<%BANNER%>

Impact Analysis of Site Development and Mileage Fee

Permanent Link: http://ufdc.ufl.edu/UFE0044466/00001

Material Information

Title: Impact Analysis of Site Development and Mileage Fee
Physical Description: 1 online resource (129 p.)
Language: english
Creator: Mamun, Md Shahid
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: average -- distribution -- fees -- generator -- link -- meue -- mileage -- nonuniqueness -- solution -- special -- tia -- vmt
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre: Civil Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation focuses on two types of impact studies: one is related to site developments; the other examines the socioeconomic effects of mileage fees. More specifically, in the first part, the link distribution percentage and special generator methods for performing traffic impact analysis are compared and enhanced; and in the second part, the impacts of adopting a mileage fee in Florida are assessed. In Florida, both the link distribution percentage and special generator methods are used to conduct traffic impact analysis. However, there is no systematic research to show whether these two methods produce similar results or if one outperforms the other. This dissertation describes an empirical study that compares these two methods. Based on the study, these two methods are observed to produce fairly consistent estimates of traffic impacts caused by the chosen hypothetical scenarios. As the link distribution percentage approach is easier to implement,this dissertation recommends this less cumbersome approach. However, both of the above mentioned approaches estimate development trips on each link from the path flow or origin-destination (O-D) specific link flow distribution. Since these two flow distributions may not be uniquely determined, an open question remains regarding the selection of a particular flow distribution as the basis for traffic impact studies. This dissertation suggests using the mean of all the path or O-D specific user equilibrium solutions as the basis for traffic impact studies. The second part of the dissertation examines the impacts of implementing mileage fees in Florida. Four different mileage fee structures are tested. The result shows that the distributional impacts of the revenue-neutral fee are negligible. However, flat fees are found to be regressive at higher rates. In contrast, step fee, a two-level tariff structure is found to be less regressive. Fees based on vehicle fuel efficiency and vehicle type are found to be environmentally friendly, but areas regressive as flat fees. This dissertation suggests that a complex mileage fee structure is needed to balance the spatial distribution of the impacts,reduce the regressive nature of the fee, generate sufficient revenue, protect the environment, and achieve other objectives simultaneously.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Md Shahid Mamun.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Yin, Yafeng.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044466:00001

Permanent Link: http://ufdc.ufl.edu/UFE0044466/00001

Material Information

Title: Impact Analysis of Site Development and Mileage Fee
Physical Description: 1 online resource (129 p.)
Language: english
Creator: Mamun, Md Shahid
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: average -- distribution -- fees -- generator -- link -- meue -- mileage -- nonuniqueness -- solution -- special -- tia -- vmt
Civil and Coastal Engineering -- Dissertations, Academic -- UF
Genre: Civil Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: This dissertation focuses on two types of impact studies: one is related to site developments; the other examines the socioeconomic effects of mileage fees. More specifically, in the first part, the link distribution percentage and special generator methods for performing traffic impact analysis are compared and enhanced; and in the second part, the impacts of adopting a mileage fee in Florida are assessed. In Florida, both the link distribution percentage and special generator methods are used to conduct traffic impact analysis. However, there is no systematic research to show whether these two methods produce similar results or if one outperforms the other. This dissertation describes an empirical study that compares these two methods. Based on the study, these two methods are observed to produce fairly consistent estimates of traffic impacts caused by the chosen hypothetical scenarios. As the link distribution percentage approach is easier to implement,this dissertation recommends this less cumbersome approach. However, both of the above mentioned approaches estimate development trips on each link from the path flow or origin-destination (O-D) specific link flow distribution. Since these two flow distributions may not be uniquely determined, an open question remains regarding the selection of a particular flow distribution as the basis for traffic impact studies. This dissertation suggests using the mean of all the path or O-D specific user equilibrium solutions as the basis for traffic impact studies. The second part of the dissertation examines the impacts of implementing mileage fees in Florida. Four different mileage fee structures are tested. The result shows that the distributional impacts of the revenue-neutral fee are negligible. However, flat fees are found to be regressive at higher rates. In contrast, step fee, a two-level tariff structure is found to be less regressive. Fees based on vehicle fuel efficiency and vehicle type are found to be environmentally friendly, but areas regressive as flat fees. This dissertation suggests that a complex mileage fee structure is needed to balance the spatial distribution of the impacts,reduce the regressive nature of the fee, generate sufficient revenue, protect the environment, and achieve other objectives simultaneously.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Md Shahid Mamun.
Thesis: Thesis (Ph.D.)--University of Florida, 2012.
Local: Adviser: Yin, Yafeng.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0044466:00001


This item has the following downloads:


Full Text

PAGE 1

1 IMPACT ANALYSIS OF SITE DEVELOPMENT AND MILEAGE FEE By MD SHAHID MAMUN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Md Shahid Mamun

PAGE 3

3 To my parents

PAGE 4

4 ACKNOWLEDGMENTS I would like to express my sincere gratitude to my advisor Dr. Yafeng Yin for his guidance, constant inspiration and invaluable suggestions thro ughout the course of my Ph D study. He has been a great teacher, advisor and friend. Without his support and encouragement, this dissertation would never have been finished. I am also extremely grateful to Dr. Sivaramakrishnan Srinivasan and Dr. Siriphon g Lawphongpanich for their guidance and helpful suggestions during the various stages of this research. I would also like to thank my other committee members Dr. Lily Elefteriadou and Dr. Scott Washburn for their valuable comments, feedback and suggestions I would like to thank the University of Florida for providing me such a quality education and fine research facilities. Additional thanks are given to Nancy Mcllrath, Ines Aviles Spadoni and Tony Murphy My special thanks are given to all my friends in t he Transporta t ion Research Center e specially Hongli Xu, Dr. Ashish Kulshrestha Dr. Di Wu Dr. Lihui Zhang Dr. Ziqi Song Dimitra Michalaka and Vipul Modi Also, I greatly appreciate the efforts of Dr. Kathryn Williams for carefully editing the manuscrip t and making suggestions for improvem e nts. Finally, I would like to thank my parents, brothers and wife for their continuous support inspiration and sacrifice

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 1.1 Background ................................ ................................ ................................ ... 15 1.2 Research Objectives ................................ ................................ ..................... 19 1.3 Dissertation Outline ................................ ................................ ....................... 20 2 LITERATURE REVIEW ................................ ................................ .......................... 21 2.1 Impact Analysis of Site Development ................................ ............................ 21 2.1.1 General Procedure for Traffic Impact Analysis ................................ .... 22 2.1.2 Methods Used by FDOT ................................ ................................ ..... 25 2.1.2.1 Man ual method ................................ ................................ ............. 26 2.1.2.2 Travel demand model method ................................ ...................... 26 2.1.3 Select Zone Analysis ................................ ................................ ........... 2 8 2.1.4 Summary ................................ ................................ ............................. 30 2.2 Impact Analysis of Mileage Fees ................................ ................................ ... 31 2.2.1 Need for Mileage Fees ................................ ................................ ........ 31 2.2.2 Pilot Studies on Mileage Fees ................................ ............................. 35 2.2.3 Impact Studies on Mileage Fees ................................ ......................... 38 2.2.4 Other Studies on Mileage Fees ................................ ........................... 40 2.2.5 Summary ................................ ................................ ............................. 41 3 COMPARISON OF TRAFFIC IMPACT ANALYSIS METHODS ............................. 43 3.1 Background ................................ ................................ ................................ ... 43 3.2 Qualitative Comparison of the Two Methods ................................ ................. 43 3.3 Empirical Study ................................ ................................ ............................. 44 3.3.1 Study Site ................................ ................................ ............................ 44 3.3.2 The Alachua/Gainesville Model ................................ ........................... 46 3.3.3 Implementation of the Two Methods ................................ ................... 47 3.4 Results ................................ ................................ ................................ .......... 49 3.5 Summary ................................ ................................ ................................ ....... 62 4 ENHANCING SELECT ZONE ANALYSIS ................................ .............................. 64

PAGE 6

6 4.1 Background ................................ ................................ ................................ ... 64 4.2 Impact of Nonuniqueness on Select Zone Analysis ................................ ...... 65 4.3 Entropy Maximiz ing Flow Distribution ................................ ........................... 68 4.4 Average Flow Distribution ................................ ................................ ............. 72 4.4.1 Definition ................................ ................................ ............................. 72 4.4.2 Stability ................................ ................................ ............................... 73 4.4.3 Computation ................................ ................................ ........................ 74 4.5 Numerical Example ................................ ................................ ....................... 76 4.6 Summary ................................ ................................ ................................ ....... 80 5 IMPACT ASSESSMENT OF MILEAGE BASED USER FEES IN FLORIDA .......... 82 5.1 Background ................................ ................................ ................................ ... 82 5.2 Data ................................ ................................ ................................ .............. 82 5.2.1 Description of the Data ................................ ................................ ........ 82 5.2.2 Descriptive Statistics of the Data ................................ ........................ 83 5.3 Methodology ................................ ................................ ................................ .. 86 5.3.1 Model Estimation ................................ ................................ ................. 86 5.3.2 Socioeconomic Measures ................................ ................................ ... 88 5.4 Impact Analysis ................................ ................................ ............................. 89 5.4.1 Flat Mileage Fee ................................ ................................ ................. 89 5.4.2 Step Mileage Fee ................................ ................................ .............. 103 5.4.3 Mileage Fee Based on Fuel Efficiency ................................ .............. 108 5.4.4 Mileage Fee Based on Vehicle Type ................................ ................. 113 5. 5 Summary ................................ ................................ ................................ ..... 115 6 CONCLUSIONS ................................ ................................ ................................ ... 119 APPENDIX A NETWORK CHARACTERISTICS OF SIOUX FALLS NETWORK ....................... 123 B O D DEMANDS OF SIOUX FALLS NETWORK ................................ ................... 124 LIST OF REFERENCES ................................ ................................ ............................. 125 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 129

PAGE 7

7 LIST OF TABLES Table page 3 1 Input data for hypothetical development scenarios ................................ ............. 47 3 2 Current situation in TAZs 225 a nd 148 ................................ ............................... 48 3 3 Trip rates recommended by FDOT ................................ ................................ ..... 48 3 4 Adjustment for special generator ................................ ................................ ........ 49 3 5 Summary of trip generation (TAZ 225) ................................ ............................... 50 3 6 Summary of trip generation (TAZ 148) ................................ ............................... 50 3 7 Link volumes of top 10 link s (TAZ 225 with scenario 1) ................................ ...... 56 3 8 Link volumes of top 10 links (TAZ 148 with scenario 1) ................................ ...... 56 3 9 Link volumes of top 10 links (TAZ 225 with scenario 8) ................................ ...... 57 3 10 Link volumes of top 10 links (TAZ 148 with scenario 8) ................................ ...... 57 3 11 Sum of total development trips on each link ( TAZ 225) ................................ ...... 57 3 12 Sum of total development trips on each link (TAZ 148) ................................ ...... 58 3 13 Comparison of two methods (TAZ 225) ................................ .............................. 59 3 14 Comparison of two methods (TAZ 148) ................................ .............................. 59 3 15 Link percentages of top 10 links (TAZ 225) ................................ ........................ 61 3 16 Link percentages of top 10 links (TAZ 148) ................................ ........................ 61 3 17 Variations of link distribution percentages of different scenarios ........................ 61 4 1 Link uses of node 1 ................................ ................................ ............................ 68 4 2 Comparisons of link uses of node 1 ................................ ................................ .... 77 4 3 Comparisons of link uses of node 10 ................................ ................................ .. 79 5 1 Descriptive statistics by location ................................ ................................ ......... 84 5 2 Descriptive statistics by income group ................................ ................................ 84 5 3 Perce nt of vehicles by type and location ................................ ............................ 84

PAGE 8

8 5 4 Percent of vehicles by type and income group ................................ ................... 84 5 5 Descriptive statistics by county ................................ ................................ ........... 85 5 6 Estimated model ................................ ................................ ................................ 88 5 7 Elasticity by income group based on average income ................................ ........ 88 5 8 Changes in consumer surplus, revenue, social welfare and VMT under different mileage fees ................................ ................................ ......................... 90 5 9 Average changes in consumer surplus, revenue and social welfare by income group ................................ ................................ ................................ ...... 91 5 10 Average changes in consumer surplus, revenue and social welfare by location ................................ ................................ ................................ ............... 91 5 11 Average changes in consumer surplus, revenue and so cial welfare by county .. 92 5 12 Average changes in consumer surplus, revenue and social welfare by income group ................................ ................................ ................................ ...... 95 5 13 Average changes in consumer surplus, revenue and social welfare by location ................................ ................................ ................................ ............... 95 5 14 Average changes in consumer surplus, revenue and social welfare by county .. 96 5 15 Average changes in consumer surplus, revenue and social welfare by income group ................................ ................................ ................................ ...... 99 5 16 Average changes in consumer surplus, revenue and social welfare by location ................................ ................................ ................................ ............... 99 5 17 Average changes in consumer surplus, revenue and social welfare by county 100 5 18 Total change in consumer surplus, revenue, social welfare and VMT under different mileage fees ................................ ................................ ....................... 104 5 19 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 1) ................................ ................................ ................ 104 5 20 Average changes in consumer surplus, revenue and social welfare by location (Scheme 1) ................................ ................................ .......................... 104 5 21 Average changes in consumer surplus, revenue and social welfare by income group (Schem e 2) ................................ ................................ ................ 105 5 22 Average changes in consumer surplus, revenue and social welfare by location (Scheme 2) ................................ ................................ .......................... 105

PAGE 9

9 5 23 Average changes in consu mer surplus, revenue and social welfare by income group (Scheme 3) ................................ ................................ ................ 105 5 24 Average changes in consumer surplus, revenue and social welfare by location (Scheme 3) ................................ ................................ .......................... 105 5 25 Average changes in consumer surplus as a percent of average income by income group ................................ ................................ ................................ .... 106 5 26 Total change in consumer surplus, revenue, social welfare and VMT under different mileage fees ................................ ................................ ....................... 106 5 27 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 4) ................................ ................................ ................ 107 5 28 Average changes in consumer surplus, revenue and social welfare by location (Scheme 4) ................................ ................................ .......................... 107 5 29 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 5) ................................ ................................ ................ 107 5 30 Average changes in consumer surplus, revenue and social welfare by location (Scheme 5) ................................ ................................ .......................... 107 5 31 Average changes in consumer surp lus, revenue and social welfare by income group (Scheme 6) ................................ ................................ ................ 108 5 32 Average changes in consumer surplus, revenue and social welfare by location (scheme 6) ................................ ................................ .......................... 108 5 33 Average changes in consumer surplus as a percent of average income by income group ................................ ................................ ................................ .... 108 5 34 Total change in consumer surplus, revenue, social welfare and VMT under d ifferent mileage fee ................................ ................................ ......................... 110 5 35 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 1) ................................ ................................ ................ 110 5 36 Aver age changes in consumer surplus, revenue and social welfare by location (Scheme 1) ................................ ................................ .......................... 110 5 37 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 2) ................................ ................................ ................ 110 5 38 Average changes in consumer surplus, revenue and social welfare by location (Scheme 2) ................................ ................................ .......................... 111 5 39 Average gasoline consumption by income gr oup ................................ ............. 111

PAGE 10

10 5 40 Total change in consumer surplus, revenue, social welfare and VMT under different mileage fee ................................ ................................ ......................... 111 5 41 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 3) ................................ ................................ ................ 112 5 42 Average changes in consumer surplus, revenue and social welfare by location (Scheme 3) ................................ ................................ .......................... 112 5 43 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 4) ................................ ................................ ................ 112 5 44 Average changes in consumer surplus, revenue and so cial welfare by location (Scheme 4) ................................ ................................ .......................... 112 5 45 Average gasoline consumption by income group ................................ ............. 113 5 46 Multiplying factors for different v ehicle type ................................ ...................... 114 5 47 Total change in consumer surplus, revenue, social welfare and gasoline consumption under different mileage fee ................................ .......................... 114 5 48 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 1) ................................ ................................ ................ 114 5 49 Average changes in consumer surplus, revenue and social welfare by location (Scheme 1) ................................ ................................ .......................... 115 5 50 Average changes in consumer surplus, revenue and social welfare by income group (Scheme 2) ................................ ................................ ................ 115 5 51 Average changes in consume r surplus, revenue and social welfare by location (Scheme 2) ................................ ................................ .......................... 115

PAGE 11

11 LIST OF FIGURES Figure page 2 1 Typical basic framework for traffic impact analysis ................................ ............. 23 3 1 Map of the study site ................................ ................................ .......................... 45 3 2 TAZ configuration in the Alachua/Gainesville MPO model ................................ 46 3 3 Flow distribution without new development ................................ ........................ 51 3 4 Flow distribution with development scenario 8 in TAZ 225 before special generator adjustments ................................ ................................ ........................ 51 3 5 Flow distribution with development scenario 8 in TAZ 225 after special generator adjustments ................................ ................................ ........................ 52 3 6 Flow distribution with development scenario 8 in TAZ 148 before special generator adjustments ................................ ................................ ........................ 52 3 7 Flow distribution with development scenario 8 in TAZ 148 after special generator adjustments ................................ ................................ ........................ 53 3 8 Traffic volumes from TAZ 225 with development scenario 8 before special generator adjustments ................................ ................................ ........................ 53 3 9 Traffic volumes from TAZ 225 with development scenario 8 after special generator ad justments ................................ ................................ ........................ 54 3 10 Traffic volumes from TAZ 148 with development scenario 8 before special generator adjustments ................................ ................................ ........................ 54 3 11 Traffic volume s from TAZ 148 with development scenario 8 after special generator adjustments ................................ ................................ ........................ 55 3 12 RMSE of development link flows from two methods (TAZ 225) .......................... 59 3 13 RMSE of development link flows from two methods (TAZ 148) .......................... 60 3 14 Variations of link distribution percentages of different scenarios ........................ 62 4 1 The Nine node network ................................ ................................ ...................... 67 4 2 Sioux Falls network ................................ ................................ ............................ 78 5 1 Spatial distribution of impacts of mileage fee of 1.6 1 cents/mile ......................... 94 5 2 Spatial distribution of impacts of mileage fee of 2.8 cents/mile ........................... 98

PAGE 12

12 5 3 Spatial distribution of impacts of mileage fee of 4.1 cents/mile ......................... 102

PAGE 13

13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy IMPACT ANAL YSIS OF SITE DEVELOPMENT AND MILEAGE FEE By Md Shahid Mamun August 2012 Chair: Yafeng Yin Major: Civil Engineering This dissertation focuses on two types of impact studies : one is related to site developments ; the other examines the socioeconomic effec ts of mileage fee s More specifically, in the first part, the link distribution percentage a nd special generator method s for performing traffic impact analysis are compared and enhanced; and in the second part, the impacts of adopting a mileage fee in Flor ida are assessed In Florida, b oth the link distribution percentage and special generator methods are used to conduct traffic impact analysis However, there is no systematic research to show whether these two methods produce similar results or if one outp erform s the other. This dissertation describes an empirical study that compare s these two methods Based on the study, t hese two methods are observed to produce fairly consistent estimates of traffic impacts caused by the chosen hypothetical scenarios. As the link distribution percentage approach is easier to implement this dissertation recommends this less cumbersome approach However, both o f the above mentioned approaches estimate development trips on each link from the path flow or origin destination ( O D) specific link flow distribution Since these two flow distributions may not be uniquely determined, an open question

PAGE 14

14 remains regarding the selection of a particular flow distribution as the basis for traffic impact studies T his dissertation suggests using the mean of all the path or O D specific user equilibrium solutions as the basis for traffic impact studies T he second part of the dissertation examines the impacts of implementing mileage fees in Florida. F our different mileage fee structures are t ested The result shows that the distributional impacts of the revenue neutral fee are negligible. However, flat fees are found to be regressive at higher rates. In contrast, s tep fee, a two level tariff structure is found to be less regressive. F ees based on vehicle fuel efficiency and vehicle type are found to be environment ally friendly but are as regressive as flat fees. Th is dissertation suggest s that a complex mileage fee structure is needed to balance the spatial distribution of the impacts, reduce the regressive nature of the fee, generate sufficient revenue, protect the environment and achieve other objectives simultaneously.

PAGE 15

15 CHAPTER 1 IN TRODUCTION 1.1 B ackground Good road networks are essential for any country, especially for the United States where residents are heavily dependent on automobiles. However, providing and maintaining a high standard transportation facility is not an easy task. Many factors can affect travel demand, transportation supply and the performance of road networks As an e xample, new development can generate additional travel demand s negatively affecting the level of service on the existing transportation networks. On the other hand highway improvement projects are costly and related fee structure changes ( e. g. change s i n the gasoline tax and in vehicle registration fee s implemen ta t ion of vehicle mile s travel ed fee s etc.) can imp ose substantial socioeconomic impacts. Since the government or concern ed authority must clearly understand all these ramifications in o rder to take appropriate action, impact studies are essential f or the decision making process. T his dissertation involves two types of impact studies : o ne examine s the influence of site development on vehicular traffic patterns; the other assesses the socioeconomi c effects of mileage fee s on Florida residents. A n ew site development generates additional travel demand s that may stress the adequacy of existing road networks. U nder the concurrency law of Florida (s. 163.3180, FS), a local government cannot approve a new development unless there will be sufficient transportation facilities to serve the traffic created as a result The law requires that the level of service (LOS) of the transportation system should not decline when the new development becomes active If the new development will produce a significant increase in the amount of traffic and impair s the performance of the affected

PAGE 16

16 transportation infrastructures, the developer is responsible for any improvement ( e.g., additional lane (s) n ew signal ( s ) and othe r transportation facilities) necessary to maintain the original LOS. Therefore, traffic impact analysis (TIA) is required to assess the impact s of the traffic generated from the new development according to the guideline s specified by the concern ed s tate a nd local government. The Institute of Transportation Engineers (ITE) has provided general guidelines for performing TIA for a ny site development (ITE, 2010), but usually s tate and local governments have more specific requirements and procedures. In Florid a the D epartment of T ransportation has its own site impact handbook (FDOT, 1997) for performing TIA in the state. According to the handbook, the TIA may be performed either via a manual technique or by using a travel demand model (FDOT, 1997). In the form er, the total number of trips generated from the proposed development is Generation report (ITE, 2008) and the size of the new development. Then trip distribution a nd assignment are performed to obtain the development trips on each link. One the other hand, when a travel demand model is used, only the required input variables (e.g., number of dwelling units, number of employees, size of the shopping center) are neede d as input, and the development traffic on each link is obtained from the model output. Using a travel demand model minimizes the potential bias from a manual procedure and gives better consideration of the level of growth in the region and potential impr ovements to the transportation network. However the number of development trips one of the major requireme nts for TIA by many authorities, generated from the model often does not match the number calculated manually using

PAGE 17

17 As a remedy, two methods are practiced in Florida the link distribution percentage and the special generator approach es In the former, the numbers of dwelling units and employ ees of a new development are estimated and inserted into the trip generation i nput file, followed by a travel demand model run to derive the development traffic percentage for each link in the impact area. The percentages are then applied to the external ITE based trip generation for the new development to quantify its traffic impac t. In the special generator approach, the new development is treated as a special generator whose attraction and possible production are manually calculated based on the ITE Trip Generation report. These are then further adjusted using an iterative method until the trips reported in the model match those in the ITE based trip generation. Traffic assignment is subsequently conducted to quantify the impact of the proposed development on the traffic network. Both approaches have pros and cons, and there has b een no systematic research to show whether the two approaches generate similar results or if one outperforms the other. There is also another problem in using the travel demand model. Irrespective of the method, commercial software uses the S elect Z one A module to obtain the development trips (trips to/from the development site) on each link. During the model run, the path flow information is stored and the development trips are determined from this path flow However, it is well known that the p ath flow distributions may not be uniquely determined from the deterministic user equilibrium (UE) assignment even though the formulation has a unique aggregate link flow solution under mild conditions (Smith, 1979). Despite the occurrence of multiple sol utions the path flow distribution

PAGE 18

18 obtained from the select zone analysis is commonly used in practice to conduct traffic impact analysis Consequently, the results may only represent one out of a number of possible spatial impacts. TIA is a practical too l to assess the impact s from a new development on the transportation network and to determine the impact fees charged to the developers. But based on the current practice the impact results may be different for different methods and different software, ma king it difficult for the concern ed authority to defend its analysis. Therefore, one aim of this dissertation is to assess and enhance the current TIA methodology. Another area of concern is the possible implementation of mileage fees, especially in the s tate of Florida. The m ileage fee concept has gained considerable attention among federal and state governments, because the revenue collected from the gasoline tax is not sufficient to meet transportation funding need in the USA. Due to the unadjusted gaso line tax rate, increasing construction cost s, and the introduction of more fuel efficient and hybrid vehicles, the gap between the revenue generated from gas oline tax es and the revenue required for transportation investment is increasing. According to the National Surface Transportation Infrastructure Financing Commission (NSTIFC) formed by the C ongress, the total funding gap will be $2.3 trillion for the period 2010 to 2035 (NSTIFC, 2009). Several federal and state studies (TRB, 2006; NSTPRSC, 2007; NSTIF C, 2009; NCF 2005; NCHRP, 2006) ha ve recommend ed a user fee, more specifically a mileage fee or vehicle mile s travel ed (VMT) fee as an alternative to the gasoline tax for long term transportation needs Pilot program have demonstrated the technical feasibility and public acceptability of

PAGE 19

19 such fees as the fuel tax replacement However, equity issue s remain as a matter of concern for actual implemen ta ti on of a mileage fee in real life. The new fee should be equitable a mong group s who have different income s and reside in different location s A lthough a f ew stud ies ( e.g., O DOT, 2008 ) ha ve been performed to assess the socioeconomic impact s of such a fee no impact study has been conducted for Florida and results from othe r states may not be applicable. Given that Florida is one of the s t ates that participated in the mileage fee pilot program and has a strong interest in the mileage fee system, the second aim of this dissertation i s the assessment of the effects of mileage fee s on drivers in Florida. 1.2 Research Objectives The objectives of this dissertation are twofold : (1) to examine the methodology of the traffic impact analys i s for new site development s ; (2) to assess the socioeconomic impact s of implementing statewide mileage fee s on residents in Florida More details of the objectives are described in the following paragraphs. Impact s of site development s In this part we assess and enhance the procedure s of traffic impact analysis for new site development s In parti cular, we perform the following two tasks: (i ) c ompare two modeling approaches, i.e., the link distribution percentage method and the special generator method, for performing traffic impact analyses for pr oposed developments in Florida; (ii ) p rovid e specif ic paths or O D specific user equilibrium solutions that can be used as the basis for conducting traffic impact analysis. Impact s of mileage fee s In this part we assess the socioeconomic impact s of implementing mileage fee s in Florida. Four mileage fee structures (flat fees, step fees, fees based on vehicle fuel efficiency and fees based on vehicle type) are examined

PAGE 20

20 The income and spatial equity effects are assessed for different mileage fee scenarios in Florida. 1.3 Dissertation Outline This disserta tion is organized as follows. Chapter 2 reviews the existing literature on traffic impact analysis and the implementation and assessment of mileage fee s Chapter 3 compares two methods of performing TIA : the link distribution percentage and the special gen erator approaches Enhancement of the TIA technique is described in Chapter 4. Chapter 5 presents the impact a ssessment s of mileage fee s in Florida Finally, conclusions and recommendation s are provided in Chapter 6.

PAGE 21

21 CHAPTER 2 LITERATURE REVIEW Traffic i mpact studies are very importa nt in transportation sectors. Without the proper understanding and techniques it is very difficult to assess the impact s of changes and to improve the system. T his dissertation focus es on two impacts on transportation systems : site development and the socioeconomic impacts of mileage fee implementation This chapter reviews the current Traffic Impact Analysis ( TIA ) procedure s and literature related to site impact s and mileage fee s 2.1 Impact Analysis of Site Development Ever y new development generates additional traffic which o ften creates congestion and requires improvement s in the existing transportation facilities. For that reason traffic impact analys e s are performed to assess the effects of site development and to miti gate the negative impacts. Traffic impact analysis which is also known as a site impact stud y, is defined as follows: A ny effort by the Department to prepare an analysis of or conduct review of an analysis prepared by another party to estimate and quantif y the specific transportation related impacts of a development proposal, regardless of who initiates the development proposal, on the surrounding transportation network (FDOT, 1997). Under the concurrency law of Florida (s. 163.3180, FS), a local governmen t cannot approve a new development unless transportation facilities are sufficient to serve the traffic created from the new development at the time of occupancy. Moreover, the law requires that the level of service (LOS) of the transportation system not w orsen due to the new development. If, however, the new development will produce a large amount of traffic impairing the performance of the transportation infrastructures, the developer is responsible for any improvement necessary to maintain the original LOS.

PAGE 22

22 Often, developers are required to pay traffic impact fees for improving roads, installing new signals and providing other transportation facilities. Traffic impact fees are charged to the developer s and are used by the local government s These fees ar e now considered as a good source of financing for highway networks in many parts of the USA (CMS 2010, PennDOT 2009). Impact fees V ariable fee location of the development. F or example the city of Orlando has a set fee rate (City of Orlando, 2012). Variable fee s are more desirable as the developers are charged based on the actual impact of the development, and w hen multiple developments are occurring simultaneously developers are charged according to their portion s of the impact s Therefore it is very importa nt to determine the traffic which will arise from the development site. 2.1.1 General Procedure for Traffic Impact Analysis T raffic impact analysis is performed under the general guidelines recommended by the Institute of Trans portation Engineers (ITE, 201 0), although s tates and local government s have their own guidelines and procedures A typical basic framework for TIA is provided in Figure 2 1 which is adopted from the FDOT's Site Impact Handbook (FDOT, 1997). In the first step the developers propose the methodology to perform TIA according to the guidelines of the local government agency. This should be conducted prior to starting of the actual development.

PAGE 23

23 Figure 2 1 Typical b asic f ramework for t raffic i mpact analysis (Sourc e: FDOT, 1997) No Yes Methodology Development Existing Cond itions Analysis Background Traffic Trip Generation Trip Distribution Mode Split Assignment Future Conditions Analysis Is LOS ok Mitigation Analysis Site Access, Circulation & Parking Preview & Permitting

PAGE 24

24 The next step is a naly sis of the existing conditions including identification of the physical characteristics of the transportation system and traffic operating conditions ( i.e., LOS) of roadways and intersections. These analysis results a re used as a basis for comparing the condition after the proposed development. The third step is the determination of background traffic which is the expected future traffic (without the traffic from development site) at the end of development completion. This can be obtained manua l l y or as part of the modeling process. I n determining the impacts of the development o n the transportation system, this b ackground traffic is used as the base condition The trip generation step estimates the amount of travel as sociated with the proposed land use based on the generation rates (ITE, 2008). The internal capture (percentage of trips that occur within the site) is also determined at this step using a predetermined pe rcentage. After the trip generation, trip distribution is performed At this step, total generated trips are allocated to origin s destination s and external site s In the manual process, trip distribution can be performed at the same time as the assignm ent. Pass by trips which are external to the development but are already on the transportation system (i.e., not new trips on the roadway) are then estimated Since these trips enter the site as an intermediate stop they are not considered for the TIA. T he m ode split step estimates the usages of the various modes available to the site. In the manual method, the amount of travel that uses modes other than automobiles is estimated from regional and local guidelines based on the existing transit usage.

PAGE 25

25 Foll owing the mode split step, an assignment of vehicle trips and transit riders (person trips) to the transportation system is performed manually or by using a computer aided travel demand model. The manual assignment process is performed based on engineering judgment. After the traffic assignment, the impacts of the development generated traffic on the transportation system are assessed using the LOS guidelines and standards. If the development causes unacceptable LOS on a roadway, the effects of the traffic impacts should be mitigated (i.e. through physical or operational improvements, travel demand management strategies, fair share contributions, or a combination of these and other strategies). The site access, site circulation and parking plan are sometim es modified as part of the mitigation analysis. This is an important element in the preparation and review of site impact analyses. Access points are designed in accordance with access management and driveway permitting requirements. Parking is considered if on street parking will be employed or parking operations have the potential to impact other functions Finally, traffic impact analysis is reviewed by the concern ed authority for the final approval. 2.1.2 Methods Used by FDOT In Florida, TIA follow s the basic framework described in the previous section, but the implementation may be performed manually (manual calculation) or by using a sophisticated computer aided travel demand model (FDOT, 1997). The following sections will discuss both of the method s and their pros and cons.

PAGE 26

26 2.1.2.1 Manual method In the manual method, forecasted background traffic is estimated using existing traffic data trends. Development trips are then determined by applying the Institute of eneration rates (ITE, 2008) according to the size of the development (e.g. gross leasable area). Internal capture and pass by trips are deducted from the development trips to obtain the net development trips. Finally, net development trips are added to the background trips and trip distribution and assignment s are performed based on experience and judgments. The manual method has some advantages. It is usually more reliable when development horizons are less than ten years in the future and development si ze is small (less than 500 peak hour trips). Another advantage is that the calculations can be performed by technical personnel in a reasonable period of time. However, this method also has disadvantages. This process assumes that the proposed development will not cause significant diversions in background traffic flow patterns, but this is not always true. More importantly, this method cannot be used for large networks and as most developments are large in size, this method has little practical use. 2.1 .2.2 Travel demand model method D emand model ing is convenient for large developments with extensive street system s and numerous traffic analysis zones (TAZs). But in this method, the number of development trips a major requirement for TIA by many authori ties generated by the trip generation module usually does not match the total number of trips calculated manually using the ITE trip generation rates (ITE based trips). As a remedy, two methods are practiced: the link distribution percentage approach and the special

PAGE 27

27 generator approach (FDOT, 1997). These two methods are presented in the following subsections. Link distribution percentage method Every travel demand model needs specific input variables for production and attraction estimation (MTPO, 2005). Typical production related variables are the number of dwelling units, total population, etc, and typical attraction related variables are the size of the shopping mall, the number of employees and so on. These input variables are required for every Traff ic Analysis Zone (TAZ) of the study area. When a new development is proposed in a specific TAZ, the input variables of that TAZ need to be updated. The travel demand model uses predefined production and attraction rates to calculate the number of productio ns and attractions (FDOT, 1997 and FDOT, 1980). Trip balancing, trip distribution, mode split and traffic assignment are performed according to the model structure. After the model run, total trips (productions plus attractions) generated from the developm ent site are obtained from the trip generation summary and the trips on each link from the number trips on each link contributed by the proposed development site. T he t otal number of trips generated from the development site obtained from the model usually do es not match the value calculated manually using ITE trip generation rates. c annot be used directly for TIA. In this case, the number of development trips on each link is divided by the total number of generated trips produced by the model to obtain the percentage of development trips on each link. Finally, the ITE based trips are multiplied by the link percentages to obtain the development trips on each link.

PAGE 28

28 Special generator method In this approach, the new development is treated as a special generator whose trip generation characteristics are not fully captured by the trip ge neration sub model. The first phase of the method is same as that for the link distribution percentage method. The input variables for production s and attractions for the proposed development are inserted into the model. After the model run, the total numb er of trips generated from the development site is compared with the ITE based total. If they do not match, adjustment is performed through the special generator input files, according to the algorithm of the model (FDOT, 2010). Typical adjustment is made by adding/subtracting some number of productions and/or attractions from the TAZ of the proposed development zone. The number added or subtracted is then divided among different trip purposes. Often, this adjustment is an iterative process until the trips reported from the model match the ITE then used to obtain the development trips on each link. 2.1.3 Select Zone Analysis to determi ne the number of trips associated with any particular zone. During the traffic impact analysis flows originat ing from and/or destined for the new development site are determined by using this tool. The software stores the path flows or origin destination (O D) specific link flow distribution across the network with the new development as one of the centroids. Subsequently, the stored path flow information is used to construct the link flows across the network for use in the traffic impact analysis. Therefo re the entire TIA depend s on the path flow information generated during the deterministic user equilibrium (UE) assignment. However, it is well known that these path flow and O D specific link flow distributions may not be uniquely determined from the det erministic

PAGE 29

29 user equilibrium (UE) assignment, although the formulation has a unique aggregate link flow solution under mild conditions (Smith, 1979). Despite the nonuniqueness, the path flow distribution obtained during the traffic assignment procedure is c ommonly used in practice to conduct the select zone analysis. Consequently, the results may only represent one of the multiple possible spatial impacts. Variants of the select zone analysis, i.e., the select node or link analysis, suffer the same limitatio n. As a remedy, Rossi et al. (1989) proposed an entropy maximization approach to identify the most likely path flow distribution which forms the basis for traffic impact studies. Their formulation was path based. In order to overcome the time consuming t ask of path enumeration Akamatsu (1997) presented a link based formulation for finding the entropy maximizing O D specific link flow distribution. The entropy maximization approach wa s recently further explored by Bar Gera and Boyce (1999), Lu and Nie (20 10) and Bar Gera (2010) among others. In p articular, Bar Gera and Boyce (1999) introduced an interesting behavioral interpretation for the entropy maximizing path flow distribution : travelers should distribute in the same proportion on each of the two alte rnative segments regardless of their origins or destinations. Lu and Nie (2010) showed that under certain continuity and strict monotonicity assumptions of the link cost function, the entropy maximizing path flow distribution is a continuous function of th e inputs to a traffic assignment problem, namely the travel demand and parameters in the link cost function. Consequently, small perturbations to those inputs only lead to small changes in the path flow distribution. In this sense, the entropy maximizing p ath flow distribution is stable. Bar Gera (2010) presented an informative review and discussion

PAGE 30

30 on computing UE solutions and developed a new algorithm that achieves quick precision and practically equivalent entropy maximizing solution s Despite the abov e connection and the behavioral interpretation of proportionality, it remains an open question whether the real travel behavior leads to an entropy maximizing path flow. In fact, entropy maximization is a concept originat ing from statistical mechanics. To our best knowledge, there is no behavioral evidence to support its applicability to the identification of a path or O D specific flow solution for traffic impact analysis. Moreover, according to Leung and Yan (1997) compared with the systems in statistica l mechanics whose scale are usually in the order of more than 10 19 the scale of the spatial interaction system is generally small (e.g., the O D demand is at most of the order of 10 6 while the path flow is relatively large) The consequence is that the a ctual probability of the most probable path flow distribution under the entropy maximizing principle is actually very small, although it is large relative to other probabilities. Using an example of O D trip distribution estimation, Leung and Yan (1997) es timated that the probability of the entropy maximizing O D distribution is less than e 20 if the total demand is 5 10 6 and there are 10 origins and 10 destinations. With such a scale of probability, it is difficult to defend use of the entropy maximizatio n solution as a basis for traffic impact studies : the traffic impact fees are charged based on a flow distribution that may never actually be realized. 2.1.4 Summary TIA is a mandatory task before the approval of any new development. TIA can be performed e ither by a manual technique or by a travel demand model. As the manual method is suitable only for small size d developments in small networks, it has no practical use for large developments or networks. Therefore, a better option for TIA is

PAGE 31

31 the use of a tr avel demand model in which all analyses are performed on an integrated model network. This process minimizes any potential bias that may result from a manual method. A nalysis with the travel demand model is generally performed in two ways: one is the link distribution percentage method and the other is the special generator method. Both methods have pros and cons and there is no systematic research to show whether the two approaches generate similar results or one outperforms the other. On the other hand, obtain the number of development trips on each link. Because uses the path flow or O D specific link flow information, the number of development trips obtained from this method m ay correspo nd to one of many possible solutions. As t he developers are charged impact fees based on TIA i f different methods yield different results, it is very difficult for the concern ed authorit ies to justify the ir decisions. Therefore, this dissertation attempt s to answer the following questions: Between the link distribution percentage and special generator methods which method should be recommended for TIA? Is there a better way or more sound technology to conduct ? 2.2 Impact Analys is of Mileage Fee s 2.2.1 Need for Mileage Fee s Since the 1920s the primary means of collecting revenue to finance construction, operation and maintenance of US highways has been the fuel tax. But currently revenue collected from the gasoline tax is not s ufficient for highway financing in the USA. Due to political reason s, the gasoline tax ha s not been increased in proportion to the inflation. The s ame federal tax rate (18.4 cents per gallon) has been in use since 1993 while construction costs have increa sed many times. The federal gas oline tax has

PAGE 32

32 experienced a cumulative loss in purchasing power of 33 % since 1993 (NSTIFC, 2009). In addition, introduction of more fuel efficient and hybrid vehicles cause s the tax revenue to decrease even more Therefore t he gap between the revenue and required funding is increasing daily According to the National Surface Transportation Policy and Revenue Study Commission (NSTPRSC) the highway account balance (which was positive $9.2 billion in 2006 ) was projected to be n egative $26 billion by 2012 (NSTPRSC, 2007). The National Surface Transportation Infrastructure Financing Commission (NSTIFC) showed that total funding gap will be $2.3 trillion for the period 2010 to 2035 (NSTIFC, 2009). In order to overcome the deficit f ederal and s tate governments are seeking alternatives to current transportation revenue sources with a special focus on an alternative to the fuel tax. Several studies have been performed on this issue and t he recommendations of the major studies are pre sented below: The fuel tax and alternatives for transportation funding (TRB, 2006) The main goals of the study were to assess the long term viability of fuel taxes for transportation finance and to identify finance alternative s That report concluded: A r eduction of 20 percent in average fuel consumption per vehicle mile is possible by 2025 if fuel economy improvement is driven by regulation or sustained fuel price increases .The willingness of legislatures to enact increases (in fuel tax rates to comp ensate for reductions in fuel consumption) may be in question. Although the present highway finance system can remain viable for some time, travelers and the public would benefit greatly from a transition to a fee structure that more directly charged v ehicle operators for their actual use of roads Ultimately, in the fee system that would provide the greatest public benefit, charges would depend on mileage, road and vehicle characteristics, and traffic conditions, and they would be set to reflect t he cost of each trip to the highway agency Road use metering and mileage charging appear to be the most promising approach to this reform within a comprehensive fee scheme that will generate revenues to cover the cost of an efficient highwa y program in a fair and practical manner

PAGE 33

33 Transportation for tomorrow (NSTPRSC, 2007) The National Surface Transportation Policy and Revenue Study Commission (NSTPRSC) was established in the Safe, Accountable, F lexible and Efficient Transportation Equity Act A Legacy for Users (SAFETEA LU). One of the goals of the commission was to find long term alternatives to replace or supplement the fuel tax as the principal revenue source to support the highway trust fund. Some related comments include : The Commissi on agrees with others who have looked at long term alternatives to the fuel tax that a VMT [Vehicle Mile Travel] fee has many promising features; but, until more is known about collection and administrative costs, ways to minimize evasion and the acceptabi lity of such a mechanism to the tax payers, it is premature to rule out other types of taxes and fees to supplement traditional fuel tax revenues The Commission recommends that the next surface transportation authorization act should fund a major national study to develop a strategy for transitioning to an alternative to the fuel tax to fund highway and transit programs. Paying our way a new framework for transportation finance (NSTIFC, 2009) The National Surface Transportation Infrastructure Financing (NSTIF) Commission was established by Congress to provide recommendations for policy and action. The task of the commission was to assess future federal highway and transit investment needs, evaluate the future of the federal Highway Trust Fund, and explo re alternative funding and financing mechanisms for surface transportation. The report concluded: The current federal surface transportation funding structure that relies primarily on taxes imposed on petroleum derived vehicle fuels is not sustainable in t he long term and is likely to erode more quickly than A federal funding system based on more direct forms of user pay charges, in the form of a charge for each mile driven (commonly referred to as a vehicle miles traveled or VMT fee sy stem), has emerged as the consensus choice for the future to a new, more direct user charge system as soon as possible and commit e stablish VMT technology standards and require original eq uipment vehicle manufacturers to install standardized technology by a date certain that will accommodate the desired 2020 comprehensive implementation

PAGE 34

34 Future highway and public transportation financing (NCF, 2005) The future highway and public transporta tion finance study was commissioned by the U.S. Chamber of Commerce through the National Chamber Foundation (NCF). The objective of the study was to identify funding mechanisms to meet national highway and transit investment needs. The report concluded: Sh ort Term Strategies: The study finds that indexing federal motor fuel taxes would have the most immediate impact. The motor fuel tax is the only new approach to transportation user f Term Strategies: The federal government should provide leadership for state and local governments to implement new systems of financing transportation funding that reduce relian ce on the motor fuels tax the federal government should provide incentives for the states to develop and test new mileage based revenue systems. This process could lead to the eventual phasing out of the federal motor fuel tax and replacing it with a fede ral VMT tax Future financing options meet highway and transit needs (NCHRP, 2006) The National Cooperative Highway Research Program (NCHRP) conducted research studies on th is topic as requested by the Association of State Highway and Transportation Offi cials (AASHTO). The objective was to present options for all level s of government to reduce the highway and transit funding deficit. The report concluded: For the l onger term, fuel taxes will be vulnerable to fuel efficiency improvements and penetration of alternative fuels and propulsion systems for motor vehicles. Further, continuing reliance on more use of fossil fuel will likely run counter to long term environmental and energy needs and policies. Several recent national policy studies have recommended shifting to nonfuel based revenue sources such as VMT fees over the next 15 to 20 years The Road User Fee Task Force in Oregon also examined 28 alternative highway financing mechanisms and concluded that a mileage fee was the only broad revenue source tha t could ultimately replace the fuel tax (ODOT, 2005). But implementing a mileage fee will not be an easy task. There are issues related to technology, institution

PAGE 35

35 and public acceptance. In order to assess the feasibility of impl eme nting a mileage fee seve ral pilot studies ha ve been carried out as described in the next section 2.2.2 Pilot Studies on Mileage Fee s Oregon's pilot program (ODOT, 2007) The Oregon Department of Transportation (ODOT) launched a 12 month pilot program in April of 2006 to test th e technological and administrative feasibility of implementing the mileage fee concept. The program also tested the feasibility of using this system to collect congestion charges. In Portland, 285 vehicles, 299 motorists and two service stations were recru ited for the test. In the test ODOT implemented the Vehicle Miles Traveled Collected at Retail (VMTCAR) system. Under this system both mileage data and fee collection occurred at the gas pump. When a vehicle arrived for gas, a central reader at the stat ion detect ed whether the vehicle was equipped with the mileage fee technology. If no n equipped vehicles were detected, the vehicles we re served as usual and charged using the existing gasoline tax. If an equipped vehicle entered the stored mileage totals driven in each zone we re electronically (short point of sale (POS) system for application of the mileage fee rates. For that purpose test vehicles were outfitted with GPS based receiver s that identifie d zones for allocation of miles driven within various predefined regions. Alt hough the position of the vehicle was identified by the global positioning system the number of miles driven was calculated using the e d with the mileage reading for each zone. By using the mileage fee rates from the central ings

PAGE 36

36 the total mileage fees we re calculated. After the fuel transaction the customers we re given a bill for payment that include d both the mileage fee and the fuel purchase price less the state fuel tax. In the Portland study a flat rate of 1.2 cents pe r mile was charged which was said to be equivalent to the existing state fuel tax (24 cents per gallon). The major findings from the study are provided below: The concept was found viable with 91 % of the pilot program participants agree ing to continue paying the mileage fee if implemented statewide. The p ilot mileage fee system was successfully integrated with the service station point of sale system and the current gasoline tax collection system The mileage fee could be phased in gradually alongsi de the gas tax. Congestion and other pricing options were also found viable. The r ush hour group (charged extra fees for driving during a peak period) reduced peak period travel by about 22 % relative to the mileage group (no peak period charge). As no specific vehicle point location or trip data w ere stored privacy was protected The study also found the system to present minimal burden on business es minimal evasion potential and low cost to implement and administer. Alt hough the main objective of the pilot study was to assess the feasibility of implementing a change s due to the mileage fee. The study showed that the total mileage driven by the group of participants (who were charged a mileage fee equivalent to the existing gas tax) was reduced by 11 % (ODOT, 2007 and Rufolo et al., 2008). Puget Sound's pilot project (PSRC, 2008) The Puget Sound Regional Council conducted a pilot project from 2005 to 2007 to assess how travelers change their travel behavior ( e.g., number, mode, route, and time of vehicle trips) in response to variable charges for road use (variable or congestion based tolling). I n the greater Seattle R egion 450 vehicles from 275 households were equipped with onboard un its (GPS

PAGE 37

37 receivers, digital roadmaps and cellular communications) to conduct the study. The participants were given a particular budget that they could use. In this study the tolls were varied by time and location. The onboard units recorded the travel a nd corresponding charges were subtracted from the pre allotted travel budget. The r emaining balance from a participant's pre allotted budget was given to him/her at the end of the study. Some primary findings are s t ated below: Variable charges can signif icantly reduce traffic congestion and raise revenues for investment. The core technology for a satellite based toll system was found mature and reliable. A p roven system, viable business model and public acceptance will be required for large scale depl oyment. University of Iowa road user study (Report not available) The University of Iowa Public Policy Center conducted a federally funded study in twelve different states to assess the feasibility and public acceptance of a mileage based charging system. The field trials were conducted in two phases : the first phase (1200 participants from Austin, Baltimore, Boise, eastern Iowa and the Research Triangle in North Carolina) ended in August 2009 ; the second phase (1450 participants from Albuquerque, Billings Chicago, Miami, Portland and Wichita) ended in July 2010. The study considered technology, robustness, privacy and security, transition/phase in, public policy ramifications and public acceptance to assess the feasibility and efficiency. For the 10 month study, each vehicle was equipped with a n electrical unit consisting of an on board computer system, a global positioning system (GPS) receiver, a simple geographic information system identifying the boundaries of road use charge jurisdictions, an associat e d rate table containing the current per mile charge rate for

PAGE 38

38 each vehicle, and a cellular wireless transmitter receiver. As the vehicle travel ed the charge was calculated by the electrical unit and maintained for each jurisdiction in which the vehicle tr avel ed The charges we re periodically uploaded to a billing and dispersal center via the wireless communication link. In th is system no detailed route or time information was collected. Moreover data encryption techniques were used to further enhance sys tem privacy and security. Within the system a different number of vehicle classes c ould be created and i ntegrated with the electric tolling system. As the final report is not available we are including some preliminary observations: The l evel of accept ance of the mileage based fee increased among the participants after a few months of participation in the study. GPS was markedly less accurate than the vehicle odometer for measuring the number of miles traveled. Retrofitting the onboard unit to a wid e variety of vehicle s was very difficult. The m ileage fee wa s being considered to be the most promising alternative to the gas oline tax as a long term financial mechanism But there were some issues related to technology, institution and public acceptance. One of the most difficult hurdles was overcom ing the technological barrier. Alt hough several studies (Forkenbrock, 2004; Porter et al., 2005) provided technological solutions to implement ation of a mileage fee, most government and transport officials were skeptical about the real life implementation until completion of pilot studies. The findings of the pilot studies in dicat e that a mileage fee can be implemented in real life with some modification and enhancement of the technology. 2.2.3 Impact Studies o n Mileage Fee s Alt hough the pilot studies indicated that the implementation of a mileage fee is technically possible and feasible there are still some issues which need to be

PAGE 39

39 addressed. One of the major concerns of policymaker s is the equity issue. The ne w mileage fee should not adversely affect low income people and those in rural area s should not suffer more than urban dwellers Many studies ha ve been undertaken to address th e se issues. By using 2001 National Household Travel Survey (NHTS) data Zhang et al. (2009) showed that short term and long term distributional effects of a $0.012/mile flat would be small for people with different income levels or in different location s However the study expressed concern about the flat rate mileage fee as this m ay discourage car owners from buying fuel efficient vehicle s, making it difficult to ac hiev e carbon reduction in the future. In order to promote more fuel efficient vehicle s McMullen et al. (2010) tried a in which l ower fuel efficien cy vehicle s were charged more than the higher fuel efficien cy vehicle s With 2001 NHTS data from Oregon the study showed that the step fee structure was more regressive than t he flat mileage fee as low income drivers owned less efficient vehicles. Larsen et al. (2012) found that an environmental ly friendly fee would be horizontally less equitable (rural household contribut ing a higher percentage of revenue) than a gas oline ta x al though they found it equitable among different economic classes. A spatial ly equitable mileage fee was also assessed and found to be vertically equitable. For their study the 2009 NHTS data of Texas w ere used. Instead of a flat mileage fee, Sana et al. (2010) used mileage based fees that varied with vehicle type and time of day (peak and of f peak). N ationwide 2001 NHTS

PAGE 40

40 data w ere used in their study. By trial and error they set fees that could generate sufficient fund s to replace the federal gas tax of $0.184/gallon. The authors found that the vehicle mile s traveled by every economic class was decreased with the mileage fee. And the mileage fee was also found equitable among the different income classes. Zhang et al. (2012) assessed the impact of a m arginal mileage fee. By using 2009 NHTS data from Oregon, the authors showed that the marginal mileage fee was significantly higher than a mileage fee equivalent to the existing state gas tax. Alt hough the marginal mileage fee would reduce vehicle mile s tr avel ed by 27%, the fee was found to be somewhat regressive in nature. 2.2.4 Other Studies on Mileage Fee s By using travel demand and highway expenditure data from the State of Indiana, Oh et al. (2007) studied the mileage fee rate for self financing highwa y pricing schemes. The study showed that with federal aid a mileage fee of 2.9 cents per mile would cover current expenditures for state administered highways in the absence of any other revenue source and that a fee of 2.2 cents per mile would be suffi cient if revenue from vehicle registration w ere maintained. The Minnesota Department of Transportation conducted a pay as you drive experiment to observe the change s in driving behavior as a result of a mileage fee (Abou Zeid et al., 2008 and MnDOT, 2006) The study found that the fee had the largest effect on weekend and peak weekday travel. Litman (1999) showed that total vehicle miles travel ed would decline by approximately 25% by imposing different types of mileage fee s : w eight distance charges 7.6%, d istance based insurance 12.6%, d istance based registration fees 3.3% and e mission fees 6.6%.

PAGE 41

41 DeCorla Souza (2002) mentioned that converting fixed vehicle charges (such as taxes, insurance, registration and lease fees) into distance based charges could re duce vehicle mile s travel and generate revenue of $44 billion in 20 years. Moreover this fee would be more equitable and affordable. 2.2.5 Summary The fuel tax has proven to be a viable mechanism for highway financing for many years. However, many studies (TRB, 2006, NSTPRSC, 2007, NSTIFC, 2009 and others) have indicate d that the existing gas oline tax rate is not sufficient to meet the financial need s for building and maintaining road networks in the USA. Short term deficit s can be minimized by increasing the fuel tax b ut due to the increas ing number of fuel efficient and hybrid vehicles, the conventional fuel tax is not a viable financing mechanism for the long term. Therefore a road user based fee, more specifically mileage fee or VMT fee, was thought t o be most promising alternative to the gas oline tax by most of the study committees. possible and feasible to implement and that the fee can be integrated with the current gas o line tax collection system with ease and small cost. M any states including Florida, are now considering the possibilities of switching from a gasoline tax to a mileage fee (FDOT, 20 0 5). I t is critical to know the possible impacts of the new system before its implemen ta ti o n Alt hough a few studies ha ve been performed on the impacts of mileage fee implementation, none was conducted in the context of Florida. As the s tate gas oline tax es are different in different s tates, the required mileage fee to replace th e gas oline tax would be different. Moreover, the equity impact would be different in different

PAGE 42

42 demographic location s Therefore this dissertation attempts to answer the following questions: What would be the mileage fee s te and local gasoline tax? What would be the socioeconomic impact s if Florida switches from a gasoline tax to mileage fee s ? Can we design a mileage fee structure that is more environment ally friendly and equitable ?

PAGE 43

43 CHAPTER 3 COMPARISON OF TRAFFI C IM PACT ANALYSIS METHOD S 3.1 Background In Florida traffic impact analysis is performed both manually and by using travel demand model s Using t he Florida Standard Urban Transportation Model Structure (FSUTMS) which is often used for large network s with lar ge development s the analysis can be performed in two ways: the link distribution percentage approach and the special generator approach. Detailed description s of the methods and their pros and cons are presented in C hapter 2. In this chapter we describe an empirical study that compare s these two methods to determine whether these two approaches generate similar results or if one outperforms the other. The Alachua/Gainesville MPO model i s used as the test bed. A number of scenarios of new developments a re created by changing various characteristics of two hypothetical developments. The traffic impacts of those hypothetical developments are estimated by implementing these two methods respectively. A qualitative comparison between these two methods is also p resented. 3.2 Qualitative Comparison of the T wo Methods In general, the link distribution percentage method is easier to implement. A single model run is sufficient and the resulting link distribution percentage pattern can be used in different scenarios. However, this method makes an implicit assumption that the link distribution percentage pattern remains the same even if a larger number of trips is generated in the new development. The assumption may not be valid, particularly when the network is conge sted, and the estimates of trip production from the model and the ITE Trip Generation report are substantially different.

PAGE 44

44 The special generator method does not require such an assumption. However, for the special generator the number of trips need s to be adjusted iteratively until the numbers reported from the travel demand model match the estimates based on the ITE rates. Moreover, the distribution of trip purpose needs to be estimated externally while in the link distribution percentage method i t is a utomatically determined by the model. In addition, there are several precautions needed when using the special generator adjustment in a travel demand model: (1) during the adjustment process for special generators, the model should not double count the tr ips of special generators (one from the regular model and the other from the special generator); (2) if the balancing is performed on the total number of adjusted productions and attractions, significant addition or deletion of trip attractions for special generators will impact the number of trip attractions for zones without special generators; (3) when determining special generator rates from the ITE Trip Generation report it is important to note that the ITE rates provide vehicle trips while travel de mand models deal with person trips. 3.3 Empirical Study 3.3.1 Study Site The Alachua/Gainesville MPO (MTPO, 2005) model wa s selected given that we we re most familiar with the region and had no success in obtaining a real world new development from another region. The model wa s built upon Cube Voyager and has been validated using Year 2000 data. In this model, the region is divided into 446 TAZs. With an intention to find an under developed TAZ to locate the hypothetical new development, the area was c arefu lly searched in Google Map and all the TAZs having a small amount of trip production and attraction in the model were examined A few TAZs were identified as potential sites for the case study, located in the northwestern,

PAGE 45

45 southwestern, northeastern and s outheastern part s of the study area respectively. Among those potential sites, TAZs 225 and 148 we re selected for the empirical study. TAZ 225 is located in the northeast part of the town, near NE Waldo R oad, with a total production equal to 243 person tr ips and total attraction of 90 person trips TAZ 148 is located in the southeast section near SE Williston R oad and S Main S treet, with a total production equal to 64 person trips and total attraction of 158 person trips. Although both zones are currently under developed, their surrounding areas and road characteristics are different. TAZ 225 is on the outskirt of the city and the surrounding areas are all under developed. In contrast, TAZ148 is near downtown where there is substantial business developme nt and the road network nearby is dense. The study site from Google Map is presented in Figure 3 1 and the TAZ configuration is presented in Figure 3 2. Figure 3 1. Map of the s tudy s ite ( Source: G oogle m ap)

PAGE 46

46 Figure 3 2 TAZ c onfiguration in the Alach ua/Gainesville MPO m odel 3.3.2 The Alachua/Gainesville Model The Alachua/Gainesville model follows the framework of the Florida Standard Urban Transportation Model Structure (FSUTMS) which is a traditional four step model (FDOT, 1997). The model use s fou r zonal data files: ZDATA1 (trip production data), ZDATA2 (trip attraction data), ZDATA3 (special generator data) and ZDATA4 (internal external production data). The model parameters and trip production and attraction rates are provided within the model By using the zonal files, parameters and rates t he GEN module generate s person trips for seven trip purposes (home based work, home based shop, home based social/recreation, home based others, non home based, truck taxi and internal external). Alt hough t he GEN module contains a set of default trip attraction ra tes for all seven trip purposes, c ustomized rates can also be specified in the second part of the GRATE file. After calculating the number of trip productions and

PAGE 47

47 attractions by zone and trip purpos e using user supplied or default trip rates, the GEN module adds the special generator trips specified in the ZDATA3 file. The GEN module then adjusts the number of trip attractions in each travel analysis zone s o that total number of trip attractions for each purpose matches the trip production totals for the same purpose. 3.3. 3 Implementation of the T wo Methods In our case study, the proposed development is a shopping center, an attraction only site. Therefore, the input variables related to attraction ( ZDATA2 file) need to be updated before executing the model The relevant input variables are : manufacturing industrial employment by place of work (MFGEMP), commercial employment by place of work (COMEMP), service employment by place of work (SERVEMP) and total employment by place of work (TOTEMP). Eight scenarios a re created for the new development at either TAZ 225 or 148. The maximum development size i s made the same as that of TAZ 237, which contains the Oaks Mall (largest shopping mall in Gainesville) while the other seven a re determined arbitrarily with the size of the shopping mall ranging from 50,000 to 100,000 square feet. The input data for these eight scenarios and the current situation of both TAZs are provided in Table s 3 1 and 3 2, respectiv ely Table 3 1 Input d ata for h ypothetical d evelopment s cenarios Scenario OIEMP MGEMP COMEMP SEREMP TOTEMP 1 0 10 170 20 200 2 0 20 250 30 300 3 0 26 500 50 576 4 0 26 700 70 796 5 0 36 1000 100 1136 6 0 36 1200 130 1366 7 0 36 1500 150 1686 8 0 36 2358 238 2632

PAGE 48

48 Table 3 2 Current situation in TAZs 225 and 148 TAZ OIEMP MGEMP COMEMP SEREMP TOTEMP 225 0 6 0 0 6 148 0 0 0 30 30 Link distribution percentage method The employment data for the development site i s updated in ZDATA2 file according to a specific development scenario. After executing the model, the development traffic on each link attributable to the new development is generation (both production and attraction) of the developm ent site i s retrieved from the development trips coming to each link, is calculated as the ratio between the development traffic on each link and the total generation of t he site. Finally, the real number of development trips on each link is obtained by multiplying an external estimate of the total trips generated from the new development by the link percentages. The external estimate can be made with reference to the ITE Trip Generation report In this analysis, the estimates a re made using the rates recommended by FDOT (FDOT, 1980) instead of the ITE rate s The relevant shopping trip rates are summarized in Table 3 3. Table 3 3. Trip r ates r ecommended by FDOT (Source: FDOT, 1980) Retail s hopping c enters Recommended a ttraction t rip r ates Recommended m ajor t rip p urposes 200,000 sq. ft. or more 13 Trips/Employee Home Based Shop 100,000 200,000 sq. ft. 33 Trips/Employee Home Based Shop 50,000 100,000 sq. ft. 30 Trips/E mployee Home Based Shop Assuming that the size of the hypothetical shopping center is between 50,000 and 100,000 sq. ft, the rate of 30 trips per employee i s used in our analysis.

PAGE 49

49 Special generator method Similar to the link distribution percentage a pproach, the employment data for the development site a re updated in ZDATA2. The new development i s then treated as a special generator and its attraction i s further adjusted in ZDATA3 during iterative runs of the model until the number of trips reported from the model match es the external estimate of ITE trip generation. Instead of assigning all the adjusted trips to be home based shopping trips, those attractions a re distributed among five trip purposes as follows: 14% for home based work, 44% for home b ased shopping, 4% for home based social recreation, 12% for home based others and 26% for non home based. Final adjustments made in ZDATA3 file are provided in Table 3 4. After the final adjustment the development traffic on each link attributable to the new development i Table 3 4 Adjustment for s pecial g enerator (Number of trips added) TAZ Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 Scenario 6 Scenario 7 Scenario 8 225 2800 4500 8000 11000 16000 19 000 23000 36000 148 3000 4300 8500 11000 16000 19000 23500 36000 3.4 Results The traffic impacts of the proposed eight hypothetical scenarios in two different TAZs (TAZ 148 and TAZ 225) we re estimated by implementing both the link distribution percentag e and special generator methods The number of productions and attractions of th e development TAZs obtained from the model run and manually calculated number of ITE based trips are summarized in Tables 3 5 and 3 6. As expected, the number of trip attractio ns predicted from the MPO model do es not match the number of ITE based trip attractions, thereby justify ing the need for these two methods.

PAGE 50

50 Table 3 5 Summary of t rip g eneration (TAZ 225) Scenario n o. Total p roduction from m odel Total a ttraction from m od el Total ITE a ttraction 1 1055 1908 6000 2 1444 2781 9000 3 2620 5388 17280 4 3564 7476 23880 5 2988 10608 34080 6 5954 12724 40980 7 7350 15752 50580 8 11409 24473 78960 T able 3 6 Summary of t rip g eneration (TAZ 148) Scenario n o. Total p roduct ion from m odel Total a ttraction from m odel Total ITE a ttraction 1 817 1831 6000 2 1205 2703 9000 3 2379 5310 17280 4 3325 7395 23880 5 4748 10530 34080 6 5715 12648 40980 7 7110 15675 50580 8 11169 24395 78960 Figures 3 3 to 3 11 show g raphical r epresentations of the flow distribution across the network and the volumes from TAZ 225 or 148 both without and with a new development (o nly scenario 8) before and after the special generator trip adjustments From Figures 3 3 to 3 7 we can see that the traffic pattern across the network looks similar before and after the development even though the maximum size development was used This is due to the fact that additional trip s generated from the development sites are distributed within the entire netw ork giving a false impression about the effect of new development. For that reason select zone analysis is performed to determine the development traffic from the new site. F rom Figures 3 8 to 3 11 th e traffic flow distribution from the new development site is not same before and after the special

PAGE 51

51 generator adjustment The differences in these before and after comparisons may be a source of concern when using the link distribution percentage method. Figure 3 3 Flow d istribution without n ew d evelopmen t Figure 3 4 Flow d istribution with d evelopment s cenario 8 in TAZ 225 before s pecial g enerator a djustments

PAGE 52

52 Figure 3 5 Flow d istribution with d evelopment s cenario 8 in TAZ 225 after s pecial g enerator a djustments Figure 3 6 Flow d istribution wi th d evelopment s cenario 8 in TAZ 148 before s pecial g enerator a djustments

PAGE 53

53 Figure 3 7 Flow d istribution with d evelopment s cenario 8 in TAZ 148 after s pecial g enerator a djustments Figure 3 8 Traffic v olumes from TAZ 225 with d evelopment s cenario 8 b efore s pecial g enerator a djustments

PAGE 54

54 Figure 3 9 Traffic v olumes from TAZ 225 with d evelopment s cenario 8 after s pecial g enerator a djustments Figure 3 10 Traffic v olumes from TAZ 148 with d evelopment s cenario 8 before s pecial g enerator a djustments

PAGE 55

55 Figure 3 11 Traffic v olumes from TAZ 148 with d evelopment s cenario 8 after s pecial g enerator a djustments The l ink volumes obtained from the link distribution percentage and special generator methods are provided in Tables 3 7 to 3 10. Note that there a re 6252 links in the network and the analyses are performed for all the 8 scenarios. However, o nly the top 10 links with large development traffic for scenario s 1 (smallest size development) and 8 (largest size development) are presented here T he numbers of development trips on each link obtained from both methods are consistent up to scenario 7 for both of the sites ( TAZ 148 and TAZ 225 ) However, for scenario 8, the number of development trips on each link for TAZ 148 is consistent, but the re is a signi ficant difference for TAZ 225 (Table 3 9) Although both of the sites use the same development size, the i r locations are different ; TAZ 148 is located in a developed area and TAZ 225 is located in an under developed area. With a large scale development, th e assumption of the constant

PAGE 56

56 link distribution percentage pattern unlikely holds in the link distribution percentage method in TAZ 225 On the other hand, the large amount of attraction to the new development may cause the special generator approach to pro duce a distorted trip distribution pattern for the originally under developed area (TAZ 225) Both factors and possibly others act together and result in the observable discrepancy The sum of the total number of development trips obtained from both method s is also compared and found to be consistent with the ITE based trips (Table s 3 11 and 3 12) Table 3 7 Link v olumes of t op 10 l inks (TAZ 225 with s cenario 1) Rank Link Development v olume on e ach l ink Node A Node B Link p ercentage m ethod Special g ene rator m ethod 1 225 2873 2637 2741 2 2873 225 2637 2741 3 2802 2799 2096 2161 4 2873 2802 2096 2161 5 2799 2802 2095 2159 6 2802 2873 2095 2159 7 2799 2675 1590 1596 8 2675 2799 1581 1582 9 2675 2657 1129 1069 10 2657 2675 1153 1040 Table 3 8 Link v olumes of t op 10 l inks (TAZ 148 with s cenario 1) Rank Link Development v olume on e ach l ink Node A Node B Link p ercentage m ethod Special g enerator m ethod 1 148 2445 2653 2868 2 2445 148 2653 2868 3 2583 2445 1644 1794 4 2586 2583 1644 1794 5 24 45 2583 1621 1772 6 2583 2586 1621 1772 7 2431 2432 1132 1125 8 2432 2434 1132 1125 9 2434 2586 1132 1125 10 2432 2431 1106 1102

PAGE 57

57 T able 3 9 Link v olumes of t op 10 l inks (TAZ 225 with s cenario 8) Rank Link Development v olume on e ach l ink Node A Nod e B Link p ercentage m ethod Special g enerator m ethod 1 2873 225 27106 24150 2 225 2873 27106 23915 3 2802 2799 18055 14585 4 2873 2802 18055 14585 5 2799 2802 18037 14506 6 2802 2873 18037 14506 7 2927 2938 4412 10314 8 2938 2927 4393 10235 9 2899 2901 4025 10099 10 2901 2927 4025 10099 T able 3 10 Link v olumes of t op 10 l inks (TAZ 148 with s cenario 8) Rank Link Development v olume on e ach l ink Node A Node B Link p ercentage m ethod Special g enerator m ethod 1 148 2445 31043 32261 2 2445 148 310 43 32261 3 2583 2445 18925 19015 4 2586 2583 18925 19015 5 2445 2583 18679 18834 6 2583 2586 18679 18834 7 2445 2325 11995 13028 8 2325 2445 11748 12847 9 2201 2185 11414 11813 10 2203 2201 11414 11813 Table 3 11 Sum of t otal d evelopment t rips on e ach l ink (TAZ 225) Scenario Link d istribution p ercentage m ethod Special g enerator m ethod 1 137879 139387 2 201779 207687 3 368514 368992 4 495303 499519 5 742180 701632 6 802607 831169 7 963182 1014026 8 1429331 1629316

PAGE 58

58 Table 3 12. Sum of t o tal d evelopment t rips on e ach l ink (TAZ 148 ) Scenario Link d istribution p ercentage m ethod Special g enerator m ethod 1 138166 149192 2 205705 214804 3 389945 405600 4 532974 556633 5 750133 796274 6 892270 951532 7 1095832 1171383 8 1631178 1818451 The number s of d evelopment trips on each link obtained from the link distribution percentage and special generator methods are compared for every link in the network. For this purpose, the root mean square errors (RMSEs), defined below (Equation 3 1) are calculated for every scenario using the special generator method as the base case: where denotes the development trips on link by the special generator method (base case) and denotes the development trips on link by the link distribution percentage method Calculated RMSEs are presented in Tables 3 13 and 3 14 and in Figures 3 12 and 3 13. These tables and figures indicate that these two met hods produce fairly consistent estimates of traffic impacts caused by the different development scenarios in both of the two hypothetical development s ites (TAZ 148 and TAZ 225) The RMSEs between the results from these two approaches a re very small, rangi ng from 0.0058 to 0.0219. (3 1)

PAGE 59

59 T able 3 13 Comparison of t wo m ethods (TAZ 225) Scenario ITE a ttractions of d evelopment z one RMSE 1 6000 0.0063 2 9000 0.0063 3 17280 0.0107 4 23880 0.0145 5 34080 0.0185 6 40980 0.0166 7 50580 0.0163 8 78960 0.0219 T able 3 14 Comparison of t wo m ethods (TAZ 148) Scenario ITE a ttractions of d evelopment z one RMSE 1 6000 0.0064 2 9000 0.0058 3 17280 0.0059 4 23880 0.0063 5 34080 0.0073 6 40980 0.0075 7 50580 0.0081 8 78960 0.0103 F igure 3 12 RMSE of d eve lopment l ink f lows from t wo m ethods (TAZ 225) 0.0000 0.0050 0.0100 0.0150 0.0200 0.0250 0 20000 40000 60000 80000 100000 RMSE ITE attractions (no. of trips)

PAGE 60

60 F igure 3 13 RMSE of d evelopment l ink f lows from t wo m ethods (TAZ 148) To investigate how the link percentage pattern varies across different scenarios, the link percentages obtained from different scenario s are also compared. Only the 10 links with the large st percentage of development traffic are presented in Tables 3 15 and 3 16. RMSEs are calculated for every scenario relative to the link percentages from s cenario 1. The comparisons are presented in Tabl e 3 17 and Figure 3 14. The RMSE values are very small in this case study b ut the RMSE increases with increas ing development size This implies that the link percentages obtained from different scenarios are fairly consistent as long as the difference bet ween the number of trips generated from the model and number of ITE based trips is not significant. 0.0000 0.0020 0.0040 0.0060 0.0080 0.0100 0.0120 0 20000 40000 60000 80000 100000 RMSE ITE attractions (no. of trips)

PAGE 61

61 Table 3 15 Link p ercentages of t op 10 l inks (TAZ 225) Node A Node B Link % (S1) Link % (S2) Link % (S3) Link % (S4) Link % (S5) Link % (S6) Link % (S7) Link % (S8) 225 2873 37.38 36.94 35.49 34.52 38.25 32.67 31.83 29.99 2873 225 37.38 36.94 35.49 34.52 38.25 32.67 31.83 29.99 2802 2799 29.71 29.24 27.89 26.86 29.11 24.50 22.74 19.98 2873 2802 29.71 29.24 27.89 26.86 29.11 24.50 22.74 19.98 2799 280 2 29.70 29.24 27.19 26.86 28.97 23.83 22.62 19.96 2802 2873 29.70 29.24 27.19 26.86 28.97 23.83 22.62 19.96 2799 2675 22.54 22.14 21.10 20.42 22.63 19.27 18.65 14.95 2675 2799 22.41 21.95 21.09 20.44 22.63 19.29 18.71 15.05 2657 2675 16.34 15.88 14.98 14.33 15.36 13.08 12.23 8.92 2675 2657 16.00 16.05 14.89 14.08 14.62 12.35 12.32 8.86 Table 3 16 Link p ercentages of t op 10 l inks (TAZ 148) Node A Node B Link % (S1) Link% (S2) Link % (S3) Link % (S4) Link % (S5) Link % (S6) Link % (S7) Link % (S8) 1 48 2445 38.92 38.71 37.99 37.47 36.80 36.38 35.81 34.44 2445 148 38.92 38.71 37.99 37.47 36.80 36.38 35.81 34.44 2583 2445 24.11 23.97 23.64 23.19 22.90 22.44 22.19 21.00 2586 2583 24.11 23.97 23.64 23.19 22.90 22.44 22.19 21.00 2445 2583 23.77 23.61 2 3.15 22.87 22.49 22.15 21.84 20.72 2583 2586 23.77 23.61 23.15 22.87 22.49 22.15 21.84 20.72 2431 2432 16.60 16.48 16.21 15.76 15.54 15.23 14.98 13.87 2432 2434 16.60 16.48 16.21 15.76 15.54 15.23 14.98 13.87 2434 2586 16.60 16.48 16.21 15.76 15.54 15. 23 14.98 13.87 2432 2431 16.22 16.10 15.74 15.54 15.24 14.90 14.60 13.54 T able 3 1 7. Variations of link distribution percentages of different scenarios Scenario RMSE (TAZ 225) RMSE (TAZ 148) 1 N/A (Base Case) N/A (Base Case) 2 0.0028 0.0022 3 0.0049 0.0027 4 0.0063 0.0035 5 0.0065 0.0041 6 0.0117 0.0049 7 0.0147 0.0055 8 0.0207 0.0083

PAGE 62

62 F igure 3 14 Variations of l ink d istribution p ercentages of d ifferent s cenarios 3. 5 Summary Both the link distribution percentage and special generator metho ds are used for performing traffic impact studies in Florida In this chapter we present an empirical study to compare these two methods. The Alachua/Gainesville MPO model used for this study is a traditional four step model built upon C ube V oyager softwa re. Eight hypothetical development scenarios are built in two sites to compare these two methods. Based on this empirical study, the link distribution percentage method and the special generator method produce fairly consistent estimates of traffic impact s caused by different scenarios created in two hypothetical development sites The RMSEs between the results from these two approaches a re very small, ranging from 0.0058 to 0.0219. As the link distribution percentage method is easier to implement than the special generator method, we recommend the link distribution percentage method for TIA 0 0.005 0.01 0.015 0.02 0.025 0 1 2 3 4 5 6 7 8 RMSE Scenario no. TAZ 225 TAZ 148

PAGE 63

63 The link distribution percentage method makes an implicit assumption that the link distribution percentage pattern remains the same even if a larger number of trips is generated in the new development. The assumption may not be valid, particularly when the network is congested, and the estimates of trip production from the travel demand model and the ITE Trip Generation report are substantially different. However, thi s is not evident in our experiment. More specifically, the link percentage patterns obtained for the different scenarios a re also very consistent with the RMSE ranging from 0.0022 to 0.0207. The quality of the results produced by the link distribution pe rcentage method depends on how well the trip generation module replicates the real scenario of the modeling area. With a well developed trip generation module, there is no need to rely on the ITE Trip Generation report to estimate the trip generation from the new development. Consequently, the link distribution percentage method will produce accurate estimates. Indeed, since there is no need to introduce a special generator to capture the difference, these two methods coincide again. Howeve r, both of the ab ove mentioned assignment procedure and then estimates the number of development trips based on the stored path flows. S ince th e path flow patterns from traffic assignment are not unique, the estimates may represent one of many possibilities. The analysis should be performed with extreme caution (Bar Gera et al., 2010). As a remedy, we p ropos e using the average path flow distribut ion (see C

PAGE 64

64 CHAPTER 4 ENHANCING SELECT ZON E ANALYSIS 4.1 Background S elect Z one A and plays a critical role in traffic impact studies. Mor e specifically, it estimates the link flow s that originate from and/or are destined for the new development ( i.e., the link use s created by the new development ) Doing so requires knowledge of the path flow distribution or origin destination (O D) specific link flow distribution across the network with the new development as one of the centroids. However, it is well known that these two flow distributions may not be uniquely determined from a deterministic user equilibrium (UE) assignment, although the form ulation has a unique aggregate link flow solution under mild conditions (Smith, 1979). As a remedy, Rossi et al. (1989) proposed an entropy maximization approach to identify the most likely path flow distribution and suggested using it as the basis for tra ffic impact studies. After a further discussion on the entropy maximization approach, we propose using another path flow or O D specific link flow distribution as the basis for the select zone analysis and the subsequent traffic impact studies. Given that there is no behavioral evidence to favor one particular solution over another, in this dissertation all solutions are treated equally with the same merit. In other words, each solution is assumed to have an equal probability of occurrence. Consequently, th e mean of all the path or O D specific link flow solutions, which is essentially the center of gravity of the UE polyhedron, seems a logical selection for the basis of traffic impact studies. Numerical examples are provided to demonstrate the proposed appr oach and to compare it with the entropy maximizing approach.

PAGE 65

65 4.2 Impact of Nonuniqueness on Select Zone Analysis Bar Gera and Luzon (2007) conducted a computational study on the Chicago network to demonstrate the nonuniqueness of UE path flow solutions. The network consists of 93,513 O D pairs and 127,248 UE paths. Their computational study revealed that flows on 41% of the UE paths carrying 13% of the total demand are not uniquely determined. In a similar sp i rit, this section presents a small example to highlight the impact of nonunique path flow solutions on the outcome of the select zone analysis. We estimate the upper and lower bounds of the link use by a new development on each link. Coincidence of the two bounds of one link implies that the link u se can be uniquely determined for that particular link. Otherwise, the distance between these two bounds reflect s the severity of the nonuniqueness. To present the mathematical formulations for estimating these bounds, we define as a strongly connected road network with and being the sets of nodes and links respectively. There are two methods for representing the traffic flows on a road network, namely link based and path based. For the former, let denote the node link incidence matrix of the road network. By definition, must be of size where denotes the cardinality of a set. We use as the index for O D pairs and as the set of all O D pairs with denot ing the travel demand for O D pair It is assumed that is given for all Let and represent the origin and destination of O D pair and the demand vector whose elements satisfy the requirement that or and for all the other nodes

PAGE 66

66 Additionally, link is d enote d as or by the pair of its starting and ending nodes, i.e., For O D pair or represents the traffic flow on link from users traveling from the origin to the destination of O D pair The vector having as its elements denotes the link flow vector for the O D pair with a s a vector with the link flows for all O D pairs as its elements. The sum is the associated aggregate link flow vector. Then, the set of all feasible flow distributions can be described as follows: Let denote the amount of flow on path that connects O D pair and (equals 0 or 1) indicates whether link is on path Then, the following set is equivalent to : where is the set of paths f or O D pair Let denote the time or cost to traverse link and is a vector of these link travel times. It is further assumed that is strictl y monotone. It is well known that is the UE flow distribution. We further define as the set of all O D specific link flow solutions associated with the aggregate UE flow distribution and as the set of UE path flow solutions. Both and are polyhedral convex sets.

PAGE 67

67 Assuming that the new development is located in node the upper (maximum) and lower (minimum) bounds of the use of link by node can be estimated by solving the following linear programs: The objective function is to maximize or minimize the sum of the O D specific flows whose origin or destination is node This linear program is solved for every link in the network to obtain the upper and lower bound of the development flow on the link. For comparison, we also compute the link uses estimated based on the maximum entropy user equilibrium (MEUE) solution to the following nonlinear convex program (Akamatsu, 1997): The nine node network, as illustrated in Figure 4 1, is used in our numerical example. The free flow travel time s and capacit ies of each link ar e provided in parenthesis in the figure. The BPR (Bureau of Public Road) travel cost function is used, and the O D demands are , and F igure 4 1 The Nine n ode n etwork With node 1 being the new development, Table 4 1 presents different estimates of number of link uses including the upper bound lower bound and the values based on the notion of entropy maximization. The results show that the number of link uses of 10 (5, 12) 1 2 5 6 7 8 3 4 9 (2, 11) (3, 25) (9, 35) (6, 33) (6, 43) (6, 18) (3, 35) (9, 20) (4, 11) (2, 19) (4, 36) (8, 39) (6, 24) (8, 26) (4, 26) (7, 32) (2, 30)

PAGE 68

68 links (out of 18 total) by node 1 cannot be uniquely determined. The distance s between the upper and lower bounds are substantial, indicating that there will be a larger error if we arbitrarily select one particular path or O D specific flow solution to conduct the select zone analysis. On the other hand, the estimates based on the entropy maximizing solution are always between the u pper and lower bounds, which is entropy maximizing solution described by Bar Gera (2010). T able 4 1 Link u ses of n ode 1 Link UE a ggregate l ink f low Upper b ound of l ink u se Lower b ound of l ink u se Link u se based on MEUE 1 5 11.22 11.22 11.22 11.22 1 6 28.78 28.78 28.78 28.78 2 5 50.32 0.00 0.00 0.00 2 6 39.68 0.00 0.00 0.00 5 6 0.00 0.00 0.00 0.00 5 7 31.11 11.22 0.00 2.52 5 9 40.42 11.22 0.00 8.70 6 5 0.00 0.00 0.00 0.00 6 8 52.36 28.78 0.00 17.09 6 9 36.10 28.78 0.00 11.69 7 3 36.21 10.00 0.00 6.26 7 4 10.84 10.85 0.00 2.41 7 8 0.00 0.00 0.00 0.00 8 3 23.79 10.00 0.00 3.74 8 4 59.16 20.00 9.16 17.59 8 7 0.00 0.00 0.00 0.00 9 7 35.94 20.85 0.00 6.16 9 8 40.58 40.00 0.00 14.23 4.3 Entropy Maximizing Flow Distribution Before presenting an alternative basis for the select zone analysis, this section offers a further discussion on the maximum entropy solution. The notion of maximum entropy has been w idely used in spatial distribution modeling (e.g., Wilson, 1970). In our context, it estimates the most probable path or

PAGE 69

69 O D specific link solution. Previous research efforts have produced efficient solution algorithms for computing the flow distributio n (see Bar Gera, 2010 and references therein). At the same time, Bar Gera and Boyce (1999) introduced a behavioral interpretation for the solution, i.e., the proportionality assumption. T he following discussion of the connection between the stochastic user equilibrium (SUE) flow distribution and the entropy maximizing flow distribution may pro vide additional insight into the entropy maximizing approach Consider the following logit based SUE formulation (Fisk, 1980): where is a positive dispersion parameter and the expression is assigned the value zero at It is known that is strictly convex with respect to f and thus both the path and link flo w solutions, denoted as and are unique (Fisk, 1980). As the first term in dominates the second and reduces to The refore, the solution will approach the deterministic user equilibrium solution, i.e., (Fisk, 1980). We naturally wonder whether the unique path flow distribution converges to a particular path flow solution. Larsson et al. (2001) pro ved that it approaches the MEUE path solution. In the following, we present another proof that where Note that the solution is unique (e.g., Rossi et al., 1989).

PAGE 70

70 Lemma 1 The entropy of the SUE path flow distr ibution is the upper bound of the entropy associated with the UE path flow solution, i.e., Proof: Given that is a feasible path flow distribution, by definition we have which is equiva lent to the following: Since is in user equilibrium and is a feasible path flow distribution, we obtain: Combining ( 4 1) and ( 4 2) yields: Because it implies that Theorem 1. The entropy maximizing path flow distribution is the limit of the SUE path flow distribution as the dispersion parameter approaches infinity, i.e., Proof: Since is a continuous function with respect to and is bounded, we have Given that is a UE path flow solution, i.e., We then prove by contradiction. To do so, we assume that Because is continuous with respect to the path flow and is bounded, From Lemma 1, we know This contradicts (4 1) (4 2)

PAGE 71

71 the fact that is a unique minimum solution to the maximum entropy problem. Therefore, the assumption is false, and thus It can be interpreted tha t the above theorem provides additional support for using the maximum entropy flow distribution as the basis for the select zone analysis, since it is the limit of the SUE path flow, which is based upon a sound behavioral foundation. The connection also pr ovides another perspective for the proportionality assumption, which seems to relate to the property of independence of irrelevant alternatives (IIA) of the multinomial logit model. Following the definition of the proportionality described by Bar Gera (201 0), we assume that and represent an arbitrary pair of alternative segments in the network; is a path segment from origin to the diverge node of and ; is a path segment from the merge node of and to destination Consequently, is a path between O D pair using segment and similarly is another path from to using instead. Denot ing the flow on these two paths as and respectively i t is easy to show that at SUE, where and are the travel times of segments and The ratio is ind ependent of the O D pair w and the other path segments or This implies that the proportionality property exists in a regular SUE path flow distribution. Therefore, i t is not surprising to observe the same prope rty at its limit. Despite the above connection and the behavioral interpretation of proportionality, it remains an open question whether real travel behavior leads to an entropy maximizing path flow. As noted in C hapter 2 entropy maximization is a concep t originat ing from statistical mechanics in which the scale of computation is much higher than that of a

PAGE 72

72 spatial flow distribution system ( Leung and Yan 1997). The consequence is that the actual probability of the most probable path flow distribution und er the entropy maximizing principle is actually very small, less than for a total demand of with origins and destinations ( Leung and Yan 1997) Moreover, there is no behavioral evidence to support its applicability to the identification of a path or O D specific flow solution for traffic impact analysis. 4.4 Average Flow Distribution Given that there is no behavioral foundation to favor one particular solution over another, it seems plausible to view each solution as having the same merit. In other words, we assume that the solutions are uniformly distributed in the solution polyhedron. We then propose using the mean of all the solutions as the basis for the select zone analysis. In the following, we focus on the O D specific link flow solution to develop the idea. The average of all the O D specific solutions is essentially the center of gravity of the polyhedron 4.4.1 Definition To facilitat e the presentation, we first rewrite in a more concise format. Define as a matrix consisting of identity matrices, i.e., = Let = diag with node link incidence matrices in total. Let denote a column vector with as its element. is of size Fu rther define and Then can be written as Let de note the mean of O D specific link flow solutions over the polyhedron Mathematically,

PAGE 73

73 where is the measure of The analytic center of the polyhedron may be also of interest, e.g., for comparison purposes. The center is the solution to the following pr oblem: 4.4.2 Stability Stability here implies that small perturbations of model inputs will lead only to small changes in the outcome of the select zone analysis. It is an important property first discussed by Lu and Nie (2010). If the basis (a particular selection of O D specific solution) for the select zone analysis is not stable, a small perturbation to the model inputs could result in a dramatic change in the outcome, thereby causing doubts on the credibility of the analysis. Lu and Nie (2010) showed that the entropy maximizing solution is a continuous function of the model inputs and is thus stable. In the following, we demonstrate that is also stable. Let denote the parameters in the link cost function (and denote the O D travel demand). Suppose that the link travel cost is continuous with respect to According to Theorem 1 in Lu and Nie (2010), the UE link flow distribution is a continuous function with respect to and Moreover, is a continuous multifunction with respect to and (Theorem 2, Lu and Nie 2010). Additionally, the measure function is c ontinuous with respect to Given bounded dom and dom the (4 3) (4 4)

PAGE 74

74 polyhedron is also bounded. Therefor e and are bounded. Further, is continuous with respect to Theorem 2. Assume that the link travel cost function is strictly monotone to link flow and continuous wit h respect to and The average function is continuous with respect to and Proof : Let be a sequence converging to We need to prove that Note that is continuous on dom dom then Since is a con tinuous function and is bounded on dom dom converges to (Theorem 4.26, Rockafellar and Wets, 1998). Consequently, is continuous with respect to and Continuity of the average O D specific flow solution implies that small perturbations to model inputs lead only to small changes in the flow distribution. I f we use the average O D specific link flow solution as the basis for the select zone analysis, the estimate of the spatial impacts of a new development will therefore b e stable. Note that the analytic center of the polyhedron can also be proved to be stab le in a similar manner. 4.4. 3 Computation The average O D specific flow solution can be numerically computed via sampling. In this study we adopt the and et al. (2009). The algorithm was initially pr oposed by Smith (1984) and is a Monte Carlo procedure to generate samples from a full dimensional convex set. Starting from an

PAGE 75

75 initial point the extended hit and run algorithm generates a random direction base d on the basis of the null space of the matrix and then obtains a new sample as where is a scalar randomly generated while ensuring The iteration proceeds until a sufficient number of samples ha s been generated from Instead of working with the full system, we remove the links that are not utilized in the UE solution and generate a reduced system This improves the sam pling efficiency to a great extent. Let denote the j th element of a vector and the sampling algorithm is as follows. Step 1: (Construct ion of the reduced system and the basis) The problem of for each O D pair and each link is solved If the maximum solution is zero, the O D specific link flow is removed from the initial system. A reduced system is created when all the zero flow O D specific links are removed. T he ba sis of the null of the matrix is computed a nd denote d as which is a ssume d to consist of a number of column vectors, i.e., Step 2: ( Initialization ) A n initial point in is generated and and are both set zero Step 3: ( Generat ion of a random direction ) A set of random numbers is generated from the standard normal distribution say, The random direction can then be calculated as:

PAGE 76

76 Step 4: ( Gene rat ion of a step size and new sample ) The minimal and maximal step size s are calculated as follows: A random scalar is generated that follows the uniform distribution between [ ]. The new sample is thus Step 5: ( Verification of the convergence ) If the average of the previous samples is used to calculate the standard deviation of the sample mean, de noted as If <0.00001, convergence is complete the sample mean is returned Otherwise, is set t o and the process is repeated from Step 3. Ban et al. (2009) demonstrat ed that the samples generated via the above procedure follow a uniform distribution over the polyhedron 4.5 Numerical Example We applied the above sampling procedure to the nine node network in Section 4.2 to obtain the average O D specific link flow solution and then estimated the number of link uses of node 1 (the new development) based on the average The estimates are reported in Table 4 2 together with the estimates based on the entropy maximizing (MEUE) solution and the anal ytic center. The estimates are given only for those 10 links whose numbers of link uses cannot be uniquely determined. For each link, the table also presents the relative difference (RD) calculated as: (4 5)

PAGE 77

77 where, Number of link uses based on the average solution Number of link uses based on either the MEUE solution or analytic center Upper bound of link uses Lower bound of link uses It can be obse rved that the number of link uses based on the average and MEUE solutions are substantially different. The maximum relative difference is as high as 15% and the average is about 6%. The average relative difference from the analytic center is 36%. We also c omputed t he average relative difference between the MEUE solution and analytic center which is 38%. T able 4 2 Comparisons of l ink u ses of n ode 1 Link UE a ggregate l ink f low Upper b ound of l ink u se Lower b ound of l ink u se Link u se based on avg. solution MEUE Analytic center Link u se Relative d ifference Link u se Relative d ifference 5 7 31.11 11.22 0.00 4.17 2.52 15% 10.00 52 % 5 9 40.42 11.22 0.00 7.05 8.70 15% 1.22 52 % 6 8 52.36 28.78 0.00 17.36 17.09 1% 28.78 40 % 6 9 36.10 28.78 0.00 11.42 11.69 1% 0.00 40 % 7 3 36.21 10.00 0.00 5.65 6.26 6% 10.00 44 % 7 4 10.84 10.85 0.00 2.92 2.41 5% 0.85 19 % 8 3 23.79 10.00 0.00 4.35 3.74 6% 0.00 44 % 8 4 59.16 20.00 9.16 17.08 17.59 5% 19.15 19 % 9 7 35.94 20.85 0.00 4.41 6.16 8% 0.85 17 % 9 8 40.58 40 .00 0.00 14.06 14.23 0% 0.37 34 % We conducted the same analysis for the Sioux Falls network (Figure 4 2) to further examine how average flow and MEUE solutions differ in a larger network. In our example, the network topology is the same as the one in LeB lanc et al. (1975). The network characteristics including the free flow travel time and capacity of each link, and the O D demands are provided in appendi ces A and B It is assumed that a new

PAGE 78

78 development is located in node 10. Our computation shows that t he number of uses of 56 links by the new development cannot be uniquely determined. For these links, Table 4 3 presents the estimates of the link uses including the upper and lower bounds, as well as the link uses based on the analytic center, entropy maxi mizing and average solutions. The results confirm the observation from the first example that the average and entropy maximizing solutions lead to substantially different estimates of spatial impacts of the new development. The maximum and average relative difference s are approximately 51% and 16% respectively The average relative difference from the analytic center is 21%. Figure 4 2 Sioux Falls n etwork

PAGE 79

79 T able 4 3 Comparisons of l ink u ses of n ode 10 Link UE a ggregate l ink f low Upper b ound of l ink u se Lower b ound of l ink u se Link u se based on avg. Solution MEUE Analytic center Link u se Relative d ifference Link u se Relative d ifference 1 2 27.21 0.21 0.00 0.07 0.03 19% 0.18 52 % 3 1 49.55 3.21 3.00 3.07 3.03 19% 3.18 52 % 3 4 89.63 8.46 7.00 7 .09 7.08 1% 7.00 6 % 4 3 85.35 5.21 5.00 5.07 5.03 19% 5.18 52 % 4 5 90.68 12.46 0.00 8.24 5.33 23% 4.97 2 6 % 4 11 38.62 11.19 0.00 2.85 5.75 26% 6.03 2 8 % 5 4 96.30 9.21 0.35 7.83 9.00 13% 9.18 1 5 % 5 6 68.76 8.00 2.07 7.02 7.60 10% 7.15 2 % 5 9 50.06 20. 08 5.00 13.33 10.33 20% 9.97 2 2 % 6 2 33.45 4.00 3.79 3.93 3.97 19% 3.82 52 % 6 5 41.70 4.08 0.00 0.09 0.00 2% 0.00 2% 6 8 100.61 8.00 3.92 7.91 8.00 2% 8.00 2% 8 6 59.99 5.72 0.00 0.91 0.37 9% 0.66 4 % 8 9 54.90 13.00 8.92 12.91 13.00 2% 13.00 2% 9 5 7 7.76 20.00 5.63 17.85 19.60 12% 19.34 1 0 % 9 8 39.52 9.72 4.00 4.91 4.38 9% 4.66 4 % 9 10 97.64 36.46 24.00 32.24 29.33 23% 28.97 2 6 % 10 9 107.95 30.00 21.35 28.76 29.98 14% 30.00 14% 10 11 75.53 35.65 14.00 17.73 15.58 10% 14.90 1 3 % 10 15 107.56 35.00 14.00 29.93 28.20 8% 26.10 18 % 10 16 40.46 14.00 12.69 13.92 13.99 5% 14.00 6 % 10 17 33.31 14.31 5.00 7.65 10.25 28% 13.00 57 % 11 4 33.71 8.65 0.00 1.24 0.03 14% 0.00 14% 11 10 79.13 39.82 13.54 19.16 28.97 37% 30.95 45 % 11 14 41.61 13.00 0.00 2.50 1. 54 7% 0.90 12 % 12 3 76.63 1.46 0.00 0.09 0.08 1% 0.00 6 % 12 11 43.31 10.00 8.54 9.91 9.92 1% 10.00 6 % 14 11 47.59 13.82 0.00 1.40 8.30 50% 9.92 62 % 14 15 42.10 14.00 0.00 7.09 1.72 38% 0.08 50 % 14 23 32.56 10.00 0.00 5.11 2.37 27% 2.84 23 % 15 10 116. 38 41.00 21.06 35.68 27.70 40% 26.08 4 8 % 15 14 36.77 13.00 0.00 5.61 3.83 14% 4.94 5 % 15 19 31.95 8.00 0.00 3.39 2.75 8% 0.00 42 % 15 22 59.12 23.00 9.00 15.93 16.63 5% 16.16 2 % 16 10 48.83 18.00 11.82 17.77 18.00 4% 18.00 4% 16 17 46.18 6.18 0.00 0.23 0.00 4% 0.00 4% 17 10 39.83 17.88 6.00 10.15 11.00 7% 11.00 7% 17 16 36.97 1.31 0.00 0.08 0.00 6% 0.00 6% 17 19 47.05 8.00 0.00 2.57 5.25 34% 8.00 68 %

PAGE 80

80 Table 4 3 C ontinued Link UE a ggregate l ink f low Upper b ound of l ink u se Lower b ound of l ink u se Lin k u se based on avg. solution MEUE Analytic center Link u se Relative d ifference Link u se Relative d ifference 19 15 21.39 6.12 0.00 1.20 0.00 20% 0.00 20% 19 17 57.35 6.12 0.00 3.92 5.00 18% 5.00 18% 19 20 34.55 4.00 0.00 1.96 4.00 51% 4.00 51% 20 19 21.29 1.12 0.00 0.13 0.00 12% 0.00 12% 20 21 39.95 2.95 0.00 0.03 0.01 1% 0.00 1% 20 22 44.74 6.00 1.93 5.84 5.99 4% 6.00 4% 21 22 39.60 11.95 5.00 5.70 5.01 10% 5.00 10% 21 24 43.81 5.00 0.00 1.64 2.67 21% 2.16 10 % 22 15 80.16 26.00 15.88 22.39 2 0.98 14% 21.00 14% 22 20 35.80 4.00 0.00 2.04 0.00 51% 0.00 51% 22 21 26.13 9.00 4.00 5.64 6.67 21% 6.16 10 % 22 23 40.14 10.00 0.00 3.25 4.95 17% 5.00 1 8 % 23 14 35.88 9.00 0.00 3.48 5.02 17% 5.00 17 % 23 22 41.76 9.00 0.00 4.85 3.98 10% 4.00 9 % 23 24 33.81 5.00 0.00 3.36 2.33 21% 2.84 10 % 24 21 42.32 4.00 0.00 0.67 0.00 17% 0.00 17% 24 23 33.75 4.00 0.00 3.33 4.00 17% 4.00 17% 4.6 Summary S elect Z one A and requires knowledge of the path flow distribution or origin destination (O D) specific link flow distribution. Since these two flow distributions may not be uniquely determined from the deterministic user equilibrium assignment, selection of a particular flow distribution as the basis f or the select zone analysis remains an open question T his dissertation suggests using the mean of all the path s or the O D specific user equilibrium solutions as the basis for traffic impact analysis and proves its stability. A modified extended hit and r un sampling algorithm is proposed to compute the average O D specific link flow distribution. It is noted that such an alternative basis is proposed as a remedy in situation s where there is no empirical evidence to support a particular selection of solutio n.

PAGE 81

81 Compared with the entropy maximizing basis, we believe that the proposed basis is more intuitively appealing to practitioners, and the result should be easier to defend. However, it is more computationally intensive given that efficient algorithms exis t to solve for the entropy maximizing solutions. The major challenge is to construct a line a r system of very large size for the sampling procedure.

PAGE 82

82 CHAPTER 5 IMPACT ASSESSMENT OF MILEAGE BASED USER FEES IN F LORIDA 5.1 Background The combination of ever i mproving fuel efficiency of gasoline powered vehicles and the increasing use of those with alternative fuels or other sources of energy have eroded the fuel tax system in the United States. In order to overcome the revenue deficit, v ehicle mileage fees hav e been suggested as one of the most promising mechanisms to replace the fuel tax (Chapter 2) In this system, drivers would be charged based on the miles travel ed instead of paying the conventional gasoline tax. As discussed in Chapter 2, t he proposed sys tem has many positive attributes over the gasoline tax Studies completed for Oregon, Iowa Texas, etc, have also demonstrated the technical feasibility and public acceptability of such fees as the fuel tax replacement. However, no impact study has been co nducted for Florida and results from other states may not be applicable in this state In this chapter we analyze the socioeconomic impacts of mileage fees in Florida Four different mileage fee structures (flat fee, step fee, fee based on vehicle fuel e fficiency and fee based on vehicle type) are assessed for different scenarios While the step mileage fee aims to achieve a more equitable structure, a fee based on vehicle fuel efficiency and vehicle type will be more environment ally friendly 5.2 Data 5. 2.1 Description of the Data The 2009 National Household Travel Survey (NHTS) data for Florida are used for this impact analysis. The data consist of four files : household, person, trip and vehicle files. There are 15884 household entries, 30952 person entr ies, 114910 trip entries and

PAGE 83

83 29457 vehicle entries in the dataset. Our analysis is performed at the household level. Consequently, some attributes from the vehicle and person data files are merged into the household data file. After data cleaning, 13086 ho usehold data are used for the analysis. The key points in data preparation are provided below: The NHTS data does not have the specific household income value. For the model estimation the average value of the income range is used. For example, if the hou sehold is in income range $5000 to $9999, $7500 is taken as household income for the analysis. In the type of vehicle field there are two different types of trucks However, for our analysis we consider both pickup and truck as truck In the vehicle d ata file there are two types of annual mile data, one is ANNMILES (self reported annualized mile estimate) and other is BESTMILE (best estimate of annual miles). We use BESTMILE for our analysis. In our analysis Energy Information Administration (EIA) f uel efficiency measures are used rather than Environmental Protection Agency (EPA) fuel efficiency. If any information used in the model is missing in the household data file that household is excluded from the analysis. If any vehicle information for a household used in the model is missing that corresponding household is excluded from the analysis. The weighted average of the vehicle characteristics are used in the analysis for the households having more than one vehicle. Mileage fee has not been im plemented yet. Therefore the data do not contain any information about a mileage fee. We do not know how people will react and change their travel demand s with the new fee. For our analysis, we assume that the travel demand sensitivity with respect to mil eage fee is similar to the sensitivity with respect to gasoline fees. 5.2.2 Descriptive Statistics of the Data Descriptive statistics of the data are presented in Tables 5 1 to 5 5 From the tables, we can observe that the average fuel efficiency of vehicl es per household is 20.79 MPG in the rural area and 21.40 MPG in the urban area. The average fuel efficienc ies are also s imilar among different income groups. However, the number of

PAGE 84

84 vehicles and total annual mile s driven per household are both greater in t he rural area and among high er income group s For vehicle composition, the households in rural area s own a higher percentage of trucks and a l ower percentage of cars compared to households in urban area s The lower income groups have a higher percentage o f cars and l ower percentage of SUVs. The average fuel efficiency and average annual vehicle mile s driven for each county are presented in Table 5 5 Table 5 1 Descriptive statistics by location No.HH Tot. annual income per HH ($) No. veh. per HH Avg. v eh. MPG per HH Tot. annual VMT per HH Rural 2775 59984 2.2 0 20.79 25552 Urban 10311 63706 1.88 21.40 19905 Overall Avg. 62917 1.93 21.27 21056 Table 5 2 Descriptive statistics by income group Income group Household No. veh. per HH Avg. veh. MPG p er HH Tot. annual VMT per HH Total HH Rural HH Urban HH $0 $19,999 2119 475 1644 1.40 20.64 12187 $20,000 $39,999 3288 737 2551 1.64 21.04 15926 $40,000 $59,999 2468 537 1931 1.91 21.49 21035 $60,000 $79,999 1800 373 1427 2.16 21.78 24650 $80,000 $200,000 3411 653 2758 2.42 21.46 29629 Table 5 3. Percent of vehicles by type and location Car Van SUV Truck RV Motor Cycle Total Rural 42.69 8.02 18.81 25.76 0.78 3.94 100 Urban 55.17 8.32 19.61 13.23 0.68 2.98 100 Table 5 4 Percent of vehicles by type and income group Income group Car Van SUV Truck RV Motor Cycle Total $0 $19,999 60.11 9.29 11.51 16.56 0.54 1.99 100 $20,000 $39,999 54.91 9.53 15.12 17.12 0.71 2.62 100 $40,000 $59,999 52.09 8.28 18.09 17.69 0.8 5 3.01 100 $60,000 $79,999 49.78 7.85 20.58 17.08 0.80 3.91 100 $80,000 $200,000 48.97 7.22 25.32 14.04 0.64 3.81 100

PAGE 85

85 Table 5 5 Descriptive statistics by county County name No. of HH Avg. MPG Avg.VMT County name No. of HH Avg. MPG Avg.VM T Alach ua 203 21.74 22215 Lee 420 21.08 19261 Baker 35 19.6 0 28692 Leon 274 21.56 25573 Bay 157 20.27 24143 Levy 65 21.08 32177 Bradford 42 20.37 31243 Liberty 5 19.82 31286 Brevard 373 21.56 18338 Madison 28 19.67 21641 Broward 1091 22.13 20500 Manate e 180 20.75 16495 Calhoun 20 20.14 28195 Marion 257 20.94 19098 Charlotte 129 22.1 0 20010 Martin 117 20.94 22000 Citrus 299 20.88 21920 Miami dade 1256 21.74 19607 Clay 175 20.92 25895 Monroe 170 20.8 0 21281 Collier 209 21.17 18020 Nassau 67 21.7 1 28517 Columbia 66 21.25 26898 Okaloosa 182 20.92 24101 DeSoto 44 19.54 19390 Okeechobee 51 20.1 0 19154 Dixie 21 18.74 32041 Orange 432 21.92 23271 Duval 523 21.97 22077 Osceola 104 21.5 0 25155 Escambia 280 21.31 22062 Palm Beach 815 21.92 18712 Flagler 124 23.05 24678 Pasco 309 20.75 19315 Franklin 24 18.62 20832 Pinellas 690 21.32 15846 Gadsden 35 21.59 24346 Polk 340 21.23 19687 Gilcrist 25 20.54 48320 Putnam 129 20.87 26888 Glades 15 19.36 25541 St. Johns 178 21.33 26101 Gulf 29 19 .72 23260 St. Lucie 235 21.45 19940 Hamilton 25 19.4 0 27880 Santa Rosa 170 20.67 25298 Hardee 22 20.98 27697 Sarasota 249 21.95 19347 Hendry 38 20.44 23109 Seminole 250 21.19 20663 Hernando 109 21.77 20598 Sumter 62 21.33 16216 Highlands 224 20.5 0 18780 Suwannee 92 19.99 26018 Hillsborough 657 21.42 21346 Taylor 32 19.4 0 41091 Holmes 35 20.71 24735 Union 14 18.93 24759 Indian River 94 20.64 23064 Volusia 384 21.38 18281 Jackson 69 19.7 0 28926 Wakulla 21 21.83 32001 Jefferson 35 21.24 228 25 Walton 62 21.77 25177 Lafayette 9 18.49 23346 Washington 41 21.26 23709 Lake 169 21.18 20774

PAGE 86

86 5.3 Methodology 5.3.1 Model Estimation Different modeling structures ha ve been used in previous studies. F or example, McMullen et al. (2010) used bo th a static model and a multiple regression model; Zhang et al. (2009) used a discrete choice model ; and Sana et al. (2010) used previously estimated elasticity values to estimate the mileage fee impacts. Z hang and McMullen (2010) compared different modeli ng structures from the perspective of behavioral realism, policy sensitivity and practicality While the static model s are very easy to use, they perform poor ly with regard to behavioral realism. On the other hand discrete continuous models capture behavi oral realism well but they are more data demanding and difficult to use Multiple regression models are capable of capturing behavioral realism and are easy to execute, and they work well for policy analys e s. Therefore in this research we use a multiple regression model. The total annual mile s driven by a household, i.e., the travel demand of a household, is assumed to be a function of the social characteristics of the household and attributes of the transportation services. For simplicity, it is assumed that the household vehicle ownership is fixed. A log log linear regression model is used to avoid a negative value of vehicle mile s driven by a household. Generally p eople with high income s are less sensitive to price changes than those at the low end of the income scale In order to capture this effect, an interaction term between the fuel cost per mile and income i s included in the model. If a household has more than one vehicle, the household has the option to switch vehicles When the fuel price is hig h they can drive the more fuel efficient vehicle and vice versa. Th erefore this type of household is less sensitive to the fuel price than a household with a single type of vehicle and this effect

PAGE 87

87 is captured by an interaction term between fuel cost pe r mile and the substitute dummy variable. The functional form is described below: w here, Total a nnual mile s driven by all vehicle s in a household (mile) W eighted average of fuel cost in per mile for a household ($/mile) calculated as follows: = / where = + + ( represents individual vehicles in a househol d) when ($/mile) when ($/mile) T otal annual household income ($) H ousehold vehicle count if household is located in urban area, 0 otherwise if household has different types of vehicles among car, van, SUV, truck and R V ; 0 otherwise N umber of workers in a household N umber of c hildren in a household The model is estimated with the clean data set described in Section 5.2 The adjusted R square is 0.56 and all the coefficients have the correct sign and are statistically significant at the 99% confidence level The coefficient s of the estimated model are provided in Table 5 6 and the resulting demand elasticity is provided in (5 1)

PAGE 88

88 Table 5 7 From the elasticity values we can observe that lower income people are more sensitive to fuel cost than those with high er income s and households with only one type of vehicle are more sensitive than household s with multiple types of vehicle s The elasticity is calculated as follows: Table 5 6 Estimated m odel Variable Name Coefficient Std. e rror t stati stic Constant 2.4 787 0.6472 3.8298 ln(PM) 5.4067 0.3416 15.8274 ln(hhtotinc) 0.7612 0.0620 12.2745 ln(hhvehcnt) 0.9188 0.0196 46.9485 U 0.1821 0.0147 12.3627 ln(hhtotinc)*ln(PM) 0.3449 0.0327 10.5518 SUB 1.6881 0.1163 14.5149 ln(PM)*SUB 0.7499 0.0607 12.3453 w rkcnt 0.1509 0.0084 18.0693 hhchild 0.1023 0.0079 12.8797 Table 5 7 Elasticity by i ncome g roup b ased on a verage i ncome Income group Avg. income ($) Elasticity with SUB Elasticity without SUB $0 $19,999 10000 1.40 2.15 $20,000 $39,999 30000 1. 10 1.85 $40,000 $59,999 50000 0.92 1.67 $60,000 $79,999 70000 0.80 1.55 $80,000 $200,000 140000 0.59 1.34 5.3.2 Socioeconomic Measures In this study, we consider the change in consumer surplus, revenue and social welfare to estimate the s ocioeconomic impacts. The c hange in consumer surplus is estimated to capture the impacts of mileage fees on households, whereas the change in revenue is estimated to see the feasibility of the new system. T he total change in social welfare is simply the su m of change in consumer surplus and change in revenue. With (5 2)

PAGE 89

8 9 the estimated model, the changes in consumer surplus, revenue and social welfare are estimated as follows: w here, under gasoline tax ($) under mileage fee ($) Annual mile s driven by a household under gasoline tax (mile) Annual mile s driven by a household under mileage fee (mile) Change in consumer surplus ($) Change in r evenue ($) Change in social welfare ($) Income and spatial equit y are also assessed. In previous studies (e.g., Zhang et al., 200 9 ; McMullen et al., 20 10; Zhang and Lu, 2012), equity is assessed mainly based on the change in consumer surplus and change in consumer surplus as a percent age of average income. In this study, we use the lat t er method to judge the equity. For county wise distributional impact evaluation, the Lowrenz curve and Gini coefficient (Larsen et al., 2012) are used. 5.4 Impact Analysis 5.4.1 Flat Mileage Fee The current average gasoline tax in Florida is 52.9 cents / gallon with the federal tax and 34.5 cents/gallon without the federal t ax, i.e., the sum of the state and county taxes. (5 3) (5 4) (5 5)

PAGE 90

90 Using 21 mile s per gallon (MPG) as the average fuel efficiency of vehicles in Florida, the current state gas tax of 34.5 cents / gallon is equivalent to 1.64 cents / mile if all other factors r emain the same Note that the above calculation does not consider travel behavior changes due to change s in the gas oline price. To obtain a revenue neutral impact fee and fees for other purposes, t he model is executed multiple times and the resulting socioeconomic impact s are summarized in Table 5 8 Table 5 8 Changes in consumer surplus, revenue, social welfare and VMT under different mileage fees Mileage f ee (cents/mile) Total c hange in consumer surplus ($) Total c hange in r evenue ($) Total c hange in social w elfare ($ ) % VMT reduction 1.60 151780 16177 135602 0.49 1.61 127431 5030 132461 0.57 1.62 103102 26204 129306 0.65 1.63 78791 47345 126136 0.73 1.64 54500 68452 122953 0.81 1.70 90849 194409 103560 1.30 2.00 807692 807027 666 3.70 2.20 1276853 1200298 76555 5.29 2.80 2646108 2313856 332252 10.04 4.1 0 5445869 4437238 1008631 20.14 It can be observed from Table 5 8 that a flat mileage fee of 1.61 cents/mile is sufficient to maintain the current revenue level (without considering the difference i n the collection and administrative cost s ). The mileage fees of 2.8 and 4.1 cents/mile can reduce annual vehicle mile s travel ed by approximately 10 % and 20 % respectively. In the following section we present the socioeconomic impacts of 1.61, 2.8 and 4.1 cents/mile mileage fees across different income groups and counties.

PAGE 91

91 Mileage fee of 1.61 cents/mile (Revenue neutral fee) The socioeconomic impacts are presented in Tables 5 9 to 5 1 1 and in Figure 5 1. It can be observed that the average change in consum er surplus and the average change in consumer surplus as a percentage of average income are negligible in all income groups (Table 5 9 ) The residents in rural area s receiv e slightly more benefits than those in the urban area s (Table 5 1 0 ), although the di fference is not significant. Across different counties, the average change in consumer surplus ranges from $ 13.92 to $ 77.93 (Table 5 1 1 ). Other than Flagler, Hernando, Broward and Charlotte counties, the average changes in consumer surplus are positive. D ue to the higher fuel efficiency of vehicles those counties are negatively affected to a small extent On the other hand, counties with less fuel efficien t vehicles are benefitting from the new policy. Table 5 9 Average c hanges in c onsumer s urplus, r eve nue and s ocial w elfare by i ncome g roup Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 2.18 (45.50) 0.02 0.49 2. 67 $ 20,000 $ 39,999 4.47 (64.20) 0.01 1.12 5.59 $ 40,000 $ 59,999 5.94 (72.90) 0.01 2.99 8.94 $ 60,000 $ 79,999 8.62 (86.82) 0.01 5.24 13.86 $ 80,000 $ 200,000 22.85 (101.47) 0.02 4.84 18.01 Table 5 1 0 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 22.30 (95.21) 5.73 16.57 Urban 6.36 (72.94) 2.03 8.39

PAGE 92

92 Table 5 1 1 Average c hanges in c on sumer s urplus, r evenue and social w elfare by county Sl. n o. County n ame No. of HH Avg. c hange in c onsumer s urplus ($) Avg. c hange in r evenue ($) Avg. c hange in social w elfare ($) 1 Alachua 203 10.75 4.47 15.22 2 Baker 35 54.44 30.51 23.94 3 Bay 157 25. 27 12.34 12.92 4 Bradford 42 25.52 0.54 24.98 5 Brevard 373 2.89 6.39 9.27 6 Broward 1091 1.46 7.92 6.46 7 Calhoun 20 32.15 19.49 12.66 8 Charlotte 129 1.04 8.68 7.64 9 Citrus 299 13.36 2.36 11.00 10 Clay 175 13.75 1.76 12.00 11 Collier 209 11.99 1.37 10.62 12 Columbia 66 15.59 3.41 12.18 13 DeSoto 44 33.99 20.55 13.44 14 Dixie 21 77.93 47.34 30.59 15 Duval 523 4.67 5.00 9.67 16 Escambia 280 10.52 2.47 12.99 17 Flagler 124 13.92 24.14 10.21 18 Franklin 24 51.43 29.89 21.54 19 Ga dsden 35 0.13 8.27 8.40 20 Gilcrist 25 74.57 33.37 41.20 21 Glades 15 35.10 19.52 15.57 22 Gulf 29 39.75 30.35 9.40 23 Hamilton 25 31.18 17.34 13.84 24 Hardee 22 27.27 9.53 17.74 25 Hendry 38 31.18 10.54 20.64 26 Hernando 109 4.78 13.59 8.80 27 Highlands 224 20.20 10.99 9.21 28 Hillsborough 657 13.18 3.27 9.92 29 Holmes 35 14.88 5.72 9.16 30 Indian River 94 20.50 5.87 14.63 31 Jackson 69 42.87 20.13 22.74 32 Jefferson 35 1.28 9.52 10.79 33 Lafayette 9 61.30 40.97 20.34 34 Lake 16 9 6.92 1.33 8.26

PAGE 93

93 Table 5 11 C ontinued Sl. n o. County name No. of HH Avg. change in consumer surplus ($) Avg. change in revenue ($) Avg. change in social welfare ($) 35 Lee 420 11.75 2.55 9.20 36 Leon 274 10.80 3.27 14.07 37 Levy 65 20.91 0.09 20.82 38 Liberty 5 40.92 23.11 17.81 39 Madison 28 41.71 24.43 17.27 40 Manatee 180 11.95 4.16 7.79 41 Marion 257 9.67 2.70 6.97 42 Martin 117 14.14 3.95 10.19 43 Miami dade 1256 1.12 4.78 5.90 44 Monroe 170 27.72 7.45 20.27 45 Nassau 67 13.61 0. 54 13.07 46 Okaloosa 182 29.96 12.76 17.21 47 Okeechobee 51 25.52 9.40 16.12 48 Orange 432 2.78 5.10 7.88 49 Osceola 104 19.46 5.53 13.93 50 Palm Beach 815 1.35 5.14 6.50 51 Pasco 309 10.83 2.46 8.37 52 Pinellas 690 5.89 0.33 5.56 53 Polk 340 11.44 1.84 9.60 54 Putnam 129 12.33 1.32 11.01 55 St. Johns 178 19.65 2.74 16.90 56 St. Lucie 235 4.93 5.04 9.97 57 Santa Rosa 170 28.40 11.24 17.16 58 Sarasota 249 8.73 2.60 11.32 59 Seminole 250 13.37 2.32 11.06 60 Sumter 62 4.33 3.62 7.95 6 1 Suwannee 92 29.09 8.98 20.11 62 Taylor 32 59.73 43.39 16.35 63 Union 14 67.21 41.73 25.48 64 Volusia 384 7.22 3.21 10.43 65 Wakulla 21 39.15 6.93 46.08 66 Walton 62 18.39 9.70 28.09 67 Washington 41 23.05 5.43 17.63

PAGE 94

94 Figure 5 1 Spatial d is tribution of i mpacts of m ileage f ee of 1.61 c ents/ m ile

PAGE 95

95 Mileage fee of 2.8 cents/mile (10% VMT reduction) The impacts of a flat mileage fee of 2.8 cents/mile are presented in Tables 5 1 2 to 5 1 4 and in Figure 5 2. Such a high fee would lead to negative cha nges in consumer surpluses for all income group s and locations. The impacts are regressive, i.e., the lower income people suffer more than those with higher income and residents in rural area s suffer more than those in urban area s The distributional impa cts are presented in Table 5 1 4 and in Figure 5 2. The dispersion of the impacts yields a Gini coefficient of 0. 17 suggesting that the spati al equity is not a major concern. Note that the NHTS data has a weight factor for each entry to avoid sampling bias However, in this study we have not use d the weight factor s to estimate the model. We believe that the amount of data in each income group is sufficiently large that the average effect is not affected by the sampling bias. However, as many counties have s mall number s of data entries the changes in consumer surpluses are adjusted according to weight factors before computing the Gini coefficient. Table 5 1 2 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup Income group Avg. change in consumer surplus in $ (std. dev.) Avg. change in consumer surplus as % of avg. income Avg. change in revenue ($) Avg. change in social welfare ($) $0 $19,999 106.43 (104.14) 0.84 77.84 28.59 $20,000 $39,999 149.67 (145.02) 0.49 120. 95 28.72 $40,000 $59,999 200.53 (160.81) 0.40 171.68 28.84 $60,000 $79,999 246.95 (180.34) 0.35 221.42 25.53 $80,000 $200,000 289.97 (201.56) 0.22 272.35 17.62 Table 5 1 3 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfar e by l ocation Location Avg. change in consumer surplus in $ (std. dev.) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 238.21 (211.16) 214.76 23.44 Urban 192.52 (165.70) 166.61 25.91

PAGE 96

96 Table 5 1 4 Average changes in consumer su rplus, revenue and social welfare by county Sl. n o. County n ame No. of HH Avg. c hange in c onsumer s urplus ($) Avg. c hange in r evenue ($) Avg. c hange in social w elfare ($) 1 Alachua 203 258.80 232.58 26.28 2 Baker 35 222.00 208.79 12.33 3 Bay 157 18 6.40 167.44 19.01 4 Bradford 42 267.90 252.05 15.87 5 Brevard 373 198.50 172.90 25.69 6 Broward 1091 215.30 184.28 31.06 7 Calhoun 20 200.50 179.42 21.08 8 Charlotte 129 201.40 170.42 31.04 9 Citrus 299 189.00 163.24 25.80 10 Clay 175 230.90 206.54 24.42 11 Collier 209 184.50 162.80 21.75 12 Columbia 66 211.60 185.21 26.43 13 DeSoto 44 146.20 129.10 17.19 14 Dixie 21 211.00 200.72 10.28 15 Duval 523 224.30 196.19 28.12 16 Escambia 280 209.70 187.48 22.30 17 Flagler 1 24 252.60 220.78 31.84 18 Franklin 24 188.30 176.67 11.63 19 Gadsden 35 228.80 193.36 35.45 20 Gilcrist 25 169.10 175.67 6.48 21 Glades 15 182.90 169.29 13.68 22 Gulf 29 161.20 136.43 24.78 23 Hamilton 25 204.20 183.20 20.02 24 Hardee 2 2 216.90 195.84 21.14 25 Hendry 38 189.90 175.77 14.20 26 Hernando 109 195.80 166.49 29.31 27 Highlands 224 150.00 129.31 20.75 28 Hillsborough 657 201.50 177.10 24.42 29 Holmes 35 190.10 161.52 28.66 30 Indian River 94 181.10 163.99 17 .20 31 Jackson 69 222.00 208.89 13.15 32 Jefferson 35 248.50 221.49 27.04 33 Lafayette 9 179.60 171.55 8.12

PAGE 97

97 Table 5 14 C ontinued Sl. n o. County name No. of HH Avg. change in consumer surplus ($) Avg. change in revenue ($) Avg. change in social w elfare ($) 34 Lake 169 176.20 151.59 24.63 35 Lee 420 183.60 160.13 23.48 36 Leon 274 241.80 220.20 21.63 37 Levy 65 275.50 247.03 28.55 38 Liberty 5 216.00 194.12 21.94 39 Madison 28 170.30 154.04 16.29 40 Manatee 180 162.70 141.34 21 .37 41 Marion 257 172.70 145.05 27.69 42 Martin 117 197.60 175.08 22.53 43 Miami dade 1256 207.00 175.62 31.43 44 Monroe 170 189.20 178.59 10.70 45 Nassau 67 242.60 215.94 26.76 46 Okaloosa 182 226.80 208.67 18.16 47 Okeechobee 51 157.4 0 144.12 13.30 48 Orange 432 230.20 199.29 30.97 49 Osceola 104 237.80 210.65 27.24 50 Palm Beach 815 187.80 160.32 27.53 51 Pasco 309 185.90 159.79 26.12 52 Pinellas 690 165.50 139.95 25.60 53 Polk 340 189.10 163.02 26.10 54 Putnam 129 223.10 196.71 26.47 55 St. Johns 178 238.90 219.93 19.00 56 St. Lucie 235 190.90 168.45 22.46 57 Santa Rosa 170 236.20 216.10 20.10 58 Sarasota 249 189.20 166.23 23.01 59 Seminole 250 212.70 188.64 24.09 60 Sumter 62 140.00 117.46 22.5 4 61 Suwannee 92 223.40 203.83 19.66 62 Taylor 32 227.20 208.90 18.33 63 Union 14 214.70 205.59 9.19 64 Volusia 384 180.30 156.20 24.10 65 Wakulla 21 269.00 269.43 0.42 66 Walton 62 245.00 233.37 11.68 67 Washington 41 234.00 204.99 29. 20

PAGE 98

98 Figure 5 2 Spatial d istribution of i mpacts of m ileage f ee of 2.8 c ents/ m ile

PAGE 99

99 Mileage fee of 4.1cents/mile (20% VMT reduction) The impacts of a f lat fee of 4.1 cents/mile are presented in Tables 5 1 5 to 5 1 7 and in Figure 5 3. The impacts among diff erent income groups are similar to those for the 2.8 cents/mile fee, but are more regressive. The difference between the average change in consumer surpluses as a percent age of average income of the lowest income and highest income groups is 1.19, compared to 0.62 for a mileage fee of 2.8 cents/mile. The rural residents again suffer more than the people in urban area s The average change in consumer surplus es are $ 502.54 in the rural area and $ 392.91 in the urban area, compared to $ 238.21 and $ 192.52 fo r the fee of 2.8 cents/mile. As the residents of rural area s drive more than the residents of urban area s the disparities increase as the fee level increases However, the spatial distribution of the impacts is fairly uniform, yielding the same Gini coeff icient of 0. 17 Table 5 1 5 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Av g. c hange in s ocial w elfare ($) $ 0 $ 19,999 210.52 (179.25) 1.66 138.34 72.17 $ 20,000 $ 39,999 301.51 (252.60) 1.00 223.69 77.82 $ 40,000 $ 59,999 407.39 (285.95) 0.81 323.56 83.84 $ 60,000 $ 79,999 506.37 (318.62) 0.71 422.81 83.55 $ 80, 000 $ 200,000 613.16 (367.14) 0.47 542.07 71.10 Table 5 1 6 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfa re ($) Rural 502.54 (392.64) 420.32 82.22 Urban 392.91 (303.21) 317.22 75.69

PAGE 100

100 Table 5 1 7 Average c hanges in c onsumer s urplus, r evenue and social w elfare by county Sl. n o. County n ame No. of HH Avg. c hange in c onsumer s urplus ($) Avg. c hange in r evenue ($) Avg. c hange in social w elfare ($) 1 Alachua 203 532.60 445.50 87.10 2 Baker 35 502.90 436.29 66.62 3 Bay 157 401.30 335.30 66.14 4 Bradford 42 567.00 490.27 76.79 5 Brevard 373 401.40 324.93 76.50 6 Broward 1091 430.60 345.41 8 5.25 7 Calhoun 20 437.10 366.23 70.95 8 Charlotte 129 402.90 317.04 85.93 9 Citrus 299 392.00 312.99 79.08 10 Clay 175 479.70 401.70 78.06 11 Collier 209 383.00 314.05 69.00 12 Columbia 66 440.90 358.41 82.54 13 DeSoto 44 327.50 265.56 61.96 14 Dixie 21 504.70 433.48 71.30 15 Duval 523 456.00 373.04 82.97 16 Escambia 280 432.60 358.54 74.08 17 Flagler 124 492.80 400.01 92.86 18 Franklin 24 432.40 371.10 61.38 19 Gadsden 35 457.50 358.89 98.64 20 Gilcrist 25 417.60 37 3.01 44.59 21 Glades 15 405.60 348.32 57.36 22 Gulf 29 363.40 288.83 74.61 23 Hamilton 25 443.50 373.04 70.50 24 Hardee 22 463.70 385.19 78.60 25 Hendry 38 413.80 348.46 65.39 26 Hernando 109 386.60 302.76 83.88 27 Highlands 224 321.00 256.69 64.32 28 Hillsborough 657 418.90 344.45 74.53 29 Holmes 35 395.30 311.96 83.37 30 Indian River 94 385.10 321.17 64.02 31 Jackson 69 492.50 425.76 66.77 32 Jefferson 35 502.30 419.54 82.79 33 Lafayette 9 427.50 376.72 50.84

PAGE 101

101 Table 5 17 C ontinued Sl. n o. County name No. of HH Avg. change in consumer surplus ($) Avg. change in revenue ($) Avg. change in social welfare ($) 34 Lake 169 359.80 287.35 72.48 35 Lee 420 380.70 309.57 71.15 36 Leon 274 499.60 425.23 74.38 37 Levy 65 574.30 473.14 101.10 38 Liberty 5 475.60 394.32 81.30 39 Madison 28 384.90 319.66 65.27 40 Manatee 180 338.90 275.02 63.89 41 Marion 257 355.00 277.34 77.71 42 Martin 117 412.40 341.91 70.54 43 Miami dade 1256 416.30 331.02 85.32 4 4 Monroe 170 410.60 354.32 56.34 45 Nassau 67 503.00 418.21 84.80 46 Okaloosa 182 489.20 418.82 70.47 47 Okeechobee 51 342.00 285.45 56.58 48 Orange 432 465.60 378.16 87.49 49 Osceola 104 498.50 411.27 87.30 50 Palm Beach 815 377.90 301. 20 76.72 51 Pasco 309 383.90 307.88 76.05 52 Pinellas 690 337.50 266.93 70.64 53 Polk 340 390.70 313.02 77.75 54 Putnam 129 461.60 380.19 81.42 55 St. Johns 178 503.20 431.18 72.02 56 St. Lucie 235 388.50 318.63 69.91 57 Santa Rosa 170 506.10 430.94 75.22 58 Sarasota 249 388.50 315.64 72.92 59 Seminole 250 442.10 366.59 75.53 60 Sumter 62 282.60 216.10 66.50 61 Suwannee 92 479.00 400.88 78.19 62 Taylor 32 522.20 451.91 70.33 63 Union 14 504.30 442.99 61.33 64 Volusia 384 368.10 294.12 74.07 65 Wakulla 21 582.30 515.13 67.17 66 Walton 62 512.30 441.71 70.65 67 Washington 41 492.00 395.52 96.52

PAGE 102

102 Figure 5 3. Spatial d istribution of i mpacts of m ileage f ee of 4.1 c ents/ m ile

PAGE 103

103 5.4.2 Step Mileage Fee The above an alysis reveals that if a high mileage fee is implemented in Florida, its socioeconomic impacts are likely to be regressive. To make the fee less regressive, step mileage fee structure s are tested below. From the descriptive statistics, we observe that the low income group generally travels less than the group with high er income s Therefore, a step fee structure has potential to reduce the regressiveness of the mileage fees in Florida. In the structure, vehicles are charged with a lower fee up to a certain v ehicle mile s traveled ( VMT ) level, and then a higher fee is charged on the miles beyond the threshold. In this study, the threshold VMT level s a re set around the average yearly VMT s of the lowest two income groups and the fees ar e set by trial and error to achieve the desired level of yearly VMT reductions. Step fee to substitute for a flat fee of 2.8 cents/mile (10% VMT reduction) In order to substitute for a flat fee of 2.8 cents/mile, three different step fee schemes are implemented In the first schem e, the threshold is set to 12000 miles, and the low er and high er rate s are set to 2.3 and 3.5 cents/mile respectively. I n the third scheme the threshold is set to 15000 miles and the low er and hi gh er rate s are set to 2.35 and 4.1 cents/mile respectivel y. Scheme 2 is sit uated between schemes 1 and 3. The resulting changes in consumer surplus, revenue, social welfare and percent VMT reduction are presented in Table 5 18 The table shows that each scheme is able to reduce the annual VMT by the same amount, with similar changes in consumer surplus, revenue and social welfare. The distributional impacts of the step fee schemes presented in Tables 5 19 to 5 2 4 are fairly similar to those of the flat fee s Table 5 2 5 compares the average changes in consumer sur plus as a percent age of average income and demonstrates that, although the step fee schemes are still regressive in nature, the

PAGE 104

104 disparity is less than that of a flat fee. It can be observed from the tables that scheme 3 is the best from the equity point of view, w ith the maximum difference between the low er and high er rate s. Table 5 18 Total change in consumer surplus, revenue, social welfare and VMT under different mileage f ee s VMT fee Scheme Threshold (mile) VMT f ee up to threshold (cents/ mile) VMT f ee beyond threshold (cents/ mile) Total change in consumer surplus ($) Total change in revenue ($) Total change in welfare ($) % VMT reduction Flat f ee 2.8 0 2646108 2313856 332252 10.04 Step f ee s cheme 1 10000 2.30 3.50 2855742 2512368 343374 10.04 S tep f ee scheme 2 12000 2.30 3.75 2894320 2547032 347287 10.04 Step f ee scheme 3 15000 2.35 4.10 2911453 2561617 349836 10.03 Table 5 19 Average c hanges in c onsumer s urplus, r evenue and social w elfare by i ncome g roup (Scheme 1) Income g roup Avg. c h ange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 85.69 (121.41) 0.67 64.47 21.22 $ 20,000 $ 39,999 139.70 (185.78) 0.46 114.27 25.43 $ 40,000 $ 59,999 210.00 (214.18) 0.42 180.00 30.01 $ 60,000 $ 79,999 278.00 (244.79) 0.39 247.39 30.61 $ 80,000 $ 200,000 350.60 (284.77) 0.27 325.56 25.10 Table 5 2 0 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare b y l ocation (Scheme 1) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 276.30 (296.18) 248.20 28.16 Urban 202.50 (222.59) 176.86 25.72

PAGE 105

105 Table 5 2 1 Average c hanges in c o nsumer s urplus, r evenue and social w elfare by i ncome g roup (Scheme 2) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19 ,999 82.64 (122.96) 0.65 62.33 20.30 $ 20,000 $ 39,999 137.10 (194.61) 0.45 112.26 24.91 $ 40,000 $ 59,999 210.40 (227.31) 0.42 180.28 30.13 $ 60,000 $ 79,999 282.80 (263.20) 0.40 251.30 31.55 $ 80,000 $ 200,000 363.40 (312.26) 0.28 336. 73 26.73 Table 5 2 2 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 2) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 285.10 (321.72) 255.70 29.45 Urban 203.90 (237.10) 178.20 25.76 Table 5 2 3 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 3) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 81.60 (122.62) 0.64 61.50 20.10 $ 20,000 $ 39,999 134.80 (202.20) 0.45 110.25 24.57 $ 40,000 $ 59,999 208.20 (239.97) 0.41 178.33 29.87 $ 60,000 $ 79 ,999 283.80 (283.63) 0.40 251.86 31.95 $ 80,000 $ 200,000 372.50 (347.32) 0.29 344.58 27.92 Table 5 2 4 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 3) Location Avg. change in c onsumer surplus in $ ( std. de v. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 292.40 (352.30) 261.79 30.64 Urban 203.60 (251.90) 177.98 25.68

PAGE 106

106 Table 5 2 5 Average c hanges in c onsumer s urplus as a p ercent of a verage i ncome by i ncome g roup Income g roup H H# Flat m ileage f ee of 2.8 cents/mile Step m ileage f ee scheme 1 Step m ileage f ee scheme 2 Step m ileage f ee scheme 3 $ 0 $ 19,999 2119 0.84 0.67 0.65 0.64 $ 20,000 $ 39,999 3288 0.49 0.46 0.45 0.45 $ 40,000 $ 59,999 2468 0.40 0.42 0.42 0.41 $ 60 ,000 $ 79,999 1800 0.35 0.39 0.40 0.40 $ 80,000 $ 200,000 3411 0.22 0.27 0.28 0.29 Step fee to substitute for a flat fee of 4.1 cents/mile (20% VMT reduction) Similarly, we have investigate d three different step fee schemes to substitute for a f lat fee of 4.1 cents/mile. The results are presented in Tables 5 2 6 to 5 3 3 All three schemes reduce the annual VMT by 20%, with similar change in consumer surplus, revenue and social welfare. Again, the schemes are found to be less regressive than the fl at fees. It is also found that the scheme with the maximum difference between the lower and higher rates gives the least regressive structure. The general conclusion is that a well designed step fee structure is able to reduce the inequality of the impacts of mileage fees in Florida. Table 5 2 6 Total c hange in c onsumer s urplus, r evenue, social w elfare and VMT under d ifferent m ileage f ee s VMT fee scheme Threshold (mile) VMT f ee up to t hreshold ( c ents/ mile) VMT f ee beyond t hreshold ( c ents/ mile) Total c han ge in c onsumer s urplus ($) Total c hange in r evenue ($) Total c hange in w elfare ($) % VMT r eduction Flat f ee 4.1 0 5445869 4437238 1008631 20.14 Step f ee scheme 4 10000 3.50 5.07 5719999 4697700 1022299 20.14 Step f ee scheme 5 12000 3.50 5.44 5775401 4746358 1029044 20.15 Step f ee scheme 6 15000 3.55 6.00 5802717 4770269 1032448 20.11

PAGE 107

107 Table 5 2 7 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 4) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. d ev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 185.30 (196.45) 1.46 125.78 59.57 $ 20,000 $ 39,999 288.60 (301.88) 0.95 217.10 71.53 $ 40,000 $ 59,999 418.40 (354 .63) 0.83 333.46 84.96 $ 60,000 $ 79,999 545.20 (405.22) 0.77 453.70 91.54 $ 80,000 $ 200,000 693.00 (484.74) 0.53 609.12 83.97 Table 5 28 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 4) Location Avg. c hange in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 551.70 (508.28) 461.75 89.72 Urban 406.20 (378.56) 331.33 74.93 Table 5 29 Average c hanges in c onsumer s urplus, r evenue and s ocial w e lfare by i ncome g roup (Scheme 5) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 182.10 (197.07) 1.43 123.85 58.32 $ 20,000 $ 39,999 285.40 (311.90) 0.94 214.83 70.58 $ 40,000 $ 59,999 418.50 (371.23) 0.83 333.62 84.95 $ 60,000 $ 79,999 551.40 (430.00) 0.78 458.45 92.99 $ 80,000 $ 200,000 711.00 (524.43) 0.54 624.15 86.88 Table 5 3 0 Average c h anges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 5) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 563.70 (543.79) 471.49 92.27 Urban 408.30 (397. 97) 333.43 74.97

PAGE 108

108 Table 5 3 1 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 6) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change i n r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 180.90 (195.44) 1.42 122.92 58.02 $ 20,000 $ 39,999 282.20 (320.31) 0.93 212.42 69.86 $ 40,000 $ 59,999 414.90 (387.24) 0.82 330.82 84.10 $ 60,000 $ 79,999 552.30 (458.45) 0.78 458 .94 93.39 $ 80,000 $ 200,000 724.90 (577.74) 0.56 635.82 89.16 Table 5 3 2 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation ( s cheme 6) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 574.10 (588.89) 479.80 94.38 Urban 408.20 (418.69) 333.51 74.73 Table 5 3 3 Average c hanges in c onsumer s urplus as a p ercent of a verage i ncome by i ncome g roup Income g roup HH# Flat m ileage f ee of 2.8 cents /mile Step m ileage f ee scheme 1 Step m ileage f ee scheme 2 Step m ileage f ee scheme 3 $ 0 $ 19,999 2119 1.66 1.46 1.43 1.42 $ 20,000 $ 39,999 3288 1.00 0.95 0.94 0.93 $ 40,000 $ 59,999 2468 0.81 0.83 0.83 0.82 $ 60,000 $ 79,999 1800 0.71 0.77 0.78 0.78 $ 80,000 $ 200,000 3411 0.47 0.53 0.54 0.56 5.4.3 Mileage Fee B ased on Fuel Efficiency In the previous two sections, we examined the impacts of flat and step mile age fee s As the flat fees are more regressive in nature, step fees are prop osed to reduce the discrepancy However none of the flat and step fees is environment ally friendly because the vehicles with le s s fuel efficiency benefi t more in the mileage fee system Many studies (e.g., Zhang et al., 2009; Larsen et al., 2012 ; Zhang a nd McMullen, 1010)

PAGE 109

109 have also show n concern that a flat mileage fee will lead to less incentive to switch to more fuel efficient vehicles. In order to promote an environment ally friendly policy, fees can be charged based on vehicle fuel efficiency. In this structure vehicle s with MPG lower than the state average are charged more than vehicles with MPG higher than state average However this cause s a concern from the equity point of view as low income households generally own less fuel efficient vehicle s than those with high er income s The refore in this section we evaluate the impacts of th is type of fee s for 10 % and 20 % VMT reduction scenarios. Fees to substitute for a flat fee of 2.8 cents/mile (10% VMT reduction) The a verage fuel efficiency of vehicl es in Florida is 21 MPG. Taking 21 MPG as a threshold we have tested two different fee schemes for this scenario. In the first scheme, the difference between the less efficient vehicle's fee and more efficient vehicle's fee is less than 1 cent, while in t he second scheme the difference is more than 1 cent. The fee structures, total change in consumer surplus, revenue and social welfare are provided in Table 5 3 4 From the table, we can observe that the total change s in consumer surplus, revenue and welfare are similar to th ose of flat fee s The distributional impacts are also similar to those of flat fee s (Tables 5 3 5 to 5 38 ). The gasoline consumptions under different fee schemes are provided in Table 5 39 From the table we can observe that the gasoline c onsumptions are reduc ed by 0.33% and 0.73% from those in the schemes 1 and 2 respectively. As the amount of gasoline consumption is directly related to greenhouse gas ( GHG ) emission, the test confirms that the new fee schemes are more environment ally frie ndly. Moreover, as the new system charges more for less efficient

PAGE 110

110 vehicles, there will be an obvious incentive to switch to vehicles with higher fuel efficien cy Table 5 3 4 Total change in consumer surplus, revenue, social welfare and VMT under different mileage f ee VMT fee Scheme Thres hold (MPG) VMT f ee for MPG> threshold (cents/ mile) VMT f ee threshold (cents/ mile) Total change in consumer surplus ($) Total change in revenue ($) Total change in welfare ($) % VMT reduction Flat fee 2.8 0 264 6108 2313856 332252 10.04 Scheme 1 21 2.55 3.20 2800080 2462436 337643 10.06 Scheme 2 21 2.20 3.78 3021144 2660339 360806 10.02 Table 5 3 5 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 1 ) Income group Av g. change in consumer surplus in $ (std. dev.) Avg. change in consumer surplus as % of avg. income Avg. change in revenue ($) Avg. change in social welfare ($) $0 $19,999 109.60 (94.96) 0.86 81.40 28.24 $20,000 $39,999 156.40 (135.08) 0.52 12 7.69 28.72 $40,000 $59,999 209.30 (150.45) 0.42 180.44 28.94 $60,000 $79,999 260.80 (170.92) 0.37 234.70 26.14 $80,000 $200,000 312.80 (191.47) 0.24 293.84 19.02 Table 5 3 6 Average c hanges in c onsumer s urplus, r evenue and s ocial w elf are by l ocation (Scheme 1 ) Location Avg. change in consumer surplus in $ (std. dev.) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 259.10 (206.24) 233.65 25.48 Urban 201.80 (158.38) 175.93 25.89 Table 5 3 7 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 2 ) Income group Avg. change in consumer surplus in $ (std. dev.) Avg. change in consumer surplus as % of avg. income Avg. change in revenue ($) Avg. change in social welfare ($) $0 $19,999 113.60 (94.86) 0.89 84.90 28.75 $20,000 $39,999 165.70 (138.28) 0.55 135.85 29.89 $40,000 $59,999 222.00 (159.25) 0.44 191.70 30.33 $60,000 $79,999 281.00 (186.20) 0.40 252.75 28.28 $80,000 $200,000 346.30 (213.62) 0.27 324.16 22.24

PAGE 111

111 Table 5 38 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 2 ) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 289.20 (227. 18) 259.46 29.78 Urban 215.10 (170.75) 188.18 26.98 Table 5 39 Average gasoline consumption by income group Income g roup HH# Flat f ee of 2.80 cents/mile (gallon) Based on f uel e fficiency Sc h eme 1 (gallon) Sc h eme 2 (gallon) $ 0 $ 19,999 21 19 394.31 392.89 391.75 $ 20,000 $ 39,999 3288 568.81 566.70 564.40 $ 40,000 $ 59,999 2468 765.36 763.18 760.65 $ 60,000 $ 79,999 1800 950.64 947.64 943.79 $ 80,000 $ 200,000 3411 1208.80 1204.80 1199.50 Total 10429071 10394855 10353424 Fee s to substitute for a flat fee of 4.1 cents/mile (20% VMT reduction) Similarly, we have investigate d t wo different fee schemes based on vehicle fuel efficiency to substitute for a flat fee of 4.1 cents/mile and t he results are presented in Tables 5 4 0 to 5 4 5 The findings are the same as those for the 2.8 cents/mile The general conclusion is that a well designed mileage fee based on vehicle fuel efficiency is able to maintain the desired revenue and reduce GHG emission s Table 5 4 0 Total change in co nsumer surplus, revenue, social welfare and VMT under different mileage f ee VMT fee Scheme Threshold (MPG) VMT fee for MPG> threshold (cents/ mile) VMT fee threshold (cents/ mile) Total change in consumer surplus ($) Total change in revenue ($) Total change in welfare ($) % VMT reduction Flat fee 4.10 5445869 4437238 1008631 20.14 Scheme 1 21 3.83 4.50 5575444 4564875 1010569 20.16 Sc heme 2 21 3.50 5.00 5734839 4711769 1023070 20.10

PAGE 112

112 Table 5 4 1 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 3 ) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus a s % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 212.80 (172.80) 1.67 141.44 71.43 $ 20,000 $ 39,999 306.90 (246.05) 1.01 229.49 77.47 $ 40,000 $ 59,999 414.40 (279.58) 0.82 330.95 83.54 $ 60,000 $ 79 ,999 517.90 (314.25) 0.73 434.05 83.91 $ 80,000 $ 200,000 633.10 (363.84) 0.49 560.70 72.49 Table 5 4 2 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 3 ) Location Avg. change in c onsumer surplus in $ ( std. de v. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 520.80 (393.26) 436.42 84.41 Urban 400.50 (300.07) 325.26 75.29 Table 5 4 3 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 4 ) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 215.40 (169.77) 1.70 123.85 58.32 $ 20,000 $ 39,999 313.40 (244.88 ) 1.03 214.83 70.58 $ 40,000 $ 59,999 423.10 (281.23) 0.84 333.62 84.95 $ 60,000 $ 79,999 532.30 (320.85) 0.75 458.45 92.99 $ 80,000 $ 200,000 658.20 (375.08) 0.50 624.15 86.88 Table 5 4 4 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 4 ) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 543.40 (405.34) 455.41 88.03 Urban 409.90 (305.40) 334.40 75.53

PAGE 113

113 Table 5 4 5 Aver age gasoline consumption by income group Income g roup HH# Flat f ee of 4.10 cents/mile (gallon) Based on f uel e fficiency Sc h eme 1 (gallon) Sc h eme 2 (gallon) $ 0 $ 19,999 2119 341.19 340.01 339.17 $ 20,000 $ 39,999 3288 509.15 507.42 505.80 $ 40, 000 $ 59,999 2468 698.38 696.66 694.98 $ 60,000 $ 79,999 1800 880.09 877.67 875.04 $ 80,000 $ 200,000 3411 1141.10 1137.90 1134.10 Total 9597180 9569291 9540431 5.4. 4 Mileage Fee B ased on Vehicle Type In the last section we examined the impact s of mileage fees based on vehicle fuel efficiency. Alt hough such a fee structure would be environmental ly friendly, it is difficult to implement in practice as the fuel efficiency varies among different vehicles One alternate option is to charge vehicle s based on vehicle type. Generally, larger vehicles consume more fuel and contribute more emissions. According to the Federal Highway Cost Allocation Study (FHW A 2000) the air pollution cost attributable to automobiles is 1.1 cents/mile ; for pickup s and vans the cost is 2.6 cents/mile ; and for vehicles weighing more than 8500 pounds is 3 cents/mile The se rates suggest that pickup s and vans are contributing air pollution at a rate about 2.5 time s higher than cars and th at vehicles weighing more than 8500 pounds are contributing air pollution at a rate about 3 times that of cars Alt hough the report does not give exact values for motorcycle s SUV s truck s and RVs, we assume that the effect of a motorcycle would be about 0.5 times that of a car ; SUV s and tr uck s would be the same as pickup s and vans (i.e., 2.5 times that of a car) ; and RV s would be 3 time s that of car s (as RVs are more than 8500 pounds) In our study, the fees for cars are set to 1.9 0 and 2.76 cents/mile for the 10%

PAGE 114

114 and 20% VMT reduction scen arios, respectively. The fees for other vehicles are obtained by multiplying by the factors provided in Table 5 46. Table 5 4 6 Multiplying factors for different vehicle type Car Van SUV Truck RV Motorcycle 1 (base case) 2.5 2.5 2.5 3 0.5 The total ch ange s in consumer surplus, revenue, social welfare and gasoline consumption under different fee schemes are provided in Table 5 4 7 From the table we can observe that both s chemes produc e more revenue than the flat fee s while also reducing gasoline consu mption by about 1% which is favorable from an emission control point of view. The distributional impacts are very similar to those of flat fees (5 48 to 5 5 1 ). Table 5 4 7 Total change in consumer surplus, revenue, social welfare and gasoline consump tion under different mileage f ee Scenario Fees type Total change in consumer surplus ($) Total c hange in r evenue ($) Total c hange in w elfare ($) Gasoline consumption (gallon) 10 % VMT reduction Flat fee (2.80 cents/mile) 2646108 2313856 332252 10429071 Fee based on veh. type scheme 1 34 63708 30 29526 43 4181 1032 7291 20 % VMT reduction Flat fee (4.10 cents/mile) 5445869 4437238 1008631 9597180 Fee based on veh. type scheme 2 6 514604 53 31646 118 2958 948 5232 Table 5 48 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 1 ) Income group Avg. change in consumer surplus in $ (std. dev.) Avg. change in consumer surplus as % of avg. income Avg. change in revenue ($) Avg. change in social welfare ($) $0 $ 19,999 107. 6 0 (138. 06 ) 0.85 79. 63 28.0 1 $20,000 $39,999 17 4 7 0 (201. 10 ) 0.58 142. 27 32. 46 $40,000 $59,999 25 6 8 0 (248. 86 ) 0.51 219. 0 8 37. 79 $60,000 $79,999 334. 2 0 (277. 40 ) 0.47 297. 00 37. 2 5 $80,000 $200,000 41 7 9 0 (32 6 7 4) 0. 32 38 6 .3 2 31. 60

PAGE 115

115 Table 5 49 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 1 ) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 35 6 5 0 (32 6 90 ) 313. 03 43. 53 Urban 2 39 9 0 (25 4 97 ) 2 09 57 30. 39 Table 5 5 0 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by i ncome g roup (Scheme 2 ) Income g roup Avg. c hange in c onsumer s urplus in $ ( std. dev.) Avg. c hange in c onsumer s urplus as % of a vg. i ncome Avg. change in r evenue ($) Avg. c hange in s ocial w elfare ($) $ 0 $ 19,999 210. 4 0 (22 3 94 ) 1.66 140. 38 70. 04 $ 20,000 $ 39,999 33 2 5 0 (3 29 84 ) 1.10 249. 02 83. 55 $ 40,000 $ 59,999 4 79 80 (408. 31 ) 0.9 5 38 0 8 4 99. 03 $ 60, 000 $ 79,999 62 0 7 0 (456.4 3 ) 0.88 51 6 47 104. 2 0 $ 80,000 $ 200,000 78 3 8 0 (54 5 29 ) 0.60 68 7 72 96. 10 Table 5 5 1 Average c hanges in c onsumer s urplus, r evenue and s ocial w elfare by l ocation (Scheme 2 ) Location Avg. change in c onsumer surplus in $ ( std. dev. ) Avg. change in revenue ($) Avg. change in social welfare ($) Rural 65 8 1 0 (55 5 20 ) 54 0 .4 7 117. 6 0 Urban 45 4 6 0 (428. 17 ) 37 1 6 3 83. 07 5. 5 Summary A m ileage fee is now being considered as the most viable alternative to the conventional gasoline tax. However, there is always a concern from the equity standpoint that low er income group s should not be disadvantaged under the new system. In t his chapter we perform ed a quantitative assessment of the impacts of implementing a mileage fee sys tem in Florida. A regression model wa s constructed using the NHTS 2009 data of Florida Four different mileage fee structures under different scenarios we re tested.

PAGE 116

116 In this study only the state and local portion of the gasoline tax ( i.e. not including th e federal tax) was replaced by a mileage fee and it was found that a flat fee of 1.61 cents/mile would be sufficient to maintain the current level of revenue ( Note that we d id not consider any cost difference for installation and operations ) If the new s ystem requires additional cost s the revenue neutral fee would be higher. W ith a flat fee of 1.61 cents/mile, the average change in consumer surplus as a percentage of average income is negligible in all income groups with those in rural area s receiv ing s lightly more benefi t than those in urban area s although the difference is not significant. Across different counties, the average change in consumer surp lus ranges from $13.92 to $77.93 F lat fee s of 2.8 and 4.1 cent s /mile we re considered to achieve 10% and 20% VMT reductions respectively Under th ese higher fees the impacts are regressive in nature and the disparity increases with the fee rate, with people in rural area s suffer ing more than u rban dwellers However, distributional impacts among counties are fairly uniform with the Gini coefficient being only 0. 17 As the flat fee is regressive in nature, a step mileage fee structure (two mileage fees with the lower being charged for travel within a certain total miles and the higher one for additional m iles) wa s tested From the result s, we f i nd that the step fees are less regressive than the flat fees and are capable of generating the same amount of revenue as the flat fee s Two environmental ly friendly fees (fee based on vehicle fuel efficiency and fe e based on vehicle type) we re tested as well The results reveal that both fee structures are as regressive as flat fees, but they are capable of reducing gasoline consumption, thereby reducing environmental emission s

PAGE 117

117 The purpose of our study is to provid e the policymakers with insight s into different mileage fee structure s and their impacts on society. Based on our empirical study, we f in d that under a flat rate mileage fee, households with less fuel efficien t vehicles benefit, while those with h igher fue l efficien cy vehicles are negatively affected. Thus the flat mileage fees provid e no incentive to use environment ally friendly more fuel efficient vehicle s or to reduce consumer travel Fees based on vehicle fuel efficiency can be favorable from the p ers pective of environmental protection, but as the low income group s own relatively less fuel efficient vehicles, the fees are regressive in nature. The s tep mileage fee structure is a n excellent way to discourage people from unnecessary trips, thereby relie ving traffic congestion and reducing emission s As the average yearly VMT of the low income group is less in Florida, the step fee s are less regressive than the flat fee s A f ee based on vehicle type seems to be better from the marginal pricing point of vi ew, as the larger vehicles generate more externalities (e.g., produce more emissions) than the smaller one s However this fee structure is also as regressive as flat fees. W e conclude that a step fee is the be s t option among the four fee structures exami ned. Furthermore w e believe that i n order to achieve multiple objectives, a more complex fee structure is needed. F or example, the step mileage fee can be charged based on vehicle type. This type of fee will allow the state to generate sufficient revenue, and it will be less regressive in nature and environment ally friendly. An even m ore complex structure would be a step mileage fee based on both vehicle type and vehicle age We would like to mention that the analyses performed here are based on a regress ion model, where the total demand of a household is assumed to be function of

PAGE 118

118 travel cost and other socioeconomic variables. Th is model is adequate for impact analysis of flat f ees and step fees. However, for the fees based on vehicle fuel efficiency and v ehicle type, there is no option to incorporate different fees for different vehicles in this simple model. Therefore, we used weighted fees in the mode l to capture the average impacts It would be more appropriate to use a simultaneous regression model or a discrete continuous model to capture the travel demand changes of different types of vehicles with different fees. Moreover, in our study we have assumed that the vehicle ownership, commuter travel behavior, and land use pattern s would remain the same a fter the implementation of the mileage fee. Our future study will provide a more complete assess ment of the impacts by incorporating all the above mentioned limitations.

PAGE 119

119 CHAPTER 6 CONCLUSIONS This dissertation involved two types of impact studies : one r elated to the effects of site development on vehicular traffic patterns, the other focusing on the socioeconomic impact of a mileage fee on Florida drivers. The first part examined the methodology of traffic impact analysis. More specifically, the link dis tribution percentage method and the special generator method were compared for performing traffic impact analysis for a new site development In addition a detailed analysis of path flow and O D specific link flow distributions led to procedures to enhanc e the methodology. In the second part of the dissertation, we assessed the impacts of implementing mileage fee s in Florida on drivers in different parts of Florida and with different socioeconomic circumstances Based on our empirical study, we observed th at the link distribution percentage method and the special generator method produce similar estimates of traffic impacts caused by hypothetical developments in different scenarios We found that the link percentage patterns obtained for different scenarios we re also similar, which is consistent with the assumption that the link distribution percentages remain the same regardless of the size of the development. Alt hough both approaches are acceptable and produce similar results, for simplicity, we recommen d ed the link distribution percentage method for traffic impact analysis. On another note, the quality of the results produced by the link distribution percentage method depends on how well the trip generation module replicates the real scenario of the mode ling area. With a well developed trip generation module, the ITE trip generation rate does not have to be applied externally. Consequently, the link distribution percentage method will produce accurate estimates.

PAGE 120

120 We work ed extensively on another issue of TIA methodology. B oth the link distribution percentage and special generator methods estimate the number of development trips from the path flow or O D specific link flow information, which is However, it is well known that the se two flow distributions are not unique. As a consequence, the estimated result will be one of a number of many possible solutions. In order to obtain consistent and defendable result s we propose d u s ing the average pat h flow or O D specific link flow distribution as the basis for TIA In the average path flow or O D specific link flow solution all results a re assumed to have an equal probability of occurrence. Consequently, the mean of all the path or O D specific lin k flow solutions, which is essentially the center of gravity of the UE polyhedron, seems a logical selection for the basis of traffic impact studies. We also prove d that t he average distribution is continuous with the assignment inputs and stable. A modifi ed extended hit and run sampling algorithm wa s proposed to compute the average O D specific link flow distribution. Our e mpirical study reveal ed that the results obtained using the average O D specific link flow distribution are significantly different fro m those produced by the entropy maximizing approach W e believe that the proposed basis is more intuitively appealing to practitioners and the result should be easier to defend as we are taking the average of all possible solutions instead of a random so lution or a solution with a very low probability of occurrence However, our proposed method is more computationally intensive and we will be investigating m ore efficient sampling procedure s in the future.

PAGE 121

121 Due to the increasing concern about the adequacy of revenue for highway projects, we also assessed the possibility of implementing a vehicle mile s travel ed (VMT) fee in Florida. Four different mileage fee structure s (flat fee, step fee, fee based on vehicle fuel efficiency and fee based on vehicle type) were evaluated. Based on our study, a flat fee of 1.61 cents/mile is sufficient to maintain the current level of revenue if Florida switches from the gasoline tax to a mileage fee ( Note that here only the state and local portion of the gasoline tax is conv erted to the mileage fee and no additional cost is considered for installation and administration of the mileage fee system ) At this fee, we f ound that the change in consumer surplus as a percent age of average household income is negligible among the diff erent income group s studied The people who live rural area s gain slightly more t han the people who live in urban area s, but when 2.8 and 4.1 cents/mile fees we re tested, we f oun d the fees to be regressive in nature with the disparities becoming greater w ith higher mileage fee s Given that the flat fees we re regressive in nature, step mileage fees we re tested In this structure lower fees we re charged up to a preset threshold mile with higher fees for miles traveled in excess of the threshold. As low in come people travel less than those with higher incomes the fee structure was found to be less regressive in nature. Since none of the above mentioned fee structures wa s environment ally friendly, we examine d two other mileage fee structures: one based on vehicle fuel efficiency and the other based on vehicle type. While b oth of the fee structures reduce d overall household gasoline consumption, both structures we re found to be as regressive as flat fee s

PAGE 122

122 Based on our analysis, we observe d that one cannot a chieve multiple objective s with a simple mileage fee structure. A flat fee of 1.61 cents/mile can immediately be charge d to convert the system from gasoline tax to mileage fee. However, in order to achieve multiple objective s complex mileage fee structure s are required. F or example, step mileage fee s can be charged based on vehicle type. In that case the fee will be less regressive, and it will also reduce total vehicle miles travel overall gasoline consumption and environment al emission s The impacts of c omplex fee structure s will be investigated in the future. We believe that the outcomes of this dissertation will be very helpful for practitioners and policymakers, especially in Florida. Based on our study, the link distribution percentage approach is n ow the recommended method for performing traffic impact studies in Florida. We also hope that the average flow distribution will be used for traffic impact analysis in the future when a more efficient sampling procedure becomes available. The mileage fee impact assessment will provide useful insights about the socioeconomic consequences of implementing a mileage fee system in Florida. While considering implementation of a mileage fee system, the results of our study will guide policymakers in fixing the ap propriate fee structure to generate sufficient revenue without compromising equity and other policy goals (e.g. emission reduction).

PAGE 123

123 APPENDIX A NETWORK CHARACTERISTICS OF S IOUX FALLS NETWORK Link FFTT* Cap** Link FFTT Cap Link FFTT Cap Link FFTT C ap 1 2 3.6 6.02 8 7 1.8 15.68 13 24 2.4 10.18 19 17 1.2 9.65 1 3 2.4 9.01 8 9 2.0 10.10 14 11 2.4 9.75 19 20 2.4 10.01 2 1 3.6 12.02 8 16 3.0 10.09 14 15 3.0 10.26 20 18 2.4 8.11 2 6 3.0 15.92 9 5 3.0 20.00 14 23 2.4 9.85 20 19 2.4 6.05 3 1 2.4 46.81 9 8 2.0 10.10 15 10 3.6 27.02 20 21 3.6 10.12 3 4 2.4 34.22 9 10 1.8 27.83 15 14 3.0 10.26 20 22 3.0 10.15 3 12 2.4 46.81 10 9 1.8 27.83 15 19 2.4 9.64 21 20 3.6 10.12 4 3 2.4 25.82 10 11 3.0 20.00 15 22 2.4 20.63 21 22 1.2 10.46 4 5 1.2 28.25 10 15 3. 6 27.02 16 8 3.0 10.09 21 24 1.8 9.77 4 11 3.6 9.04 10 16 3.0 10.27 16 10 3.0 10.27 22 15 2.4 20.63 5 4 1.2 46.85 10 17 4.2 9.99 16 17 1.2 10.46 22 20 3.0 10.15 5 6 2.4 13.86 11 4 3.6 9.82 16 18 1.8 39.36 22 21 1.2 10.46 5 9 3.0 10.52 11 10 3.0 20.00 1 7 10 4.2 9.99 22 23 2.4 10.00 6 2 3.0 9.92 11 12 3.6 9.82 17 16 1.2 10.46 23 14 2.4 9.85 6 5 2.4 9.90 11 14 2.4 9.75 17 19 1.2 9.65 23 22 2.4 10.00 6 8 1.2 21.62 12 3 2.4 46.81 18 7 1.2 46.81 23 24 1.2 10.16 7 8 1.8 15.68 12 11 3.6 9.82 18 16 1.8 39.36 24 13 2.4 11.38 7 18 1.2 46.81 12 13 1.8 51.80 18 20 2.4 8.11 24 21 1.8 9.77 8 6 1.2 9.80 13 12 1.8 51.80 19 15 2.4 4.42 24 23 1.2 10.16 *: Free flow travel time in minutes **: Link capacity in 10 3 veh/hr

PAGE 124

124 APPENDIX B O D DEMANDS OF SIOUX F ALLS NETWORK O \ D 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 1 0 1 1 3 2 3 3 2 2 4 2 2 3 2 3 3 2 1 2 2 1 3 2 1 2 1 0 1 2 1 2 2 4 2 4 2 1 3 1 1 3 2 0 1 1 0 1 0 0 3 1 1 0 2 1 3 1 2 1 3 3 2 1 1 1 2 1 0 0 0 0 1 1 0 4 3 2 2 0 3 2 2 3 3 4 3 2 4 3 3 5 4 1 1 2 1 3 4 1 5 2 1 1 3 0 2 2 3 4 5 3 2 2 1 2 4 2 0 1 1 1 2 1 0 6 2 3 2 3 2 0 2 4 2 4 2 2 2 1 2 5 3 1 2 3 1 2 1 1 7 3 2 1 2 2 2 0 4 3 5 3 3 3 2 4 5 4 1 2 3 1 3 1 0 8 4 2 2 4 3 4 3 0 4 5 4 3 4 2 3 5 5 3 2 3 2 3 2 1 9 3 2 1 4 4 2 3 3 0 6 4 4 3 3 4 4 4 4 2 3 2 4 3 2 10 3 4 2 4 3 4 5 4 6 0 6 5 3 3 5 5 5 4 4 4 4 5 5 5 11 2 1 2 3 2 1 3 4 4 5 0 3 3 4 4 4 3 1 2 3 2 3 3 2 12 2 1 2 4 2 2 4 3 3 4 3 0 4 3 3 3 2 1 3 4 3 3 4 4 13 3 3 1 4 2 2 2 3 3 6 4 4 0 3 4 3 2 1 2 3 3 4 3 3 14 3 1 1 3 1 1 2 3 4 5 4 3 3 0 4 3 2 1 2 4 3 4 4 2 15 3 1 1 3 2 2 1 2 3 5 3 2 3 4 0 5 4 2 3 4 3 5 3 2 16 2 1 1 2 2 3 3 4 3 6 3 2 3 4 4 0 5 2 4 3 2 3 2 1 17 2 2 1 3 2 3 3 3 2 6 4 3 2 3 4 4 0 1 4 3 2 3 2 1 18 1 0 0 1 0 1 2 3 2 7 2 2 1 1 2 3 4 0 2 4 1 2 1 0 19 3 1 0 2 1 2 4 5 4 5 4 3 2 2 3 4 5 2 0 3 2 3 2 1 20 3 1 0 3 1 3 3 5 3 6 4 3 3 2 3 4 4 2 3 0 3 4 2 1 21 1 0 0 2 1 1 2 4 3 5 4 3 3 2 4 4 5 1 2 4 0 4 2 2 22 3 1 1 3 2 2 3 3 4 6 4 2 3 3 4 4 5 1 3 3 4 0 3 3 23 3 0 1 4 1 1 2 3 5 5 3 4 2 3 4 2 2 1 2 3 2 4 0 2 24 1 0 0 2 0 1 1 2 2 4 3 3 3 2 2 1 1 0 1 2 3 4 3 0 Unit: 10 3 veh/hr

PAGE 125

125 LIST OF REFERENCES Abou Zeid, M., Ben Akiva, M. Tierney, K. Buckeye, K.R. Buxbaum, J.N. 2008. Minnesota p ay as y ou d rive p ricing e xperiment. Transportation Research Record 2079, 8 14 Akamatsu, T. 1997 Decomposition of p ath c hoice e ntropy in g eneral t ransport n etworks. Transportation Science 31 ( 4 ) 349 362. Ban, X., Ferris, M.C. Tang, L. 2009. Risk Neutral Second Best Toll Pricing. Technical Report RPI WP 200902 < http://www.rpi.edu/~banx/publications/RNSBTP.pdf > (a ccessed 26 .05. 2010 ) Bar Gera, H. 2010. Traffic a ssignment by p aired a lternative s egments. Transportation Research Part B 44 ( 8 9 ) 1022 1046. Bar Gera, H. Boyce, D. 1999. Ro ute f low e ntropy m aximization in o rigin b ased t raffic a ssignment. In: Proceedings of the 14th International Symposium on Transportation and Traffic Theory (A. Ceder, ed.) Jerusalem, Israel, Elsevier Science, Oxford, UK, 397 415 Bar Gera, H. Luzon, A. 2 007. Difference among r oute f low s olutions for the u ser e quilibrium t raffic a ssignment p roblem. Journal of Transportation Engineering 133 ( 4 ) 232 239. Beckmann, M., McGuire, C.B. Winsten, C.B. 1955. Studies in the Economics of Transportation. Yale Unive rsity Press, New Haven, Connecticut. Also published as Rand RM 1488 PR, Rand Corporation, Santa Monica, CA Center for Multimodal Solutions for Congestion Mitigation (CMS) 2010 VMT Based Traffic Impact Assessment: Development of a Trip Length Model. Univ ersity of Florida, Gainesville, Florida. City of Orlando 2012 Information a o fficial w eb site < http://www.cityoforl ando.net/econom ic/impactfee.htm> (a ccessed 03 .01. 2012 ) DeCorla Souza, P. 2002. Estimating the b enefits from m ileage b ased v ehicle i nsurance, t axes, and f ees. Transportation Research Record 1812, 171 1 78. Federal Highway Administration (FHWA) 2000 Addendum to the 199 7 Federal Highway Cost Allocation Study, Final report, U. S. Department of Transportation Fisk, C. 1980. Some d evelopments in e quilibrium t raffic a ssignment. Transportation Research Part B 14, 243 255. Fisk, C. Brown, G.R. 1975. A n ote on the e ntropy f o rmulation of d istribution m odels. Operational Research Quarterly 26 ( 4 ) 755 758.

PAGE 126

126 Florida Department of Transportation (FDOT) 1980 Urban Transportation Planning Model Update: Task B, Review & Refinement of Standard Trip Generation Model. Final Report. Fl orida Department of Transportation (FDOT) 1997 Site Impact Handbook. Tallahassee, Florida. Florida Department of Transportation (FDOT) 1997 Documentation and Procedural Updates to the Florida Standard Urban Transportation Model Structure (FSUTMS): FSUT MS Trip Generation Model (GEN). Final Technical Report No. 3. Central Office, Systems Planning. Florida Department of Transportation (FDOT) 2005 2025 Florida Transportation Plan, Office of Policy Planning, Tallahassee, Florida. Florida Department of Tran sportation (FDOT) 2010 Trip Generation Characteristics of Special Generators. Final Report. Research Office, Tallahassee, Florida. Forkenbrock D.J. 2004 Mileage b ased r oad u ser c harge c oncept. Transportation Research Record 1864, 1 8. Institute of Tra nsportation Engineers (ITE) 2008 Trip Generation, 8 th Edition: An ITE Informal Report. Washington, D.C. Institute of Transportation Engineers (ITE) 2010 Transportation Impact Analyses for Site Development: An ITE Recommended Practice. Washington, D.C. Larsen, L., Burris, M., Pearson, D. Ellis, P. 2012. Equity e valuation of v ehicle m iles t raveled f ees in Texas. In: P roceedings of the 91 st Annual Meeting of the Transportation Research Board, Washington, D.C. Larsson, T., Lundgren, J., Patriksson, M. R ydergren, C. 2001. Most likely traffic equilibrium route flows: a nalysis and computation. Equilibrium p roblems: n onsmooth o ptimization and v ariational i nequality m odels. In: Proceedings of Intern ational Workshop in Memory of Marino De Luca (F. Giannessi, A. Maugeri and P.M. Pardalos, eds.) Taormina, Italy, December 3 5, 1998 Kluwer Academic Publishers, Dordrecht, Netherlands, 129 159. LeBlanc, L.J., Morlok, E.K. Pierskalla, W.P. 1975. An e fficient a pproach to s olving the r oad n etwork e quilibrium t raffi c a ssignment p roblem. Transportation Research 9, 309 318. Leung, Y. Yan, J. 1997. A n ote on the f luctuation of f lows u nder the e ntropy p rinciple. Transportation Research Part B 31 ( 5 ) 417 423. Litman, T. 1999. Distance Based Charges; A Practical Strate gy for More Optimal Vehicle Pricing. Victoria Transport Policy Institute, Victoria, BC, Canada.

PAGE 127

127 Lu, S. Nie, Y. 2010. Stability of u ser e quilibrium r oute f low s olutions for the t raffic a ssignment p roblem. Transportation Research Part B 44 ( 4 ) 609 617. Mc Mullen, B.S., Zhang, L. Nakahara, K. 2010. Distributional i mpacts of c hanging from a g asoline t ax to a v ehicle m ile t ax for l ight v ehicles: A c ase s tudy of Oregon. Transport Policy 17, 359 366. Metropolitan Transportation Planning Organization (MTPO) 20 05 Gainesville Urbanized Area Year 2025 Long Range Transportation Plan Update: Gainesville Urbanized Area Model Update. Technical Report 4, Gainesville, Florida Minnesota Department of Transportation (MnDOT) 2006 Mileage Based User Fee Demonstration Pr oject: Pay As You Drive Experimental Findings. Final Report 2006 39A. St. Paul, 2006. National Chamber Foundation (NCF) 2005 Future Highway and Public Transportation Financing: Study Release Event. Executive Summary. National Cooperative Highway Research Program (NCHRP) 2006 Future Financing Options to Meet Highway and Transit Needs. Project 20 24 (49). National Surface Transportation Infrastructure Financing Commission (NSTIFC) 2009 Paying Our Way: A New Framework for Transportation Finance. Final Report of the National Surface Transportation Infrastructure Financing Commission. National Surface Transportation Policy and Revenue Study Commission (NSTPRSC) 2007 Transportation for Tomorrow. Report of the National Su rface Transportation Policy and Revenue Study Commission. Nguyen, S. Dupuis, C. 1984. An e fficient m ethod for c omputing t raffic e quilibria in n etworks with a symmetric t ransportation c osts. Transportation Science 1 8 ( 2 ) 185 202. Oh, J.U., Labi S. Sin ha K.C., 2007. Implementation and e valuation of s elf f inancing h ighway p ricing s chemes, a c ase s tudy. Transportation Research Record 1996, 25 33. Oregon Department of Transportation (ODOT) 2005 and Road User Fee Pilot Progr am: Report to the 73 rd Oregon Legislative Assembly. Salem, Oregon. Oregon Department of Transportation (ODOT) 2007 and Road User Fee Pilot Program: Final Report. Salem, Oregon. Oregon Department of Transportation (ODOT ) 2008 Techniques for Assessing the Socio Economic Effects of Vehicle Mileage Fees. OTREC Final Report. Salem, Oregon.

PAGE 128

128 Pennsylvania Department of Transportation (PENNDOT) 2009 Transportation Impact Fees: Porter, J.D., Kim, D.S., Vergara, H.A., Whitty, J., Svadlenak, J., Larsen, N.C., Sexton, C.B. Capps, D.F. 2005. Development and p erformance e valuation of a r evenue c ollection s ystem b ased on v ehicle m iles t raveled. Transportation Research Record 1932, 9 15. Pug et Sound Regional Council (PSGC) 2008 Traffic Choice Study Summary Report. Seattle, WA. Rockafellar, R.T. Wets, R.J. 1998. Variational Analysis. Springer Verlag, Berlin. Rossi, T., Mcneil, S. Hendrickson, C. 1989. Entropy m odel for c onsistent i mpa ct f ee a ssessment. Journal of Urban Planning and Development 115 ( 2 ) 51 63. Rufolo, A. M., Kimpel, T.J. 2008. e xperiment in r oad p ricing. Transportation Research Record 2079, 1 7. Sana, B., Konduri, K.C. Pendyala, R.M. 2010. Quant itative a nalysis of i mpacts of m oving t oward a v ehicle m ileage b ased u ser f ee. Transportation Research Record 2187, 29 35. Smith M.J. 1979. The e xistence, u niqueness and s tability of t raffic e quilibrium. Transportation Science 13 ( 4 ) 295 304. Smith, R.L 1984. Efficient Monte Carlo p rocedures for g enerating p oints u niformly d istributed over b ounded r egions. Operation Research 36 ( 6 ) 1296 1308. Transportation Research Board (TRB) 2006. The Fuel Tax and Alternatives for Transportation Funding. Special R eport 285. Transportation Research Board of the National Academies. Washington, D.C. Wilson, A.G. 1970. Entropy in Urban and Regional Modeling. Pion, London. Zhang, L. Lu, Y. 2012. Marginal c ost v ehicle m ileage f ee. In: Proceedings of the 91 st Annual M eeting of the Transportation Research Board, Washington, D.C. Zhang, L. McMullen, B.S. 2010. Green v ehicle m ileage f ees: c oncept, e valuation m ethodology, r evenue i mpact, and u ser r esponses. In: Proceedings of the 89th Annual Meeting of the Transportation Research Board, Washington, D.C. Zhang, L., McMullen, B.S., Valluri, D. Nakahara, K. 2009. Vehicle m ileage f ee on i ncome and s patial e quity: s hort and l ong r un i mpacts. Transportation Research Record 2115, 110 118.

PAGE 129

129 BIOGRAPHICAL SKETCH Md Shahid Mamun was born in Satkhira, Bangladesh He received his Bachelor of Science degree in c ivil e ngineering from Bangladesh University of Engineering and T echnology (BUET) in 2000 and h e earned his Master of Applied Science degree in c ivil e ngineering from the Uni versity of Toronto Canada in 2003. He then moved back to Bangladesh to join the faculty of the Department of Civil Engineering at Ahsanullah University of Science and Technology. In 2008, Mamun entered the University of Florida for his Ph.D degree in civ il e ngineering under the supervision of Dr. Yafeng Yin During his Ph.D. program Mamun was appointed as a research assistant and worked on several FDOT pr o jects. His research interests include traffic impact analysis, transportation policy analysis, network modeling and travel demand modeling.