On Form Versus Function: Will the "New Urbanism" Reduce Traffic or Increase It?

A major attraction of the popular and influential planning movements known as ’the new urbanism’, ’transit-oriented development’, and ’neotraditional planning’ are their presumed transportation benefits. Though the architects and planners promoting these ideas are usually careful to emphasize the many ingredients necessary to obtain desired results -the straightening of streets to open the local network, the ’calming’ of traffic, the better integration of land uses and densities, and so on w a growing literature and number of plans feature virtually any combination of these elements as axiomatic improvements. The potential problem is that the traffic impacts of the new plans are generally indeterminate, and it is unclear designers understand the reasons well enough to avoid unintended results. This paper proposes a simple behavioral model to identify and assess the tradeoffs these ideas impose on transportation and subdivision planners. *I am grateful to Marta Baillet, Marlon Boarnet, Robert Cervero, Richard Crepeau, Ralph Gakenheimer, Peter Gordon, Robert Noland, Don Pickrell, Sherry Ryan, Lois Takahashi, Brian Taylor, and Martin Wachs for very heIpful feedback on this topic, and to the U.S. and California Departments of Transportation and the University of California Transportation Center for financial support. m Introduction Transportation problems seem to offer no end of interesting policy wrinkles and engineering challenges, but despite the promise of each new technological k, movation, financial windfall, and dazzling social science breakthrough, planners have not fared well. Even as air pollution, fuel, and traffic congestion costs mount o the point where the benefits of making any headway appear substantial, more freeway lanes are dedicated to carpoolers and toll-ways, and new transit systems continue to soak up many billion of dollars, getting people to ’improve’ their driving behavior remains the ultimate planning brick wall. Increasing evidence suggests that tr~msportation management schemes have extremely limited effectiveness, in the sense that only marginal and perhaps even cost-ineffective changes can be expected from most of the tools applied thus far (e.g., Giuliano, Hwang and Wachs 1993; Wachs 1993a; Deakin and Harvey, et al. 1994). While one view is that the planner’s arsenal of transportation management tools has proven iargely ineffective in dealing with traffic congestion especially, the somewhat more optimistic a(-count of some planners and architects is that attention has been focused on symptoms rather than the disease itself. The vanguard of such urban design schools as ’the new urbanism’, ’neotraditional planning’, and ’transit-oriented development’ have collectively argued that the way we organize space has profound implications not only for traffic patterns but perhaps for our sense of self and modem civilization as a whole. Such prominent urban designers as Andres Duany and Elizabeth Plater-Zyberk (1991, 1992), best known for their work on the neotraditional community of Seaside, Florida, and Peter Calthorpe (1993), the author of the transit-oriented ’pedestrian pocket’ concept, forcefully claim their planning strategies will, among other things, improve traffic conditions, reduce home prices and generally increase the quality of residential life. 1 Of course, this is just talk. As bold and stirring as these claims may be, they are mainly meant to get us thinking afresh about where and how improvements can be made -not as cold hard facts. Most transportation planners probably recognize that blanket statements of this nature are overly simplistic. (See for example the questions raised concerning the scope for transportation policy to influence land use, or vice-versa, in Boamet and Crane (1995), G6mez-Ib~fiez (1985), Giuliano (1989), and Wachs (1993a, 1993b).) Even the architects and planners promoting ideas are usually careful to emphasize the many ingredients necessary to obtain desired results: The straightening of streets to open the local network, the ’calming’ of traffic, the better integration of land uses and densities, and so on. The new designs have many elements, and while their purported transportation benefits are often featured, it is by no means the primary component. Still, these ideas appear to have had made a great impact on modem city planning thought and practice. Perhaps in their frustration to find policy tools that can make a difference in the struggle with automobiles in the city, a good many planners and communities have taken the transportation themes of the new urbanism to heart as among the most feasible and promising. Within a few short years, the new urbanism has become perhaps the most visible intellectual paradigm in urban design theory circles and has steadily increased its influence among subdivision and transportation planners as well. A growing number of general and specific plans feature various combinations of these elements as self-evident improvements (e.g., see Calavita 1994; Los Angeles 1993; San Bemardino 1993; San Diego 1992), and the claim that virtually any one such element has transportation benefits has rarely been chaJlenged in either the practitioner or scholarly literatures. It is somewhat surprising, then, to find the empirical support for these transportation benefits to be meager and their behavioral foundations obscure (Crane 1995; Gordon and Richardson 1995). Only a few studies -discussed below -have examined these issues head on, and interpreting their mixed and contrary results is difficult owing to the lack of a consistent analytical framework. 3 The purpose of this paper is to suggest an organizing scheme, and to then explore the behavioral implications of the new plans for travel. It presents a simple model of travel demand identifying the interaction of the main factors under debate. Under very general circumstances, we find the net effect of the new plans can either increase or decrease the number of car trips as well as overall vehicle miles traveled. The result in any instance depends in part on how sensitive the demand for each mode is to changes in the time required for each trip, how well one mode substitutes for another, and the particular manner in which the plan is implemented. There is no theoretical basis for stating that the new urbanism will unambiguously reduce car travel. However well intentioned, the new designs can thus cause problems when naively applied A second purpose of this paper is to suggest how such problems can be avoided. The behavioral fr~unework can be used as the basis for comparing the impacts of different plan elements on traffic and pedestrian travel. The new urbanism is a hopeful school of thought, brimming with promise, and should be encouraged in those respects in which it succeeds. The purpose of this work is to make that task easier. m Linking Neighborhood Design to Travel As essentially architecture, the new urbanism is part philosophy, part art, part economics, ~td part social engineering. Still, a key to its popular acceptance is the open embrace of conventional and even conservative standards of neighborhood form, scale and creed. Neotraditional planning in particular self-consciously recalls small town settings where your neighbors walk to get a haircut, and stop on the way to chat as you sit on the front porch watching the kids play. The attraction of these ideas is subjective, personal, yet pervasive: surveys indicate their appeal among suburban residents is especially solid (haman 1993). After all, in principle, what’s not to like about pretty homes in quiet, friendly and functional neighborhoods? But will they improve the traffic? Available evidence on the transportation benefits of any feature of these plans is mixed at best and often contrary. Most studies are grounded in either dubious assumptions, poorly flamed questions, or comparisons of dissimilar communities. Studies of actual neotraditional developments have not been published, as virtually none are fully built out at this time. Hence, even careful quantitative evaluations tend to be based on hypothetical situations, as in the case of engineering simulations, or data obtained from older ’traditional ~ communities haring some characteristics with proposed ’neotraditional’ communities. Those studies supportive of automobile-suppressing properties of either grid street patterns or pedestrian friendly characteristics tend assume trip frequencies do not vary from one design to another, or fail to isolate the independent influence of each feature on travel behavior: Simulation studies, such as Kulash, Angtin and Marks (1990), McNally and Ryan (1993), Rabiega and Howe (I 994), and Stone, Foster and Johnson (1992), have tended to focus whether a more grid-like street pattern reduces vehicle miles traveled (VMT)o They model the new plans as essentially moving trip origins and destinations closer together, but most hold the number of trips fixed. (Stone, Foster and Johnson (1992) let trip generation rates change based asstmaed differences in the land use mix in each scenario, and then apply fixed trip rates for each use based on published engineering standards.) Thus the studies tend to ask "If a trip becomes shorter, will people drive as far?" The answer is "no", but what we learn from the exercise about the expected impact of these schemes is unclear. The pivotal question is whether there will be a behavioral response, such as a change of modes or a change in tdp frequency. These studies typically assume away such responses -apart from what engineering standards divine -though they would seem to be key to predicting what will happen in practice. (Figures 1, 2 and 3 about here.) Empirical studies can’t assume away behavior; rather they must explain it. The research strategy in most analyses has been to simply search for correlations among neighborhood features and observed travel -in some cases controlling for other relevant factors and in others not. For example, studies such as Cervero and Gorham (1995), Friedman, Gordon and Peers (1992), Hanson and Schwab (1987), Guy and Wrigley (1987), Handy (1992b, I994), Holtzclaw (1994), Ki’tamura, Mokhtarian and Laidet (1994), and 1000 Friends (1993), have collectively reported more ’traditional’ neighborhoods are associated either fewer or more automobile trips than °nontraditional’ environments, with the result that overall VMT can either fall or rise. Again, however, interpreting this range of results in any one case is problematic since the causal theory is not clearly established outside the design rhetoric. What is generalizable about the factors in one environment hat generate more car trips, and in another less? While some studies based on observed behavior do attempt to control for different trip purposes (e.g., shopping versus commuting), trip lengths (neighborhood versus regional), and demographic variables likely associated with trip demand (income, age, etc.), the approach is typically adhoc. Further, there is the question of what the range of outcomes found in this work suggests about the ability of the new urbanism to deliver the transportation benefits it promises. The point of departure for this paper is the argument that the literature on the transportation impacts of neotraditional or other new urbanism designs has yet to employ a strong conceptual framework when investigating these issues, making both supportive and contrary empirical results difficult to compare or interpret. In particular, an analysis of trip frequency and mode choice requires a discussion of the demand for trips, but this is often lacking in planning and land use studies at even a superficial level. That approach would permit us to explore the behavioral question, for example, of how a change in trip distance influences the individual desire and ability to take trips by each mode. The tools of microeconomics provide perhaps the most straightforward framework for such a discussion, by emphasizing how overall resource constraints enforce tradeoffs among available alternatives, such as travel modes, and how the relative attractiveness of those altematives in turn depends on relative costs, such as trip times. The demand approach assumes that individuals make choices, either alone or as part of a family or other group, based on their preferences over the goods in question, the relative costs of those goods, and available resources (e.g., Kreps 1990). Preferences include attitudes and tastes, for example regarding the experience of driving versus walking, and are likely correlated with demographic and other persona/idiosyncratic characteristics. But the decision to take a trip to the coffee shop by car or by foot depends not only on how one feels about those options, but also on external factors over which one has no or only limited control; i.e., on the cost of one mode versus the other. I may prefer to drive, but if the gasoline or parking expenses of doing so are high enough, walking may appear to be the better choice. Thus the demand for walking trips is explained not only by one’s preferences across modes but also on the cost of walking relative to the cost of driving, etc. That, simply put, is the economic theory of demand. The rote of accounting for available resources is mainly to fix the importance of costs; the impact of a $5 parking charge on your decision to drive to the coffee shop depends on what funds you’ll have left for that double expresso needed to get you through the afternoon. Note the framework applies just as well to any situation where decisions are made concerning the allocation of scarce resources, whether they involve actual money or not. In the model presented below, for example, the scarce resource is time, and each mode is compared in terms of the time consumed rather than the cash. Note also that this framework does not explain preferences, it only explains how one makes informed decisions given those tastes together with costs. While this approach can appear artificially abstract and vexing at times, depending on the problem at hand, it does have the substantial advantage of laving out the tradeoffs of interest in a rel.atively clear-cut fashion once the fundamentals are understood. What would we expect if the cost of one mode rises while the other falls, for example? It is not necessary to follow the model or the derivation of the results to understand the argument hey support, but the details do determine the credibility of the argument. The usefulness of the analysis is another matter. Modeling design features in this framework requires some simplification and standardization. That naturally depends on the appropriateness of the model to the problem under study, and oversimplification can obviously be fatal in that respect. This is no less true in the present instance, and some care has been taken to capture the main elements of the neighborhood travel environment. The results are summarized in the last paragraph of this section and the conclusion. A Model of the Influence of Neighborhood Design on Trip Demand To focus on the behavior of interest, consider the problem of individuals making choices over five uses of time: the number of trips they complete by car, those they complete by foot, those by transit, those by some other transportation mode, and a composite good representing all other uses of time. (That is, the model abstracts from other decisions, which is different from asstuning they don’t happen but does imply they aren’t a central feature of the story.) For most p~uposes, a trip is a ’derived’ demand, meaning that people typically travel as a means to an end, not as an end in itself. A ’trip’ is thus defined as a hedonic index of the quantity and kinds of goods one obtains during each sort of trip. We ignore non-time constraints to emphasize the restriction imposed by the time required for a trip in each mode on the choice of the number of trips in each mode. (These simplifications substantially streamline the exposition while not affecting the qualitative results.) In this case, the decision process behind the choice of the number oft-tips may be formalized as the constrained maxirnizafion problem of choosing the number of trips by each mode to maximize the benefit of travel by mode, subject to a budget constraint reflecting travel costs and available time. In the standard functional notation of this modeling approach, the problem statement is to assume that individuals choose their desired number of trips by each mode to maximize subject to U(a, w, b,x) y = x ÷ apa + WPw ÷ bPb, where U is an index of the benefits of using time for each purpose, a is a vector of the number of trips by automobile for each purpose, w is a vector of the number of trips by walldng for each purpose, b is a vector of the number of trips by bus or other transit for each purpose, x is a composite of the time spent on other activities, the Pi are the respective vectors of times for each trip type in each mode (i ~a,w,b), and is total available time. Forexample, say there are 10 different trip purposes which we index by j = 1,2,3,...,10. Perhaps the first purpose is grocery shopping, the second trips to school, etc. Then a -~ (a1 ,a2,a3 .... ,aI0) = (number of car trips to grocery store, number of car trips to school,....). The total number of car trips taken for all purposes during the reference time period (a week say) is ,~-vj° laJ, and the total time spent traveling by car was aPa = ~°laJp~. Note further the time per trip is the quotient of trip length mi and speed t i; i.e., where Pi E mi / ti for i ~ a,w,b, for any particular trip purpose (i.e., with superscripts suppressed for simplicity). The solution to the choice problem is summarized by the trip demand functions a( pa , Pw , Pb , Y) , w( pa , Pw , Pb , Y) and If Pa , Pw, Pb , Thes e functions have many useful properties, but their practical value for the problem at hand is that for any given set of travel preferences, they transparently relate changes in trip costs to the number of trips desired, by trip purpose. For example, they can be used to estimate the impact of an urban design change that lowered the time (or other cost) of a trip by foot on the number of trips by foot, the number trips by car, and the number of transit trips -for each trip purpose. This information could thus be used to calculate how vehicle miles traveled respond to increased pedestrian, transit, or auto access due to a change in street patterns. Estimable forms of these demand functions for empirical application to specific data may be obtained by specifying a particular form for U (e.g., Small 1992; Crane and Crepeau 1995). However, the basic theoretical implications of the behavioral model can be explored v,~ithout data. The potential inconsistency regarding the transportation benefits of the new urbanism is internal to these design principles. To show this most clearly, the paper is restricted to deriving some basic implications of the behavioral model via the method of comparative statics. 3 In the context of the model presented above, how can the pivotal features of the new plans be represented? Rather than attempt a comprehensive r view, the analysis is restricted to the three most common design elements with assumed transportation benefits: a grid-like street pattern intended to reduce the distance between local trip origins and destinations, traffic ’calming’ measures intended to slow cars down, and integrated land uses at higher densities intended to combine more trip destinations at single locations. The role of the grid in these plans is multifaceted, ranging from increasing the ’legibility’ of the neighborhood to improving the connection of people and places. Among the ideas that have been reborn in the new urbanism, the renewed popularity of the grid is both the most fi’equently mentioned by traffic analysts and perhaps the most compatible with modem street and subdivision codes. For transportation purposes, its major function seems to shorten local trips. The relationship between the time required for each trip in each mode and land use is assumed to be captured by way of a ’grid’ shift parameter y, where an increase in y (more grid-like) decreases trip lengths. That is, the derivative d~ni < 0 for i = a,w,b. Notice this parameter could represent the effect of any land use change that made a trip shorter. It is also compatible with a specification assuming transit or pedestrian trips are shortened more than car trips, where dma draw > ~, or other possibilities. dr dr Traffic calming refers to the narrowing of streets and intersections, and other means as well, that slow cars down (e.g., Untermann 1984; Ben-Joseph 1995). We model this effect with ’calming’ parameter X, where an increase in Z slows car speeds down; i.e., dta < 0. Finally, ,Ix mixing, combining and intensifying the density of land uses to make any one trip potentially serve more than one purpose might affect trip demand in at least three ways: It can essentially increase the consumption associated with amp directly and it can also lower the cost of any ’chained’ trip. Defining aa increase m the shift parameter /~ to symbolize an increase in land use mixing or more Oa < 0 and the latter by intense use of a destmation site, the former effect can be represented by O~ dpi < 0 for i = a,w,b. More intensive use can also increase congestion, such that dta < 0. du 10 The Grid With these design features so defined, we can investigate their individual and collective theoretical impacts on travel behavior via comparative statics. Note first that total VMT for all V~° aimj Hence, an approximate measure of the effect on VMT for one trips is am a = ~ j.: a ̄ 11 particular purpose due to a move toward a grid street pattern is simply, dVMT a dm a da +Pa dy t a dy -~y (1) (This approach treats trips as approximately continuous. They are not, of course, and the modifications necessary to account for the discrete nature of the trip decision are described in BenAkiva and Lerman (I985) and Small (1992).) This equation succinctly summarizes the automobile travel behavior of an individual benefiting from a more grid-like street network that in tam leads to shorter trips. The first term on the fight-hand side of (1) measures the effect of shorter trips for the number of car trips prior to the street network change. It enters (1) negatively by assumption, and summarizes the results of the studies which have held trip frequency fixed. The latter term is the induced effect on the number of car trips. Do we expect trips to increase or fall? To see this note the number of car trips responds to a small change in trip length according to the total derivative, da ~ dma 1 o~a dmw 1 Oa dmb 1 + + (2) dy o~na dy t a Ornw dy t w ~nb dy t b 3~e first term on the right-hand side is the change in the desired number of car trips induced by the time savings per trip. This is likely positive, as can be seen from the Slutksy decomposition for da/o~a, which breaks down the price change impact into two parts: the impact due to a change in relative prices, and the impact due to a change in overall costs: o~a t~c 8a @0 @,~ ay ~c °~Pa ,Pw,U) where --~ 0pa o~a < 0 is the change in demand ue to the change in relative prices (the ’compensated’ effect) and ~a/a), is the impact of having time freed up by the shorter trips. If automobile trips are a normal good (i.e., the demand for auto trips increases with resources), then ~a/~ > 0 and 8a/o~pa must be negative. Thus, the demand curve for automobile trips is typically downward sloping and the first term in (2) is positive: All things considered as the time per trip falls, due in this case to a shorter trip, people will tend to want to take more trips. The number of car trips can fall with a decrease in trip length, however, if the sum of the second and third terms in (2) is sufficiently negative° These represent the cross-price substitution effects of shorter walking and transit trips on car trips. As walking trip lengths fall, owing to a better system of walkways or more direct street patterns, etc., we might expect people to substitute walking trips for car trips. Indeed, pedestrian trips are more influenced by trip length (and purpose) than by trip time, especially when compared to motorized transport. Evidence that walking trips fall off dramatically after trip distance of a half-mile suggests that the second term in (2) is highly elastic near that figure, and zero for longer distances (e.g., Untermann 1984). Shorter transit trips have a less clear effect, again depending on the trip purpose and other particulars not explicitly modeled here -though the time of the trip is probably a more important single indicator that the trip length. Hence, if automobile trips are a normal good then 0a is negative and the sign of (2) indeterminate. If the new street network is such that people tend to substitute walking or transit for car trips compared to an alternative plan, and the demand for car trips is relatively insensitive to the length of the trip, the number of car trips can fall. But if these conditions are not met, car trips can rise. Whether VMT rises or falls is a separate matter. VMT is the product of the number of trips and their length. If trip lengths fall, as implied by a move to a grid, (2) shows that the nmmber of trips could rise -especially where few transit or walking trips are substituted for car trips and if car trips are sensitive to their length. If the number of car trips rise enough, then VMT could rise as well. 12