,790 6.13 12101 PASCO 737,814 0.37 2,240,961 1.09 12103 PINELLAS 6,156,937 3.09 8,568,408 4.18 12105 POLK 9,925,488 4.99 10,884,632 5.30 12107 PUTNAM 958,859 0.48 1,387,111 0.68 12109 SAINT JOHNS 675,341 0.34 734,102 0.36 12111 SAINT LUCIE 1,744,147 0.88 1,348,677 0.66 12113 SANTA ROSA 1,030,924 0.52 1,140,550 0.56 12115 SARASOTA 3,070,364 1.54 3,766,440 1.84 12117 SEMINOLE 1,626,777 0.82 2,908,305 1.42 12119 SUMTER 183,144 0.09 309,011 0.15 12121 SUWANNEE 1,663,375 0.84 418,270 0.20 12123 TAYLOR 874,370 0.44 2,463,105 1.20 12125 UNION 99,309 0.05 86,566 0.04 12127 VOLUSIA 4,832,479 2.43 4,496,695 2.19 12129 WAKULLA 367,052 0.18 404,591 0.20 12131 WALTON 499,521 0.25 372,797 0.18 12133 WASHINGTON 74,631 0.04 135,476 0.07 Total 198,998,890 100.00 205,198,942 100.00

PAGE 91

75 Table 6.10 Distribution of Rail Outflows and Inflows by County FIPS COUNTY Rail Outflows Percentage Rail Inflows Percentage 12001 ALACHUA 8,329 0.01 644,890 0.63 12003 BAKER 0 0.00 53,492 0.05 12005 BAY 415,300 0.58 865,046 0.84 12007 BRADFORD 421,135 0.59 56,026 0.05 12009 BREVARD 11,506 0.02 2,911,315 2.83 12011 BROWARD 294,628 0.41 2,319,017 2.25 12013 CALHOUN 0 0.00 0 0.00 12015 CHARLOTTE 0 0.00 0 0.00 12017 CITRUS 0 0.00 3,636,619 3.53 12019 CLAY 178,783 0.25 64,112 0.06 12021 COLLIER 0 0.00 0 0.00 12023 COLUMBIA 3,089 0.00 840,128 0.82 12027 DE SOTO 24,325 0.03 0 0.00 12029 DIXIE 0 0.00 0 0.00 12031 DUVAL 3,524,830 4.90 13,099,361 12.72 12033 ESCAMBIA 412,694 0.57 1,365,736 1.33 12035 FLAGLER 0 0.00 162,438 0.16 12037 FRANKLIN 0 0.00 0 0.00 12039 GADSDEN 3,278,541 4.56 41,616 0.04 12041 GILCHRIST 0 0.00 0 0.00 12043 GLADES 80,212 0.11 0 0.00 12045 GULF 403,908 0.56 502,126 0.49 12047 HAMILTON 3,931,039 5.47 2,935,368 2.85 12049 HARDEE 3,726,854 5.18 19,446 0.02 12051 HENDRY 232,127 0.32 43,885 0.04 12053 HERNANDO 816,837 1.14 625,641 0.61 12055 HIGHLANDS 1,948 0.00 70,287 0.07 12057 HILLSBOROUGH 3,467,516 4.82 28,790,424 27.97 12059 HOLMES 0 0.00 0 0.00 12061 INDIAN RIVER 0 0.00 142,593 0.14 12063 JACKSON 0 0.00 241,356 0.23

PAGE 92

76 Table 6.10 (Continued) 12065 JEFFERSON 0 0.00 0 0.00 12067 LAFAYETTE 0 0.00 0 0.00 12069 LAKE 38,342 0.05 0 0.00 12071 LEE 2,47 3 0.00 0 0.00 12073 LEON 0 0.00 29,989 0.03 12075 LEVY 0 0.00 0 0.00 12077 LIBERTY 324,130 0.45 24,616 0.02 12079 MADISON 102,606 0.14 33 0.00 12081 MANATEE 972,6 48 1.35 912, 069 0.89 12083 MARION 249,789 0.35 102,350 0.10 12085 MARTIN 165,519 0.23 1,376,753 1.34 12086 MIAMI-DADE 12,206,759 16.97 4,872,463 4.73 12087 MONROE 0 0.00 0 0.00 12089 NASSAU 404,769 0.56 1,256,861 1.22 12091 OKALOOSA 0 0.00 42,863 0.04 12093 OKEECHOBEE 0 0.00 200,640 0.19 12095 ORANGE 355,129 0.49 5,695,140 5.53 12097 OSCEOLA 0 0.00 35,613 0.03 12099 PALM BEACH 539,415 0.75 1,892,704 1.84 12101 PASCO 7,485 0.01 207,532 0.20 12103 PINELLAS 363,692 0.51 297,027 0.29 12105 POLK 32,629,452 45.37 16,861,858 16.38 12107 PUTNAM 254,609 0.35 4,053,628 3.94 12109 SAINT JOHNS 35,232 0.05 148,115 0.14 12111 SAINT LUCIE 24,081 0.03 1,129,999 1.10 12113 SANTA ROSA 434,723 0.60 448,662 0.44 12115 SARASOTA 36,332 0.05 0 0.00 12117 SEMINOLE 46,698 0.06 169,137 0.16 12119 SUMTER 114,026 0.16 257,127 0.25 12121 SUWANNEE 0 0.00 179,076 0.17 12123 TAYLOR 570,448 0.79 329,576 0.32 12125 UNION 0 0.00 0 0.00 12127 VOLUSIA 794,628 1.11 2,648,693 2.57 12129 WAKULLA 0 0.00 0 0.00 12131 WALTON 4,550 0.01 326,433 0.32 12133 WASHINGTON 0 0.00 12,704 0.01 Total 71,911,136 100.00 102,942,583 100.00

PAGE 93

77 Table 6.11 Distribution of Air Outflows and Inflows by County FIPS COUNTY Air Outflows Percentage Air Inflows Percentage 12001 ALACHUA 502 0.13 595 0.08 12003 BAKER 0 0.00 0 0.00 12005 BAY 0 0.00 0 0.00 12007 BRADFORD 0 0.00 0 0.00 12009 BREVARD 308 0.08 548 0.08 12011 BROWARD 61,669 15.83 72,242 9.90 12013 CALHOUN 0 0.00 0 0.00 12015 CHARLOTTE 1,369 0.35 255 0.03 12017 CITRUS 0 0.00 0 0.00 12019 CLAY 0 0.00 0 0.00 12021 COLLIER 1 0.00 0 0.00 12023 COLUMBIA 0 0.00 0 0.00 12027 DE SOTO 0 0.00 0 0.00 12029 DIXIE 0 0.00 0 0.00 12031 DUVAL 15,655 4.02 37,212 5.10 12033 ESCAMBIA 1,822 0.47 3,874 0.53 12035 FLAGLER 0 0.00 0 0.00 12037 FRANKLIN 0 0.00 0 0.00 12039 GADSDEN 0 0.00 0 0.00 12041 GILCHRIST 0 0.00 0 0.00 12043 GLADES 0 0.00 0 0.00 12045 GULF 0 0.00 0 0.00 12047 HAMILTON 0 0.00 0 0.00 12049 HARDEE 0 0.00 0 0.00 12051 HENDRY 0 0.00 0 0.00 12053 HERNANDO 0 0.00 0 0.00 12055 HIGHLANDS 0 0.00 0 0.00 12057 HILLSBOROUGH 42,964 11.03 107,370 14.72 12059 HOLMES 0 0.00 0 0.00 12061 INDIAN RIVER 0 0.00 0 0.00 12063 JACKSON 0 0.00 0 0.00

PAGE 94

78 Table 6.11 (Continued) 12065 JEFFERSON 0 0.00 0 0.00 12067 LAFAYETTE 0 0.00 0 0.00 12069 LAKE 0 0.00 0 0.00 12071 LEE 4,574 1.17 6,85 5 0.94 12073 LEON 2,064 0.53 2,633 0.36 12075 LEVY 0 0.00 0 0.00 12077 LIBERTY 0 0.00 0 0.00 12079 MADISON 0 0.00 0 0.00 12081 MANATEE 0 0.00 0 0.00 12083 MARION 22 0.01 0 0.00 12085 MARTIN 0 0.00 0 0.00 12086 MIAMI-DADE 200,522 51.48 356,013 48.80 12087 MONROE 0 0.00 0 0.00 12089 NASSAU 0 0.00 0 0.00 12091 OKALOOSA 258 0.07 1,786 0.24 12093 OKEECHOBEE 0 0.00 0 0.00 12095 ORANGE 48,664 12.49 118,739 16.27 12097 OSCEOLA 0 0.00 0 0.00 12099 PALM BEACH 8,161 2.10 19,631 2.69 12101 PASCO 0 0.00 0 0.00 12103 PINELLAS 0 0.00 0 0.00 12105 POLK 0 0.00 0 0.00 12107 PUTNAM 0 0.00 0 0.00 12109 SAINT JOHNS 0 0.00 0 0.00 12111 SAINT LUCIE 346 0.09 0 0.00 12113 SANTA ROSA 0 0.00 0 0.00 12115 SARASOTA 0 0.00 1,030 0.14 12117 SEMINOLE 0 0.00 0 0.00 12119 SUMTER 0 0.00 0 0.00 12121 SUWANNEE 0 0.00 0 0.00 12123 TAYLOR 0 0.00 0 0.00 12125 UNION 0 0.00 0 0.00 12127 VOLUSIA 601 0.15 806 0.11 12129 WAKULLA 0 0.00 0 0.00 12131 WALTON 0 0.00 0 0.00 12133 WASHINGTON 0 0.00 0 0.00 Total 389,502 100.00 729,589 100.00

PAGE 95

79 Table 6.12 Distribution of Water Outflow s and Inflows by County FIPS COUNTY Water Outflows Percentage Water Inflows Percentage 12001 ALACHUA 0 0.00 0 0.00 12003 BAKER 0 0.00 0 0.00 12005 BAY 403,984 2.57 1,781,836 3.07 12007 BRADFORD 0 0.00 0 0.00 12009 BREVARD 440,571 2.81 269,924 0.47 12011 BROWARD 863 0.01 327,323 0.56 12013 CALHOUN 9,234 0.06 0 0.00 12015 CHARLOTTE 464,839 2.96 144,529 0.25 12017 CITRUS 0 0.00 0 0.00 12019 CLAY 0 0.00 0 0.00 12021 COLLIER 188,541 1.20 6,469,318 11.16 12023 COLUMBIA 0 0.00 0 0.00 12027 DE SOTO 0 0.00 0 0.00 12029 DIXIE 0 0.00 0 0.00 12031 DUVAL 1,757,972 11.20 8,431,755 14.54 12033 ESCAMBIA 585,178 3.73 2,500,678 4.31 12035 FLAGLER 0 0.00 448 0.00 12037 FRANKLIN 855 0.01 0 0.00 12039 GADSDEN 0 0.00 102,833 0.18 12041 GILCHRIST 0 0.00 0 0.00 12043 GLADES 17,237 0.11 1,659,719 2.86 12045 GULF 0 0.00 0 0.00 12047 HAMILTON 0 0.00 31,561 0.05 12049 HARDEE 0 0.00 0 0.00 12051 HENDRY 0 0.00 0 0.00 12053 HERNANDO 0 0.00 4,866 0.01 12055 HIGHLANDS 0 0.00 0 0.00 12057 HILLSBOROUGH 6,979,686 44.45 22,823,023 39.37 12059 HOLMES 0 0.00 0 0.00 12061 INDIAN RIVER 0 0.00 97,567 0.17 12063 JACKSON 0 0.00 0 0.00

PAGE 96

80 Table 6.12 (Continued) 12065 JEFFERSON 0 0.00 0 0.00 12067 LAFAYETTE 0 0.00 0 0.00 12069 LAKE 0 0.00 0 0.00 12071 LEE 0 0. 00 6,046 0.01 12073 LEON 0 0.00 0 0.00 12075 LEVY 0 0.00 0 0.00 12077 LIBERTY 0 0.00 0 0.00 12079 MADISON 0 0.00 0 0.00 12081 MANATEE 2,18 7 0.01 1, 002 0.00 12083 MARION 0 0.00 0 0.00 12085 MARTIN 0 0.00 952 0.00 12086 MIAMI-DADE 940,047 5.99 3,742,486 6.46 12087 MONROE 0 0.00 1,715,190 2.96 12089 NASSAU 0 0.00 210,711 0.36 12091 OKALOOSA 0 0.00 0 0.00 12093 OKEECHOBEE 0 0.00 0 0.00 12095 ORANGE 12 0.00 9,720 0.02 12097 OSCEOLA 0 0.00 0 0.00 12099 PALM BEACH 129,707 0.83 4,637,171 8.00 12101 PASCO 50 0.00 5,659 0.01 12103 PINELLAS 3,042,498 19.38 13,599 0.02 12105 POLK 0 0.00 11,850 0.02 12107 PUTNAM 0 0.00 143,002 0.25 12109 SAINT JOHNS 37,569 0.24 830,420 1.43 12111 SAINT LUCIE 0 0.00 4,088 0.01 12113 SANTA ROSA 0 0.00 0 0.00 12115 SARASOTA 2,339 0.01 0 0.00 12117 SEMINOLE 0 0.00 2,711 0.00 12119 SUMTER 0 0.00 0 0.00 12121 SUWANNEE 0 0.00 0 0.00 12123 TAYLOR 0 0.00 0 0.00 12125 UNION 0 0.00 0 0.00 12127 VOLUSIA 0 0.00 6,931 0.01 12129 WAKULLA 689,617 4.39 1,937,929 3.34 12131 WALTON 7,871 0.05 45,574 0.08 12133 WASHINGTON 0 0.00 0 0.00 Total 15,700,857 100.00 57,970,421 100.00

PAGE 97

81 6.5 Distribution of Freight Flows by Regi on and Socio-Economic Characteristics The population distribution in Fl orida at the zipcode level is shown in the Appendix A in Figure A.31. Figure A.32 depicts freight export in annual tons per person at the zipcode level in Florida, while Figure A.33 depicts fr eight import in annual to ns per person at the zipcode level in Florida. Similar figures at the county level are al so included in the Appendix A. The employer locations, as given in the In foUSA database, are also shown in Appendix A. Figure A.39 presents the total employ ment by zipcode, while Figures A.40 – A.49 present the employment by various industry types at the zipcode level. Figure A.50 presents the freight exports in annual tons per employee at the zipcode level in Florida, while Figure A.51 presents the freight imports in annual tons per employee at the zipcode level in Florida. 6.6 Conclusions As it is determined to model each commodity group separately, databases for each of the 17 commodity groups were created. Each of th ese 17 databases used in this study consist of 859,329 records. Each record consists of flows by various modes for each origindestination zipcode pair in Florida along with the socio-ec onomic characteristics of the origin & destination and di stance between the origin and destination zipcodes. The freight flow variables in each of th ese databases had high number of zeros (around 97%) as many zipcode pairs do not exchange freight flows. Thus, these freight flow

PAGE 98

82 variables in the databases unduly skew the variable distribution. However, the socioeconomic variables in the databa ses were normally distributed. In conclusion, the exploratory analysis condu cted on the database suggests that the database offers variables with plausible sta tistical distributions and summaries consistent with expectations. Although th e TRANSEARCH freight database has its share of errors and omissions, many states are i nvesting in the purchase of th is data to develop statewide freight travel demand models. In this context, it was felt th at it is not inappropriate to develop models of freight flow using the TRANSEARCH databases, as the objective of this paper is to develop practical models of fr eight flow that utilize data available at many state and local agencies. However, readers should note the potential limitations of the database used in this study and interpret model results presented in the next section with appropriate caution.

PAGE 99

83 Chapter 7 Modeling Methodology 7.1 Introduction The modeling of multimodal freight moveme nts involves dealing with multiple endogenous variables in a simultaneous e quations framework. Commodity flows on different modes are travel demand related endog enous variables that ar e inter-related with one another. When modeling the interactions among severa l inter-dependent endogenous variables, simultaneous equations systems offer an appropriate framework for model development and hypothesis testing (Bollen, 1989) In this study, the structural equations modeling methodology is adopted for estimati ng simultaneous equations systems that capture the inter-dependencies amon g multimodal freight movements. Thus, the modeling methodology adopted in th is paper is centered on the structural equations modeling framework that can be us ed to determine and model relationships among several dependent (endogenous) vari ables simultaneously. As the model framework described in Chapter 3 includes a number of endogenous variables (freight movements by mode), it was considered appr opriate to adopt this modeling methodology. In a structural analysis approach, also known as causal analysis, path analysis, or simply simultaneous equations, the phenomenon under study is cause-and-effect relationships. The relationships are either unidirectional, that is, they each postulate that one variable

PAGE 100

84 influences another, or reciproc al where relationships are spec ified in both directions. In this way, many structural equation models incorporate both direct and feedback influences. This chapter attempts a review of the best current practice in specifying and estimating such sophisticated models. However, as described in the previous chap ter, it is found that there are many origindestination pairs that do not exchange freight flows of a certain commodity at all. Thus, there is a high number of zero flows in th e database. As the presence of zeros unduly skews the dependent variable distribution (a spik e at zero in the frei ght flow distribution), this study employs a structural equations estimation methodology that accommodates skewed non-normal endogenous variables. 7.2 Structural Equations Modeling A typical structural equations model (with ‘G’ number of endogenous variables) is defined by a matrix equation syst em as shown in Equation 1. Y Y YX BGG 11. . . (7.1) This can be rewritten as YBYX (7.2) (or) YIBX()()1 (7.3) where Y is a column vector of endogenous variables,

PAGE 101

85 B is a matrix of parameters associated with right-hand-side endogenous variables, X is a column vector of exogenous variables, is a matrix of parameters associ ated with exogenous variables, and is a column vector of error terms a ssociated with the endogenous variables. Structural equations systems are estimated by covariance-based structural analysis, also called method of moments. In this me thodology, instead of minimizing the sum of squared differences between observed and pr edicted individual va lues, the difference between the sample covariances and the covariances predicted by the model is minimized. The observed covariances minus the predicted covariances form the residuals. The fundamental hypothesis for th e covariances-based es timation procedures is that the covariance matrix of the observed variables is a f unction of a set of parameters as shown in Equation 4: = () (7.4) where is the population covariance matrix of observed variables, is a vector that contains the model parameters, and () is the covariance matrix written as a function of Equation 4 implies that each element of the covariance matrix is a function of one or more model parameters. The relation of to () is basic to an understanding of identification, estimation, and assessmen ts of model fit. The matrix () has three

PAGE 102

86 components, namely, the covariance matrix of Y the covariance matrix of X with Y and the covariance matrix of X Let = covariance matrix of X and = covariance matrix of Then it can be shown that [29]: () ()()()() () IBIBIB IB111 1 (7.5) Before estimating model parameters, it is firs t necessary to ensure that the model is identified. Model identification in simultaneous structural e quations systems is concerned with the ability to obtain unique estimates of the structural parameters. The identification problem is typically resolved by using theoretical knowledge of the phenomenon under investigation to place re strictions on model parameters. The restrictions usually employed are zero restrictions where se lected endogenous variables and certain exogenous variables do not appear on the right hand side of certain equations and selected error correlations are specified to be zero. There are several rules that can be used to check whether a structural equation model system is identified. Detailed discussions on these identification rules may be f ound in Bollen (1989), Judge et al (1985) and Johnston et al (1997). The unknown parameters in B, , and are estimated so that the implied covariance matrix, is as close as possible to the sample covariance matrix, S. In order to achieve this, a fitting function F(S, ()) which is to be minimized is defined. The fitting

PAGE 103

87 function has the properties of being scalar, grea ter than or equal to zero if and only if () = S, and continuous in S and (). 7.3 ADF-WLS Estimation Available methods for parameter estim ation include maximum likelihood (ML), unweighted least squares (ULS), generalized least squares (G LS), scale free least squares (SLS), and asymptotically distribution-free we ighted least squares (ADF-WLS). Each of these methods minimizes the fitting function and leads to consistent estimators of The ADF-WLS method of estimation was used to es timate parameters of structural equations models as the univariate distri butions of the endo genous variables are non-normal in that there are substantial numbers of observations for each variable with zero value, which denotes no commodity flow between a zip c ode pair. For such distributions, the ML coefficient estimates will be consistent, but th e estimates of parameter standard errors and the overall model 2 goodness-of-fit will likely be bias ed (Golob et. al., 1997). Unbiased estimates of standard errors and goodness-of -fit can be generated using the ADF-WLS method (Golob et. al., 1997). The ADF-WLS estimation method proceeds in thr ee distinct steps. First, it is assumed that each observed endogenous variable is generated by an unobserved normally distributed latent variable. If the latent variable is greater than a censoring level, it is observed; otherwise the censoring level is observed. Each latent variable is assumed to be conditional on the other variables in the sy stem. The problem is to determine the conditional unknown mean and variance of each censored latent variable. This can be

PAGE 104

88 done using the Tobit model. An appropria te maximum likelihood estimation procedure for the Tobit model is described in Maddala (1983). Second, estimates of the correlations between the latent censored endogenous variab les, and the correlations between each of the latent variables and the continuous e xogenous variables in the system are derived. Finally, parameters of the structural equati on model are estimated such that the modelimplied correlation matrix is as close as possible to the sample correlation matrix, where the sample correlation matrix is determined in the previous steps. The fitting function is then: FWLS = [s ( )]’ W-1[s – ( )] (7.6) where s is a vector of censor ed correlation coefficients for all pairs of endogenous and exogenous variables, ( ) is a vector of model-implied co rrelations for the same variable pairs, and W is a positive-definite weight matrix. Minimizing FWLS implies that the parameter estimates are those that minimize th e weighted sum of squa red deviations of s from ( ). This is analogous to weighted least squares re gression, but here the observed and predicted values are variances and covari ances rather than raw observations. The best choice of the weight matrix is a consistent estimator of the asymptotic covariance matrix of s: W = ACOV(sij, sgh) (7.7) Under very general conditions: ) ( 1gh ij ijghs s s N W (7.8) is a consistent estimator, where sijgh denotes the fourth-order moments of the variables around their means, and sij and sgh denote covariances. Browne (1984) demonstrated that

PAGE 105

89 FWLS with such a weight matrix will yield consistent estimates, which are asymptotically efficient with correct parameter test statisti cs. These properties ho ld for very general conditions, and consequently such estimators are known as arbitrary distribution function, or asymptotically distribution free (ADF) es timators. ADF-WLS estimators are available in several structural equation model estima tion packages including AMOS (Arbuckle, 2000) and LISREL (Joreskog et. al ., 1993). AMOS was used to estimate the structural equations models for each of the 17 comm odity groups in this study. The estimation results are presented in the next chapter.

PAGE 106

90 Chapter 8 Model Estimation Results 8.1 Introduction This chapter describes the model specification and estimation results for each of the 17 structural equations models developed. The models employed a host of exogenous (explanatory) and endogenous (dependent) vari ables to model freigh t flows by zip code pair. Exogenous variables may be divided into three groups: population demographic characteristics of the origin and destination, employment characteristics of the origin and destination, and the impedance (distance) between the origin and destination. All population demographic variables were derived from the 2000 Census and all employment characteristics were derived from the InfoUSA 2000 database. Exogenous variables to be included in the models were selected based on earlie r research (Sivakumar et. al., 2002; Brogan et. al., 2001; Cambridge Systematics, 1996; Jack Faucett Associates, 1999b; Garrido, undated). Endogenous variable s are commodity flows between origindestination zip codes by various modes. The commodity flow on each mode is a different endogenous variable. As mentioned earlier, th e distributions of the endogenous variables are highly skewed and non-normal with a la rge number of zero observations. Nearly 97 percent of the observations are zero observations in the data set. Even within the context of the ADF-WLS estimation method, such a h eavily zero-inflated distribution leads to computational intractability. To help with computational tr actability, log transformations

PAGE 107

91 of the variables are used in the estimation pr ocess. For all observations and variables, unity was added to the raw variable value to avoid having to take the logarithm of zero which is undefined. Thus all zero observati ons appear as zeros in the log-transformed data set as well because the logarithm of unity is zero. 8.2 Model Estimation Results In this study, models were estimated for all commodity groups except the coal commodity group. This commodity group had only two zip-code pairs that exchange coal flow between them. Thus, the demand for coal was not estimated. Models were estimated for the remaining 16 commodity groups. For these 16 commodity groups, various mode l structures with different variable combinations were considered to test for a st ructure that performs the best. It was found that aggregate population and employment of the origin and destin ation zipcodes were enough to be used in the model structures. All the other socio-economic variables listed in Chapter 4 were insignificant in descri bing the modal flows between an origindestination pair. Overall employment was found to be significant as opposed to employment for the industry which produc es the commodity, because many different types of industrial, service and household se ctors may each consume some amount of that commodity. Thus, the final models developed for each of the 16 commodity groups consist of 5 exogenous variables: Destin ation Employment, Destinat ion Population, Origin

PAGE 108

92 Employment, Origin Populati on, and Distance. The endogenous variables were the Total Freight Flow between the O-D pair, and the Freight Flows by various modes between the O-D pair. Table 8.1 presents the structural equation model estimation results for all commodities combined. Tables B.1 – B.16 in Appendix B present the structural equation model estimation results for all th e 16 commodity groups. The path diagram showing the relationships depicted in Table 8.1 is shown in Figure 8.1, wh ile those depicted in Tables B.1 – B.16 (Appendix B) are shown in Figur es C.1 – C.16 (Appendix C). The models provided excellent goodness-of -fit measures with the 2 statistic indicating that the model can not be rejected with a high degree of c onfidence (95 percent or higher) and with the goodness-of-fit index (GFI) equal to unity. T hus the models are clearly capable of capturing the key relationships influencing frei ght flows, even within the context of a large database (more than 859,329 record s) where endogenous variables are highly skewed, zero-inflated, and non-normal. The indications provided by all the models ar e quite consistent w ith expectations and plausible. The tables show the direct effects and total effects that constitute relationships among variables. A direct effect is one where a variable directly a ffects another variable as depicted by a direct arrow linking the two variables in the path diagram. On the other hand, an indirect effect is one where a variable influences another variab le through a mediating variable. For example, in Figure 8. 1, one can see that origin employment does not directly affect the total freight moveme nt by rail. However, origin employment

PAGE 109

93 affects both total flow and total truck flow. In turn, total flow and total truck flow affect total flow by rail. Thus origin employment affects the total flow by rail through the intermediate variables, total flow and total truck flow. In some cases, a variable may have both a direct and an indirect effect on a nother variable. Then the total effect is the sum of the direct and indirect effects. Only Total and Direct effects are shown in tables presented. Indirect effects can be obtained as the difference between the total and direct effects. From Table 8.1, which presents the Structur al Equations Model estimation results for all commodity groups, some of the important findings are as follows: Employment, both at the origin and des tination end, has a positive impact on freight flows by various modes Population, both at the origin and destin ation, has a negative impact on freight flows by various modes Distance has a negative impact on freight flows by various modes Total flow affects the total truck and rail flows with coefficients less than one. The total flow by truck has a negativ e effect on the total flow by rail All the other commodity groups yi elded models rather similar to this model (Tables B.1 – B.16). As such, this model may be considered illustrative of the types of the models that can be developed and applied using the data base and methods described in this study.

PAGE 110

94Table 8.1 Structural Equations Model Estimation Results for All Commodity Groups Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Rail Flow Full Truck Load Private Truck Total Flow -0.370 Total 0.100 -0.025 -0.012 0.099 -0.026 0.000 0.000 0.000 0.000 0.000 Direct 0.100 -0.025 -0.012 0.099 -0.026 0.000 0.000 0.000 0.000 0.000 Total Truck Flow -0.011 Total 0.098 -0.025 -0.009 0.097 -0.025 0.978 0.000 0.000 0.000 0.000 Direct 0.000 0.000 0.004 0.000 0.000 0.978 0.000 0.000 0.000 0.000 Total Rail Flow 0.009 Total 0.018 -0.004 -0.011 0.019 -0.006 0.192 -1.545 0.000 0.000 0.000 Direct -0.001 0.001 -0.003 0.000 -0.001 1.702 -1.545 0.000 0.000 0.000 Full Truck Load 0.003 Total 0.084 -0.021 -0.007 0.083 -0.021 0.845 0.886 0.028 0.000 0.000 Direct -0.001 0.000 0.000 -0.001 0.000 -0.069 0.930 0.028 0.000 0.000 Private Truck -0.001 Total 0.090 -0.023 -0.007 0.090 -0.023 0.899 0.949 0.017 -0.106 0.000 Direct -0.001 0.000 0.001 0.000 0.000 -0.065 1.074 0.020 -0.106 0.000 Rail Car Load 0.001 Total 0.017 -0.004 -0.011 0.018 -0.006 0.191 -1.505 0.999 -0.008 -0.006 Direct 0.000 0.000 0.000 0.000 0.000 -0.039 0.052 1.000 -0.008 -0.006 Note: N = 859,329; chi-square = 1.064 with df = 6; p-value = 0.983; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 111

95 Figure 8.1 Path Diagram for the Total Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Rail flow Log of FTL Flow Log of PVT Flow Log of Rail Car Load Flow Log of Total Truck Flow 1 2 3 4 5 6

PAGE 112

968.3 Validity of the Model Estimation Results The hypothesis commonly used in freight tr ansportation planning is that population is assumed to affect the attraction of freight to an area, and industry employment is assumed to affect the generation of freight in an area. The generation of a particular commodity in an area has traditionally been tied to employment within the commodity’s industry. Thus, the positive effect of employment on flows by various modes is plausible and is as expected. One can hypothesize that as the employment increases, the total flow as well as modal flows between an origin-d estination pair increase. The variation in the total freight flow as a result of the vari ation in employment is estimated in the next chapter. Distance is found to have a ne gative impact on freight flow s between origin-destination pairs. Once again, this finding is consistent with expectations as distance constitutes a measure of impedance. While there are ce rtainly strategic leve l decisions regarding facility location and cu stomer clustering that tends to ma ke distance a secondary variable in influencing freight flows, one can not ignor e the possibility that distance is correlated with the quantity of freight flow between an origin-destination pair. For the state of Florida, recent Commodity Flow Surveys have indicated that about 60 percent of freight movements by value and 80 percent of freight movements by weight occur within the state. Clearly, distance is playing a major ro le in shaping the dist ribution and quantity of freight flows in Florida. In fact, about 70 percent (by weight) of all commodity flows originating in Florida travel less than 100 m iles. The distance variable in the models simply reflects this tendency in the freight flow database and is found to offer statistically significant and intuitively plausible coefficients.

PAGE 113

97 Moreover, as expected, total flow affects the total truck and rail flows with coefficients less than one. These coefficients represen t how the total flow between an origindestination pair contributes to the different types of modal flows between an origindestination pair. A rather surprising finding is that the or igin and destination population variables are found to have a negative impact on freight fl ows in both the models. It was originally expected that population variables would have a positive impact on the quantity of flow. However, estimation results show that populat ion variables are associated with negative coefficients. On the other hand, the employ ment variables have po sitive coefficients. Thus, it appears that employment is the key driv er of freight flow activity while resident population is not a key driver of statewide freight flow activ ity. This could be explained by the following arguments. Business establishments, manufacturing & pr oduction operations, an d other industrial land uses contribute to heavier volumes of fr eight flow. Many of these industrial sites are located in zip codes with minimal residentia l population, but attrac t and generate large amounts of freight flow. The negative coeffici ents for population variables may be due to the fact that freight is more likely to be pr oduced and attracted in such rural areas with small populations that have more land avai lable to support larg e-scale manufacturing activities. Thus, the presence of a residential population does not necessarily contribute positively to freight flows between origin-des tination pairs at a statewide level.

PAGE 114

98 Freight flows made up of finish ed products may be expected to be shipped directly to consumers rather than being transported to warehouses or other industries for further distribution. Within an urba n area context, where one is concerned with movement of finished goods, one may conjecture that bot h business establishments and residential population contribute positively to truck trip ge neration. However, within the context of a statewide freight flow analysis where the fr eight flows are mostly industrial raw goods, residential population is not likely to attract fr eight trips. Thus, it appears that this finding may have some merit in the statewide modeli ng context. This findi ng also lends credence to the approach taken by many states and urban areas that try to attract “jobs” to their area to promote economic activity. The notion is that people will then come to where the “jobs” are located. Moreover, zipcodes with higher populations ha ve lesser growth related activities, thus diminishing the demand for construction related commodities. This explains the negative coefficients for population in the case of c onstruction related commodity groups. Thus, as mentioned in Chapter 3, inclusion of growth re lated variables such as data on estimates of future dwelling construction and other major construction sites (e.g. new road or rail links, or major urban redevelopment sites) could be used in modeling the demand for construction related commodities. In the future, growth variables should be included in the model specification for construction relate d commodity groups to make sure that the model is sensitive to these attributes. Future researchers may also wish to investigate the shipment characteristics of commodities to provide a more precis e explanation of the negative coefficients of population variables.

PAGE 115

99 Finally, it should be noted that previous research in the development of statewide freight trip generation models also found negative coefficients as sociated with the population variables. In a similar piece of work, Brogan, et al (2001) provides freight trip generation equations (single production a nd attraction equations by commodity group) estimated on the TRANSEARCH database. In their equati ons, the population va riables are found to have negative coefficients and employ ment variables are found to have positive coefficients. Thus the models developed in this study appear to provide very robust indications of the effects of residential population on orig in-destination freight flows by commodity and mode. Sensitivity analysis can be performed on these estimated models to estimate the effects of population along with the other exogenous variables on the freight flows. Overall, the SEMs specified and estimated in this chapter corrobor ated their potential effectiveness in unraveling complex struct ural relationships among socio-economic characteristics, modal level of service char acteristics and freight fl ows on various modes. It is also found that freight travel demand can well be ad dressed using the structural equations framework. The ensuing chapter focuses on the sensitivity analysis for these models to estimate the variati ons in freight demand in respon se to hypothetic al variations in the exogenous variables.

PAGE 116

100 Chapter 9 Sensitivity Analysis 9.1 Background Mathematical and computational models are us ed in a variety of settings and purposes, often to gain insight of possible outcomes of one or more courses of action. These courses of action may be a policy action, the assessment of industrial practices or environmental impacts. Sensitivity analysis is the study of how the vari ation in the output of a model (numerical or otherwise) can be apportioned, qualitativel y or quantitatively, to different sources of variation. Sensitivity analysis aims to ascertain how the model depends upon the information fed into it, upon its structure and upon the framing assumptions made to build it. This information can be invaluable, as Different level of acceptance (by the de cision-makers and stakeholders) may be attached to different types of uncertainty. Different uncertainties impact differently on the reliability, the robustness and the efficiency of the model. Originally, sensitivity analysis was created to deal simply with uncertainties in the input variables and model parameters. Over the course of time the ideas have been extended to

PAGE 117

101 incorporate model conceptual uncertainty, i.e. uncertainty in model structures, assumptions and specifications. As a whole, sensitivity analysis is now being used to increase the confidence in the model and its predictions, by providing an understanding of how the model response vari ables respond to changes in the inputs, be they data used to calibrate it, model structures, or factor s, i.e. the model i ndependent variables. Sensitivity analysis is thus closely linked to uncertainty analysis, which aims to quantify the overall uncertainty associated with the re sponse as a result of uncertainties in the model input. In this chapter, sensitivity analysis is perfor med to examine changes in the total flow of a commodity brought about by changes in explanator y variables. The inte nt of this chapter is to demonstrate the applicability of th e 17 SEM models described in the previous chapter. 9.2 Sensitivity Analysis In order to perform sensitivity analysis, vari ations in commodity flows with regard to different hypothetical increments of the ex planatory variables (d estination employment, origin employment, distance, destination popul ation, and origin population) are predicted using the models developed. In all, increm ents from 10% to 100% of the explanatory variables are considered. Two base cases have been considered: Base Case I: All the explanatory variables are at their mean values and Base Case II: All explanatory vari ables are at their maximum values

PAGE 118

102 In the data set that was prepared for modeling, the mean value for origin and destination employment is 6739, while the mean distance between the zipcodes is 153 miles and the mean value for origin and destination popul ation is 17238 (used in Base Case I). The maximum value for origin and destination employment is 53604, while the maximum distance between the zipcodes is 509 mile s and the maximum value for origin and destination population is 74476 (u sed in Base Case II). 9.3 Results of Sensitivity Analysis The sensitivity analysis for all commodity gr oups combined for the first base case where all explanatory variables are at their m ean values is presented in Table 9.1. Table 9.1 Percentage Increase in Total F reight Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.30 6.33 9.15 11.77 14.23 16.55 18.74 20.82 22.80 24.68 Origin Employment 3.26 6.26 9.05 11.64 14.07 16.37 18.53 20.59 22.54 24.40 Distance -0.39 -0.76 -1.10 -1.41 -1.70 -1.97 -2.23 -2.47 -2.69 -2.91 Destination Population -0.82 -1.57 -2.27 -2.91 -3.50 -4.06 -4.58 -5.07 -5.53 -5.97 Origin Population -0.83 -1.59 -2.29 -2.94 -3.54 -4.11 -4.63 -5.13 -5.60 -6.04 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 From Table 9.1, it can be seen that the to tal freight flow increases by 3.3%, as the destination employment increases by 10%. Li kewise, as the destination employment increases by 100%, the total freight flow in creases by 24.68%. The increments in total

PAGE 119

103 freight flow for other increments in the de stination employment can be seen from the table. Similarly, as the origin employment increases by 10%, the total freight flow increases by 3.26%. Also, as the origin em ployment increases by 100 %, the total freight flow increases by 24.4%. Thus, it can be seen th at the increments in the total freight flow are roughly the same for origin and destination employments. It can also be seen from Table 9.1 that in crease in distance by 10% decreases the total freight flow by 0.39%. The decrease in the total freight flow is only 2.91% for 100% increase in distance. For the destination population, the decreas e in the total freight flow is 0.82% and 5.97% for an in crease by 10% and 100% respectivel y. Quite similar to these figures is the origin populat ion which when increased by 10% and 100% respectively, decreases the total freight flow by 0.83% and 6.04% respectively. Thus, it can be clearly seen that employment, both at the origin and destination is a key driver of freight flows. When a region’s employment at the mean value is doubled, the total freight flow increases by around 25%. The distance has a very small effect on the total freight flow, as can be seen from the f act that in spite of doubling the distance of travel from the mean value, the total fr eight flow decreases by a mere 3%. Also, population, both at the origin and destination has a nominal effect by decreasing the total freight flow by 6% in spite of doubling the population from the mean value. One finds the same kind of effects from Table 9.2, which presents the sensitivity analysis for all commodity groups combined for th e second base case where all explanatory

PAGE 120

104 variables are at their maximum values. In this table, it can be noticed that the effects of all the variables remain same. However, the magnitudes of the effects are considerably lesser than those found in the base case wher e all the explanatory variables are assumed to be at their mean values. In this case, it can be observed that the total freight flow increases by 14.76% when the destination employment increases by 100%. The origin employment also has similar magnitude and eff ect in that the total freight flow increases by 14.60% when the origin employment increa ses by 100%. Distance has a negligible effect on the freight flow, as an increase in distance by 100% decreas es the freight flow only by 1.75%. Destination and origin populati ons also have a negligible effect. An increase in destination population by 100% decreases the freight flow only by 3.57%. Similarly, an increase in destination populat ion by 100% decreases th e freight flow only by 3.61%. Table 9.2 Percentage Increase in Total F reight Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.97 3.79 5.47 7.04 8.51 9.90 11.21 12.45 13.64 14.76 Origin Employment 1.95 3.75 5.41 6.96 8.42 9.79 11.08 12.31 13.48 14.60 Distance -0.24 -0.46 -0.66 -0.85 -1.02 -1.18 -1.34 -1.48 -1.62 -1.75 Destination Population -0.49 -0.94 -1.36 -1.74 -2.10 -2.43 -2.74 -3.03 -3.31 -3.57 Origin Population -0.50 -0.95 -1.37 -1.76 -2.12 -2.46 -2.77 -3.07 -3.35 -3.61 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 121

105 It can be also seen easily that the change s in the freight flow are higher when the explanatory variables are around their means th an at maximum values. For example, the destination employment when incremented by 100% from the mean increases the freight flow roughly by 25%, while the same when incremented by 100% from the maximum value increases the freight flow roughly by 15%. Tables D.1 to D.32 present the sensitivity analysis for each of the 16 co mmodity groups for which SEM models have been estimated. The results are quite simila r to the results in Tables 9.1 and 9.2. The results presented in this chapter clearly ad d evidence to the fact that employment is the key factor influencing freight flows between two regions. It is to be noted that shifts in freight flows due to hypothetical increments in employment are very high compared to the other variables considered. The distan ce and population, both at the origin and destination only have a small effect. Nevertheless, significa nt causal relationships among socio-economics, modal level of service char acteristics, and frei ght flows by various modes are discerned which are critical to the designing of new and complex transportation policies. In conclusion, this chapter has established the role of the structural equation modeling methodology in assessing impacts of socio-eco nomic characteristics and modal level of service characteristics on freight flows betw een two regions. The sensitivity analysis discussed in this chapter ha s thus demonstrated that th e SEM methodology is capable of providing a practical tool for es timating freight flows in the c ontext of a statewide freight demand modeling framework.

PAGE 122

106 Chapter 10 Conclusions and Further Research 10.1 Background Freight transportation lies at th e heart of our economic life. In industrialized countries, it accounts for significant share of the gross national product. In developing countries, it is the essential ingredient of sustainable deve lopment. With free trade zones emerging in several parts of the world and with the gl obalization of the economic system, freight transportation will in all like lihood play an even more majo r role in years to come. The trend towards larger, more integrated and more efficient transportation systems is likely to remain and should create the need for bette r planning at the stra tegic, tactical and operational levels. Thus, major advances have been made in the development of freight transportation modeling methods and framewor ks to assist in fr eight transportation planning efforts. Freight transportation model de velopment is now a critical component of the overall transportation planning process as urban areas, states, and th e nation consider mobility strategies for enhancing the safety and effici ency of freight transportation. Metropolitan Planning Organizations, transportation planne rs and researchers have attempted to forecast future freight supply and demand in order to estimate future needs. However, the successful implementation of freight dema nd modeling is hamper ed by the lack of

PAGE 123

107 appropriate transportation modeling methodol ogies. The advanced methods developed have not seen application in pr actice partially due to the lack of adequate data to support their estimation and application to forecasting. The most significant hurdle to the inclus ion of freight transportation into the transportation modeling process is related to the prevailing lack of knowledge on the fundamental mechanisms conditioning freight demand and supply. In order to develop reliable freight demand models, it is esse ntial to understand the mechanisms driving freight demand and incorporate the behavior into the modeling process. However, the development of behavioral freight demand mo dels faces significant hurdles which are a consequence of the inherent complexity of the mechanisms driving freight demand. A number of factors add complexity to frei ght demand: a) there are multiple dimensions (i.e., value, weight, volume, trips) to be c onsidered; b) there are multiple decision makers (e.g., drivers, dispatchers) that interact dynami cally and take decisions that affect freight demand; c) these interactions takes place in a private context, for the most part not accessible to transportation planners; d) the opportunity costs of the cargoes exhibit a wide range, resulting in multiple user classe s ranging from products such as gypsum with a market value of $9/ton; to products such as computer chips that cost in excess of $500,000/ton; and e) freight demand data is for the most part considered to be commercially sensitive. Freight transportation data has been traditionally difficult to collect due to the proprietary nature of the data and due to the difficulty with identifying the proper entity to which a freight transpor tation survey needs to be administered.

PAGE 124

108 In this regard, the state of freight dema nd modeling suffers from a compromise between the behavioral validity and the data requirement s. It has been identified in the literature review that the models that are able to captu re the behavioural aspects of freight demand suffered with two main drawbacks: the data requirement and the computational complexity of the solving process. In contrast those models that are simple and practical do not capture the behavioral aspects of freight transportation in a comprehensive manner. 10.2 Conclusions In this context, the objective of this study was to propose a relative ly data simple, but behaviorally robust statewide modeling framework for the state of Florid a, in the spirit of an aggregate level four-step planning pro cess. The model formulation and empirical analysis in this study were specifically targeted toward the trip generation, trip distribution and mode choice steps. The goa l was to propose a modeling framework that can quantify and predict freight flows by various modes betw een origin-destination pairs in the state of Florida. These origin-destination pairs may be traffic analysis zones, census tracts, zip codes, cities, counties, or ev en states depending on the particular freight transportation planning context of interest. But, it was desired to c onduct this study at the microscopic level of a zipcode unlike any of the earlier studies. The zip code level is considered an appropriate level of disaggrega tion where the data can be considered to be reliable avoiding large amounts of missing data.

PAGE 125

109 Contrary to passenger transportation, in whic h there is only one unit of demand, (i.e., the passenger who is also the decision maker) in freight transportation there are multiple dimensions (e.g., volume, wei ght, and vehicle-trips) to be taken into account when modeling freight movements. The multiple variables that could be used to measure and define freight demand, have given rise to two major modeling platforms: commoditybased and trip-based modeling. In order to develop a behaviora lly robust statewide modeling framework that considers the cargoe s’ economic characteristics, commoditybased modeling has been adopted in this study. The hypothesis used in this study is larg ely in line with paradigms and freight transportation demand-supply relationships identif ied in the literature. It is assumed that freight movement is fundamentally gene rated by the demand for consumption of commodities at the destination (or attracti on) region, which is met by the flow of commodities from one or more origin (or pr oduction) regions. Thus a model concept that predicts freight flows on various modes between two zipcodes based on the population characteristics, employment characteristics, an d the modal level of service characteristics has been developed. The model framework is simple & practical, but behaviorally robust and can therefore be easily estimated on a da tabase that can be a ssembled by any public agency that has resources to purch ase some commercial databases. After a review of the various freight data sources, it wa s found that the TRANSEARCH database suits best for statewide freight demand modeling of Florida. Thus, the model development in this study is based on the TR ANSEARCH freight flow database that is

PAGE 126

110 commercially available from Reebie Associat es. This database, pr oviding freight flow information at the zip code level, was me rged with population information from the Census 2000 database and employment info rmation from the InfoUSA 2000 database. The resulting database constituted a comprehensive database for modeling freight flows between origin-destination pairs. The only mi ssing component in the database is modal level of service attributes that would poten tially influence frei ght flows by mode (by commodity) between origin-destination pairs. The process of merg ing modal level of service attributes is currently ongoing and will result in further enhancement of the models developed in this paper. However, simple map distance between zipcodes has been used in this study. The exploratory analysis conducted on the data base suggested that the database offers variables with plausible statistical dist ributions and summaries consistent with expectations. The modeling methodology consis ted of a structural equations modeling framework that can accommodate multiple dependent variables simultaneously. This structural equations model can be applied to all origin-destination zip code pairs in a region. In this model system, explanatory va riables representing origin and destination population and employment characteristics a nd impedance (distance between the origindestination pair) are included. The models for various commodity groups are found to offer statistically valid indications and plausible interp retations suggesting that thes e models may be suitable for application in freight tran sportation demand forecasting a pplications. Likelihood ratio 2

PAGE 127

111 tests showed that population, employment a nd distance are all important and significant in explaining the freight demand and its mode choice. The analysis results for all the commodity groups demonstrated positive relationship between freight demand and employment both at the origin and destination. Consistent significant positive relationships between em ployment and freight flow for all the commodity groups indicate that employment has a strong influence on all kinds of commodity flows. Similarly, for all the co mmodity groups, distance is found to have a negative impact on freight flows by various modes as expected. A rather surprising finding was that the or igin and destination population variables are found to have a negative impact on freight flows in all the models. Based on the hypothesis used in this study, it was originally expected that popula tion variables would have a positive impact on the quantity of flow However, it appears that this finding may have some merit in the statewide modeling context. Within the context of a statewide freight flow analysis where th e freight flows are mostly i ndustrial raw goods, residential population is not likely to attrac t freight trips. However, a more in-depth exploration is required to base this conclu sion on a better standpoint. Sensitivity analysis has been conducted in order to examine changes in the total flow of a commodity brought about by changes in expl anatory variables. The results obtained clearly add evidence to the fact that employ ment is the key factor influencing freight flows between two regions. Shifts in freight flows due to hypothetical increments in

PAGE 128

112 employment were very high compared to the other variables considered. This indicates that employment variables ar e extremely important in expl aining the freight demand and that their omission from the models may be more serious than the omission of demographic and modal level of service variables. In conclusion, sensitivity anal ysis has established the role of the structural equation modeling methodology in assessi ng impacts of socio-economic characteristics and modal level of service characteristics on freight fl ows between two regions. Significant causal relationships among socio-economics, modal leve l of service characteristics, and freight flows by various modes were discerned which are critical to the designing of new and complex transportation policies. Thus it has been demonstrat ed that the SEM methodology is capable of providing a practical tool for estimating freight flows in the context of a statewide freight demand modeling framework. 10.3 Role in the Overall Planning Process This research focuses on a statewide freight demand model for the state of Florida that takes the structure of a co mmodity-based model. The Structural Equations Modeling methodology developed in this study is speci fically targeted toward the commodity generation, commodity distribution and commodity mode split steps. These three steps in a typical commodity-based model are combin ed in a unique SEM methodology. Thus the outputs of this model system are Origin-D estination commodity volume matrices for various modes. The inputs to the modeling sy stem are the socio-ec onomic characteristics

PAGE 129

113 and the modal level of service characteri stics. The role of SEM methodology in the overall planning process is depicted in Figure 10.1. Figure 10.1 Flow Diagram for the Planning Process The outputs of the SEM model system, as can be seen from Figure 10.1 are OriginDestination commodity volume matrices by vari ous modes. The next step in the overall planning process is the traffi c assignment phase which is a combination of vehicle loading models and complementary models th at capture empty trips, using the origindestination matrices by mode. These vehicle trips are then assigned to the network. Structural Equations Modeling Commodity Generation Commodity Distribution Commodity Mode Split O-D Commodity Volumes by Mode Vehicle-trip Estimation Traffic Assignment Socio-Economic Characteristics Modal Level of Service Characteristics

PAGE 130

11410.4 Model Responsiveness The model responsiveness in comparison to other methodologies is presented in Table 10.1. As tabulated in this table, SEM has mo dest data requirements and is behaviorally robust compared to other pract ical approaches such as gr owth factor models and loglinear regression. Table 10.1 Model Responsiveness in Comparison to Other Methodologies Structural Equations Modeling Growth Factor Models Log-Linear Regression Aggregate Cost Function Approaches Data Needs Modest Minimal Minimal Large Practicality Modest Easy-to-use Easy-to-use Complex Computational Complexity Modest Low Low High Theoretical Foundation Modest Low Partial High Predictive Capability Modest Low Low High Sensitivity to Socio-Demographics Yes No Yes Yes Modal LOS Yes No No Yes Industrial Organization No No No Yes 10.5 Further Research A limitation of this research is that some important variables were not included in the modeling in order to strike a balance between the behavioral capture and data simplicity of the statewide model framework. For exampl e, data on estimates of future dwelling construction and other major construction site s (e.g. new road or rail links, or major urban redevelopment sites) could be used to estimate the demand for construction related commodities. Moreover, land use data may also be used for estimating the agricultural freight demand. Inclusion of such variable s will enhance the model reliability, even though the modeling framework might tend to be data intensive.

PAGE 131

115 The development of freight databases and coll ection of freight movement data continues to be a challenge for model development and estimation. There is always a degree of uncertainty regarding the coverage of the da tabase with respect to geography, commodity groups, and modes and regarding the accuracy of the data as one goes to greater levels of spatial detail. Undoubtedly, further research in understanding the underlying behavior of freight flows is important. Data collection effo rts can be targeted based on such research. These days, massive volumes of data are being collected everyday around the globe, hence the co-ordination between different act ors collecting similar data has become a new challenge, as the same data measured by different entities do not always match. The development of freight transportation mode ls is making great strides, but there is some question as to how transferable these models are between geographic contexts and between geographic scales within the same cont ext. How applicable is it to use a model system estimated at the zip code level at anot her level of aggregation such as census tract or traffic analysis zone? Research into thes e issues will greatly enhance our ability to develop freight transportation models a nd estimate freight flows accurately while analyzing the effects of alternative freight mobility strategies and policies.

PAGE 132

116 References Allen, B. W. The Demand for Freight Transp ortation: A Micro Appr oach. Transportation Research 11, 9-14, 1977. Arbuckle, J. L. AMOS Users Guide Version 4.01. Small Waters Corporation, Chicago, IL, 2000. Baumol, W. J., and Vinod, H. D. An invent ory theoretic model of freight transport demand. Management Science, 16, 413-421, 1970. Bollen, K. A. Structural Equations with Latent Variables. Wiley, New York, 1989. Boyer, K. D. Minimum rate regu lation, modal split sensitivitie s, and the railroad problem. Journal of Political Economy, 85, 493-512, 1977. Brogan, J., Stephen, C. B., and Demetsky, M. J. Application of a Statewide Intermodal Freight Planning Methodology. Virginia Transportation Research Council, VTRC, 02-R5, August 2001. Bronzini, M. S. Evolution of a Multimodal Freight Transportation Network Model. Proceedings of the Transportation Research Forum, 21(1), 475-485, 1980. Browne, M. W. Asymptotically Distribu tion Free Methods for the Analysis of Covariance Structures. British Jour nal of Mathematical and Statistical Psychology 37, 62-83, 1984. Bureau of Transportation St atistics. Commodity Flow Surv ey 1997, US Department of Transportation, 1999. Cambridge Systematics, Inc. A Guidebook for Forecasting Freight Transportation Demand. NCHRP Report 388, Transpor tation Research Board, National Research Council, Washington, D. C., 1997. Cambridge Systematics, Inc., COMSIS Cor poration, and University of WisconsinMilwaukee. Quick Response Freight Manua l. Final Report DOT-T-97-10, U.S. Department of Transportation, 1997.

PAGE 133

117 Cassing, J. Transport costs in international trade theory: a comparison with the analysis of non-traded goods. Quarterly Journal of Economics, Volume 92, Number 4, pp. 535-550, 1978. Crainic, T. G. & Laporte G. Planning models for freight transporta tion European Journal of Operational Research 97, 409-438, 1997. Crainic, T. Operations Research Models of Intercity Freight Transportation: The Current State and Future Research Issues. Logist ics and Transportati on Review, 23(2), pp. 189-206, 1987. Cutler M. L., Grenseback, R. E., Paquette, D., Beagan, K., Proussaloglou, Jonnalagadda, N., Williamson, M., Schrieber, J. (2000) The Assessment of Market Demand for Cross-Harbor Rail Freight Service in th e New York Metropolitan Region, 79th TRB Annual Meeting, Washington D.C. Czerniak, R. J., Lahsene, J. S., and Chatte rjee, A. Urban Freight Movement, What Form will it Take?, Transportation in the Ne w Millennium, 7p, A1B07: Committee on Urban Goods Movement, Washington D. C., Transportation Research Board, 2000. Das, C. (1974). Choice of transport servic e: An inventory theoretic approach. The Logistics and Transporta tion Review, 10(2), 181-187. DeWitt, W., and Clinger, J. Intermodal Freight Transportation, Transportation in the New Millennium, A1B05: Committee on Interm odal Freight Transport, Washington, D.C., Transportation Research Board, 2000. Este, G. D. (2002). Urban Freight Move ment Modeling, Handbook of Transportation Modeling, Edited by D. A. Hensher and K. J. Button. Faris, J. M., and Ismart, D. Freight Mode ling Techniques for Small and Medium Sized Areas. Sixth National Conference on Tr ansportation Planning for Small and Medium-Sized Communities, Washington State Department of Transportation, 1999. Friesz, T. Strategic Freight Network Pl anning Models. In: Handbook of Transport Modeling, 527-537, Elsevier Science Publishers, 2000. Friesz, T., Tobin, R. and Harker, P. Predictive intercity network models: the state of the art. Transportation Research A, Volume 17, 409-417, 1983. Garrido, R. A. Spatial interaction between the truck flows through the Mexico-Texas border”. Transportation Research A, Volume 34, Number 1, 23-33, 2000

PAGE 134

118 Golob, T. F., and McNally, M. G. A Model of Activity Participation and Travel Interactions between Household Heads. Tr ansportation Research B, Vol. 31, No. 3, 1997, pp. 177-194. Hancock, K. L., Karthikeyan, S., and Sreek anth, A. Freight Transportation Planning: Converting Commodity Volumes to Vehicl es, Final Report, 128. University of Massachusetts, Amherst, 2000. Harker, P. T. The State of the Art in the Pr edictive Analysis of Freight Transportation Systems. Transport Reviews 5(3), pp. 44-62, 1985. Harker, P. T., and Friesz, T. L. Prediction of Intercity Freight Flows. Part I: Theory and Part II: Mathematical Formulations. Tr ansportation Research B 20 (2), pp. 139153, 155-174, 1986. Hautzinger, H. The Prediction of Interr egional Goods Vehicle Flows: Some New Modeling Concepts. Ninth Internati onal Symposium on Transportation and Traffic Theory, VNU Science Press, 375-396, 1984. Holguin-Veras, J., and Jara-Diaz, S. Optim al space allocation and pricing for priority service at container por ts. Transport Research Part B 33 (2), 81–106, 1999. Holguin-Veras, J., and Thorson, E. Modeling commercial vehicle empt y trips with a first order trip chain model. Transportati on Research Part B, 37, 129–148, 2003. Holguin-Veras, J., and Thors on, E. Trip Length Distribut ions in Commodity-Based and Trip-Based Freight Demand Modeling. In Transportation Research Record 1707, TRB, National Research Council, Wa shington, D. C., pp. 37-48, 2000. Hu, P. (2001). Developing an E-sensible tran sportation system: what are the research and data needs? Transportation Research News 216: 4–8, September–October 2001, Transportation Research board, Washington DC. Inamura, H. and Srisurapanon, V. Freight flow forecasting model using a rectangular input-output system. 8th World Conferen ce on Transport Research Conference, 12-17 July, 1998, Antwerp, Belgium, 1998. Jack Faucett Associates (b). Research a nd Development of Destination, Mode, and Routing Choice Models for Freight, Final Report, Prepared for DOT SBIR Office, DTS-22, May 20, 1999. Johnston, J., and DiNardo, J. Econometric Methods. Fourth Edition, Wiley, New York, 1997.

PAGE 135

119 Joreskog, K. G., and Sorbom, D. LISREL 8 Users Reference Guide. Scientific Software, Inc., Chicago, IL, 1993. Judge, G., Griffiths, W. E., Hill, R. C., Lu tkepohl, H., and Lee, T. C. The Theory and Practice of Econometrics. Second Edition, Wiley, New York, 1985. Kumar, A. S., and Bhat, C. R. Freight M odal Split Modeling: Conceptual Framework, Model Structure and Data Sources. Cent er for Transportation Research, CTR, 1833-5, 2002. Lahsene, S. New economy, new vision for tran sportation: prominent role for intermodal Freight, Transportation Research News 216: 9–11, Transportation Research Board, Washington D. C., September–October 2001. Leontief, W. W. The Structure of the Am erican Economy 1919-1939, 2nd Edition, 1951. New York: Oxford University Press, 1941. Levin, R. C. Allocation in surface freight transportation: Does rate regulation matter? Bell Journal of Economics, 9, 18-45, 1978. List, G. F., and Turnquist, M. A. Estimating Truck Travel Patterns in Urban Areas. In Transportation Research Record 1430, TRB, National Research Council, Washington, D. C., pp. 87-93, 1995. Maddala, G. S. Limited Dependent and Qualitative Variables in Econometrics. Cambridge University Press, Cambridge, MA, 1983. Markusen, J. R., and Venables, A. J. Mul tinational firms and the new trade theory. Journal of Internationa l Economics 46, 183-203, 1998. McFadden, D. and Winston, C. Joint estima tion of discrete and continuous choices in freight transportation. Proceedings of the 1981 meeting of the Econometric Society, 1981. McFadden, D., Winston, C., and Boersh-S upan, A. Joint estimation of freight transportation decisions under nonrandom sa mpling. In: Analytical Studies in Transport Economics (Daughety A. F. ed .), Cambridge: Cambridge University Press, 137-157, 1985. Morton, A. A statistical sketch of intercity freight demand. Highway Research Record, 166, 1969. Ogden, K. W. Urban Goods Movement: A Gu ide to Policy and Planning. ISBN 1-85742029-2. Ashgate Publishing Limited, Aldershot, England, 1992.

PAGE 136

120 Ogden, K. W. Truck Movement and Access in Urban Areas. Journal of Transportation Engineering, ASCE, Vol. 117, No. 1, pp. 71-90, 1991. Ogden, K.W. The Distributi on of Truck Trips and Commod ity Flow in Urban Areas: A Gravity Model Analysis. Transportation Research, Vol. 12, 131-137, 1978. Oum, T. H. (1979). A Warning on the Use of Linear Logit Models in Transport Mode Choice Studies. Bell Journal of Economics, 10(1), 374-388. Owoc, M. & Sargious, M., A. The role of tr ansportation in free trade competition, in: N. Waters, ed., Canadian Transportation: Competing in a Global Context, 23-32, 1992. Pendyala, R. M., Shankar, V. N., and McCu llough, R. G. Frei ght Travel Demand Modeling: Synthesis of Approaches a nd Development of a Framework. In Transportation Research Record 1725, J ournal of the Transportation Research Board, TRB, National Research Council, Washington, D. C., 9-16, 2000. Quandt, R. E. and Baumol, W. J. The demand for abstract transport modes: Theory and measurement. Journal of Regi onal Science, 6(2), 13-26, 1966. Regan, A. C., Holguin-Veras, J., Garland, C., and Miles, H. S. Freight Transportation Planning and Logistics, Transportation in the New Millennium, 5p, A1B02: Committee on Freight Transpor tation Planning and Logist ics, Washington, D. C., Transportation Research Board, 2000. Regan, A., C. & Garrido, R., A. Modeling Freight Demand and Shi pper Behavior: State of the Art Future Directions, Institute of Transportation Studi es, University of California, Irvine, 2002. Shankar, V. N., and Pendyala, R. M. Frei ght Travel Demand Modeling: Econometric Issues in Multi-level Approaches. In D. Hensher, ed., The Leading Edge of Travel Behaviour Research, Elsevi er Science Publishers, B. V., The Netherlands, Chapter 38, 629-644, 2001. Slavin, H. L. Demand for Urban Goods Vehicl e Trips. Transportation Research Record 591, National Research Council, Washington, D.C., 32-38, 1976. Souleyrette, R., Maze, T. H, Strauss, T., Prei ssig, D., and Smadi, A. G., Freight Planning Typology. Transportation Research Record 1613, 12-19, 1998. Southworth, F., and Peterson, B. E. Inte rnational and Interm odal Freight Network Modeling, Transportation Re search C 8(1), 147-166, 2000.

PAGE 137

121 Strong, C.K., Harrison, R., and Mahmassani H. S. A Methodology for Determining the Freight Border Transportation Impact of the North American Free Trade Agreement, University of Texas, Cent er for Transportation Research, Report 1319-4, 1996. Watson P. L. Urban Goods Movement, A Di saggregate Approach, D.C. Health and Company, 1975. Winston, C. The demand for freight tran sportation: models and applications. Transportation Research Part A, 17(6), 419-427, 1983. Zalatan, P. Economic cycles, structural cha nge and the transportation sector, in: B.G. Bison, ed., A look Back from the Y ear 2000, 28th Annual Meeting of the Canadian Transportation Research Forum, 111-121, 1993.

PAGE 138

122 Appendices

PAGE 139

123Appendix A Figure A.1 Freight Outflows in Annual Tons by Zipcode

PAGE 140

124Appendix A (Continued) Figure A.2 Freight Inflows in Annual Tons by Zipcode

PAGE 141

125Appendix A (Continued) Figure A.3 Ratio of Freight Outflows to Inflows by Zipcode

PAGE 142

126Appendix A (Continued) Figure A.4 Truck Outflows in Annual Tons by Zipcode

PAGE 143

127Appendix A (Continued) Figure A.5 Truck Inflows in Annual Tons by Zipcode

PAGE 144

128Appendix A (Continued) Figure A.6 Ratio of Truck Outflows to Truck Inflows by Zipcode

PAGE 145

129Appendix A (Continued) Figure A.7 Rail Outflows in Annual Tons by Zipcode

PAGE 146

130Appendix A (Continued) Figure A.8 Rail Inflows in Annual Tons by Zipcode

PAGE 147

131Appendix A (Continued) Figure A.9 Ratio of Rail Outflows to Rail Inflows by Zipcode

PAGE 148

132Appendix A (Continued) Figure A.10 Water Outflows in Annual Tons by Zipcode

PAGE 149

133Appendix A (Continued) Figure A.11 Water Inflows in Annual Tons by Zipcode

PAGE 150

134Appendix A (Continued) Figure A.12 Ratio of Water Outflows to Water Inflows by Zipcode

PAGE 151

135Appendix A (Continued) Figure A.13 Air Outflows in Annual Tons by Zipcode

PAGE 152

136Appendix A (Continued) Figure A.14 Air Inflows in Annual Tons by Zipcode

PAGE 153

137Appendix A (Continued) Figure A.15 Ratio of Air Outflows to Air Inflows by Zipcode

PAGE 154

138Appendix A (Continued) Figure A.16 Freight Outflows in Annual Tons by County

PAGE 155

139Appendix A (Continued) Figure A.17 Freight Inflows in Annual Tons by County

PAGE 156

140Appendix A (Continued) Figure A.18 Ratio of Freight Outflows to Inflows by County

PAGE 157

141Appendix A (Continued) Figure A.19 Truck Outflows in Annual Tons by County

PAGE 158

142Appendix A (Continued) Figure A.20 Truck Inflows in Annual Tons by County

PAGE 159

143Appendix A (Continued) Figure A.21 Ratio of Truck Outflows to Truck Inflows by County

PAGE 160

144Appendix A (Continued) Figure A.22 Rail Outflows in Annual Tons by County

PAGE 161

145Appendix A (Continued) Figure A.23 Rail Inflows in Annual Tons by County

PAGE 162

146Appendix A (Continued) Figure A.24 Ratio of Rail Outflows to Rail Inflows by County

PAGE 163

147Appendix A (Continued) Figure A.25 Water Outflows in Annual Tons by County

PAGE 164

148Appendix A (Continued) Figure A.26 Water Inflows in Annual Tons by County

PAGE 165

149Appendix A (Continued) Figure A.27 Ratio of Water Outflows to Water Inflows by County

PAGE 166

150Appendix A (Continued) Figure A.28 Air Outflows in Annual Tons by County

PAGE 167

151Appendix A (Continued) Figure A.29 Air Inflows in Annual Tons by County

PAGE 168

152Appendix A (Continued) Figure A.30 Ratio of Air Outflows to Ai r Inflows Ratio by County

PAGE 169

153Appendix A (Continued) Figure A.31 Population by Zipcode

PAGE 170

154Appendix A (Continued) Figure A.32 Annual Tons Exported per Person by Zipcode

PAGE 171

155Appendix A (Continued) Figure A.33 Annual Tons Imported per Person by Zipcode

PAGE 172

156Appendix A (Continued) Figure A.34 Population by County

PAGE 173

157Appendix A (Continued) Figure A.35 Population Density by County

PAGE 174

158Appendix A (Continued) Figure A.36 Annual Tons Exported per Person by County

PAGE 175

159Appendix A (Continued) Figure A.37 Annual Tons Imported per Person by County

PAGE 176

160Appendix A (Continued) Figure A.38 Employer Locations in Florida

PAGE 177

161Appendix A (Continued) Figure A.39 Employment by Zipcode

PAGE 178

162Appendix A (Continued) Figure A.40 Agricultural, Forestry and Fishery Employment by Zipcode

PAGE 179

163Appendix A (Continued) Figure A.41 Mining and Construction Products Employment by Zipcode

PAGE 180

164Appendix A (Continued) Figure A.42 Light Manufactured Products Employment by Zipcode

PAGE 181

165Appendix A (Continued) Figure A.43 Heavy Manufactured Products Employment by Zipcode

PAGE 182

166Appendix A (Continued) Figure A.44 Transportation, Communication and Utilities Employment by Zipcode

PAGE 183

167Appendix A (Continued) Figure A.45 Wholesale and Retail Trade Employment by Zipcode

PAGE 184

168Appendix A (Continued) Figure A.46 Finance, Insurance and Real Estate Employment by Zipcode

PAGE 185

169Appendix A (Continued) Figure A.47 Entertainment, Accommodation and F ood Services Employment by Zipcode

PAGE 186

170Appendix A (Continued) Figure A.48 Other Services Employment by Zipcode

PAGE 187

171Appendix A (Continued) Figure A.49 Public Administration Employment by Zipcode

PAGE 188

172Appendix A (Continued) Figure A.50 Freight Exports in Annual To ns per Employee by Zipcode

PAGE 189

173Appendix A (Continued) Figure A.51 Freight Imports in Annual Tons per Employee by Zipcode

PAGE 190

174Appendix B Table B.1 Structural Equations Model Estimation Results for Agriculture Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Total Flow -0.0025 Total 0.0018 -0.0008 -0.0007 0.0015 -0.0006 0 Direct 0.0018 -0.0008 -0.0007 0.0015 -0.0006 0 Total Truck Flow 0.0000 Total 0.0018 -0.0008 -0.0007 0.0014 -0.0006 0.9901 Direct 0.0000 0.0000 0.0000 0.0000 0.0000 0.9901 Note: N = 859,329; chi-square = 11.041 with df = 9; p-value = 0.273; CFI = 1; RMSEA = 0.001 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 191

175Appendix B (Continued) Table B.2 Structural Equations Model Estimation Re sults for Other Minerals Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Total Flow -0.0241 Total 0.0101 -0.0019 -0.0092 0.0130 -0.0048 0.0000 Direct 0.0101 -0.0019 -0.0092 0.0130 -0.0048 0.0000 Total Rail Flow 0.0000 Total 0.0101 -0.0019 -0.0092 0.0130 -0.0048 0.9991 Direct 0.0000 0.0000 0.0000 0.0000 0.0000 0.9991 Note: N = 859,329; chi-square = 5.743 with df = 9; p-value = 0.765; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 192

176Appendix B (Continued) Table B.3 Structural Equations Model Estimation Results for Food Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Rail Flow Full Truck Load Private Truck Total Flow -0.155 Total 0.047 -0.015 -0.006 0.043 -0.013 0.000 0.000 0.000 0.000 0.000 Direct 0.047 -0.015 -0.006 0.043 -0.013 0.000 0.000 0.000 0.000 0.000 Total Truck Flow -0.001 Total 0.047 -0.015 -0.005 0.043 -0.013 0.994 0.000 0.000 0.000 0.000 Direct 0.000 0.000 0.000 0.000 0.000 0.994 0.000 0.000 0.000 0.000 Total Rail Flow -0.001 Total 0.002 -0.001 0.000 0.002 -0.001 0.051 -1.835 0.000 0.000 0.000 Direct 0.000 0.000 0.001 0.000 0.000 1.876 -1.835 0.000 0.000 0.000 Full Truck Load 0.003 Total 0.036 -0.011 -0.004 0.033 -0.010 0.778 0.830 0.064 0.000 0.000 Direct -0.001 0.000 0.000 -0.001 0.000 -0.168 0.948 0.064 0.000 0.000 Private Truck 0.001 Total 0.043 -0.013 -0.005 0.039 -0.012 0.914 0.928 0.026 0.130 0.000 Direct -0.001 0.000 0.000 0.000 0.000 -0.036 0.853 0.018 0.130 0.000 Rail Car Load 0.000 Total 0.002 -0.001 0.000 0.002 -0.001 0.050 -1.568 1.010 -0.011 -0.014 Direct 0.000 0.000 0.000 0.000 0.000 -0.288 0.307 1.011 -0.009 -0.014 Note: N = 859,329; chi-square = 0.619 with df = 8; p-value = 1; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 193

177Appendix B (Continued) Table B.4 Structural Equations Model Estimation Results for Non-Durable Manufacturing Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Full Truck Load Total Flow -0.0392 Total 0.0192 -0.0074 -0.0105 0.0173 -0.0055 0.0000 0.0000 0.0000 Direct 0.0192 -0.0074 -0.0105 0.0173 -0.0055 0.0000 0.0000 0.0000 Total Truck Flow -0.0009 Total 0.0191 -0.0074 -0.0103 0.0173 -0.0055 0.9964 0.0000 0.0000 Direct 0.0000 0.0000 0.0002 0.0000 0.0000 0.9964 0.0000 0.0000 Full Truck Load 0.0172 Total 0.0047 -0.0024 -0.0089 0.0039 -0.0017 0.2454 0.2463 0.0000 Direct 0.0000 -0.0006 -0.0064 -0.0003 -0.0003 0.0000 0.2463 0.0000 Private Truck -0.0034 Total 0.0179 -0.0068 -0.0066 0.0161 -0.0049 0.9444 0.9478 -0.2302 Direct -0.0002 0.0001 0.0016 -0.0003 0.0003 0.0000 1.0045 -0.2302 Note: N = 859,329; chi-square = 7.344 with df = 9; p-value = 0.601; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 194

178Appendix B (Continued) Table B.5 Structural Equations Model Estimation Results for Lumber Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Rail Flow Full Truck Load Total Flow -0.1032 Total 0.0322 -0.0108 -0.0019 0.0260 -0.0066 0.0000 0.0000 0.0000 0.0000 Direct 0.0322 -0.0108 -0.0019 0.0260 -0.0066 0.0000 0.0000 0.0000 0.0000 Total Truck Flow -0.0002 Total 0.0320 -0.0107 -0.0019 0.0260 -0.0066 0.9932 0.0000 0.0000 0.0000 Direct 0.0001 0.0000 0.0000 0.0001 0.0000 0.9932 0.0000 0.0000 0.0000 Total Rail Flow -0.0017 Total 0.0010 -0.0003 0.0011 0.0005 0.0000 0.0380 -2.1077 0.0000 0.0000 Direct -0.0001 0.0000 0.0012 -0.0003 0.0002 2.1315 -2.1077 0.0000 0.0000 Full Truck Load 0.0013 Total 0.0264 -0.0088 -0.0014 0.0215 -0.0056 0.8380 0.8516 0.0388 0.0000 Direct -0.0006 0.0003 0.0002 -0.0004 0.0000 -0.0905 0.9333 0.0388 0.0000 Private Truck 0.0002 Total 0.0267 -0.0089 -0.0015 0.0218 -0.0056 0.8460 0.8569 0.0386 0.9202 Direct 0.0000 0.0000 -0.0001 0.0000 0.0000 -0.0042 0.0795 0.0030 0.9202 Note: N = 859,329; chi-square = 5.033 with df = 8; p-value = 0.754; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 195

179Appendix B (Continued) Table B.6 Structural Equations Model Estimation Results for Paper Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Rail Flow Full Truck Load Total Flow -0.0512 Total 0.0154 -0.0058 0.0000 0.0146 -0.0046 0.0000 0.0000 0.0000 0.0000 Direct 0.0154 -0.0058 0.0000 0.0146 -0.0046 0.0000 0.0000 0.0000 0.0000 Total Truck Flow -0.0009 Total 0.0152 -0.0057 0.0003 0.0145 -0.0046 0.9825 0.0000 0.0000 0.0000 Direct 0.0000 0.0000 0.0003 0.0002 -0.0001 0.9825 0.0000 0.0000 0.0000 Total Rail Flow -0.0027 Total 0.0010 -0.0002 0.0011 0.0008 -0.0002 0.0862 -1.2036 0.0000 0.0000 Direct -0.0003 0.0003 0.0014 -0.0003 0.0001 1.2688 -1.2036 0.0000 0.0000 Full Truck Load 0.0011 Total 0.0107 -0.0040 0.0003 0.0103 -0.0033 0.7142 0.7236 0.0620 0.0000 Direct -0.0003 0.0001 0.0000 -0.0002 0.0000 -0.0754 0.7982 0.0620 0.0000 Private Truck 0.0000 Total 0.0131 -0.0049 0.0003 0.0126 -0.0040 0.8565 0.8651 0.0276 0.3787 Direct -0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.5961 0.0041 0.3787 Note: N = 859,329; chi-square = 11.248 with df = 10; p-value = 0.339; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 196

180Appendix B (Continued) Table B.7 Structural Equations Model Estimation Results for Chemicals Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Rail Flow Full Truck Load Total Flow -0.1483 Total 0.0459 -0.0135 -0.0113 0.0431 -0.0112 0.0000 0.0000 0.0000 0.0000 Direct 0.0459 -0.0135 -0.0113 0.0431 -0.0112 0.0000 0.0000 0.0000 0.0000 Total Truck Flow -0.0105 Total 0.0449 -0.0133 -0.0073 0.0423 -0.0109 0.9643 0.0000 0.0000 0.0000 Direct 0.0006 -0.0003 0.0037 0.0008 -0.0001 0.9643 0.0000 0.0000 0.0000 Total Rail Flow 0.0080 Total 0.0027 -0.0005 -0.0109 0.0023 -0.0003 0.0960 -1.6782 0.0000 0.0000 Direct -0.0007 0.0004 -0.0036 -0.0006 0.0006 1.7142 -1.6782 0.0000 0.0000 Full Truck Load 0.0000 Total 0.0448 -0.0133 -0.0072 0.0422 -0.0109 0.9626 0.9990 0.0004 0.0000 Direct 0.0000 0.0000 0.0000 0.0000 0.0000 -0.0013 0.9996 0.0004 0.0000 Less Than Truck Load 0.0060 Total 0.0059 -0.0017 -0.0026 0.0061 -0.0017 0.1507 0.1486 0.0124 -4.6235 Direct -0.0011 0.0004 -0.0013 -0.0005 0.0000 -0.0199 4.7910 0.0140 -4.6235 Note: N = 859,329; chi-square = 9.503 with df = 7; p-value = 0.219; CFI = 1; RMSEA = 0.001 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 197

181Appendix B (Continued) Table B.8 Structural Equations Model Estimation Results for Petroleum Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Private Truck Total Flow -0.0379 Total 0.0135 -0.0048 -0.0046 0.0133 -0.0039 0.0000 0.0000 0.0000 Direct 0.0135 -0.0048 -0.0046 0.0133 -0.0039 0.0000 0.0000 0.0000 Total Truck Flow -0.0022 Total 0.0132 -0.0045 -0.0044 0.0129 -0.0036 0.9560 0.0000 0.0000 Direct 0.0003 0.0000 0.0000 0.0002 0.0001 0.9560 0.0000 0.0000 Private Truck 0.0003 Total 0.0122 -0.0042 -0.0041 0.0120 -0.0034 0.8930 0.9341 0.0000 Direct -0.0001 0.0001 0.0000 -0.0001 0.0000 0.0000 0.9341 0.0000 Water 0.0000 Total 0.0003 -0.0002 -0.0002 0.0004 -0.0003 0.0443 -1.0068 0.0155 Direct 0.0000 0.0000 0.0000 0.0000 0.0000 1.0068 -1.0212 0.0155 Note: N = 859,329; chi-square = 9.359 with df = 14; p-value = 0.807; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 198

182Appendix B (Continued) Table B.9 Structural Equations Model Estimation Re sults for Rubber Plastics Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Full Truck Load Total Flow -0.0119 Total 0.0043 -0.0019 0.0000 0.0038 -0.0015 0.0000 0.0000 Direct 0.0043 -0.0019 0.0000 0.0038 -0.0015 0.0000 0.0000 Full Truck Load 0.0009 Total 0.0024 -0.0010 -0.0002 0.0021 -0.0009 0.6149 0.0000 Direct -0.0003 0.0002 -0.0002 -0.0002 0.0001 0.6149 0.0000 Private Truck -0.0001 Total 0.0038 -0.0017 0.0000 0.0033 -0.0013 0.8901 0.3257 Direct 0.0001 0.0000 0.0000 0.0000 0.0000 0.6898 0.3257 Note: N = 859,329; chi-square = 2.162 with df = 5; p-value = 0.826; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 199

183Appendix B (Continued) Table B.10 Structural Equations Model Estimation Result s for Durable Manufacturing Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Total Flow 0.0014 Total 0.0014 -0.0007 -0.0017 0.0013 -0.0006 0.0000 Direct 0.0014 -0.0007 -0.0017 0.0013 -0.0006 0.0000 Total Truck Flow 0.9023 Total 0.0013 -0.0006 -0.0012 0.0012 -0.0005 0.9023 Direct 0.0001 0.0000 0.0003 0.0000 0.0000 0.9023 Note: N = 859,329; chi-square = 0.974 with df = 6; p-value = 0.987; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 200

184Appendix B (Continued) Table B.11 Structural Equations Model Estimation Result s for Clay, Concrete & Glass Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Full Truck Load Total Flow -0.2728 Total 0.0744 -0.0208 -0.0079 0.0723 -0.0177 0.0000 0.0000 Direct 0.0744 -0.0208 -0.0079 0.0723 -0.0177 0.0000 0.0000 Full Truck Load 0.0026 Total 0.0578 -0.0161 -0.0058 0.0563 -0.0137 0.7922 0.0000 Direct -0.0011 0.0003 0.0005 -0.0010 0.0003 0.7922 0.0000 Private Truck 0.0006 Total 0.0689 -0.0192 -0.0075 0.0672 -0.0164 0.9340 0.3037 Direct -0.0002 0.0002 -0.0002 0.0000 0.0000 0.6934 0.3037 Note: N = 859,329; chi-square = 7.252 with df = 5; p-value = 0.203; CFI = 1; RMSEA = 0.001 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 201

185Appendix B (Continued) Table B.12 Structural Equations Model Estimation Re sults for Primary Metals Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Total Flow -0.0276 Total 0.0097 -0.0038 -0.0010 0.0087 -0.0031 0.0000 Direct 0.0097 -0.0038 -0.0010 0.0087 -0.0031 0.0000 Full Truck Load -0.0002 Total 0.0096 -0.0038 -0.0010 0.0086 -0.0030 0.9842 Direct 0.0000 0.0000 0.0000 0.0000 0.0000 0.9842 Note: N = 859,329; chi-square = 1.856 with df = 7; p-value = 0.967; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 202

186Appendix B (Continued) Table B.13 Structural Equations Model Estimation Results for Fabricated Metal Products Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Truck Flow Full Truck Load Total Flow -0.0443 Total 0.0143 -0.0058 0.0000 0.0126 -0.0040 0.0000 0.0000 0.0000 Direct 0.0143 -0.0058 0.0000 0.0126 -0.0040 0.0000 0.0000 0.0000 Total Truck Flow -0.0011 Total 0.0142 -0.0058 0.0003 0.0125 -0.0040 0.9937 0.0000 0.0000 Direct 0.0000 0.0000 0.0003 0.0000 0.0000 0.9937 0.0000 0.0000 Full Truck Load 0.0008 Total 0.0112 -0.0045 0.0002 0.0100 -0.0032 0.8044 0.8095 0.0000 Direct -0.0003 0.0002 0.0000 -0.0002 0.0000 0.0000 0.8095 0.0000 Private Truck -0.0002 Total 0.0118 -0.0047 0.0003 0.0105 -0.0033 0.8409 0.8462 0.7162 Direct 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.2664 0.7162 Note: N = 859,329; chi-square = 11.399 with df = 12; p-value = 0.495; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 203

187Appendix B (Continued) Table B.14 Structural Equations Model Estimation Results for Transportation Equipment Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Total Flow -0.0161 Total 0.0063 -0.0025 -0.0009 0.0055 -0.0020 0.0000 Direct 0.0063 -0.0025 -0.0009 0.0055 -0.0020 0.0000 Total Truck Flow -0.0014 Total 0.0062 -0.0025 -0.0004 0.0053 -0.0020 0.9717 Direct 0.0001 0.0000 0.0004 0.0000 0.0000 0.9717 Note: N = 859,329; chi-square = 4.59 with df = 6; p-value = 0.597; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 204

188Appendix B (Continued) Table B.15 Structural Equations Model Estimation Result s for Miscellaneous Freight Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Rail Flow Total Flow -0.0052 Total 0.0025 -0.0012 -0.0007 0.0021 -0.0006 0.0000 0.0000 Direct 0.0025 -0.0012 -0.0007 0.0021 -0.0006 0.0000 0.0000 Total Rail Flow -0.0019 Total 0.0024 -0.0012 -0.0001 0.0020 -0.0006 0.9237 0.0000 Direct 0.0001 0.0000 0.0005 0.0001 0.0000 0.9237 0.0000 Rail Car Load -0.0002 Total 0.0022 -0.0010 -0.0002 0.0019 -0.0005 0.8372 0.9064 Direct 0.0001 0.0000 -0.0001 0.0000 0.0001 0.0000 0.9064 Note: N = 859,329; chi-square = 3.463 with df = 8; p-value = 0.902; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 205

189Appendix B (Continued) Table B.16 Structural Equations Model Estimation Results for Warehousing Commodity Group Endogenous Variable Intercept Effect Destination Employment Destination Population Distance Origin Employment Origin Population Total Flow Full Truck Load Total Flow -0.2304 Total 0.0625 -0.0157 -0.0116 0.0685 -0.0208 0.0000 0.0000 Direct 0.0625 -0.0157 -0.0116 0.0685 -0.0208 0.0000 0.0000 Full Truck Load 0.0053 Total 0.0524 -0.0131 -0.0097 0.0577 -0.0177 0.8598 0.0000 Direct -0.0014 0.0003 0.0003 -0.0012 0.0002 0.8598 0.0000 Private Truck -0.0030 Total 0.0564 -0.0140 -0.0087 0.0621 -0.0188 0.9148 0.1912 Direct -0.0005 0.0003 0.0019 -0.0004 0.0002 0.7505 0.1912 Note: N = 859,329; chi-square = 0 with df = 4; p-value = 1; CFI = 1; RMSEA = 0.000 All Variables Signifi cant at 95% level All Variables are in Logarithmic Form

PAGE 206

190Appendix C Figure C.1 Path Diagram for the Agriculture Commodi ty Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Truck Flow 1 2

PAGE 207

191Appendix C (Continued) Figure C.2 Path Diagram for the Other Minerals Comm odity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Rail Flow 1 2

PAGE 208

192Appendix C (Continued) Figure C.3 Path Diagram for the Food Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Rail flow Log of FTL Flow Log of PVT Flow Log of Rail Car Load Flow Log of Total Truck Flow 1 2 3 4 5 6

PAGE 209

193Appendix C (Continued) Figure C.4 Path Diagram for the Non-Durable Manufacturing Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of FTL Flow Log of PVT Flow Log of Total Truck Flow 1 2 4 5

PAGE 210

194Appendix C (Continued) Figure C.5 Path Diagram for the Lumber Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Rail flow Log of FTL Flow Log of PVT Flow Log of Total Truck Flow 1 2 3 4 5

PAGE 211

195Appendix C (Continued) Figure C.6 Path Diagram for the Paper Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Rail flow Log of FTL Flow Log of PVT Flow Log of Total Truck Flow 1 2 3 4 5

PAGE 212

196Appendix C (Continued) Figure C.7 Path Diagram for the Chemicals Commodi ty Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Rail flow Log of FTL Flow Log of LTL Flow Log of Total Truck Flow 1 2 3 4 5

PAGE 213

197Appendix C (Continued) Figure C.8 Path Diagram for the Petroleum Commodi ty Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of PVT flow Log of Water Flow Log of Total Truck Flow 1 2 3 4

PAGE 214

198Appendix C (Continued) Figure C.9 Path Diagram for the Rubber Plastics Commo dity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of PVT Flow Log of FTL Flow 1 2 3

PAGE 215

199Appendix C (Continued) Figure C.10 Path Diagram for the Durable Manufacturing Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Truck Flow 1 2

PAGE 216

200Appendix C (Continued) Figure C.11 Path Diagram for the Clay, Concrete & Glass Co mmodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of PVT Flow Log of FTL Flow 1 2 3

PAGE 217

201Appendix C (Continued) Figure C.12 Path Diagram for the Primary Metals Commo dity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of FTL Flow 1 2

PAGE 218

202Appendix C (Continued) Figure C.13 Path Diagram for the Fabricated Metal Products Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of FTL flow Log of PVT Flow Log of Total Truck Flow 1 2 3 4

PAGE 219

203Appendix C (Continued) Figure C.14 Path Diagram for the Transportation Equipmen t Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of Total Truck Flow 1 2

PAGE 220

204Appendix C (Continued) Figure C.15 Path Diagram for the Miscellaneous Freight Commodity Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of CL flow Log of Total Rail Flow 1 2 3

PAGE 221

205Appendix C (Continued) Figure C.16 Path Diagram for the Warehousing Commodi ty Group Structural Equations Model Log of Origin Population Log of Origin Employment Log of Destination Population Log of Destination Employment Log of Distance Log of Total Flow Log of PVT flow Log of Total Truck Flow 1 2 3

PAGE 222

206Appendix D Table D.1 Percentage Increase in Agriculture Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 2.33 4.87 7.21 9.38 11.39 13.28 15.05 16.72 18.31 19.81 Origin Employment 1.87 3.99 5.93 7.74 9.42 10.99 12.47 13.86 15.18 16.43 Distance -1.53 -2.51 -3.42 -4.25 -5.03 -5.76 -6.45 -7.10 -7.71 -8.29 Destination Population -1.69 -2.82 -3.86 -4.82 -5.72 -6.55 -7.34 -8.08 -8.78 -9.45 Origin Population -1.38 -2.23 -3.01 -3.73 -4.40 -5.03 -5.62 -6.17 -6.70 -7.20 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.2 Percentage Increase in Agriculture Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.31 2.86 4.28 5.61 6.84 7.99 9.07 10.09 11.05 11.97 Origin Employment 1.02 2.31 3.50 4.61 5.63 6.59 7.49 8.34 9.15 9.91 Distance -1.05 -1.65 -2.21 -2.72 -3.20 -3.64 -4.06 -4.46 -4.83 -5.19 Destination Population -1.15 -1.84 -2.47 -3.06 -3.61 -4.12 -4.60 -5.05 -5.48 -5.88 Origin Population -0.96 -1.48 -1.95 -2.39 -2.80 -3.19 -3.55 -3.89 -4.21 -4.51 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 223

207Appendix D (Continued) Table D.3 Percentage Increase in Other Mi nerals Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 2.77 5.23 7.49 9.59 11.54 13.37 15.10 16.72 18.25 19.71 Origin Employment 3.54 6.71 9.62 12.33 14.85 17.21 19.42 21.51 23.50 25.38 Distance -2.36 -4.58 -6.63 -8.52 -10.28 -11.92 -13.47 -14.93 -16.31 -17.61 Destination Population -0.43 -0.89 -1.32 -1.71 -2.08 -2.42 -2.74 -3.05 -3.33 -3.60 Origin Population -1.20 -2.37 -3.44 -4.44 -5.36 -6.22 -7.04 -7.80 -8.52 -9.21 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.4 Percentage Increase in Other Mi nerals Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.55 2.98 4.30 5.52 6.65 7.71 8.71 9.66 10.55 11.40 Origin Employment 2.00 3.84 5.54 7.11 8.57 9.94 11.23 12.45 13.60 14.69 Distance -1.43 -2.72 -3.92 -5.02 -6.04 -7.00 -7.91 -8.75 -9.56 -10.32 Destination Population -0.31 -0.58 -0.82 -1.05 -1.27 -1.46 -1.65 -1.83 -1.99 -2.15 Origin Population -0.76 -1.44 -2.06 -2.64 -3.17 -3.67 -4.15 -4.59 -5.01 -5.41 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 224

208Appendix D (Continued) Table D.5 Percentage Increase in Food Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.63 6.93 9.98 12.81 15.46 17.94 20.28 22.50 24.59 26.59 Origin Employment 3.32 6.34 9.13 11.72 14.13 16.40 18.54 20.56 22.47 24.29 Distance -0.43 -0.85 -1.23 -1.58 -1.91 -2.22 -2.51 -2.79 -3.04 -3.29 Destination Population -1.12 -2.17 -3.13 -4.01 -4.84 -5.61 -6.33 -7.02 -7.66 -8.27 Origin Population -0.97 -1.87 -2.71 -3.48 -4.19 -4.86 -5.49 -6.08 -6.64 -7.17 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.6 Percentage Increase in Food Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.89 3.62 5.21 6.70 8.09 9.39 10.61 11.77 12.87 13.92 Origin Employment 1.72 3.31 4.77 6.12 7.39 8.58 9.70 10.76 11.76 12.72 Distance -0.24 -0.46 -0.66 -0.85 -1.02 -1.18 -1.33 -1.48 -1.61 -1.74 Destination Population -0.60 -1.15 -1.65 -2.12 -2.55 -2.96 -3.33 -3.69 -4.03 -4.35 Origin Population -0.52 -1.00 -1.43 -1.84 -2.21 -2.56 -2.89 -3.20 -3.50 -3.77 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 225

209Appendix D (Continued) Table D.7 Percentage Increase in Non-Durable Ma nufacturing Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.58 6.83 9.84 12.62 15.21 17.64 19.93 22.09 24.13 26.07 Origin Employment 3.22 6.16 8.86 11.37 13.70 15.89 17.95 19.89 21.73 23.48 Distance -1.92 -3.68 -5.31 -6.81 -8.21 -9.51 -10.74 -11.89 -12.99 -14.02 Destination Population -1.36 -2.61 -3.76 -4.82 -5.81 -6.74 -7.61 -8.43 -9.20 -9.94 Origin Population -1.01 -1.94 -2.79 -3.58 -4.32 -5.01 -5.65 -6.26 -6.84 -7.39 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.8 Percentage Increase in Non-Durable Ma nufacturing Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.96 3.77 5.44 6.99 8.43 9.78 11.06 12.25 13.39 14.47 Origin Employment 1.77 3.40 4.90 6.29 7.59 8.81 9.95 11.03 12.06 13.03 Distance -1.09 -2.08 -2.98 -3.82 -4.60 -5.33 -6.01 -6.66 -7.26 -7.84 Destination Population -0.78 -1.48 -2.12 -2.71 -3.26 -3.77 -4.26 -4.71 -5.14 -5.55 Origin Population -0.58 -1.10 -1.58 -2.02 -2.43 -2.81 -3.17 -3.51 -3.83 -4.13 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 226

210Appendix D (Continued) Table D.9 Percentage Increase in Lumb er Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.37 6.44 9.28 11.91 14.36 16.66 18.82 20.87 22.80 24.65 Origin Employment 2.73 5.21 7.49 9.61 11.58 13.44 15.18 16.82 18.38 19.86 Distance -0.17 -0.35 -0.52 -0.67 -0.81 -0.95 -1.07 -1.19 -1.30 -1.41 Destination Population -1.10 -2.12 -3.07 -3.94 -4.75 -5.50 -6.21 -6.88 -7.52 -8.12 Origin Population -0.66 -1.29 -1.87 -2.40 -2.89 -3.36 -3.79 -4.20 -4.59 -4.96 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.10 Percentage Increase in Lumb er Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.76 3.39 4.90 6.30 7.60 8.83 9.98 11.06 12.09 13.07 Origin Employment 1.42 2.73 3.95 5.08 6.13 7.11 8.04 8.91 9.74 10.53 Distance -0.13 -0.22 -0.31 -0.39 -0.47 -0.54 -0.61 -0.67 -0.73 -0.78 Destination Population -0.62 -1.16 -1.66 -2.13 -2.56 -2.96 -3.34 -3.69 -4.03 -4.35 Origin Population -0.39 -0.72 -1.03 -1.31 -1.57 -1.82 -2.05 -2.27 -2.48 -2.67 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 227

211Appendix D (Continued) Table D.11 Percentage Increase in Pap er Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.25 6.29 9.09 11.69 14.11 16.37 18.50 20.51 22.41 24.22 Origin Employment 3.08 5.96 8.61 11.07 13.37 15.51 17.53 19.44 21.24 22.95 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -1.33 -2.47 -3.52 -4.49 -5.40 -6.24 -7.03 -7.78 -8.49 -9.16 Origin Population -1.07 -1.97 -2.81 -3.58 -4.30 -4.97 -5.60 -6.19 -6.75 -7.28 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.12 Percentage Increase in Pap er Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.67 3.19 4.60 5.90 7.11 8.25 9.32 10.33 11.28 12.19 Origin Employment 1.58 3.02 4.36 5.59 6.74 7.82 8.83 9.79 10.69 11.55 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -0.63 -1.20 -1.73 -2.22 -2.67 -3.10 -3.50 -3.87 -4.23 -4.56 Origin Population -0.50 -0.96 -1.37 -1.76 -2.12 -2.46 -2.77 -3.07 -3.35 -3.62 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 228

212Appendix D (Continued) Table D.13 Percentage Increase in Chemicals Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.39 6.52 9.40 12.09 14.60 16.95 19.16 21.26 23.25 25.14 Origin Employment 3.18 6.11 8.83 11.34 13.70 15.90 17.98 19.94 21.81 23.58 Distance -0.86 -1.62 -2.32 -2.96 -3.56 -4.13 -4.66 -5.15 -5.62 -6.07 Destination Population -1.03 -1.94 -2.78 -3.55 -4.27 -4.95 -5.58 -6.18 -6.74 -7.27 Origin Population -0.92 -1.73 -2.47 -3.16 -3.80 -4.40 -4.97 -5.50 -6.00 -6.47 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.14 Percentage Increase in Chemicals Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.84 3.53 5.10 6.55 7.91 9.18 10.38 11.51 12.58 13.61 Origin Employment 1.73 3.32 4.78 6.15 7.42 8.61 9.74 10.80 11.81 12.76 Distance -0.45 -0.87 -1.25 -1.60 -1.92 -2.23 -2.51 -2.78 -3.04 -3.28 Destination Population -0.54 -1.04 -1.49 -1.91 -2.30 -2.67 -3.01 -3.33 -3.64 -3.92 Origin Population -0.48 -0.92 -1.33 -1.70 -2.05 -2.37 -2.68 -2.96 -3.23 -3.49 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 229

213Appendix D (Continued) Table D.15 Percentage Increase in Petroleum Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.23 6.16 8.86 11.37 13.70 15.88 17.94 19.87 21.71 23.45 Origin Employment 3.18 6.07 8.73 11.20 13.50 15.65 17.67 19.58 21.39 23.10 Distance -1.07 -2.06 -2.97 -3.81 -4.60 -5.33 -6.03 -6.68 -7.29 -7.88 Destination Population -1.12 -2.16 -3.12 -4.00 -4.83 -5.60 -6.32 -7.00 -7.65 -8.26 Origin Population -0.91 -1.75 -2.53 -3.25 -3.92 -4.54 -5.13 -5.69 -6.21 -6.71 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.16 Percentage Increase in Petroleum Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.75 3.32 4.76 6.09 7.33 8.50 9.59 10.63 11.60 12.53 Origin Employment 1.73 3.27 4.69 6.00 7.23 8.37 9.45 10.47 11.43 12.35 Distance -0.54 -1.07 -1.55 -2.01 -2.43 -2.82 -3.19 -3.54 -3.87 -4.18 Destination Population -0.57 -1.12 -1.63 -2.10 -2.54 -2.95 -3.34 -3.70 -4.04 -4.37 Origin Population -0.45 -0.90 -1.32 -1.70 -2.06 -2.39 -2.70 -3.00 -3.28 -3.54 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 230

214Appendix D (Continued) Table D.17 Percentage Increase in Rubber Pl astics Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.80 7.27 10.46 13.42 16.18 18.76 21.18 23.46 25.63 27.68 Origin Employment 3.35 6.42 9.25 11.86 14.29 16.57 18.71 20.73 22.64 24.45 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -1.68 -3.22 -4.63 -5.93 -7.15 -8.28 -9.35 -10.36 -11.31 -12.21 Origin Population -1.33 -2.54 -3.65 -4.68 -5.64 -6.54 -7.38 -8.18 -8.93 -9.64 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.18 Percentage Increase in Rubber Pl astics Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.92 3.59 5.13 6.55 7.88 9.12 10.29 11.39 12.43 13.42 Origin Employment 1.71 3.18 4.54 5.80 6.98 8.07 9.10 10.08 11.00 11.87 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -0.72 -1.46 -2.14 -2.77 -3.35 -3.90 -4.41 -4.90 -5.36 -5.79 Origin Population -0.55 -1.13 -1.67 -2.17 -2.63 -3.06 -3.47 -3.85 -4.21 -4.55 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 231

215Appendix D (Continued) Table D.19 Percentage Increase in Durable Manuf acturing Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 2.29 4.41 6.35 8.15 9.83 11.39 12.87 14.25 15.57 16.81 Origin Employment 2.13 4.09 5.89 7.57 9.12 10.58 11.95 13.23 14.45 15.61 Distance -2.81 -5.36 -7.71 -9.88 -11.91 -13.80 -15.58 -17.25 -18.84 -20.35 Destination Population -1.18 -2.23 -3.20 -4.10 -4.94 -5.72 -6.46 -7.15 -7.81 -8.43 Origin Population -1.01 -1.92 -2.75 -3.52 -4.24 -4.91 -5.54 -6.13 -6.70 -7.23 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.20 Percentage Increase in Durable Manuf acturing Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.39 3.02 4.53 5.92 7.22 8.44 9.58 10.65 11.67 12.64 Origin Employment 1.26 2.78 4.18 5.47 6.68 7.81 8.86 9.86 10.81 11.70 Distance -2.57 -4.55 -6.37 -8.06 -9.63 -11.10 -12.49 -13.79 -15.02 -16.19 Destination Population -1.30 -2.12 -2.87 -3.57 -4.22 -4.82 -5.39 -5.93 -6.44 -6.92 Origin Population -1.17 -1.87 -2.52 -3.12 -3.67 -4.19 -4.68 -5.14 -5.58 -5.99 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 232

216Appendix D (Continued) Table D.21 Percentage Increase in Clay, Concret e & Glass Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.23 6.19 8.93 11.48 13.87 16.12 18.23 20.24 22.14 23.96 Origin Employment 3.14 6.02 8.68 11.16 13.47 15.65 17.71 19.66 21.51 23.27 Distance -0.32 -0.63 -0.92 -1.18 -1.43 -1.65 -1.87 -2.07 -2.26 -2.45 Destination Population -0.88 -1.70 -2.45 -3.14 -3.79 -4.39 -4.95 -5.48 -5.99 -6.46 Origin Population -0.75 -1.44 -2.08 -2.67 -3.22 -3.73 -4.22 -4.67 -5.10 -5.50 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.22 Percentage Increase in Clay, Concret e & Glass Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.85 3.54 5.11 6.57 7.93 9.22 10.43 11.58 12.67 13.71 Origin Employment 1.80 3.44 4.96 6.38 7.71 8.95 10.13 11.25 12.30 13.31 Distance -0.19 -0.37 -0.53 -0.68 -0.82 -0.95 -1.08 -1.19 -1.30 -1.41 Destination Population -0.51 -0.97 -1.40 -1.80 -2.17 -2.51 -2.84 -3.14 -3.43 -3.70 Origin Population -0.43 -0.83 -1.19 -1.53 -1.85 -2.14 -2.42 -2.67 -2.92 -3.15 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 233

217Appendix D (Continued) Table D.23 Percentage Increase in Primary Metals Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.42 6.67 9.66 12.43 15.01 17.43 19.70 21.84 23.87 25.80 Origin Employment 3.05 5.96 8.65 11.13 13.45 15.61 17.65 19.57 21.39 23.12 Distance -0.50 -0.84 -1.14 -1.43 -1.69 -1.94 -2.17 -2.39 -2.59 -2.79 Destination Population -1.53 -2.80 -3.97 -5.05 -6.06 -7.00 -7.88 -8.72 -9.51 -10.25 Origin Population -1.27 -2.31 -3.26 -4.15 -4.97 -5.74 -6.46 -7.14 -7.78 -8.39 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.24 Percentage Increase in Primary Metals Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.81 3.44 4.94 6.34 7.63 8.85 9.99 11.07 12.08 13.05 Origin Employment 1.62 3.09 4.44 5.68 6.85 7.94 8.96 9.92 10.84 11.70 Distance -0.16 -0.33 -0.48 -0.63 -0.76 -0.88 -1.00 -1.11 -1.21 -1.31 Destination Population -0.68 -1.32 -1.90 -2.45 -2.95 -3.43 -3.87 -4.29 -4.68 -5.06 Origin Population -0.55 -1.07 -1.55 -1.99 -2.40 -2.79 -3.15 -3.49 -3.82 -4.12 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 234

218Appendix D (Continued) Table D.25 Percentage Increase in Fabricated Met al Products Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.61 6.82 9.79 12.54 15.10 17.49 19.75 21.87 23.89 25.80 Origin Employment 3.19 6.02 8.63 11.05 13.31 15.42 17.40 19.27 21.05 22.73 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -1.34 -2.64 -3.84 -4.95 -5.98 -6.94 -7.85 -8.70 -9.51 -10.27 Origin Population -0.90 -1.80 -2.62 -3.39 -4.10 -4.76 -5.39 -5.98 -6.53 -7.06 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.26 Percentage Increase in Fabricated Met al Products Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.69 3.30 4.77 6.14 7.41 8.61 9.73 10.79 11.79 12.74 Origin Employment 1.49 2.90 4.20 5.40 6.52 7.57 8.56 9.49 10.37 11.21 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -0.77 -1.42 -2.01 -2.56 -3.08 -3.56 -4.01 -4.43 -4.83 -5.21 Origin Population -0.55 -0.99 -1.41 -1.79 -2.14 -2.47 -2.78 -3.08 -3.35 -3.62 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 235

219Appendix D (Continued) Table D.27 Percentage Increase in Transportation Equipment Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.26 6.25 9.00 11.54 13.92 16.14 18.22 20.19 22.06 23.82 Origin Employment 2.84 5.45 7.85 10.07 12.15 14.08 15.90 17.62 19.25 20.79 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -1.31 -2.50 -3.59 -4.60 -5.53 -6.41 -7.24 -8.01 -8.75 -9.45 Origin Population -1.05 -2.00 -2.87 -3.68 -4.43 -5.13 -5.79 -6.41 -7.00 -7.56 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.28 Percentage Increase in Transportation Equipment Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.62 3.20 4.65 6.00 7.25 8.43 9.53 10.57 11.55 12.49 Origin Employment 1.40 2.78 4.05 5.22 6.32 7.34 8.30 9.21 10.07 10.88 Distance 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Destination Population -0.79 -1.42 -1.99 -2.53 -3.02 -3.49 -3.92 -4.33 -4.72 -5.09 Origin Population -0.66 -1.16 -1.62 -2.04 -2.44 -2.81 -3.16 -3.49 -3.80 -4.09 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 236

220Appendix D (Continued) Table D.29 Percentage Increase in Miscellaneou s Freight Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.50 6.43 9.12 11.61 13.93 16.10 18.13 20.06 21.87 23.60 Origin Employment 2.99 5.45 7.71 9.80 11.74 13.57 15.28 16.89 18.42 19.87 Distance -0.59 97.00 -2.15 -2.84 -3.49 -4.09 -4.66 -5.20 -5.70 -6.18 Destination Population -1.23 -2.64 -3.93 -5.12 -6.23 -7.27 -8.25 -9.17 -10.04 -10.87 Origin Population -0.47 -1.17 -1.81 -2.41 -2.96 -3.48 -3.97 -4.43 -4.87 -5.28 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.30 Percentage Increase in Miscellaneou s Freight Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.42 3.03 4.51 5.88 7.16 8.35 9.48 10.54 11.54 12.49 Origin Employment 1.14 2.49 3.73 4.89 5.96 6.96 7.91 8.79 9.64 10.43 Distance -0.83 -1.28 -1.70 -2.08 -2.44 -2.77 -3.09 -3.38 -3.66 -3.93 Destination Population -1.19 -1.96 -2.67 -3.33 -3.94 -4.52 -5.05 -5.56 -6.04 -6.50 Origin Population -0.77 -1.15 -1.51 -1.84 -2.14 -2.43 -2.70 -2.95 -3.19 -3.42 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476

PAGE 237

221Appendix D (Continued) Table D.31 Percentage Increase in Warehou sing Flow (Base Case I) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 3.16 6.06 8.74 11.23 13.57 15.76 17.82 19.77 21.63 23.40 Origin Employment 3.47 6.65 9.59 12.32 14.89 17.29 19.56 21.71 23.75 25.69 Distance -0.57 97.00 -1.59 -2.04 -2.46 -2.85 -3.22 -3.56 -3.89 -4.20 Destination Population -0.78 -1.50 -2.16 -2.77 -3.34 -3.87 -4.37 -4.84 -5.29 -5.71 Origin Population -1.04 -1.99 -2.87 -3.68 -4.43 -5.13 -5.79 -6.41 -7.00 -7.55 Base Case: Destination Employment = 6739, Origin Employment = 6739, Distance = 153 miles, Destination Population = 17238, Origin Population = 17238 Table D.32 Percentage Increase in Warehou sing Flow (Base Case II) Percentage Increase in the Explanatory Variable Explanatory Variable 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Destination Employment 1.75 3.37 4.87 6.26 7.56 8.78 9.94 11.03 12.06 13.05 Origin Employment 1.92 3.70 5.34 6.87 8.30 9.64 10.91 12.11 13.25 14.33 Distance -0.33 -0.63 -0.90 -1.15 -1.39 -1.61 -1.81 -2.01 -2.19 -2.36 Destination Population -0.45 -0.85 -1.22 -1.56 -1.88 -2.17 -2.45 -2.72 -2.96 -3.20 Origin Population -0.59 -1.12 -1.61 -2.06 -2.48 -2.87 -3.24 -3.59 -3.92 -4.23 Base Case: Destination Employment = 53604, Origin Employment = 53604, Distance = 509 miles, Destination Population = 74476, Origin Population = 74476


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001469376
003 fts
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 040524s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000278
035
(OCoLC)55644090
9
AJR1130
b SE
SFE0000278
040
FHM
c FHM
090
TA145
1 100
Jonnavithula, Siva S.
0 245
Development of structural equations models of statewide freight flows
h [electronic resource] /
by Siva S. Jonnavithula.
260
[Tampa, Fla.] :
University of South Florida,
2004.
502
Thesis (M.S.C.E.)--University of South Florida, 2004.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 237 pages.
520
ABSTRACT: The modeling of freight travel demand has gained increasing attention in the recent past due to the importance of efficient and safe freight transportation to regional economic growth. Despite the attention paid to the modeling of freight travel demand, advances in modeling methods and the development of practical tools for forecasting freight flows have been limited. The development of freight demand models that incorporate the behavioral aspects of freight demand face significant hurdles, partially due to the data requirements, which are a consequence of the inherent complexity of the mechanisms driving freight demand. This research attempts to make a contribution in this context by proposing a relatively data simple, but behaviorally robust statewide modeling framework for the state of Florida, in the spirit of an aggregate level four-step planning process. The modeling framework that is developed in this research can be applied to the modeling of freight travel demand using data contained in readily available commercial databases such as the Reebie TRANSEARCH database and the InfoUSA employer database. The modeling methodology consists of a structural equations modeling framework that can accommodate multiple dependent variables simultaneously. This framework predicts freight flows on various modes between two zipcodes based on the socio-economic characteristics and the modal level of service characteristics. Separate models have been developed for various commodity groups. The estimated models for various commodity groups are found to offer statistically valid indications and plausible interpretations suggesting that these models may be suitable for application in freight transportation demand forecasting applications. The sensitivity analysis conducted on these models clearly added evidence to the fact that employment is the key factor influencing freight flows between two regions.
590
Adviser: Pendyala, Ram M.
653
sensitivity analysis.
ADF-WLS estimation.
employment.
population.
commodity groups.
data requirements.
690
Dissertations, Academic
z USF
x Civil Engineering
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.278