MTI REPORT 01-19
LAND USE AND TRANSPORTATION ALTERNATIVES:
CONSTRAINT
OR EXPANSION OF
HOUSEHOLD CHOICE?
JUNE 2002
DR. JONATHAN
LEVINE
DR. ASEEM INAM
DR. RICHARD WERBEL
DR. GWO-WEI TORNG
a publication of the
Mineta Transportation Institute
College of Business
San José State University
San Jose, CA 95192-0219
Created by Congress in 1991
FHWA/CA/OR-2001-31
executive summary ...................... 1
chapter one: introduction--AN ALTERNATIVE RATIONALE FOR LAND USE TRANSPORTATION POLICIES........... 3
CHAPTER two: ACCESSIBILITY-BASED LAND USE FORMS..........7
CHAPTER THREE: NEIGHBORHOOD FORMS AND CHARACTERISTICS .................... 17
CHAPTER FOUR: STUDY METHODOLOGY ........23
CHAPTER FIVE: LAND USE AND CHARACTERISTICS OF METROPOLITAN BOSTON AND METROPOLITAN aTLANTA ...... 35
CHAPTER SIX:cASE STUDIES OF NEIGHBORHOOD CLUSTERS .................... 53
CHAPTER SEVEN: NEIGHBORHOOD PREFERENCES AND NEIGHBORHOOD CHOICES ........91
CHAPTER EIGHT: CONCLUSIONS ......107
APPENDIX a: SURVEY QUESTIONNAIRE ......111
APPENDIX b: kEY MAPS .................. 121
BIBLIOGRAPHY .................. 123
Table 1: Examples of Accessibility-Based Land Use Innovations 10
Table 2: Sources for Geo-Referenced Data 24
Table 3: Land Use Categories Used for Land Use Mix Measures 28
Table 4: Variables Used to Characterize Neighborhoods 29
Table 5: Examples of Tradeoff-Styled Survey Questions 33
Table 6: Outcome of Attempted Survey Interviews 34
Table 7: Calculation of Weighting Factors 34
Table 8: Average Values of Variables in the Five Clusters 49
Table 9: Households by Cluster, Atlanta and Boston 52
Table 10: Percent of Respondents Reporting Moderate or Strong
Preferences for Various Neighborhood Options 96
Table 11: Proportion Preferring Pedestrian- and Transit-Oriented
Neighborhood over Auto-Oriented Neighborhood 97
Table 12: Factor Loadings of Transit/Pedestrian Preference Factor 99
Table 13: Multinomial Logit Model of Choice by Neighborhood Type 105
Figure 1: Population Density on Residential Land by TAZ, Atlanta and
Figure 2: Employment Density by TAZ, Atlanta and Boston, 1995 37
Figure 3: Percent of Three-Way Intersections, Atlanta and Boston 38
Figure 4: Intersection Density by TAZ, Atlanta and Boston 39
Figure 5: Average Peak Hour Speeds of Major Roads by TAZ, Atlanta
Figure 6: Average Number of Lanes on Major Roads by TAZ, Atlanta
Figure 7: Two-Mile Land Use Intensity, Atlanta and Boston 43
Figure 8: Two-Mile Land Use Variety, Atlanta and Boston 44
Figure 9: Employment Accessibility, Atlanta and Boston 45
Figure 10: Population Distribution of Employment Accessibility via Automobile, Atlanta and Boston 46
Figure 11: Population Distribution of Employment Accessibility via
Transit, Atlanta and Boston 47
Figure 12: Clustering Results for Boston 50
Figure 13: Clustering Results for Atlanta 51
Figure 14: Aerial Photograph, near Downtown Atlanta 54
Figure 15: Aerial Photograph, Georgia State University and Grady
Figure 16: Street Level, Linden Avenue, Atlanta 55
Figure 17: Aerial Photograph, Northeastern University and Prudential
Center Neighborhood, Boston 56
Figure 18: Westland Avenue, Boston 57
Figure 19: Huntington Avenue, Boston 57
Figure 20: Aerial Photograph, Castle Square Neighborhood, Boston 58
Figure 21: Castle Square Neighborhood, Boston 59
Figure 22: Aerial Photograph, Hwy. I-75 and I-85, Atlanta 60
Figure 23: Single-Family Home, Wycliff Road, Atlanta 60
Figure 24: Neighborhood Between I-75 and I-85, Atlanta 61
Figure 25: Virginia and Ponce de Leon Avenues, Atlanta 62
Figure 26: Ponce de Leon Avenue, Atlanta 62
Figure 27: Aerial Photograph, Columbia Road and Massachusetts
Figure 28: Edward Everett Square, Boston 64
Figure 29: East Cottage Street, Boston 64
Figure 30: Aerial Photograph, Cambridge Neighborhood, Boston 65
Figure 31: Intersection of Second and Charles Street, Cambridge 66
Figure 32: Aerial Photograph, Turner Stadium Neighborhood,
Figure 33: Oakland Avenue, Atlanta 67
Figure 34: McDonough Boulevard, Atlanta 68
Figure 35: Aerial Photograph, Virginia Highland Neighborhood,
Figure 36: North Highland Avenue, Atlanta 69
Figure 37: Amsterdam Avenue, Atlanta 70
Figure 38: Aerial Photograph, Brookline Neighborhood, Boston 71
Figure 39: Beacon at Winthrop Street, Brookline Neighborhood,
Figure 40: Hyslop Road anad Fisher Avenue, Boston 72
Figure 41: Aerial Photograph, Waltham Neighborhood, Boston 73
Figure 42: Barton Street, Boston 73
Figure 43: Newton and Main Streets, Boston 74
Figure 44: Aerial Photograph, Hapeville, Atlanta 75
Figure 45: Hapeville Neighborhood, Atlanta 75
Figure 46: Myrtle Road, Atlanta 76
Figure 47: Aerial Photograph, Tucker Neighborhood, Atlanta 77
Figure 48: Tucker Neighborhood, Atlanta 77
Figure 49: Tucker Neighborhood, Atlanta 78
Figure 50: Danvers Neighborhood, Boston 79
Figure 51: Danvers Neighborhood, Boston 79
Figure 52: Weston Neighborhood, Boston 80
Figure 53: Village Green and Common Near Weston, Boston 81
Figure 54: Kendal Green Rail Station, Weston, Boston 81
Figure 55: Aerial Photograph, Roswell Neighborhood, Atlanta 82
Figure 56: Single-Family Home, Roswell Neighborhood, Atlanta 83
Figure 57: Houze Road, Atlanta 83
Figure 58: Aerial Photograph, Union City, Atlanta 84
Figure 59: Stonewall Drive, Atlanta 85
Figure 60: MARTA Stop, Union City Neighborhood, Atlanta 85
Figure 61: State Route 117, Bolton Neighborhood, Boston 87
Figure 62: Rural Boulton Neighborhood, Boston 87
Figure 63: Bolton Office Park, Boston 88
Figure 64: Hudson Neighborhood, Boston 89
Figure 65: Hudson Neighborhood, Boston 89
Figure 66: Ostego Drive, Boston 90
Figure 67: Mean Pedestrian, Transit, Auto and Overall Ratings of Atlanta
and Boston Neighborhoods by their Residents 93
Figure 68: Neighborhood Preference Score Distributions, Atlanta and
Figure 69: Relationship of Transit-Pedestrial Preference to Residence in Transit-and Pedestrian-Friendly Zones 101
Figure 71: Atlanta Key Map 124
Transportation and land use research that considers such alternatives as New Urbanist development, jobs-housing balance, transit villages, or "smart growth" most typically tests the capacity of such physical forms to reduce vehicle miles traveled (VMT) or bring about other desired outcomes in the modification of travel behavior. Establishing such causality is broadly seen as a precondition for the urban planning interventions that are presumed to be necessary to bring these forms about. But such a view neglects the extent to which current interventions--notably zoning and transportation regulations--tend to preclude the development of such innovations in areas of high accessibility where they can potentially be of the greatest benefit.
Payoffs in VMT reduction, though desirable, are hardly the necessary precondition for the relaxation of such regulations. Instead, the increased land use and transportation choice that such liberalization can engender is self-justifying in that it allows households to forge a closer link between their land use and transportation preferences on the one hand and their actual choices on the other.
This framework is examined here through a comparison of two metropolitan areas: Boston, which offers its residents relatively rich opportunities for residence in transit- and pedestrian-friendly areas, and Atlanta, which offers many fewer such opportunities. The study is based on three principal components: A clustering of neighborhoods throughout each metropolitan area according to their transit and pedestrian characteristics; an urban design analysis of selected neighborhoods in each region; and a survey of 1,600 households regarding their preferences for neighborhood environments. The study concludes that while residents of Atlanta are considerably less interested in transit- and pedestrian-friendly neighborhoods than their Boston counterparts, the difference in preference is insufficient to explain the difference in the transit and pedestrian quality of the neighborhoods the two groups inhabit. The neighborhood choices of the Boston residents was, as a consequence, considerably more sensitive to their transportation and land use preferences than the choices of their Atlanta counterparts. By providing a greater range of neighborhood transit/pedestrian-friendly and automobile- oriented zones, Boston enabled its residents to forge a closer fit between preferences and choices.
Introduction: an alternative rationale for land use transportation policies
One of the most controversial issues in transportation policy currently is the implication of alternative land use practices for transportation outcomes. A number of researchers and planning practitioners have sought to establish a connection between innovative land use development practice and a reduced demand for automotive transport that would ostensibly justify such practice (Cervero and Kockelman 1997, Cervero 1996, Frank and Pivo 1994). Others have questioned, both theoretically and empirically, the capacity of land use policy to induce changes in vehicle miles traveled (VMT), congestion, or other relevant transportation outcomes (Giuliano 1993, Gordon, Richardson and Jun 1991; see reviews at Badoe and Miller 2000, Boarnet and Crane 2001). To the extent that uncertainties remain in the relationship between provision of alternative land use forms and travel behavior outcomes, some of these observers would question the planning rationale for such policy directions as New Urbanist development, job-housing balance, transit villages, "smart growth," and related initiatives.
This study examines the transportation policy implications of alternative land use practices, but from a different perspective. Providing for broad range of alternative development forms, including rich alternatives to single family, low-density, land use-segregated, auto-oriented suburban development, is seen here as desirable not so much for a capacity to reduce automobile dependence or to moderate the growth in VMT, however sought after those outcomes may be. Instead, the primary impact of such alternatives is seen in their potential to allow households to forge a closer link between their preferences for land use and transportation environments on the one hand, and their actual choices on the other. In many growing areas of the United States, little variation in physical development is provided for, with the vast majority of housing being located in zones that were laid out with automotive accessibility dominantly in mind. In areas such as these, individuals with preferences for pedestrian or transit-friendly neighborhoods--or those interested in affordable housing close to their work and non-work destinations--may be impeded in their ability to select the neighborhood environment that matches their preferences or needs. In contrast, where broader ranges of neighborhood types are allowed to develop, households should have a greater ability to sort themselves by their environmental preferences and needs for accessibility. One of the most desirable values in urban form is the access to choice: a choice of people, of jobs, of physical settings, of institutions such as schools or churches, of entertainment, and so forth. A diversity of behavioral and physical settings implies that it is easier for an individual or group to find one that is congenial to them, or to become enriched in new ways. An extension of this philosophy is the notion that a desirable region is one that affords a relatively close fit between the preferences of its residents on the one hand, and their actual choices on the other.
If municipal land use and transportation regulation constitutes a significant barrier to the development of accessible, transit- and pedestrian-friendly alternatives, the question of whether automotive travel is reduced by these land use forms is not the logical precondition to their development. Rather, such development is to be desired where it fits the needs and preferences of its current and potential future residents, including their preferences for land use and transportation environments. Reducing barriers to development of this kind is self-justifying on the basis of expansion of households' range of effective choice, with reduction in demand for automotive travel being a desirable--if uncertain--side effect. Scientific uncertainty regarding beneficial travel behavior impacts would not seem to be a reasonable basis for excluding such developments from areas where markets for them exist. In areas where markets do not support the provision of such alternatives, one hardly needs policy involvement to exclude them, as the absence of profits will tend to accomplish this much more effectively.
This study tests this framework with three principal elements. First, the entire territory of metropolitan Boston and metropolitan Atlanta was classified into zones, five in each area, ranging from the most accessible, and transit- and pedestrian-friendly zones to the most automobile-oriented zones. Zones were designed to be as comparable as possible between the two regions. Second, detailed on-site urban design case studies, ten neighborhoods in each metropolitan region, were developed for selected neighborhoods in each of the zones in the two regions. These served both as a kind of "ground truthing" to verify the classification of neighborhoods into zone, and to explore in greater detail the urban environmental characteristics of families of neighborhood types in each of the two metropolitan areas. Finally, a survey was conducted of a randomly selected eight hundred households in each region. The survey focused on people's stated preferences, regardless of where they actually lived, for residence in alternative neighborhood types. Surveys were geocoded and characteristics of survey respondents were analyzed jointly with characteristics of neighborhoods in which they live. The study suggests that while Atlanta residents were considerably less interested in transit- and pedestrian-oriented neighborhoods than the Bostonians; the latter achieved a closer fit between their preferences and their choices of land use and transportation environments. The study attributes this improved match to the wider distribution of neighborhood types their region offered to the residents of metropolitan Boston.
It is not suggested that this wider distribution of neighborhood types in metropolitan Boston is the product of conscious policy; rather the development of the two regions in different historical periods was the driving factor. But developing as it did largely in the post-World War II era, metropolitan Atlanta was subject to a number of self-imposed barriers to the development of denser, more accessible, and more mixed-use alternatives. These barriers come in a number of forms and include banks' lending practices, developers' inclinations to stay with demonstrably successful formulas, and opposition from local communities, amongst others. This study is principally interested in the barriers that regulatory policy itself may place in the way of such development. Of these, local land use regulation in the form of zoning and negotiated agreements may be the most significant. Many economists recognize that suburban zoning conflicts most frequently take the form of disagreements between residents who prefer low-density land uses and developers interested in building higher density uses (Bogart 1998, Fischel 1985); minimum lot zoning as fostering metropolitan sprawl (e.g., Pasha 1996), and more exclusivity than would arise in the absence of such regulations (e.g., Wheaton 1993). Yet following Tiebout (1956), many economists frequently view these effects in a relatively positive light, as the acknowledged capacity of zoning regulations to exclude on the basis of income is seen as fostering an efficient sorting of the population in terms of demand for public goods, leading to presumably more efficient, homogenous jurisdictional units than would otherwise arise.
The casual observer might distinguish these phenomena from other cases of governmental regulation, arguing that it is simply the wishes of the neighbors that lead to the exclusion of higher density development from neighborhoods. Yet clearly regulation is at play here; the neighbors--when successful--are only able to implement their desires for a low-density environment by harnessing the regulatory power that the state confers upon the city. Direct action to accomplish exclusion is generally illegal, though hardly unheard of; threatened or actual violence against minorities relocating into white communities would be an example. Exclusion without regulatory intervention can, in principle, be accomplished on the basis of contract law, as in the cases of deed restrictions on land. Yet since the normal operation of contract law depends on aggrieved parties bringing their cases to court,1 it is difficult to see a non-regulatory approach that would enjoy the ubiquity of enforcement that zoning does.
This study does not examine land use regulations directly. Instead, the notion that land use regulations can constrain choice is examined through a comparison of two areas chosen for their distinctiveness one from the other--a distinctiveness that is the product of the differing historical eras in which each developed. Metropolitan Boston was selected as an area offering its households relatively rich opportunities for residence in a range of neighborhood types, including low-density automobile oriented areas and transit- and pedestrian-friendly zones. Metropolitan Atlanta was selected as an area offering many fewer opportunities for residence in transit- and pedestrian-oriented areas, with a much greater share of its territory developed according to post-World War II automobile-oriented principles. The study tests the notion, largely unexplored in the planning and transportation literature, that residents of an area offering a great variety in neighborhood types--represented here by the Boston area--will exhibit a closer fit between their neighborhood preferences and their actual neighborhood choices than will residents of an area, such as Atlanta, that is developed in a more uniformly automobile- dependent manner. To the extent that this hypothesis is borne out, it would tend to support efforts to overcome regulatory and other barriers to the development of alternatives to low-density automobile-oriented neighborhoods independently of any proof of benefit in VMT reduction.
Accessibility-Based Land Use Forms
A fundamental understanding in the field of transportation demand analysis is the notion that the demand for travel is derived (Meyer and Miller 1984); that is, with few exceptions, people travel not for the pleasure of motion per se, but in order to access opportunities available at their destinations. This understanding was developed in order to establish the formal link between land use and transportation in order to enable modeling of transportation system characteristics on the basis of distributions of residential and non-residential land uses across the landscape, especially in metropolitan areas.
While this use of the "derived demand" concept has become well established in professional transportation practice, another implication of the framework has been routinely overlooked. The transportation professions have most traditionally seen themselves as ensuring the mobility--or often the automobility--of the populations they served. In some cases, however, policies seeking to improve mobility--such as extensive highway construction--may have induced outward movement of land uses ("sprawl") in such a way that travel distances grew. Where this happens, mobility improvements can potentially be associated with increasing total time and money costs of travel. If the "derived" nature of transportation demand is taken seriously, then a set of policies that increases the time and money cost of travel per destination would not be desirable, even if it reduced the time and money cost of travel per mile. This is because it would leave travelers with less time and money to spend at their destinations. Thus a conscientious application of the "derived" framework would lead to the conclusion that improvements in mobility per se--reductions in the time and money cost of transportation per mile--are not necessarily desirable. Rather it is improvements in accessibility--reductions in the time and money cost of travel per destination--which should be sought by transportation policy. Under this framework, mobility improvements are desired when they enhance accessibility. Efforts at improving mobility that detract from accessibility in the longer run (by inducing land use change and growing travel distances that increase the time and money cost of transportation) are to be avoided, based on the "derived demand" nature of transportation.
Clearly, measurement of accessibility improvements is more complex than gauging change in mobility (Handy and Niemeier 1997). While mobility is readily gauged with the transportation engineer's "level of service" measurements of highway flow (Transportation Research Board 1992), accessibility is observed through interactions between residences, businesses, and other destinations. Thus policies relating to accessibility cannot focus on transportation system characteristics alone, but require significant attention to the distribution of land uses. The guiding notion of accessibility-based policies is that land uses can be configured--whether through directed planning, market forces, or both--in such a way as to either increase or decrease the need for transportation in general, and automotive transportation in particular.
Implications of an accessibility-based--as opposed to a mobility-based--land use policy are far reaching. Mobility-based thinking frequently leads to use of the land use regulatory power to exclude or reduce the developers' desired density of construction in an area, so as to avoid overloading the local automotive network. But when this scenario is played out many times over throughout the development of a region, it can amount to a regulatory-based recipe for low-density, automobile-oriented growth. In contrast, accessibility-based land use policies seek to facilitate denser development in areas of high accessibility. This accessibility may take a number of forms, including proximity to a high quality transit system, to areas of high job concentrations, or to shopping or cultural destinations. Proximity here is a relative and flexible concept; at one extreme, it may imply walkable distances of up to one quarter of a mile or so. At the other extreme, territory lying, for example, ten miles from a major employment center may be viewed as relatively accessible if its development affords households alternatives to commuting much greater distances.
A family of accessibility-based land use approaches is described in Table 1. These approaches, which are distinct but not mutually exclusive, can be characterized by the type of travel they aspire to reduce--whether the work trip or the non-work trip--and by the scale at which they operate. A "local" scale refers to an option that can readily be portrayed through neighborhood or site plans; alternatives operating at a regional scale are those for which the dimension of regional accessibility--not readily observable at a strictly local scale--is paramount.
Local scale, non-work travel: "New urbanism" is a planning and development concept that draws inspiration from traditions of American town planning of the early 20th century (Calthorpe 1993). Village scale, mixed uses, and walkability are central elements of this approach, which tends to be local in scale (it can be readily represented as a neighborhood plan), oriented toward reducing the need for non-work travel. That is, New Urbanist developments are not necessarily located in the immediate vicinity of major job centers, but seek to offer residents, among other benefits, opportunities to reduce the need for automotive travel for shopping, social, and cultural destinations.
Local scale, work travel: Inspired by higher density development in Europe that clusters around transit stations, "transit villages" seek to offer mixed used environments within easy walking distance of high quality public transit (Bernick and Cervero 1997). Where New Urbanist densities tend to range between eight and fifteen residential units per acre, transit villages may reach twice those densities or more in a city-like environment. In contrast to New Urbanist development, their primary transportation orientation, based on their proximity to regional transit, is towards facilitating non-automotive commuting by their residents.
Regional scale, work travel: Job-housing balance is a concept with a long planning pedigree: the notion of co-location of jobs and affordable residences to facilitate commute reducing choices by households (Cervero 1996, Levine 1998). Under this analysis, the problem is that as a consequence of municipal regulatory practices, such as fiscal or exclusionary zoning, certain subregions of metropolitan areas are systematically rich in jobs but poor in affordable housing. These areas become magnets for commute trips that are presumably longer than those that would have held had ample supplies of housing not been zoned out.
Regional scale, non-work travel: In the United States, the principle of using accessibility--and particularly transit accessibility to guide the location of major regional facilities--is more observed in the breach than in practice. Other countries have developed this approach more fully. For example, under the Dutch ABC system of land use planning, the central government can withhold funding from developments that do not meet guidelines under which shops and offices are concentrated in "A" areas easily accessible to public transport. Areas accessible by car and transit are designated "B" and are earmarked for office development. Areas with only automotive access are designated "C"; their uses are restricted to transport or land-intensive activities, such as agriculture, leisure parks, etc. (Monzon and Echeverria 1997).
Table 1: Examples of Accessibility-Based Land Use Innovations
These families of approaches are referred to in this study as the "accessibility-based" land use forms. Despite their obvious differences, in transportation terms, they share a common underpinning: seeking to reduce the need for some automotive travel, rather than accommodating car use alone. While in some planning and transportation circles development of these options appears to be self-evidently desirable, they remain controversial in the broader research and policy communities. A number of observers have questioned the capacity of these options to produce the desired transportation outcomes, such as reduction in VMT or congestion. As a consequence, the planning interventions into market processes that are presumed to be required to bring these options to fruition are seen by some as lacking sufficient scientific basis. Where such uncertainty remains, many would argue, planning should leave outcomes to the market, and not focus on imposing accessibility-based development forms.
Those who have advocated accessibility-based development forms as a key element of transportation policy have generally accepted similar terms of debate. Like observers more skeptical of reliance on these land use approaches (e.g. Crane 1996, Giuliano and Small 1993), proponents (e.g., Cervero 1996, Frank, Stone and Bachman, 2000) have explored whether benefits in modification of people's travel behavior are established with sufficient scientific confidence to justify the planning interventions presumed to be required to bring these development forms about. The difference lies in the conclusion reached, not especially in the question asked. The divergent policy conclusions stem either from differing perceptions of the quality of the scientific evidence of the travel-reducing capacity of these development forms, or from normative differences in the evidentiary threshold required to justify planning interventions. The fundamental question of whether the transportation payoffs justify the planning interventions is shared by many of the parties of this debate.
In contrast, this study does not seek to resolve claims regarding travel behavior impacts of the accessibility-based development forms. Instead, we suggest that scientific evidence regarding a complex phenomenon that is not amenable to controlled experimentation, such as that of travel behavior impacts of alternative land use forms, is likely to remain mixed, and subject to conflicting interpretations. But this scientific uncertainty hardly precludes the alteration of exclusionary policies that maintain the status quo in terms of metropolitan development.
The perspective of the transportation and land use analysts contrasts markedly with the views of the economists described above. The former seem to be asking whether or not there is enough science to justify interventions into the land use market to bring about innovations including increased residential densities. The latter tend to accept that existing interventions, in the form of land use regulation, have the effect of keeping densities below market levels--but view that outcome positively on the basis of efficiencies of public good provision. This study seeks to merge the perspectives by acknowledging the limitations on denser (and by extension, transit- and pedestrian-friendly) development that current regulatory practices impose, but questioning the fundamental desirability of that regulatory outcome.
At first blush, the notion of policy and planning being employed to exclude accessibility-based alternatives may appear surprising. A conventional view seems to equate sprawling metropolitan forms with uncontrolled market forces, with the planning function seeking to encourage alternatives to sprawl. But anecdotal observations of the workings of the municipal planning function appear inconsistent with this view. In general, zoning ordinances limit densities or floor-area ratios to a given maximum, rather than setting a floor. In most areas, land use regulation still seeks to separate land uses, limiting mixing of housing with commercial uses, or even housing of different income levels or physical forms. Minimum lot size requirements are a particularly pervasive form of regulatory control in newly developing areas. Transportation regulations frequently specify wide street widths and minimum parking requirements. In other words, embedded in the regulations of scores of thousands of units of local government is a design template that is largely inimical to the accessibility-based innovations described above. Only when that regulatory template is relaxed can innovative development appear.
Examples of this phenomenon can be found in numerous reports from around the United States of developers seeking to build in a more compact, accessible, or mixed-use fashion than regulations allow, then having their designs rejected or modified through the planning process to conform to locally desired low-density patterns. Frequently the prescription from the planning authorities is to return with a plan for conventional single family development on the site in question. Several examples follow:
From The Tennessean, October 8, 1999
An unusual proposal to plant a mini-village on a country road west of Murfreesboro is dead. Murfreesboro planning commissioners voted Wednesday night to deny a zoning plan that would mix stores, offices, and homes on 250 country acres off Florence Road...Commissioners asked developer Roy Waldron to return with a zoning plan for single-family homes, city-planning commissioner Chris Bratcher said... The commission's decision effectively kills a proposal reminiscent of an increasingly popular form of planning. In this kind of planning, the developer creates a village by mixing stores with apartments and homes of various sizes on variously sized lots. Sometimes the village has a school, or a village green. For example, Walt Disney Co. built such a town in Florida in which every detail was meant to foster a sense of community...
From The Atlanta Constitution, October 8, 1999
"Smart growth"... means building higher density, mixed use developments closer into town and easily accessible by transit. So MARTA and BellSouth tried to do just that, planning a 50-acre complex of offices, residences and shops that would surround the Lindbergh MARTA station. Is everybody happy then? Nope... the Buckhead Neighborhood Planning Unit has voted 19-7 to reject the plan." (Authors' note: a scaled-back version of the transit village was ultimately approved, but a 39-story condominium complex was eliminated from it).
From the Albuquerque Journal, December 14, 1999:
Councilors rejected a plan Monday that would have allowed a high-density housing development, despite arguments the project fulfills a new city growth policy... Much of the discussion Monday centered on Resolution 70, a growth plan approved by the council in September that emphasizes infill development in established areas as a way of discouraging sprawl. The plan also calls for encouraging higher density housing along transportation corridors... Ronald Bohannan, a consultant to project developer Sean Gilligan, said the project fulfills the intent of R-70 by providing new high-density housing in an established area of the city along bustling Tramway Boulevard. Councilor Alan Armijo, the only councilor to express support for the development, said the city needs to begin approving decisions that promote infill development. "We're never going to solve the infill problem if we don't start somewhere," Armijo said... After the vote, Planning Commissioner Susan Johnson said the proposal's defeat shows the difficulties the city faces in implementing infill policies.
These stories describe the planning function being used to reject or limit New Urbanist-inspired development (in the Tennessee case), transit village construction (Atlanta), and conventional infill development (Albuquerque). In all cases, the agents of densification were not public planners, but developers seeking profits from a market they judged to support such dense construction. If these stories are representative, they suggest that the debate on the accessibility-based development forms has unwittingly turned the planning versus the market argument on its ear. In growing, highly accessible areas, it may be that the land development market tends towards greater density, accessibility and mixed-use development than planning regulations and practice allow. If this were the case, then analysis of the impact of urban form on travel behavior--while an interesting and valid scientific endeavor--would not be especially relevant as justification for the allowance of such alternatives.
None of this is to suggest that planning take a hands-off approach to development of the accessibility-based alternatives. Clearly, the planning function can be employed to designate and reserve areas for higher density or transit-oriented development. This function can be particularly useful in order to achieve critical mass in such areas; a laissez-faire approach would subject territory that might be appropriate for accessibility-based development to the vagaries of business cycles and the habits of a particular developer. But while planning can designate and enable such development, it requires private developers who see profits to be made to actually carry out the development. Planning can facilitate market forces that tend towards accessible development, but can hardly create such development when no market exists for it. An example is found in the following story:
From the Denver Rocky Mountain News, October 4, 1999
The city wants to build a "test" development project that includes a mix of residential housing with retail and commercial uses. The idea is called "new urbanism" because it harkens back to the old mixed use neighborhood... On Monday the city council gave the go-ahead to seek proposals from developers for a mixed use project. One developer has expressed an interest to do such a project.
Thus, the city can signal its interest, but then requires private developers who see profits to be made before the concept can be implemented. In this sense, the actions of the planning authorities are permissive in nature rather than command-and-control; if such development offered developers only subnormal profits, they would seek greener, and more profitable, pastures elsewhere. Where a city misjudges and designates a transit village or New Urbanist zone for which there is insufficient market, the land would presumably remain vacant until the profit situation changed, or the city relented with its designation.
In this case, the city apparently judged the potential for New Urbanist development correctly, based on a report a year and a half after the original:
From the Denver Post, June 10, 2001
Continuum Partners starts construction this week on a $220 million urban village in an unlikely place: an undeveloped field in Westminster. The Denver real estate developer hopes to carve a new pattern in the land of suburban sprawl. With its narrow streets, front porches, back alleys and urban-style townhouses, the 120-acre Bradburn development will bring a new look to suburban development... In late April, the Westminster City Council agreed to revamp its zoning rules to let Continuum build its brand of denser, more urban-style development at West 120th Avenue between Interstate 25 and U.S. 36.
It is notable that in order to implement its 1999 decision supporting such development in principle, the Westminister City Council needed in 2001 to relax restrictive zoning that would have prevented the development firm from building to its desired densities. In other words, successful development of a new urbanist neighborhood was contingent on liberalization, not tightening, of land use regulations.
The argument about planning regulations limiting innovation in land use development is hardly a new one. But it seems barely to have infiltrated the debate surrounding the accessibility-based development forms, whose legitimacy is still broadly construed to hinge on demonstrable travel behavior impacts. Only when the two issues are juxtaposed does the alternative rationale for accessibility-based development--one based in household choice, rather than in VMT reduction per se--become apparent.
Household Choice: From Developers and Planners to (Missing) Residents
When planning regulations exclude or limit accessibility-based development forms, they restrict the ability of those households that would have occupied such neighborhoods from getting what they want in a transportation and land use environment. Developers become successful by accurately judging markets; those who fail at this task too often will go out of business. Thus one can conceive of an action excluding a high-density transit village, for example, as the equivalent of denying several hundred households the opportunity to reside in what would have been their preferred residential environment. In municipal political processes, these households hardly constitute a potent political force, as they are likely not to be current residents of the community in question. This force is routinely overlooked in local debate with (usually long-time) residents over new types of residential development and in discussion about appropriate land use policies, geared as they are towards past patterns rather than future trends.
Moreover, these households are not likely to understand the process by which they had been excluded or even to identify themselves as excluded by governmental regulation. Instead, these households' perception of the processes involved will be filtered through the market and translated into unaffordable prices or rents for the kind of housing they might prefer. Households that would have occupied dense housing near a transit station--for example near the 39-story condominium tower that was excluded by regulation from the area of the Lindbergh MARTA station in Atlanta--might find its desired neighborhood unaffordable, but would probably not understand the process by which it was excluded from its first choice. Such a household, in all likelihood, would quietly opt instead for lower cost housing elsewhere; given the paucity of transit-oriented development, such location would probably be in automobile-oriented districts. Having now located in such an area, households may well find themselves opposing proposals for higher density development in their neighborhood, thus completing the systematic cycle of exclusion of denser, more accessible development forms. If the processes hypothesized here are highly influential, then households' choices could be constrained to the point that choices become a deteriorated indicator of actual preferences. This would limit the capacity of studies of revealed preference to impute the motivation of households for choosing between given zones or housing types from their actual choices.
However, the phenomenon of constrained residential choices would be observable as a weak connection between households' preferences for transportation and land use environments and their actual residence in such environments. Where choices are less constrained, households should be able to forge a better "fit" between their preferences and their choices. Thus the relative impact of choice-reducing constraints on development of alternative neighborhood forms may be observed as a weaker linkage between preferences and choices in the more constrained area as compared with the less constrained. In other words, where a range of choices of neighborhood types is readily available, households can be expected to sort themselves out according to their preference; where constraints limit the availability of alternative choices, less of this self-directed sorting would go on. This perspective can partially overcome the limitation of revealed preference studies that are restricted by constrained choice sets.
In order to assess the degree of fit between household preferences and neighborhood choices, it was necessary to classify different urban forms at the neighborhood level and the physical characteristics that typify them. This question is explored in the next section.
Neighborhood Forms and Characteristics
A metropolitan-wide classification of neighborhoods according to their transit/pedestrian or automobile orientation will necessarily rely on spatial data that are available consistently across the regions under study. This project seeks, however, to base its use of such data on current thinking in urban design. Brower (2000) proposes a classification of neighborhoods based primarily on qualities perceived by residents and tested through surveys. The four types which emerge from this research are: center--a cosmopolitan, active, lively type of neighborhood; small town--a settled, familiar, friendly type of neighborhood; residential partnership--an exclusive, homogeneous, family-directed type of neighborhood; and retreat--a type of place where one can find respite from people and pressures. While these neighborhoods describe the general correlation between physical characteristics and lifestyle patterns, our focus in this study is to examine the physical characteristics, or variables, that distinguish the nature of neighborhoods (e.g., pedestrian-oriented, transit-oriented, automobile-oriented, or a mix of orientations).
By dissecting classifications of neighborhood types (e.g., Brower 2000), we focus more specifically on several aspects, such as density, land use, layout, and amenities. The bases for neighborhood types by physical characteristics, then, are combinations of these aspects. However, neighborhood types are neither clear nor absolute; rather they are combinations of aspects that serve to highlight major physical characteristics and the differences amongst those combinations. In developing a cluster of significant and relevant neighborhood typologies, one has to acknowledge physical characteristics that residents recognize explicitly (e.g., whether they have to cross a busy arterial or walk through parking lots), or those found to be implicit that serve as proxies (e.g., higher density as a proxy for potentially higher social interaction, and abundance of sidewalks and footpaths as a proxy for potentially higher walkability). We also have to go beyond standard, empirical--yet narrow--measures such as degree of density and number of different uses, and incorporate variables such as distance to major destinations and variety of transportation modes in order to more fully grasp the richness and complexity of neighborhoods. With these qualifications in mind, we can now examine some of the major physical characteristics of residential neighborhoods that are most relevant to the study at hand--especially those that create pedestrian-oriented, transit-oriented, and/or automobile-oriented neighborhood forms.
Objective density in residential neighborhoods refers to gross residential density as measured by people per acre and housing units per acre and provides an overall sense of density in terms of proximity to neighbors, but also to work, school, retail, and other services. Other common measures of objective density includes: net residential density, which is total households per residential acre; gross population density, which is the total population per total acres; net population density, which is the total population per residential acre; gross employment density, which is the total employment per total acre; and net employment density, which is the total employment per commercial and industrial acre.
Subjective density refers to residents' experience of density--measured, for example, as percentage of neighborhood area as open space, percentage of neighborhood area as green space such as parks and gardens, a sense of scale such as absolute dimensions of open spaces where 40 feet in any single direction is intimate, 80 feet is human, and a maximum of 450 feet constitutes a successful urban square, and a sense of proportion via height/width proportions of enclosures such as buildings and streets (Lynch and Hack 1984). The quality of density may be measured by the amount of vegetation and cover (as seen from aerial photographs and plans which indicate green spaces, greenways, and landscape treatments); the grain or density of street network (e.g., average block size); land subdivision--pattern of lots (average parcel size); lot coverage--percentage of lots covered by built objects; size of land parcel-- smallest, median, largest acres or square foot areas; and street-widths as measured by right-of-way dimension (property line to property line on either side of the street) and number of lanes.
The types of land uses most significant for neighborhoods are residential and those closely related to it, such as public buildings, institutional (e.g., school or civic), and retail businesses. The number and variety of land uses in and near a particular neighborhood are also important. The most relevant uses to residential areas are employment (e.g., offices and commercial), retail (e.g., grocery and convenience stores, pharmacies, laundries, barbers, restaurants, shopping malls, banks), recreation (e.g., parks and recreation centers such as gymnasiums), education (e.g., day care centers, schools, universities), and services (e.g., libraries, health clinics).
According to a survey in Florida (Audriac 1999), the order of importance for residents to access other land uses is as follows: (1) parks, (2) shopping (e.g., grocery and convenience stores, pharmacies, laundry and dry cleaning), (3) community services (e.g., post office, church, library), (4) employment, (5) cafes and restaurants, and (6) entertainment (e.g., movies, theatres). Accessibility to such uses can be measured by walking distance (average 5 minutes or 1/4 mile and maximum 10 minutes or 1/2 mile), within walking and transit distance (30 minutes or so), and within automobile reach (10-15 minutes driving time, which equals 5-8 miles driving distance at 30 miles per hour). A variety of land uses can also reflect the diversity of a community--for example, heterogeneous lifestyles as reflected in housing types (e.g., income levels, marital status reflected in housing size and tenure such as rental versus ownership), and heterogeneous life cycle stages as reflected in housing types (e.g., singles, families with children, empty-nesters as reflected in small apartments, single-family detached homes, or assisted living complexes).
The layout of a neighborhood includes (a) spread (e.g., distance between destinations, distance between buildings); (b) grain (e.g., average lot sizes and average house sizes including smallest, median, and largest); (c) origin/ destination travel patterns (e.g., diffused or concentrated, concentration/dispersal of employment); (d) grid pattern of streets (e.g., easier pedestrian and car access, but also higher traffic and thus less attractive for families with children); (e) spatial quality (e.g., looseness such as free-form and objects floating in space versus tightness such as defining streets, providing definition to open spaces, establishing edges); and (f) road system orientation (e.g., feeding onto limited arteries and freeways, or shuttling vehicles within the area, or a combination thereof).
Other aspects of layout include (g) geometric pattern (e.g., linear, radial, grid, cluster); (h) legibility in terms of orientation (e.g., principal entries and exits, relationship to surrounding areas, location within neighborhood); (i) identity (e.g., distinct character of neighborhood, social and historical associations with physical place); (j) grain (e.g., intersections per square mile, blocks per square mile or average block size, building coverage--figure-ground, and number of lots of land or average lot size). In the book Great Streets (Jacobs 1993), the grain of urban fabric is demonstrated by intersections per square mile, where fine grain is 200 and above (e.g., Portland 351, central San Francisco 274, Boston 261, downtown Manhattan 218, central Oakland 208); medium grain is 150 to 200 (e.g., Santa Monica 185, central Los Angeles 171, midtown Manhattan 159, Washington D.C. 155); and a coarse grain urban fabric is 150 and below (e.g., residential areas of Irvine 119, central Walnut Creek 116, suburban Los Angeles 81, business complex at Irvine 15). The looseness or tightness of the built fabric can also determine grain, such as that of free form objects floating in space versus defined edges in a continuous fabric. A sense of scale refers to: (a) dimensions, (b) definition (e.g., sense of enclosure determined by height/width proportions, and edge as determined by tangible boundaries), and (c) proportion (e.g., width of streets and open spaces to height of surrounding buildings).
Amenities--such as prominence of natural features, type of landscaping, and style of architecture--are crucial to the qualitative, and often hard to quantify, aspects of neighborhoods. Landscaped features may be determined by examining number of open spaces per square mile, open space coverage as percentage of total neighborhood area, and amount of as well as types of vegetation. Often, neighborhoods are dominated by or designed around major natural features, such as a lake, river, hill, or woods. At a smaller scale, one can examine the type of landscaping which is prominent in a neighborhood, such as a lush and green type or a paved hardscape with street furniture. However, landscaping is most effective when serving both a utilitarian and an aesthetic purpose; for example, mitigating high vehicular traffic areas via visual and noise barriers of dense plantings. The quality of architecture impacts the overall feel of a neighborhood by the degree of prestige associated with architectural style, richness of materials and details (i.e. durability and variety), and designated historic district. However, amenities can also influence accessibility--a major focus of this study--by supporting or detracting from choices of modes of travel (e.g., walking, bicycling, automobile, bus, train, van, etc.). Thus, bicycle paths may be absent or present, limited or extensive. In the following paragraphs, we discuss amenities that are pedestrian-oriented, transit-oriented, and automobile-oriented.
Pedestrian-oriented amenities include: (a) design of separate lanes for slow and fast traffic, slowing traffic and pedestrian safety via curbside parking, landscaped medians and sidewalk, necking of intersections, special paving or creation of pedestrian tables at street crossing, etc.; (b) absence, presence or abundance of activities in public spaces such as walking, talking, sitting, jogging, bike riding, walking pets, rollerblading, and children playing; (c) traffic mitigation: absence or presence of landscaping (trees, shrubs, planters, grass) as buffer between traffic and pedestrians for screening visual impact, noise, and providing privacy to housing units facing road; (d) quality of experience (visual stimulation: landscaping, views, pavement design); sidewalk length in proportion to roadway length; and (e) nine percent parking; that is, when more than 9 percent of a 10 acre area is devoted to surface and garage parking, people feel that it is no longer pedestrian-oriented--an environment for people versus one for cars (Alexander et al. 1977). An example of a pedestrian-oriented amenity which contributes to the character of a neighborhood is an accessible green, which is a public open green space (e.g., park, garden, trail) about 3-5 minutes from every house (Alexander et al. 1977). Pedestrian orientation is also reflected in the adequate design of paths, where they are 3 feet wide for one-way walk, 3 feet for bench, 4 feet for two-way walk, 8 feet for two-way walk for 4 pedestrians, and 2 feet 6 inches for planter or curb next to road (Hoke Jr. 2000).
Transit-oriented amenities can be measured by the number of bus routes, number of bus stops, number of subway/light rail routes, number of subway/light rail stations; and the quality of transit stops--for example, simply a sign attached to a pole planted in the ground as a bus stop, or a bus shelter with a bench, lighting, garbage can, newspaper kiosk, public telephone and paved (rather than dirt or gravel) ground. A public transit interchange can be surrounded by housing and workplaces that cater to those who need public transit, be continuous with the pedestrian network of sidewalks and paths, and not interrupted by large parking lots or other barriers, and with a transfer distance of 300 feet if possible and a maximum of 600 feet (Alexander et al. 1977). At a larger scale, an entire transit village (Bernick and Cervero 1997) includes physical characteristics such as the congregation of housing, retail, and employment around transit stations, pedestrian-oriented amenities such as ample sidewalks and walkable destinations from the transit station, a mix of housing types which includes affordable housing, and places for public gathering, celebrations, parades, performances, and protests.
Automobile-oriented amenities include: (a) parking--percentage of parking in terms of open space and in terms of total neighborhood area, and quality of parking in terms of preponderance of large empty parking lots or small landscaped ones; (b) automobile surface areas--percentage of total neighborhood area and total open space area as roads and parking; (c) proximity to major roads and regional arteries; (d) traffic flow--presence of high volume roads, cars per day, average speed or speed limit; and (e) presence and/or dominance of roads--average street widths/right of way.
A good qualitative measure of the relative automobile, transit, or pedestrian orientation of a neighborhood is to observe not only the physical character, but also the activity at major intersections of a neighborhood. For example, it is not uncommon to observe wide roads, long crossing times, and few sidewalks in the suburban areas of American cities. Furthermore, one can sense the trepidation of pedestrians--especially the elderly--in attempting to cross wide arteries with fast moving traffic, even with crosswalks and signals. Similarly, a transit stop at an intersection which consists of just a pole and a sign, versus one which includes a shelter, bench, lighting, trash can, newspaper vending machine, and a public telephone communicates vast differences in the priorities of that neighborhood toward transit.
We examined some of the characteristics listed above in the clustering of the different types of areas using geographic information systems (e.g., population and employment densities) as well as the on-the-ground photographic documentation and analysis of case study neighborhoods within each type of cluster (e.g., presence of pedestrian-, transit-, and automobile-oriented amenities), as will be seen in the following sections.
Development of Neighborhood Typologies
In order to analyze the closeness of fit between households' preferences for land use and transportation environments on the one hand and their choices on the other, a scheme needed to be developed by which the territory of metropolitan Atlanta and metropolitan Boston would be divided into neighborhood types. Ideally the transportation and land use meaning of a given neighborhood type would be as close as possible between the two regions. This process was implemented through a cluster analysis that defined five neighborhood types in each of the two regions--corresponding roughly to "Central Business District," "other central city," "inner suburban," "middle suburban," and "outer suburban/exurban"--and assigned all neighborhoods to one of the five types.
Development of such neighborhood typologies was based on geo-referenced data that were available throughout the two metropolitan areas. These did not include all data that would assist in such neighborhood characterization; some desired items, such as sidewalk presence and continuity, were not available on a metropolitan-wide scale. Thus while this study has the advantage of analyzing all neighborhoods throughout two metropolitan areas (as opposed to neighborhood specific studies), it is more limited in the site-specific data that it can employ. Spatially referenced geographic information system (GIS) base coverages and sources for each metropolitan area are detailed in Table 2.
Table 2: Sources for Geo-Referenced Data
ARC: Atlanta Regional Council (Atlanta MPO)
CTPS: Central Transportation Planning Staff (Boston MPO)
ESRI: Environmental Systems Research Institute Data & Maps CD set dogwood.gis.gatech.edu: a web site maintained by Dr. William Backman of the Georgia Institute of Technology.
www.magnet.state.ma.us: Massachusetts state GIS web site
Tabular data assembled for the study included: 1995 population, household, and employment by travel analysis zone; congested speed (estimates) and number of lanes for the transportation modeling network; zone-to-zone travel time by automobile and public transportation; and transportation friction factors by travel time. The sources for all these data items were the respective Metropolitan Planning Organizations (MPOs) for the two regions.
The study area is defined as the ten-county area of the Atlanta Regional Council and the 101-town region of Metropolitan Boston, the area of the Boston MPO. The two areas are roughly comparable in population, with 1.1 million households in metropolitan Atlanta, and 930,000 in the Boston region, though the land area of the much more densely built Boston is considerably less: 1,400 square miles as opposed to 3,000 for metropolitan Atlanta. The geographical unit of analysis is the travel analysis zone (TAZ). TAZs are geographical units developed for transportation modeling purposes; they are sized to contain roughly 2,000 residents and/or employees, and to serve as a logical neighborhood unit for purposes of transportation analysis. The Atlanta study area is divided into 928 TAZs, while the Boston region contains 613 such zones.
Based on the urban design concepts described above, variables used to characterize the neighborhoods are listed in Table 4. Three classes of variables were employed: density variables, road network characteristics, and regional and local accessibility indicators.
Population and employment densities are the primary gauges of concentration of activity within a TAZ. The two were computed differently; whereas the denominator for population density was total residential land in a TAZ, employment was counted on the basis of total area. This was because of the differing meaning of the two types of density. A clustered village surrounded by open space is considered here to be a relatively dense living environment; hence, the nonresidential land was netted out of the residential density calculations. In contrast, a small island of dense employment does not render the neighborhood a significant job center; for this reason, employment divided by total land area was used.
As described above, urban form is significantly shaped by the characteristics of a neighborhood's street network. Walkable neighborhoods tend to be characterized by a fine-grained street network, indicated here by intersection density and street length density. Greater connectedness of a street network can shorten walking distances; thus, the percentage of "T" intersections can distinguish more connected grid or grid-like networks from those dominated by cul-de-sacs or other partially connected forms.
For transportation-modeling purposes, both MPOs maintain a transportation-modeling network that contains all major roadways in the region. These modeling networks include average number of lanes and estimated peak hour speeds as two basic link attributes. This study takes advantage of these modeling network attributes to calculate average speed and number of lanes by TAZ. Walkable neighborhoods tend to be characterized by relatively narrow and slow speed streets, which are easier and more inviting for pedestrians to cross than fast, wide arterials. For this reason, the travel speeds and number of lanes on links within the region's transportation modeling network were included as characteristics of the neighborhood these main routes border or traverse.
Metrics were developed to characterize the ease with which residents in a given TAZ could access destinations locally and regionally. Three scales were considered here: the walking scale of destinations within one quarter mile; the cycling or very short transit or auto trip of destinations within two miles, and the regional scale, incorporating accessibility to employment destinations throughout the respective regions.
At the very local scale, accessibility is viewed here as a function of land use mixing, or the coarseness or fineness of the "grain" with which an area is developed. For example, strictly land use separated residential areas that offer few nonresidential destinations within a short distance (defined here alternatively as one-quarter mile and two miles) are viewed here as offering poor local accessibility. Two dimensions, the intensity and the variety of land use mixing are captured here. "Intensity" refers to the extent to which a land use referred to is proximate to different land uses in Table 3. That is, the intensity measure for a grid cell--both the quarter-mile and two-mile cells--is the total number of cells surrounding it with different land uses. Thus, a residential island in the midst of a business district would score very highly on this measure. However, the measure does not capture the variety of opportunities available from a particular locale; for example, it fails to distinguish the residential zone surrounded by commercial land uses from a similar zone adjacent to commercial, recreational and institutional uses. For this reason, the "variety" measure was developed, and is equal to the number of uses in surrounding cells that are different from the use of the a cell in question. For any TAZ, the score on these measures was the average score of the cells in the zone.
In contrast, measures of regional accessibility require a view beyond the immediate neighborhood and need to capture the ease with which one can access destinations throughout the entire metropolitan area. Clearly, more remote destinations contribute less to the accessibility of a zone than closer destinations; similarly, proximity to larger concentrations would contribute more than access to smaller ones. The specific tradeoff between amount of distance to a particular destination is gauged here through the use of "friction factors," a product of travel demand modeling. People's travel between zones is observed to increase with decreasing distance of the zones and increasing travel opportunities at the destination zone; friction factors are the empirically fitted parameters designed to describe this variability in interaction between zones as a component of the regional transportation modeling process. As such, they were seen as a basis for gauging the impact that other zones throughout the region have on the accessibility of a given zone.
Numbers of jobs were used as the definition of the size of the destination zones. This was done for two considerations: (1) despite the growth in nonwork travel, accessibility to employment has been shown to be the single most influential determinant of residential location within a one-hour commuter shed (Levine 1998); (2) employment can also serve as an indicator for non-job related travel destinations. That is, zones containing schools, shopping, or recreational destinations will also display the jobs located at those sites.
Two measures of regional accessibility are utilized within the clustering model. The first is based on highway travel time and indicates regional accessibility for people traveling by automobile. The second is the ratio of transit accessibility to automobile accessibility, which is intended as a measure of the relative position of transit for a given zone.
Table 3: Land Use Categories Used for Land Use Mix Measures
|
Residential, single family--high density (lot size < 0.25 acre) |
|
Residential, single family--low density (lot size > 0.5 acre) |
Table 4: Variables Used to Characterize Neighborhoods
Implementation of the research design required the characterization of neighborhoods throughout the two metropolitan areas into neighborhood classes, based on similarity of transportation and urban design characteristics. In order to accomplish this characterization, a K-Mean cluster analysis (Aldenderfer and Blashfield 1984) was performed on TAZs of the two regions, utilizing the thirteen variables listed in Table 4. Cluster analysis is a family of techniques designed to group like cases on the basis of similarity across multiple dimensions. In order to render the meaning of the clusters as consistent as possible between the two areas, clustering was done for Atlanta and Boston in the same analysis; that is, the TAZs of the two regions were combined in a single data set, and clusters created without regard to the region in which they were located. In this fashion, the statistical meaning--if not the perceived land use and transportation implications--of a given cluster is the same between the two regions.
In initial clustering trials, the variables were all entered without transformation, with the exception of the regional accessibility variable, which tended to dominate other variables in defining the clusters because of its regular and smooth distribution. These initial trials yielded highly lopsided distributions that did not serve the purpose of creating useable neighborhood classes. These lopsided clusterings were the result of the presence of three highly skewed variables: density of population, employment, and roadway intersections. In order to reduce the skewness and create a more even distribution of TAZs between clusters, the natural logarithm of these variables was used as the basis for clustering. After these transformations were performed, the regional accessibility variable was re-inserted, and the final set of clusters developed.
A survey was developed and pretested for conducting by telephone (see Appendix A for the detailed questionnaire). The survey focused on respondents' neighborhood and transportation preferences, regardless of the neighborhood in which the respondents actually reside. Many of the key questions were phrased in tradeoff format. For example, Table 5 lists pairs of statements; respondents were asked to indicate which statement they agreed with more, and then to assign a degree of intensity regarding their selected statement. The guiding philosophy was that many people hold a set of preferences that is internally contradictory; for example, they may want walkability on the one hand, but only low-density, land-use separated development forms on the other. The idea of the tradeoff-styled questions was to force them into a choice between potentially contradictory elements of their preferences in order to ascertain which was a higher priority.
Table 5: Examples of Tradeoff-Styled Survey Questions
The survey sample was developed through a random selection of individuals from the database of Experian, the credit reporting company. An initial check was performed to ensure that the distribution of individuals in the sample matched the distribution of population by community throughout each region; a chi-square analysis confirmed the match at greater than 99 percent confidence.
Initially, samples of 5,600 individuals were drawn in each of the two metropolitan areas. Contact was attempted with approximately 3,000 households in each sample (first by postcard to alert respondents, then by phone call) with about 2,000 individuals actually contacted in each area (Table 6). Numbers that were unanswered were attempted at least four times at different times of the day before the phone number was abandoned. Overall, 1,607 individuals completed the survey for a response rate of 38.9 percent.
Table 6: Outcome of Attempted Survey Interviews
In order to fashion a survey sample representative of the population in each region, it was necessary for the sampling percentages in each neighborhood cluster (Table 7) to match population percentages in the same area. Because of differing response rates in different areas of the metropolitan regions, weights were applied to construct a sample that would match this outcome. Calculation of weights (the ratio of population proportion to sample proportion by neighborhood cluster) is specified in Table 7.
Table 7: Calculation of Weighting Factors