Sprawl, I Hardly Know Ye

I am behind with other things so to fill space & time I’m recycling my 1997 letter to the editors of JAPA, commenting on the now famous point/counterpoint articles on sprawl by Gordon & Richardson and Ewing in the previous issue. It is kind of hard to find otherwise (if anyone was looking). This letter was prompted by Marty Wachs, who asked, “What did you think of these two articles?” I was caught off guard as I hadn’t seen them yet, so probably read with more vigor than I would otherwise. The citation is “Sprawl, I Hardly Know Ye,” Journal of the American Planning Association 63 (2), 278-9.

p.s. If you have issues with an article in an academic journal, write a letter to the editor and see what happens.

p.p.s. On the topic of sprawl, Rolf Pendall and two of the authors of the forthcoming QJE paper on measuring sprawl I discussed in this blog, Diego Puga and Matthew Turner, have added extensive comments on that post regarding data sources, etc.

Dear Editors,
I read the recent JAPA “Point/Counterpoint” articles by Peter Gordon and Harry Richardson (1997) and Reid Ewing (1997) with great expectations. Several basic questions for planners of various stripes were addressed, both positive (e.g., what is sprawl, what does infrastructure cost, how does land use affect travel behavior and the quality of life?) and normative (e.g., should development be required to be more compact, is sprawl good or bad?). Unfortunately, I admit to being somewhat discouraged by the apparent Hobson’s choice. The two papers may span the range of views and evidence on these matters but there is little actual exchange in the exchange. The articles do identify several important questions for both research and practice but I wonder how plain that is to those outside the literature. In particular, the gap between what we know and don’t know receives precious little notice.

What is the Question?
As Ewing usefully indicates, “sprawl” has numerous elements and is many things to many people. Gordon and Richardson (G & R) mainly argue against regulated increases in densities while Ewing questions the desirability of a particular form of expansive low density land use pattern, perhaps with leapfrogging, or what he calls “Los Angeles-style.” In either case they chiefly ask whether planners should intervene by (a) clustering activities and/or (b) moving potential origins closer to potential destinations, all things considered. Put compactly, should trips be made shorter?

G & R say, for the most part, “no.” They reason there’s plenty of land and fuel, people prefer less to more crowding, there are few aggregate benefits such as increased productivity, cost savings, or income redistribution, and besides there’s little wrong with low densities that the market can’t mediate. They mention for example that subsidies to car use are relatively small and lower densities may increase trip lengths but don’t contribute to traffic congestion, generally speaking. The implicit message is that specious meddling by planners, however well intended, could worsen rather than improve things. Even perhaps in L.A.

Ewing says, in many instances, “yes.” He argues that employment centers continue to offer spatial advantages, cars impose massive social costs, open space is undervalued by the market, and that low density sprawl indeed does inflate infrastructure and psychic “deprivation” costs, as well as impairing residents’ sense of community. Thus, in addition to healthy market incentives, there are social benefits to improving “accessibility” and promoting higher densities in many cases.

The two papers wander because there is lots of messy terrain to cover. Still, the questions are all important and the surveys make for good reading, apart and together. It would be a fun class exercise to compare and contrast them in some detail, but here is my dilemma: Presented with these alternatives, I can commit to neither. And I don’t think JAPA readers should either. To keep things short, consider one example:

How Does Density Affect Behavior?
We don’t really know. This is a remarkable void given all the attention the question received in these articles and elsewhere. There is no shortage of research on the issue; rather there is a shortage of credibility. The problems are two fold. Data are generally poor and the empirical strategies to date are inadequate to the task.

Unfortunately for planners and researchers alike, travel behavior is complex. It has typically proven very difficult to explain even a quarter of the variation in either aggregate or individual level travel data. The variation that is explained is difficult to interpret in a standard manner. While recent studies (e.g., Cervero and Kockelman 1996) make great strides in measuring and characterizing land use variables, for example, they rarely possess even rudimentary behavioral foundations. Instead they employ various measures of “accessibility”, pedestrian “friendliness”, density, and the like as control variables in ad hoc regression specifications or ANOVAs with little systematic behavioral content.

One consequence is that variables are dropped and added, and the results compared, without any discussion of the dangers of specification bias. A conventional approach to evaluating demand would specify behavior as a function of relative prices, resources, tastes, and other relevant controls. It is surprising then to note that none of the studies cited by either G & R or Ewing include relative trip costs in any form. It is skating on thin ice indeed to base policy on a density coefficient estimated by a model that simply assumes away a pivotal factor in mode choice, trip length, and the decision to travel. Moreover, density and accessibility measures implicitly vary with respect to both demand and supply variables, and until the data and models adequately capture these differences it will be hard to assess or interpret a given estimate of how much density affects VMT. The relationship is too aggregate in nature to be reliable as a policy variable at this stage.

Worse yet, nearly all these studies ignore the truncated nature of the data. People who live in a neighborhood of one sort (defined by street pattern, density, or level of access) cannot be reliably compared directly with people who do not. They self-selected, at least somewhat, and thereby revealed their preferences across available travel environments. People who want to walk or bike or drive or travel by train relatively more, and thus do, will seek out settings where they can. For example, those who live near commuter rail stations may well take the train more often than those who do not and thus are more likely to live there to begin with. The sample data reported in comparisons of this kind are systematically biased.

The problem can be statistically finessed. Boarnet and Sarmiento (1996) explicitly model this set of joint choices — where to live and then how to travel — and find that on net land use variables do not influence travel in their southern California sample. This work is still in its infancy, however, and it is far too soon to generalize on the basis of any one study or set of studies for policy purposes.

What Should Planners Do?
Should we try to compact cities? I don’t know. And I don’t think anyone knows. It is still unclear how the generic benefits of doing so compare with the generic costs. Certainly the evidence on how travel behavior would respond is both mixed and untrustworthy. It also seems rather intrepid to elevate a particular kind of development pattern, like compactness, to near paradigm status without stronger support. On the other hand there does appear to be some obvious merit in permitting the marketplace to provide such opportunities if they don’t impose other social costs, as Fulton (1996) suggests, if only for the value of greater product diversity. Beyond that, and absent blind faith, the profession should be wary.

G & R also point to the dearth of data on some of these questions but do not suggest how to improve the situation. It is enough, apparently, to find a lack of empirical support for compact plans. In other cases, their choice of supportive data is debatable. Ewing seems to think many of these issues are settled but in doing so relies on studies that some might find insubstantial. In other instances (e.g., regarding “psychic costs”), he favors data consistent with his story.

Rather than lean on data to emphasize where our values differ it might be useful to focus on where better knowledge would clarify things. Perhaps the next Point/Counterpoint on planning goals could emphasize the quality of and, where appropriate, the need for research as much as the contrary nature of raw results.

Boarnet, Marlon and Sharon Sarmiento. 1996. “Can land use policy really affect travel behavior? A study of the link between non-work travel and land use characteristics,” Urban Planning Working Paper No. 51, University of California at Irvine, December.

Cervero, Robert and Kara Kockelman. 1996. “Travel demand and the three Ds: Density, diversity, and design,” IURD Working Paper No. 674, University of California at Berkeley, July.

Gordon, Peter and Harry Richardson. 1997. “Are compact cities a desirable planning goal?” Journal of the American Planning Association 63, Winter: 95-106.

Ewing, Reid. 1997. “Is Los Angeles-type sprawl desirable?” Journal of the American Planning Association 63, Winter: 107-126.

Fulton, William. 1996. The New Urbanism: Hope or Hype for American Communities? Cambridge, MA: Lincoln Institute of Land Policy.

Thursday, March 9th, 2006
randall Crane