Abstract: Resolving the sprawl vs traffic debate. Not really.
By R. Crane (UCLA) and D. Chatman (Rutgers).
(Note: This essay originally appeared in Access magazine #23, Fall 2003, so I've updated the references to their subsequently published versions. The essay was in turn largely drawn from an article in Planning & Markets, later republished in slightly revised form in a chapter of the book, Urban Sprawl in Western Europe and the United States, Ashgate, 2004.)
The single clearest spatial fact in the structural evolution of metropolitan areas may be the continuing decentralization of jobs and people to the suburbs and beyond, aka “sprawl.” What are the consequences for travel behavior and transportation planning? Surprisingly, perhaps, we have little to say on these matters and a lot to learn.
The sprawl debate, like its namesake, often seems to chaotically spill forth without check – but two chief complaints are environmental and social. The former contends, among other things, that sprawl is responsible for the excess conversion of open land to urban uses, and the air and water pollution associated with increased auto use. Leading scholars also argue that suburbs leave their residents isolated and alienated, in part as lengthening commutes leave less time and energy for social interaction.
But the transportation literature, key to both arguments, can be mainly noted for its several pronounced gaps. UCLA recently initiated a major research effort aimed at addressing many of these, ranging from investment strategies to access equity to finance. This article focuses only at one of these questions: How does job sprawl influence commute length?
The classic theories of urban structure suggest that people would normally prefer to live near where they work to reduce commute time and cost, but that land markets force a tradeoff in the form of housing costs. Land prices end up reflecting the value of job access; one can better afford a larger house and yard further from central employment locations. People will then choose a commute length that matches their budget and preferences for space.
In a model with centralized employment, the suburbs materialize for those who want more affordable larger homes (and other suburban amenities).
What if jobs suburbanize too? In real cities, firms do not all locate downtown. They also face tradeoffs between the benefits of clustering (agglomeration economies), the cost of land (versus access to markets), and from being nearer their workers (to reduce commute costs and thus potentially wages). In this story, job decentralization occurs when the benefits of being near workers are sufficiently high compared to the benefits of firm clustering. Shorter commutes therefore go hand in hand with decentralization. (See Figure 1.)
Of course, real cities are more complex still, leaving the theoretical influence of job decentralization on commute length ambiguous. Consider three complications.
First, workers change jobs over time, and households often have more than one worker. As jobs decentralize, the location of one’s next job is increasingly uncertain. Choosing where to live on the basis of proximity to the work location becomes a gamble, especially since it is costly to move. Workers may well hedge their bets by locating at some intermediate location to minimize average expected commute costs. Similarly, when households have more than one worker, and their current or expected future jobs are in different locations, the residential location decision is not a simple matter. In either case, decentralization could increase rather than decrease commute distances.
Second, there are other possible benefits to firms of decentralization, including underused local transportation capacity in outlying areas, better access to key transport nodes to external markets, reduced parcel assembly and demolition costs, lower labor unionization rates, lower taxes, and proximity to other suburbanizing firms and residents. If these factors are driving job decentralization, then job decentralization does not necessarily imply shorter commutes. This is particularly likely if the pattern of decentralization is polycentric rather than simply dispersed from a single city center.
Third, households think about more than their jobs when choosing where to live. These factors include access to shopping and other non-work activities, various qualities of the local neighborhood, schools and other public service quality, and the resale value of the property (which reflects all these). By definition, these other motivations for residential location will tend to increase commute length from that expected under the classic model.
So, in theory, sprawl may lengthen or shorten average commute length. While not answering our question definitively, these stories do substantially clarify the contributing factors and the nature of their tradeoffs. If commutes are longer in more suburbanized areas, this would imply that certain factors dominate, and vice versa. The role of these factors in transportation planning should then more transparent.
To find out what happens in practice, we need to examine actual travel data.
The observed trend is that in many metropolitan areas, commute times have risen (as shown in Census data in Figure 2.) However, we do not know why, or in particular what the contributing role of employment suburbanization might be.
Relatively little empirical work exists. In an important study 15 years ago, Peter Gordon, Ajay Kumar and Harry Richardson used county-level data for metropolitan areas to investigate this question. They found that commutes in spatially larger cities took more time, while shorter commutes were associated with a higher proportion of industrial employment. Both higher overall residential density and the share of employment in the central city were highly associated with commutes of longer duration. They concluded that both residential and employment dispersion lead to shorter commutes.
Other research using individual-level data has found that decentralization lengthens the commute under some circumstances or for particular household groups. For example, in his UC Berkeley dissertation (and subsequent research), Jonathan Levine found that the commute distances of low-income households grew as employment suburbanized, in part due to shortages of affordable housing nearby.
For a number of reasons, this research cannot be considered conclusive. In part, the behavioral data are too aggregate or otherwise limited to explore individual choice margins, and the statistical models are insufficiently developed. More to the point, these phenomena are far too complex to ever be fully understood. Our purpose is more to move the discussion along than to settle such questions.
Our approach improves on the data and methods of earlier work in several respects. We start with the American Housing Survey (AHS), an extensive, nationwide longitudinal sample of housing units administered every two years by the Census Bureau. We then merge the AHS data with county-level employment data from the Bureau of Economic Analysis. Finally, we estimate panel regressions of worker commute length on a conceptually sound model of commute choice that conforms to urban structure theory, as described below.
We used a subset of seven waves of data from the AHS from 1985 to 1997 consisting of 185,085 total observations, which in turn represent 42,380 distinct housing units.
Unlike typical measures of decentralization based on distance from a city center, our measure allows non-contiguous or spatially extensive urbanization of sufficient density to be designated as core urbanization. Only employment in counties without large enough agglomerations or sufficiently high density is designated as decentralized. The measure is therefore quite conservative in terms of including portions of polycentric urban areas within the urbanized area, although it is somewhat crude due to its reliance on county geography. It can be thought of the share of employment in the most dispersed form—outside of urbanized counties altogether, though adjacent to and functionally related to the core urban areas.
Consistent with other studies, average commute distance in our AHS sample showed an increasing trend over time, men have longer commutes than women, and residents of owner-occupied units have longer commutes than renters. The share of employment in outlying counties for four industries defined using one-digit Standard Industrial Classification codes. Nationwide, manufacturing and FIRE (finance, insurance, and real estate) are substantially more dispersed than construction and wholesale employment.
The statistical methods in the next section permit us to explore the explanations for these aggregate trends, and isolate the independent role of the latter (job suburbanization) on the former (commute length).
Two main implications of the theoretical discussion above bear on the empirical strategy.
First, if job decentralization causes greater uncertainty about future job location, average commute lengths may vary depending on worker occupations. (Job mobility may be higher for some occupations, such as construction workers, and lower for others, such as university professors.)
Also, those with higher expected moving costs, such as older households, larger households, and households with children in school, may also be more likely to have to take into account possible locations of future jobs when making a residential location. Therefore, both households with higher moving costs, and households with workers in high-turnover occupations, should be expected to have longer commutes in a decentralized metropolitan area in comparison to others.
Second, the benefits of decentralization may vary by industry. For example, the decentralization of manufacturing jobs may have been driven by the search for larger land parcels needed to carry out efficient operations given new technology, while over time the importance of firm clustering may be actually increasing in other industries such as software production, clothing design, and film making. Because the BEA data is available by the old-style Standard Industrial Classification codes, we can perform a crude test of the theory that different types of firm engage in different kinds of decentralization, and therefore have different aggregate impacts on average commute length of residents at fixed locations within a metropolitan area.
Based on conventional economic models of travel demand, such as developed in McFadden’s Nobel Prize winning research, we modeled individual commute distance as a function of various household and community factors. In addition to the variables described above—the share of metropolitan employment located in outlying counties, as broken out by broad industrial category, as well as proxies for moving costs, the presence of children and housing tenure status—demand factors included household income, the presence of more than one earner in a household, travel costs, tastes, and other standard factors.
As noted above, household income and housing rents are expected to be causally associated with commute distances. We control for this using a two-stage least squares procedure. Finally, we use pooled panel techniques to account for the fact that the panel data set consists of repeated observations over time on housing units.
The results of our analysis are shown in simplified form in Figure 3. The explanatory variables are listed in the left-hand column, and the estimated direction of their influence on commute length is given by the plus or minus signs in the next two columns. These latter two columns correspond to two versions of the statistical model, the first being a naïve pooled model estimated via Ordinary Least Squares, and the last via Two-stage Least Squares to account for the panel structure of the data and the potential endogeneity of housing costs in residential location choice.
The first several variables are conventional demand factors. Our main interest here is the sign of the estimated effect of employment suburbanization, which is negative in both the naïve and more elaborate model. The estimated coefficient on that variable is -0.335, implying that a five percent increase in the amount of employment in a metropolitan area’s outlying counties is associated with a 1.5 percent reduction in the average commute distance. This is equivalent to a reduction of about a tenth of a mile in our sample.
When the suburban employment share is broken out by industry some interesting patterns emerge (Figure 4). Construction and wholesale employment dispersal are associated with shorter commutes, while manufacturing and government employment dispersal are associated with longer commutes. Retail and service employment does not appear to be strongly associated with commute length.
These differences may be due to the pattern of clustering characteristic of these industries. Construction and wholesale employment may not be as clustered within a given county-level pattern of dispersion, while manufacturing and government employment may be. There is evidence that certain kinds of manufacturing firms (particularly, small manufacturers in the more technologically advanced industries) tend to cluster to realize Marshallian agglomeration economies. Meanwhile, retail and service firms do cluster to some extent, but because they are population-serving they tend to also be pulled out to follow the more dispersed pattern of the residential development they serve.
Our evidence supports the argument that decentralized employment is associated with shorter commutes on average. This is not to say that commutes are shortening as cities expand their footprint; indeed, they seem to be slowly lengthening. Contrary to some conventional wisdom, however, the marginal effect of job suburbanization appears to be to bring jobs and workers closer.
Even then, there are strong differences by industry. The suburbanization of construction, wholesale, and service employment is associated with shorter commutes, while manufacturing and finance deconcentration (weakly) explain longer commutes. These results may reflect a combination of industry agglomeration effects, differential job location stability by industry, and historical transitions.
What does this mean? First, our study is exploratory only. Many loose ends remain, suggesting numerous ways to refine and extend our understanding of these relationships.
For example, the AHS data do not allow us to test the determinants of commute times, since only commute distance is available over this period. This ignores the substantial role of congestion in urban form and behavior debates: If congestion is lower (higher) in suburban areas, the dispersal of employment to outlying counties within metropolitan areas might then reduce commute times more (less) than it does commute distance.
Larger questions remain as well. Neither the influence of urban form on travel behavior nor the merits of concentrated versus dispersed urban growth are well understood. The former is yet another complex set of nuanced behaviors awaiting better data and empirical strategies. Regarding the latter, we do not know how the social and economic costs of sprawl, however measured, compare with their benefits. Our work is but a small piece of this bigger puzzle.
Boarnet, Marlon and Randall Crane (2001) Travel by Design: The Influence of Urban Form on Travel, Oxford University Press.
Crane, Randall (1996) “The influence of Uncertain Job Location on Urban Form and the Journey to Work,” Journal of Urban Economics 39, 342-356.
Crane, Randall and Daniel Chatman (2003) “Traffic and Sprawl: Evidence from U.S. Commuting, 1985-1997,” Planning & Markets 6.
Crane, Randall and Roz Greenstein (2002) “Sprawl: What Don’t We Know and How Can We Know It?” UCLA, Los Angeles.
Glaeser, Edward and Matthew Kahn (2001) “Decentralized Employment and the Transformation of the American City,” in Gale and Pack (editors) Papers on Urban Affairs, Brookings Institution Press, Washington, 1-65.
Gordon, Peter, Ajay Kumar and Harry Richardson (1989) “The Influence of Metropolitan Spatial Structure on Commuting Time,” Journal of Urban Economics 26, 138-151.
Kahn, Matthew (2000) “The Environmental Impact of Suburbanization,” Journal of Policy Analysis and Management 19, 569–586.
Levine, Jonathan (1990) “Employment Suburbanization and the Journey to Work,” Ph.D. Dissertation, Department of City Planning, UC Berkeley.
McFadden, Daniel (1974) “The Measurement of Urban Travel Demand,” Journal of Public Economics 3, 303-328.
Wheaton, William C. (2004) “Commuting, Congestion, and Employment Dispersal in Cities with Mixed Land Use,” Journal of Urban Economics 55, 417-438