Evaluation of Land Use and Transportation Strategies to Increase Suburban Transit Ridership in the Short Term

Final Report Evaluation of Land Use and Transportation Strategies to Increase Suburban Transit Ridership in the Short Term

Gregory L. Thompson, Professor Jeffrey R. Brown, Associate Professor

with Torsha Bhattacharya as Co‐Author for Chapter 4

Department of Urban and Regional Planning Florida State University

A Report Made to:

The Public Transit Office Florida Department of Transportation 30 April 2010

The Florida State University Department of Urban and Regional Planning Room 351 BEL 113 Collegiate Loop PO Box 3062280 Tallahassee, Florida 32306-2280

+1.850.644.4510 http://www.fsu.edu/~durp +1.850.644.8514 direct +1.850.645.4841 fax [email protected] http://garnet.acns.fsu.edu/~gthompsn/my_web/default.htm

30 April 2010

Ms Diane Quigley Public Transit Office Florida Department of Transportation Tallahassee, Florida

Dear Ms. Quigley:

On behalf of Professor Jeffrey Brown and myself, I am pleased to submit to you the final report, “Evaluation of Land Use and Transportation Strategies to Increase Suburban Transit Ridership in the Short Term.” I think that you will find that we addressed all of your comments satisfactorily.

We also took the opportunity to refine our statistical analysis, and we reference its results to an onboard survey conducted of Broward County Transit passengers. This additional work shows a more robust statistical analysis supporting the overall conclusion that reducing transit travel times between all pairs of major origins and destinations is the most fruitful path to increasing transit ridership. There are many ways in which public policy can encourage shorter transit times, including the promotion of TODs on both the origins and destinations of trips. Such TOD policies would have the result of shortening walking time, which is an important component of the overall time spent in traveling from an origin to a destination. With the passage of time, if public policy directs most population and employment growth to TODs, urban regions will be more compact than they otherwise would be, and such compactness will increase transit patronage even more.

Sincerely yours,

Gregory L. Thompson Professor

Evaluation of Land Use and Transportation Strategies to Increase Suburban Transit Ridership in the Short Term Executive Summary

This study seeks to understand the relative efficacy of two classes of policies intended to increase the ridership and productivity of public transit service in Florida. One class of policies seeks to improve transit effectiveness by freezing transit service in the older parts of metropolitan areas (where it is thought that higher densities of population and employment and the presence of pedestrian amenities induce higher levels of transit demand) and directing new population and employment growth to redeveloped areas around transit stops in the older areas. The other class of policies seeks to connect employment and population, wherever it locates, as directly as possible by transit routes. The thrust of transit development of this second category of policies is in the newer rather than older parts of metropolitan areas, because it is in the newer areas where most population and employment growth is located. The study uses two methods, both focused on transit service in Broward County, Florida. The first method, presented in Chapter 4, is statistical and seeks to examine transit ridership between every pair of traffic analysis zones in Broward County in order to understand the importance of variables that might give rise to that ridership. The variables that we used give insight into both hypotheses; the purpose of the statistical analysis is to understand which of the variables are more important. We conducted our analysis with data for 2005, when there were 921 traffic analysis zones in Broward County and over 800,000 pairs of zones. Because of the fact that transit service did not exist between every pair of zones and the further fact that the Census Bureau suppressed data from some zones for confidentiality reasons, we actually analyzed transit ridership between about 550,000 pairs of zones. The statistical analysis developed a relatively weak model for predicting work transit trips between an origin zone and a destination zone, but that model none‐the‐less speaks clearly about variables that increase transit ridership and those that have little impact. In general, the model supports the efficacy of the second set of policies. The most important consideration in attracting transit ridership is to directly connect population and employment. The analysis shows that it does not matter where the population or the employment are located. Reducing travel time from places where people live to places that they want to go, measured by employment, is by far the most important thing policy can do to increase transit ridership. Policy can shorten transit travel time by restructuring routes, by improving headways, by extending coverage, and by increasing speed. It is not important where the employment is located; that located in the CBD does not have a particularly greater draw than that located elsewhere. It is important to serve it all. The conclusions about the ability of TOD developments to increase transit ridership are clouded by the fact that there are no TOD developments in Broward County, and our efforts to identify TAZs with development that is similar to TOD development were not successful. However, our results from the model clearly indicate that shorter walking times to and from transit are highly important for increasing transit ridership. TODs, if designed properly, will reduce walking time to and from transit and thus will increase transit ridership significantly. An implication of this finding is that planning methods that focus on the relationship of developments to stops will be effective if they take into account how well the stops are connected to all destinations in the region. Creating short walking times along attractive paths will boost transit ridership

iii

if the transit stops to which the paths connect are well‐connected to population and employment throughout the region. Another implication is that because both population and employment are dispersed, planners cannot achieve time reductions by implementing direct routes between every pair of origins and destinations. Planners need to think in terms of networks of routes that depend upon transfers. Ideally routes would be frequent, and if the areas traversed are large, routes would be speedy, as well. Transfer points should be designed for easy movement betwee n routes, and fare structures should facilitate transferring. Running express buses from many neighborhoods to CBDs would be ineffective, because CBDs account for so little of regional employment. However, in larger regions an overlay of a regional grid of limited‐stop routes offering much higher scheduled speeds than local buses, interconnecting all important employment concentrations in a region, is an important component of a transit network that achieves higher ridership. The second method used in this study, presented in Chapter 5, is a case study analysis that comes to similar conclusions to those drawn from the statistical analysis of Chapter 4. The case study compares transit development in Broward County with that in Tarrant County, Texas, where Ft. Worth is located. Both counties are the second counties in their respective metropolitan areas in terms of population and employment. Both counties have similarly sized populations, and both counties have grown at about the same rate over the past several decades. Transit service in both counties connects with relatively recently‐created rail commuter service originating in the dominant county of the respective areas. There are major differences in transit policy between the two counties, however. Broward County has no historic central business district, and the transit system has a county‐wide focus. The route structure is a grid that serves all population and employment concentrations in the county. County residents can get from most parts of the county to most other parts where employment is located. Tarrant County, however, contains the Ft. Worth central business district, and transit service historically developed in Ft. Worth as streetcars focused on that CBD. Transit technology in Ft. Worth now is bus, but the route structure still is largely radial in nature focused on the CBD. There also is a CBD‐focused express bus system super‐imposed on the local routes. Many areas of major employment growth in Tarrant County outside of the CBD remain un‐served by transit, however. The city of Arlington, which contains tens of thousands of jobs, remains the largest urban area in the United States without transit service. So, here we have two transit systems laid out according to two different transit policies. Transit in Broward County attempts to connect most origins to most destinations scattered throughout the county with a grid of routes, requiring many passengers to transfer. Transit in Tarrant County attempts to connect many neighborhoods to the CBD, where large numbers of jobs are located. Both local buses and peak period express buses focus on the Ft. Worth CBD. The idea is to serve one destination well, and the destination that is chosen has well‐developed pedestrian connections to jobs. Which policy is the more effective for attracting transit riders? The case study comparison points to the strategy of connecting all population to all jobs throughout the urban region as being the more effective in stimulating transit ridership. Broward County is an environment where transit is not supposed to work. There is no downtown and employment is scattered. Yet, transit in Broward County carries almost 400 percent more ridership per capita than does transit in Tarrant County, while each bus mile operated in Broward County carries about 35 percent more passengers. In summary, we provide two analyses, one statistical and one a case study comparison. Both analyses point in the same direction. The most effective policies for increasing transit ridership and productivity are those oriented to connecting together population and employment that is decentralized throughout metropolitan regions in Florida. It need not be one policy or the other, however. TOD policies can be important in decentralized areas such as Florida by shortening the walk and improving its iv

attractiveness on each end of the transit trip. Over time, to the extent that TOD policies will promote the creation of more compact metropolitan areas with shorter distances between origins and destinations, such policies will further stimulate transit ridership.

Table of Contents

Cover Letter ...... ii Executive Summary ...... iii Table of Contents ...... vi List of Tables ...... vii List of Figures ...... viii List of Maps ...... ix Acknowledgments ...... x Chapter 1. Introduction ...... 1 Chapter 2. The Literature Review ...... 3 Chapter 3. Questions, Hypotheses, and Research Methods ...... 13 Chapter 4. The Statistical Analysis ...... 15 Chapter 5. The Case Study ...... 39 Chapter 6. Discussion and Conclusions ...... 54 References ...... 56 Annotated Bibliography ...... 67

List of Tables

Table 1. Summary of Data Set with Some Observations Removed ...... 34 Table 2. Pair‐Wise Correlations between Independent Variables ...... 35 Table 3. Estimation of Model ...... 36 Table 4. BCT and The T Bus Service, 1984 ‐ 2006 ...... 49 Table 5. Transit Performance by Service Type: BCT and The T ...... 52

vii

List of Figures

Figure 1. Broward and Tarrant Counties Have Similar Population and Growth Rates ...... 40 Figure 2. Productivity (Passenger Miles per Bus Mile) ...... 50 Figure 3. Riding Habit (Passenger Miles per Capita) ...... 50 Figure 4. Efficiency (Cost per Passenger Mile) ...... 51

viii

List of Maps

Map 1. Distribution of 2005 Population in Broward County by TAZ ...... 17 Map 2. Distribution of 2005 Population Density in Broward County ...... 18 Map 3. Distribution of 2005 Median Household Income in Broward County by TAZ ...... 19 Map 4. Distribution of 2005 Percentages of Households with Children in Broward County by TAZ ...... 20

Map 5. Distribution of 2005 Average Household Auto Ownership in Broward County by TAZ ...... 21

Map 6. Distribution of 2005 Employment in Broward County by TAZ ...... 21

Map 7. Distribution of 2005 Employment Density in Broward County by TAZ ...... 24

Map 8. Broward County Routes Superimposed over Employment Density ...... 25

Map 9. Distribution of Long Term Parking Fees in Broward County ...... 26

Map 10. Walkability of Broward County TAZs per SERPM05 Definitions ...... 30

Map 11. Distribution of Neighborhood TODs ...... 31

Map 12. Distribution of Urban TODs ...... 32

Map 13. BCT Adjacent to Miami‐Dade Transit and Tri‐Rail ...... 42

Map 14. The T Adjacent to Dallas Area Rapid Transit and Trinity Rail Express ...... 43

Map 15. BCT Serves Many Destinations; The T Serves One Destination Well ...... 46

Acknowledgments

We could not have completed this research without the assistance of numerous individuals. Assisting in obtaining transit skims and other data from the SERPM05 models were Dave Schmitt of AECOM, Roberto Miquel of Wilbur Smith & Associates, Jeff Weidner, Min Tang Li and Shi‐Chiang Li of FDOT District 4, and Dan Harris of Cambridge Systematics. We are grateful to Spencer Stoleson of Broward County Transit and Nancy Amos (Senior Vice President) and Carla Forman (Assistant Vice President) of the Ft. Worth Transit Authority for granting us interviews and making available data for their systems. We also thank for GIS and other information Lina Kulikowski, Ed Sirianni, and Osama Alaschkar of the Broward Metropolitan Planning Organization, and Jonathan Robertson of Broward County Transit, as well as additional individuals in Broward County Transit and the North Central Texas Council of Governments. We thank Minxing Chen for help is assembling our data set for Stata analysis. Finally we thank Mark Ellis of the Department of Geography, University of Washington for offering insights on interpretation of negative binomial regression models. We are immensely thankful for the cooperation that we received from all of these individuals. We remain responsible, however, for possible errors and omissions as well as for opinion expressed in this document.

Evaluation of Land Use and Transportation Strategies to Increase Suburban Transit Ridership in the Short Term: An Empirical Analysis Based on Broward County

Chapter 1. Introduction

Recent discourse over high energy prices and deleterious climate effects of carbon dioxide emissions has renewed an ongoing public debate over the merits of possible policies for improving the efficiency and effectiveness of public transportation in U.S. urban regions. Some analysts see improved transit as a possible substitute for auto travel, though they are troubled by the marginal role that transit plays today in meeting urban transportation demands in most urban regions. What policies possibly could pull transit out of such marginalia? One line of the debate addresses this concern by proposing that urban form needs to be changed, in order to make transit a possible substitute for personal transportation. The argument goes that the fixed bus and rail routes of transit cannot connect most trip beginnings and ends in sprawling suburban development; development thus needs to be reorganized and densified around transit stops. If that happened, transit would substitute for many auto trips, as it is thought to do in many European cities. There is a counter argument, however, suggesting a different line of policy. It is not development that needs to be reorganized, according to the counter argument, but transit services. Fixed route transit services can, in fact, serve sprawling land uses if the routes are not primarily focused on a central business district. Though it is doubtful that transit’s marginal role can be overcome entirely, it is possible through route restructuring to double or triple transit usage in many metropolitan regions while improving transit productivity. The purpose of this research is to examine the relative efficacy of these two arguments.

The examination focuses on transit in Broward County, Florida. Bus transit in the county is organized in a grid pattern along the county’s major arterials. Bus service is designed so that residents from most parts of the county can reach employment in most parts of the county, even though most residences and most jobs are sprawled. There is, however, considerable variation in a person’s ability to reach jobs using transit, depending upon where they are located in the county. There also are clusters of higher density employment in the county in areas where urban design has created walkable communities. Transit serves such clusters. The examination has two parts:

1. One part of the study compares the transit performance of transit service in Broward County to that in another region of the country that is similar to Broward County but uses a different approach to providing transit service. The comparison region is Tarrant County, Texas, in which is situated Ft. Worth. Both counties contain the main secondary centers of economic activity and population in their respective metropolitan areas. Both counties have approximately the same population size that has been growing at approximately the same rapid rate over the previous twenty years. Transit service is organized differently, however. Broward’s is a grid of routes that serves sprawling jobs and residences. Tarrant’s, on the other hand, is largely a radial system of routes focused on a traditional central business district. Although Tarrant has a traditional central business district, which has heavy employment, some residences, and a design making walking possible, (Broward lacks such a center), much of the population and employment in Tarrant County is sprawled. Which transit service design is the more effective and efficient, and why?

2. The second part of the study examines variations of transit usage within different parts of the county in comparison to how easy people can reach jobs from those parts, or in comparison to the density, diversity, and design of the neighborhoods around which the usage occurs. To what extent does access to jobs determine usage? To what extent does urban design determine usage?

Our study presents what we know about the topic from other studies that have been done, ending with the questions that the study addresses. It then elaborates upon the two methods for addressing the questions with information from Broward and Tarrant Counties. The next two chapters carry out the methods, followed by a discussion of the results, and finally, policy conclusions.

Chapter 2. Literature Review

The literature review discussion proceeds as follows: 1) works that comment on the role of policy for increasing the importance of transit is U.S. metropolitan areas; 2) works that provide a descriptive overview of transit ridership; 3) works that emphasize external factors that affect ridership (including land use); and 4) works that emphasize internal factors that affect ridership (including service planning decisions).

Role of Policy in Affecting the Role of Transit and the Automobile Based on History and European Example In market economies, of which the U.S. is an example, consumer sovereignty is relied upon as the principle instrument for determining the mix and prices of goods and services to be provided to society. How far can public policy affect consumer choice in such societies? This is the question taken up by Pucher and Levefre (1996) and Jones (2008). Both works examine public behavior in travel and locations for living over a long period of time in many different market societies. The societies that they examine reflected many cultures and embraced many different policies for the provision of transportation systems and the regulation of land uses. Both works come to the same conclusion: as people become wealthier, regardless of cultural background or public policy, they choose the mobility afforded by the automobile and light trucks, and they prefer more dispersed living. Public policy can have an effect on public behavior on the margin, however. Public policies that impose costs on those choosing to drive to account for the social costs of driving and who provide high quality public transportation systems result in marginally fewer VMT per capita compared to those that do not. How far can public policy go in this direction? Growing Cooler (Ewing, et al 2008) argues that it is possible to accommodate most new population and employment in the U.S. between now and 2050 within existing urban areas at higher densities and in mixed use development, including revived CBDs through changed land use regulations. By doing so, they argue that VMT would be reduced by up to 40 percent compared to accommodating new development in low density sprawl on the edges of urban areas. Driving and the Built Environment (Committee on the Study on . . . 2009) counters Growing Cooler. Driving and the Built Environment argues that existing patterns of sprawled development are a reflection of public preference and are unlikely to be changed. The history of land use regulations intended to reverse the pattern of development has shown such policies to be ineffective. Public policy, such investing in transit systems to mold future development, also has had limited impact on development. The only rail transit investments that have had an apparent effect on development patterns are those that have been built in areas already chosen by the private market as ripe for development. What policy can be effective? Pricing that reflects the social cost of choices. Some believe in the power of public policy to alter the form of society, because they believe that past public policy created the existing auto‐dominated transportation system in the U.S.. Rail transit, for example, almost disappeared in the U.S. because governments at all levels allowed General Motors and other automobile interests to buy viable rail transit systems and then junk them, forcing patrons to turn to automobiles. Post (2007), however, provides strong arguments against this position. Post documents how technological development created a competitor to the streetcar that was cheaper and offered most of the qualities that rail transit offered.

Descriptive Overview of Transit Ridership

In the postwar period, transit ridership experienced a long decline followed by a number of recent peaks and valleys. Jones (1985), Vuchic (2005), and Post (2007) provide general discussion of the longer‐term trend, emphasizing the decentralization of urban areas and competition with the automobile as among the primary causes for transit’s postwar decline. By the 1970s and 1980s, Jones (1985) observed that transit was largely limited to serving two markets: transit‐dependent individuals and commuters traveling to and from jobs in the central business districts in the nation’s largest cities. Transit‐dependent individuals are defined as individuals who for reasons of age, income, or disability lack either access to or the ability to use an automobile and thus rely on public transit as a primary means of transportation. Researchers have typically measured transit dependency using variables such as household income, age, race, ethnicity, immigrant status, and number of automobiles in the household. National transportation surveys (such as those conducted in 1983, 1990, 1995, and 2001) regularly report that individuals who fall into certain demographic group categories (defined using these variables) are disproportionately transit users. Using data from the 2001 National Household Travel Survey, Pucher and Renne (2003) found that the poor, blacks, Hispanics, and those with low levels of vehicle ownership are more likely to use transit than are other groups. Particularly important is the latter variable. The same survey found, however, that the numbers of individuals placed into the demographic categories we use to define transit dependency declined between the 1995 and 2001 surveys. The surveys also reported that even for transit dependent groups, transit is not their primary mode of transportation—the automobile is. During the mid and late 1990s, a series of articles appeared documenting a large decline in transit ridership during the early part of the decade and speculating that public transit was headed for rough times. However, in the late 1990s and on to the present, ridership (measured in terms of unlinked passenger trips, but not mode share) increased. Pucher (2002) identified the economic recession of the early 1990s, and particularly its effect on employment in New York, as the driving force behind the ridership decline of the early 1990s. He cites the economic recovery of the 1990s, rising gasoline prices, stable fares, improved service quality, and the expansion of rail transit services as among the key contributing factors for the ridership rebound of the latter part of the decade. The limitation of this article is that it is purely descriptive; Pucher makes no effort to examine other potential causes using more sophisticated multivariate techniques. Thompson and his coauthors (2006) examine the ridership trend in the nation’s largest cities. Focusing on the period between 1990 and 2000 in all metropolitan statistical areas that had more than 500,000 persons, they paint a picture of ridership that grew faster than population growth in areas that most researchers would not suspect, namely in the metropolitan areas of the auto‐oriented west. They note that service grew in most parts of the country as well. They also find that service productivity (measured in terms of load factor, or the ratio of passenger miles to vehicle miles) declined throughout the country, but experienced the smallest decline in the West. In short, western cities added a lot of service and gained a lot of riders in doing so. However, this purely descriptive piece does not explain why transit is growing in many “surprising” places.

The External Factors that Influence Transit Ridership

The external factors that influence transit ridership include: urban structure (decentralization), local land use patterns, automobile ownership levels and costs, and regional economic conditions. The particularly relevant literature for this study is that which focuses on local land use patterns, although we briefly discuss some of the other literature as well.

Urban Structure (Decentralization)

Meyer, Kain and Wohl (1965), Jones (1985), and Vuchic (2005) cite urban decentralization as one of the primary causes of the long‐term decline in transit use in the postwar period. The corollary is that transit use is positively tied to the degree of urban centralization, and in particular, the strength of the central business district (CBD) as a locus of economic activity. Mierzejewski and Ball (1990) found some support for this notion, where choice riders (those who have access to an automobile but choose to use transit) are concerned. In a survey of 4,000 persons in 17 metropolitan areas, they found that 82 percent of choice riders who used transit worked in the central city. The conventional wisdom is that transit works best when it focuses on serving the CBD commute market (Ferreri 1992; Meyer and Gomez‐Ibanez 1981; Pisarski 1996; Taylor 1991). The implication is that transit agencies should structure their service to feed the CBD and provide high quality service to that destination, because that, the literature would suggest, is where riders wish to travel. An agency decision to serve other destinations, particularly those dispersed throughout the suburbs, is criticized for being an inefficient use of public subsidy (Taylor 1991) and for resulting in low service productivity (Ferreri 1992; Meyer and Gomez‐Ibanez 1981). There have been a handful of studies that have examined the link between urban structure and transit ridership using statistical techniques. Some studies have found a close link between decentralization and transit ridership while others have found a more complicated set of relationships between these variables. Most studies have used the relative strength of the central business district as the measure of urban structure. Hendrickson (1986) examined the relationship between transit commute mode share and the number of jobs in the central business district in 1970 and 1980 for 25 large metropolitan areas using a series of multivariate models. The first multivariate model estimated ridership in 1970 as a function of CBD employment in 1970 (R square = .96), the second model estimated ridership in 1980 as a function of CBD employment in 1980 (R square = .90), and the third model estimated ridership in 1970 as a function of both CBD employment and the total number of workers in the metropolitan area (R square = .98). He then estimated two change models, one with a dummy variable for Sunbelt cities (R square =.77) and one without (R square = .66). Finally, he estimated a change model with dummy variables for both Sunbelt cities and those with fixed rail systems (R square = .81). Hendrickson (1986) found strong relationships between CBD employment and transit commute mode share. He found positive, statistically significant effects on transit commute mode share from the Sunbelt dummy variable, and negative, statistically significant effects from the fixed‐rail dummy variable. However, his study suffers from two shortcomings, which include: 1) lack of control variables and 2) mixing of cities with significant differences in both the size of the CBD and the transit commute mode share. Particularly problematic is the inclusion of New York, which dwarfs the other cities on both variables, in the same analysis. 5

Gomez‐Ibanez (1996) conducted a more sophisticated analysis of the relationship between transit ridership and decentralization in Boston. He used a time series approach that examined ridership betwee n 1970 and 1990, and included variables that controlled for fare, per capita income, and service level. His measure of decentralization was the number of jobs in the city of Boston. He found: 1) a 1 percent decline in the percent of jobs in the city of Boston was associated with between a 1.24 percent and 1.75 percent decline in ridership; 2) a one percent increase in real per‐capita incomes was associated with a 0.71 percent decline in ridership; 3) a one percent increase in fares was associated with a .22 to .23 percent decline in ridership; and 4) a one percent increase in vehicle miles of service was associated with a .30 to .36 percent increase in ridership. His models accounted for nearly 90 percent of the variation in transit ridership from 1970‐ 1990. Gomez‐Ibanez concluded that transit ridership in Boston has been strongly influenced by the decentralization of employment. However, the definition of employment is problematic and measures jobs throughout the city of Boston as opposed to jobs inside the central business districts of Boston and Cambridge, which the author states he had hoped to measure. Two recent statistical studies have found very different results. Brown and Neog (2007) examined the relationship between transit ridership and urban structure in all U.S. metropolitan statistical areas with more than 500,000 persons in 1990 and 2000. They define urban structure as the percent of MSA employment in the CBD and use two measures of transit ridership, passenger kilometers per capita and transit commute mode share. The authors controlled for variables measuring fare, service frequency, service coverage, motor fuel price, urban area population density, regional unemployment rate, and the percent of households in each metropolitan area that lacked access to an automobile. They found no statistically significant links between the percent of MSA employment in the CBD and transit ridership. The authors found the strongest links between two service variables (service frequency and service coverage) and transit ridership. They also found a strong relationship between the percent of MSA households that do not own an automobile and transit ridership. Brown and Thompson (2008a) examined the relationship between transit ridership and urban decentralization in Atlanta from 1978 to 2003. The authors used linked passenger trips as their ridership variable. They created three employment variables to measure the degree of employment decentralization: percent of employment in the CBD, percent of employment outside the CBD but inside the transit service area, and percent of employment outside the transit service area. They controlled for fare, service level, motor fuel price, and population decentralization in their time‐series analysis. They also included a variable measuring the percent of transit service delivered by rail transit. They found that transit ridership is strongly and positively linked to the strength of employment inside the transit agency service area (outside the CBD) and is strongly and negatively linked to the strength of employment beyond the transit agency service area. The authors found no association between the strength of the CBD and transit ridership in Atlanta. The authors also noted that transit ridership is more strongly linked to the decentralization of employment than to the decentralization of population, and that fare levels and the absolute amount of transit service are also associated with transit ridership. The authors infer that MARTA is successfully connecting transit patrons to dispersed employment locations within its service area, but that the failure of the service area to include some major poles of employment growth is depressing MARTA ridership.

Local Land Use Patterns (Transit‐Oriented Development)

Over the past two decades, there has been a great deal of interest in the relationship between local land use patterns near bus and rail transit lines, stops, and stations and transit ridership. Often lumped under the label of transit‐oriented development (TOD), this body of literature hypothesizes that the density, land use mix, and urban design characteristics of a neighborhood can influence individual mode choice decisions (APTA 1987; Bae 2002; Beimborn 1991; Caltrans 2003; Cervero 1998; Cervero 2006a; Cervero 2006b; Dunphy and Porter 2006; Hemily 2004; Jun 2008; Knaap, et al 2001; Schlossberg and Brown 2004; Song 2005; Urbitrans 2006). There is an extensive literature on the subject, much of which builds on work by Robert Cervero. The primary hypotheses about transit‐oriented development and its relationship to ridership are voiced in books by the team of Bernick and Cervero (1997) and Cervero (1998) on his own. Both books rely on case study analysis to argue that developments characterized by higher density, more mixed uses, and more pedestrian‐friendly designs tend to have higher transit ridership. Therefore, the suggestion is made that if metropolitan areas promote these kinds of developments they should expect to see auto use decline, while transit use, walking, and perhaps bicycling increase in importance. Indeed, Parker and co‐authors (2002) found associations between transit‐oriented development and transit mode share in their case study of transit‐oriented development in California. Lund and Willson (2005), on the other hand, found modest ridership results in their case study of transit‐oriented development along the gold line light rail line in suburban Los Angeles. They surveyed the residents in 37 multi‐family buildings located within 1/3 mile of rail stations. Of 1,595 housing units surveyed, they obtained responses from 221 units recording information about 477 trips. They found few transit‐dependent residents in their survey. Respondents were primarily white, worked in professional occupations, and owned one or more automobiles. Few residents had low incomes. About 75 percent of respondents rarely or never used transit, while 15 percent regularly used transit. The authors noted that respondents were more frequent transit users after they moved to their current place of residence, but noted that there might be a self‐selection bias at work. Essentially, they found that TOD in this particular corridor was too expensive to be occupied by transit riders and was instead occupied by wealthier professionals, who tend not be transit riders. The mismatch between TOD residential profiles and transit user profiles is frequently noted by TOD skeptics. Residential self‐ selection has also been cited by TOD skeptics who assert that the people who live in residential TODs are people who were already predisposed to engage in more use of non‐automobile transport modes. There are, however, a number of quantitative studies that have found a connection between TOD‐as sociated elements and ridership. These studies have examined the relationship between transit ridership and distance, density, diversity, and design. Cervero (1993) discussed several studies that examine the ridership characteristics of projects located near rail transit stations. He cites a 1989 San Francisco Bay Area study found that 35 to 40 percent of residents living near three Bay Area Rapid Transit District (BART) stations used public transit. He also cited a 1987 Washington, DC study found that rail and bus transit mode share declines by 0.65 percent for every 100‐foot increase in distance of a residential site from a rail transit station. The same 1987 study found that ridership was higher at downtown than at suburban work sites and that ridership declined steadily as distance to the station increased. All these studies essentially examined the correlation between transit mode share and distance to a rail station. They did not control for other factors that might influence an individual’s decision to use public transit (fare, service quality, auto access and cost, or the ease with which travelers could reach their destinations).

The Institute of Urban and Regional Development (2004) reported the descriptive results of residential studies showing that: 1) workers living near the San Francisco area’s Bay Area Rapid Transit District (BART) heavy rail line were six times more likely to use it for commute trips than the average Bay Area resident; 2) workers living near light rail transit in Silicon Valley were five times more likely to use transit for commute trips than average area residents; and 3) people living near transit in Washington, DC have high transit mode shares that decline with increased distance from a transit station. The authors also summarized a set of office and retail studies that showed: 1) 50 percent of those working within 1,000 feet of a downtown Washington Metro station used rail to get to work; 2) 60 percent of customers at a downtown San Diego shopping center located two blocks from light rail arrived either by transit or by foot; and 3) 34 percent of patrons at a downtown San Francisco shopping center that has a direct connection to BART arrived by transit. More studies have focused on the link between density and transit ridership than any other factor. These studies have their roots in early work by Pushkarev and Zupan (1977). Parsons Brinckerhoff (1996) found, in a study of 17 cities with light rail or commuter rail, that residential densities had a strong effect on transit boardings. Spillar and Rutherford (1998) also documented a density effect in their analysis of Denver, Portland, Salt Lake City, San Diego, and Seattle. They noted, however, that density appeared to have a stronger relationship with transit ridership in low‐income neighborhoods. The Institute of Urban and Regional Development (2004) also presented a set of multivariate models from studies for the San Francisco Bay Area and Arlington County, Virginia that indicate particularly strong relationships between the density of the land use and transit ridership. Overall, the authors concluded that residents living in TODs usually patronize transit five to six times as often as the typical resident of a region. The authors acknowledged that self‐selection bias might be an issue in the residential studies they discuss. Cervero (2002) found a modest density effect on ridership (elasticity between 0.2 and 0.6) in his study of Montgomery County, Maryland. Kuzmyak and his coauthors (2003) also reported that transit ridership tends to be higher at higher densities. Citing work by Parsons Brinckerhoff for Chicago, they reported that a 10 percent increase in residential density is correlated with an 11 percent increase in per‐capita transit trips and a 13 percent increase in transit mode share. Citing work by Levinson and Kumar for a national study of the U.S., they reported that density only becomes relevant to mode choice at densities higher than 7,500 persons per square mile. Citing work by Frank and Pivo in Seattle, they also noted that transit requires workplace densities of 50‐75 employees per gross acre and residential densities of 10‐15 dwelling unit per net residential acre to achieve significant commute mode shifts. Citing a study by Nelson/Nygaard for Portland, Oregon, they noted that housing density and employment density accounted for 93 percent of the variation in daily transit trip productions and attractions across the region. The authors cautioned that in many of these studies self‐selection bias may be a concern. Kuzmyak and his coauthors (2003) also presented the results of studies indicating that transit use tends to be higher in areas characterized by mixed land uses. However, they cautioned that many of these environments tend to also be characterized by higher densities, so separating the mixed use effect from the density effect is difficult. Citing work by Messenger and Ewing in Florida, they noted that more balanced (jobs and workers) areas tend to have higher transit mode share. Citing a study by Cervero of 57 suburban activity centers, the authors noted that centers with on‐site housing had 3 to 5 percent more transit, bike, and walk trips. Song and Knaap (2004) examined public preferences for mixed land uses in a hedonic price model study of single family home purchases in Washington County, Oregon. At the time of their study Washington County was the fastest growing part of the Portland metropolitan area and featured a wide 8

variety of housing choices, including mixed use transit oriented development built around light rail stops. Song and Knaap found that single family home buyers valued locations with amenities such as parks, open space, and the ability to walk to neighborhood retail. On the other hand, such buyers discounted proximity to multi‐family dwellings, large employment concentrations, and properties with small lot sizes. Transit‐oriented development is also characterized by more transit and pedestrian‐friendly urban design. Urban design is the hardest of the 3 Ds (density, diversity, design) to measure, but there have been a few studies on the effect of urban design on transit ridership (Canepa 2007: Crane and Crepeau 1998; Hess and Lombardi 2004). Cervero (2000) found that urban design, and particularly sidewalk provisions and street dimensions, significantly influence whether someone reaches a rail stop by foot or not in his study in Montgomery County, Maryland. He asserted that conversion of park‐and‐ ride lot s to transit‐oriented developments holds considerable promise for promoting walk‐and‐ride transit usage in years to come. Cervero (2006b) found a relationship between street connectivity and an individual’s decision to use transit in his study of people living near rail stations in California.

Other External Factors

The literature has also identified a number of other factors beyond the control of agency managers that can influence transit ridership. These factors include population and population growth, regional economic conditions, housing costs, and personal income (Cambridge Systematics 1995, 1998, 2005; Liu 1993; Taylor, et al 2002; Taylor, et al 2003). Some particularly important additional external factors relate to the automobile. Studies by Brown and Neog (2007), Liu (1993), and Taylor, et al (2003) have all highlighted the important relationship between the share of carless households in a metropolitan area and transit ridership. Studies by Dueker and his coauthors (1998) and Mierzejewski and Ball (1990) have noted the important role played by parking availability and cost in influencing transit use.

The Internal Factors that Influence Transit Ridership

The internal (agency‐controlled) factors that influence transit ridership include: fare policy, service frequency, service coverage, service orientation, and targeted marketing efforts. Particularly relevant for this research are those studies that consider service frequency, coverage, and orientation, which collectively help to define the level of accessibility a transit system provides to the community it serves. General Discussion

There is a sizeable descriptive literature that introduces service strategies that might influence transit ridership in particular settings—without evaluating the performance of the particular strategy. One author who has conducted significant past research in this area is Robert Cervero (1994). Cervero identified timed transfer systems, paratransit services, reverse commute and specialized runs, employer‐sponsored van pools, and high‐occupancy‐vehicle and dedicated busway facilities as transit service strategies that might result in higher ridership in decentralized areas. He reemphasized these kinds of service strategies in his international case study of transit metropolises (Cervero 1998). Working 9

with Beutler (1993) he discussed the use of bus rapid transit services and free market paratransit services as possible service strategies in certain urban environments. Using case studies of eight transit agencies in the United States and Canada, Charles River Associates (1997) identify feeder bus, fare integration, express bus, times transfer, pass programs with universities, and a fareless square as promising strategies in certain environments. However, these same authors conclude that policies that make private vehicle use less attractive will have a larger positive effect on ridership than policies that make transit more attractive. A number of authors emphasize the role of targeted marketing and market segmentation as strategies to increase ridership among specific rider groups (Elmore‐Yach 1998; Frumkin‐Rosengaus 1987; Haas 2005; Taylor, et al 2003; TranSystems 2007). Cambridge Systematics (1995, 1998, 2005) uses repeated surveys of agencies that experienced ridership increases to identify fare policies, service adjustments, and marketing efforts as key factors that affect transit ridership. Haas (2005) discusses the use of Eco pass programs, guaranteed ride home programs, day passes, and on‐line fare media sales programs. Skinner (2007) found, however, that transit services targeted toward particular ridership markets might have unexpected negative effects. Miami‐Dade Transit operates a number of routes that seek to serve the elderly population, and connect social service and other destinations to residential areas where the elderly reside. However, these routes have low elderly and non‐elderly ridership, and as a result, very poor performance, because they are slow and indirect. Far more elderly people ride the direct and frequent bus services in Dade County running on arterial roads that are targeted toward the general ridership. These are the bus routes in Dade County characterized by high performance. Finally, the California Department of Transportation (2003) uses a survey of actual and potential riders to identify service reliability, convenience, comfort, and safety as key factors that might influence an individual’s decision to ride transit. As noted above, none of these articles evaluates the performance of the strategy or factor that the authors describe. Fare Policy

There is an extensive body of literature that documents the relationship between fare levels and ridership (Brown and Thompson 2008a; Gomez‐Ibanez 1996; Kain 1997; Kohn 2000; Pucher 2002; Taylor, et al 2003). Kyte (1986) found an important relationship between fare and ridership in his study of Portland. Taylor and his coauthors (2002) documented the importance of fare policy in their U.S. national study, and so did Kohn (2000) in his Canadian study. Kain and Liu (1999) noted the importance of fares in their study of Houston and San Diego, as did McLeod, et al (1991) in their time‐series analysis of Honolulu. TRL Limited (2004) summarizes the results of an extensive set of empirical studies. They report that fare elasticities vary depending on both mode and time‐frame. Bus fare elasticities average around ‐ 0.4 in the short run, ‐0.56 in the medium run, and ‐1.0 in the long run. Rail transit elasticities tend to be higher than those for bus for suburban rail services and smaller than those for bus for heavy rail. Off‐ peak ridership tends to be twice as responsive to fare changes as peak period ridership.

Service Frequency and Coverage

There is also a large literature that documents the relationship between the service provided by an agency and transit ridership (Gomez‐Ibanez 1996; Kain 1997; Kohn 2000; Kyte 1986; Pucher 2002; Taylor, et al 2003). A smaller literature has broken down service into two components: frequency and 10

coverage. Both are hypothesized to positively influence ridership. Brown and Neog (2007), and Thompson and Brown (2006) found positive effects of both service frequency and service coverage in their national analyses of transit ridership in large U.S. metropolitan areas in 1990 and 2000. Brown and Neog (2007) report elasticities for both service and coverage in the 0.7 to 1.0 range. Evans (2004) provides an overview of empirical work on the relationship between transit service frequency and ridership. He found that ridership does respond to service frequency and schedule changes (elasticity = 0.5), and that the largest responses are found in higher income areas that previously had very infrequent service. In more traditional transit areas, the ridership response was more modest. Pratt and Evans (2004) examined the relationship between coverage and ridership in a routing study. The authors found elasticities in the range of 0.6 to 1.0. The authors noted that the largest ridership increases occurred when the system emphasized “high service level core routes, consistency in scheduling, enhancement of direct travel and ease of transferring” (Pratt and Evans 2004, 5). The authors claim that new and expanded systems of the hub‐and‐spoke variety produced slightly higher ridership than grid systems, although there were no controls for other possible variables. Taylor, et al (2003) also noted that route coverage was an import influence on transit ridership.

Service Orientation

A particular interest in this project is the role of service orientation as a factor influencing transit ridership. Regrettably, there have been few studies that explicitly examine service orientation (Hadj‐ Chikh and Thompson 1998; Mieger and Chu 2006; . Thompson and Matoff (2003) conducted an early case study analysis of nine cities in which they distinguished between radial and multidestination (grid) oriented transit systems. The authors obtained data on transit system profiles and transit performance from 1983 to 1998 for transit systems in Cleveland, Columbus, Houston, Minneapolis, Pittsburgh, Portland, Sacramento, San Diego, and Seattle. The performance measures include: cost per passenger mile, peak‐to‐base ratio, passenger miles per capita, and vehicle miles per capita. The authors then compared systems that met their definitions of multidestination versus radial service orientations on each of these measures. The authors found that multidestination systems were more effective (that is, had higher ridership), nearly as efficient (about the same cost), and more equitable (lower peak‐to‐base ratio) than radial systems. More recently, Thompson and Brown (2006) explored the relationship between service orientation and ridership using a statistical analysis. The same authors have also recently explored the relationship between service orientation and service productivity (Brown and Thompson 2008b). In their ridership study, identify and examine the key determinants of transit ridership change between 1990 and 2000 in U.S. metropolitan statistical areas (MSA) with more than 500,000 persons. Among the key variables they examine is a service orientation that distinguishes between multidestination and traditional service orientations. The authors found that transit is growing most rapidly in the non‐ traditional markets of the West but that much of the regional variation is a function of the particular service coverage, frequency, and orientation decisions made by transit agencies in this region. Service coverage and frequency are the most powerful explanatory variables for variation in ridership change among MSAs with 1 million to 5 million people, while a multidestination service orientation is the most important explanation for variation in ridership change among MSAs with 500,000 to 1 million people. A

weakness of the analysis is the definition of the service orientation variable as a binary variable, as opposed to a continuous one. Their productivity paper substitutes a quantitative variable that measures the percent of transit routes that do not serve the CBD (Brown and Thompson 2008b). They find that decentralized service orientation does not lead to diminished productivity. In fact, the signs on the coefficient for this variable in their statistical models are positive, although not statistically significant.

Chapter 3. Questions, Hypotheses, and Research Method

Synopsis of Literature Findings The literature review suggests general agreement about variables under the control of transit managers that influence transit patronage. These variables include fares, coverage of service, and frequency of service. Service orientation also is suggested by some researchers as important but not mentioned by others. The literature is not in agreement about how the design of land uses affects transit patronage. Some researchers conclude that residential areas must be developed to a minimum of seven dwelling units per acre and must be designed with pedestrian amenities before significant transit ridership will occur. This literature generally is silent about the design of the transit service to which the development would be connected; indeed, some such developments have been built with no transit at all. The idea is that transit may come eventually. Where would such service take people when it did eventually come? Researchers in this category generally do not say. They appear to assume that transit riders and routes have only one destination: the central business district. To these researchers, effective government policy for shifting motorists to transit is to give incentives to the development community to build higher density residential developments with excellent pedestrian connections to attractive‐looking bus and rail stops, regardless of whether there are any buses or trains using the stops, and if there are, how effectively the bus or rail services can take passengers where they want to go. Advocates for such developments call for a mixture of uses, but in practice the non‐residential parts of such developments typically are confined to such residential‐serving activities as convenience stores, coffee houses, and dry cleaners. Other researchers conclude that the design of residential areas is not sufficient to increasing transit use. If a resident gets to a bus stop, she then is confronted with the question, how efficiently will the bus take her to where she wants to go? The design of the transit service determines the answer to that question. That is the question to which public policy should concern itself, according to these researchers.

Hypothesis Which of the two types of public policy is more effective in increasing transit patronage? The two policies may be summarized as: 1. Encouraging Transit Oriented Development around transit stops will increase transit ridership, regardless of the quality of the transit service linking that stop with every place a person might wish to go. 2. Encouraging the creation of transit service designs that get passengers where they wish to go will increase transit patronage greatly, even if the walking environment on the origin or destination end of the trip is not addressed. Are both types of policies important? The hypothesis that is tested in this research is that, although both policies are effective to increasing transit patronage, the second is more effective.

Research Design We test the hypothesis by two methods of empirical research, each focused on bus transit in Broward County, Florida. We proposed Broward County in the research proposal, because planners in Broward County consciously designed transit service to take riders living in virtually any part of the urbanized parts of the county to reach destinations located virtually in any part of the county. We thus are able to test the hypothesis that this method of organizing transit service is effective. 13

One of the two research methods for testing the hypothesis is a case study comparing the effectiveness and efficiency of transit service in Broward County to transit service in another urban area that has focused on serving well one important destin ation. The comparison transit system is The T in Fort Worth, Texas, which focuses on serving well a traditional downtown and does not attempt to take patrons to many other important destinations in Tarrant County, the county in which Fort Worth is located. In the case study, we compare population magnitude and growth in Broward and Tarrant counties, the two counties’ places in their respective urban regions, the manner in which transit service has evolved in the two counties, and the results of the evolution in terms of passenger miles per capita, passenger miles per vehicle mile, and operating expense per passenger mile. Passenger miles per capita reveals how well each transit systems penetrates their respective markets, passenger miles per vehicle mile reveals the average number of passengers on board a bus each mile it operates and reveals productivity, while operating expense per passenger miles shows the resources devoted to moving each passenger one mile. The second research method for addressing the hypothesis is a statistical analysis of transit patronage and its relationship to service design and urban design in Broward County. It is possible to objectively measure how well transit links the population of different parts of Broward County to jobs located throughout the county, and doing so reveals a great deal of variation. Some areas of the county are well connected to jobs; some are not. There also is a great deal of variability to transit ridership within the county. Some areas of the county produce much more ridership than other areas. What accounts for the variability in ridership? Is it the variability in the ability to reach jobs? Is it variability in the urban designs of the various areas? Is it merely reflective of the distribution of population? In the statistical analysis, we construct models explaining the variability of transit patronage in the county in terms of the distribution of population and employment, the distribution of urban design features, and the distribution of the quality of transit service in each area, measured by the ability to reach employment in the county. The models allow us to test our hypothesis as well as counter hypotheses.

Chapter 4. Statistical Analysis

Introduction From the Census Transportation Planning Package we obtained Year 2000 estimates of how many transit work trip passengers traveled between each pair of the 921 TAZs in Broward County. There are approximately 848,000 pairs of TAZs (called interchanges), between which transit passengers potentially could travel. In this chapter we specify and estimate a model that explains the flows of transit work trip passengers from one zone to another. The estimated parameters of the model allow us then to test hypotheses about the relative importance of the explanatory variables in determining transit ridership. This chapter explains the specification and estimation of the model. The major transit system that carries most transit passengers within Broward County is Broward County Transit. Broward County Transit operates a grid of bus routes on most of the arterial roads serving the populated parts of the county, as well as community circulator routes, and dial‐a‐ride. Chapter 5 offers a synopsis of Broward County Transit’s current service structure and how it evolved. Here, before constructing a model to predict its work trips, we provide a thumb nail sketch of its patronage as revealed by an on‐board survey. In 1997 the Center for Urban Transportation Research conducted an on‐board survey of Broward County Transit passengers, the results of which we obtained from Broward County Transit staff (Center for Urban Transportation Research 1997). Sixty percent of the riders belonged to households earning less than $20,000 per year, while nine percent earned more than $50,000. Of the riders not coming from home, 46% were coming from work, 17% from shopping, 12% from visiting friends and relatives, 12% from non‐categorized purposes, 6% from school, and 6% from the doctor. Forty‐seven percent of riders had no car in their household while another 32 percent had one car. Slightly more than half of the riders were female. Slightly more than half of the riders were black, 35% where white, and 10% were Hispanic. Sixty‐four percent of the riders transferred one or more times to complete their trips. The survey suggests largely a transit‐dependent ridership riding between many pairs of origins and destinations. It suggests that important explanatory variables in a model predicting transit ridership would include auto ownership (more autos would lead to less ridership), income (higher income would lead to less ridership), and the presence of children in the household (more children would lead to higher ridership). Employment, employment density, and parking fees also would be important variables for the destination of a trip. Employment would be important, because almost half of the ridership is work trip related (and our dependent variable is work trips). Employment density would be important, because riders would have to have to walk from the bus to their work place destination, and walk trips likely would be shorter in higher density work environments. We think that measures of walkability of both the origins and destinations of trips would be important, because the transit dependent users would have to walk to their destination, and most likely would walk from their residence to the bus (though they could park and ride or be dropped off). We use these considerations and others in specifying our model for predicting work trips.

Model Specification Functional Form The nature of the transit patronage data influences the choice of model specification. The flow of passengers between any pair of zones is a count variable, which is a non‐negative integer. In Broward County, the flow between many pairs of TAZs is 0, and the maximum flow is 41. The mean is .015, indicating a large number of 0 flows. As a consequence, the variance (.172) is large compared to the

mean (.015). These characteristics suggest that we specify an exponential model to be estimated by negative binomial regression, a form often used in models of migration (Greene, 1993; Long and Freese, 2006).

Variables Describing Trip Production An exponential model predicting passenger flows is a form of gravity model. In gravity models, explanatory variables typically depict the mass of the originating and destination zones, respectively, and the transportation friction separating the two zones (Haynes and Fotheringham, 1984). We choose mass variables based on the hypotheses that we wish to test and the data that we can obtain. The hypotheses that we are testing are that urban form variables influence transit trips between one zone and another. Population and population density are variables typically used to indicate mass of the originating zone for work trips. In addition, we use median household income (Map 3), percentage household with children (Map 4) and average auto ownership (Map 5). We use those, which are available at the TAZ level in the Southeast Regional Planning Model, known as SERPM05 (Corradino 2008) and the Census Transportation Planning Package 2000 (CTPP). In addition, we create a multiplicative variable to denote the degree to which the originating zone has characteristics of a TOD. We describe the origin zone TOD or neighborhood TOD variable below. The distribution of population in Broward County by TAZ is shown in Map 1. The map indicates heavy population concentrations across the western part of the county, which partly reflects the larger size of traffic analysis zones in the western part of the county. Map 2 corrects for TAZ size and displays the distribution of population density. Zones with higher population density are scattered throughout the built up parts of the county. The highest density zones are cluster west and north of downtown Ft. Lauderdale, but there also are many higher density zones in the eastern part of the county. Small zones of higher population density surround the Ft. Lauderdale central business district, but there is little population located in the center of the Ft. Lauderdale CBD, which is devoted to employment, as shown in Map 6. . We also create a multiplicative variable that indicates the degree to which a TAZ simultaneously exhibits high retail, high service, and high mix of households (with and without children). We hypothesize that all of these attributes are important for generating transit trips, but if they all exist simultaneously, there is an added boost to transit patronage compared to the individual contribution of each of these variables in generating transit trips. We present the construction of this composite variable below.

Map 1. Distribution of 2005 Population in Broward County by TAZ

Source: SERPM Data (2005)

Map 2. Distribution of 2005 Population Density (Persons per Acre) in Broward County

Source: SERPM Data (2005)

Map 3. Distribution of 2000 Median Household Income in Broward County

Source: Census Transportation Planning Package (2000) 19

Map 4. Distribution of 2005 Percentage Households with Children in Broward County

Source: SERPM Data (2005) 20

Map 5. Distribution of 2005 Average Household Auto Ownership in Broward County

Source: SERPM Data (2005)

Attraction Variables On the destination end of the trip, employment, employment density, walkability and parking charges typically are used to indicate the attractiveness of the zone for transit trips. These four variables are available through SERPM05. Map 6 shows the distribution of employment, Map 7 the distribution of employment density, Map 8 the distribution of employment density with the Broward County Transit route map superimposed on it, and Map 9 a distribution of zones charging for work trip parking. In addition, we create a multiplicative variable to denote the degree to which these various attributes for attracting transit trips are present simultaneously. As we do on the origin end of the transit trip, we hypothesize that the simultaneous existence of all of the desirable destination variables in a destination TAZ would give a boost to attracting transit patronage to the zone compared to the sum of the transit attractiveness of each of the variables. We describe the construction of the interaction variable below. Map 3 indicates that employment is widely dispersed in Broward County, although not as widely dispersed as population. Higher density employment occurs on an east‐west axis in the northern part of the county running from Coral Springs to the ocean. There also are zones with sizable employment in the west central and southern parts of the county. Map 4 indicates the distribution of employment density. Zones with higher employment density also are widely scattered, though there is a greater tendency for them to be located in the eastern part of the county. The highest density zones are in the Ft. Lauderdale CBD, through which runs the north‐south Florida East Coast Railroad. North and south of the Ft. Lauderdale CBD are scattered along the railroad other zones of higher density employment, though much larger employment concentrations are found to the west. Map 8 indicates that the grid‐ like route structure of Broward County Transit serves almost all of the employment in Broward County. Long term parking fees (Map 9) tend to exist in the eastern areas of the county. Zones with the highest fees are those in the Ft. Lauderdale CBD. To the south the airport shows up as a zone with parking fees. High activity beach areas east of Fr. Lauderdale and farther to the north also show long term parking fees. To the south of Ft. Lauderdale, the airport, downtown Hollywood, and barrier islands charge for parking. In Coral Springs to the northwest, zones containing the small downtown and very large regional medical facilities also are indicated as having long term parking fees. We also examined how walkable are each of the TAZs in Broward County. FDOT in the SERPM05 model defines the walkability of the various TAZs in southeast Florida. The variable that they created is a count variable that can be 0, 1, 2, or 3 for a given TAZ. It is 0 if a TAZ has no sidewalks, no marked crossings of streets, and is assigned a rural area type. It is 1 if fewer than 10% of the streets have sidewalks, fewer than 10% of the intersections have marked crossings, and the area type is designated OBD. It is 2 if between 10 and 90 percent of the streets have sidewalks, between 10 and 90 percent of the intersections have marked crossings, and the area type is fringe or residential. It is a 3 if more than 90 percent of the streets have sidewalks, more than 90 percent of the intersections have marked crossings, and the area type is CBD. Map 10 shows the distribution of the walkability variable for Broward County and shows that in general the Ft. Lauderdale CBD, the Hollywood downtown, coastal zones and a scattering of other zones are defined as walkable (Corradino 2008, Table B32, p. B27).

Map 6. Distribution of 2005 Employment in Broward County by TAZ

Source: SERPM Data (2005) 23

Map 7. Distribution of 2005 Employment Density (Jobs per Acre) in Broward County

Source: SERPM Data (2005)

Map 8. Broward County Transit Routes Superimposed on a Map of Broward County Employment Density

Source: Broward County Transit

Map 9. Distribution of 2005 Long Term (8 hours) Parking Fees in Broward County, in Cents

Source: SERPM Data (2005)

Transit Friction Variables One hypothesis that we are testing is that the number of transit trips from one TAZ to another is influenced not only by land use characteristics of the originating and destination TAZs, but also by the quality of transit service linking the two TAZs. The literature review suggests that we may measure quality of transit service between an origin and a destination by both transit fares and transit travel time between the two points. The higher the fare or the longer the transit travel time between two TAZs, the greater the transit friction between the two TAZs, and presumably larger friction leads to fewer transit trips. We use transit travel time as the measure of transit friction in this study. This is because Broward County Transit charges a flat fare that does not vary from one pair of TAZs to another, making it impossible to measure the influence of fares on patronage. Whatever results we obtain are understood to prevail with the fare structure that was in existence in 2003. Transit travel time, on the other hand, varies widely from one pair of TAZs to another, and it is possible to test whether such variance affects the magnitude of transit travel. Thus, we focus upon transit travel time as the variable denoting quality of transit service. We use SERPM05 to obtain estimates of the time that it would take for a passenger to travel between each pair of zones in Broward County using transit. Dave Schmitt of AECOM Consult, acting on behalf of FDOT District 4, provided us with peak and base transit skims for the 2005 scenario. Each skim is a square matrix whose column and row headings are the TAZs in Dade, Broward, and Palm Beach Counties. Thus, each cell indicates a component of transit travel time between a pair of TAZs. The skims consist of several tables corresponding to each stage of the transit journey. Table 1 of the skims is the time spent walking to and from the bus at each end of the trip. Table 4 of the skims is time spent riding the local bus. Table 11 of the skims is time spent waiting for the initial bus. Table 12 of the skims is time spent waiting for all of the connecting buses if transfers are required. Perceived door‐to‐door transit travel time between two TAZs is the addition of the perceived time required by a transit traveler to negotiate each of the several stages required to complete a trip from beginning to end. A traveler has to walk from the origin of their trip to a bus stop. The traveler then has to wait for the bus. After riding the bus, the traveler may have to transfer to a second bus, which requires additional waiting time. There may be more than one transfer. Finally, the traveler has to walk to the final destination. Typically, transit users perceive the time spent waiting and walking outside of the vehicle to be more onerous than time spent on board the vehicle. Practitioners in southeast Florida weight out of vehicle time at 2.25 times greater than in‐vehicle travel time (Corradino 2008, Tables 2‐22 and 2‐23, p. 2‐39). For this research, we obtain perceived door‐to‐door travel time by weighting walking and waiting time at 2.25 compared to the time spent in riding the transit vehicle, and then adding the weighted times to in‐vehicle time.1 The transit travel time between one TAZ and another can differ by time of day. Transportation demand models that estimate transit travel times within metropolitan areas often provide estimates for the peak period and the off‐peak. The appropriate time to use depends upon the dependent variable. In

1 We first determined (by consulting with FDOT District 4 modelers) which TAZs in the 2005 scenario belonged to Broward County. These were 1751 through 2671. We then transferred those parts of the tables representing the Broward County TAZs to Excel2007. This action resulted in a set of 921 rows by 921 column tables for Broward County.

this case, the dependent variable is comprised of work trips. Most work trips occur during the peak, so we used skims from the peak travel time model to extract perceived transit travel times. There is one further modification that we made to the matrix. Some cells had zeros in them. The zeros indicate that it is not possible to use transit between the two TAZs represented by the cell. The reason is that one or both of the TAZ’s are too far from the nearest transit stop to make walking possible. Actually, zero is a poor choice to place in a cell representing two unconnected TAZs, because it indicates that a person can instantaneously travel from one to the other. In reality, the travel time should be impossibly long. For such cells, we replaced the zero with a large number (99999), which indicates that it would take an impossibly long time to use transit in traveling between the two TAZs.

The TOD Variables Planners define Transit Oriented Developments (TODs) as a mixture of land uses within walking distance of a transit stop (usually defined as a quarter mile maximum walking distance) that promote transit use. A quarter mile circle around a transit stop contains 160 acres, which is the area of a TOD. To be a TOD the 160 acre area must be designed to be walkable. It also must contain a minimum of seven dwelling units per acre (14 to 21 inhabitants per acre) organized into a mixture of housing types. The area should have employment, as well, although most literature on TODs assumes that the primary role of a TOD is to originate transit trips that are destined to a CBD, and consequently says very little about employment. Calthorpe (1993), on the other hand, argues that some TODs should be designed as regional destinations, characterized by up to 11,000 regional jobs within the 160 acre site, mixed in with one to two thousand residents. Those magnitudes of jobs and population on a 160 acre site yield a population density of 12 persons per acre and an employment density of 69 persons per acre for a regionally‐oriented TOD (Calthorpe 1993; Frank and Pivo 1994; Cervero 2002). In this study we wish to determine whether there are zones in Broward County that exhibit at least some of these characteristics, which Cervero (2002) characterizes as density, diversity and design, and if so, do those characteristics stimulate transit patronage? The unit of analysis that we have to work with is the traffic analysis zone (TAZ) and data associated with it; namely, population, households, employment, and area. Our approach is to first identify TAZs in Broward County that have household and employment densities that exceed the threshold established by Calthorpe for neighborhood and urban TODs. There are six TODs that meet the Calthorpe thresholds for neighborhood TODs (Map 11). Three are in the Ft. Lauderdale CBD, one is north of the Ft. Lauderdale CBD by about four miles, one is northwest of the Ft. Lauderdale CBD by about six miles, and one is south of the Ft. Lauderdale CBD adjacent to the Hollywood CBD. There are only three that do so for urban TODs, and all three are in the Ft. Lauderdale CBD (Map 12). We next considered how walkable these TAZs are, using Map 10 as well as Google Earth as references. According to Map 10, all TAZs in the Ft. Lauderdale CBD have the highest index (3) for walkability. On the other hand, Map 10 indicates that the three TAZs that are candidates for neighborhood TOD designation and that lie outside of the Ft. Lauderdale CBD are only minimally walkable, each with an index score of 1. We then used Google Earth to see how these three TAZs appear, and we were surprised to see that they look very walkable to us. The TAZ adjacent to Hollywood, for example, is comprised of a fine grid of mostly narrow streets lined on both sides with sidewalks. The arterial roads defining the edges of the TAZ are lined with sidewalks and have many signalized pedestrian crossings. The remaining two non‐CBD candidate TAZs for neighborhood TOD designation have pedestrian connections almost but not quite as good as the TAZ near Hollywood. We therefore considered all of the candidate TAZs to be walkable.

Based on population and employment density as well as walkability we defined as neighborhood TODs all of the six of the candidate TAZs for neighborhood TOD designation, and we defined as urban TODs all three of the candidate TAZs for urban TOD designation. We then rated each of these TAZs for the degree to which they possess mixed uses, using an index ranging from 0 to 1. The value of the index is based on an entropy index method used by several researchers of transit oriented development to indicate mixed land uses. Frank and Pivo (1994), for example, measured the square footage of various zones devoted to different land uses. They used the following formula to develop a measure ranging from 0 (there is only one land use in the zone) to 1 (there is an equal share of square footage in the zone devoted to all land use categories under consideration):

Level of land use mix (entropy value) = ‐ [single family * log10 (single family)] +[multifamily*log10

(multifamily)] + [retail and services * log10 (retail and services)] + [office * log 10 (office)] + [entertainment * log10 (entertainment)] + [institutional * log10

(institutional)] + [industrial/manufacturing * log10 (industrial/manufacturing)]/log10(number of categories). Equation 1

Because we have household and employment categories for every zone rather than land areas divided into categories that Frank and Pivo had, we modified their entropy index to use the variables that we have available to measure mix of uses. We considered that a desirable mix of uses would be equal numbers of households with and without children and equal proportions of retail and non‐retail employment. Our modified Frank and Pivo entropy index is as follows:

TOD = ‐ [population in household without children * log10 (population in household without children)] +[ population in household with children *log 10 (population in household with children)] + [retail * log (retail)] + [services * log (services)] + [other employment * log 10 10 10 (other employment)]/ log10(number of categories). Equation 2

Map 11 shows our final designation of neighborhood TODs. There are six of them. Three are in the Ft. Lauderdale CBD with a moderate mix of uses, indicated by indexes of .241, .287, and .369. The other three neighborhood TODs have comparable degrees of mixed uses as the ones in the Ft. Lauderdale CBD. All three of the non‐CBD neighborhood TODS are characterized by mixes of small one story homes on small lots and small to medium sized apartments ranging from one to four stories. Mixed in with the residential dwellings in all three zones are storage lockers, light industrial buildings, strip retail establishments, tennis courts, and baseball diamonds. Map 12 shows our final designation of urban TODs. There are three of them, and they all are located in the Ft. Lauderdale CBD. Their indexes of mixed use are .698, .764, and .878, indicating a high degree of mixed use for all three urban TODs.

Map 10. Walkability of Broward County TAZs per SERPM05 Definition

Source: SERPM Data (2005)

Map 11. Distribution of Neighborhood TODs (NTOD)

Source: SERPM Data (2005)

Map 12: Distribution of Urban TOD Attraction variable (UTODj) in Ft. Lauderdale CBD

Source: SERPM Data (2005)

The Specified Model Putting these considerations together, we arrive at the model specification shown in Equation 3:

ij exp( += 10 * i + 2 POPbPOPbbT _* DENSITYi + 3 * i + bNTODb 4 * MEDHHINCi +

5 _* i + bWLKORIb 6 %* HHWCHILDi + b7 * AVGHHAUTOi + 8 * j + 9 EMPbEMPb _* DENSITY j

+ 10 j + 11 *_* j + 12 _* j + 13 TTIMbWLKDESbUTODbLTPARKb _* PERCEIVEDij ),

Equation 3 where, Tij represents the number of transit work trips originating in zone i and terminating in zone j; TTIM_PERCEIVEDij= perceived door‐to‐door transit travel time between zone i and zone j; POPi = population in originating zone i; POP_DENSITYi=population density in originating zone i ; NTODi=Neighborhood TOD variable for originating zone i ; MEDHHINCi=Median household Income in originating zone i ORI_WLKi=Walkability index in the originating zone i %HHWCHILDi= Percentage of household with children in originating zone i AVGHHAUTOi=Average household auto ownership in originating zone i EMPj=employment in destination zone j ; EMP_DENSITYj=employment density in destination zone j ; PARK_LTj=long term parking fee in destination zone j ; UTODj=Urban TOD variable for destination zone j ; DES_WLKj=Walkability index in the destination zone j ; and, the b’s are parameters to be estimated.

In words Equation 3 tells us that work transit trips originating in zone i and destined to zone j are influenced by the population, population density, TOD characteristics of the originating zone, median household income, walkability, children in households and auto ownership of the originating zone as well as by the employment, employment density, long term parking fees, TOD characteristics, and walkability of the destination zone, and by the perceived transit time that one must endure to travel between the two zones. The question is, what is the relative importance of each of these variables for determining transit ridership? By estimating Equation 3 with population, employment, and TOD variables for each of Broward County’s TAZs, as well as with transit travel time between each pair of zones, we can answer that question. We will obtain values for each of the parameters in Equation 3, and when we have those parameters, we may test our hypotheses.

Data The data for estimating the model potentially could contain 848,241 observations, if we include as an observation travel between any pair of traffic analysis zones. Many of these observations would be invalid, however, and we removed them from the data. The Census Bureau suppressed ridership data from TAZs where the population is too small to protect confidentiality of the respondents. We estimated that the threshold population where suppression occurred is about 500 people, and we removed from the data set all observation containing traffic analysis zones with fewer than 500 people. Also, many of the observations had no transit service connecting them. Where there was no transit service connecting 33

one zone with another, we set the transit travel time value (Ttime) at a very large number of minutes, 99,999. Doing so, however, created an average transit travel time between one zone and another of 11,000 minutes. This value is absurdly large. We therefore removed all observations where transit service did not exist by eliminating observations where Ttime was more than 90,00. Upon removing observations affected by confidentiality procedures and absence of transit service, we obtained a data set for estimating the model. The modified data set is shown in Table 1. The modifications removed somewhat less than 300,000 observations, leaving about 550,000 observations for estimating Equation 3. Table 1 shows that the minimum population now is 505 people, and the maximum value of perceived transit travel time (Ttime) between an origin and a destination is 353 minutes, while the average is 139 minutes.

Table 1. Summary of Data Set With Some Observations Removed Variable Observations Mean Std. Dev. Minimum Maximum

Ridership 550,441 0.02 0.49 0.00 35.01 Population 550,441 2,598.20 1,752.30 505.00 9,799.00 Popden 550,441 9.80 5.86 0.88 70.84 Employment 550,441 827.47 962.63 0.00 7,983.00 Empden 550,441 5.88 12.85 0.00 186.20 Ttime 550,441 139.49 49.52 16.65 353.19 Park2 550,441 14.09 54.67 0.00 431.00 Median Household Income 550,441 45,286.92 24,377.78 0.00 200,000.00 Neighborhood TOD 550,441 0.00 0.03 0.00 0.37 Urban TOD 550,441 0.00 0.05 0.00 0.88 Destination Walk Index 550,441 1.378135 0.8469453 0 3 Origin Walk Index 550,441 1.46 0.75 0.00 3.00 % Household with Children 550,441 40.17 17.78 1.15 78.36 Average Household Autoownership 550,441 1.64 0.35 0.58 2.60

Co‐linearity Among Independent Variables Table 2 presents pair‐wise correlations between each pair of independent variables shown in Table 1. If two independent variables are highly correlated with each other, estimations of models containing them may result in erroneous estimated parameters for one or both of those variables. Generally correlations above an absolute value of about .7 indicate potential problems. Table 2 shows that there are no concerns with co‐linearity in this data set.

Table 2. Pair‐Wise Correlations between Independent Variables

Results We used negative binomial regression to estimate the coefficients in Equation 3 with data summarized in Table 1. Table 3 presents the results. Results indicate three pieces of information about the model. They indicate the usefulness of the model in predict ing work trip ridership between a pair of zones, they indicate the relative importance of each explanatory variable in making the prediction, and they also indicate whether negative binomial regression is an appropriate method for estimating the model. We address the last questions first. As indic ated earlier, the nature of the explanat ory variable dictates that we use a count model. The usual count model is Poisson Regression, but that choice is valid only when the dependent variable (in our case, transit work trips between two zones) has a variance that is equal to its mean. Looking at Table 1, we see that this variable, called, “Ridership,” has a variance about ten times its mean.2 In such a case, negative binomial regression is called for. The chibar2 statistic at the bottom of Table 3 indicates the probability that Poisson regression was a better procedure to use than negative binom ial regression. That probability is zero to at least three decimal places. We appear vindicated in having used negative binomial regression. The chi‐square statistic tests the hypothesis that all of the coefficients for the explanatory variables are in reality zero; that is, the explanatory variables have no explanatory power. Table 3 indicates that the probability of that being the case is zero to at least four decimal places. We reject that hypothesis and conclude that at least some of the explanatory variables are useful for predicting work trips between two zones. However, the low value of the pseudo r‐squared statistic indicates that the explanatory power of the model is not high. Turning to each of the explanatory variables in Table 3, we see the estimated coefficient for each one, and we see the 95 percent confidence interval that we can place in that coefficient. If the 95 percent confidence interval includes zero, we do not have much faith in that variable for explaining transit work trip ridership. However, we might conclude that the variable tends to have an effect if the 95 percent confidence interval just barely includes zero. We see in Table 3 that the confidence interval

2 The standard deviation of ridership is .49, so the variance is .24 (standard deviation squared). The variance is roughly ten times greater than the mean. 35

for both long term parking (Park2) and the destination TOD variable (UTODj) just barely includes zero. Both variables tend to increase transit ridership the larger they become, indicating that zones with more TOD‐like qualities or higher long term parking rates will stimulate more transit patronage. The effect is weak, however.

Table 3. Estimation of Model

Negative binomial regression No. of obs = 550,441 LR chi2(8) = 991.94 Dispersion = mean Prob > chi2 = 0.0000 Log likelihood =‐13289.766 Pseudo R2 = 0.036

Dependent variable is Ridership

Standard Elasticities Independent Variables Coefficient Error Z Statistic P>|z| 95% Confidence Interval at means

Pop 0.00021 0.00003 6.59 0.0000 0.00015 0.00027 0.54 Pop_density 0.02822 0.01299 2.17 0.0300 0.00276 0.05368 0.28 Emp 0.00065 0.00006 10.34 0.0000 0.00053 0.00078 0.54 Emp_density ‐0.01009 0.00533 ‐1.89 0.0590 ‐0.02055 0.00036 ‐0.06 Ttime ‐0.01989 0.00104 ‐19.16 0.0000 ‐0.02193 ‐0.01786 ‐2.77 Park2 0.00146 0.00098 1.50 0.1350 ‐0.00045 0.00337 0.02 MedHHIncome ‐0.00002 0.00000 ‐4.47 0.0000 ‐0.00002 ‐0.00001 ‐0.73 NTOD ‐0.89537 1.22323 ‐0.73 0.4640 ‐3.29285 1.50211 0.00 UTOD 1.26735 1.08121 1.17 0.2410 ‐0.85179 3.38649 0.00 Des_Walk 0.05244 0.05599 0.94 0.3490 ‐0.05731 0.16219 0.07 Ori_ Walk ‐0.03208 0.06976 ‐0.46 0.6460 ‐0.16880 0.10464 ‐0.05 % HH with Children 0.03491 0.00398 8.78 0.0000 0.02712 0.04270 1.40 Avg. HH Auto ‐1.49458 0.23949 ‐6.24 0.0000 ‐1.96398 ‐1.02519 ‐2.45 Constant ‐1.82443 0.35254 ‐5.18 0.0000 ‐2.51539 ‐1.13346

/lnalpha | 6.79355 0.03539 6.72419 6.86292 alpha | 892.07570 31.57108 832.2949 956.1505

Likelihood‐ratio test of alpha=0: chibar2(01) = 9.8e+04 Prob>=chibar2 = 0.000

The elasticity of the variable at means also is important in establishing which variables are important in affecting transit patronage. In an exponential model, such as Equation 3, the elasticity of demand with respect to each explanatory variable is the value of the variable multiplied by its coefficient (Manheim 1979, p. 125, Table 4.1). The elasticity at means indicates how many percentage points transit ridership will shift when there is a one percent shift in the explanatory variable away from its mean value shown in Table 1. For example, the mean value of perceived transit travel time from one zone to another in Broward County is 139 minutes, as shown in Table 1. If we reduce that travel time by one percent, we can expect an increase of transit ridership between those two zones by 2.77 percent. In 36

Table 3 we see seven explanatory variables with relatively high elasticities. Five of these are variables describing conditions in the zone where the trip originates: population , population density, median household income, percentage household with children and average household auto ownership. One describes conditions at the destination zone of the trip: employment. One variable (which turns out to be the most important variable) describes the difficulty in making the trip between the origin and destination zones: the perceived transit time between the two zones. The transit travel time has a particularly high elasticity. Not as important are employment density (in fact, it has the wrong sign), long term parking, the TOD variable at the production end of the trip (which also has the wrong sign), the TOD variable at the destination end of the trip, the walkability at the origin end (which also has the wrong sign), and the walkability at the destination end. As we indicated in our definition of neighborhood TODs earlier in this chapter, we do not have confidence in the walkability variables. We discuss the implications of the remainder of the results below.

Discussion Our confidence in the results shown in Table 3 is strengthened by the corroboration they give to the profile of Broward County Transit passengers revealed in the passenger survey summarized in the introduction to this chapter. As does the survey, Table 3 presents a picture of a highly transit‐ dependent ridership. Population magnitude does the most to explain the magnitude of passengers originating in a TAZ, but presence of household children, the size of household income, and household auto ownership all have significant and important effects on modifying the magnitude of patronage predicted by population. For every percent that the proportion of households with children increases in an originating zone, transit use increases by 1.4%. For every percent that an originating zone’s auto ownership rises, transit use declines by 2.45%. For every percent that an originating zone’s median income increases, transit use declines by 0.73%. Table 2 shows that auto ownership and income are not highly correlated (.58), so we interpret the effects of income and auto ownership on transit ridership as largely independent of each other. Households with children create more demand for travel and reduce auto availability for any given level of auto ownership. Also, for every percentage that population density in the originating zone is increased above the mean, ridership increases by .28%. The destination variables also support the profile of a transit dependent ridership. For every percent that employment in a destination TAZ is raised above the mean, ridership attracted to that zone is predicted to increase by .54%. The fact that employment density at the destinat ion zone depresses transit ridership can be interpreted as an effect of a largely transit dependent population. We suspect that higher density employment, such as that found in the Ft. Lauderdale CBD, is comprised of more affluent workers who tend to drive to their work. Transit dependent workers more likely are going to scattered employment sites that have lower density. This argument might also explain the small (though measurable) effect that long term parking rates have on boosting transit ridership to a destination zone, as well as the small positive effect that urban TOD designation has on boosting transit ridership. The variable that has the greatest effect in determining transit ridership is TTIM, which measures how much time transit riders perceive that it takes them to travel from an originating TAZ to a destination TAZ. The average TTIM in the data set is 139 minutes. For every percent that TTIM is reduced from the mean, the model predicts that transit ridership will increase by 2.77%. TTIM thus has a tremendous impact on transit ridership. The TOD variables have much less of an influence on ridership, but the important TTIM variable suggests that TOD developments could be important to transit development, particularly in the long term. The Urban TOD variable contributes marginally to higher transit ridership, while the Neighborhood TOD variable contributes marginally to depressing ridership compared to what population and population density of zones predict on their own. These small effects may stem from the fact that 37

Broward County contains no development that has all of the characteristics of neighborhood and urban TODs promulgated in the literature. We did identify several zones that are walkable, have mixes of uses, and have both high population and employment density, but these zones do not look like the TODs one sees in the literature. They either are in the Ft. Lauderdale downtown, or they are what appear to us to be down market neighborhoods. We turn now to the TTIM variable to understand what it says about desirable policies for boosting transit ridership. TTIM is comprised of several components including the time to walk from home to the bus stop, the time required to walk from the destination bus stop to the final destination, the time spent waiting for the bus, and the time spent transferring, in addition to the time spent riding the bus. Anything that policy makers can do to reduce TTIM will greatly increase transit ridership. Possible policies are: • Encouraging development into TOD configurations to reduce walk times on the originating and destination ends of transit trips; • Increasing overall regional densities, so as to shorten the distance between origins and destinations, thus reducing the length of transit trips; • Shortening circuitous transit travel, by restructuring routes from CBD‐radial to grid configurations (see the case study in the next chapter); • Reducing headways to reduce wait and transfer times; • Speeding up service on heavily‐traveled routes by increasing door capacity (so that passengers will board and alight faster) and changing fare collection systems so that passengers can board at all doors • Speeding up service by segregating transit vehicles from private vehicular traffic. In general, we can group these policies into two categories: those that change the transit system to better link origins and destinations, and those that group activity around the origins and destinations (which may be termed TOD policies). Examples of the first set of policies are restructuring of routes into grids or spider webs with relatively frequent service. Such networks would have more direct and faster connections between the dispersed population and employment clusters characteristic of Florida metropolitan areas than more conventional routing. The model results show that the more direct routes would boost ridership in several important ways. More direct routes mean less territory to cover and fewer minutes on the bus. More direct routes also mean faster buses, removing still more minutes from the time on the bus. More frequent service means less time waiting for the first bus and less time waiting at the transfer point. A drawback could be longer walks to the bus, but this is where the second set of policies comes in: grouping population and employment around the stops into TODs to shorten walking distances. The results suggest that the TOD policies are less important than the first set, but the results also suggest that TOD policies can have a significant impact if TODs are: • designed to have a large percentage of their units available to families with children; • designed to appeal to households with one or no cars available; • designed to have a significant number of units available to households of limited means; and, • designed to have short walks along attractive and safe paths to transit stops that are well‐ connected to jobs throughout the region. In the long term, if policy succeeds in directing most population and employment growth into TODs, regions will be more compact, and distances between origins and destinations will be shorter. According to the model, shorter distances would have a large effect on increasing transit patronage.

Chapter 5. The Case Study

Introduction Planners and policy makers consciously designed Broward County Transit to serve not only dispersed population, but also dispersed employment. This chapter examines differences (and similarities) between Broward County Transit and a more traditional transit system configuration focused on an historic central business district. The comparison system is Ft. Worth Transportation Authority, known as The T, serving Ft. Worth, Texas. The objective of the comparison is to gain insight into the efficacy of the two approaches to transit system design. We first examine the settings for Broward County Transit and The T before considering the historic evolution of transit in each of the two settings. We then examine performance of transit in each of the two areas, using three indicators. These are passenger miles per capita, passenger miles per vehicle mile, and operating expenses per passenger mile. Sources for this section include the Miami and Dallas regional case studies in Brown and Thompson (2009), interviews conducted with Broward County Transit (Stoleson 2008) and The T(Amos 2008) officials in 2008, data made available by Broward County Transit (Broward County Transit 2008a and 2008b) and The T (Fort Worth Transportation Authority 2008) subsequent to the interviews, and data from the National Transit Data Base, retrieved through FTIS.

The Settings Both transit systems serve counties that are the second largest in their respective metropolitan areas. Broward County, served by Broward County Transit, lies immediately north of Dade County, in which lies the City of Miami. Tarrant County, served by The T, lies immediately west of Dallas County, home to the City of Dallas. As of the most recent year for our data (2006), both Broward and Tarrant counties had similarly sized populations that grew at nearly the same high rate over the preceding 22‐ year period (Figure 1). They differ in one important way, however. Tarrant County contains a large, traditional central business district (downtown Ft. Worth) that emerged in the late nineteenth century. An electric streetcar system and an electric interurban line running between Ft. Worth and Dallas, evolved in symbiosis with downtown Ft. Worth. Broward County has no traditional central business district of the magnitude of Ft. Worth. It does have small downtowns (the largest of which is Ft. Lauderdale) that grew around stations on the Florida East Coast Railroad that linked Miami to Jacksonville in the 1890s,running near the coast, but well into the twentieth century Miami remained as the only traditional central business district of the region.

Figure 1. Broward and Tarrant Counties Have Similar Populations and Growth Rates, 1984‐2006

Broward County Transit and The T are the second largest transit systems in their respective metropolitan areas. They both are the primary transit providers in the counties that they serve, and they connect with transit systems in other counties. Broward County Transit buses enter northern Dade County where they connect with Miami Dade Transit Authority buses (Map 13). They also connect with Palm Tran buses in southern Palm Beach County. About half of BCT bus routes also cross tracks of Tri‐ Rail. Tri‐Rail, operated by the South Florida Transportation Authority, is a suburban passenger service using tracks on the old Seaboard Air Line Railroad, five or six miles inland of the Florida East Coast Railroad. Tri‐Rail trains connect Miami to West Palm Beach, stopping at seven stations within Broward County. Tri‐Rail currently runs trains hourly in both directions during the week day. These are supplemented by additional trains during peak periods. Service is every two hours on weekends. During early 2008 Tri‐Rail boarded about 14,000 passengers per day, with a little more than a third of those boarding at Broward County stations. While the Broward County train boardings are substantial, there is virtually no transfer activity between Broward County Transit buses and Tri‐Rail trains. Tri‐Rail passengers wishing to board BCT buses would pay 50 cents to do so, less than half the normal bus fare of $1.25 (as of October 2007); BCT passengers wishing to transfer to Tri‐Rail trains would pay the full Tri‐ Rail fare (which is zoned depending upon distance traveled) but would get to board BCT for free. 40

Transfers between BCT buses are free. Because of the absence of transfer activity, this study focuses on Broward County buses. The T is more insulated from other bus systems in its metropolitan area (Map 14), but it is somewhat better integrated with commuter rail service, known as Trinity Railway Express (TRE). TRE began limited service from Dallas Union Station (where it connects with Dallas Area Rapid Transit light rail trains) to a station south of the Dallas‐Ft. Worth airport in 1996; in2001 TRE service was extended westward into the Ft. Worth central business district, where it connects with The T buses in a large multi‐modal transit terminal. TRE trains now run roughly on an hourly headway Monday through Saturday, with more service during peak times and was attracting roughly 9,000 passengers per day in March 2008, rising to over 12,000 passenger per day in July 2008 as gas prices rose. The T and Dallas Area Rapid Transit share in ownership of TRE, and there are free transfers between The T buses and TRE trains. There is some amount of transfer activity between The T buses and TRE trains, but not much. Trinity Rail Express serves few trips within Tarrant County, so this study focuses on The T.

Map 13. Both Transit Systems Fit into Their Regional Contexts Similarly: BCT Adjacent to Miami‐Dade Transit and Tri‐Rail

Map 14. The T Adjacent to Dallas Area Rapid Transit and Trinity Rail Express

Transit Development in the Broward County Prior to public involvement in the provision of transit service in Broward County, two private operators offered service in the county. One ran several routes focused on the Ft. Lauderdale downtown; the other ran several routes focused on the Hollywood downtown. Our agency contact person characterized both systems as having skeletal, circuitous routes with hourly headways. He called them, “spaghetti networks,” that attempted to go “where the riders are;” that is, routes wandered through neighborhoods where riders lived. On the other end, routes served the beaches and were designed to carry domestics who worked in condos. Our agency contact person further characterized the systems as “unreliable and inefficient.” Broward County Transit was organized to take over the two private systems in the mid‐1970s. Originally it was a division in the Office of Transportation but later was moved to Community Services, reflecting a vision of transit as being a social service. Sometime later BCT was moved back to the Office of Transportation, where it remains today. At first BCT expanded upon the route structure that already was in place. One improvement was the creation of an overlay of express bus routes that ran from various parts of the county to downtown Ft. Lauderdale and to Miami International Airport. Our agency contact person, who joined the system as that time as a bus driver, said that the system carried few riders. Even the modest ridership that the express lines initially attracted dwindled from year to year. Low ridership on all of its services prompted BCT management to reflect upon how it might do things differently. Service to Miami International Airport was suspended when Eastern Airlines shut down. The director of the system at the time, Houston Miller, determined that the system needed to be gridded, but that it would be changed over incrementally. The process began in 1980 with Operation Changeover. Base headways were reduced from 60 to 30 minutes on what were termed, “mainline routes.” Headways were shortened due to recognition that a grid would require many passengers to transfer to complete their trips. Hourly headways were felt to be too long for passengers to wait at transfer points. The gridding of the system happened over 10 to 15 years, beginning in 1980. For many years some routes still had deviations to serve destinations like condo complexes. All express routes were gone by the late 1980s. Our agency contact person said that BCT formed its routing decisions with studies by CUTR and NTI that compared BCT to other transit agencies. BCT also used common sense. Broward County has a grid pattern for its arterial roads, so the move to grid transit network seemed logical. Our agency contact person also reported that BCT received positive feedback from its early route straightening that gave it confidence to continue with them. After BCT made route straightening, they would see ridership increase. Impacts of changes appeared right away. He also pointed to population growth as a factor influencing steady increase in ridership from 13 million trips in 1984 to around 40 million today. The heaviest service today operates on U.S. 441, a high‐speed, heavily trafficked multi‐lane arterial highway that runs through the middle of the built up part of the county in a north‐south orientation. Two routes operate on this road from one end of the county to the other. Route 18 provides local service on 15 minute headways. “The Breeze” provides limited stop service, stopping every mile or so to interchange passengers with buses on heavy east‐west routes. Loads are heavy, and BCT uses articulated buses to handle them. The U.S. 441 routes serve no downtown but do serve numerous strip malls, regular malls, and big box stores. Apartment complexes generally are only one to two blocks away on either side. On the south end the U.S. 441 routes connect with Metro Dade buses. The Breeze picks up 10 to 15 passengers per trip at this point, some of whom are transferees from MDT buses. 44

When BCT eliminated route deviations by pulling buses out of neighborhoods and putting them on arterial roads, it met some political resistance from users who did not want to walk farther to reach a bus stop. The political solution to this problem was the designation of some transit operating funds to support community circulators, small buses that wander through neighborhoods, taking residents to nearby destinations and to stops on the mainline BCT routes. There are many local governments in Broward County, and evidently the local governments determine how to run the circulators in their jurisdictions. Our agency contact person states that almost all of the patronage growth for BCT has been on the mainlines on the arterial roads. The left panel of Map 15 shows BCT’s route structure in 2006 in relation to employment density in the county. The dispersal of employment sites throughout the county is readily apparent. All employment sites have gridded transit routes passing them. Residents living in most parts of the county can reach employment wherever it is located by using buses running in straight lines along arterial roads.

Transit Development in Tarrant County The dominance of the Ft. Worth central business district over a long period of time and differences in funding mechanisms for transit between Texas and Florida have influenced The T to evolve very differently than Broward County Transit. Streetcar lines and the Ft. Worth CBD grew hand in hand during the early twentieth century, with streetcars extending out to suburbs from the CBD in the classic radial pattern. Through the transition from streetcar to bus and to the present day, this pattern of organizing transit routes has not changed (though it has been added to), even though employment and residents have decentralized ever more throughout the region since autos ownership began rising rapidly after World War I. Finance also affects the pattern of transit development in Florida and Texas. As a county department, BCT receives subsidies from the county in sufficient magnitude to allow it to serve all of those parts of the county that are urbanized. The Florida Department of Transportation also provides some transit operating support through its gas tax. Financing is more difficult for The T and restricts the territory that it can serve. There is no state operating support for transit in Texas, where local sales tax revenues provide the primary source of subsidy for transit operating deficits. Texas transit systems must appeal to individual communities for sales tax revenues, but Texas law imposes a sales tax cap on communities of 8.25 percent. Many communities already were at the limit before transit agencies approached them for funding. If a community chooses not to provide sales tax funding for transit, it gets no service. Because The T historically served the city of Ft. Worth and was a city department before becoming an authority in 1983, it receives tax support from the city (population today of about 700,000). At the time it became an authority, it received a dedicated ¼ cent sales tax from the city to support transit. The T also receives support from the city of Richland (population 7,000). The city of Arlington (population 300,000), in contrast, refuses to provide sales tax funding to either The T or to Dallas Area Rapid Transit; Arlington thus receives no transit service. Unfortunately, some of the heaviest employment concentrations and most rapid employment growth in Tarrant County are in Arlington. Thus, The T does not serve significant parts of the urbanized areas in Tarrant County. The T’s route structure today is largely radial in nature. The two most heavily‐traveled routes operate in straight lines on arterial roads from one side of the city to the other, one north‐south and the other east‐west. These operate every 15 minutes during weekdays. The two routes intersect in the CBD at the Intermodal Transportation Center, where Trinity Rail Express also stops. Schedules are coordinated so that passengers may transfer in both directions between the two routes and with Trinity Rail trains. The outer ends of these routes serve transit centers from which fan community circulator routes. Again, connections are coordinated. Other routes wind through neighborhoods not served by 45

the first two routes on their way to the CBD. Some operate every 30 minutes; others hourly. There is a major route that was implemented relatively recently and operates on arterial roads as it connects transit centers on the east, south, and west ends of the city. This belt route, which operates every 30 minutes, does not serve the CBD but does serve malls. It is the third most heavily patronized of The T’s routes, and its patronage has been growing briskly. During peak hours, seven express buses operate from outer neighborhoods and transit centers to the CBD. Most express routes consist of a handful of trips in the peak direction during the peak hours. In addition to regular route services, The T operates vans during shift changes between some major employment centers (particularly in the north) and transit centers. The right panel of Map 15 shows The T’s route structure in 2006 in relationship to the distribution of employment in Tarrant County. Although Ft. Worth is a central business district, employment is widely scattered throughout the county. While radial routes of The T pass by many of the suburban centers residents in many parts of Tarrant County cannot reach the jobs without first traveling out of direction to the CBD transfer center, transferring, and then riding back out into the suburbs in another direction. There also are major job concentrations that routes of The T do not touch at all. Those in Arlington are along the eastern border of Tarrant County.

Map 15. BCT Serves Many Destinations; The T Serves One Destination Well

Comparative Transit Performance We collected operating statistics for both systems showing performance from 1984 through 2006. These are summarized in Table 4. BCT has been more generously funded than The T, and this is apparent in Table 4 in the amount of service provided, measured as revenue miles. A revenue mile is a bus running one mile in revenue service. In 1984 BCT operated slightly more than twice the revenue miles that The T operated. By 2006 BCT operated almost four times as many revenue miles as The T. Often times a system that provides much more service than another will be less productive, because it has saturated the market. This is not the case of BCT compared to The T. Service productivity measures the average number of passengers on board the bus at any given time. For much of the period BCT buses were one and a half times to twice as full as The T buses, though productivity for The T increased rapidly in 2005 and 2006, greatly narrowing the gap.3 Figure 2 visually shows the productivity trends. We suspect that the greater productivity of BCT buses arises from the wider array of destinations that they serve relatively well. As a consequence of offering four times as much service combined with the greater productivity of each mile of service, BCT penetrates the travel market in its area to a much greater extent than does The T. We denote the penetration of the travel market as riding habit, a term that the U.S. transit industry once used to this purpose. Historically the transit industry defined riding habit as revenue passengers divided by population served. The industry no longer collects the statistic of revenue passengers (that it now calls linked trips), so we define the term as revenue passenger miles divided by population served. We also define the population served as that in the county. Even if the transit system does not serve all of the county, residents that it does serve want to reach destinations throughout the county, so county population is a fair measure. On that basis, we see in Figure 3 that riding habit now is nearly five times greater in Broward County than it is in Tarrant County. We also see in Figure 4 that because of its greater productivity, BCT spends significantly less to move a passenger one mile than does The T. To gain additional insight into the relative performance of the two transit systems, we examine in Table 5 the performance of their various categories of services. At the time we collected data, BCT distinguished only two categories of service: the gridded fixed bus routes operating on arterial roads, and community bus services, circulating through neighborhoods. The top panel of Table 5 shows that the fixed routes are far more productive than are the community routes while accounting for about 15‐ fold more patronage than the community services. Moreover, our agency contact person for BCT states that all of the patronage growth for the system has been accounted for by the gridded mainline routes on arterial roads. The T operates a wider array of services. Our examination of the performance of individual routes (FWTA 2008) shows only three routes with heavy patronage. The well‐performing routes include the east‐west and north‐south routes that intersection in the CBD and the belt line that connects the east and south suburban transit centers with suburban destinations while intersecting with all routes operating to the CBD from the east, south, and west. These three routes account for just more than 50% of the patronage of the fixed route system in FY 2008. Other radial routes, crosstown routes, circulator routes, and express routes have much lower patronage. The seven express routes contributed only 2.7% of system patronage. Table 2 reflects the widely differing performance level in each category

3 The improvement in productivity for The T is not the result of more passengers riding the system, but the result of passengers riding longer distances. We asked our agency contact person why passengers were riding longer distances; we noted that express bus ridership was not increasing. Our agency contact persons did not know the reason. 47

of service by showing large differences between mean and median performance in most categories. The mean is heavily weighted by the one or two routes that do well in the crosstown and radial categories respectively, whereas the median reflects the performance of the remaining routes in each category. As in Broward County, in Tarrant County the routes that perform the best are those that operate in relatively straight lines on major arterial roads, serving a relatively large array of destinations.

Table 4. BCT and The T Bus Service, 1984‐2006

Broward County Transit (BCT) Operating County Passenger Revenue Riding Service Expense per Year Population Miles Miles Habit Productivity Pa sse nge r Mi le (2006$) 1984 1,110,862 72,755,935 6,771,663 65.50 10.74 $0.50 1985 1,132,921 84,264,996 7,437,699 74.38 11.33 $0.49 1986 1,154,494 78,991,384 8,375,628 68.42 9.43 $0.59 1987 1,180,921 61,379,078 8,875,849 51.98 6.92 $0.78 1988 1,208,428 75,028,484 8,910,748 62.09 8.42 $0.68 1989 1,233,040 67,589,568 8,973,206 54.82 7.53 $0.80 1990 1,263,301 81,992,838 8,947,336 64.90 9.16 $0.68 1991 1,296,261 81,118,030 9,120,846 62.58 8.89 $0.68 1992 1,325,375 97,622,366 9,134,271 73.66 10.69 $0.55 1993 1,372,526 96,753,748 9,111,227 70.49 10.62 $0.56 1994 1,412,641 103,822,086 9,662,692 73.50 10.74 $0.53 1995 1,447,124 111,004,429 9,767,690 76.71 11.36 $0.50 1996 1,481,333 109,542,370 9,832,227 73.95 11.14 $0.51 1997 1,522,179 110,289,977 9,801,046 72.46 11.25 $0.50 1998 1,560,649 111,568,312 10,410,633 71.49 10.72 $0.53 1999 1,594,130 114,736,758 10,598,450 71.97 10.83 $0.51 2000 1,623,018 119,986,652 12,013,192 73.93 9.99 $0.53 2001 1,670,494 137,200,475 13,245,365 82.13 10.36 $0.51 2002 1,703,998 142,999,966 14,687,845 83.92 9.74 $0.53 2003 1,728,336 153,883,282 15,392,404 89.04 10.00 $0.55 2004 1,753,000 162,009,619 15,314,924 92.42 10.58 $0.54 2005 1,777,638 162,688,826 15,760,508 91.52 10.32 $0.53 2006 1,787,636 168,100,759 16,013,518 94.04 10.50 $0.53

Fort Worth Transportation Authority (The T) Operating County Passenger Revenue Riding Service Expense per Year Population Miles Miles Habit Productivity Pa sse nge r Mi le (2006$) 1984 1,001,836 25,996,998 3,146,409 25.95 8.26 $0.62 1985 1,043,207 23,787,695 3,826,627 22.80 6.22 $0.75 1986 1,083,641 27,286,469 3,729,784 25.18 7.32 $0.72 1987 1,116,110 26,077,602 3,513,866 23.36 7.42 $0.73 1988 1,133,193 21,543,916 3,596,248 19.01 5.99 $0.84 1989 1,149,530 31,693,345 3,606,597 27.57 8.79 $0.60 1990 1,177,220 48,894,085 4,217,180 41.53 11.59 $0.39 1991 1,205,887 41,969,177 4,597,108 34.80 9.13 $0.50 1992 1,225,543 27,569,034 4,516,312 22.50 6.10 $0.82 1993 1,243,884 32,344,667 4,827,258 26.00 6.70 $0.73 1994 1,270,639 34,797,556 4,992,711 27.39 6.97 $0.69 1995 1,294,453 30,474,382 4,993,480 23.54 6.10 $0.77 1996 1,323,207 30,275,663 4,754,570 22.88 6.37 $0.73 1997 1,355,318 28,706,617 4,940,493 21.18 5.81 $0.82 1998 1,388,366 24,962,373 4,597,262 17.98 5.43 $0.92 1999 1,422,372 25,373,686 4,657,887 17.84 5.45 $1.00 2000 1,446,219 27,266,081 4,740,854 18.85 5.75 $0.96 2001 1,488,780 30,617,583 4,868,114 20.57 6.29 $1.00 2002 1,525,317 27,632,150 4,750,862 18.12 5.82 $1.17 2003 1,557,128 24,048,649 3,923,945 15.44 6.13 $1.14 2004 1,587,019 21,537,919 3,879,328 13.57 5.55 $1.18 2005 1,620,479 29,106,436 4,459,345 17.96 6.53 $0.89 2006 1,671,295 31,615,080 4,063,813 18.92 7.78 $0.85 Sources: FTIS (2008), U.S. Census Bureau (2008)

Comparative Performance

Figure 2. Productivity (Passenger Miles per Bus Mile, 1984 ‐ 2006)

Figure 3. Riding Habit (Passenger Miles per Capita, 1984 ‐ 2006)

Figure 4. Efficiency (Cost per Passenger Mile, 1984 ‐ 2006)

Table 5. Transit Performance by Service Type, BCT and The T

Broward County Transit (BCT) Service Type Monthly Monthly Boardings per Revenue Boardings Revenue Hour Hours Median Average Route Fixed-Route Bus 3,209,681 87,317 36.76 31.72 Community Bus 214,085 21,183 10.11 8.84

Fort Worth Transportation Authority (The T) Service Type Monthly Monthly Boardings per Revenue Boardings Revenue Hour Hours Median Average Route Radial Routes 355,389 20,036 17.74 13.72 Crosstown Routes 67,247 4,562 14.74 10.61 Express Routes 11,372 1,023 11.12 12.94 Feeder Routes 59,313 4,657 12.74 8.54 Circulator Routes 15,798 675 23.39 7.73

All CBD-serving Routes 366,360 21,311 17.19 13.09 All Non-CBD Routes 142,759 9,642 14.81 11.78

All Fixed-Route Bus 509,119 30,953 16.45 12.81

Sources: BCT (2008), FWTA (2008). Note: BCT and The T statistics are for January 2008.

Conclusions According to much of the literature, Tarrant County offers a better built environment to support greater transit demand than does Broward County. Tarrant County possesses a traditional central business district and surrounding inner suburbs, whose form took shape when streetcars were the dominant urban transport mode. While most of Tarrant County’s growth took place after the automobile became the domi nant form of transporta tion, there exis ts in Tarrant County a core whose land uses were shaped around transit and that presumab ly today offers a hospitable environment in which transit can prosper. Planners for The T have taken advantage of this situation and have continued to focus transit routes as connectors between suburban residences and CBD jobs. They further have enhanced transit service by overlaying a network of express buses between outlying neighborhoods and the CBD during week day peak travel periods. In contrast, no such central business district existed in Broward County, which consisted during the pre‐auto era of very small towns strung out along a railroad line. The urban form of Broward County began to take shape later, long after the private automobile was the dominant form of urban transportation. No central business district then emerged. Instead, employment as it grew in Broward County scattered about the county. Private transit se rvice that survived into the 1970s connected residential areas with the small downtown of Ft. Lauderdale, but the private service attracted few riders, prompting planners to think of another way of serving the market when the county took over the service. So, based on urban form we would expect transit to perform much better in Tarrant County than in Broward. And yet, just the opposite has transpired. Part of the explanation for this unexpected result derives from organization and funding. As a county‐wide agency, Broward County Transit is compelled to think of ways of serving the entire county, not just the small downtowns. The T, in contrast, has its organization roots in the city of Ft. Worth, and othe r jurisdictions in the county do not want to pay for service provided by The T. Consequently, The T thinks of its market much differently than BCT. We think that it is the difference in thinking that accounts for the rest of the difference in performance between the two systems. Large areas of employment in Tarrant County remain un‐served by transit, and much of the suburban employment that is served is done so ineffectively because of circuitous routing. The T serves the Ft. Worth CBD well but other possible destinations less well. In contrast, BCT with its grid route structure on major arterial roads serves most destinations tolerably directly. This contrast suggests to us that how a transit system uses its route structure to connect together origins and destinations is more important to developing ridership than is the design of the origins and destinations. This is not to say that policies for concentrating development around stops at both the origin and destination of transit trips would not boost transit ridership. The modeling effort of Chapter 4 clearly shows that if walk trips to and from buses were shorter in Broward County, without sacrificing route speeds or headways to accomplish the shorter walks, ridership would increase markedly. One way for shortening walks is through transit oriented development. Over time, if policy through the large scale application of TODs can accommodate population and employment growth in smaller urban regions than otherwise would be the case, transit ridership would increase substantially.

Chapter 6. Discussion and Conclusions

This study seeks to understand the relative efficacy of two classes of policies intended to increase the ridership and productivity of public transit service in Florida. One class of policies seeks to improve transit effectiveness by freezing transit service in the older parts of metropolitan areas (where it is alleged that higher densities of population and employment and the presence of pedestrian amenities induce higher levels of transit demand) and directing new population and employment growth to redeveloped areas around transit stops in the older areas. The other class of policies seeks to connect employment and population, wherever it locates, as directly as possible by transit routes. The thrust of transit development of this second category of policies is in the newer rather than older parts of metropolitan areas, because it is in the newer areas where most population and employment growth is located. The study uses two methods, both focused on transit service in Broward County, Florida. The first method, presented in Chapter 4, is statistical and seeks to examine transit ridership between every pair of traffic analysis zones in Broward County in order to understand the importance of variables that might give rise to that ridership. The variables that we used give insight into both hypotheses; the purpose of the statistical analysis is to understand which of the variables are more important. We conducted our analysis with data for 2005, when there were 921 traffic analysis zones in Broward County and over 800,000 pairs of zones. Because of the fact that transit service did not exist between every pair of zones and the further fact that the Census Bureau suppressed data from some zones for confidentiality reasons, we actually analyzed transit ridership between about 550,000 pairs of zones. The statistical analysis developed a relatively weak model for predicting work transit trips between an origin zone and a destination zone, but that model none‐the‐less speaks clearly about variables that increase transit ridership and those that have little impact. In general, the model supports the efficacy of the second set of policies. The most important consideration in attracting transit ridership is to directly connect population and employment. The analysis shows that it does not matter where the population or the employment are located. Reducing travel time from places where people live to places that they want to go, measured by employment, is by far the most important thing policy can do to increase transit ridership. Policy can shorten transit travel time by restructuring routes, by improving headways, by extending coverage, and by increasing speed. It is not important where the employment is located; that located in the CBD does not have a particularly greater draw than that located elsewhere. It is important to serve it all. The conclusions about the ability of TOD developments to increase transit ridership are clouded by the fact that there are no TOD developments in Broward County, and our efforts to identify TAZs with development that is similar to TOD development were not successful. However, our results from the model clearly indicate that shorter walking times to and from transit are highly important for increasing transit ridership. TODs, if designed properly, will reduce walking time to and from transit and thus will increase transit ridership significantly. An implication of this finding is that planning methods that focus on the relationship of developments to stops will be effective if they take into account how well the stops are connected to all destinations in the region. Creating short walking times along attractive paths will boost transit ridership if the transit stops to which the paths connect are well‐connected to population and employment throughout the region. Another implication is that because both population and employment are dispersed, planners cannot achieve time reductions by implementing direct routes between every pair of origins and destinations. Planners need to think in terms of networks of routes that depend upon transfers. Ideally

routes would be frequent, and if the areas traversed are large, routes would be speedy, as well. Transfer points should be designed for easy movement between routes, and fare structures should facilitate transferring. Running express buses from many neighborhoods to CBDs would be ineffective, because CBDs account for so little of regional employment. However, in larger regions an overlay of a regional grid of limited‐stop routes offering much higher scheduled speeds than local buses, interconnecting all important employment concentrations in a region, is an important component of a transit network that achieves higher ridership. The second method used in this study, presented in Chapter 5, is a case study analysis that comes to similar conclusions to those drawn from the statistical analysis of Chapter 4. The case study compares transit development in Broward County with that in Tarrant County, Texas, where Ft. Worth is located. Both counties are the second counties in their respective metropolitan areas in terms of population and employment. Both counties have similarly sized populations, and both counties have grown at about the same rate over the past several decades. Transit service in both counties connects with relatively recently‐created rail commuter service originating in the dominant county of the respective areas. There are major differences in transit policy between the two counties, however. Broward County has no historic central business district, and the transit system has a county‐wide focus. The route structure is a grid that serves all population and employment concentrations in the county. County residents can get from most parts of the county to most other parts where employment is located. Tarrant County, however, contains the Ft. Worth central business district, and transit service historically developed in Ft. Worth as streetcars focused on that CBD. Transit technology in Ft. Worth now is bus, but the route structure still is largely radial in nature focused on the CBD. There also is a CBD‐focused express bus system super‐imposed on the local routes. Many areas of major employment growth in Tarrant County outside of the CBD remain un‐served by transit, however. The city of Arlington, which contains tens of thousands of jobs, remains the largest urban area in the United States without transit service. So, here we have two transit systems laid out according to two different transit policies. Transit in Broward County attempts to connect most origins to most destinations scattered throughout the county with a grid of routes, requiring many passengers to transfer. Transit in Tarrant County attempts to connect many neighborhoods to the CBD, where large numbers of jobs are located. Both local buses and peak period express buses focus on the Ft. Worth CBD. The idea is to serve one destination well, and the destination that is chosen has well‐developed pedestrian connections to jobs. Which policy is the more effective for attracting transit riders? The case study comparison points to the strategy of connecting all population to all jobs throughout the urban region as being the more effective in stimulating transit ridership. Broward County is an environment where transit is not supposed to work. There is no downtown and employment is scattered. Yet, transit in Broward County carries almost 400 percent more ridership per capita than does transit in Tarrant County, while each bus mile operated in Broward County carries about 35 percent more passengers. In summary, we provide two analyses, one statistical and one a case study comparison. Both analyses point in the same direction. The most effective policies for increasing transit ridership and productivity are those oriented to connecting together population and employment that is decentralized throughout metropolitan regions in Florida. It need not be one policy or the other, however. TOD policies can be important in decentralized areas such as Florida by shortening the walk and improving its attractiveness on each end of the transit trip. Over time, to the extent that TOD policies will promote the creation of more compact metropolitan areas with shorter distances between origins and destinations, such policies will further stimulate transit ridership.

References

American Public Transit Association. Building Better Communities. Washington, D.C.: American Public Transit Association, 1987.

Amos, Nancy (Senior Vice President) and Carla Forman (Assistant VP) of Ft. Worth Transit Authority (The T). Interview conducted by phone with Ms. Amos and Ms. Forman in their office in Ft. Worth by Greg Thompson and Jeff Brown in Tallahassee, 16 July 2008.

Bae, Christine, Chang‐Hee. Orenco Station, Portland, Oregon: A Successful Transit Oriented Development Experiment? Transportation Quarterly, Vol. 56, No. 3, Summer 2002 (9‐18).

Beimborn, Edward. Guidelines for Transit Sensitive Suburban Land Use Design. Washington, D.C.: U.S. Department of Transportation, 1991.

Broward County Transit 2008a. “Route by Route Monthly Ridership October 2007 ‐ September 2008.” Ft. Lauderdale, 4 June 2008.

Broward County Transit 2008b. “Route by Route Hours and Miles, October 2007 through January 2008.” Ft. Lauderdale, 4 June 2008.

Bernick, Michael and Robert Cervero. Transit Villages in the 21st century. New York: McGraw‐Hill, 1997.

Brown, Jeffrey and Dristi Neog. “Reexamining the Link Between Urban Structure and Transit Ridership in the United States.” Tallahassee, FL: Florida Planning and Development Lab, Florida State University, 2007.

Brown, Jeffrey and Gregory L. Thompson. “The Relationship Between Transit Ridership and Urban Decentralization: Insights from Atlanta.” Urban Studie s 45 (5&6): 1119‐1139, 2008. Cited as 2008a.

Brown, Jeffrey and Gregory L. Thompson. “Examining the Influence of Multidestination Service Orientation on Transit Service Productivity: A Multivariate Analysis.” Transportation 35 (2): 237‐252, 2008. Cited as 2008b.

Brown, Jeffrey and Gregory L. Thompson. The Influence of Service Planning Decisions on Rail Transit Success or Failure. San Jose, CA: Mineta Transportation Institute, 2009, cited as 2009a.

Brown, Jeffrey and Gregory .L. Thompson., “Express Bus versus Rail Transit: How the Marriage of Mode and Mission Affects Transit Performance.” Transportation Research Record 2110: 45‐54, 2009, cited as 2009b.

California Department of Transportation. An Analysis of Public Transportation to Attract Non‐Traditional Riders in California. Sacramento, CA: California Department of Transportation, 2003.

Calthorpe, Peter. The Next American Metropolis: Ecology, Community, and the American Dream. New York: Princeton Architectural Press, 1993.

Cambridge Systematic, Inc. Transit Ridership Initiative. Transit Cooperative Research Program Research Results Digest Number 4. Washington, DC: Transportation Research Board, National Research Council, 1995.

Cambridge Systematics, Inc. Continuing Examination of Successful Transit Ridership Initiatives. Transit Cooperative Research Program Research Results Digest Number 29. Washington, DC: Transportation Research Board, National Research Council, 1998.

Cambridge Systematics, Inc. Evaluation of Recent Ridership Increases. Transit Cooperative Research Program Research Results Digest Number 69. Washington, DC: Transportation Research Board, National Research Council, 2005.

Canepa, Brian. Bursting the Bubble ‐ Determining the Transit‐Oriented Development’s Walkable Limits. Transportation Research Record: Journal of the Transportation Research Board, No. 1992, Transportation Research Board of the National Academies, Washington, D.C., 2007, pp. 28–34.

Center for Transit‐Oriented Development (Jeffrey Wood, Mariia Zimmerman, and Shelley Poticha, principal authors). Destinations Matter‐‐‐Building Transit Success, Report FTA CA‐26‐1007. Washington, D.C.: Federal Transit Administration, U.S. Department of Transportation, May 2009.

Cervero, Robert. Ridership Impacts of Transit‐Focused Development in California. Working Paper No. 176, Chapter 2. Berkeley, CA: University of California Transportation Center, 1993.

Cervero, Robert. 1994. “Making Transit Work in the Suburbs.” Transportation Research Record 1451 (1994): 3‐11.

Cervero, Robert. The Transit Metropolis: A Global Inquiry. Washington, D.C.: Island Press, 1998.

Cervero, Robert. “Walk‐and‐Ride: Factors Influencing Pedestrian Access to Transit.” Journal of Public Transportation 3, no. 4 (2000): 1‐23.

Cervero, Robert. “Built Environment and Mode Choice: Toward a Normative Framework.” Transportation Research Part D 7, no. 4 (2002): 262‐284.

Cervero, Robert. “Office Development, Rail Transit, and Commuting Choices.” Journal of Public Transportation 9, no. 5 (2006): 41‐55. Cited as 2006a.

Cervero, Robert. “Transit‐Oriented Development’s Ridership Bonus: A Product of Self‐Selection and Public Policies.” Forthcoming in Environment and Planning A (2006). Cited as 2006b.

Cervero, Robert and John Beutler. Adaptive Transit: Enhancing Suburban Transit Services. Berkeley, CA: University of California Transportation Center, 1993.

Charles River Associates, Inc. Building Transit Ridership: An Exploration of Transit’s Market Share and the Public Policies that Influence It. Transit Cooperative Research Program Report 27. Washington, DC: Transportation Research Board, National Research Council, 1997.

Committee for the Study on the Relationships Among Development Patterns, Vehicle Miles Traveled, and Energy Consumption. Driving and the Built Environment: The Effects of Compact Development on Motorized Travel, Energy Use, and CO2 Emissions – Special Report 298. Washington, D.C.: Transportation Research Board, National Research Council, 2009.

The Corradino Group, in association with AECOM Consult, Inc. Southeast Regional Planning Model VI, 2000 and 2030 Models, Technical Report 3, Model Application Guidelines. Ft. Lauderdale: Florida Department of Transportation, District IV, January 2008.

Crane, Randall and Richard Crepeau. Does Neighborhood Design Influence Travel?: A Behavioral Analysis of Travel Diary and GIS Data. Transportation Research D, Vol, 3, No. 4 (1998): 225‐238.

Dueker, Kenneth, James Strathman, and Martha J. Bianco. Strategies to Attract Auto Users to Public Transportation. Transit Cooperative Research Program Report 40. Washington, DC: Transportation Research Board, National Research Council, 1998.

Dunphy, Robert T. and Douglas R. Porter. Manifestations of Development Goals in Transit‐Oriented Projects. Transportation Research Recorder 1977 (2006).

Elmore‐Yalch, Rebecca. Using Market Segmentation to Increase Transit Ridership. Transit Cooperative Research Program Report 36. Washington, DC: Transportation Research Board, National Research Council, 1998.

Evans, John. Transit Scheduling and Frequency. Transit Cooperative Research Program Report 95, Chapter 9. Washington, DC: Transportation Research Board, National Research Council, 2004.

Ewing, Reid, Keith Bartholomew, Steve Winkelman, Jerry Walters, Don Chen. Growing Cooler: The Evidence on Urban Development and Climate Change. Washington, D.C.: The Urban Land Institute, 2008.

Ferreri, Michael. “Comparative Costs”. In Public Transportation, edited by G.E. Gray and L.A. Hoel. 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 1992.

Frank, Lawrence D. and Gary Pivo. “Impacts of Mixed Use Density on Utilization of Three Modes of Travel: Single‐Occupant Vehicle, Transit, and Walking.” Transportation Research Record 1466 (1994): 44‐ 52.

FWTA, Fort Worth Transportation Authority. “Year‐to‐Date Ridership Comparison by Route.” Fort Worth: Fort Worth Transportation Authority, 31 August 2008.

Frumkin‐Rosengaus, Michelle. Increasing Transit Ridership through a Targeted Transit Marketing Approach. University Microfilms International, 1987.

Gomez‐Ibanez, Jose A. “Big‐City Transit Ridership, Deficits, and Politics: Avoiding Reality in Boston.” Journal of the American Planning Association 62, no. 1 (1996): 30‐50.

Greene, William H.. Econometric Analysis, Second Edition. New York: Macmillan, 1993. Chapter 21, Models with Discrete Dependent Variables, pp. 635‐681.

Haas, Peter. “Ridership Enhancement Quick Study.” Mineta Transportation Institute, San José State University, 2005.

Hadj‐Chikh, Gibran J. and Gregory L. Thompson. “Reaching Jobs in the Suburbs: Tri‐Rail in South Florida.” Transportation Research Record 1618 (1998): 14‐21. 60

Haynes, K. and S. Fotheringham. Gravity and Spatial Interaction Models. Sage Publications, 1984. Pp. 1‐ 28, 40‐47.

Hemily, Brendon. Trends Affecting Public Transit’s Effectiveness: A Review and Proposed Actions. Washington, DC: American Public Transportation Association, 2004.

Hendrickson, Chris. “A Note on Trends in Transit Commuting in the United States Relating to Employment in the Central Business District.” Transportation Research Part A 20, no. 1 (1986): 33‐37.

Hess, Daniel B. and Peter A. Lombardi. Policy Support for and Barriers to Transit‐Oriented Development in the Inner City ‐ Literature Review. Transportation Research Record: Journal of the Transportation Research Board, No. 1887, TRB, National Research Council, Washington, D.C., 2004, pp. 26–33.

Institute of Urban and Regional Development, Parsons Brinckerhoff Quade and Douglas, Inc., Bay Area Economics, and Urban land Institute. Transit‐Oriented Development in the United States: Experiences, Challenges, and Prospects. Transit Cooperative Research Program Report 102, Chapter 8. Washington, DC: Transportation Research Board, National Research Council, 2004.

Jones, David. Mass Motorization + Mass Transit: An American History and Policy Analysis. Bloomington & Indianapolis: Indiana University Press, 2008.

Jones, David. Urban Transit Policy: An Economic and Political History. Englewood Cliffs, NJ: Prentice‐Hall, 1985.

Jun, Myung‐Jin. Are Portland's Smart Growth Policies Related to Reduce Automobile Dependence? Journal of Planning Education and Research 2008.

Kain, John. “Cost‐Effective Alternatives to Atlanta’s Rail Rapid Transit System.” Journal of Transport Economics and Policy (January 1997): 25‐49. 61

Kain, John F. and Zhi Liu. “Secrets of Success: Assessing the Large Increases in Transit Ridership Achieved by Houston and San Diego Transit Providers.” Transportation Research Part A 33, nos. 7/8 (1999): 601‐ 624.

Knaap, Gerrit J., Ding, Chengri & Lewis D. Hopkins. Do Plans Matter? The Effects of Light Rail Plans on Land Values in Station Areas (2001). Journal of Planning Education and Research 21:32‐39.

Kohn, Harold M. “Factors Affecting Urban Transit Ridership.” Paper presented at the Canadian Transportation Research Forum Conference, 2000.

Kuzmyak, J. Richard, Richard Pratt, G. Bruce Douglas, and Frank Spielberg. Land Use and Site Design. Transit Cooperative Research Program Report 95, Chapter 15. Washington, DC: Transportation Research Board, National Research Council, 2003.

Kyte, Michael. Measuring Change in Public Transportation Usage: an Analysis of the Factors Influencing Transit. University Microfilms International, 1986.

Liu, Zhi. Determinants of Public Transit Ridership: Analysis of Post‐World War II Trends and Evaluation of Alternative Networks. Cambridge, MA: Harvard University, 1993.

Long, J. Scott and Jeremy Freese. Regression Models for Categorical Dependent Variables Using Stata. College Station, Texas: A Stata Press Publication, 2006. Chapter 8, Models for Count Outcomes, pp. 349‐ 414.

Lund, Hollie and Richard W. Willson. The Pasadena Gold Line: Development Strategies, Location Decisions, and Travel Characteristics along a New Rail Line in the Los Angeles Region. Mineta Transportation Institute, San Jose State University, 2005.

McLeod, Malcolm, Kevin Flannelly, Laura Flannelly, and Robert Behnke. “Multivariate Time‐Series Model of Transit Ridership Based on Historical, Aggregate Data: The Past, Present, and Future of Honolulu.” Transportation Research Record 1297 (1991): 76‐84.

Manheim, Marvin L.. Fundamentals of Transportation Systems Analysis, Vol. 1, Basic Concepts. Cambridge, MA: The MIT Press, 1979.

Meyer, John and Jose Gomez‐Ibanez. Autos, Transit, and Cities. Cambridge, MA: Harvard University Press, 1981.

Meyer, John, John F. Kain, and Martin Wohl. The Urban Transportation Problem. Cambridge, MA: Harvard University Press, 1965.

Mieger, David and Chaushie Chu. “The Los Angles Metro Green Line: Why Are People Riding the Line to Nowhere?” Paper presented at the 86th Annual Meeting of the Transportation Research Board, 2006.

Mierzejewski, Edward and William Ball. 1990. “New Findings on Factors Related to Transit Use.” ITE Journal (February 1990): 34‐39.

Parker, Terry, Mike McKeever, G.B. Arrington, and Janet Smith‐Heimer. Statewide Transit‐Oriented Development Study: Factors for Success in California. Sacramento, CA: California Department of Transportation, 2002.

Parsons Brinckerhoff Quade & Douglas, Inc. Transit and Urban Form. Transit Cooperative Research Program Report 10, Chapters 1 and 2. Washington, DC: Transportation Research Board, National Research Council, 1996.

Pisarski, Alan. Commuting in America II. Washington, D.C.: Eno Foundation, 1996.

Post, Robert C. Urban Mass Transit: The Life Story of a Technology. Westport, Connecticut, 2007.

Pratt, Richard and John Evans. Bus Routing and Coverage. Transit Cooperative Research Program Report 95, Chapter 10. Washington, DC: Transportation Research Board, National Research Council, 2004.

Pucher, John. “Renaissance for Public Transport in the United States?” Transportation Quarterly 56, no. 1 (2002): 33‐49.

Pucher, John and John Renne. “Socioeconomics of Urban Travel: Evidence from the 2001 NHTS.” Transportation Quarterly 57, no. 3 (2003): 49‐78.

Pucher, John and Christian Lefevre. The Urban Transport Crisis in Europe and North America. London: MacMillan Press Ltd., 1996.

Pushkarev, B. and J. Zupan. Public Transportation and Land Use Policy. Bloomington, IN: Indiana University Press, 1977.

Schlossberg, Marc A and Nathaniel Brown. “Comparing Transit Oriented Developments Based on Walkability Indicators”. Transportation Research Record 1887 (2004): 34‐42.

Skinner, Jon. Elderly and Youth Bus Ridership: A Comparison of Routes in Miami‐Dade County. Master’s degree paper, Department of Urban and Regional Planning, Florida State University, 2007.

Song, Yan and Gerrit‐Jan Knaap. Measuring the Effects of Mixed Land Uses on Housing Values. Regional Science and Urban Economics 34 (2004): 663‐680.

Song, Yan. Smart Growth and Urban Development Pattern: A Comparative Study. International Regional Science Review 28, 2: 239–265 (April 2005).

Spillar, Robert and G. Scott Rutherford. “The Effects of Population Density and Income on Per Capita Transit Ridership in Western American Cities.” Paper presented at the 60th Annual Meeting of the Institute of Transportation Engineers, 1998.

Stoleson, Spencer of Broward County Transit. Interview conducted by phone with Mr. Stoleson in his office in Broward County by Greg Thompson and Jeff Brown in Tallahassee, 13 June 2008.

Taylor, Brian D. “Unjust Equity: An Examination of California’s Transportation Development Act.” Transportation Research Record 1297 (1991): 85‐92.

Taylor, Brian, Peter Haas, Brent Boyd, Daniel Hess, Hiroyuki Iseki, and Alison Yoh. Increasing Transit Ridership: Lessons from the Most Successful Transit Systems in the 1990s. San Jose, CA: Mineta Transportation Institute, 2002.

Taylor, Brian D., Douglas Miller, Hiroyuki Iseki, and Camille Fink. “Analyzing the Determinants of Transit Ridership Using a Two‐Stage Least Squares Regression in a National Sample of Urbanized Areas.” Presented at the 2004 Annual Meeting of the Transportation Research Board, 2003.

Thompson, Gregory, Jeffrey Brown, Rupa Sharma, and Samuel Scheib. “Where Transit Use is Growing: Surprising Results.” Journal of Public Transportation 9, no. 2 (2006): 25‐43.

Thompson, Gregory and Jeffrey Brown. “Explaining Variation in Transit Ridership in U.S. Metropolitan Areas Between 1990 and 2000: A Multivariate Analysis.” Transportation Research Record 1986 (2006): 172‐181.

Thompson, Gregory L. and T. G. Matoff. “Keeping Up with the Joneses: Planning for Transit in Decentralizing Regions.” Journal of the American Planning Association 69, no. 3 (2003): 296‐312.

TRL Limited. The Demand for Public Transport: A Practical Guide. TRL Report TRL 593, 2004.

Thompson, Gregory L.. “Identifying Gainers and Losers from Transit Service Change: A Method Applied to Sacramento.” Journal of Planning Education and Research 17 (Winter,1998): 125‐136.

TranSystems, Inc., Planners Collaborative, and Tom Crikelair Associates. Elements Needed to Create High Ridership Transit Systems. Transit Cooperative Research Program Report 111. Washington, DC: Transportation Research Board, National Research Council, 2007.

Urbitran Associates, Inc., Cambridge Systematics, Inc., Kittelson and Associates, Inc., Pittman and Associates, Inc. and Center for Urban Transportation Research. Guidebook for Evaluating, Selecting, and Implementing Suburban Transit Services. Transit Cooperative Research Program Report 116. Washington, DC: Transportation Research Board, National Research Council, 2006.

Vuchic, Vukan. Urban Transit: Operations, Planning, and Economics. Hoboken, NJ: John Wiley and Sons, 2005.

ANNOTATED BIBLIOGRAPHY

American Public Transit Association. Building Better Communities. Washington, D.C.: American Public Transit Association, 1987.

This study focuses on a host of land use strategies that can integrate with transit planning. The authors favor integration of land use and transit planning, although it may require changes to local ordinances, regulations, building codes and procedures. They suggest many land use strategies that allow public agencies and developers to integrate the impact of mass transit investments and private sector financial participation. They also explain efficient strategies for developers like subdivision and activity design strategies, travel demand strategies, and transportation management associations. They include designing policies, working with the investment community, urban design considerations, ordinances and regulations, comprehensive planning, developer‐furnished improvements, adequate public facilities, etc. Finally, they conclude that if the land use planning is transit supportive then it can bring about an increase in transit ridership.

Bae, Christine, Chang‐Hee. Orenco Station, Portland, Oregon: A Successful Transit Oriented Development Experiment? Transportation Quarterly, Vol. 56, No. 3, Summer 2002 (9‐18).

The authors undertake a review of existing literature on transit‐oriented development and use Orenco Station in Portland as a study site against which to apply the literature’s principles of successful TOD. Orenco Station is different from many TODs in that much of the development is some distance away from the rail station (very little is within a quarter mile), an artifact of the preexisting land ownership situation. Based on their literature review, the authors assert that successful TODs have certain key requirements: the need for supportive land policies around rail (or bus) stations and terminals; the promotion of high density residential development near stations; some commercial and mixed‐use development; and pedestrian design elements. Established in an area of market gardens, the authors note that Orenco Station had few amenities to claim as a locational advantage. Pacific Trust decided that access to the MAX station was its key amenity. However, for most residents it remains an "option demand;" it is there if we need it, we may use it, but we probably never will. The real attraction of access to MAX from Orenco Station may not be as a commuting mode, but rather for an evening visit to downtown Portland for a concert or dinner.

Beimborn, Edward. Guidelines for Transit Sensitive Suburban Land Use Design. Washington, D.C.: U.S. Department of Transportation, 1991.

Transit ridership keeps declining, partly due to its failure to capture riders in the suburbs. The dispersed land use pattern that exists there is the major reason responsible for the transit failure in suburbia. This guidebook introduces elements of successful transit and criteria for transit‐sensitive suburban land use design. It presents a list of transit‐oriented and transit‐

compatible land uses to be included in an area served by transit and which should be located elsewhere. It presents guidelines for land use policies, access policies and transit policies under two major frameworks: system planning and district planning. It further outlines administrative and policy guidelines for transit agencies and local government. It also presents implementation methods as well as a case study wherein the guidelines were applied to develop a successful transit‐oriented development in an emerging suburban area in the City of Milwaukee.

Bernick, Michael and Robert Cervero. Transit Villages in the 21st century. New York: McGraw‐Hill, 1997.

The focus of the book is the emergence of transit villages (i.e. transit‐oriented development) as a reaction to declining quality of life. The authors see the transit village concept as a way of achieving a host of social benefits, ranging from air quality to quality of life. The authors open the book with a discussion of the historic influences on contemporary transit villages. They then make their case for the numerous benefits of a transit village approach to urban development. The primary source of many benefits is the mode shift from solo auto use to transit and non‐ solo auto modes. The authors acknowledge that recent rail transit ridership forecasts have been very inaccurate, but they argue that the numbers might materialize if auto use was priced at its full social cost. They then summarize a host of earlier empirical and qualitative case study research, including their own work, on transit oriented development. Much of the work, including the same detailed case studies, can also be found in Cervero’s The Transit Metropolis published a year later. The lessons are similar to Cervero’s other work on the subject—namely that transit‐oriented developments can lead to increased transit ridership and also promote a wide array of societal benefits.

Controlling for urban area density, unemployment rate, motor fuel prices, transit service frequency, transit service coverage, and the percent of households that do not own an automobile, this study examines the relationship between urban structure (defined as percent of MSA employment in the CBD) and two measures of transit patronage (passenger kilometers per capita, transit journey‐to‐work mode share) in 1990 and 2000 for all US metropolitan statistical areas (MSAs) with more than 500,000 persons. The authors collected data from the US Bureau of Economic Analysis, US Bureau of Labor Statistics, US Census Bureau, and National Transit Database. They obtained the following variables:

PKM_PC = Passenger kilometers per capita (aggregated for all agencies in the MSA)

JTW_MS = Transit journey‐to‐work mode share (aggregated for all counties in the MSA)

CBDEMPSHARE = percent of MSA employment in the CBD

FARE_KM = Fare revenue per passenger kilometer (inflation‐adjusted to 2005 dollars, aggregated for all agencies in the MSA)

FREQUENCY = Service frequency, defined as the ratio of vehicle kilometers to route kilometers (aggregated for all agencies in the MSA)

COVERAGE = Service coverage, defined as the ratio of route kilometers to population (aggregated for all agencies in the MSA)

CARLESS_HH = Percent of MSA households that do not own an automobile (aggregated for all counties in the MSA)

UNEMPLOY = Unemployment rate (by MSA)

FUEL = Motor fuel price index (by MSA)

UZADENS_KM = Urbanized area density, defined as persons per square kilometer

The authors used these variables to estimate multivariate models for each of the two transit ridership variables for 1990 and 2000 for three different groups of MSAs: all MSAs, MSAs with 1 million to 5 million persons, and MSAs with 500,000 to 1 million persons. The authors used the natural log transformations of the variables in order to interpret the coefficients as elasticities. To illustrate, the model for passenger miles per capita read as follows:

LN (PKM_PC) = Constant + LN (CBDEMPSHARE) + LN (FARE_KM) + LN (FREQUENCY) + LN (COVERAGE) + LN (CARLESS_HH) + LN (UNEMPLOY) + LN (FUEL) + LN (UZADENS_KM)

The authors find no statistically significant links between the percent of MSA employment in the CBD and transit ridership. The authors find the strongest links between the two service variables (service frequency and service coverage) and transit ridership. Both of these variables are at least partially under the control of transit agency managers. The other consistently significant variable is the percent of MSA households that do not own an automobile. This is an external factor beyond the control of agency managers. The other external factor variables reveal inconsistent relationships across the dependent variables, across time, and across the MSA groups. All the authors’ models had high R squared values and large F statistics.

Brown, Jeffrey and Gregory L. Thompson. “The Relationship Between Transit Ridership and Urban Decentral ization: Insights from Atlanta.” Urban Studies 45 (5&6): 1119‐1139, 2008. Cited as 2008a.

Controlling for passenger fare, service levels, and the proportion of transit service provided by rail, this study examines the relationship between transit ridership and the decentralization of population and employment in Atlanta from 1978 to 2003.

The authors collected data from the Atlanta Regional Commission, U.S. Bureau of Labor Statistics, U.S. Census Bureau, Metropolitan Atlanta Rapid Transit Authority, and National Transit Database. The following are the key variables:

LPT = annual linked passenger trips VKM = annual vehicle kilometers of service PCTRAIL = percent of vehicle kilometers that are railcar miles FARE = average fare per linked trip (in inflation‐adjusted 2005 dollars) FUEL = an index of motor fuel prices EMPMARTA = the level of non‐CBD employment within the MARTA service area RATIO_EMP = ratio of employment outside MARTA service area to employment inside MARTA service area (including CBD) RATIO_POP = ratio of population outside MARTA service area to population inside MARTA service area (including CBD) OLYMPICS = a dummy variable denoting 1996 as the year of the Atlanta Olympics The authors used these variables to estimate the following time‐series model: β1 β 2 β 3 β 4 β 5 LPTt ⎡ VKM t ⎤ ⎡ PCTRAILt ⎤ ⎡ FARESt ⎤ ⎡ FUELt ⎤ ⎡ EMPMARTAt ⎤ = ⎢ ⎥ * ⎢ ⎥ * ⎢ ⎥ * ⎢ ⎥ * ⎢ ⎥ * LPT VKM PCTRAIL FARES FUEL EMPMARTA t−1 ⎣ t−1 ⎦ ⎣ t−1 ⎦ ⎣ t−1 ⎦ ⎣ t−1 ⎦ ⎣ t−1 ⎦ β 6 β 7 ⎡ RATIO _ EMPt ⎤ ⎡ RATIO _ POPt ⎤ ⎢ ⎥ * ⎢ ⎥ {}β *8exp* OLYMPICSt . ⎣ RATIO _ EMPt−1 ⎦ ⎣ RATIO _ POPt−1 ⎦

The authors found that transit ridership is strongly and positively linked to the strength of employment inside the transit agency service area (outside the CBD) and is strongly and negatively linked to the strength of employment of employment beyond the transit agency service area. The authors report no association between the strength of the CBD and transit ridership in Atlanta. The authors note that transit ridership is more strongly linked to the decentralization of employment than to the decentralization of employment. Finally, the authors observe that fare levels and the absolute amount of transit service are also associated with transit ridership. The authors rely on their analysis and anecdotal evidence gleaned from interviews with local planners to infer that MARTA is successfully linking transit patrons to dispersed employment locations.

Between 1990 and 2000, U.S. transit agencies added service and increased ridership, but the ridership increase failed to keep pace with the service increase. The result was a decline in service effectiveness (or productivity). This marks the continuation of a long‐running and often‐ studied trend. The scholarly literature attributes this phenomenon, at least in part, to transit agency decisions to decentralize their service rather than focus on serving the traditional CBD market. Many scholars argue that a decentralized service orientation is both ineffective and inefficient because it attracts few riders and requires large per‐rider subsidies. This research

tests whether a non‐traditional, decentralized service orientation, called multidestination service, results in reduced service productivity. Contrary to what the literature suggests, we find that MSAs whose transit agencies pursued a multidestination service orientation did not experience lower productivity. These results indicate that policies that have encouraged the growth of decentralized transit services have not necessarily been detrimental to the industry.

Brown, Jeffrey and Gregory L. Thompson. The Influence of Service Planning Decisions on Rail Transit Success or Failure. San Jose, CA: Mineta Transportation Institute, 2009, cited as 2009a.

This study examines the evolution of regional transit service and performance between roughly 1975 and 2005 in eleven U. S. metropolitan regions that either adoped rail transit during that period or began the period with rail transit. The eleven metropolitan areas are Atlanta, Dallas‐Ft. Worth, Denver, Miami, Minneapolis‐St. Paul, Pittsburgh, Portland, Sacramento, Salt Lake City, San Diego, and San Jose. The objective was to investigate the relationship between how rail transit was used in the overall regional transit network and transit performance. The method was case studies involving interviews with one to two persons with long‐term knowledge of the evolution of each system, set against study of planning documents and compilation of service and performance statistics going as far back into the period as possible. The study found a wide range in the way rail transit was incorporated into regional transit service strategies. It also found a side range in transit performance. The study found that region’s who used radial strategies to focus on preservation of the region’s CBD generally failed. Such systems were not networks but collection of routes connecfting suburbs with CBDs. Rail lines in such systems were just viewed as another route. The most successful systems were designed to serve many important destinations in their respective metropolitan areas. Rail systems typically served the CBD, but bus routes in the rail corridors were refocused from serving the CBD to serving suburban rail stations and suburban destinations, instead. Rather than forcing bus passengers onto trains, such systems, reducing patronage, such systems opened up new travel opportunities. The overall result was expanded transit patronage per capita and productivity at the top of the systems studied. Most systems studied fell between these extremes.

This paper focuses on the relative roles that express buses and rail transit play in regional transit development and performance between about 1984 and 2006 in four metropolitan statistical areas that make use of both modes. The regions examined are Atlanta, San Diego, Minneapolis‐St. Paul and Pittsburgh. The paper contrasts different missions that different urban areas call upon their transit systems to play, and it shows how express buses and rail transit can fit into those missions. Express buses fit in best to those regional systems that are conceived of as providing excellent service from suburban neighborhoods to CBDs during peak periods; rail transit fits in best with those systems conceptualized as providing multi‐ destination service throughout the day. The paper then examines relative performance of rail 71

transit and express buses in San Diego and Atlanta, which exemplify the multi‐destination mission in at least part of their respective regions, and Minneaplis‐St.Paul and Pittsburgh, which exemplify the radial‐CBD mission in at least part of their regions. It finds that in all four cases rail provides a relatively small percentage of regional service but accounts for larger proportions of passengers carried, whereas the obverse is true for express buses. In the two multi‐destination cases rail carries between 30 and 50 percent of transit passenger miles and is up to ten fold more productive that express buses in terms of passenger miles carried for each service hour provided. Rail is far more productive than express buses in the radial context, as well.

California Department of Transportation. An Analysis of Public Transportation to Attract Non‐ Traditional Riders in California. Sacramento, CA: California Department of Transportation, 2003.

The study sought to determine customer expectations and needs regarding transit and to develop strategies to increase transit ridership. The authors used a combination of literature review, a survey of 3,302 California residents, and focus groups to identify expectations and needs. The authors then used geographic information systems (GIS) analysis to identify locations in the state with the best potential to attract riders. The authors note that external factors (land use patterns, parking availability, and aging population) are significant influences on transit ridership and can hinder efforts to increase ridership. The authors observed that both riders and non‐riders have similar high expectations about service reliability, convenience, comfort, and safety. They also observed that non‐riders are not very likely to commit to using transit even when these high expectations are met. This poses real challenges for agencies seeking to attract more choice riders. The authors identified the state’s four largest metropolitan areas as the regions with the highest potential to attract new riders.

Drawing on interviews with 40 transit agency managers, the authors make observations about the factors that contributed to transit ridership increases between 1991 and 1993. The authors collected and analyzed data on ridership from American Public Transit Association reports to identify candidate systems. The authors then interviewed senior staff at the transit agencies (via telephone) to elicit their comments about the factors they believed accounted for the ridership increases experienced by their agencies. Based on their analyses and interviews, the authors assert that external factors, which are those beyond the control of agency managers, typically have a larger effect on ridership than internal factors, which are those within the control of agency managers. The authors identify population changes, regional economic conditions, and development trends as key external factors that affect transit ridership. The authors identify fare policies, service adjustments, and marketing efforts as key internal factors that affect transit ridership. The authors concede that these findings are based on agency staff perceptions of the influences on transit ridership as opposed any statistical analysis of these candidate factors.

This study is a follow‐up to the 1995 “Transit Ridership Initiative” study. The authors conducted follow‐up interviews with staff at agencies contacted for the earlier study and added a set of additional agencies for a total of more than 50 transit system managers. The interviews focused on agency ridership experiences from 1994 through 1996. The authors followed the same methodology as in the earlier study to identify additional candidate agencies and interviewed the same agencies they had contacted for the prior study. The authors found that the factors identified in the earlier study continued to be commonly cited during the interview process as important determinants of transit ridership.

This study is the third and final report in a series of studies that identify the key factors and initiatives that led to ridership increases at a set of transit agencies. This report focuses on ridership increases from 2000 to 2002 at 28 agencies. The authors used the American Public Transportation Association’s Quarterly Transit Ridership Reports to identify 31 systems with the largest reported ridership increases, including 15 systems that experienced ridership increases from 1994 to 1996 and continued to enjoy ridership increases from 2000 to 2002. The authors then conducted telephone interviews with staff at 28 of the 31 systems. The authors found that the most significant ridership increases were the result of a combination of factors or initiatives. The key initiatives fell into five categories: service adjustments, fare and pricing adaptations, marketing and information initiatives, and new efforts in service coordination, collaboration, and partnering. The authors note that most of the 18 systems that experienced the highest ridership growth improved their ability to serve more riders with greater efficiency.

This analysis seeks to examine whether the established half‐mile TOD radius is accurate. The walkable radius of a TOD is critical not only for its potential to affect sprawl and other negative externalities but also because of the large amount of land value at stake. This paper demonstrates that by expanding the TOD radius by 66% and subsequently allowing for greater density within that space, the amount of available land can be nearly tripled. This expansion could have significant impacts on investment; an area such as Arlington County, Virginia, could see its Metro corridor office space increase from 18.3 million ft2 to roughly 50 million ft2 and residential units increase from 22,500 to 62,500. This paper begins by examining the basis of the half‐mile radius and then considering which traits (including 73

controllable urban factors such as density and form) contribute to its potential size, along with their relative significance. From this information, one can determine why the half‐mile boundary fluctuates in size, the implications for local and regional TOD planning, and the impact the half‐ mile boundary may have on transit use and urban growth. Evidence from Australia and the United States demonstrates that pedestrians are prepared to travel more than 0.5 mi if an accommodating atmosphere prevails. Although variables such as density and design have been determined to play a role in the success of transit developments, researchers have found, “It is difficult to untangle the effects of land use mix and urban design from the effects of density”. This difficulty arises primarily because well‐ connected urban areas also tend to be of higher densities and more suitable to pedestrians. Efforts have been made to sort out findings between the variables, with some success, but questions still remain as to whether mode choice is an effective way to gauge the impact of the given variables. Despite the uncertainty of factors within transit sites, some planners have begun to find that the traditional stance of viewing developments as independent units may complicate the radius dilemma and that a regional context must be established to comprehend the impacts of communities on one another.

The Center for Transit‐Oriented Development is a division of Reconnecting America, an organization dedicated to the development of alternative forms of transportation. This study examines factors affecting rail transit patronage for systems throughout the U.S. and concludes that the dominant variable are the number of jobs served. The report’s authors note that successful systems serve large numbers of jobs in the suburbs; less successful systems do not. The authors conclude that new rail lines need to be constructed where the jobs are; not where rights of way are cheap. The authors do not examine how rail systems and bus systems may be integrated, so that rail line passengers may reach suburban jobs not immediately near rail stations. They dismiss such concerns by stating (erroneously) that bus lines typically are not shifted to make use of rail line investments. The report thus overlooks high ridership light rail lines, such as the Blue Line in Los Angeles (the highest light rail patronage in the U.S.) and the Blue Line in San Diego, both of which use cheap rights of way, but allowing their patrons to access suburban jobs some distance from rail stations through bus connections.

Cervero, Robert. Ridership Impacts of Transit‐Focused Development in California. Working Paper No. 176, Chapter 2. Berkeley, CA: University of California Transportation Center, 1993.

The author provides a literature review of several studies that examine the transit ridership characteristics of residential and commercial projects located near rail transit stations. The literature employs surveys of residents and workers in the San Francisco and Washington metropolitan areas. A 1991 San Francisco Bay Area study reported no relationship between distance to the transit station and transit mode split for housing located within 1/3 mile of the station. A 1989

San Francisco Bay Area study found that 35 to 40 percent of residents living near three Bay Area Rapid Transit District (BART) stations used public transit. A 1987 Washington, DC study found that rail and bus transit mode share declines by 0.65 percent for every 100‐foot increase in distance of a residential site from a rail transit station. The same 1987 study found that ridership was higher at downtown than at suburban work sites and that ridership declined steadily as distance to the station increased. All these studies essentially examined the correlation between transit mode share and distance to a rail station. They did not control for other factors that might influence an individual’s decision to rider transit (fare, service quality, auto access and cost, etc).

Cervero, Robert. 1994. “Making Transit Work in the Suburbs.” Transportation Research Record 1451 (1994): 3‐11.

Rapid decentralization of population and employment over the past several decades has chipped away at the U.S. transit industry's market share. The implications of decentralization on the ridership, operating performance, and fiscal health of the nation's largest transit operators are examined. On the basis of the results of a national survey, a number of service strategies that offer hope for reversing transit's decline are explored, including timed transfers, paratransit services, reverse commute and specialized runs, employer‐sponsored van pools, and high‐ occupancy‐vehicle and dedicated busway facilities. Land use options, like traditional neighborhood designs and transit‐based housing, are also examined. A discussion of various institutional, pricing and organizational considerations when implementing suburban‐targeted service reforms and land use initiatives is also provided. Century‐old models involving joint public‐private development of communities and transit facilities, it is argued, also deserve reconsideration.

Cervero, Robert. The Transit Metropolis: A Global Inquiry. Washington, D.C.: Island Press, 1998.

The author observes that there is a global decline in transit use due to competition with the automobile and continued decentralization of urban areas. However, he notes a dozen metropolitan areas (transit metropolises) that seem to be doing well. The objective of the book is to determine why these cities’ transit systems are so successful. His hypothesis is that they have matched their transit services with their land use patterns. The book is a series of case studies. The author classifies the cities into four categories: 1) adaptive cities (cities using rail transit to guide urban growth, which include: Stockholm—rail‐ served satellite cities; Copenhagen—suburban communities along radial rail lines; Tokyo—new towns served by rail transit; Singapore—strong land use and transport planning); 2) hybrid cities (cities that are tailoring transit to serve their urban forms and adapting urban form using transit: Munich—leveraging existing pro‐transit development patterns; Ottawa—strong use of busway; Curitiba—linear city oriented around bus rapid transit); 3) strong core cities (cities that integrate transit with strong centralized development patterns: Zurich—auto restraint plus pro‐transit policies; Melbourne—using transit to encourage centralized urban pattern); and 4) adaptive transit (cities that adapt transit to serve decentralized urban form: Karlsruhe, Germany—use of adaptive light rail transit; Adelaide, Australia—use of bus ways; Mexico City—hierarchy of transit

services). The author then gathers quantitative and qualitative data to paint a portrait of the city and its use of transit. All the cases are success stories. The author offers fifteen lessons: 1) transit metropolises evolve from a well‐articulated vision of the future; 2)transit metropolises need inspired leadership; 3) they need efficient institutional structures (especially at regional level); 4) they need pro‐active planning processes; 5) they need to maintain strong, viable CBDs; 6) they need balanced traffic flows; 7) the transit agencies need to have an ethos of competition to provide efficient, low‐cost service; 8) they need to give transit priority over the automobile; 9) they take incremental steps; 10) they have people‐friendly urban design; 11) they have policies to restrain automobile ownership and use; 12) they have integrated transit services; 13) they have flexible transit services—give a strong role for buses; 14) they embrace innovation in service delivery; and 15) they take advantage of serendipitous developments. The author closes by briefly discussing five North American cities that he sees as following in the footsteps of the transit metropolises: Portland, Oregon; Vancouver, British Columbia; San Diego, California; St. Louis, Missouri; and Houston, Texas.

Cervero, Robert. “Walk‐and‐Ride: Factors Influencing Pedestrian Access to Transit.” Journal of Public Transportation 3, no. 4 (2000): 1‐23.

The predominant means of reaching suburban rail stations in the United States is by private car. Transit villages strive, among other things, to convert larger shares of rail access trips to walk‐ and‐ride, bike‐and‐ride, and bus‐and‐ride. Empirical evidence on how built environments influence walk‐access to rail transit remains sketchy. In this article, analyses are carried out at two resolutions to address this question. Aggregate data from the San Francisco Bay Area reveal compact, mixed‐use settings with minimal obstructions are conducive to walk‐and‐ride rail patronage. A disaggregate‐level analysis of access trips to Washington Metrorail services by residents of Montgomery County, Maryland, shows that urban design, and particularly sidewalk provisions and street dimensions, significantly influence whether someone reaches a rail stop by foot or not. Elasticities are presented that summarize findings. The article concludes that conversion of park‐and‐ride lots to transit‐oriented developments holds considerable promise for promoting walk‐and‐ride transit usage in years to come.

Cervero, Robert. “Built Environment and Mode Choice: Toward a Normative Framework.” Transportation Research Part D 7, no. 4 (2002): 262‐284.

The author examines the effect of built environment variables measuring density, diversity, and design, as well as generalized modal cost and socioeconomic variables, on individual mode choice in Montgomery County, Maryland. The author obtains trip data from the 1994 Household Travel Survey compiled for the Metropolitan Washington Council of Government and both travel time and land use data from databases compiled for use in area travel forecasting models. The author then estimates probabilistic models for a trip being made by each of three modes of transportation (solo auto, group‐ride auto, and transit) as a function of a vector of land use variables and utility functions associated with making a trip from point A to point B using that mode of travel.

The author finds that land use density and diversity have moderate, inelastic (in the .2 to .6 range) effects on transit ridership. The authors note that design variables have more modest , yet measurable, effects on transit use.

Cervero, Robert. “Office Development, Rail Transit, and Commuting Choices.” Journal of Public Transportation 9, no. 5 (2006): 41‐55. Cited as 2006a.

The article examines commuting behavior in workplace environments served by rail transit. The author compiles information from a number of his empirical studies that explored differences in transit mode share in different kinds of work place environments. The author finds that people working in office buildings near rail transit are three times more likely to use transit than those working further away from rail transit stations. The author argues that the presence of feeder bus services, employer transit subsidies, and scarce parking are all key factors influencing the mode choice decision. The author advises policymakers to promote the use of feeder buses, employer‐based transit subsidies, and flexible parking policies in these near‐station work environments.

Cervero, Robert. “Transit‐Oriented Development’s Ridership Bonus: A Product of Self‐Selection and Public Policies.” Forthcoming in Environment and Planning A (2006). Cited as 2006b.

The author examines what he terms the “ridership bonus” among people living near California rail stations in California by comparing their behavior to people who live beyond comfortable walking distance of the stations. The author used a database on travel behavior and other attributes of 1000 people living in 26 housing projects within ½ mile of urban rail stations in California. He estimated binomial logit models for predicting transit mode choice for residents’ commute trips as a function of travel times, regional accessibility, workplace job and parking policies, neighborhood design, auto ownership levels, and a variable measuring transit lifestyle preference. He also estimated a binomial logit model predicting non‐motorized access to rail stations as a function of income, ownership, and the density of street lighting. Finally, he estimated a pair of nested logit models for location choice and mode choice as a function of an array of location, transportation, household, neighborhood, and individual attributes. The author finds that residential self‐selection (lifestyle preference), employer‐based parking policies, and destination‐area street connectivity are among the key factors that influence residents’ decision to ride transit. The author calls for an array of regulatory (zoning) and market‐based strategies to take advantage of these findings and promote more “transit‐ based” housing.

Cervero, Robert and John Beutler. Adaptive Transit: Enhancing Suburban Transit Services. Berkeley, CA: University of California Transportation Center, 1993.

The authors set out to identify places where transit agencies have implemented services that have allowed them to adapt to the changing population and employment patterns of their metropolitan areas. The authors emphasize the use of seamless services that avoid transferring. They distinguish between three types of adaptive services: technological innovations, bus‐based 77

service innovations, and small‐vehicle Paratransit services. The authors identified ten case studies (including cases in the United States, Canada, Australia, Germany, and Puerto Rico) that involved the use of some form of adaptive service. The authors caution that many of the cases have yet to yield data that would permit a detailed effectiveness evaluation. The authors do not attempt to develop overall lessons, but rely on their individual case studies to provide insights to policymakers and transit managers (the authors’ intended audiences). Among the more promising services identified here were bus rapid transit (then a not widely discussed phenomenon) and free‐market Paratransit services. Interestingly, the authors do not investigate the importance of integrating bus with rail transit, although they include both bus and rail case studies.

The report discusses strategies that have been used to help increase public transit ridership and travel market share. The authors conducted a survey of 50 transit agencies in the United States and Canada, detailed case studies of eight agencies, and a general analysis of the state of the transit industry. The authors had hoped to conduct a quantitative analysis but were unable to do so because of insufficient data and resources. The authors found that ridership growth has not been a priority for the surveyed agencies; they have been focused more on serving existing customers. The survey also found that transit‐related initiatives alone were not sufficient to shift significant numbers of people from the automobile. The report followed the survey with a more detailed investigation of eight case study sites. These included: feeder bus (Metro North), fare integration (Toronto), Express bus (Minneapolis), times transfer (Norfolk), U Pass (Seattle), fareless square (Portland), land use (Toronto), and road pricing. The experiences were judged positive in the cases of Metro North, Toronto, and Seattle. Flat ridership results were reported in Portland. The other cases lacked sufficient data to make a definitive judgment. The authors conclude by noting that policies that make private vehicle use less attractive will have a larger positive effect on ridership than policies that make transit more attractive.

Committee for the Study on the Relationships Among Development Patterns, Vehicle Miles Traveled, and Energy Consumption. Driving and the Built Environment: The Effects of Compact Development on Motorized Travel, Energy Use, and CO2 Emissions – Special Report 298. Washington, D.C.: Transpor tation Research Board, National Research Council, 2009.

The study reported in, Driving and the Built Environment, was requested by the Energy Policy Act of 2005 and was conduted by a committee of the Transportation Research Board of the National Academy of Sciences. Members of the committee include leading researchers on the effects of public policy on the pattern of the built environment, as well as on how different patterns of built environment affect travel choices. Other noted scholars provided commissioned papers, that were literature reviews of different aspects of this topic, including the impact of transportation investments on the pattern of the built environment, the impact of 78

the built environment on VMT, among other topics. The report concludes that if most growth between now and the future could be accomplished within existing urban areas, VMT might be reduced, but that the likelihood of achieving such a radically different pattern of growth from that which has occurred over the past one hundred years and which continues unabated today despite rhetoric and public policy to the contrary, is next to nil. The counterview to this report is contained in, Growing Cooler, (Ewing, et al 2008).

Crane, Randall and Richard Crepeau. Does Neighborhood Design Influence Travel?: A Behavioral Analysis of Travel Diary and GIS Data. Transportation Research D, Vol, 3, No. 4 (1998): 225‐238.

This article attempts to determine to what extent the street layout of a traditional neighbourhood will curb the number of automobile trips. Using data obtained from a survey for travel and socio‐economic variables, and data obtained via GIS for land‐use variables, multiple regression models were run in order to determine in what way neighbourhood design influences mode choice. The land‐use variables were: connected street pattern, mixed street pattern, street network density, residential share of census tract, commercial share of tract, vacant share of census tract, distance to downtown, and squared distance to downtown.An analyses of the results of the calculations revealed that the variables associated with more traditional neighbourhoods (grid‐based streets, higher density, mixed use) were actually more likely to be associated with a higher frequency of automobile trips.

This report examines the effectiveness of automobile parking policies, alone and in conjunction with changes to transit service policy, in attracting automobile users to public transportation. The authors employed a literature review followed by statistical modeling (based on the 1990 NPTS) of likely effects of policy changes, and then conducted an extensive set of case study interviews to capture locations where one or more of eight defined parking policies had been employed. The authors collected information about the specific parking policy and its implementation, transit service changes, other public policy interventions, transit ridership, and the general socio‐economic and land use profile of the case study sites. The authors found strong relationships between parking prices and transit use.

Dunphy, Robert T. and Douglas R. Porter. Manifestations of Development Goals in Transit‐Oriented Projects. Transportation Research Record 1977 (2006).

The paper examines a number of projects generally considered to be effective TODs and reviews their strengths and weaknesses, on the basis of several specific principles representing a TOD perspective. The authors draw conclusions from recent updates and their long‐term familiarity with the cited TOD experiences. The paper demonstrates that the TOD concept is alive and well and evolving toward the described ideals.

To planners, the qualities of development desired for areas adjacent to transit— especially stations or terminals—generally involve the four Ds: density, diversity (mixed uses), design, and distance from nearby development to transit facilities . When 10 transit agencies were asked to define “transit‐oriented development,” most emphasized the importance of high‐ quality walking environments. Four called for mixed uses, and, interestingly, only two mentioned high density. The response of Bay Area Rapid transit (BART) in California to the survey mentioned higher density development with a mix of residential, employment, and shopping designed for pedestrians (without excluding the automobile). The Washington Metropolitan Area Transit Authority emphasized results—smart growth development that would reduce reliance on the car, encourage pedestrian and bike access, foster safe station environments, enhance physical connections, and provide a vibrant mix of land use activities.

This document is a guidebook that covers issues, procedures, and strategies associated with the use of market segmentation to tailor ridership initiatives to particular markets of transit customers. The guidebook discusses the application of market segmentation strategies in Boise, Milwaukee, and Washington, DC. The discussion does not contain data that would allow one to directly connect the strategy to increased ridership. The guidebook is simply designed to introduce agency managers to market segmentation concepts and their application.

Evans, John. Transit Scheduling and Frequency. Transit Cooperative Research Program Report 95, Chapter 9. Washington, DC: Transportation Research Board, National Research Council, 2004.

This chapter is part of a larger study of traveler responses to transportation system changes. This chapter examines changes to transit schedules and frequencies. It does not examine changes to transit service structures. The authors recount the results of a series of studies dating from the 1960s to 2000s on different service schedule and/or frequency changes and the ridership results. The authors use this information to calculate the elasticity of ridership with respect to the particular service change. The author found that ridership does respond to service frequency or schedule changes (elasticity = 0.5), and that the largest responses are found in higher income areas that previously had very infrequent service. In more traditional transit areas, the ridership response was more modest. The author use the results of rider surveys to note that between one half and one third of the new transit riders would have previously driven cars to make their trip.

Ewing, Reid, Keith Bartholomew, Steve Winkelman, Jerry Walters, Don Chen. Growing Cooler: The Evidence on Urban Development and Climate Change. Washington, D.C.: The Urban Land Institute, 2008.

The argument of this report is that it is impossible to achieve objectives for reducing greenhouse gas emissions without reducing vehicle miles traveled (VMT) by automobiles and light trucks in 80

U.S. metropolitan areas. The authors argue further that existing patterns of sprawled urban develop do not permit VMT reduction. However, by accommodating most growth between now and 2050 within existing urban areas through dense, mixed use development with revitalized CBDs, VMT will be reduced by up to 40 percent, compared to an alternative scenario of accommodating the same growth in sprawled development on the edges of metropolitan areas. The reduced VMT would come from shorter trip lengths and more trips made on public transportation and by bikes and walking. The authors argue that the alternative form of development, that they term, “smart growth,” is achievable and is preferable to imposing the social cost of driving on the U.S.public. This point is refuted in Driving and the Built Environment (Committee for the Study . . . 2009). See above.

Ferreri, Michael. “Comparative Costs”. In Public Transportation, edited by G.E. Gray and L.A. Hoel. 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 1992.

Synopsis This chapter in the Gray and Hoel text discusses the various components of operating and capital transit costs. Its usefulness for our study is in its assertions that transit is best suited to serving the CBD and other traditional transit markets. The chapter attributes the decline in transit service productivity to decentralizing urban forms and the dispersion of activities throughout the urban area. It notes that transit has a particularly difficult time effectively serving this kind of urban environment. The chapter is therefore reflective of the traditional view in the literature.

Frumkin‐Rosengaus, Michelle. Increasing Transit Ridership through a Targeted Transit Marketing Approach. University Microfilms International, 1987.

This dissertation concentrates on commuters as the target market segment, analyzing their response to transit marketing at the place of employment. This dissertation tested two marketing theories. The first is a Peer Pressure Theory proposing that it is more effective for a marketing campaign to target areas of existing high ridership. The second is a Utilitarian Theory suggesting that marketing campaigns will have an effect regardless of the area's previous ridership trends. Santa Clara County Transit was used as the case study. The major employment centers used for the analysis were Varian, Lockheed and several companies located within Moffett Park. Information was obtained for 545 transit riders. After nine months of marketing campaigns, 21 percent of the transit riders were new riders and 79 percent were riders who were utilizing the transit service before the marketing efforts began. Several multivariate statistical techniques were used to analyze the data. A correlation analysis showed a positive, though small, correlation between the residential areas of new riders and old riders. A principal components analysis indicated that 88 percent of the variation of new riders could be explained with fourteen variables combined into three components. A multiple regression analysis showed that new riders could be predicted with a standard error of 1.6, yielding a multiple correlation coefficient of 0.8 between the number of predicted and observed new riders. The research findings indicate that the response to transit marketing is, in fact, related to a peer pressure effect and to the diffusion of information, but there were other important factors as well. A long distance from place of residence to place of 81

employment, in terms of commute time, was a key variable. High ridership areas were also characterized by a concentration of high household incomes, a predominance of white collar workers and the existence of a conveniently located park‐and‐ride lot.

Gomez‐Ibanez, Jose A. “Big‐City Transit Ridership, Deficits, and Politics: Avoiding Reality in Boston.” Journal of the American Planning Association 62, no. 1 (1996): 30‐50.

The author examines changes in ridership and agency deficits for the Massachusetts Bay Transportation Authority in Boston from 1970‐1990. He employs a multivariate statistical analysis that tests the effects of internal factors (fare and service policies) and external factors (per‐capita income and employment in the city of Boston) on ridership levels. The author obtained data for the time period 1970 to 1990. The author estimated multivariate models that include the following variables:

Income = Real per‐capita income for the MSA (in log form)

Employment = Jobs in the city of Boston (in log form)

Fare = Real average fare per passenger trip (in log form, one‐year lag)

Vehicle miles = Vehicle miles of service operated by MBTA (in log form, one‐year lag)

Dummy variable for 1980‐1981 = Dummy for year in which MBTA service was reduced considerably

In one multivariate model, the author substituted a trend variable for the income variable, with marginal effects on either model performance or the significance levels and elasticities associated with the explanatory variables. The author found: 1) a 1 percent decline in the percent of jobs in the city of Boston was associated with between a 1.24 percent and 1.75 percent decline in ridership; 2) a one percent increase in real per‐capita incomes was associated with a 0.71 percent decline in ridership; 3) a one percent increase in fares was associated with a .22 to .23 percent decline in ridership; and 4) a one percent increase in vehicle miles of service was associated with a .30 to .36 percent increase in ridership. The authors’ models accounted for nearly 90 percent of the variation in MBTA ridership from 1970‐ 1990. Durbin‐Watson statistics indicate that the models are appropriately specified. The author uses the model results to state that transit ridership in Boston has been strongly influenced by factors beyond the agency’s control (particularly the decentralization of employment). However, the definition of employment is problematic and measures jobs throughout the city of Boston as opposed to jobs inside the central business districts of Boston and Cambridge, which the author had hoped to measure. The authors findings are considerably different from those obtained by Brown and Thompson (2006) for Atlanta, but there are considerable differences in the definition and treatment of employment in these studies.

Haas, Peter. “Ridership Enhancement Quick Study.” Mineta Transportation Institute, San José State University, 2005.

The author identifies and discusses the specific characteristics or factors that might lead an agency to adopt one or more of four strategies (ECO pass programs, guaranteed ride home programs, day passes, and on‐line fare media sales programs) that are frequently cited as effective ways to boost ridership. The author focuses on the 150 largest transit agencies in the United States. The author used a preliminary search of agency websites to identify agencies that use these strategies and then conducted interviews with managers at each of the agencies. The author identified a number of service, urban structure, and travel characteristics that seem to act as barriers to the introduction of these strategies (low density, system size, service hours, etc). The author then identified a number of agencies that he believes represent likely candidates for the successful introduction and adoption of these strategies.

Hadj‐Chikh, Gibran J. and Gregory L. Thompson. “Reaching Jobs in the Suburbs: Tri‐Rail in South Florida.” Transportation Research Record 1618 (1998): 14‐21.

The authors examine traffic patterns on the Tri‐Rail commuter rail system in south Florida. The station siting process led to the construction of some stations that seemed well‐suited to serving suburban transit markets as opposed to the central business district‐bound market. The authors compare the degree to which people are using the service to reach suburban destinations versus the central business district. The authors gathered ridership data from Tri‐Rail staff. These data provided ridership between all pairs of stations (from automated ticket machines) for one work week during a twelve‐hour period (4 a.m. to 4 p.m.). The authors classified station pairs as serving the suburb‐ to‐suburb or suburb‐to‐CBD market. They made comparisons between the two markets for six distance categories. The authors evaluated three hypotheses. The first hypothesis tested whether suburban jobs could support commuter rail to the same degree as CBD jobs. They estimated a gravity model as part of the process of testing this hypothesis. The second hypothesis tested the ability of stations to serve their potential market. They estimated an index of market penetration to evaluate this hypothesis. The third hypothesis tested whether the degree of market penetration of a station pair was related to the distance between the stations. They estimated a multivariate model to evaluate this hypothesis. The authors find that both markets have comparable total potential ridership. They identify potential ridership all along the Tri‐Rail corridor, not just where the CBD is the destination. The authors found that Tri‐Rail penetrates the suburb‐to‐CBD market about twice as much as the average suburb‐to‐suburb market. The authors also found that market penetration increased with distance, although the model left a considerable amount of unexplained variation in the dependent variable. The authors use the results to highlight the existence of sizeable suburb‐to‐suburb demand for commuter rail service. They further observe that commuter rail planners who are developing their systems to serve CBD markets might be able to tap this potential market at very little additional cost.

Hemily, Brendon. Trends Affecting Public Transit’s Effectiveness: A Review and Proposed Actions. Washington, DC: American Public Transportation Association, 2004.

The author reviews a wide range of data, including socio‐economic trends, changes in land use and mobility patterns, societal changes, and emerging professional practices to distill the “challenges they create for transit system effectiveness and for the industry as a whole, and to identify some questions, opportunities, and potential actions for consideration in the formulation of future strategic directions for transit in the community (vii). The author compiles literature and data from a wide range of sources to paint a portrait of the continuing evolution of communities and the implications these continuing changes to patterns of residential location, employment location, and mobility desires and needs have for the transit industry. The author uses the review to identify a new vision for transit’s role in the community. This vision is “[a] transportation system that meets the needs for mobility and accessibility while balancing the current and long‐term goals of economic growth, environmental quality, and social equity” (viii). The author identifies three key actions that should be pursued to achieve the vision: provision of new transit infrastructure, a focus at all levels of government on smart growth and sustainable land use planning, and more use of market segmentation strategies that are designed to tailor transit services to the specific needs of different rider groups.

Hendrickson, Chris. “A Note on Trends in Transit Commuting in the United States Relating to Employment in the Central Business District.” Transportation Research Part A 20, no. 1 (1986): 33‐37.

The author uses basic statistical analysis to examine the link between public transit ridership and number of jobs in the central business district in 1970 and 1980, and the change between 1970 and 1980. The author uses transit commute mode share as the measure of ridership. The sample consists of 25 large metropolitan areas in the U.S. The author gathers data from the US Census Bureau to estimate a series of multivariate models . The first multivariate model estimates ridership in 1970 as a function of CBD employment in 1970 (R square = .96), the second model estimates ridership in 1980 as a function of CBD employment in 1980 (R square = .90), and the third model estimates ridership in 1970 as a function of both CBD employment and the total number of workers in the metropolitan area (R square = .98). The author then estimates two change models, one with a dummy variable for Sunbelt cities (R square =.77) and one without (R square = .66). Finally, he estimates a change model including dummy variables for both Sunbelt cities and those with fixed rail systems (R square = .81). The author finds strong relationships between CBD employment and transit commute mode share. The author finds positive, statistically significant effects on transit commute mode share from the Sunbelt dummy variable, and negative, statistically significant effects from the fixed‐rail dummy variable. The study’s shortcomings include: 1) the lack of control variables and 2) the mixing of cities with significant differences in both the size of the CBD and the transit commu te mode share. Particularly problematic is the inclusion of New York, which dwarfs other cities on both variables, in the data set.

Hess, Daniel B. and Peter A. Lombardi. Policy Support for and Barriers to Transit‐Oriented Development in the Inner City ‐ Literature Review. Transportation Research Record: Journal of the 84

Transportation Research Board, No. 1887, TRB, National Research Council, Washington, D.C., 2004, pp. 26–33.

The policies that are widely believed to be supportive of TOD are examined, the gap in knowledge about TOD in established city neighborhoods is addressed, and the challenges of TOD in different urban settings are compared. The authors find that (a) the literature appears to be consistent and confident in outlining the public policies that encourage TOD; (b) researchers tend to focus on TODs in suburban and greenfield areas of fast‐growing regions in the western and southern United States; (c) TODs in older cities are not well publicized and are largely ignored by the literature; and (d) researchers who study inner‐city TOD usually focus on the lack of it, or any type of development, in economically depressed areas. The conclusion of several researchers that a strong local economy is key to successful TOD offers a clue as to why recently built TOD is largely absent from many older, slow‐growth cities like Buffalo, New York, and St. Louis, Missouri. It also offers some insight into why the TOD trend is strongest in high‐growth metropolitan areas like San Diego, California, and why it seems to skip struggling neighborhoods within them, like South Central Los Angeles, California.

Institute of Urban and Regional Development, Parsons Brinckerhoff Quade and Douglas, Inc., Bay Area Economics, and Urban land Institute. Transit‐Oriented Development in the United States: Experiences, Challenges, and Prospects. Transit Cooperative Research Program Report 102, Chapter 8. Washington, DC: Transpo rtation Research Board, National Research Council, 2004.

This chapter from a TCRP report on transit‐oriented development examines evidence about the ridership effects. The authors query an extensive literature that examines transit ridership at both residential and employment‐related land uses that meet the characteristics of transit‐ oriented development. The authors report the descriptive results of residential studies showing that: 1) workers living near BART were six times more likely to use it for commute trips than the average Bay Area resident; 2) workers living near light rail transit in Silicon Valley were five times more likely to use transit for commute trips than average area residents; and 3) people living near transit in Washington, DC have high transit mode shares that decline with increased distance from a transit station. The authors also summarize a set of office and retail studies that showed: 1) 50 percent of those working within 1,000 feet of a downtown Washington Metro station used rail to get to work; 2) 60 percent of customers at a downtown San Diego shopping center located two blocks from light rail arrived either by transit or by foot; and 3) 34 percent of patrons at a downtown San Francisco shopping center that has a direct connection to BART arrived by transit. The authors also present a set of multivariate models from studies for the San Francisco Bay Area and Arlington County, Virginia that indicate particularly strong relationships between the density of the land use and transit ridership. Overall, the authors conclude that residents living in Transit‐Oriented Development usually patronize transit five to six times as often as the typical resident of a region. The authors acknowledge that self‐selection bias might be an issue in the residential studies they discuss.

Jones, David. Mass Motorization + Mass Transit: An American History and Policy Analysis. Bloomington & Indianapolis: Indiana University Press, 2008.

Jones examines the rise and fall of mass transit and of mass motorization in the U.S., Canada, and in several European and Asian countries. Many planners hope that the U.S. eventually will see the U.S. following the better European examples in terms of land use and transportation policy, but his impressive compilation of the historical record demonstrates that the opposite is happening, with disastrous world‐wide environmental consequences. Jones argues that peoples of all cultures choose suburbanization and automobile transportation when their incomes rise to the point where they can make such choices. He further argues that most societies behave similarly at the same level of auto ownership. Europe has looked attractive because it has lagged the U.S. in auto ownership, but the travel and land use decisions of its populations is becoming more like those in the U.S. as its levels of auto ownership approach those in the U.S.. None‐the‐ less, Jones concludes that on the margin European societies will use their autos somewhat less that U.S. people, because of higher taxes on driving, much better transit systems, and denser land uses, even in new suburban areas that also are served by transit. Jones recommends better auto technology and smaller, lighter autos to lessen their environmental impacts, heavy pricing and taxation of auto and truck transportation, and in some cases rail transit systems serving TOD type developments.

Jones, David. Urban Transit Policy: An Economic and Political History. Englewood Cliffs, NJ: Prentice‐ Hall, 1985.

Jones’ book is an account of the past several decades of public transit history. He focuses a great deal of attention on the loss of most transit markets to the automobile during the period from the 1920s to the 1950s and its shrinking to focus primarily on the CBD‐bound commuter and transit dependent riders.

Kain, John. “Cost‐Effective Alternatives to Atlanta’s Rail Rapid Transit System.” Journal of Transport Economics and Policy (January 1997): 25‐49.

The article examines the policy history of rail transit in Atlanta, estimates a multivariate time‐ series model to explain ridership change from 1972 to 1993, and uses the estimated model to examine the likely performance of alternatives to rail rapid transit development. The paper estimates multivariate time series models that predict ridership as a function of fares, service miles, vehicle size, fuel prices, regional employment, and a trend variable that functions as a proxy for decentralization. Most of the variables are inserted in the model in their natural log forms. The models indicate that ridership was strongly influenced by fares, service, vehicle size, and fuel prices. The trend variable also proved statistically significant, although the elasticity is quite small. Brown and Thompson (2006) present a model that extends Kain’s work and incorporates more direct measures of population and employment decentralization.

Jun, Myung‐Jin. Are Portland's Smart Growth Policies Related to Reduce Automobile Dependence? Journal of Planning Education and Research 2008.

This study investigates the effects of smart growth policies on commuters’ choice to drive alone with emphasis on four types of smart growth policies implemented in Portland: the UGB, public transit such as the MAX light rail system and bus service, mixed land use, and TODs. This study attempts to assess the impact of Portland’s smart growth on automobile dependence by building logistic regression models, after controlling other variables which may affect mode choice in commuting. Empirical evidence reveals mixed results for smart growth proponents. Higher accessibility to the MAX light rail and bus service and more mixed uses of land were significantly associated with higher probabilities of commuting by the alternative modes to private vehicles, while TOD and higher residential and employment densities were hardly related to a reduction in the choice to drive alone. In addition, higher accessibility to freeway interchanges and a higher share of single family residential land resulted in a greater likelihood of driving alone. Thus, more diversified land use in neighborhoods, more extensive provision of public transit service, and decreasing accessibility to freeway interchanges were associated with fewer choices of driving alone, while making settlements compact via the UGB and TODs has no clear relationship with reducing the choice to drive alone. Empirical analyses also suggest that the provision of public transit service and mixed land use implemented at residential zones (origins) were more effective in reducing automobile dependence than those implemented at places of work (destinations). The empirical analysis has substantial policy implications. First, densification of population and employment via the UGB had no direct impact on a reduction in automobile dependence, while additional mixed land use in the place of residence would be an effective smart growth tool for reducing single‐occupant commuter vehicles. Second, the negative relationship of transit accessibility and the positive relationship of freeway accessibility with automobile dependence suggest an argument for increased subsidy and investment in public transit to reduce automobile dependence.

The authors examine the experiences of transit systems in Houston and San Diego that achieved large ridership increases during a period when transit systems in most other metropolitan areas experienced large ridership declines. They develop a series of multivariate models that seek to explain variation the variation in ridership over time as a function of fares, service, automobile variables, per‐capita income, and regional employment variables. The authors estimate time‐series models to explain variation in ridership from year‐to‐ year, and then use the model estimates to investigate the likely ridership effects of different fare and service strategies. The authors discovered that the variables most strongly connected to ridership are service levels, fare levels, and metropolitan employment and population growth. This study is an update of a study the authors conducted in 1995 for the Federal Transit Administration.

The authors used their models to develop estimates of operating and total costs per passenger boarding and per passenger mile for Houston’s bus transit system and San Diego’s bus and light rail transit operators. These estimates suggest that the bus systems are more cost‐ effective than the light rail systems when evaluated on the basis of total costs per passenger.

Knaap, Gerrit J., Ding, Chengri & Lewis D. Hopkins. Do Plans Matter? The Effects of Light Rail Plans on Land Values in Station Areas (2001). Journal of Planning Education and Research 21:32‐39.

The ultimate goal of this study is to determine whether automobile usage will be significantly reduced by TOD, given the variety of activities afforded consumers, the freedom developers and business owners have to choose store locations, and the constraints on public investments. The authors use data on land sales in Washington County, Oregon, which contains the Western corridor of the Portland metropolitan area. By focusing our analysis on this corridor of the metropolitan area, we seek to evaluate whether the information in plans is capitalized into land values and thus plays a role in altering development patterns. The relationships between land values and most of the independent variables are highly significant and conform with expectations. As in most studies, land values per acre decrease with parcel size, reflecting diminishing marginal utility of lot size or economies of scale in the subdivision process. Land values also decrease with distance from a public park or open space, the property tax rate, and with distance from downtown Portland. Interestingly, the effects of density zoning are insignificant. Land values are lower for parcels located in a floodplain and for parcels adjoining a major or minor road. Land values increase with time, per pupil expenditure by the pertinent school district, and the median housing value of the pertinent census tract. Finally, land values increased on average approximately 11 percent per year.

Kohn, Harold M. “Factors Affecting Urban Transit Ridership.” Paper presented at the Canadian Transportation Research Forum Conference, 2000.

The author uses a study of 85 Canadian transit companies to determine the importance of fares, population size, and service variables as predictors of transit ridership. The author collects data for 85 transit agencies covering the period 1992 to 1998. He then tests alternate multivariate models to arrive at the best model to predict transit ridership. The author finds that the best predictors of transit ridership (R square = 0.97, F = 7190) are average fare and vehicle revenue hours. The author leaves unexamined many external factors (urban density, urban area size, socioeconomic characteristics) that might be associated with both of his explanatory variables.

Kuzmyak, J. Richard, Richard Pratt, G. Bruce Douglas, and Frank Spielberg. Land Use and Site Design. Transit Coopera tive Research Program Report 95, Chapter 15. Washington, DC: Transportation Research Board, National Research Council, 2003.

This chapter is part of a larger study of traveler responses to transportation system changes. This chapter examines traveler responses to various dimensions of land use and site design. The report presents a compilation of empirical studies on the topic.

The authors report that transit ridership tends to be higher at higher densities. Citing work by Parsons Brinckerhoff, et al (1996) for Chicago, they report that a 10 percent increase in residential density is correlated with an 11 percent increase in per‐capita transit trips and a 13 percent increase in transit mode share. Citing work by Levinson and Kumar for a national study of the U.S., the authors report that density only becomes relevant to mode choice at densities higher than 7,500 persons per square mile. Citing work by Frank and Pivo (1994) in Seattle, the authors note that transit requires workplace densities of 50‐75 employees per gross acre and residential densities of 10‐15 dwelling unit per net residential acre to achieve significant commute mode shifts. Citing a study by Nelson/Nygaard (1995) for Portland, Oregon, the authors note that housing density and employment density accounted for 93 percent of the variation in daily transit trip productions and attractions across the region. The authors also present the results of studies indicating that transit use tends to be higher in areas characterized by mixed land uses. However, the authors caution that many of these environments tend to also be characterized by higher densities, so separating the mixed use effect from the density effect is difficult. Citing work by Messenger and Ewing (1996) in Florida, the authors note that more balanced (jobs and workers) areas tend to have higher transit mode share. Citing a study by Cervero (1989) for 57 suburban activity centers, the authors note that centers with on‐site housing had 3 to 5 percent more transit, bike, and walk trips. Finally, in terms of the influence of site design, the authors note that in more transit and pedestrian friendly environments transit use tends to be higher. The authors cite studies by Cervero (1988, 1989, 1991), Cambridge Systematics (1994), Comsis (1994), and Hooper (1989) that show modestly higher transit mode shares in areas that are characterized by a more pedestrian and transit friendly environment. However, many of these environments also tend to possess higher densities and mixed uses, so isolating the effects of design can be difficult. The authors caution that in many of these studies self‐selection bias may be a concern, particularly in studies of residential uses.

Kyte, Michael. Measuring Change in Public Transportation Usage: an Analysis of the Factors Influencing Transit. University Microfilms International, 1986.

The focus of this research is the development of a methodology for analyzing changes in public transportation usage over time. The methodology includes three elements: (1) the development of a set of models that relate transit demand to level of service, cost, and market size, (2) assessment of the impacts of past service and fare changes, and (3) forecasting the effects of future service and fare changes on transit ridership. The statistical methodology used here is the time‐series analysis and modeling approach of Box and Jenkins. This methodology was applied to data describing transit usage in Portland, Oregon from 1971 through 1982. Three levels of data aggregation were used: system level, sector level, and route level. Five different classes of time‐series models were developed. The following conclusions can be drawn from this research: (1) Service level, cost, and market size adequately explained both past and future variations in transit ridership. The effects of service level and fare changes on transit ridership are not instantaneous but are delayed and distributed over specific periods of time. (2) The models were consistent, in terms of lag structure and elasticities, among the three data aggregation levels. (3) Impact analysis using intervention models provided an assessment of nine system‐wide events and 78 individual route 89

level service changes. This research represents an important extension of previous work in this area. The use of three different aggregations of data has yielded important perspectives on the relative effectiveness of system vs. route level models. Lag structures have been more clearly identified here than in any previous study. In addition, the study of all service and fare changes implemented in Portland between 1971 and 1982 has provided important information on how elasticities can vary over time and according to the specific situation of a given change.

Liu, Zhi. Determinants of Public Transit Ridership: Analysis of Post‐World War II Trends and Evaluation of Alternative Networks. Cambridge, MA: Harvard University, 1993.

The author estimates a series of multivariate models to explain transit ridership in Portland, Oregon between 1950 and 1990. The author estimates multivariate models that predict transit ridership (trips per capita) as a function of passenger car registrations, per capita transit subsidies, percent of population in the central city, city population, gasoline price, passenger fare, MSA employment, transit vehicle miles of service, and a time trend variable. Variables are entered in their log forms. The author’s key finding is that income, passenger car registrations, and central city population all have strong effects on ridership. The automobile variable is problematic, however, in that the total number of vehicle registrations variable does not tell us anything about the level of household vehicle ownership, in particular the number of households that do not own an automobile. This variable has been found to be a strong predictor of transit ridership. The author uses insights from the analysis to predict the likely ridership results of individual variable trends on ridership.

The authors examine travel behavior, attitudes, and other individual characteristics at transit‐ oriented residential developments along the Gold Line light rail transit line in Los Angeles. The authors observe a boom in transit‐oriented development activity but lower than expected ridership (one half of forecast). The authors survey all residents in 37 multi‐family buildings located within 1/3 mile of rail stations. Of 1,595 housing units surveyed, they obtained responses from 221 units recording information about 477 trips. The authors interviewed ten developers and five property managers. The authors gathered neighborhood population and housing profile data from the U.S. Census Bureau. The authors also conducted site visits to assess the local pedestrian environment. The authors found few transit‐dependent residents in their survey. Respondents were primarily white, worked in professional occupations, and owned one or more automobiles. Few residents had low incomes. About 75 percent of respondents rarely or never used transit, while 15 percent regularly used transit. The authors noted that respondents were more frequent transit users after they moved to their current place of residence, but noted that there might be a self‐selection bias at work.

The interviews with developers and property managers elicited a widespread sense that having their property near the transit line led to a rent and/or market value premium. However, there is also significant demand for housing in these communities, so the effect of location cannot be isolated from these larger market forces.

McLeod, Malcolm, Kevin Flannelly, Laura Flannelly, and Robert Behnke. “Multivariate Time‐ Series Model of Transit Ridership Based on Historical, Aggregate Data: The Past, Present, and Future of Honolulu.” Transportation Research Record 1297 (1991): 76‐84.

The authors estimate multivariate models to determine the principal influences on transit ridership in Honolulu between 1956 and 1984. The authors develop a multivariate model that predicts transit ridership (revenue trips) as a function of the number of civilian jobs, per capita income, fare, the number of buses, and a dummy variable identifying years in which a strike occurred. All but the last variable are transformed into their natural log forms. The authors then estimated a similar model that substituted linked passenger trips as the dependent variable. The multivariate models explained more than 97 percent of the variation in transit ridership over the study period. However, there are some cautions. The income variable is at best an imperfect gauge of either overall regional economic activity or individual household welfare. The service variable (number of buses) is not the most desirable means of tracking service – more appropriate would be to use vehicle hours or vehicle miles.

Meyer, John and Jose Gomez‐Ibanez. Autos, Transit, and Cities. Cambridge, MA: Harvard University Press, 1981.

The authors argue that political decisions have resulted in the redistribution of transit service from core areas to low‐density suburbs. The consequence has been a decline in service productivity, as measured on a cost per unit of service basis. The authors attribute these political decisions to a combination of a desire to broaden the political base for mass transportation subsidies and a sincere belief in the social benefits of these services. The authors’ work is typical of a large body of literature calling for privatization as the only way to avoid these kinds of policy decisions.

Meyer, John, John F. Kain, and Martin Wohl. The Urban Transportation Problem. Cambridge, MA: Harvard University Press, 1965.

In this classic work, the authors document decentralization of various populations in all sizes of metropolitan areas in the United States over the course of several decades. They also document the declining importance of transit in urban regions and attribute the decline to decentralization. They argue that transit performs best where it links high density suburbs to large and dense central business districts, both of which are environments that are in relative decline in almost all metropolitan areas. The authors do not address the question of whether fixed route transit can serve other types of markets but implicitly assume that it cannot.

The paper examines the Metro Green Line in Los Angeles, which has been criticized for being a ‘Line to Nowhere’. The authors address criticisms that the Green Line does not connect major activity centers and was not likely to generate sufficient ridership to justify the investment by noting that, in fact, it serves major employment and carries more riders than the critics would expect. One important reason is the line’s important role as a connector to both the Blue Line and bus lines in its service area. The authors use internal agency ridership numbers to track the growth of ridership on the Green Line versus the other rail lines operated by the Los Angeles County Metropolitan Transportation Authority (MTA). Ridership on the Green Line increased from 13,650 average weekday boardings in 1996 to 37,487 a decade later, an average annual growth rate of 12 percent. The authors also collect line‐by‐line bus route ridership and station boardings‐and‐ alightings to illustrate the important role that bus‐to‐rail and rail‐to‐bus transfers are playing in increasing Green Line ridership. The authors find that the many Green Line riders use the line as a feeder to the Blue Line, which provides service between downtown Los Angeles and Long Beach, or as a trunk line fed by the strong arterial bus routes that cross the Green Line. The authors conclude that the Green Line is succeeding by serving non‐traditional transit markets.

Mierzejewski, Edward and William Ball. 1990. “New Findings on Factors Related to Transit Use.” ITE Journal (February 1990): 34‐39.

The authors identify the choice factors that affect individuals’ decisions to use transit. The authors conducted a telephone survey of 4,000 persons in 17 selected MSAs who had public transportation available within one‐half mile of their homes. The authors found that the attractiveness of the automobile was the primary deterrent to transit use, although 22 percent of respondents reported that their place of employment was not served by transit. The survey results also confirm the traditional view that CBD‐bound commuters are an important transit market. Of the choice riders, 82 percent worked in the central city and the majority of them listed parking availability as the main reason for using transit.

The purpose of the study was to define the concept of transit‐oriented development, identify its potential benefits, identify barriers to its widespread implementation, document what appears to be working well, and develop strategies to promote more widespread use of the concept. The underpinning of the review is a set of 12 detailed case studies of transit‐oriented developments in the state. For each development, the authors obtained land use, socioeconomic, and travel

data as well as information about TOD‐supportive public policies and records of development activity. The report encompasses descriptions of each TOD site (which were used to build a web‐ based database) and recommendations about policy that should promote more use of TOD. The authors rely on descriptive statistics to make their case that TOD sites have higher transit ridership, but there is no attempt made to control for other potential influences. The authors distinguish between the types of transit available at each site, but they do not discuss larger service structure issues.

Parsons Brincke rhoff Quade & Douglas, Inc. Transit and Urban Form. Transit Cooperative Research Program Report 10, Chapters 1 and 2. Washington, DC: Transportation Research Board, National Research Council, 1996.

This report examines the relationship between urban form (which consists of urban structure, density, land use mix, and land use design) and transit ridership. The report is essentially a literature review compilation of an extensive set of prior empirical work on the topic. The authors note a number of key findings from their own and other research: From their own study of 17 cities with light rail and/or commuter rail, the authors report that residential densities have a strong influence on rail transit boardings and that CBD size and density is also a strong influence on rail ridership. From their own study of Chicago and San Francisco, the authors note that residents of higher density residential areas are more likely to walk to access transit. From their own study of Chicago and San Francisco, the authors note that residents of more traditional (pre‐1950s) neighborhoods are more likely to use non‐automobile modes than residents of suburban (post‐1950s) neighborhoods. The authors also report extensively from other literature on the link between the CBD and transit ridership, the roles of employment clusters (other than the CBD) as ridership attractors, the importance of higher residential and employment density in correlating with higher transit ridership and/or mode shares, and the roles of land use mix and design in enabling transit to be a more viable mode for trips that might otherwise be undertaken by automobile. However, the authors note that density is often correlated with land use mix and design, and that separating the effects of these factors from the effects of density is often quite difficult.

Pisarski, Alan. Commuting in America II. Washington, D.C.: Eno Foundation, 1996.

The author provides a portrait of commute travel in the United States using data obtained from the 1990 Census. The author points to the decentralization of population and employment in U.S. metropolitan areas as a primary cause of the decline in transit mode share. The report implies that transit is tied to a traditional, mono‐centric urban form, and that, as this urban form disappears, transit will decline. But there are exceptions, as the author notes in the cases of Orlando, Tampa, Phoenix, San Diego, Houston, and Los Angeles.

Post, Robert C. Urban Mass Transit: The Life Story of a Technology. Westport, Connecticut, 2007.

Post, Curator Emeritus of Transportation at the Smithsonian Institution, accounts for the beginnings of the U.S. mass transit industry, its choice of technology throughout its history, its 93

decline as private enterprise, and its rebirth and technological choices as public enterprise. In accounting for the transition from streetcar to bus, Post examines both the G.M. conspiracy as well as the relative economics of the two modes as the bus evolved technologically at the end of the 1930s. Post concludes that performance and economics favored the bus over even the modern PCC streetcar, which he none‐the‐less saw as one of the greatest technological innovations of the U.S. transit industry.

Pratt, Richard and John Evans. Bus Routing and Coverage. Transit Cooperative Research Program Report 95, Chapter 10. Washington, DC: Transportation Research Board, National Research Council, 2004.

This chapter is part of a larger study of traveler responses to transportation system changes. This chapter examines rider responses to changes in bus transit routing. These changes include: new bus systems and system closures, bus system expansion and contraction, changes in geographic coverage, and routing and coverage changes that might be made in tandem with fare changes. The authors provide an overview of literature on the topic from the 1970s to the end of the 1990s, and report elasticities of ridership with respect to each of the routing and coverage changes. The authors also provide more detailed case studies for several cities. The authors found elasticities in the range of 0.6 to 1.0. The authors noted that the largest ridership increases occurred when the system emphasized “high service level core routes, consistency in scheduling, enhancement of direct travel and ease of transferring” (5). The authors claim that new and expanded systems of the hub‐and‐spoke variety produced slightly higher ridership than grid systems, although there were no controls for other possible variables.

Pucher, John. “Renaissance for Public Transport in the United States?” Transportation Quarterly 56, no. 1 (2002): 33‐49.

During the mid and late 1990s, a series of articles appeared documenting a large decline in transit ridership during the early part of the decade. This study examines ridership over the entirety of the decade to identify trends and possible causes for those trends. The author collects data from the American Public Transportation Association and the National Transit Database, including unlinked passenger trips, vehicle miles of service, fares, and subsidies, and data from the Census Bureau and Bureau of Labor Statistics, including population and employment statistics. The author uses these data to prevent a descriptive account of transit ridership trends. The author emphasizes the crucial role played by transit ridership in the New York metropolitan area in driving national transit statistics. He identifies the economic recession of the early 1990s, and particularly its effect on employment in New York, as the driving force behind the ridership decline of the early 1990s. He cites the economic recovery of the 1990s, rising gasoline prices, stable fares, improved service quality, and the expansion of rail transit services as among the key contributing factors for the ridership rebound of the latter part of the decade. The limitation of this article is that it is purely descriptive; it makes no effort to examine the ridership trend and its potential causes using more sophisticated multivariate techniques.

Pucher, John and John Renne. “Socioeconomics of Urban Travel: Evidence from the 2001 NHTS.” Transportation Quarterly 57, no. 3 (2003): 49‐78.

The authors analyze the results of the 2001 National Household Travel Survey to document urban travel trends and differences in travel behavior among different socio‐economic groups. The authors extract descriptive tables from the 2001 NHTS to identify differences in travel behavior based on geography, income level, auto ownership, race, and ethnicity of individual travelers. The authors document a continued decline in transit use and corresponding growth in vehicle travel. The authors find that the poor, blacks, Hispanics, and those with low levels of vehicle ownership are more likely to use transit than are other groups.

Pucher, John and Christian Lefevre. The Urban Transport Crisis in Europe and North America. London: MacMillan Press Ltd., 1996. Jones (2008) ploughed the same ground, but much more finely, as did Pucher and Lefevre did a decade earlier in this pioneering work. Pucher and Lefevre compare transit and auto policy and their consequences in the U.S., Canada, and various western and eastern European countries over a period of a number of years, attempting to use comparable metrics in the various countries. They found that Europe, particularly eastern Europe, was motorizing much more rapidly than anticipated, and as various countries motorized, their land uses were decentralizing, as well. Pucher and Lefevre questioned, given enough time, whether European transportation and land uses would come to resemble those in the U.S. rather than the other way around? Despite this question, they did see that European policy, particularly in taxation of auto travel, the absence of inner city motorways, the excellent transit systems (in western but not eastern Europe), and the higher density, planned suburban development, would have the effect of keeping Europe less motorized than in the U.S., even under high levels of personal income.

Pushkarev, B. and J. Zupan. Public Transportation and Land Use Policy. Bloomington, IN: Indiana University Press, 1977.

This book examines the relationship between transit service supply, transit demand, and urban density. It is based on an earlier study prepared for the Regional Plan Association. The key insights, from the perspective of transit ridership and system performance, are that transit use is higher at higher urban densities. The authors also point out that auto ownership is lower (even when controlling for income) at higher densities.

Schlossberg, Marc A and Nathaniel Brown. “Comparing Transit Oriented Developments Based on Walkability Indicators”. Transportation Research Record 1887 (2004): 34‐42.

This article uses twelve GIS based walkability variables to evaluate and compare eleven TOD sites in Portland, Oregon.The researchers used three primary techniques in comparing the TOD sites: 1) Network classification. This reflects the road network that is effectively available to the pedestrian after the heavily automobile trafficked streets have been removed.

2) Pedestrian catchment areas. These represent the actual areas that can be walked within a five minute (quarter mile) or ten minute (half mile) time. Two different versions of these are used; the second maps the distance that can be walked excluding high‐speed, high‐volume roads. 3) Impedand‐based intersection intensities. These reflect the how many intersections are available to the pedestrian. Based on these methods, six variables were used to rank the TODs: quantity of accessible paths (high/low) quantity of impedance paths (high/low) PCA ranking (good/poor) IPCA ranking (good/poor) intersection density (high/low) density of dead ends (high/low)

The authors conclude that the pedestrian infrastructure can vary greatly from one TOD to another, even within the same city. GIS maps of the most accessible TOD (Gresham Central Transit Center) and the least accessible (Beaverton Transit Center) provide a visual representation of the pedestrian environment. The authors note that a clearer picture could be obtained if detailed data sets reflecting the more various types of sidewalks, intersections and roads encountered in real life were available.

Skinner, Jon. Elderly and Youth Bus Ridership: A Comparison of Routes in Miami‐Dade County. Master’s degree paper, Department of Urban and Regional Planning, Florida State University, 2007.

The author examines the performance of routes classified on the basis of the percentage of elderly or youth riders in order to distinguish routes with disproportionate numbers of these riders from other routes. He then examines these routes in terms of ridership, productivity, and the extent to which the route meanders. He finds that routes with high percentage of elderly riders have lower ridership and poorer performance than other routes and also tend to characterized by significant route meandering that is indicative of service that diverts from arterials to serve neighborhoods and provide more ‘front door’ type service. These routes tend to repel both elderly and non‐elderly patrons. The author notes that larger numbers of elderly patrons actually use the traditional routes that provide more direct service along arterials. The preferences of elderly patrons tend to be a lot like other transit users —they, too, value more direct and higher speed service.

Song, Yan. Smart Growth and Urban Development Pattern: A Comparative Study. International Regional Science Review 28, 2: 239–265 (April 2005).

This article provides a comparative study to examine the influences of the smart growth instruments on urban development patterns in Portland, Oregon; Orange County, Florida; and Montgomery County, Maryland. The article examines the impacts that various planning frameworks have on five dimensions of compact and traditional development: street network connectivity, density, land use mix, access, and pedestrian walkability.

The authors find that neighborhoods are becoming better internally connected in all five counties—especially after the late 1980s or the early 1990s—but also less externally connected. Neighborhoods have been developed at higher density in all five counties since as early as the late 1970s or the 1980s. Single‐family homes have been developed similarly in smaller lots and larger homes across three regions. A mixture of land uses within the residential neighborhoods appears to be absent in all five counties, distances from single family houses to commercial stores and transit appears to be increasing in all counties, and pedestrian accessibility (except Multnomah and Washington County, Oregon) to commercial land uses and bus stops appear to be falling over the study period. These results suggest that single‐family residential neighborhoods across three regions remain relatively homogeneous in land uses that commercial uses remain separated, and that neighborhoods remain isolated from transit.

Song, Yan and Gerrit‐Jan Knaap. Measuring the Effects of Mixed Land Uses on Housing Values. Regional Science and Urban Economics 34 (2004): 663‐680.

This article examines public preferences for mixed land uses in a hedonic price model study of single family home purchases in Washington County, Oregon. At the time of their study Washington County was the fastest growing part of the Portland metropolitan area and featured a wide variety of housing choices, including mixed use transit oriented development built around light rail stops. Song and Knaap found that single family home buyers valued locations with amenities such as parks, open space, and the ability to walk to neighborhood retail. On the other hand, such buyers discounted proximity to multi‐family dwellings, large employment concentrations, and properties with small lot sizes.

The authors examine the relationships between both residential densities and income on transit ridership in Denver, Portland, Salt Lake City, San Diego, and Seattle. The authors obtain data on per capita transit use, total population, annual income, and geographic acreage from the 1980 U.S. census and local data sources. They then estimate multivariate models that predict transit ridership at the neighborhood level. The authors find a strong density effect, however the effect varies depending on the income of the neighborhood. Density appears to have a stronger effect in lower income neighborhoods.

Taylor, Brian D. “Unjust Equity: An Examination of California’s Transportation Development Act.” Transportation Research Record 1297 (1991): 85‐92.

This paper examines the consequences of California’s Transportation Development Act, which provided dedicated transit funding for all counties, on subsidy and performance of urban versus

suburban operators. The author compares suburban and center‐city operators on a number of performance dimensions. The author argues that the allocation formulas of the Act have strongly favored lightly‐ patronized suburban service over more heavily‐patronized urban services. The result has been a proliferation of new, well‐funded, and expanding suburban operators that attract few riders while older, more heavily‐patronized central city operators are forced to cut service because of funding shortfalls. The author calls for a redirection of subsidy to central city operators. This recommendation is in line with the traditional view that transit should focus on serving a CBD and central city market.

The authors identify and analyze strategies used by transit agencies that enjoyed ridership increases between 1995 and 1999. The authors conducted a survey of 103 agencies and learned that a majority had expanded services, restructured routes, and developed new marketing strategies, including promotion of partnerships with universities, large employers, and other major activity centers. Surveyed agencies also cited the importance of population growth and economic conditions as factors that strongly influenced transit ridership. The authors followed the initial survey with more detailed case studies of 12 systems. These case studies revealed that among the most important internal policy initiatives undertaken were: fare restructuring, coordination with employers, and route restructuring. Route restructuring included elimination of low‐productivity routes, suburb‐to‐suburb commuter services, and the introduction of specialized services (welfare‐to‐work transportation, medical transportation). The authors conclude that while many factors that affect transit ridership are beyond the control of agencies, creative managers can still employ a combination of strategies and enjoy positive results.

The authors investigate the factors that explain transit ridership in 265 urbanized areas in the year 2000. The authors estimate a two‐stage, least squares regression model that predicts transit ridership as a function of regional geographic characteristics (population, population density), metropolitan economic characteristics (household income, housing prices), population characteristics (race, age, percent below poverty), auto and highway characteristics (fuel prices, percent carless households), and transit system characteristics (fares, coverage, frequency). The authors find that the most important determinants of transit ridership variability among the urbanized areas are: metropolitan area size, median housing costs, and percent of households that do not own an automobile. They also find that transit service levels and fares are also associated with ridership, with elasticities generally within ranges cited in the literature.

The authors identify and examine the key determinants of transit ridership change between 1990 and 2000 in U.S. metropolitan statistical areas (MSA) with more than 500,000 persons. Among the key variables they examine is a service orientation that distinguishes between multidestination and traditional service orientations. The authors obtained data from the US Census Bureau, US Bureau of Labor Statistics, and National Transit Database. They estimated multivariate models for the percent change in ridership (passenger miles per capita) between 1990 and 2000 for three different MSA groups: all MSAs, medium MSAs (1 million to 5 million persons), and small MSAs (500,000 to 1 million persons). The explanatory variables included: 1) 1990 passenger miles per capita; 2) percent change in urbanized area density; 3) West region (dummy variable); 4) change in ratio of rail service to total service; 5) percent change in service frequency; 6) percent change in service coverage; 7) percent change in MSA population; 8) percent change in unemployment rate; 9) percent change in black population share; 10) percent change in Hispanic population share; and 11) multidestination service orientation (dummy variable). The authors found that transit is growing most rapidly in the non‐traditional markets of the West but that much of the regional variation is a function of the particular service coverage, frequency, and orientation decisions made by transit agencies in this region. Service coverage and frequency are the most powerful explanatory variables for variation in ridership change among MSAs with 1 million to 5 million people, while a multidestination service orientation is the most important explanation for variation in ridership change among MSAs with 500,000 to 1 million people. A weakness of the analysis is the definition of the multidestination variable as a binary variable, as opposed to a continuous one.

Thompson, Gregory, Jeffrey Brown, Rupa Sharma, and Samuel Scheib. “Where Transit Use is Growing: Surprising Results.” Journal of Public Transportation 9, no. 2 (2006): 25‐43.

Using data obtained from the National Transit Database, the authors identify the kinds of metropolitan areas where transit ridership increased from 1990 to 2000. The authors report descriptive statistics for ridership, service, and service productivity by Census region and MSA population size class to identify places where transit use is growing. This paper essentially investigates whether transit’s fate is tied to the last vestiges of old urban forms or whether transit is finding niches in the new, largely suburban urban forms that increasingly have manifested themselves since the 1920s. The hypothesis is that most growth is in census regions with the strongest vestiges of older urban forms centered on CBDs. The method to test the hypothesis is to document how transit performance changed between 1990 and 2000 in U.S. metropolitan areas with more than 500,000 people in the year 2000. The results show that for MSAs with fewer than 5 million people, transit use has been growing faster than very rapid population growth in the West region, but not elsewhere in the country. The conclusion is that transit growth is not tied to old urban forms.

Thompson, Gregory L. and T. G. Matoff. “Keeping Up with the Joneses: Planning for Transit in Decentralizing Regions.” Journal of the American Planning Association 69, no. 3 (2003): 296‐312.

The authors investigate the relationship between service orientation and transit system performance using comparative case studies of transit systems in decentralized metropolitan areas that have pursued multidestination versus radial service approaches. The authors obtained data on transit system profiles and transit performance from 1983 to 1998 for transit systems in Cleveland, Columbus, Houston, Minneapolis, Pittsburgh, Portland, Sacramento, San Diego, and Seattle. The performance measures include: cost per passenger mile, peak‐to‐base ratio, passenger miles per capita, and vehicle miles per capita. The authors then compared systems that met their definitions of multidestination versus radial service orientations on each of these measures. The authors found that multidestination systems were more effective (higher ridership), nearly as efficient (about the same cost), and more equitable (lower peak‐to‐ base ratio) than radial systems.

TranSystems, Inc., Planners Collaborative, and Tom Crikelair Associates. Elements Needed to Create High Ridership Transit Systems. Transit Cooperative Research Program Report 111. Washington, DC: Transpor tation Research Board, National Research Council, 2007.

The report includes case studies that focus on the internal and external elements that contributed to successful ridership increases and examines how the transit agencies influenced or overcame internal and external challenges to increase ridership. The report is simply a list of the different strategies employed with no evaluation of performance of the strategy. Most strategies relate to fare policy or the development of services targeted at specific customer subgroups through marketing.

TRL Limited. The Demand for Public Transport: A Practical Guide. TRL Report TRL 593, 2004.

This study is an update of the 1980 report The Demand for Public Transport. The report presents the results of numerous studies on the factors influencing the demand for public transportation. The report is a compilation of numerous other studies. The report presents the study results as elasticities of ridership with respect to the specific set of factors that are discussed. The report includes chapters on fares, time (travel, access, and wait), other aspects of service quality, income , car ownership, and land use. Among the key findings are the following:

Urbitran Associates, Inc., Multisystems, Inc., SG Associates, Inc., and Robert Cervero. Guidelines for Enhancing Suburban Mobility Using Public Transportation. Transit Cooperative Research Program Report 55. Washington, DC: Transportation Research Board, National Research Council, 1999.

The authors seek to provide guidance to transit operators and local policymakers to enhance suburban mobility through traditional and non‐traditional transit services. The authors use an extensive literature review to develop categories of suburban land‐use environments and a typology of service strategies. They then conducted detailed case studies of 11 US and Canadian

100

transit operators to determine the kinds of strategies that appear to be most effective in suburban environments. The authors used their analyses to develop 12 key findings about transit in suburban environments that can serve as guidance to operators and local policymakers. Their recommendations include: 1) develop service around focal points; 2) operate along moderately dense suburban corridors; 3) continue to serve transit’s traditional demographics; 4) link suburban services to the regional line‐haul network; 5) target markets appropriately; 6) economize on expense; 7) adapt vehicle fleets to customer demand; 8) creatively adapt transit service practices to the landscape; 9) obtain private sector support; 10) plan with the community; 11) establish realistic goals; and 12) develop supportive policies, plans, and regulations, especially as pertains to land use and development policies.

This report is an update of TCRP Report 55 and is paired with a web‐only document that details the eight case studies that are briefly presented in the guidebook. The purpose of the study is to examine the current status of suburban transit, from both operations and land‐use perspectives, and to develop guidelines for evaluating, selecting, and implementing these services (1). The authors consulted literature as preparation for conducting 28 preliminary case studies scattered throughout the United States. The authors interviewed key agency contact persons at each of the systems, and collected information about the types of services offered, subsidy policies, the land use patterns of the area in which the agency operated, and information about the policy objectives underlying the specific types of services that are offered. The authors used insights gained from these preliminary case studies to develop a procedure for gaining more detailed information about eight systems that offer a range of suburban services and are located throughout the United States. The authors used the information obtained from both the preliminary and detailed case studies to develop a set of lessons about suburban transit services. They found: 1) the best performing services (measured in passengers per hour) are among the least flexible; 2) the best performing routes are among those serving the most balanced mix of land uses; and 3) services that target specific groups (such as seniors or students) seem to be among the most productive. The authors call for additional research on suburban alternatives to traditional fixed‐route transit service.

Vuchic, Vukan. Urban Transit: Operations, Planning, and Economics. Hoboken, NJ: John Wiley and Sons, 2005.

This is a textbook on public transportation. It includes discussions of transit system operations and networks, transit agency economics and organization, and transit systems planning and mode selection. The book’s discussions of transit users and transit network structures are most relevant to our examination. It offers discussions of factors that influence transit travel (especially level of auto ownership), and it differentiates between route structures and service 101

philosophies. The remainder of the book is more useful to practicing transit planners than to researchers.

Zhang, Shaoming. Feasibility Study on Transit‐Oriented Development, Using Urban‐Form and Non‐ Urban‐Form Variables. Paper presented at the 2005 ESRI International User Conference.

This paper suggests quantitative ways to measure urban form, which could be useful when considering where investment in TODs should take place. The author uses two linear regression models to calculate two variables that would come into play when evaluating how successful a transit joint development (TJD) would be: property value, and transit ridership. The independent variables used are average travel time, square footage, housing units, median number of rooms, level of land us mix, population density, average floor area ratio, median HH income, per capita income and year built. Calculations were made for areas surrounding stops along Atlanta’s MARTA. An improved version of Frank and Pivo’s Entropy Index was used to calculate the level of land‐use mix:

Level of land‐use mix = ‐A / LN (N) Where: A= Σ {b(n)/a * LN [b(n)/a]} With b(n)/a = proportion of building floor area of each land use among total square feet of all the land uses present in a buffer (when building floor area of one specific land use equals to 0, value of 0.01 will be given for its calculation); N = Number of land use categories used in the research.

Contrary to what one would normally expect, the author found that the amount of land‐use mix was actually negatively related to the number of workers using transit to travel. The research did confirm, however, that higher densities are positively related to transit ridership. Higher levels of both land‐use mix and population density resulted in increased property values.

102