USF Libraries
USF Digital Collections

Danger afoot

MISSING IMAGE

Material Information

Title:
Danger afoot sidewalks, environmental justice, and pedestrian safety in Pinellas County, Florida
Physical Description:
Book
Language:
English
Creator:
Harmak, Craig W
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Automobile accidents
Urban planning
Hazards
Transportation
Scale
Dissertations, Academic -- Geography -- Masters -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Though often taken for granted, few everyday activities involve so much genuine danger as the hazards associated with motor vehicles. Urban areas are built, modified, and/or deconstructed with motoring in mind. Also true is that few are at as much risk, as are those pedestrians who dare to cross paths with motor vehicles. Unfortunately, all too often, pedestrians are casualties of encounters with the ubiquitous automobile. The Tampa-St. Petersburg-Clearwater, Florida metropolitan statistical area (MSA) has recently been deemed, by one study, to be the nation's second most dangerous MSA for pedestrians. Using information on pedestrian/motor vehicle accident sites, sidewalk location and density, and U.S. Census demographic data, this project focuses on Pinellas County--the most densely populated county in the state of Florida. Issues that were investigated in this case study include: (a) the spatial distribution of pedestrian accident risk within the county, (b) the relationship between the presence of sidewalks and Pedestrian Related Motor Vehicle Accidents (PRMVAs), and (c) the environmental justice implications of these PRMVAs. This thesis seeks to identify spatial and socio-economic trends associated with pedestrian accidents and thus provide an improved understanding of the nature of the danger experienced by pedestrians in the heavily motorized world of west-central Florida.
Thesis:
Thesis (M.A.)--University of South Florida, 2007.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Craig W. Harmak.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 161 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001916908
oclc - 181158377
usfldc doi - E14-SFE0002013
usfldc handle - e14.2013
System ID:
SFS0026331:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Danger Afoot: Sidewalks, Environmental Justi ce, and Pedestrian Safety in Pinellas County, Florida by Craig W. Harmak A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Geography College of Arts and Sciences University of South Florida Major Professor: Jayajit Chakraborty, Ph.D. Graham Tobin, Ph.D. M. Martin Bosman, Ph.D. Date of Approval: April 6, 2007 Keywords: automobile accidents, urban planning, hazards, transportation, scale Copyright 2007, Craig W. Harmak

PAGE 2

ACKNOWLEDGEMENTS I would like to thank Dr Jayajit Chakraborty for hi s patience, attention to detail, and expertise in all th ings thesis-related. Without his aid and persistence, this thesis would likely still be far from complete. As an advisor, I recommend him highly. I also thank my committee members, Dr. Graham Tobin, and Dr. M. Martin Bosman. Your advice and recommendations were invaluab le. I very much appreciate the time and effort you have each put into making this thesis a better product. Sincere thanks as well go to Jim Burd and Shannon Whaley at the Fish and Wildlife Research Institute. Also thanks to Clay and Sarah. You each offered solid advice and counsel. Your efforts were very much worthwhile and always appreciate d. Thanks also to my family and friends who never stopped asking, "Are you still working on that?" If nothing else, it always served as a useful bit of insp iration. Sincerely, thanks. Th anks especially to Mom, who has been there without exception, supportive, lo ving, and irreplaceable. It is a wonderful thing to know I always have someplace to go, when I need someplace to go. Finally, thanks to Lisa. You keep me going when I need to keep going, you worry when I struggle, and you care. You are also the best of editors. I Love you.

PAGE 3

TABLE OF CONTENTS List of Tables iii List of Figures v Abstract vii Chapter 1 Introduction and Research Questions 1 Chapter 2 Literature Review 6 2.1 The Demographics of Pedestrians 7 2.2 Environmental Justice and the Pedestrian 8 2.3 Organization of the Built Environment 11 2.4 Pedestrian Perceptions 14 2.5 Politics and Pedestrians 17 2.6 Summary 19 Chapter 3 Data Collection 21 3.1 Area Density 22 3.2 Income Levels within Pinellas County 25 3.3 Pedestrian-Related Motor Vehicl e Accidents 27 3.4 Pinellas County Sidewalks 29 3.5 Pinellas County Roadways 31 Chapter 4 Research Tasks & Methodology 33 4.1 Research Tasks for Question #1 33 4.2 Research Tasks for Question #2 36 4.3 Research Tasks for Question #3 42 Chapter 5 Results: Characteristics of Pedest rian-Related Motor Veh Acc 44 5.1 Temporal Characteristics 45 5.2 Structural Conditions 47 5.3 People and Pedestrian-Related Moto r Vehicle Accidents 50 i

PAGE 4

Chapter 6 Results: Analysis of Sidewalk Density 54 6.1 100-Meter Radius 55 6.2 500-Meter Radius 59 6.3 1-Kilometer Radius 64 6.4 Sidewalk Density Index Statisti cal Analysis 69 Chapter 7 Results: Socio-Demographi c Factors 73 7.1 Results: Non-White 74 100-Meter Radius 75 500-Meter Radius 78 1-Kilometer Radius 82 Non-White Statistical Analysis 87 Summary 90 7.2 Results: Black 91 100-Meter Radius 92 500-Meter Radius 94 1-Kilometer Radius 98 Black Statistical Analysis 102 Summary 105 7.3 Results: Hispanic 106 100-Meter Radius 107 500-Meter Radius 110 1-Kilometer Radius 114 Hispanic Statistical Analysis 118 Summary 120 7.4 Results: Below Poverty 121 100-Meter Radius 122 500-Meter Radius 124 1-Kilometer Radius 127 Below Poverty Statistical Analysis 131 Summary 133 7.5 Socio-Demographic Factors: Summary 134 Chapter 8 Conclusions and Discussion 136 Afterword 144 References 146 Appendices 154 Appendix A 155 Appendix B 156 Appendix C 158 Appendix D 160 ii

PAGE 5

LIST OF TABLES Table 5.3a Pedestrian Desc ription for Pinellas County PRMVAs 50 Table 5.3b Causes for Pinellas County PRMVAs 51 Table 5.3c PRMVAs and Pedestrian Injury 52 Table 5.3d PRMVAs and Pedestrian Fatality 52 Table 6.1a Mean SDI for All 100-Mete r Cells 57 Table 6.1b Mean SDI for Aggregated 100-Meter Cells 58 Table 6.2a Mean SDI for All 500-Mete r Cells 61 Table 6.2b Mean SDI for Aggregated 500Meter Cells 62 Table 6.3a Mean SDI for All 1-Kilometer Cells 66 Table 6.3b Mean SDI for Aggregated 1Kilometer Cells 67 Table 6.4a Logistic Regression for SDI 69 Table 6.4b Ordinary Least Squares Regr ession for SDI 71 Table 7.1a Mean Non-White for All 100Meter Cells 76 Table 7.1b Mean Non-White for Aggregated 100-Meter Cells 77 Table 7.1c Mean Non-White for All 500Meter Cells 80 Table 7.1d Mean Non-White for Aggregated 500-Meter Cells 81 Table 7.1e Mean Non-White for All 1-K ilometer Cells 84 Table 7.1f Mean Non-White for Aggregated 1-Kilometer Cells 85 Table 7.1g Logistic Regression for Non-White 88 Table 7.1h Ordinary Least Squares Regre ssion for Non-White 89 Table 7.2a Mean Black for All 100-Me ter Cells 93 Table 7.2b Mean Black for Aggregated 100-Meter Cells 93 Table 7.2c Mean Black for All 500-Mete r Cells 95 Table 7.2d Mean Black for Aggregated 500-Meter Cells 96 iii

PAGE 6

Table 7.2e Mean Black for All 1-Kilometer Cells 99 Table 7.2f Mean Black for Aggregated 1-Kilomete r Cells 100 Table 7.2g Logistic Regression for Black 102 Table 7.2h Ordinary Least Squares Regression fo r Black 103 Table 7.3a Mean Hispanic for All 100-Meter Ce lls 108 Table 7.3b Mean Hispanic for Aggregated 100-Meter Cells 109 Table 7.3c Mean Hispanic for All 500-Meter Cells 111 Table 7.3d Mean Hispanic for Aggregated 500-Meter Cells 112 Table 7.3e Mean Hispanic for All 1-Kilometer Cells 115 Table 7.3f Mean Hispanic for Aggregated 1-Kilome ter Cells 116 Table 7.3g Logistic Regression for Hispanic 118 Table 7.3h Ordinary Least Squares Regression for Hispanic 119 Table 7.4a Mean Below Poverty for All 100-Meter Cells 123 Table 7.4b Mean Below Poverty for Aggregated 100-Meter Cells 123 Table 7.4c Mean Below Poverty for All 500-Meter Cells 125 Table 7.4d Mean Below Poverty for Aggregated 500-Me ter Cells 126 Table 7.4e Mean Below Poverty for All 1-Kilometer Cells 128 Table 7.4f Mean Below Poverty for Aggregated 1-Kilo meter Cells 129 Table 7.4g Logistic Regression for Below Poverty 131 Table 7.4h Ordinary Least Squares Regression for Be low Poverty 132 iv

PAGE 7

LIST OF FIGURES Figure 3.1a Florida Population Density by County 23 Figure 3.1b Pinellas County Population Dens ity by Census Block 24 Figure 3.2 Pinellas County Per Capita Income 26 Figure 3.3 Pedestrian-Related Motor Vehicl e Accident Locations 28 Figure 3.4 Pinellas County Sidewalks 30 Figure 3.5 Pinellas County Roadways 32 Figure 5.1a PRMVAs Aggregated by Month 45 Figure 5.1b PRMVAs Aggregated by Time of Day 46 Figure 5.1c PRMVAs by Light Conditions 47 Figure 5.2a PRMVAs by Moisture Condition 48 Figure 5.2b PRMVAs by Traffic Control Device 48 Figure 5.2c Pinellas County Site Location Factors 49 Figure 6.1a SDI at 100-Meter Scale 56 Figure 6.1b SDI 100-Meter PRMVA Frequency 57 Figure 6.1c SDI 100-Meter PRMVA Frequenc y Aggregated 58 Figure 6.2a SDI at 500-Meter Scale 60 Figure 6.2b SDI 500-Meter PRMVA Frequency 62 Figure 6.2c SDI 500-Meter PRMVA Frequenc y Aggregated 63 Figure 6.3a SDI at 1-Kilometer Scale 65 Figure 6.3b SDI 1-Kilometer PRMVA Frequency 67 Figure 6.3c SDI 1-Kilometer PRMVA Freque ncy Aggregated 68 Figure 7.1a Percent Non-White at 100-Meter Scale 75 Figure 7.1b Percent Non-White at 500-Meter Scale 79 v

PAGE 8

Figure 7.1c Percent Non-White 500-Meter PRMVA Frequency 80 Figure 7.1d Percent Non-White 500-Meter PRMVA Frequency Aggr 81 Figure 7.1e Percent Non-White at 1-Kilometer Scale 83 Figure 7.1f Percent Non-White 1-Kilometer PRMVA Frequency 85 Figure 7.1g Percent Non-White 1-Kilometer PRMVA Frequency Aggr 86 Figure 7.2a Percent Black at 100-Meter Scale 92 Figure 7.2b Percent Black at 500-Meter Scale 94 Figure 7.2c Percent Black 500-Meter PR MVA Frequency 96 Figure 7.2d Percent Black 500-Meter PRMVA Fr equency Aggregated 97 Figure 7.2e Percent Black at 1-Kilometer Scale 98 Figure 7.2f Percent Black 1-Kilometer PRMVA Fre quency 100 Figure 7.2g Percent Black 1-Kilomete r PRMVA Frequency Aggregated 101 Figure 7.3a Percent Hispanic at 100-Meter Scale 107 Figure 7.3b Percent Hispanic at 500-Meter Scale 110 Figure 7.3c Percent Hispanic 500-Meter PRMVA Frequency 112 Figure 7.3d Percent Hispanic 500-Meter PR MVA Frequency Aggregated 113 Figure 7.3e Percent Hispanic at 1-Kilometer Scale 114 Figure 7.3f Percent Hispanic 1-Kilometer PRMVA Frequency 116 Figure 7.3g Percent Hispanic 1-Kilometer PRMVA Frequency Aggr 117 Figure 7.4a Percent Below Poverty at 100-Meter Scale 122 Figure 7.4b Percent Below Poverty at 500-Meter Scale 124 Figure 7.4c Percent Below Poverty 500-Meter PRMVA Frequency 125 Figure 7.4d Percent Below Poverty 500-Meter PRMVA Frequency Aggr 126 Figure 7.4e Percent Below Poverty at 1-Kilometer Scale 127 Figure 7.4f Percent Below Poverty 1-Kilometer PRM VA Frequency 129 Figure 7.4g Percent Below Poverty 1-Kilometer PRMVA Frequency Aggr 130 vi

PAGE 9

DANGER AFOOT: SIDEWALKS, ENVIRONMENTAL JUSTICE, AND PEDESTRIAN SAFETY IN PINELLAS COUNTY, FLORIDA Craig W. Harmak ABSTRACT Though often taken for granted, few ev eryday activities involve so much genuine danger as the hazards associated w ith motor vehicles. Urban areas are built, modified, and/or deconstructed with motoring in mind. Also tr ue is that few are at as much risk, as are those pedestrians who da re to cross paths with motor vehicles. Unfortunately, all too often, pedestrians are ca sualties of encounters with the ubiquitous automobile. The Tampa-St. Petersburg-Clearwater, Florida metropolitan statistical area (MSA) has recently been deemed, by one study, to be the nations second most dangerous MSA for pedestrians. Using information on pedestrian/motor vehicle accident sites, sidewalk location and density, and U.S. Census demographic data, this project focuses on Pinellas County --the most densely populated co unty in the state of Florida. Issues that were investigated in this case study include: (a) the spatial distribution of pedestrian accident risk within the county, (b) the relationship between the presence of sidewalks and Pedestrian Related Motor Vehicle Acci dents (PRMVAs), and (c) the environmental vii

PAGE 10

justice implications of these PRMVAs. This thesis seeks to identify spatial and socioeconomic trends associated with pedestrian accidents and thus provide an improved understanding of the nature of the danger experienced by pedestrians in the heavily motorized world of west-central Florida. viii

PAGE 11

CHAPTER 1 INTRODUCTION AND RESEARCH QUESTIONS When looking at the city, it is readily apparent that urban spatial structure is intrinsically tied to the networks that li nk the component parts. In early 21st century America, transportation networ ks and automobile roadways, in particular, are the most dominant features influenci ng the expanding shape and size of urban environments. Between 1980 and 2003, total roadway miles in the U.S. have increased by more than 100,000 miles to 3,974,107 miles (US-DOT, 2005). This increase has largely occurred in urban areas which experienced a nearly 50 perc ent growth in roadway mileage. In fact, rural roadways have actually decreased nearly 200,000 miles as rural areas have been engulfed by expanding urban zones (US-DOT, 2 005). This expansion of roadways has been fueled, in part, by a nearly 50 percent increase in the number of motor vehicles plying the nations roadways between 1980 and 2003 (US-DOT, 2005). Motor vehicles are clearly a significant part of the fabric of American life and as su ch, receive substantial attention by policymakers and industry. With all this attention and funding given to motorists, often overlooked are the means for those who must, or merely choose to, travel in a non-motorized fashion. Sidewalks offer, and perhaps even encourage, an alternative mode of transport. While automobile tra ffic arteries are constr ucted and expanded, the

PAGE 12

more environmentally-friendl y options are often forgotte n. At best, the option of improving pedestrian travel routes is given scant consider ation, but simply not deemed worthy, leaving the trans it choices limited. Pedestrians are particularly at risk when moving about in urban environments. When compared to travel by automobile, foot travel is as much as 50 times as dangerous (Aultman-Hall, Kaltenecker 1999). Despite this extreme difference in safety, frequently little is done to improve this situation. A recent study (Ernst, 2004), indicates Florida has the four most dange rous metropolitan areas in the nation (the Orlando MSA ranks first, Tampa-Saint Pete rsburg-Clearwater is second). This study used a weighted formula based on pedestrian miles traveled to account for differing regions and their varied volumes of pedestrian traffic. Even in light of these facts, the streets remain a frequently under-funded and al l too often downright dangerous place for those who walk and bicycle. A useful indicator of recent trends in walking and biking may be seen in U.S. Census data on journey-to-work tre nds. In 1990, there were almost 300,000 people commuting to work in Pinellas County. Of these, just 4.4 percent chose to bike or walk. In 2000, the County had well over 400,000 commuter s but the percentage of bikers and pedestrian commuters declined to a mere 2.8 pe rcent. So severe was this downturn, that despite a robust 40.7 percent increase in th e number of commuters, the absolute number of walking and bicycling commuters actually decreased from 13,149 in 1990 to 11,854 in 2000 (U.S. Census, 2000). While several fact ors could have cont ributed to this precipitous drop, the dire lack of viable pedestrian arteries is likely to be an important contributing factor. 2

PAGE 13

Significant too, are the dangers faced by those without the means to travel by automobile. City dwellers can often be seen traveling afoot or on bicycles, going to work, getting consumables from the market, or simply visiting friends and family. The more affluent may do this because they choose to, or perhaps seek to enjoy the fresh air, and get some exercise, or possibly even due to a concern for the environmental degradation that is otherwise perpetrated by their sport utility vehicle. The less prosperous city dweller may do this because they have no choice. They cannot afford to do otherwise. They have no car. For this lo w-income urbanite, the pathways for walking and biking are not a luxury of recreation or convenience. The foot and bicycle-oriented routes are an absolute necessity to functi on in their everyday life. For these folks, walking and biking will be done whether there are sidewalks and bike lanes present or not. For them, the only question is the level of danger each trip will entail. This predicament may well be an issue of environm ental injustice. In 1994, President Clinton issued Executive Order 12898 defining a fede ral environmental justice policy. This Order mandated that all Federal agenci es consider and act to alleviate any disproportionate adverse health and envir onmental effects that may be incurred by minorities and those of lower income. In 1997, the United States Department of Transportation furthered this with their ow n policy notice stating that environmental justice concerns should be incorporated in all transportation plans, programs, and policies. Issues of concern for environmen tal justice have historically and primarily focused on the disproportionate distribution of dis-amenities such as polluting facilities or locally unwanted land uses (L ULUs). Inequities in the provision of 3

PAGE 14

amenities, such as sidewalks and related passageways for pedestrians, may also represent a violation of environmental just ice principles. Therefore, it is important to examine if racial/ethnic minorities and low-income indi viduals may be adversely affected by the structure of the urban pedestrian environment. In summary, there is an ongoing n eed for research in to the spatial and practical constructs that inhibit pedestrian public safety. In particular, the importance that available sidewalks may have upon the walki ng class has been under-investigated. The thesis project addresses these issues by exploring which locations within the transportation network present disproportiona te hazards for pedestrians. Potential associations between sidewalk availability a nd pedestrian related motor vehicle accidents (PRMVAs) are examined. Finally, the environm ental justice implications of pedestrian accidents and their safety are investigated by analyzing the spatia l relationship between PRMVA incidence and the proportion of r acial/ethnic minorities and low-income individuals. This research project examines seve ral key issues and problems associated with pedestrian and bicyclist safety in an urban environment dominated by the presence of motor vehicles. The following research qu estions were investigated, based on a case study conducted in Pinellas County, Florida: 1. What is the spatial distribution of pe destrian-related motor vehicle accidents (PRMVAs) within the study area? Which parts of the transp ortation network in Pinellas County are more likely to experien ce an incidence of PRMVAs? Under what conditions are PRMVAs most likely to occur? 4

PAGE 15

2. Is there a spatial association between sidewalk/roadway ratio and the incidence of PRMVAs? Are PRMVAs more or less likely to occur in areas where more sidewalks are present relative to roadways? 3. What are the environmental justice implicati ons of PRMVA incidences in the county? Are PRMVAs more likely to occur in ar eas containing a disproportionately large number of racial/ethnic minorities and/or low-income individuals? A combination of seve ral statistical and spatial analytic methods were used in conjunction with geographic informati on systems (GIS) software to explore these research questions. This study focused on pedestrian rela ted motor vehicle accidents. For the purposes of this project, the accidents re ferenced encompass motor vehicle accidents involving exclusively pedestri ans on foot. While consider ation was given to include bicyclists as well, it was determined that the safety concerns of those riding bicycles are substantially different enough to be best se rved by another study focusing exclusively upon bicycle safety. The motor vehicles me ntioned include all forms of motorized vehicles including cars, truc ks, and motorcycles. 5

PAGE 16

CHAPTER 2 LITERATURE REVIEW The material that follows divides th e existing literature related to the subject into several different, though inevitably related issues. The Demographics of Pedestrians is a brief overview of some studies that highlight specific groups of people that may be collectively in greater danger than others. Environmental Justice and the Pedestrian looks at the issue of environmental injustice and literature that has addressed this issue of distributive equity. Organiz ation of the Built Environment focuses on the structural elements that play such an integral role in the safety and welfare of pedestrians. Pedestrian Perceptions highlights why impr oved pedestrian fac ilities are necessary, with special emphasis on how pedestrians may be imperiling themselves and what may be done to improve their safety. The final portion, Politics and Pedestrians, touches on the practical matters of politic s. This section reviews some elements of how, why, (and why not) municipalities choose to direct their often pede strian, pedestrian-oriented efforts. 6

PAGE 17

2.1 The Demographics of Pedestrians Often the best way to better unders tand the nature of a problem is by studying the characteristics of the people invo lved. This section delves into several studies that chose to focus on particul ar classifications of pedestrians. A 1998 study reviewed pedestrian acci dents in Florida (Baltes, 1998). The paper sought to categorize accidents using a number of factors including demographics. Five years of crash data were utilized. The results indicated that elderly pedestrians were the most likely to be involved in an accident and also suffered the most severe injuries. A recent study of Arizona pedestrians considered the race/ethnicity of those who were fatally injured (Campos-O utcalt, Bay, Dellapenna, Cota, 2002). This article, while featuring an et hnic group not appearing in significant numbers in most of Florida (Native Americans), does present the unfortunate, and apparently unevenly distributed incidence of pedestrian fatality amongst minority groups. Using the national Fatality Analysis Reporting System (FARS) coupled with death certificate information, the authors compiled data covering the pe riod of 1990-1996. Their results showed a significantly greater rate of Native American pedestrian fatalities when compared to the population as a whole. Interestingly, alcohol played a disproportionate role in many of these fatalities, the rationale given that becau se no alcohol can be sold on reservations, a long walk is often necessary to get alcohol. This study represents one of the few that sought to incorporate a racia l/ethnic consideratio n into the dangers for pedestrians. Another category of imperiled pedest rians is perhaps surprising to some extent. In a Virginia study, it was discovered that older males, walking in rural areas, and 7

PAGE 18

who have been drinking are in greater da nger than the typical pedestrian (Hebert Martinez, Porter, 2004). The authors review ed crash trends from 1990-1999. While the aforementioned males were most likely to die in an accident, urban areas were the most dangerous, and those aged 5-19 were most likely to be involved in an accident. Children are in increased danger as pe destrians. The demographic agents at issue were investigated in a recent ar ticle featuring four California communities (LaScala, Gruenewald, Johnson, 2004). The notion that school areas feature large numbers of pedestrian children is pursued. Reviewing demographic data, it is discovered that in the communities studied, pedestrian child ren injuries were greatest in areas with high youth populations, more unemployment, fewer high-income households, and more traffic flow. Most of this is not surprising, but the analysis does provide confirmation for what otherwise would be mere assumptions. 2.2 Environmental Justice and the Pedestrian The environmental justice movement in the U.S. and related research has primarily focused on environmental pollution and the siting of undesirable land uses. Presidential Executive Order 12898 mandated th at each Federal agency shall make achieving environmental justi ce part of its mission by identifying and addressing, as appropriate, disproportionately high and advers e human health or environmental effects of its programs, policies, and activitie s on minority populations and low-income populations (Clinton, 1994). This edict se rves to expand the potential reach of environmental justice and the ability to corr ect injustices. The U.S. Department of Transportation further accentuated the potenti al reach of environm ental justice. The 8

PAGE 19

objective of this Order is the development of a process that integrates the existing statutory and regulatory requireme nts in a manner that helps en sure that the interests and well being of minority populat ions and low-income populations are considered and addressed during transportation decision making (US Department of Transportation Order on Environmental Justice, 1997). From the practical founding of th e environmental justice movement through the preponderance of the most current literature, exposure to air, water, and ground pollution has been covered most thor oughly (see reviews by Cutter, 1995; Bowen, 2002). There has also been some attention given to the disproportionate effects levied upon the built environment of minorities and the economically disadvantaged (Greenberg, 1993; Liu, 2001). In one survey of existing literature, studies of environmental justice are subdivided into three categorie s with particular focus on tr ansportation is sues: processbased, benefit-based, and cost-based claims (Schweitzer, Valenzuela, 2004). Processbased issues deal with the decision-making process and th e lack of influence low-income and minorities hold. Benefit-based issues in clude concerns such as access to economic opportunity (or lack thereof) and the equitabl e enforcement of environmental regulations. Cost-based issues refer to the siting of locally unwanted land uses (LULUs) and the inequitable distribution of environmental ex ternalities, including more accidents and greater pollution. While each of these subdivi sions reflects an intriguing aspect of environmental justice, none specifically addr esses the safety of pedestrians in these situations. The focus tends to be not on the dangers that automobiles pose for pedestrians, but rather the s econdary and tertiary effects of motor vehicles. These include 9

PAGE 20

such effects as the resultant air pollutants and the structural damage neighborhoods incur concurrent with inconsiderate placem ent of divisive major roadways. An expansion of environmental justi ce focus is exhibited in a study using the technology of geographic information sy stems (GIS) to look at the impacts of vehicle-generated po llutants and the often neglecte d noise effects (Chakraborty, Schweitzer, Forkenbrock, 1999). This paper highlights the advantag es of using newer technologically based tools to better quantif y, present, and underst and the effects of transportation networks upon minorities and the poor. Another report delves into the politic al structures that direct the planning of urban transportation environments in occas ionally inequitable fashion (Sanchez, Wolf, 2005). Metropolitan Planning Organizations (MPOs) are the conduits whereby funds are routed. These MPOs are, as most any e ssentially political entity, subject to bias, misrepresentation, and outright corruption. The efforts toward a level of effective transportation equity likely require appropriate representation within MPOs. This concern for the structures of politics reflects a philosophical, rather than more directly practical consideration of environmental justice. Yet another environmental justice persp ective is that of differing levels of physical fitness (Greenberg, Renne, 2005). This study offers the reasoning that African Americans residing in self-def ined fair or poor quality neighborhoods were less apt to walk or bike and consequently were less he althy. The notion of walkability is one necessarily fraught with a vague ness inherent in the form. Th ere is also a mention that even if a neighborhood were more walkable, would that necessari ly lead to more walking and ultimately a more physically fit populace? This focus, while differing from 10

PAGE 21

the planned direction of this thesis, does seemingly indicate minorities perceive a greater incidence of substandard pedestrian infrastructure in their neighborhoods. 2.3 Organization of the Built Environment The structural elements of neighborhoods are certainly important when considering pedestrian safety. Rural and small urban areas are examined in a recent article (Ossenbruggen, Pendharkar Ivan, 2001). The authors had an expectation that areas where greater number of pedestrians was traveling would likely result in an increased level of hazard. They compared re sidential zones, shoppi ng areas, and village zones. The village zone represents an area wi th businesses that rely heavily on pedestrian traffic for patronage. Reviewing police reports, they discovered the ar ea with the greatest number of pedestrians was also the safest area. The village area experiences slowest average speeds and has the significant fact or of readily availa ble sidewalks. The significance of the infrastructure and the resultant increased safety should not be underestimated. There is, as with any public expenditure, a concern for the costs of adding to the infrastructure of a neighborhood. In a Norwegian study, a look at cost-benefit analysis is made (Elvik, 2000). This arti cle chooses to focus on what policy makers consider and, as this study i ndicates, fail to consider. The conclusions of this study show that with a failure to account for the valuab le benefits occurring with improvements, such as pedestrian traffic signals and separate bike and pedestrian pa thways, there is an incomplete cost-benefit analysis at work. Al so being overlooked are the more subtle and 11

PAGE 22

less quantifiable health, polluti on, and general welfare effects that occur with increased pedestrian and bicycling activity. A Canadian study (Aultman-Hall, Hall, 1998) sought to improve understanding of where bicyclists were in most peril of accidents. The authors looked at three possible danger classification classes: on -road, off-road pathways, and sidewalks. There were also three separate categories of results for accidents: collision, fall (a noncontact result, often incurred while avoiding a collision), and injury. They attached questionnaires to parked bicycl es at apparent work and sc hool locations, choosing only to focus on those that used their bikes for co mmuting purposes. The response rate was greater than 50 percent, with an impressive total of over 1600 surveys returned. A significant finding was that only 15 percent of self-reported a ccidents had been reported to the police. Notable also was the fact that bicyclists who had experienced fatality, as well as those who perhaps were involved in an incident that prevents them from cycling anymore, were each clearly not receiving th ese surveys. The survey also included questions of bicycling fre quency and length, important because the numbers were crunched to determine the distance traveled pe r event. The results showed that there were statistically significant differences betw een the rates of injury and rates for falls, while the rate for collisions was not statistically significantly different. The statistically safest way to travel is on the road, followed by bike paths, with th e least safe being the sidewalk. An important consideration also is that while sidewalk travel had the greatest chance of danger in the Ottawa area, it was also complete ly bereft of any reported significant injuries. Overall, the rate of bicycle accidents was determined to be between 10 and 41 times as great as automobile travel. 12

PAGE 23

A similar study, done by one of the same authors as the previously mentioned article, looked at the larger metropolis of Toronto (Aultman-Hall, Kaltenecker, 1999). This studys results were largely the same but to a somewhat more severe extent. As it turns out, the dangers were even greater in Toront o for bicyclists. In this case, however, sidewalk accidents indicat ed the greatest likelihood of major injuries as well. In Toronto the bicycle danger was an even larger 26-68 times that of automobile related accidents. Not to be overlooked is the factor of the automobile, and how drivers behave and the resultant effects. In a pape r about the effect of automobile speed and other variables (Garder, 2004), pedestrian re lated accidents in Maine were analyzed. Predicted and actual crash numbers were compared. Predicted numbers were determined by considering pedestrian and vehicular volumes on randomly chosen roadways. Swedish and English prediction models were th en used to compute an expected number. It is explained that there are no U.S. models to perform this task. Actual crash information was gleaned from police reports. These European models delivered results that fell short of actual accidents. This may be due to the significantly different nature of European and American roadways, automobile s, and perhaps even the attitudes of the drivers. The study did yield confirmation that high speeds lead to more pedestrian related accidents. Another apparently related factor wa s that wider roads also were less safe than predicted. Also confirmed was the idea that marked crosswalks are safer than those not marked, though the author is quick to point out, crosswalks are often placed at the safest possible location to begin with. 13

PAGE 24

2.4 Pedestrian Perceptions Pedestrians are often at issue with motorists. Motorists, by the same token, are often at odds with pedestrians. A recent article reviewed these attitudes and suggested some ideas for remedy (Redmon, 2003) According to the author, pedestrians showed a lack of understanding for their rights of way, were disturbed by frequent lack of sidewalks, and did not understand the value of their own visibili ty. Motorists were bothered by too-slow pedestrians, also did not understand their right of way, and notably, often drove because they felt safer drivi ng than walking. This piece offers a good summary of the misunderstandings that ofte n plague automobile-pedestrian encounters and what can be done to improve these situations. An issue can sometimes be the c onditions under which pedestrians and bicyclists view their own visibility. According to a 2004 study (Tyrrell, Wood, Carberry, 2004), pedestrians misunderstanding of their ow n visibility may be a factor in the danger that they find themselves in after dark. In this work, the auth ors had a collection of people walk in place on the shoulder of a road after dark. When these people felt they were visible to an oncoming driver, they pressed a button. The self-reported visibility indicated that the pedestrians typically believed they were visible long before the driver actually saw them. In this experiment, pedest rians on average believed they were visible at 443.0 feet. The drivers on average actua lly saw them at just 251.0 feet an astonishingly significant difference. These findings accentuate the idea that in many cases, pedestrians do not realize the extent of their danger. Perception differs markedly from reality. 14

PAGE 25

In a summary of visibility aids, a recent paper (Kwan, Mapstone, 2004) reviewed findings that shed some light on what may be viable to aid in visibility, and thus safety, for pedestrians. The findings generally yielded predictable re sults. Bright colors are more visible. A flashing or strobe light increased chances of being seen sooner. Red and yellow reflective colors were more visible. In most cases, motion helped to improve visibility as well. As of the time of this review, the authors were not aware on any study that focused on pedestrian/bicyclist-motor ve hicle accidents that compared a before and after effect of visibility aids The results reported in this paper, in conjunction with the aforementioned paper above (Tyrrell, et al), cl early make a case for educating pedestrians and bicyclists of their potential danger as well as creating and/or maintaining an infrastructure that is conducive to their safety. In some cases, there are dangers where they are not expected. By a similar token, there can be cases wher e, although dangers are perceive d, there is little actual occurrence of accidents. In a recent paper (Schneider, Ryznar, Khattak, 2004), this issue was investigated. These authors reviewed accidents occurring around the University of North Carolina at Chapel Hill, a college town with a substantial mix of pedestrian, bicycle, and automobile traffic. The focu s of this study was to develop an idea on a proactive solution to prevent further incidence of pedestrian related accidents. They looked from two angles. On one side, they reviewed police reports to find out about actual accident details. On the other, they ad ministered surveys to determine what people thought of as existing dangerous conditions. Using geographic information systems, they plotted the actual accident sites around campus. They then plotted the perceived danger spots on campus as based on survey respondent s. Using chi-squared analysis, the two 15

PAGE 26

sets of points were compared. The results i ndicated a statistically significant difference between actual and perceived danger. A nearest neighbor cluster analysis revealed similarly significant results. An important c onsideration is that wh ile this study revealed that accidents tended not to ha ppen as often in areas that were perceived to dangerous, this should not imply that the perception of danger was incorrect. Quite the contrary, when people are in areas they think may be mo re dangerous, they tend to behave with an increased level of caution. This greater cau tion would conceivably se em to lead to an actual lessening of accident occurrences. This study appears to back up this notion. An unfortunate potential irony is th e relative increase in occurrence of accidents at less likely places may well lead to a greater concern for thos e areas. This is at the expense of areas that, due to more cautious behavior, experien ce fewer accidents. The result is dangerous places become safer with behavioral modi fications, while apparently safe areas experience lax concern for safety although the area may in fact be intrinsically far more dangerous. The fear of crime committed upon a pe destrian can be oppressive. When on foot, there is no safety barrier around to s upply a feeling of security. A U.K. article looks at how the fear of crime can be impact ed by the addition of improved street lighting (Painter, 1996). This type of study supplies firm credence to the expected idea that lighting improves safety and though in this ca se the danger considered is crime rather than automobiles, the positive impact of im proved infrastructure is still reinforced. In a study of California, New York, and No rth Carolina pedestrian and bicyclist accidents, injury and spatial factors of accident s were reviewed (Stutts, Hunter, 1999). In this case, data were collected via surveys from several hospitals in these locations. 16

PAGE 27

Perhaps the most important result of this paper was a reiteration that police-reported pedestrian and bicyclist accidents are far less th an the actual incidenc e of these types of accidents. The danger is far greater th an any accident statistics indicate. Race and perception of safety is the subject of another paper (Reed, Parikh, 2004). This study revealed the partic ularly salient point th at although minorities tend to have a strong view that they are behavi ng safely as pedestrians, they represent a disproportionate number of PRMVAs. The im plication may be that because minorities are more likely to live in more densely pack ed urban environments, they are more likely to be both more aware of their dangers and be more subject to th e hazards of being a pedestrian in the city. It t oo is conceivable that the infr astructure that may support the urban minority pedestrian (UMP) is insuffici ent thus resulting in greater UMP danger. 2.5 Politics and Pedestrians No matter the good intentions of t hose performing studies, policy changes are largely done via politics. A study was performed recently reviewing responsible parties who determine specific pedestrian policy within Utah municipal governments (Librett, Yore, Schmid, 2003). Using surveys mailed to localities, the authors sought to find if ordinances existed that promoted physical activity (including, but not limited to such things as sidewalks, bike lanes, and greenways). Di viding cities into high, medium, and slow growth, planning ordinances were found to be most common in fast growth areas. In fact, the faster the projected growth, the greater the chance that these sort of ordinances existed. This further emphasizes the point that planning is never more important than in areas experi encing substantial change. Like the area of focus for this 17

PAGE 28

study, Pinellas County is an area undergoing tremendous growth. It is unclear if Pinellas will show the necessary attenti on to such planning issues. Sometimes government needs to be steered in a more productive direction to avoid further deterioration of the environm ent. In a study covering the U.K., transport policy is considered with a look toward sustainable planning (Owens, 1995). Three planning notions are brought forth. Predict and Pr ovide aims to create the infrastructure that is expected to be need ed. This is the policy most likely in place currently (as of the papers publ ication). The Price is Ri ght has users paying for the costs as they go. Here, the nature of cost s and their worth can make this a subtle proposition. The Planning Panacea is the third concept. This is oriented more with an idea of manipulating use toward the operation of less polluting means of transport. Each of these philosophies has valid attributes and in reality can contribu te to practical policy making. The issue becomes one of the appr opriate proportions each idea would best contribute. Seattle has a reputation as a civical ly progressive city. An article explaining Seattles Comprehensive Plan goes far, in theory, at how city planning can work for the benefit of pedestrians (Antiput Gray, Woods, 1996). S eattles plan entailed (as much as was politically possible), a re-cre ation of an early 20th century-style streetcar environment. This scene was one where urban villages were organized in a pedestrianfriendly fashion. Also part of their Ped estrian Programme was a decided effort to incorporate safety and design elements with pe destrians in mind. This demonstrates that under the right circumstances, a pedestri an-oriented municipality is possible. 18

PAGE 29

2.6 Literature Review Summary: Pedestrians and bicyclists may not ev en be entirely aware of the dangers present in an urban environment dominated by motor vehicles. While every pedestrian and bicyclist has likely had so me form of near-miss involving automobiles, the dangers are often gravely underestimated. Neighbor hoods of differing affluence often have a differing infrastructure. Of pa rticular importance for the bicyclist and pedestrian is the presence or absence of safety-oriented features such as sidewalks, bicycle-specific lanes, well-marked roadways, and ample lighting, among other things. There may be a difference between those neighborhoods that have these attributes and those that do not. This difference may well contribute to an increased hazard for the low-income and nonwhite, of whom frequently are less likely to have alternative forms of transportation beyond traveling on foot or via bicycle. The safety concerns for pedestrians ar e well documented in the literature. Many are focused on the adverse health effect s of injured pedestrians (Stutts, Hunter, 1999) or physical fitness concerns (Greenbe rg, Renne, 2005). Govern mental authorities, such as the U.S. Department of Transpor tation and the National Highway Transportation Safety Administration, often simply quantify th e numbers. Some do look at the structural elements such as dangerous intersections (L eden, 2002) or the potential for structural improvements to lessen pedestrian dangers (Clarke, Hummer, Dutt, 1996). What is lacking, however, is a more detailed inve stigation into how the urban hardware, specifically sidewalks, may affect pedestrian safety. Much more research is needed to address specifically how the presence of sidewalks may affect the number, type, and severity of pedestrian related motor vehicle accidents. This study aims to fill some of 19

PAGE 30

that void by addressing how the oft-neglecte d urban amenity--the sidewalk, may impact the safety of pedestrians. Pre-existing environmental justice research has addressed a great many concerns associated with dangerous pollutant s and a wide range of chemical and other environmental hazards. What remains to be investigated is whether the structural elements of the built environment show pa tterns of environmental injustice. The significant question remains: are the neighborhoods of raci al/ethnic minorities and the less affluent subject to stru ctural surroundings that may put the pedestrian in some measure of disproportionate danger? 20

PAGE 31

CHAPTER 3 DATA COLLECTION Pinellas County is located on west-central Floridas tourist-friendly Suncoast. This county, part of the Tamp a-Saint Petersburg-Clearwater MSA, has been well-documented as an exceptionally dangerous place to be a pedestrian (Ernst, 2004). Pinellas has Saint Petersburg occupying the southern portion, and Clearwater, to the north, as the two largest citi es. Pinellas County is a smaller peninsula on the Florida peninsula, bordered by the Gulf of Mexico to the west and Tampa Bay to the east. Due to this confined condition, and what is generally seen as an attractive climat e, it is the states most densely populated county (U.S. Census 2000). The population density of Pinellas County (3,292 per square mile) is twice that of nearby Hillsborough C ounty. It is also nearly three times that of Miami-Dade C ounty and more than tw ice that of south Floridas Broward County. Pinell as is more than eleven times as densely populated than the state as a whole. Significant too is the robust eigh t percent rate of population growth between 1990 and 2000 (U.S. Census, 2000). Th is largely urbanized and geographically restricted area continues to grow at an ove rwhelming pace. This growing population of exceptional density, when coupled with the real ity of an exceptionally high incidence of Pedestrian Related Motor Vehicle Accide nts (PRMVAs), makes Pinellas County a 21

PAGE 32

particularly suitable area for this study. Al so important is the avai lability of relevant data. Pinellas County government has the necessa ry accident data as well as substantial socio-demographic information in various forms. The County also has approximately 4 ,347 miles (6,995 km) of roadways (U.S. Census, 2000) in place to accommodate a resident populace of 921,482 (U.S. Census, 2000), in addition to countless more who visit th e area as tourists. With these types of numbers, it may be expected that there will tend to be a considerable collection of motor vehicle versus pedestrian accidents. This study addresses these issues by reviewing accident site data in conjunction with infrastruc ture data (specifically sidewalk locations). These data were spatially analyzed in conjunc tion with socioeconomic data (specifically assessing the geography of those below povert y) and ethnic/racial data (specifically concerning Blacks, Hispanics, and a more generalized category including all NonWhites). These data were gathered at th e census block group levels of aggregation, obtained from the 2000 U.S. Census of Population and Housing. The intention was to search for possible environmental injustice patter ns reflected in an inferior infrastructure in these neighborhoods, and the possible result ant increased danger for the less affluent. 3.1 Area Density According to the U.S. Census, a 200 4 estimate puts the total population of Florida at 17,397,161 (U.S. Census, 2005) with the population density at 3,292 people per square mile. The following density map (Figure 3.1a), shows population density based on 2004 estimates. The darker shades serve to strongly indicate Floridas 22

PAGE 33

Figure 3.1a: Florida Population D ensity by County, 2000 concentrated urban environments which, not coincidentally, are also among the most hazardous nationwide for pedestrians. Pinellas County is the most densel y populated county in Florida. Figure 3.1b depicts the spatial distribution of population within Pine llas County, on the basis of census block level data. The map indicates th at although the county as a whole is indeed densely settled, there are certainly portions more crowded than others with Saint Petersburg's central areas and Clearwater further north being more densely populated. 23

PAGE 34

Figure 3.1b: Pinellas County Population Dens ity by Census Block, 2000 24

PAGE 35

3.2 Income Levels within Pinellas County Not unlike most any region, the distribution of wealth within the county is varied and clustered. Figure 3.2 disp lays the per capita income as of the 2000 U.S. Census and portrays a useful idea of the distribution of affluence in Pinellas County. The barrier islands on the Gulf coast of the peninsula, as well as the Snell Isle and Shore Acres neighborhoods on Saint Petersburg's Tampa Bay shoreline are areas of higher incomes. Midtown Saint Petersburg and smaller region s in northern Clearwater are less affluent. 25

PAGE 36

Figure 3.2: Pinellas County Per Capita Income, 2000 26

PAGE 37

3.3 Pedestrian Related Motor Vehicle Accident Locations Pedestrian related motor vehicle accident site data were obtained from the Pinellas County Information Systems Office. It was selected from a much larger dataset that included all types of automobile accidents. These data covers all 691 reported traffic incidents involving pedestrians and moto r vehicles from January 1, 2002 through December 31, 2003. Among the included information are injuries and fatalities, the apparent cause of the accident, whether or not drugs and/or alcohol was a factor, the time of day when the accident occurred, and the pa rticular location of the accident (driveway, sidewalk, road, etc.) The spatial distribution of these accidents in Pinellas County is depicted in Figure 3.3. An unfor tunate limitation of the data may lie in the likelihood that a great many accidents may be unreported, partic ularly if no injury is involved (Stutts, Williamson, Whitley, Sheldon, 1990). An assumption here will be that the distribution of these unreported accidents does not differ spatiall y from those accidents that are reported. 27

PAGE 38

Figure 3.3: Pinellas County PRMVA Locations, 2002-2003 28

PAGE 39

3.4 Pinellas County Sidewalks This dataset was also obtained from Pine llas County Information Systems. The data represented indicates improved sidewalks in existence at the time of the survey. This information is typically updated at irregular intervals as new information is gathered. The update utilized for this study was dated July 13, 2005. Previous revisions that might more accurately match the period of PRMVAs in this study were not available. It is assumed, however, that any additions or deletions of sidewalk extents are likely relatively small and as a practical matter, inco nsequential. A review of this sidewalk dataset was completed within ArcGIS soft ware utilizing an aerial photograph backdrop from the 2004 United States Geological Survey. The sidewalk dataset was found to have a number of omissions and a s cattering of duplications. Thes e errors were judiciously corrected utilizing the ArcEd itor extension within an ArcG IS environment. It is important to note that sidewalk improveme nts and maintenance may not be done at predictable intervals. As a result, existing sidewalks may be in markedly differing states of condition. Some may be mapped but may be in such a state of disrepair as to make a simple activity, such as pushing a baby stroller, nearly impossibl e and effectively useless. Without a thorough and (at this point) unf easible round of ground-truthing, these concerns will unfortunately be ignored. The corrected dataset, shown in Figure 3.4, does seem to indicate a spatial pattern largely pa ralleling major automobile arteries. Also, significantly, there is apparently less sidewalk coverage than there is automobile arterial coverage. 29

PAGE 40

Figure 3.4: Pinellas County Sidewalk Extents, 2005 30

PAGE 41

3.5 Pinellas County Roadways These data were obtained from ESRIs StreetMap USA. This set covers all roadways including, Interstates, U.S., state, and county highways, as well as local streets. As a matter of course, this dataset was reviewed in conjunction with 2004 U. S. Geological Survey aerial photogr aphy within an ArcGIS environment. Here it was discovered that although the dataset is largely complete and accurate, it did have several irregularities, inconsistencies and omissions. These issues were rectified using the ArcEditor extension associated with ArcGIS software. While every effort was made to correct flaws within this data set, it should be understood th at minor imperfections could still exist. The map (Figure 3.5) below represents the correc ted roadway layer and clearly indicates the varying, though generally high density of roadways throughout much of Pinellas County. 31

PAGE 42

Figure 3.5: Pinellas County Roadways, 2005 32

PAGE 43

CHAPTER 4 RESEARCH TASKS AND METHODOLOGY Introduction The tasks necessary to investigate the potential relationships identified in the primary research questions required the use of various statistical techniques, as well as the utilization of geographic information syst ems (GIS) software. Each task and its associated steps are outlined in detail below: 4.1 Research Tasks for Question 1: What is the spatial distribution of pede strian-related motor vehicle accidents (PRMVAs) within the study area? Which parts of the transportation network in Pinellas County are more li kely to experience an inci dence of PRMVAs? Under what conditions are PRMVAs most likely to occur? Accident Characteristics The purpose of this task was to better grasp the characteristics of PRMVAs in Pinellas County. The Pinell as County Information Systems Office assembled tabular information covering the 691 reported accidents at issue during the 33

PAGE 44

period of this study. Variables included the date, time, and location of the incident, the road and weather conditions, the age and state of inebriation (or not) of involved parties, and any citations issued, among many other things. The results were statistically analyzed to determine which conditions, circ umstances, and accident characteristics may be significantly associated with the incide nce of PRMVAs. Tabular summaries of the basic elements as gathered and described in the associated police reports were assembled and are presented in Chapter 5. Seasonal variation was the first factor examined. Florida, and in particular Pinellas County experiences an annual regula r influx of temporary residents. Many are relatively unfamiliar with the area. Some are tourists, coming to explore the many area attractions. Many are elderly and perhaps slow to respond to an unexpected pedestrian crossing their path. Seasonal flux in PR MVAs was seen as a great likelihood. The time of day when ac cidents happen was also of interest. Periods when automobile traffic is highest, such as dur ing morning and late afternoon rush hours, is also a period when corresponding pedestrian traffic may also be higher. During these times of day, hazards were greatest. A factor more related to the physical environment rather than the yeoman, workaday, clock-centric world, was that of the light conditions when the PRMVA took place. Dawn, daylight, dusk, and dark (with and without artifici al light) are each characteristics gathered for each accident. Information was also included on whet her there was some structural issue with the roadway, or whether the road was we t, dry, or simply "slippery." There were also additional potential factors included in this database such as whether there might be 34

PAGE 45

environmental influences at work. These ma y have included visually obstructive trees and shrubs, "inclement weather," as well as signage that may have limited the view of those involved in these accidents. Ther e was even "glare" listed as a possible contributing factor. Roadway components that may have played a part in these PRMVAs are listed as well. The presence or absence of traffic signals, stop si gns, crossing guards are mentioned as are if these incidents transpired in a school zone or some sort of no passing zone or other "special speed zone." There was information about what component of the roadway experienced this accident. Inters ections, driveways, entry and exit ramps, bridges, parking lots and severa l other possibilities were included. An important component is that of a ny injuries or fatalities that befell the involved parties. This study did not seek to specifically focus on the casualties of PRMVAs but nonetheless, the happenstance of such misfortune can be telling and serve as a reminder of the very real severity of this hazard. Citations and the presence of any alc ohol or drugs of the involved players was yet another valuable element of informati on gathered in this database. While again, not the focus of this project this knowledge helped to better understand why these accidents materialize. 35

PAGE 46

4.2 Research Tasks for Question 2: Is there a spatial associa tion between sidewalk/roadw ay ratio and the incidence of PRMVAs? Are PRMVAs more or less likely to occur in areas where more sidewalks are present relative to roadways? Spatial Subdivision of the Study Area To best ascertain the spatial incidence of PRMVAs, the County was subdivided into a large number of equally-s ized polygons (zones) of uniform shape. Census enumeration units (e.g., census tracts, block groups) were not appropriate for this purpose because they lack the spatial consis tency required for this study. Boundaries of census units also tend to be coincident with traffic thoroughfares, which tend to be where significant numbers of PRMVAs occur. Large numbers of PRMVAs that occurred on study area borders may compromise an intent of this study. To address these problems, this study utilized polygons of uniform size and shape whose boundaries do not coincide with roadways. A hexagonal grid, when aligned properly, helps minimize the occurren ce of significant numbers of data sites falling on borders. Compared to square or rect angular grid cells, th e hexagonal grid has been deemed particularly statistically eff ective at representing a subdivided space (Olea, 1984, Cox, Cox, Ensor, 1997). The hexagonal shape also replicates a circular zone, but without any voids or overlap of grid cells. This circular cell shape better serves to explain areas, rather than mere poi nts of potential pedestrian danger. Choosing a particular size for the he xagon grid is particularly important. Too large a grid and the risk of overlooking specific dangerous areas may become an 36

PAGE 47

issue. In these instances, dangerous areas may become muted by falling within the same hexagonal cell as areas of relative pedestrian safety. An opposing potential concern may be choosing a cell size that is too small. In these cases, dangerous areas may be too restricted, creating an assumption of nearby safe pedestrian z ones when in fact something so simple (and unfortunately possible) as digitizing variation and errors may place an accident site in an adjacent cel l, rather than reflecting the ac tual location of an accident. Attention must be given to the realities of the digital spat ial placement of the pedestrian accident sites. Specific site locations were determined by Pinellas County Information Systems Office personnel based on information gleaned from offi cial police reports. As a practical matter, reports of this nature might often have somewhat dubious location information (i.e. "near the Burger King on 44th Ave"), making specific accident site geocoding similarly dubious. For this study, hexagonal cell size s were created at three radii: 100 meters, 500 meters, and 1000 meters. These numbers represent the radius from the centroid to the corner of each hexagonal cell. The choice of these radii was based somewhat on an intuitive notion of the appropriate size an urban pedestrian "zone" might conceivably be and is cert ainly open for further discussion and consideration. To create the desired hexagonal grid, it was necessa ry to utilize an older edition of Geographic Information Systems (GIS) software as there was no ArcGIS capability found to create such a grid. ESRI 's ArcView 3.3 was used in conjunction with a "Build Hexagons" extension compatible exclus ively with this version of GIS software. The hexagon grids were constructed at each of the 100 meter, 500 meter, and 1000 meter radius scales and then saved as ArcGIS 9.1-compatible shapefiles. 37

PAGE 48

Within ArcGIS, the aforementioned hex grid shapefiles were clipped to match a 1:40,000 Pinellas County vector shorel ine obtained from the Florida Fish and Wildlife Conservation Commission (FWC). The resulting areas reflect a reasonably accurate shoreline with lakes, rivers, and other significant water bodies removed. It is very important that these water features were removed as the accurate land area of each cell provided a valuable basis fo r subsequent computations. Sidewalks The sidewalk data layer was next added to an ArcGIS session in conjunction with the 100, 500, and 1000 meter hexagonal grid layers. With the GIS map projection set to Albers Equal Area, sidewalk lengths were th en calculated. It was here necessary to subdivide the sidewalks such that the appropriate lengths of sidewalk were spatially assigned to the appropr iate grid cells. The ArcGIS intersect tool served this purpose. Finally, the lengths of sidewalk located in each grid cell, at each scale, were calculated using the analytical capabilities of GIS software. Roadways Roadway data layers were then added. Only those roadways that may reasonably be expected to have sidewalks parall eling them were of interest for this study. Consequently, interstate highway roadways we re removed from this dataset, as these types of arteries are specifi cally designed for motor vehicle traffic at the exclusion of pedestrians. Roadway lengths were assigned and mileage was then calculated using an identical set of processes as in 4.2b. 38

PAGE 49

The Sidewalk Density Index The previous two steps resulted in lengths being summed for sidewalks and roadways within each 100, 500, and 1000 me ter radius hexagonal grid cells in Pinellas County. The Sidewalk Density Index is the result of a simp le formula where the length of sidewalk miles is compared to the total roadway miles inside each grid cell. The index was computed for each hexagonal zone using the following formula: Sidewalk Density Index (SDI) = S/R where S = sidewalk miles within that zone and R = roadway miles within the zone. The result yields a value ge nerally between 0 and 1, with 0 representing a zone that contains absolutely no sidewalks and 1 repres enting a complete match of sidewalks with roadways. Contended here is that in id eal circumstances, every road will have a corresponding sidewalk paralleling it. This resulting Sidewalk Density Index serves to give better indications of wh at specific areas have a relative abundance or paucity of sidewalks. There are rare in stances where the SDI value can exceed 1. These results are most likely to occur when examining pa rk-type settings and were found to be exceedingly rare. Pedestrian-Related Motor Vehicle Accident Sites The geocoded locations of PRMVAs were added next to an ArcGIS session along with the 100, 500, and 100 meter radius hexagon layers. Utilizing the 39

PAGE 50

point-in-polygon functionality of the Hawth's Tools GIS ex tension within an ArcGIS setting, the number of PRMVAs was counted fo r each hexagonal grid cell. This tallied number represents the basis for which pedest rian danger was determined for each cell. The values are the PRMVA frequency values and are simply the number of PRMVAs per hexagonal cell. These accident sites were then separated into two classes: those cells with PRMVAs and those without. Bivariate Regression Analysis Procedure The strength and significance of th e statistical association between the PRMVA frequency (the dependent variable) and each explanatory variable (the SDI, percent Non-White, Black, Hispanic, and Below Poverty) was analyzed at each of the three scales of interest (1 00-meter, 500-meter, and 1-kilometer radius) on all hexagonal grid cells using SPSS statistical software. A si mple application of ordinary least squares (OLS) regression was not considered to be a ppropriate because: (1) the distribution of the dependent variable (PRMVAs) was non-normal; and (2) a large proportion of hexagonal grid cells at each scale showed no PRMVA occurrence (values of th e dependent variable equal zero). In order to obtain the best fit for the data, a twoequation approach was utilized (Duan, Manning, Morris, Newhouse, 1983; Daniels, Friedman, 1999). This twostep approach partitions the dependent variab le into two observed random variables. In the first step, logistic regr ession is used to analyze th e dichotomous event of PRMVA occurrence or non-occurrence (dependent variable coded as either 1 or 0) for each grid cell. In the second step, OLS regression is used to model the frequency (number) of 40

PAGE 51

PRMVAs for only those grid ce lls where accidents occurre d. The general form of the model is as follows: STEP 1: Logistic Regression: x yP yPe 10 )0( )1( log where: y = 0 when accident frequency = 0 and y = 1 when accident frequency >0 STEP 2: Ordinary Least Squares Regression x yE10)( where: y > 1 This implies: E (accidents) = P (accidents>0)*E (accidents|accidents>0). 41

PAGE 52

4.3 Research Tasks for Question 3: What are the environmental justice imp lications of PRMVA incidences in the county? Are PRMVAs more likely to occur in areas containing a disproportionately large number of racial /ethnic minorities and/or low-income individuals? Data of Racial/Ethnic Minority and Poverty Status This was obtained via the application of U.S. Census block group level socio-demographic and racial/eth nic data, as available. Vari ables typically considered in previous environmental justice studies have included the percentage non-white, percentage African-American, pe rcentage Hispanic, and percentage below poverty within a given community. Here too, these demogra phic groups were considered. A process of areal interpolation (Gregory, 2002; Brindl ey, Wise, Maheswarean, Haining, 2005) was applied to transfer population data from censu s enumeration units to the aforementioned hexagonal grid cells created in section 4.2, above. It is here where the significance of an accurate measure of land area becomes clear. The demographic elements within census units were applied proportionately to the appr opriate hexagonal cell. For example, if a hexagonal cell was spatially composed of thr ee different census units, representing areas of say, 50 percent, 35 percent, and 15 percent, then the cell's demographic values were apportioned and calculated in matching proportions. This procedure makes the significant, though practical, assumption that there is a uniform distribution of people within the relevant census unit. 42

PAGE 53

It should be noted here that this study focuses on the socio-demographic attributes of neighborhoods at varying scales. The intent of this study is not to address the specific socio-demographic standing of the individuals involved in PRMVAs. The statistical association between PRMVA incidences and the proportion of individuals in each socio-demographic ca tegory was analyzed at each scale using the two-step regression approach de scribed previously. Binary logistic regression was first used to analyze the effect of the Non-White /Black/Hispanic/Below Poverty percentage on the probability of a grid cell experiencing an accident. Next, Or dinary Least Squares (OLS) Regression was used to examine the linear relationship between PRMVA frequency and Non-White percentages at each scale, on cells with at least one accident. 43

PAGE 54

CHAPTER 5 RESULTS: CHARACTERISTICS OF PE SDESTRIAN-RELATED MOTOR VEHICLE ACCIDENTS Introduction This chapter provides a summary of the basic factors and elements associated with pedestrian-related motor vehicle accident (PRMVA) occurrences in Pinellas County, based upon data assembled by the Pinellas County Information Systems Office. The information was compiled from po lice reports covering th e two years of this study (2002-2003) and clearly is of inconsistent detail and quality. As a result of the nature of this diverse assortment of reports this summary is not concise but rather a useful collection of potentially meaningful facts and figures. PRMVA characteristics can be qui te varied. Seasonal fluctuations, temporal variations, weather, structural, and several human element-type factors can play a part in the occurrence of a PRMVA. The chapter will first look at temporal factors to better understand effects beyond those of a spatia l nature. Next, structural elements will be reviewed to better grasp the character of the place where the PRMVA happened. Finally, the individuals involved in the accide nts will be considered, including factors 44

PAGE 55

such as their sobriety and how the involved party's actions may have played a role, as well as whether any injuries occurred. 5.1 Temporal Characteristics Florida's population varies over the y ear thanks largely to a seasonal influx of temporary residents as well as a consid erable volume of tourists. Florida also experiences weather conditions that, while refreshing and soothing to some, can also make the extended summer months unreasona bly warm for those who may otherwise commute afoot. As Figure 5.1a shows, the period when seasonal residents appear, is also the time where the number of PRMVAs are higher. The complementary period is a trough during the broiling summer months when the often overwhelming heat and vicious daily thunderstorms surely keep many wouldbe pedestrians indoors, thus conceivably reducing summertime pedestrian numbers and c onsequently, diminishing their chances of being involved in a PRMVA. Figure 5.1a: Pinellas County PRMVA occurr ence (2002-2003) aggregated by month 0 10 20 30 40 50 60 70 80 JANFEBMARAPRMAYJUNJULAUGSEPOCTNOVDECPRMVAs 45

PAGE 56

During the two year period this study covers (2002-2003), reported PRMVAs have occurred at all hours. The following chart gives some indication of how pedestrian danger varied throughout the da y (Figure 5.1b). Notably, the number of PRMVAs peaked during the morning and la te afternoon rush hour periods. Not surprisingly, the wee hours of the morning were the least likely to experience an accident. Figure 5.1b : Pinellas County PRMVA occurr ence (2002-2003) aggregated by time-of-day 0 10 20 30 40 50 60 70 0010020030040050060070080090010001100120013001400150016001700180019002000210022002300 HOURPRMVAs It should be noted that the hourly breakdown may reflect how the work and school days function rather than any particul ar visibility issues that may be salient for a particular period of the day. A better way of considering the effects of visibility may be to review whether accidents happened in some disproportion during daylight, dark, or the periods somewhere in between. 46

PAGE 57

As shown in Figure 5.1c, 40 percen t of Pinellas County PRMVAs happen during periods of marginal or no natural light Certainly pedestrians can be difficult to see under the best of circumstances, but remain in substantial danger even when the sun goes down. Figure 5.1c: Pinellas County PRMVA occurr ence by light conditions Daylight 59% Dusk 2% Dawn 2% Dark (Street Light) 30% Dark (No Street Light) 7% 5.2 Structural Conditions The wet and dry road conditions can also be a factor for pedestrians and their encounters with motor vehicles alt hough as we can see below, (Figure 5.2a), the vast majority of PRMVAs occur when condition s are dry rather than wet or "slippery." In all likelihood, forbidding condi tions can be expected to keep those afoot, indoors. 47

PAGE 58

Figure 5.2a: Pinellas County PRMVA occurrence by moisture condition Dry 93.1% Wet 6.6% Slippery 0.3% Traffic control conditions are summ arized in Figure 5.2b. Here we see that areas with no traffic controls are the pl aces with the greatest chance of tallying a Figure 5.2b: Pinellas County PRMVA occurrence by traffic control device No Control 56% Special Speed Zones 9% Traffic Light 22% Stop/Yield Signs 11% All Other 1% Officer/Guard/ Flagperson 1% 48

PAGE 59

PRMVA. What is less than cl ear is why this occurs. Wit hout concise pedestrian travel information, the answer to this question is difficult to deduce. One of the goals of this project is to address the issue of the structural elements potentially contributing to pedestri an safety, specifically sidewalks. The aggregated police reports do help provide some information about the location of PRMVAs with respect to structural roadwa y element at the accident site. Figure 5.2c displays the site location factors and rev eals that although almost half of PRMVAs happen away from intersections, railroad cr ossings and bridges, almost 40 percent are around intersections. Clearly the intersection is a dangerous place to be a pedestrian. Figure 5.2c: Pinellas County Site Location Factors Bridge/ Overpass 0.6% Entrance/ Exit Ramp 0.3% Private Property 0.7% Service Road 0.3% All Other 1.2% Railroad Crossing 0.1% NOT At Intersection/ RR Xing/ Bridge 49.0% At or Influenced by Intersection 39.6% Driveway Access 8.3% 49

PAGE 60

5.3 The People and Pedestrian-Related Motor Vehicle Accidents An interesting, even fascinating, view of pedestrian-related accidents can be gleaned by perusing the police reports' "Pedestr ian Description" category (Table 5.3a). Table 5.3a: Pedestrian Description for Pi nellas County PRMVAs, 2002-2003 PEDESTRIAN DESCRIPTION # Midblock-Other 79 Intersection-Other 69 Vehicle/Turn Merge 54 Midblock Dash 34 Intersection Dash 30 Walking Along Rd/With Traffic 30 Intersection-Driver Violation 27 Ped Not In Roadway 24 Ped Walks into Vehicle-Midblock 14 Other-Weird 12 Play Vehicle Related 12 Backing Vehicle 11 Dart Out Second Half 10 Disabled Vehicle Related 9 Walking Along Rd/Against Traffic 9 Working On Roadway 8 Ped Walks Into Vehicle-At Intersection 7 Dart Out First Half 6 Expressway Crossing 5 Ped Waiting To Cross At/Near Curb 5 Playing In Roadway 5 Exiting/Entering Parked Vehicle 4 Trapped 4 Inadequate Information 3 Mult Threat Not At Intersection 3 Vendor/Ice Cream Truck 3 Cannot Specify 2 Dart Out Cannot Specify 2 Driverless Vehicle 2 Emergency/Police Vehicle Related 2 Hot Pursuit 2 School Bus-Related 2 Commercial Bus-Related 1 Mult Threat-At Intersection 1 Walking To/From Disabled Vehicle 1 50

PAGE 61

The table gathers sometimes interesting information about the circumstances of these PRMVAs. This represents a sort of catch-all for what may have been the unofficial cause of the accident in question. The most common responses ("Midblock-Other," and Intersection-Other") have all the specificity of someone filling out a police report with great haste and little attention to any real consideration. Intriguing though, and included without further explanation, ar e such esoteric de scriptions as "Trapped," and "OtherWeird." The official causes listed on accident reports can also prove useful in helping to understand the nature of the respective accidents. Table 5.3b shows the Table 5.3b: Causes for Pinellas County PRMVAs, 2002-2003 CAUSE # Failed to Yield Right-of-Way 363 Careless Driving 183 No Improper Driving/Action 73 All Other 59 Alcohol-Under Influence 28 Obstructing Traffic 24 Disregarded Other Traffic Control 23 Alcohol and Drug-Under Influence 19 Disregarded Traffic Signal 12 Improper Backing 11 Driver Distraction 6 Exceeded Stated Speed Limit 6 Exceeded Safe Speed Limit 6 Driving Wrong Side/Way 3 Followed Too Closely 3 Fleeing Police 3 Failed to Maintain Equip./Vehicle 3 Drove Left of Center 3 Improper Turn 2 Improper Passing 2 Disregarded Stop Sign 2 Vehicle Modified 1 Improper Load 1 Improper Lane Change 1 51

PAGE 62

breakdown of causes as determined by the au thorities associated with the studied accidents. In many cases, multiple causes were listed for a single PRMVA. "Failure to Yield Right-of-Way" and "Careless Driving" were the most common offenses. There were only 73 incidences of "No Impr oper Driving/Action" among the 691 reported PRMVAs during 2002-2003. In the vast majority of cases, the motorists' actions were considered in some measure causal for PRMVAs in Pinellas County. Decidedly important, but not the part icular focus of this study, is the physical harm incurred upon the pedestrian in these PRMVAs. As shown in Table 5.3c, nearly 80 percent of accidents resulted in so me number of injuries. A significant nine percent of PRMVAs led to fatalities (Table 5.3d). Most pedestrians involved in reported PRMVAs walk away with some measure of injury or worse. Table 5.3c: PRMVAs and Pedestrian Injury, 2002-2003 INJURED # 0 132 19.1% 1 523 75.7% 2 29 4.2% 3 5 0.7% 4 1 0.1% 7 1 0.1% Table 5.3d: PRMVAs and Pedestrian Fatality, 2002-2003. DEATHS # 0 629 91.0% 1 62 9.0% 52

PAGE 63

PRMVA Characteristics Results in Summary Pedestrian-related motor vehicle accidents in Pinellas County occur at all times of day and night, throughout every seas on, under significantly varying conditions, and can be extremely dangerous. PRMVAs ar e more likely to occur during the rush hour periods of the day and less frequently during the summer months. The vast majority of accidents occur when conditions are dry. Most occur away from intersections (although a considerable amount transpire ne ar intersections). The larg est number take place where there is no traffic control mechanism. Some sort of motorist ma lfeasance, inaction, or neglect is generally considered the cause of these accidents. Pedestrian health also fairs poorly when a PRMVA is involved. Simply put, the pedestrian is in dire danger at all times in most places in the heavily motorized urban environment. 53

PAGE 64

CHAPTER 6 RESULTS: ANALYSIS OF SIDEWALK DENSITY This chapter focuses on the relationship between the occurrence of Pedestrian-Related Motor Vehicle Accide nts (PRMVAs) and a measure of Sidewalk Density. This relationship was examined by analyzing the spatial association between these two variables at each of three resolutions in sequence, 100-meter, 500-meter, and 1kilometer radius hexagonal grid cells. In the first portion of this chap ter, a review of the constituent data at each scale is presented. The data were first tabulated at each PRMVA level and then aggregated into appropriate cate gories to better identify any trends. In the second section, the data were further analy zed using statistical inferential tests to investigate the significance of the observed relationships between pedestrian accident occurrence and sidewalk density. An effective way to measure how sidewalks and roadways relate in a given area is the Sidewalk Density Index (SD I), explained previously in Chapter 5. The SDI is calculated as the sum of the lengths of sidewalks divided by the sum of the lengths of roadways, in a given area (in these cases, hexagonal cells at each of the three studied scales). 54

PAGE 65

6.1 Sidewalk Density I ndex: 100-Meter Radius Sidewalk density may be expected to match urban settlement patterns. Such is largely the case in Pinellas County. Figure 6.1a depicts the density of sidewalks within the study area. This version is ba sed on hexagonal cells subdividing the county on a fine 100-meter radius. Darker areas on the map represent a greater density of sidewalks relative to the roadways in th e respective 100-meter radius ce lls. Evident are clustered areas of sidewalks where the population is also predictably more dense. The map displays SDI patterns that woul d seem to indicate that sidewa lk density parallels that of roadway density. There are also a few aberrant cells where the SDI exceeds 1, indicating a greater sum of lengths of sidewalks in a ce ll than that of roadways. Further visual investigation using aerial photography has shown these are typically areas of public spaces such as parkland and areas near schools. 55

PAGE 66

Figure 6.1a: Pinellas County Sidewalk Densit y Index, 100-Meter Radius Cells 56

PAGE 67

The tables to follow indicate how the number of accident occurrences per cell relates to the SDI. In these tables, th e SDI values listed represent the mean SDI for all cells at the corresponding acc ident occurrence level. Table 6.1a describes the relationship between the SDI values and accident frequency for the 100-meter radius cells. At this scale, there is a notable upward trend in the SDI as the number of accidents per cell increases (0 to 4), implying a positive association between accident fr equency and the SDI. With the exception of one cell which experienced five accidents, the SDI va lues in all categorie s exceed that of the Pinellas County mean SDI ( 0.4566) as shown in Figure 6.1b. Table 6.1a: 100-Meter Radius Cells. Mean Si dewalk Density Index (SDI) at each PRMVA level. AX/CELL # CELLS SDI 0 30511 0.4094 1 519 0.6451 2 60 0.7499 3 14 0.7133 4 1 1.0400 5 1 0.2635 Figure 6.1b: 100-Meter Radius Cells SDI and PRMVA/Cell at each PRMVA level. The red horizontal line represents the mean SDI, countywide 0.0000 0.2000 0.4000 0.6000 0.8000 1.0000 1.2000 012345PRMVA/CELLSDI 57

PAGE 68

Table 6.1b and Figure 6.1c show an a ggregated summary of the same data to aid in the identification of a more consistent trend. With this type of classification, outliers are grouped into more ma nageable categories and the resulting trend is decidedly more reliable. Here we can see that without exception, as the accidents per cell increases, so too does the mean Sidewalk Density Index. It should also be noted that at such a fine scale, fully 98 percent of the 31,106 100-mete r radius cells experience no accident occurrence at all. Still, the results are intri guing but it is important to look at other scales to better grasp what these numbers may sugge st. The exceptionally fine radius of these cells may serve to imply a descriptive accuracy beyond the realistic digitization quality of the represented roadways and pedestrian arteries. Table 6.1b: 100-Meter Radius Cells. Mean Si dewalk Density Index (SDI) at aggregated accident levels. AX/CELL # CELLS SDI 0 30511 0.409394 1 519 0.645066 2+ 76 0.740577 Figure 6.1c: 100-Meter Radius Cells SDI and PRMVA/Cell at aggregated PRMVA levels. The red horizontal li ne represents the mean SDI, countywide 0.0000 0.1000 0.2000 0.3000 0.4000 0.5000 0.6000 0.7000 0.8000 012 +PRMVA/CELLSDI 58

PAGE 69

6.2 Sidewalk Density Index: 500-Meter Radius Figure 6.2a shows the same basic SDI dataset as Figure 6.1a, except the data were interpolated/extrapo lated using 500-meter radius cells At this coarser scale, each full hexagon (unreduced by substantial water bodies) covers almost 650,000 square meters in area. This is substantially larger than the area of a mere 26,000 square meters occupied by a 100-meter radius hexagon but considerably smaller than the nearly 2.6 million square meter area covered by the 1-k ilometer radius hexagons. With an increase in radii, the areas of hexagons, as with a more traditional square grid, increase exponentially. 59

PAGE 70

Figure 6.2a: Pinellas County Sidewalk Densit y Index, 500-Meter Radius Cells 60

PAGE 71

At the 500-meter radius, it was reasona ble to expect a s ubstantial decrease in the number of cells with no PRMVAs, and i ndeed this was the case. Of the 1,458 cells of a 500-meter radius, just 75 percent have a zero value or no accidents (Table 6.2a). While still a large number, it is a notabl e reduction from the 98 percent cells without accidents observed at the 100-meter scale. When looking at all the mean SDI values, there is some corresponding upw ard inclination, but the outlier s mask much of the real variation. Notably, however, the mean SDI at each PRMVA level with at least two accidents exceeds the countywide SDI aver age of 0.4566 by significant margins. Table 6.2a: 500-Meter Radius Cells. Mean Si dewalk Density Index (SDI) at each PRMVA level. AX/CELL # CELLS SDI 0 1104 0.3292 1 200 0.4287 2 80 0.5316 3 31 0.5314 4 18 0.6072 5 11 0.5866 6 6 0.6984 7 1 0.5771 8 4 0.5265 9 2 0.7291 15 1 0.9000 The bar graph (Figure 6.2b) reveals this upward trend in the association between PRMVAs and sidewalk density thoug h the trough at the higher accident levels may serve to obscure th is pattern slightly. 61

PAGE 72

Figure 6.2b: 500-Meter Radius Cells SDI and PRMVA/Cell at each PRMVA level. The red horizontal line r epresents the mean SDI, countywide. 0.0000 0.1000 0.2000 0.3000 0.4000 0.5000 0.6000 0.7000 0.8000 0.9000 1.0000 012345678915PRMVA/CELLSDI An aggregated version of this same 500-meter data, (Table 6.2b), reveals a pattern that is far more consistent and clearly indicates a positive co rrespondence between accident prevalence and sidewalk presence. Here, the SDI values yield a mean higher than the countywide average (0.4566) by a substantial margin at the PRMVA classifications with at least two accidents. Table 6.2b: 500-Meter Radius Cells. Mean Si dewalk Density Index (SDI) at aggregated accident levels. AX/CELL # CELLS SDI 0 1104 0.3292 1 200 0.4287 2-4 129 0.5421 5+ 25 0.6274 When a bar chart (Figure 6.2c) is produced, the clarity of the positive association between and PRMVA/cell is more apparent when then PRMVA values are aggregated. 62

PAGE 73

Figure 6.2c: 500-Meter Radius Cells SDI and PRMVA/Cell Aggregated -The red horizontal line represent s the mean SDI, countywide 0.0000 0.1000 0.2000 0.3000 0.4000 0.5000 0.6000 0.7000 012 45 +PRMVA/CELLSDI 63

PAGE 74

6.3 Sidewalk Density Index: 1-Kilometer Radius The final scale that this study cove rs is based on hexagonal grid cells at the 1000-meter radius (Figure 6.3a). A concern at this scale is that the substantially large cell size may mask the effects of particular areas, and at a fully one kilometer radius, these cells are fairly expansive. As noted ea rlier, hexagonal cells of this size have an area that is four times (2,600,000 versus 650,000 square meters) as expansive as that of the 500-meter cells. When compared to the 100-meter radius cells, the difference is monumental. The 1-kilometer cells have 100 tim es the area when contrasted with that of 100-meter cells. Notably however, the coar ser 1-kilometer scale results in only 55 percent of the cells without a single PRMVA. 64

PAGE 75

Figure 6.3a: Pinellas County Sidewalk Density Index, 1 Kilometer Radius Cells 65

PAGE 76

When looking at the mean SDI values for each accident level at the 1000meter radius, (Table 6.3a), a linear trend between SDI and PRMVA frequency is less obvious. Table 6.3a: 1 Kilometer Radius Cells. Me an Sidewalk Density Index (SDI) at each accident level. AX/CELL # CELLS SDI 0 237 0.2761 1 62 0.4504 2 37 0.4466 3 26 0.4105 4 24 0.4625 5 7 0.5053 6 11 0.4467 7 2 0.3491 8 5 0.5350 9 2 0.3330 10 3 0.6906 11 4 0.6183 12 2 0.1708 13 1 0.7826 15 1 0.8414 16 1 0.5535 20 1 0.2808 23 2 0.8917 When these 1-kilometer radius values are summarized graphically, evidence of positive association is present though the trend is marred by some aberrant elements (Figure 6.3b). 66

PAGE 77

Figure 6.3b: 1 Kilometer Radius Cells SDI and PRMVA/Cell at each PRMVA level. The red horizontal line represents the mean SDI, countywide 0.0000 0.1000 0.2000 0.3000 0.4000 0.5000 0.6000 0.7000 0.8000 0.9000 1.0000 01234567891011121315162023PRMVA/CELLSDI The aggregation of values is only sligh tly useful at this 1000-meter scale. Table 6.3b and Figure 6.3c demonstrate that hi dden amongst the outliers, aberrations, and oddities, there is only a modestly identifia ble consonant trend be tween the increasing volume of accidents in a cell and an increasing SDI value. Table 6.3b: 1-Kilometer radius cells. Mean Sidewalk Density Index (SDI) at aggregated accident levels. AX/CELL # CELLS SDI 0 237 0.2761 1 62 0.4504 2-4 87 0.4402 5-9 27 0.4626 10+ 15 0.6491 67

PAGE 78

Figure 6.3c: 1000-meter radius cells SDI and PRMVA/Cell Aggregated 0.0000 0.1000 0.2000 0.3000 0.4000 0.5000 0.6000 0.7000 0 1 2-4 5-9 10+PRMVA/CELLSDI The results of the preliminary analysis of these three scales serve to remind that the choice of proper geographi c resolution is imperative to producing appropriate and viable results. The same three scales and methodology were used to analyze the socio-demographic factors and investigate potential environmental justice implications for the pedestrian in Chapter 7. 68

PAGE 79

6.4 Sidewalk Density Inde x: Statistical Analysis The previous portion of this analys is examined sidewalk density only on the basis of descriptive measures and summary statistics. To best examine the strength and significance of the statistical associ ation between the SDI and PRMVA frequency, two separate tests of statistic al inference were performed us ing SPSS statistical software. As described in Chapter 4, a two-step appro ach was utilized because a large proportion of grid cells experienced zero accid ents. Binary logistic regre ssion was first used to analyze the effect of sidewalk density on the probability of any grid cell experiencing an accident. Next, Ordinary Least Squares (OLS) Regression was used to examine the linear effect of the SDI on the number of pedestrian accidents at each scale, for cells that experienced one or more accidents. STEP 1: Logistic Regression The dependent variable for the logi stic regression analysis (accident occurrence) was treated as a dichotomous even t with each cell coded as either 1 or 0 to represent the presence or ab sence of a PRMVA, respectiv ely. Table 6.4a shows the results of this analysis. Table 6.4a: Logistic Regression of SDI on PRMVA Occurrence. SDI N B sig exp(B) 100M 31,106 0.006 0.156 1.006 500M 1,458 1.549 0.000 4.705 1KM 428 2.485 0.000 12.001 69

PAGE 80

For each logistic model, the natural logarithm of the odds that a grid cell experiences an accident is assumed to be a linear function of the independent variable (sidewalk density) and the maximum likelihood method is used to estimate the model. The table describes the total number of cells at the described scale (N), the va lue of the logi stic regression coefficients (B) for the independent variable and the observed level of significance (pvalues) for each model. To further aid in in terpreting the regression results, the table also provides the odds ratio or the exponent of the regression coefficient produced by each logistic model [exp(B)]. This ratio may be interpreted as a multiplier of the odds of an accident occurring in a grid cell. When th e value exceeds 1, the odds of experiencing a PRMVA increase that many times for each un it change of the SDI in the model. As the table indicates, at the 100meter scale, the association between accident occurrence and sidewalk density is not statistically significant (p>.10). At the 500-meter and 1-kilometer scales, however, the effects are statistica lly significant (p<.01) and positive. Thus, at the 500-meter radius scale, a one-unit incr ease in SDI boosts the odds of a PRMVA occurrence by a factor of 4.7. At the 1-kilometer scale the odds of a PRMVA occurrence increase by a staggering 12 times for each one-unit increase in the SDI. STEP 2: Ordinary Least Squares Regression The second step in this two-equatio n analysis focuses on only those cells that experienced at least one PRMVA a nd uses OLS Regression to analyze the relationship between the number of accident s (the dependent variable) and SDI (the independent variable). Tabl e 6.4b summarizes the results of these regression models for 70

PAGE 81

each scale of study and provides the sample si ze, the least squares coefficients for the independent variables, and the related levels of significance. Table 6.4b: Least Squares Regression for Si dewalk Density Index on PRMVA frequency for those Cells with at Least One Accident. SDI N B sig 100M 595 0.090 0.091 500M 354 1.552 0.000 1KM 191 3.914 0.002 When grid cells are considered at the 100-meter radius, the relationship between PRMVA frequency and the SDI is statistically significant and positive only at the .10 level of significance (p<.10). At th e 500-meter and 1-kilometer radius scales, the regression coefficients produced are each larg er and statistically significant (p<.01). A one-unit increase in the SDI leads to an in crease of about 1.5 accidents at the 500-meter scale and almost four accident s at the 1-kilometer scale. 71

PAGE 82

Sidewalk Density Index Results in Summary As scales increase in size, the a ppearance of trends and tendencies also increases. When looking at the SDI data thr ough the aid of aggrega tion, there appears to be some positive association between SDI and PRMVA occurrence, particularly at the 1kilometer, and even more so, the 500-meter ra dius scales. The density of sidewalks is consistently greater in areas that expe rience PRMVAs compared to areas without reported accidents. The two-step regression an alyses confirm that associations between accident occurrence and sidewalk density are both positive and, in two coarser scales, statistically significant, even at the .01 le vel of significance. The probability of PRMVA occurrence and number of such occurrences both increase significantly as sidewalk density increases, at the 500-meter and 1-kilome ter scales. As the scale increases so too do the coefficients, implying a stronger effect of sidewalk density upon PRMVA occurrence at coarser scales. 72

PAGE 83

CHAPTER 7 RESULTS: ANALYSIS OF RACE/ETHNICITY AND POVERTY STATUS This chapter examines the envi ronmental justice implications of pedestrian-related motor vehicle acciden ts (PRMVAs) by analyzing several sociodemographic categories and their relationshi p to PRMVAs. The following variables are those typically utilized in environmental justice research including: Non-White, Black, Hispanic, and Below Poverty. It is acknowle dged that these catego ries may represent groups of sometimes significant overlap. The goal here is not to examine interrelationships between these groups in te rms of accident occurr ence, but rather to isolate and analyze each of these factors w ith respect to the geographic distribution of PRMVAs within the study area. The following sections display layouts with the spatial distribution of the aforementioned socio-dem ographic groups. For each individual group, a map was produced with the relevant data inte rpolated as necessary. The sections that follow are similarly structured to that of the SDI results section (Chapter 6). It should be reiterated that the issue being studied is NOT the race/ethnic ity or poverty status of the specific pedestrians who are involved in these PRMVAs, but rather the sociodemographic characteristics of the neighborhood in which the PRMVA occurs, as defined 73

PAGE 84

by hexagonal grid cells at various scales. Th e analyses presented in this chapter are based on data gathered at the census block group level. 7.1 Results: Percent Non-White This section seeks to uncover th e spatial relationship between PRMVA occurrence and the proportion of Non-White individuals in Pi nellas County. Specifically, the objective was to determine if PRMVAs are more likely to occur in those areas with a disproportionately high number of Non-White residents. Th e association was examined at the 100-meter, 500-meter, and 1-kilometer radi us scales similar to the methods utilized with the SDI in Chapter 6. Figures 7.1a, 7.1b, and 7.1d each display the distribution of Non-White population as a percentage of the total populatio n, at three hexagonal grid scales. At the 100-meter radius scale, the di stribution largely follows that of the census block groups on which this hexagonal display is based. At th e 500-meter scale, smaller areas of unique character begin to be blended with neighbori ng cells of perhaps differing character. At the 1-kilometer scale, a significant degree of a gglomeration has served to mask all but the most essential characteristics. In Pine llas County, as of the 2000 U.S. Census, 14.1 percent of the population listed themselves as Non-White. 74

PAGE 85

Figure 7.1a: Pinellas County Percentage NonWhite, 100-Meter Radius Cells 75

PAGE 86

Non-White: 100-Meter Radius When PRMVA occurrence is examined at three different scales, some interesting differences begin to emerge. As shown in Table 7.1a, at the 100-meter scale, 98 percent of the 31,106 hexagonal cells expe rienced zero PRMVAs during the period studied. Among grid cells with one or more PRMVAs, there are few discernable patterns or trends. Notably, four out of five frequency levels experiencing accidents indicate a mean Non-White percentage of population greater than the co rresponding Pinellas County average of 14.1 percent. Table 7.1a: 100-Meter Radius Cells. Mean percentage of Non-White population at each accident level. AX/CELL # CELLS NON-WHITE 0 30511 11.0% 1 519 20.3% 2 60 16.8% 3 14 22.8% 4 1 9.8% 5 1 31.6% When accident values are viewed in aggregate at the 100-meter scale, no consistent positive trends are observed in the re lationship. It is abundantly clear that grid cells with at least one accident are likely to have a disproportionately Non-White percentage of population (Table 7.1b). 76

PAGE 87

Table 7.1b: 100-Meter Radius Cells. Mean per centage of NonWhite Population at aggregated accident levels. AX/CELL # CELLS NON-WHITE 0 30511 11.0% 1 519 20.3% 2+ 76 18.0% 77

PAGE 88

Non-White: 500-Meter Radius The countywide Non-White percenta ge distribution with the 500-meter radius scale grid cells differs from the 100-me ter view (Figure 7.1b). At this scale, the spatial delineation no longer matc hes that of the constituent census enumeration units. Wider visible spatial trends ar e gained at a likely loss of finer neighborhood details. At the 500-meter radius scale, a positive association between PRMVAs and Non-White percentage becomes evident. 78

PAGE 89

Figure 7.1b: Pinellas County Percentage Non-Wh ite, 500-Meter Radius Cells 79

PAGE 90

When looking at each accident level individually, the relationship between the percent Non-White and PRMVA occurrence is not clearly apparent (Table 7.1c). The bar graph (Figure 7.1c) also fails to show any significant trends. Table 7.1c: 500-Meter Radius Cells. Mean per centage of Non-White Population at each accident level. AX/CELL # CELLS NON-WHITE 0 1104 8.5% 1 200 15.1% 2 80 16.7% 3 31 18.6% 4 18 35.4% 5 11 26.3% 6 6 22.0% 7 1 8.0% 8 4 27.0% 9 2 15.9% 15 1 11.0% Figure 7.1c: 500-Meter Radius Cells. Mean percentage of Non-White Population at each accident level. The red horizontal line represents the countywide mean. 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0% 012345678915PRMVA/CELLPCT NON-WHITE 80

PAGE 91

When the data are aggregated, a positive associati on between the two variables emerges (Table 7.1d). As the aggreg ated table exhibits, grouping at specific levels can be a useful aid in helping to reveal trends. When this information is placed in graphical form, the trend and direction are dis tinctly evident (Figure 7.1c). The effects of changing scales appear significant. When aggregated, areas with PRMVA occurrence (AX/CELL 1) each exceed the county average for percent Non-White (14.1 percent), while grid cells without PRMVAs in dicate a substantially lower mean. Table 7.1d: 500-Meter Radius Cells. Mean per centage of Non-Wh ite Population at aggregated accident levels. AX/CELL # CELLS NON-WHITE 0 1104 8.5% 1 200 15.1% 2-4 129 19.8% 5+ 25 23.2% Figure 7.1d: 500-Meter Radius Cells. Mean percentage of Non-White Population at aggregated accident levels. The red horizontal line represents the countywide mean 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 012 45 +PRMVA/CELLPCT NON-WHITE 81

PAGE 92

Non-White: 1-Kilometer Radius The coarsest scale considered is th e 1-kilometer radius grid cell. An important justification for usi ng multiple data resolutions is to find the scale which allows trends to become evident without the undo loss of the specific and pert inent character of a neighborhood. Perhaps at a 1-kilometer radius the considerable area in question begins to approach this limit. As noted earlier, he xagonal cells of this size have an area that is four times (2,600,000 versus 650,000 square mete rs) that of the 500-meter cells. Even more notable is that the 1-k ilometer cells have 100 times the area when contrasted with that of 100-meter grid cells. Figure 7.1d shows that at the 1-kilometer grid scale, vast areas of a relatively high pe rcentage Non-White cover much of Saint Petersburg as well as Clearwater. 82

PAGE 93

Figure 7.1e: Pinellas County Percentage Non-Whit e, 1-Kilometer Radius Cells 83

PAGE 94

Table 7.1e shows that at 1-kilome ter, the scale may be too coarse to indicate any reliable associat ion. Trends are also not clea rly visible in the bar graph (Figure 7.1e). Table 7.1e: 1-Kilometer Radius Cells. Mean percentage of Non-White Population at each accident level. AX/CELL # CELLS NON-WHITE 0 237 5.9% 1 62 11.0% 2 37 9.3% 3 26 10.8% 4 24 16.7% 5 7 20.7% 6 11 16.3% 7 2 15.3% 8 5 52.6% 9 2 13.5% 10 3 6.9% 11 4 30.6% 12 2 23.0% 13 1 81.3% 15 1 46.0% 16 1 13.5% 20 1 11.3% 23 2 24.6% 84

PAGE 95

Figure 7.1f: 1-Kilometer Radius Cells. Me an percentage of Non-White at each accident level. The red horizontal line represents the countywide mean 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 01234567891011121315162023PRMVA/CELLPCT NON-WHITE Significantly, however, when these same numbers are aggregated, there indeed emerges a persistent and positive association between PRMVAs and the mean Non-White percentage. The effects of aggreg ation are visibly distinct when contrasting Figures 7.1f and 7.1g. Table 7.1f: 1-Kilometer Radius Cells. Mean percentage of Non-White Population at aggregated accident levels. AX/CELL # CELLS NONWHITE 0 237 5.9% 1 62 11.0% 2-4 87 11.8% 5-9 27 23.9% 10+ 15 26.0% The chosen values for aggregation were car efully selected and consistently applied. Importantly, appropriate discretion needs to be taken in selecting proper values for aggregation. 85

PAGE 96

Figure 7.1g: 1-Kilometer Radius Cells. Mean percentage of Non-White Population at aggregated accident levels. The red horizontal line represents the countywide mean 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 0 1 2-4 5-9 10+PRMVA/CELLPCT NON-WHITE 86

PAGE 97

Non-White: Statistical Analysis Up to this point, descriptive measures and summary statistics have been used to examine the relationship between the Non-white percentage and PRMVA occurrences at three different scales. As was explained in Chapter 4, a two-step regression approach was necessary due to th e exceptionally large number of grid cells with a zero accident occurrence. Binary logistic regression was used first to analyze the effect of the Non-White proportion on the probability of any grid cell experiencing an accident. Next, Ordinary Least Squares (O LS) Regression was used to examine the linear relationship between PRMVA frequency and Non-White percentages at each scale, on cells that had at least one accident. STEP 1: Logistic Regression The dependent variable (PRM VA occurrence) was treated as a dichotomous event, and coded as either 1 or 0 to represent presence or absence of PRMVAs, respectively. As described previously for the SDI, the natu ral logarithm of the odds that a grid cell experiences an accident is assumed to be a linear function of the independent variable (in this case, Non-White percentage) for each logistic model and the maximum likelihood method is used to estimate the model. Table 7.1g displays the results of the logistic regression models at each scale involving the percentage NonWhite variable. 87

PAGE 98

Table 7.1g: Logistic Regression for Percent Non-White on PRMVA Occurrence. NON-WHITE N B sig exp(B) 100M 31,106 0.018 0.000 1.018 500M 1,458 0.030 0.000 1.031 1KM 428 0.074 0.000 1.077 The table describes the total number of cells at the descri bed scale (N), the value of the logistic regressi on coefficient (B), and the obser ved level of significance (pvalues). To aid in interp reting the regression results, th e exponent of the regression coefficient or odds ratio produced by each logistic model [exp(B)] is also provided. The odds ratio may be interpreted as a multiplier of the odds of an accident occurring in a grid cell. When the value exceeds 1, the odds of experiencing a PRMVA increase that many times for each one percentage change in the Non-White proportion. Table 7.1g indicates that the relati onship between PRMVA occurrence and percent Non-White is positive and statistically significant (p<.001) at all scales. Also worthy of note is the positive and increasing tr end evident as the scales expand. At the 100-meter scale, a one percent increase in th e Non-White percentage increases the odds of a PRMVA by almost two percent. The odds increase by over three percent at the 500meter scale and by almost eight pe rcent at the 1-kilometer scale. STEP 2: Ordinary Least Squares Regression This step includes only those cells that had at least one ( 1) reported PRMVA and uses OLS regression to ex amine the association between PRMVA frequency (the dependent variable) and Non-White percentage (the inde pendent variable). 88

PAGE 99

Table 7.1h summarizes the result s including the sample size, l east squares coefficient for the independent variable, and the related levels of significance. Table 7.1h: Least Squares Regression for Percent Non-White on PRMVA Frequency for Cells with at Least One Accident. NON-WHITE N B sig 100M 595 0.000 0.745 500M 354 0.009 0.017 1KM 191 0.056 0.000 At the 100-meter radius scale, the results fail to achieve statistical significance (p>.10). The 500-meter and 1-kilometer scales do howev er, yield significant results (p<.05), as well as positive relationships. The least square coefficients suggest that a ten percent increase in the Non-White percentage in a gr id cell would result in an increase of only 0.1 PRMVAs at the 500-meter scale, and about 0.6 accidents at the 1-kilometer scale. 89

PAGE 100

Non-White Results Summary When basic descriptive measures are utilized, there is strong evidence of an association between the Non-White per centage and accident frequency. When viewing the summary statistics in tabular fo rmat, there appears some reliable association between PRMVAs and the percentage NonWhite. This relationship between the variables is particularly not able at finer scales. When analyzed through a two-step regression approach, the relationship betw een PRMVA occurrence and percentage NonWhite is confirmed to be consistently positive and increasing with the size of grid cells. The probability of PRMVA occurrence and the nu mber of such occurrences both increase significantly as the Non-White proportion increases, at th e 500-meter and 1-kilometer scales. Clearly, there appears to be a pos itive and statistically significant connection between the distribution of PRMVAs and percentage Non-White in Pinellas County. 90

PAGE 101

7.2 Results: Percent Black This section closely matches the form at used in the previous section that focused on the Non-White population distributio n. The objective in this section was to determine if PRMVA occurrences and frequency are more likely in areas with a disproportionately large number of Black residents. PRM VA distribution was spatially analyzed in conjunction with the percentage of the Black population in Pinellas County at each of three scales (100-meter, 500-meter, a nd 1-kilometer radius cells). These same data were then statistically analyzed using a two-step regression proc ess utilizing logistic regression and ordinary least squares regression, as described in Chapter 4, and applied in the SDI and Non-White sections. The Non-White classification examined in the previous section includes all categories of non-white popul ation of which Black (or Af rican-American) represents the largest minority. This group represented 9.0 percent of the overall Pinellas County population as of 2000 (US Census, 2000). Figures 7.2a, 7.2b, and 7.2e each display the county distribution of Black population as a percentage of the total population. Not surprisingly, this distribution closely matches that of the Non-White demographic. Also, as with that group, neighborhoods increasingly lose their robustness as the scales become coarser. 91

PAGE 102

Figure 7.2a: Pinellas County Percentage Black, 100-Meter Radius Cells 92

PAGE 103

Black: 100-Meter Radius At the 100-meter radius grid, there is certainly some apparent propensity for PRMVAs to occur disproportionately in cells with higher Black proportions. Grid cells without any PRMVAs average 6.4 per cent Black compared to the countywide proportion of 9.0 percent. This is a notable di fference although a review at each accident level (Table 7.2a) delivers results that indica te at best, an erratic trend. In aggregate (Table 7.2b), it becomes increasingly cl ear that grid cells with higher Black concentrations are to some extent more likely to experience a PRMVA. This scale leaves much to be desired for those seeking useful linear trends, however. Table 7.2a: 100-meter Radius Cells. Mean per centage of Black Population at each accident level. AX/CELL # CELLS BLACK 0 30511 6.4% 1 519 14.2% 2 60 10.7% 3 14 17.0% 4 1 1.4% 5 1 19.2% Table 7.2b: 100-meter Radius Cells. Mean per centage of Black Population at aggregated accident levels. AX/CELL # CELLS BLACK 0 30511 6.4% 1 519 14.2% 2+ 76 11.9% 93

PAGE 104

Figure 7.2b: Pinellas County Percentage Black, 500-Meter Radius Cells 94

PAGE 105

Black: 500-Meter Radius Interestingly, at this scale, those cells without PRMVAs have a mean Black population of just 4.4 percent. This is smaller than half the overall Black distribution (9.0 percent) th roughout the county. At the 500meter scale, the percent Black category again displays some irregularity and a lack of consistent relationship when considered at each PRMVA level (Table 7.2c). Table 7.2c: 500-Meter Radius Cells. Mean per centage of Black Population at each accident level. AX/CELL # CELLS BLACK 0 1104 4.4% 1 200 9.8% 2 80 10.6% 3 31 11.7% 4 18 30.1% 5 11 21.2% 6 6 17.9% 7 1 1.6% 8 4 15.1% 9 2 9.8% 15 1 7.3% A bar graph (Figure 7.2c) makes th e apparent lack of consistent association at this scale all the more evident. Intriguingly, the patt ern seems to peak at the middle accident levels with lower proporti ons of the Black population at the lowest and higher accident rates. 95

PAGE 106

Figure 7.2c: 500-Meter Radius Cells. Mean percentage of Black Population at each accident level. The red horizontal line represents the countywide mean 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 012345678915PRMVA/CELLPCT BLACK When aggregated, there material izes a heretofore unseen positive association between the cons idered factors (Table 7.2d). Areas with no PRMVAs indicate a mean percentage of the Black population that is approximate ly half that of the county average. Those grid cells with one PRMVA indicate a mean th at is close to the average for the county, while those cells with five or more accidents have a percentage Black rate of nearly twice the Pinellas County mean. Table 7.2d: 500-Meter Radius Cells. Mean per centage of Black Population at aggregated accident levels. AX/CELL # CELLS BLACK 0 1104 4.4% 1 200 9.8% 2-4 129 13.6% 5+ 25 17.2% The contrast between the non-aggreg ated and the aggregated graphs is notable. When sorted on at each PRMVA leve l, the bar graph shows little indication of a 96

PAGE 107

coherent association between PRMVA occurrence and percentage Black (Figure 7.2c). In fact, a cursory glance at the column gra ph could lead one to believe the association could be negative, in part due to a generous spike for cells with four accidents. The bar graph of the aggregated data helps to elucidate th is positive relationship (Figure 7.2d). It is useful to note that at each aggregated accident level, the number of cells with accidents decreases at a relatively consistent rate. Figure 7.2d: 500-Meter Radius Cells. Mean percentage of Black Population at aggregated accident levels. The red hor izontal line represents the countywide mean. 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% 18.0% 20.0% 012 45 +PRMVA/CELLPCT BLACK 97

PAGE 108

Figure 7.2e: Pinellas County Percentage Black, 1-Kilometer Radius Cells 98

PAGE 109

Black: 1-Kilometer Radius The apparent inconsistency of dir ection observed at the 500-meter radius for the Black category is also obvious at the 1-k ilometer radius scale (Table 7.2e). At this scale, grid cells that experienced zero PRMVAs over the two year study period were, on average, just 2.6 percent Black (county proportion: 9.0 percent Black). For cells with Table 7.2e: 1-Kilometer Radius Cells. Me an percentage of Black Population at each accident level AX/CELL # CELLS BLACK 0 237 2.6% 1 62 6.6% 2 37 5.0% 3 26 6.0% 4 24 9.6% 5 7 15.0% 6 11 11.9% 7 2 3.0% 8 5 47.9% 9 2 4.2% 10 3 1.5% 11 4 24.0% 12 2 12.6% 13 1 77.6% 15 1 36.7% 16 1 5.2% 20 1 3.0% 23 2 18.5% PRMVAs, the Black proportion ranged from a mere 1.5 percent to a tremendous 77.6 percent. The key factor is that the actual number of PRMVAs in these extreme areas is quite small. Figure 7.2f shows the irregular association graphically. 99

PAGE 110

Figure 7.2f: 1-Kilometer Radius Cells. Me an percentage of Black Population at each accident level. The red horizont al line represents the countywide mean. 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% 80.0% 90.0% 01234567891011121315162023PRMVA/CELLPCT BLACK Aggregation of these values helps to reduce the effect of outliers that may otherwise mislead (Table 7.2f). At this aggregation, as was the case at the 500-meter scale, there is a tremendous jump in Black proportion in areas with five or more PRMVAs per grid cell. Table 7.2f: 1-Kilometer Radius Cells. Mean percentage of Black Population at aggregated accident levels. AX/CELL # CELLS BLACK 0 237 2.6% 1 62 6.6% 2-4 87 6.6% 5-9 27 18.1% 10+ 15 19.3% Graphically, the effect and util ity of aggregation is clear when Figures 7.2f and 7.2g are compared. With all levels graphed individua lly, there is only the va guest of patterns. When aggregated, there appears some eviden ce of a positive association between Black proportions and PRMVA occurrences. 100

PAGE 111

Figure 7.2g: 1-Kilometer Radius Cells. Mean percentage of Black Population a mean. t aggregated accident levels. The red hor izontal line represents the countywide 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 0 1 2-4 5-9 10+PRMVA/CELLPCT BLACK 101

PAGE 112

Black: Statistical Analysis Summary statistics and descriptive measures have revealed valuable information on the association between th e percentage of the Black population and PRMVA occurrences. As with the SDI a nd the percent Non-White classifications described previously, two tests of sta tistical inference were conducted using SPSS statistical software. As described in Chap ter 4, a two-step appro ach based on logistic regression and OLS regression was utilized because a large propor tion of grid cells experienced no accidents. STEP 1: Logistic Regression Binary logistic regression was perf ormed by first coding each grid cell as a 1 or 0, representing the presence or ab sence of a PRMVA respectively. Table 7.2g displays the results of the logistic regre ssion model at each scale based on using the percent Black in each cell as the independent variable. The table contains the number of grid cells at each scale (N), the logistic regression coefficient (B), for the independent variable, and the observed level of significance (p-value). The odds ratio or the exponent of the regression coefficient produced by each logistic model to aid in interpreting the regression results [exp(B)] is also provided. Table 7.2g: Logistic Regression for Percent Black on PRMVA Occurrence. BLACK N B sig exp(B) 100M 31,106 0.016 0.000 1.016 500M 1,458 0.026 0.000 1.026 1KM 428 0.054 0.000 1.056 102

PAGE 113

As was the case with the percenta ge Non-white category (of which Black is the largest component), a statistically significant association with the odds of accident occurrence is observed at each scale (p<.001). As observed in the regression results for the Non-White percentage, the relationship is positive and increases in strength as the scale increases. At the 100meter scale, a one percent in crease in the Black proportion increases the odds of a PRMVA occurrence by 1.6 percent. With each percent increase in the Black proportion, the odds of a PRM VA increase by 2.6 percent at the 500-meter scale and 5.6 percent at the 1-kilometer scale. STEP 2: Ordinary Least Squares Regression The second step in this two-equation analysis includes only those cells that experienced at least one PRMVA and uses OLS Regression to analyze the relationship between the number of accidents (the depende nt variable) and percentage Black (the independent variable). Tabl e 7.2h contains the product of th e Ordinary Least Squares Table 7.2h: Least Squares Regression for Percent Black on PRMVA Frequency for Cells with at Least One Accident. BLACK N B sig 100M 595 0.000 0.662 500M 354 0.008 0.035 1KM 191 0.049 0.001 regression at each geographic scale. The re lationship between the two variables is not statistically significant only at the 100-mete r radius scale (p>.10). At the larger 500meter and 1-kilometer scales, there is a pos itive relationship coupled with statistical significance (p<.05). The OLS co efficients indicate that a ten percent increase in the 103

PAGE 114

proportion Black would yield an increase of just 0.1 PRMVAs at the 500-meter scale and an increase of half an accident at the 1-kilometer scale. It is notable that the coefficient is six-times higher at the 1-kilometer scale th an at the 500-meter scale indicating some increasing effect as the scale increases. 104

PAGE 115

Black: Results Summary The percent Black variable offers an intriguing set of results that are quite similar to those obtained for the Non-White category. When examining the data on the basis of summary statistics in tabular format and bar charts, there is apparent evidence of a significant spatial association between PRMVAs and the distribution of Black population as a percentage of total popul ation. Areas with the highest PRMVA occurrence levels have remarkably high proportions of Black distribution. In aggregation, these numbers become particular ly robust. When the data are analyzed through a two-step regression approach, th e statistical signif icance of the positive associations are reinforced at almost all ge ographic scales. Consis tently, at the 500-meter and 1-km scale, the probability of PR MVA occurrence and the number of such occurrences both increase significantly as the Black proportion increases. 105

PAGE 116

7.3 Results: Percent Hispanic This section of the study addre sses whether PRMVA occurrences and frequency are higher in areas with a dispropor tionate number of Hispanic residents. These relationships were spatially evalua ted using the same methodology described in previous sections with the SDI, Non-White, and Black variables. When compared to the distributi on of the Black population, Hispanics are more widely dispersed and not concentrated in specific areas. Figures 7.3a, 7.3b, and 7.3e each exhibit the distributi on of Hispanics as a percenta ge of the total population at varying scales. With just 4.6 percent of the county's population, Hispanics represent about half the overall pro portion of the Black community (U.S. Census, 2000). 106

PAGE 117

Figure 7.3a: Pinellas County percentage Hispani c, 100-Meter Radius Cells 107

PAGE 118

Hispanics at 100-Meter Radius The 100-meter radius grid cells do no t reveal clearly id entifiable patterns for this ethnic group. As Table 7.3a suggests, there is little obvious direction and only a slightly greater propensity for those cells with accidents to have a disproportionate number of Hispanic residents. While Pine llas County overall is 4.6 percent Hispanic, the mean percent Hispanic ranges from a low of 4.0 percent to a high of just 6.7 percent for cells with zero to four PRMVAs. These numbers, while marginally higher than the average countywide Hispanic percentage, di splay only a slightly positive relationship between PRMVA occurrence and percent Hispanic. Table 7.3a: 100-Meter Radius Cells. Mean percentage of Hispanic Population at each accident level. AX/CELL # CELLS HISPANIC 0 30511 4.0% 1 519 5.7% 2 60 6.7% 3 14 4.3% 4 1 6.5% 5 1 13.7% Even when aggregated, the directio n is, at best, a m odest indicator of relationship between these two factors (Tab le 7.3b). Here we see that although the direction is consistent, it is not immediately obvious. It is again an important reminder that entirely 98 percent of 100-meter radius cells had no PRMVA occurrence over the two year period of study. 108

PAGE 119

Table 7.3b: 100-Meter Radius Cells. Mean percentage of Hispanic Population at aggregated accident levels. AX/CELL # CELLS HISPANIC 0 30511 4.0% 1 519 5.7% 2+ 76 6.4% 109

PAGE 120

Figure 7.3b: Pinellas County Percentage Hispanic, 500-Meter Radius Cells 110

PAGE 121

Hispanics at 500-Meter Radius Even at the previously useful 500meter radius scale, the association between the percentage Hispanic and PRMVA occurrence appears modest. Except for an aberration at the 8 accident/cell level, no clear trends exist (Table 7.3c). When reviewing the entire breakdown of PRMVAs it does at firs t appear to exhibit some measure of trend, but at the higher PRMVA occurrence le vels, the pattern seems to waver. Table 7.3c: 500-Meter Radius Cells. Mean per centage of Hispanic Population at each accident level. AX/CELL # CELLS HISPANIC 0 1104 3.6% 1 200 4.6% 2 80 5.2% 3 31 5.6% 4 18 6.9% 5 11 5.7% 6 6 3.3% 7 1 6.2% 8 4 14.7% 9 2 4.0% 15 1 9.5% Graphically, the interesting upward tr end at the 0 to 5 PRMVA levels is quite evident, though the pattern is decidedly more obscure at higher accident levels (Figure 7.3c). 111

PAGE 122

Figure 7.3c: 500-Meter Radius Cells. Mean percentage of Hispanic Population at each accident level. The red horiz ontal line represents the countywide mean 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% 012345678915PRMVA/CELLPCT HISPANIC The trends appear unimpressive when aggregated, although there is a consistently positive and increasing pattern (T able 7.3d). Figure 7.3d is a particularly vivid demonstration of this positive trend. Although the ra nge in values is relatively slight, the direction is clear. At PRMVA acci dent levels of zero, the percentage Hispanic is below the corresponding count y proportion of 4.6 percent. Fo r cells with one accident, the county proportion is matched. At acciden t levels greater than one PRMVA, the proportion of Hispanics exceeds the Pinellas County proportion to an increasing degree. Table 7.3d: 500-Meter Radius Cells. Mean percentage of Hispanic Population at aggregated accident levels. AX/CELL # CELLS HISPANIC 0 1104 3.6% 1 200 4.6% 2-4 129 5.5% 5+ 25 6.6% 112

PAGE 123

Figure 7.3d: 500-Meter Radius Cells. Mean percentage of Hispanic Population at aggregated accident levels. The red horizontal line represents the countywide mean 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 012 45 +PRMVA/CELLPCT HISPANIC 113

PAGE 124

Figure 7.3e: Pinellas County Percentage Hispani c, 1-Kilometer Radius Cells 114

PAGE 125

Hispanics at 1-Kilometer Radius At the coarsest of scales, there is still only the slightest of adverse PRMVA consequences for Hispanics. Tabl e 7.3e indicates a ge nerally indistinct relationship between PRMVA occurrences and th e proportion of Hispanic individuals in Pinellas County. It is worthy of note that areas without PRMVA occurrences are onethird less Hispanic than the county proportion. Table 7.3e: 1-Kilometer Radius Cells. Me an percentage of Hispanic Population at each accident level. AX/CELL # CELLS HISPANIC 0 237 3.0% 1 62 3.7% 2 37 3.9% 3 26 4.7% 4 24 5.7% 5 7 5.2% 6 11 3.8% 7 2 6.1% 8 5 4.0% 9 2 4.1% 10 3 4.9% 11 4 6.6% 12 2 16.0% 13 1 2.9% 15 1 5.9% 16 1 12.5% 20 1 5.7% 23 2 9.1% When viewed graphically, there ar e clearly visible spikes in PRMVA occurrence at the 12, 16 and 23 accident levels but the other data classes fail to offer support for a readily identifiable associ ation between PRMVA occurrence and the percentage Hispanic cla ssification (Figure 7.3f). 115

PAGE 126

Figure 7.3f: 1-Kilometer Radius Cells. Me an percentage of Hispanic Population at each accident level. The red horizont al line represents the countywide mean 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% 18.0% 01234567891011121315162023PRMVA/CELLPCT HISPANIC In aggregation, some direction emerge s, but apparent indi cators are subtle and only moderate (Table 7.3f). The column graph below (Figure 7.3g) exhibits some measure of positive relations hip although only the 10+ cl ass exceeds the countywide proportion by any appreciable degree. Table 7.3f: 1-Kilometer Radius Cells. Me an percentage of Hispanic Population at aggregated accident levels. AX/CELL # CELLS HISPANIC 0 237 3.0% 1 62 3.7% 2-4 87 4.6% 5-9 27 4.4% 10+ 15 7.9% 116

PAGE 127

Figure 7.3g: 1-Kilometer Radius Cells. Mean percentage of Hispanic Population at aggregated accident levels. The red horizontal line represents the countywide mean 0.0% 1.0% 2.0% 3.0% 4.0% 5.0% 6.0% 7.0% 8.0% 9.0% 012 45 91 0 +PRMVA/CELLPCT HISPANIC 117

PAGE 128

Hispanic: Statistical Analysis To this point the percentage Hispanic classification has only been examined based on descriptive measures and summary statistics. As described previously, a two-step approach was util ized again to examine the strength and significance of the statistical association between the Hispanic category and PRMVA incidence. STEP 1: Logistic Regression The Hispanic percentage in each grid cell was used as an independent variable to analyze the dichotomous depe ndent variable (PRMVA occurrence). Table 7.3g summarizes the results provided by the lo gistic regression models at each scale. Table 7.3g: Logistic Regression for Percent Hispanic on PRMVA Occurrence HISPANIC N B sig exp(B) 100M 31,106 0.086 0.000 1.090 500M 1,458 0.137 0.000 1.146 1KM 428 0.472 0.000 1.604 At each scale, a statistically signi ficant (p<.001) and positive relationship is observed between the dependent and independ ent variables. The odds ratio values also increase persistently as the scale of consideration broadens At the 100-meter radius, a one percent increase in percentage Hispanic yields a nine percent increase in the odds of a PRMVA occurrence. The odds increase to 14.6 percent at 500-meters, and burgeon to 60 percent when the 1-kilome ter scale is considered. 118

PAGE 129

STEP 2: Ordinary Least Squares Regression The second step in this two-equatio n approach considers only those cells that experience at least one PRMVA and uses OLS Regression to analyze the relationship between the number of accidents (the depende nt variable) and the percentage Hispanic (the independent variable). Table 7.3h summarizes the results table for the OLS regressions. Table 7.3h: Least Squares Regression for Percent Hispanic on PRMVA Frequency for Cells with at Least One Accident. HISPANIC N B sig 100M 595 0.003 0.409 500M 354 0.059 0.000 1KM 191 0.569 0.000 Matching, to some degree, the results of the Bl ack category, there is a lack of statistical significance at the 100-meter scale (p>.10) but strong evid ence of a statistically significant association (p<.001) at the 500-me ter and 1-kilometer scales. Additionally, positive relationships between percent Hispanic and PRMVA frequency exist. At the 500-meter scale, the least squares coefficien t suggest that a ten percent increase in percent Hispanic would result in a 0.6 increase in PRMVAs and at the 1-kilometer scale, a ten percent Hispanic increase leads to a re latively substantial gain of almost six accidents. 119

PAGE 130

Hispanic: Results Summary Summary statistics and simple descriptive measures are less reliable for this variable compared to the SDI, percent Non-White, and percent Black. At each scale, the Hispanic category shows only limited evidence of a strong association with PRMVAs. While descriptive analysis reveals a pattern that supports a positive and increasing association with PRMVAs at each scale, aberrations are common. However, the two-step regression analys is reveals stronger associa tions between the Hispanic percentage and accident incidence. Statistical significance exists at all scales and the positive relationships increase rather drama tically with increasing scales. When examining exclusively those cells with PRMVAs at the 1-kilometer scale, the results are exceptionally strong and associations reliably positive. More than any other variable considered in this study, regression analysis serves to uncover meaningful associations between the disproportionate di stribution of Hispanic residents and both occurrence and frequency of PRMVAs. 120

PAGE 131

7.4 Results: Percent Below Poverty Unlike the three previous sections, wh ich dealt with racial or ethnic groups of Pinellas County, this section is concerned with the economic status of the population. Poverty-riddled areas in Pinell as County are frequently spatially coincident with that of Black areas. As can readily be seen in the following figures (Figures 7.4a, 7.4b, 7.4e), Saint Petersburg's Midtown neighborhood and a region just south of Safety Harbor represent two areas both dispr oportionately destitute and, in significant numbers, Black. While certain racial and ethnic groups have a greater statistical propensity to find themselves below poverty, any number of de mographic groups can also be Below Poverty. Countywide, 10.0 pe rcent of Pinellas County re sidents fell below Federal poverty standards in 2000 (U.S. Census, 2000). While not a small proportion, this is far smaller than Florida's statewide average of 14.1 percent and marginally lower than the national rate of 12.4 percent. Patterned much like the previous sections, the percent Below Poverty segment was spatially examined at each of three scales (100-meter, 500-meter, and 1kilometer radius). First, de scriptive and summary statistics were utilized. Second, a twostep regression model was conducted to better elucidate the relationship between those Below Poverty and the incidence of PRMVAs in Pinellas County. 121

PAGE 132

Figure 7.4a: Pinellas County Percentage Below Poverty, 100-Meter Radius Cells 122

PAGE 133

Below Poverty: 100-Meter Radius As with the previously examined va riables, there is some indication that this segment is disproportionately affect ed by PRMVA occurrences (Table 7.4a). Whereas countywide the poverty rate stands at 10.0 percent, cells without any PRMVA incidence over the twoyear period show an average Belo w Poverty rate of 8.2 percent. At each of the other frequency levels, th e county average was exceeded to varying degrees. An upward trend in greater accident frequency in more poverty stricken areas is possible but an aggregated versi on of this table serves to ma ke this clearer (Table 7.4b). In aggregate, areas experiencing PRMVAs on average readily exceed the Pinellas County poverty rate. Table 7.4a: 100-Meter Radius Cells. Mean percentage of Below Poverty Population at each accident level. AX/CELL # CELLS BEL POVERTY 0 30511 8.2% 1 519 13.5% 2 60 13.9% 3 14 16.3% 4 1 10.3% 5 1 23.4% Table 7.4b: 100-Meter Radius Cells. Mean percentage of Below Poverty Population at aggregated accident levels. AX/CELL # CELLS BEL POVERTY 0 30511 8.2% 1 519 13.5% 2+ 76 14.4% 123

PAGE 134

Figure 7.4b: Pinellas County Percentage Below Poverty, 500-Meter Radius Cells 124

PAGE 135

Below Poverty: 500-Meter Radius 500-meter radius grid cells reveal an even larger discord between those areas with and without accidents (Table 7.4c, Figure 7.4c). In areas without accidents, the poverty rate stands at only 7.2 percent. In grid cells with at leas t one accident, all but two frequency levels exceeded the county mean of 10.0 percent. Table 7.4c: 500-Meter Radius Cells. Mean percentage of Below Poverty Population at each accident level. AX/CELL # CELLS BEL POVERTY 0 1104 7.2% 1 200 10.4% 2 80 12.2% 3 31 14.0% 4 18 18.5% 5 11 18.0% 6 6 17.8% 7 1 9.5% 8 4 15.0% 9 2 23.3% 15 1 9.1% Figure 7.4c: 500-Meter Radius Cells. Mean percentage of Below Poverty Population at each accident level. The red horizontal line represents the countywide mean. 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 012345678915PRMVA/CELLPCT BEL POVERTY 125

PAGE 136

Without aggregation, there is some hint of a trend hidden amongst the irregularities at this scale. When placed in aggregate, a more visible pattern becomes obvious (Table 7.4d). Notably, in grid cells with at least two PR MVAs, the poverty rate in substantially exceeded. Table 7.4d: 500-Meter Radius Cells. Mean percentage of Below Poverty Population at aggregated accident levels. AX/CELL # CELLS BEL POVERTY 0 1104 7.2% 1 200 10.4% 2-4 129 13.5% 5+ 25 17.2% A comparison of Figures 7.4b and 7.4c illustrate the valu e of aggregation in these cases. Whereas the lo wer accident levels show a somewhat consistent trend, the higher accident levels are less obvious, until the values are aggregated. Figure 7.4d: 500-Meter Radius Cells. Mean percentage of Below Poverty Population at aggregated accident levels. The red horizontal line represents the countywide mean. 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% 18.0% 20.0% 012 45 +PRMVA/CELLPCT BEL POVERTY 126

PAGE 137

Figure 7.4e: Pinellas County Below Poverty di stribution, 1-Kilometer Radius Cells 127

PAGE 138

Below Poverty: 1-Kilometer Radius At the 1-kilometer radius, grid ce lls with no PRMVAs experience a Below Poverty rate of a mere 5.5 per cent, and are substantially sma ller than the Pinellas County rate of 10.0 percent. Prior to aggregati on, there is some indication of a positive association between PRMVAs and the proporti on of those Below Poverty, (Table 7.4e), particularly when the data are seen in graphical form (Figure 7.4f). Table 7.4e: 1-Kilometer Radius Cells. Mean percentage of Below Poverty Population at aggregated accident levels. AX/CELL # CELLS BEL POVERTY 0 237 5.5% 1 62 7.8% 2 37 8.4% 3 26 10.1% 4 24 11.6% 5 7 13.7% 6 11 11.3% 7 2 9.3% 8 5 19.0% 9 2 18.2% 10 3 9.6% 11 4 21.1% 12 2 17.7% 13 1 32.5% 15 1 18.5% 16 1 16.6% 20 1 14.7% 23 2 22.6% 128

PAGE 139

Figure 7.4f: 1-Kilometer Radius Cells. Mean percentage of Below Poverty Population at each accident level. The red horizontal line represents the countywide mean. 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 01234567891011121315162023PRMVA/CELLPCT BEL POVERTY Once the table is aggregated into reasonable PRMVA groupings, there is a clearly visible progression and the positive association between the factors appears impressively strong (Table 7.4f). At this 1-k ilometer level, there is quite a clear direction and some measure of consistency with th e positive relationship between the percent Below Poverty and PRMVAs. Table 7.4f: 1-Kilometer Radius Cells. Mean percentage of Below Poverty Population at aggregated accident levels. AX/CELL # CELLS BEL POVERTY 0 237 5.5% 1 62 7.8% 2-4 87 9.8% 5-9 27 13.7% 10+ 15 18.4% Graphically, the positive and linear re lationship is particularly evident and appears strong (Figure 7.4g). In aggregate, the relationship is remarkably consistent, as is 129

PAGE 140

the reliable decrease in the number of PRMVA cells as accident volume increases. The association between Below Poverty areas and PRMVA incidence is plainly discernable. Figure 7.4g: 1-Kilometer Radius Cells. Mean percentage of Below Poverty Population at aggregated accident levels. The red horizontal line represents the countywide mean. 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 16.0% 18.0% 20.0% 0 1 2-4 5-9 10+PRMVA/CELLPCT BEL POVERTY 130

PAGE 141

Below Poverty: Statistical Analysis The last portion of this analysis fo llows the format used previously for the other variables (the SDI, pe rcent Non-White, percent Black, and percent Hispanic). The strength and significance of the statistic al association betwee n the Below Poverty percentage and PRMVA incidence was exam ined using both logistic regression and ordinary least squares regression methods. STEP 1: Logistic Regression The percentage Below Poverty in each grid cell was used as an independent variable to analyze the dichotomous dependent variable (PRMVA occurrence). Table 7.4g summarizes the resu lts provided by the l ogistic regression models at each scale. Table 7.4g: Logistic Regression for Perc ent Below Poverty on PRMVA Occurrence BELOW POVERTY N B sig exp(B) 100M 31,106 0.074 0.000 1.076 500M 1,458 0.239 0.000 1.269 1KM 428 0.292 0.000 1.339 As the table indicates, the associ ation between percen t Below Poverty and PRMVA occurrence is positive and statistically significant at all scales (p<.001). Interestingly, the Below Poverty category produc es an odds ratio that is higher (and thus of greater effect) than all the racial/ethnic categories at each respective level, except Hispanics (whose ratio is higher at the 100-me ter and 1-kilometer levels). At the 100meter radius scale, a one pe rcent increase in the Below Poverty population results in a 7.6 131

PAGE 142

percent increase in the odds of a PRMVA o ccurrence. The odds of accident occurrence increase to 26.9 percent at the 500-meter scale and 34 percent at 1-kilometer radius scale. STEP 2: Ordinary Least Squares Regression The second step in this two-equatio n approach considers only those cells that experienced at least one PRMVA a nd uses OLS Regression to analyze the relationship between the number of accidents (the dependent variable) and percent Below Poverty (the independent vari able). Table 7.4h summari zes the results from the OLS regressions. Table 7.4h: Least Squares Regression for Percent Below Poverty on PRMVA Frequency for Cells with at Least One Accident. BELOW POVERTY N B sig 100M 595 0.003 0.187 500M 354 0.088 0.000 1KM 191 0.295 0.000 At the 100-meter radius scale, there is insufficient evid ence of statistical significance (p>.10). At the 500-meter and 1-kilometer scales, models indicate both statistical significance (p<.001) and a positive relationship be tween the variables. The least squares coefficients indicate that a ten percent increase in the percent Below Poverty would yield an increase of almost one accident at the 500-meter scale and nearly three accidents at the 1-kilometer scale. 132

PAGE 143

Below Poverty: Results Summary The Below Poverty category appears to have one of the strongest spatial relationships with pedestrian-related motor vehicle accidents among the variables considered in this study. Areas with PRMVAs are clearly more likely to contain a higher percentage of Below Poverty residents than the countywide proporti on. At each scale and for all grid cells with at least one accident this appears consistent. These patterns are confirmed by the two-step regression anal ysis which provides strong evidence of a statistically significant and positive relati onship between the proportion Below Poverty and accident occurrence or frequency. Areas with PRMVA incidence in the county are disproportionately populated by residents who are Below Poverty. 133

PAGE 144

7.5 Socio-Demographic Factors: Summary Based on the information uncovered in this chapter, pedestrian-related motor vehicle accidents in Pinellas County a ppear to be disproportionately distributed with respect to the proportion of individua ls in the four categ ories (Non-White, Black, Hispanic, and Below Poverty), considered in the study. Summary statistics revealed a substantially high proportion of each variable in grid cells which experienced at least one accident. At the 100-meter radius scale, patterns in most cases were not read ily apparent. When values were aggregated, tendencies toward higher numbers of the f our categories became only slightly more apparent. When the scale shifted to the 500-meter radius, both trends and increasingly positive correspondence emerged. When values were aggregated, each one of the four socio-demographic categories displayed a cons istent and increasing positive trend. Both the percentage Black and Below Poverty categories exhibited notably high disproportionate values in cells with PRM VA occurrences. At th e 1-kilometer scale, although a positive relationship remained, much of the impressive upward trends and linearity faded as apparently discrete regions began to blend with dissimilar adjacent areas. When aggregated, the socio-demographic factors generally displayed a modest association with PRMVAs, except for the Belo w Poverty classification, which retained a particularly strong relati onship at this scale. A two-step regression analysis furt her confirmed the si gnificance of the relationships between each va riable and PRMVA incidence. At the 100-meter radius scale, binary logistic regr ession revealed statistically significant, and unrelentingly positive coefficients for each explanatory variable. When OLS regression was performed 134

PAGE 145

at this 100-meter scale on t hose cells with at least one accident, there was limited evidence of a statistically significant relati onship between accident frequency and the proportion of the variable examined. The logistic regression results at the 500-meter radius scale yielded positive coefficients that considerably exceeded the corresponding values at the 100meter level. At this radius, each of the Non-White, Black, Hispanic and Below Poverty percentages produced statistically significant, positive relationships with PRMVAs. The percent Below Poverty category in particular, displayed th e strongest effect on PRMVA occurrence. The least squares regressions of cells with PRMVAs delivered statistically significant results that were again indicative of a positive relationship between the frequency of accidents and each variable considered. Intriguing results came from the 1kilometer radius l ogistic regression analysis. Here, the Hispanic odds ratio was particularly high, exceedi ng even that of the percent Below Poverty category. Least square s regression imparted reaffirming positive PRMVA relationships, particularly with Hispan ics, whose coefficient was nearly twice as high as that of the Below Poverty group. For each of the studied variables, at each of the scales considered, wherever there were statistically signifi cant results, positive relationships between pedestrian-related motor vehicle accidents and these variables were obtained. The variables considered here represent thos e classifications typically utilized in environmental justice studies and without excep tion, the evidence is clear. There appears to be significant evidence of environmenta l injustice in the spatial distribution of pedestrian-related motor vehicl e accidents in Pinellas County. 135

PAGE 146

CHAPTER 8 CONCLUSIONS AND DISCUSSION Locational hazards were the crux of the investigation presented in this thesis. It is contended here that the dangers in curred by pedestrians afoot (and for that matter, on bicycles) can be represented by a spatial pattern that may allow for a better understanding of the relevant dangers. It is imperative to the understanding of this study that accident sites are more than just a poi nt where an accident has occurred. Particular accident sites are indicative of proven dangerous areas for pedestrians. The identification of specific trends and patterns in the spa tial distribution of pedestrian-related motor vehicle accidents sites can help direct attention toward fact ors that make certain places more hazardous than others for pedestrian s and more specifically, the disadvantaged groups in society. An improved understandi ng of the geography of pedestrian danger may lead not only to a safer environment for those who may not be of the motorized set, but also, in a wider sense, would help with the planning of more func tional, efficient, and inviting urban environments. The dangers that befall pedestrian s in Pinellas Count y, Florida, may well be little different from hazards experienced el sewhere. In this county, pedestrian-related 136

PAGE 147

motor vehicle accidents (PRMVAs) occur more frequently when conditions are dry, in daylight hours, during the less warmer mont hs, and around the rush hour periods of the day. PRMVAs are as likely to happen near an intersection as other location on the transportation network. Motorist s are usually at fault due to their "failure to yield" or "careless driving." Pedestrians are frequently injured, and all too often killed in these encounters. This would seem to come as no surprise as the expected components of motorist-involved pedestrian accidents. This study also revealed that, in Pinellas County, the presence of sidewalks relates positively to the chance of a person afoot being involved in a PRMVA. It appears that the greater the volume of sidewalks, the more likely there will be a PRMVA occurrence. The results of statistical analysis suggest that sidewalk density and pedestrian danger are associated. This rela tionship should not be construed however, in any way, as being causal. This study did not se t out to address causali ty. The goal of this project was to identify spatial associations between PRMVAs and the considered factors. Sidewalks are unlikely to create danger on thei r own. Instead, they are more likely to be located in areas with pre-existing dangerous elements. If anything, sidewalks may be best described as a pull factor for pedestrians. The presen ce of a sidewalk could lure the pedestrian into choosing a particular route, even if this choice results in a longer and less convenient journey. The appearance of a sidewalk makes an intrinsically dangerous trip at least a little more comf ortable. The practical question that arises from this is to identify a remedy for this pedestrian danger. 137

PAGE 148

The study also provided evidence of locational inequities in the distribution of pedestrian accidents, with resp ect to the racial/ethnic and economic status of the residential population. In Pinellas C ounty, it is apparent that PRMVA incidences occur more frequently in predominantly NonWhite neighborhoods, in general, and Black and Hispanic communities, in particular Additionally, those Below Poverty are decidedly more predisposed to live in an ar ea with a disproportionate level of danger for pedestrians. Admittedly, each of the four categories examined in this study partially overlaps with at least one other category. The utiliza tion of overlapping sociodemographic categories was not accidental. The purpose was to view potential factors in various ways to better ascerta in the nature and magnitude of possible inequities. It was revealed that although all Non-Whites experien ce some measure of spatial inequity as pedestrians, the Black population was the most susceptible. It was also discovered that areas characterized by individua ls in poverty, regardless of race or ethnicity, were also likely to experience higher probabilities and frequencies of accide nt occurrence. The environmental injustice implications of PRMVA incidence are intriguing and considerable, and certainly warrant further study. The environmental justice (EJ) movement was a response to a set of disturbing, and in retrospect, r eadily apparent inequities in the siting of the most visible and destructive elements of Americas hea vy industry. Smokestacks, toxic waste dumps, and chemical storage facilitie s are each easily recognized as hazards with adverse human health effects. The research literature in itiated by the environmental justice movement has empirically demonstrated that spa tially, these hazardous facilities were 138

PAGE 149

disproportionately located in minority and low-income neighborhoods. This thesis has revealed patterns of disparity in the occu rrence of pedestrian-related motor vehicle accidents. Initially, this study sought to inve stigate the simple issue of the built environment and how it may impact pedestrian sa fety. The most direct pedestrian-related element of the built environment is the sidewa lk. They are installed for no other purpose than to increase pedestrian sa fety. With further examination of this issue, it has become apparent that the methodology developed for an alyzing sidewalk density in this thesis could be used to address more than just th e hardware within the city. There was an opportunity to expand the scope and relevance of this project by furt her investigating the relationship between pedestrian safety a nd neighborhood demographics. Environmental justice studies have traditionally examined the location of dis-amenities. Rather than limiting the investigation to the noxious facilities that are pres ent, this thesis expands the notion of environmental injusti ce by examining pedestrian safety in the same fashion that environmental justice research has considered industrial hazards and pollutants. Safe pedestrian travel is no less a worthwhile and deserving amenity than clean air and unpolluted drinking water. Minorities and the economically disadvantag ed are entitled to this amenity no less than the majority and the more affluent. By choosing to investigate a pervasive hazard in this way, it is hoped the definition of environmental justice can be extended to include consideration for amen ities beyond industrial pollutants and produce a better understanding of the di sparities in today's society. 139

PAGE 150

The study makes a methodological contribution by using a hexagonal approach for the spatial definition of analytical units. Prior studies have used pre-defined census units, zip code areas, overlapping circular buffers, and muni cipal boundaries to spatially subdivide areas of study. Unlike pa st approaches, this project has utilized discrete and non-overlapping hexa gonal cells, at three distinct geographic scales. This allows for greater flexibility and effectivel y avoids edge-effect problems where events occur on boundaries of areal uni ts. Often, predefined boundaries coincide with roadways which would otherwise place a great many PR MVAs on these borders. An arbitrary, consistently applied hexagon grid addresses th e problem and thus provides more reliable analytical results. Another consideration that was treat ed with great discretion was that of scale. When studying spatial attributes, it is important to consider that the results of statistical analysis are sensitive to the choice of geographic scale. This is the Modifiable Areal Unit Problem (MAUP) which suggests that the strength and significance of relationships increase as the scales becomes coarser (Sui, 1999). Instead of choosing a single spatial scale, this project deals with potential MAUP concer ns by analyzing each relationship at three distinct sc ales. Issues of spatial extrap olation and interpolation take a considerable liberty with the notion that areal characteristic s are consistently distributed throughout the zones of consideration. By wo rking with three scal es, this concern was addressed in a way that inc onsistencies would more likely be revealed with any unreasonable interpolations or extrapolations. 140

PAGE 151

Much consideration has been made ov er the possibility that the analysis may have yielded differing result s at other scales. The choice of scales that differ from those chosen could conceivably have great effect on the results. This was a major decision, but at each end there were troubling is sues. If the radius was further widened, ever-increasing scales may have rewarded th is study with increasingly strong statistical inferences, but the nature of a neighborhood (the essential focus of this project) would be lost. Smaller areas were deemed over-precise and all too likely to exceed the precision of the source data. Some consideration was made for what size of an area would reasonably represent an urban neighborhood. Somewhat arbitrarily, though with much deliberation, 100-meter, 500-meter, and 1-kilometer radii were chosen. Ultimately, it was decided practical considerations must exceed statistically pleasing results. If there was a wider reac hing, practical goal for th is project, it was to create an increased appreciation for sidewalk s and pedestrian travel It was hoped that information gleaned from this work would pr omote the need for mo re and better routes for those who do not use motorized means of transport. This study began with a preconception that that lack of sidewalks creates a public safety concern for pedestrians. It was presumed that areas with an absen ce of sidewalks would experience a greater number of pedestrian-related mo tor vehicle accidents. Surprisi ngly, this is clearly not so. In fact, the opposite appears to be more the case. Where there are more sidewalks, there are also more accidents. In retrospect, the r eason behind this seems evident. If sidewalks exist, more people are likely to choose to walk that particular route. Sidewalks act as a "pull" factor for pedestrians. Generally, side walks are laid out with some apparent regard 141

PAGE 152

for the most commonly needed routes of pedestrian travel. They are where they are, because they are needed there. This is not to assume that they are always where they are needed, but rather are generally placed appropriately. The end result of all this is that pedestrians are most likely to be hit where pedestrians are most likely to be, that is to say, on a sidewalk, or perhaps an associated z one such as a crosswalk. Still, although pedestrians are involved in more PRMVAs in areas with the greatest volume of sidewalks, it should not at al l be assumed that increasing the number of sidewalks will lead to more PRMVAs. Perhaps, an incr ease in sidewalks would allow for safer alternative routes where pedestrians could better avoid the aforementioned motorized hazards that lurk on the roadways. Then, reasonably, more sidewalks can mean more safety for those afoot. Why does there appear to be an increasing proportion of minorities and economically disadvantaged indi viduals in areas with a grea ter occurrence of PRMVAs? Perhaps it is because minorities, who are so often less affluent, are more likely to walk when they need to get somewhere. Maybe they have no access to motorized means of transport. Maybe the absence of convenient and reliable public transportation leaves these groups with no other choice. Perhaps it is the sheer numbers of walking minorities and poor that result in a great er incidence of PRMVAs. This purpose of the study was to identify spatial patterns and re lationships, but not the processe s that explain the patterns. Further research seeking to identify the causes of these trends is encouraged. 142

PAGE 153

Perhaps the most significant limitation of this study is the absence of data. Specifically, there is a lack of information due to all of the unreported PRMVAs that surely happen each day. Certainly when one of the involved parties is injured, the accident is likely to involve the authorities, and thus reported, thereby becoming one of the component statistics for this project There are undoubtedly a great many accidents that involve a pedestrian who is perhaps just grazed or bum ped by an automobile, and are fortunate enough to just "walk it off." Perhaps the automobile driver might seek to avoid having an accident on their driv ing record, and thus keep thei r car insurance rates down. And maybe the pedestrian vict im would rather not wait around for the proper authorities to show up, and subject themselves to mounds of paperwork and volumes of irretrievably lost time. The result is that the accident doesn't get reported. These incidents are not included in this study. Only accident reports filed with the proper authorities are considered here. Another important factor that this study did not (and could not) focus upon was the total volume of pedestrians along any given route. For this, there simply is no concise dataset yet produced that would cover the number of pedestrian s that travel each route in this urban environment. As a pr actical matter, it would seem unreasonable to expect any such pedestrian survey to be completely thorough, covering all routes, all times of day, throughout the entire year. Certainly it would be expected that more heavily trafficked (both afoot and motori zed) routes would likely experience more PRMVAs. But to what extent are these occu rrences happening at disp roportionate rates? At this point, the data necessary to build such a knowledge base is not available. 143

PAGE 154

AFTERWORD Initially, this study was intended to investigate the da ngers faced by both pedestrians and bicyclists. Each, it seemed, faced a sim ilar danger in the ubiquitous automobile. Alas, as it turns out, despite the author's personal (and very situational) misgivings of the practice, bicyclists and pedestrians do indeed have differing safety zones. To wit, studies show that bicyclists are in increased danger when biking on the sidewalk whereas a pedestrian is generally said to be more out of harm's way when walking on that same sidewalk (Wachtel, Lewiston, 1994, Senturia et al, 1997). The bicyclist is considered to be safer on the roadways so much so that many municipalities have legally mandated roadway travel in lieu of bicycles utilizing si dewalks. To dispute the value of the existing statistics may be f oolhardy but an intuitive thought is that when biking on the sidewalk, a bicyc list encounters potential inters ect points with automobiles at driveways, intersections, crosswalks, and perhaps a few other places. When biking on a roadway, every single length of travel repr esents a potential, and legally accessible, intersect point where the automobile and the bi cycle can readily cross paths. Intuitively, one might think the simple act of reducing th e number of potentially hazardous intersect points would be the easiest way to reduce PR MVAs. But with the legal encouragement of roadway travel for bicyclists, this study could not presume to treat pedestrians and bicyclists alike. The intended arteries fo r each are different. Additional worthwhile 144

PAGE 155

opportunities for future research rest with fu rther exploring the differing nature of the hazards encountered by the pedestrian and the bicyclist. 145

PAGE 156

REFERENCES Antiput, S., Gray, B., Woods, S., 1996. Steps Ahead: Making Streets that Work in Seattle, Washington. Landscape and Urban Planning 35: 107-122. Aultman-Hall, L., Hall, F.L., 1998. Ottawa -Carleton Commuter Cyclist On-Road and Off-Road Incident Rates. A ccident Analysis and Prevention 30(1): 29-43. Aultman-Hall, L., Kaltenecker, M.G., 1999. To ronto Bicycle Commute r Safety Rates. Accident Analysis and Prevention 31: 675-686. Bailey, T.C., Gatrell, A.C., 1995. Interactive Spatial Analysis New York, NY: Prentice Hall Baltes, M.R., 1998. Descriptive Analysis of Crashes Involving Pedestrians in Florida, 1990-1994. Transportation Research Record 1636: 138-145. Barber, G, 1995. Aggregate Characteristic s of Urban Travel. Geography of Urban Transportation (Second Edition), Susan Hanson (Ed.). NewYork: The Guilford Press. Bowen, W., 2002. An Analytical Review of Environmental Justice Research: What Do We Really Know? Environmental Management 29(1): 3-15. Brindley, P., Wise, S.M., Maheswaran, R ., Haining, R.P., 2005. The Effect of Alternative Representations of Population Location on the Areal Interpolation of Air Pollution Exposure, Computers. Environment and Urban Systems 29: 455-469. 146

PAGE 157

Bullard, R.D., 1990. Chapter One: Environmentalism and Social Justice. Dumping in Dixie: Race, Class, and Environmental Quality Boulder, Colorado: Westview Press. Campos-Outcalt, D., Bay, C., Dellapenna, A., Cota, M.K., 2002. Pedestrian Fatalities by Race/Ethnicity in Arizona, 1990-96. Ameri can Journal of Preventative Medicine 23(2): 129-135. Chakraborty, J., Schweitzer, L.A., Forke nbrock, D.J., 1999. Using GIS to Assess the Environmental Justice Consequences of Trans portation System Changes. Transactions in GIS 3(3): 239-258. Clarke, K.L., Hummer, J.E., Dutt, N., 1996. Field Evaluation of Fluorescent Strong Yellow-Green Pedestrian Warning Signs Transportation Research Record 1538: 39-46. Clinton, W.J., 1994. Federal Actions to A ddress Environmental Justice in Minority Populations and Low-Income Populations. Executive Order No. 12898. Cox, D.D., Cox, L.H., Ensor, K.B., 1997. Sp atial Sampling and the Environment: Some Issues and Directions, Environm ental and Ecological Statistics 4: 219-233. Cutter, S.L., 1995. Race, Class, and Envi ronmental Justice, Progress in Human Geography 19(1): 111-122. Cutter, S.L., Solecki, W.D., 1996. Setting Environmental Justice in Space and Place: Acute and Chronic Airborne T oxic Releases in the Southeas tern United States. Urban Geography 17(5): 380-399. Daniels, G., Friedman, S., 1999. Spatial Ine quality and the distribution of Industrial Toxic Releases: Evidence from the 1990 TRI. Social Science Quarterly, 80(2): 244-262. 147

PAGE 158

Dewey, J.F., Denslow, D., 2003. Trans portation Issues: Pedestrian Safety Gainesville, FL: Bureau of Economic and Business Research, University of Florida, State of Florida Department of Transportation. Duan, N., Manning, Jr., W.G., Morris, C.N ., Newhouse, J.P., 1983. A Comparison of Alternative Models for the Demand for Health Care. Journal of Business & Economic Statistics 1(2): 115-126. Elvik, R., 2000. Which are the Relevant Costs and Benefits of Road Safety Measures Designed for Pedestrians and Cyclists? Accident Analysis and Prevention 32: 37-45. Ernst, M., 2004. Mean Streets 2004 How Far Have We Come? Pedestrian Safety, 1994-2003 Washington, DC: Surface Tran sportation Policy Project. Florida Fish and Wildlife Cons ervation Commission, Florida Fish and Wildlife Research Institute, 2006. http://ocean.floridamarine.org/mrgis/viewer.htm Retrieved data multiple times throughout 2006. Garder, P.E., 2004. The Impact of Speed and Other Variables on Pedestrian Safety in Maine. Accident Analysis and Prevention 36: 533-542. Greenberg, M.R., 1993. Proving Environmental Inequity in Siting Locally Unwanted Land Uses. Risk Issues in Health and Safety 4: 235-252. Greenberg, M.R., Renne, J., 2005. Where Do es Walkability Matter Most? An Environmental Justice Interpre tation of New Jersey Data. Journal of Urban Health: Bulletin of the New York Academy of Medicine 82(1): 90-100. Gregory, I.N., 2002. The Accuracy of Areal Interpolation Techniques: Standardising 19 th and 20 th Century Census Data to Allow L ong-Term Comparisons. Computers, Environment and Urban Systems 26:293-314. 148

PAGE 159

Hebert Martinez, K.L., Porter, B.E., 2004. The Likelihood of Becoming a Pedestrian Fatality and Drivers Knowledge of Pede strian Rights and Responsibilities in the Commonwealth of Virginia. Tr ansportation Research Part F 7: 43-58. Hillsborough County Metropolitan Planning Organization, 2000. Bicycle Crash Data Analysis for Hillsborough County Tampa, Florida: Hillsborough County Metropolitan Planning Organization. Jackson, L.E., 2003. The Relationship of Urba n Design to Human Health and Condition. Landscape and Urban Planning 64: 191-200. Kwan, I., Mapstone, J., 2004. Visibility Aids for Pedestrians and Cyclists: A Systematic Review of Randomised Controlled Trials Accident Analysis and Prevention 36: 305312. LaScala, E.A., Gruenewald, P.J., Johns on, F.W., 2004. An Ecological Study of the Locations of Schools and Child Pedestrian In jury Collisions. Accident Analysis and Prevention 36: 569-576. Leden, L., 2002. Pedestrian Risk Decrease w ith Pedestrian Flow. A Case Study Based on Data from Signalized Intersections in Hamilton, Ontario. Accident Analysis and Prevention 34: 457-464. Librett, J.J., Yore, M.M., Schmid, T.L., 2003. Local Ordinances that Promote Physical Activity: A Survey of Municipal Policies. American Journal of Public Health 93(9): 1399-1403. Liu, F., 2001. Environmental Justice Anal ysis: Theories, Methods, and Practice CRC Press: New York. 149

PAGE 160

National Highway Transportation Safety Administration, 2006. http://www.nhtsa.dot.gov/ Retrieved multip le times between 20 March 2005 01 April 2006. Olea, R.A., 1984. Sampling Design Optimizatio n for Spatial Functions, Mathematical Geology 16: 369-92. Ossenbruggen, P.J., Pendharkar, J., Ivan, J., 2001. Roadway Safety in Rural and Small Urbanized Areas. Accident Analysis and Prevention : 33: 485-498. Owens, S., 1995. From Predict and Provide to Predict and Prevent?: Pricing and Planning in Transport Policy. Transport Policy 2(1): 43-49. Painter, K., 1996. The Influence of Street Lighting Improvements on Crime, Fear and Pedestrian Street Use, After Dark. Landscape and Urban Planning 35: 193-201. Perlin, S.A., Sexton, K., Wong, D.W.S., 1999. An Examination of Race and Poverty for Populations Living Near Industrial Sources of Air Pollution. Journal of Exposure Analysis and Environmental Epidemiology 9(1): 29-48. Pinellas County Information Systems Office, 2006. http://pubgis.co.pinellas.fl.us/public_gis/ Retrieved data and information various times during March-June, 2006. Pollock III, P.H., Vittas, M.E., 1995. W ho Bears the Burdens of Environmental Pollution? Race, Ethnicity, and Environmenta l Equity in Florida. Social Science Quarterly 76(2): 294-310. Redmon, T., 2003. Assessing the Attitudes and Be haviors of Pedestrians and Drivers in Traffic Situations. ITE Journal 73(4): 26-30. 150

PAGE 161

Reed, R., Parikh, K., 2003. Racial Differences in Senior Citize n Attitudes Toward Pedestrian Safety. TRB Annual Meeting CD-ROM. Sanchez, T.W., Wolf, J.F., 2005. Environmen tal justice and Transportation Equity: A Review of Metropolitan Planning Organizations. Brookings Institution. Schneider, R.J., Ryznar, R.M., Khattak, A.J ., 2004. An Accident Waiting to Happen: A Spatial Approach to Proactive Pedestrian Pl anning. Accident Analysis and Prevention 36: 193-211. Schweitzer, L., Valenzuela, Jr., A., 2004. E nvironmental In justice and Transportation: The Claims and the Evidence. Journal of Planning Literature 18(4): 383-398. Senturia, Y.D., Morehaead, T., LeBailly, S ., Horwitz, E., Kharasch, M., Fisher, J., Christoffel, K.K, 1997. Bicy cle-Riding Circumstances and Injuries in School-aged Children: A Case Control Study. Archives of Pediatrics & Adolescent Medicine 151(5): 485-489. Smith, T., Nelischer, M., Perkins, N., 1997. Quality of an Urban Community: A Framework for Understanding the Relationshi p between Quality and Physical Form. Landscape and Urban Planning 39: 229-241. Stutts, J.C., Hunter, W.W., 1999. Motor Vehicl e and Roadway Factors in Pedestrian and Bicyclist Injuries: An Examination Based on Emergency Department Data. Accident Analysis and Prevention 31: 505-514. Stutts, J.C., Williamson, J.E., Whitley, T., Sheldon, F.C., 1990. Bi cycle Accidents and Injuries: A Pilot Study Comparing Hospitaland Police-Reported Data. Accident Analysis and Prevention 22(1): 67-78. 151

PAGE 162

Sui, D., 1999. GIS, Environmental Equity Analysis, and the Modifiable Areal Unit Problem (MAUP). In M. Craglia and H. On srud, eds. Geographic Information Research: Transatlantic Perspective London: Taylor and Francis; 41-54. Tyrrell, R.A., Wood, J.M., Carberry, T.P., 2004. On-Road Measures of Pedestrians Estimates of Their Own Nighttime Conspicuity. Journal of Safety Research 35: 483490. United Church of Christ, Commission for Raci al Justice, 1987. Toxic Wastes and Race in the United States: A National Report on th e Racial and Socioeconomic Characteristics of Communities with Hazardous Waste Sites New York: Commission for Racial Justice, United Church of Christ. United States Census, 2005. www.census.gov Retrieved multiple times between 20 March 2005 29 March 2006. United States Department of Transportation, Bureau of Transporta tion Statistics, 2005. www.dot.gov/. Retrieved multiple time s between 20 March 05 29 March 2006. United States Department of Transportati on, Departmental Office of Civil Rights and Office of the Assistant Secretary for Trans portation Policy, 1997. U.S. Department of Transportation Order on Environmen tal Justice. Federal Register 62(72): 18377-18381. United States Environmental Protection Agenc y, Office of Environmental Justice, 2000. www.epa.gov/compliance/environmentaljustice/index.html Retrieved multiple times between 20 March 2005 29 March 2006. Wang, G., Macera, C.A., Scudder-Soucie, B., Sc hmid, T., Pratt, M., Buchner, D., Heath, G., 2004. Cost Analysis of the Built Environm ent: The Case of Bike and Pedestrian Trails in Lincoln, Nebraska. American Journal of Public Health 94(4): 549-553. 152

PAGE 163

Watchel, A., Lewiston, D., 1994. Risk Factors for Bicycle-Motor Vehicle Collisions at Intersections. ITE Journal 64(September 1994): 30-35. Zandbergen, P.A., Chakraborty, J., 2006. Improving Environmental Exposure Analysis Using Cumulative Distribution Functions and Individual Geocoding. International Journal of Health Geographics 5(23). 153

PAGE 164

APPENDICES 154

PAGE 165

APPENDIX A 100-METER RADIUS SUMMARY TABLES Table A-1: Summary of the Studied Factors 100-Meter Radius Scale Each Accident Level AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 30511 89.0% 11.0% 6. 4% 4.0% 8.2% 0.4094 1 519 79.7% 20.3% 14.2% 5.7% 13.5% 0.6451 2 60 83.2% 16.8% 10.7% 6.7% 13.9% 0.7499 3 14 77.2% 22.8% 17.0% 4.3% 16.3% 0.7133 4 1 90.2% 9.8% 1.4% 6.5% 10.3% 1.0400 5 1 68.4% 31.6% 19.2% 13.7% 23.4% 0.2635 Table A-2: Summary of the Studied Factor s 100-Meter Radius Scale Aggregated Accident Levels AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 30511 89.0% 11.0% 6. 4% 4.0% 8.2% 0.4094 1 519 79.7% 20.3% 14.2% 5.7% 13.5% 0.6451 2+ 76 82.0% 18.0% 11. 9% 6.4% 14.4% 0.7406 Table A-3: Summary of the Studied Factor s 100-Meter Radius Scale Comparing Cells With and Without PRMVAs AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 30511 89.0% 11.0% 6. 4% 4.0% 8.2% 0.4094 1+ 595 80.0% 20.0% 13. 9% 5.7% 13.6% 0.6573 155

PAGE 166

APPENDIX B 500-METER RADIUS SUMMARY TABLES Table B-1: Summary of the Studied Factors 500-Meter Radius Scale Each Accident Level AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 1104 91.5% 8.5% 4.4% 3.6% 7.2% 0.3292 1 200 84.9% 15.1% 9. 8% 4.6% 10.4% 0.4287 2 80 83.3% 16.7% 10. 6% 5.2% 12.2% 0.5316 3 31 81.4% 18.6% 11. 7% 5.6% 14.0% 0.5314 4 18 64.6% 35.4% 30. 1% 6.9% 18.5% 0.6072 5 11 73.7% 26.3% 21. 2% 5.7% 18.0% 0.5866 6 6 78.0% 22.0% 17.9% 3.3% 17.8% 0.6984 7 1 92.0% 8.0% 1.6% 6.2% 9.5% 0.5771 8 4 73.0% 27.0% 15. 1% 14.7% 15.0% 0.5265 9 2 84.1% 15.9% 9.8% 4.0% 23.3% 0.7291 15 1 89.0% 11.0% 7.3% 9.5% 9.1% 0.9000 Table B-2: Summary of the Studied Factor s 500-Meter Radius Scale Aggregated Accident Levels AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 1104 91.5% 8.5% 4.4% 3.6% 7.2% 0.3292 1 200 84.9% 15.1% 9. 8% 4.6% 10.4% 0.4287 2-4 129 80.2% 19.8% 13. 6% 5.5% 13.5% 0.5421 5+ 25 76.8% 23.2% 17. 2% 6.6% 17.2% 0.6274 156

PAGE 167

APPENDIX B (CONTINUED) Table B-3: Summary of the Studied Factor s 500-Meter Radius Scale Comparing Cells With and Without PRMVAs AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 1104 91.5% 8.5% 4.4% 3.6% 7.2% 0.3292 1+ 354 82.6% 17.4% 11. 7% 5.1% 12.0% 0.2553 157

PAGE 168

APPENDIX C 1-KILOMETER RADIUS SUMMARY TABLES Table C-1: Summary of the Studied Factors 1-Kilometer Radius Scale Each Accident Level AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 237 94.1% 5.9% 2.6% 3.0% 5.5% 0.2761 1 62 89.0% 11.0% 6.6% 3.7% 7.8% 0.4504 2 37 90.7% 9.3% 5.0% 3.9% 8.4% 0.4466 3 26 89.2% 10.8% 6.0% 4.7% 10.1% 0.4105 4 24 83.3% 16.7% 9.6% 5.7% 11.6% 0.4625 5 7 79.3% 20.7% 15.0% 5.2% 13.7% 0.5053 6 11 83.7% 16.3% 11.9% 3.8% 11.3% 0.4467 7 2 84.7% 15.3% 3.0% 6.1% 9.3% 0.3491 8 5 47.4% 52.6% 47.9% 4.0% 19.0% 0.5350 9 2 86.5% 13.5% 4.2% 4.1% 18.2% 0.3330 10 3 93.1% 6.9% 1.5% 4.9% 9.6% 0.6906 11 4 69.4% 30.6% 24.0% 6.6% 21.1% 0.6183 12 2 77.0% 23.0% 12.6% 16.0% 17.7% 0.1708 13 1 18.7% 81.3% 77.6% 2.9% 32.5% 0.7826 15 1 54.1% 46.0% 36.7% 5.9% 18.5% 0.8414 16 1 86.5% 13.5% 5.2% 12.5% 16.6% 0.5535 20 1 88.8% 11.3% 3.0% 5.7% 14.7% 0.2808 23 2 75.4% 24.6% 18.5% 9.1% 22.6% 0.8917 158

PAGE 169

APPENDIX C (CONTINUED) Table C-2: Summary of the Studied Factors 1-Kilometer Radius Scale Aggregated Accident Levels AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 237 94.1% 5.9% 2.6% 3.0% 5.5% 0.2761 1 62 89.0% 11.0% 6.6% 3.7% 7.8% 0.4504 2-4 87 88.2% 11.8% 6. 6% 4.6% 9.8% 0.4402 5-9 27 76.1% 23.9% 18. 1% 4.4% 13.7% 0.4626 10+ 15 74.0% 26.0% 19. 3% 7.9% 18.4% 0.6491 Table C-3: Summary of the Studied Factors 1-Kilometer Radius Scale Comparing Cells With and Without PRMVAs AX/CELL # CELLS WHITE NONWHITE BLACK HISPANIC BEL POVERTY SDI 0 237 94.1% 5.9% 2.6% 3.0% 5.5% 0.2761 1+ 191 85.7% 14.3% 9. 2% 4.5% 10.4% 0.4631 159

PAGE 170

APPENDIX D REGRESSION SUMMARY TABLES Table D-1: Binary Logistic Regres sion 100-Meter Radius All Factors 100-M N B sig exp(B) Non-White 31,106 0.018 0.000 1.018 Black 31,106 0.016 0.000 1.016 Hispanic 31,106 0.086 0.000 1.090 Below Poverty 31,106 0.074 0.000 1.076 Table D-2: Ordinary Leas t Squares Regression 100-Me ter Radius All Factors 100-M N B sig Non-White 595 0.000 0.745 Black 595 0.000 0.662 Hispanic 595 0.003 0.409 Below Poverty 595 0.003 0.187 SDI 595 0.090 0.091 Table D-3: Binary Logistic Regres sion 500-Meter Radius All Factors 500-M N B sig exp(B) Non-White 1,458 0.030 0.000 1.031 Black 1,458 0.026 0.000 1.026 Hispanic 1,458 0.137 0.000 1.146 Below Poverty 1,458 0.239 0.000 1.269 SDI 1,458 1.549 0.000 4.705 160

PAGE 171

APPENDIX D (CONTINUED) Table D-4: Ordinary Leas t Squares Regression 500-Me ter Radius All Factors 500-M N B sig Non-White 354 0.009 0.017 Black 354 0.008 0.035 Hispanic 354 0.059 0.000 Below Poverty 354 0.088 0.000 SDI 354 1.552 0.000 Table D-5: Binary Logis tic Regression 1-Kilometer Radius All Factors 1-KM N B sig exp(B) Non-White 428 0.074 0.000 1.077 Black 428 0.054 0.000 1.056 Hispanic 428 0.472 0.000 1.604 Below Poverty 428 0.292 0.000 1.339 SDI 428 2.485 0.000 12.001 Table D-6: Ordinary Least Squares Regression 1-Kilome ter Radius All Factors 1-KM N B sig Non-White 191 0.056 0.000 Black 191 0.049 0.001 Hispanic 191 0.569 0.000 Below Poverty 191 0.295 0.000 SDI 191 3.914 0.002 161


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001916908
003 fts
005 20071114111923.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 071114s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002013
035
(OCoLC)181158377
040
FHM
c FHM
043
n-us-fl
049
FHMM
090
G116 (ONLINE)
1 100
Harmak, Craig W.
0 245
Danger afoot :
b sidewalks, environmental justice, and pedestrian safety in Pinellas County, Florida
h [electronic resource] /
by Craig W. Harmak.
260
[Tampa, Fla.] :
University of South Florida,
2007.
3 520
ABSTRACT: Though often taken for granted, few everyday activities involve so much genuine danger as the hazards associated with motor vehicles. Urban areas are built, modified, and/or deconstructed with motoring in mind. Also true is that few are at as much risk, as are those pedestrians who dare to cross paths with motor vehicles. Unfortunately, all too often, pedestrians are casualties of encounters with the ubiquitous automobile. The Tampa-St. Petersburg-Clearwater, Florida metropolitan statistical area (MSA) has recently been deemed, by one study, to be the nation's second most dangerous MSA for pedestrians. Using information on pedestrian/motor vehicle accident sites, sidewalk location and density, and U.S. Census demographic data, this project focuses on Pinellas County--the most densely populated county in the state of Florida. Issues that were investigated in this case study include: (a) the spatial distribution of pedestrian accident risk within the county, (b) the relationship between the presence of sidewalks and Pedestrian Related Motor Vehicle Accidents (PRMVAs), and (c) the environmental justice implications of these PRMVAs. This thesis seeks to identify spatial and socio-economic trends associated with pedestrian accidents and thus provide an improved understanding of the nature of the danger experienced by pedestrians in the heavily motorized world of west-central Florida.
502
Thesis (M.A.)--University of South Florida, 2007.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 161 pages.
590
Advisor: Jayajit Chakraborty, Ph.D.
653
Automobile accidents.
Urban planning.
Hazards.
Transportation.
Scale.
690
Dissertations, Academic
z USF
x Geography
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2013