A Broader Context for Land Use and Travel Behavior, and a Research Agenda

Abstract

Problem: Planning studies of land use and travel behavior focus on regression analysis of travel as a function of traveler demographics and land use near study subjects’ residences. Methodological debates have tended to focus almost exclusively on the possibility that persons choose their residence based on how they wish to travel. This longer view steps back from the confines of the regression-based literature to explain the historical roots, methods, and results of the literature, and to assess how the land use–travel literature must be transformed to be more relevant to planning.

Purpose: There are many summaries and meta-analyses of the impact of land use on travel. The goal here is not to understand how we might better specify a regression or summarize the results of past studies, but rather to explain how a literature that has become fundamental to planning scholarship is failing to be sufficiently planning focused. At the same time, this longer view describes how the literature can be transformed to address the planning challenges of today and tomorrow.

Methods: This longer view summarizes over 100 articles, covering transportation methods from the dawn of the interstate highway era to topics that include program evaluation, land development, and cognitive aspects of travel behavior. The primary focus is on the land use and travel literature, but the review and analysis is broad ranging and places the literature and its challenges within the broader context of recent developments in the social sciences, planning, policy, and electronic data collection.

Results and conclusions: This longer view elucidates three research frontiers that will be necessary to move the land use–travel literature forward. First, behavioral models of land use and travel must expand to consider how land is developed, how places are planned, and how cities are built. Second, the land use–travel literature should build a robust retrospective program evaluation tradition, which is currently almost completely absent in a scholarly field dominated by cross-sectional hypothesis tests and forecasting models. Third, economic social welfare analysis must be carefully researched, including questions of preferences for neighborhood types and whether such preferences are fixed or malleable.

Takeaway for practice: Planning is about city building, and the literature and practice on land use and travel behavior should adapt to better support city building. This requires both a serious commitment to social science research and planning's characteristically broad view of context, problem, and place. In an era of climate change, and amidst debates about sustainability, the land use–travel literature must more aggressively examine the process of plans and place making, evaluate the increasingly innovative transportation policies being implemented at the local level, and develop methods that allow more informed discussion about the costs and benefits of transportation policies.

Research support: None.

Keywords:

• land use
• travel behavior
• transportation

Acknowledgments

He has studied land use-travel behavior relationships, urban growth patterns, and transportation planning. Effective January 1, 2012, he will be professor of policy, planning, and development and director of graduate programs in planning at the University of Southern California.

Notes

1. More advanced models iterate back to the land use allocations (e.g., zonal allocations of residents and jobs), which drive the four-step model. See, for example, Waddell (2002) or Hunt and Abraham (2003), and Putman (1983), for an early example.

2. Criteria pollutants are the six pollutants (ozone, particulate matter, carbon monoxide, nitrogen oxides, sulfur dioxide, and lead) regulated by the U.S. Environmental Protection Agency as part of the National Ambient Air Quality Standards (NAAQS). See U.S. EPA, “Six Common Air Pollutants.” Retrieved May 2011 from http://www.epa.gov/oaqps001/urbanair/. The phrase “criteria pollutants” is often used to distinguish the six NAAQS pollutants from greenhouse gas emissions, and we follow that general convention here.

3. The antecedents of the modern focus on land use as a transportation policy tool are, however, arguably as old as the field of urban planning. For a discussion, see Zegras (2010). Treating land use as a prescriptive transportation policy tool has not so much emerged in the past two decades; rather, it has substantially increased in focus. My interpretation is that the transportation field, broadly speaking, was more reactively focused on forecasting travel flows before the 1980s, and that the increasing emphasis since then on land use as a determinant of travel behavior differs in magnitude, method, and intended policy application from the role of travel models in the first two to three decades of the U.S. interstate highway era.

4. Daniel McFadden, with co-authors at times, pioneered a behavioral treatment of travel behavior based on discrete choice models derived from a random utility framework. See, for example, Domencich and McFadden (1975); and McFadden (2001, 2007). McFadden won a Nobel Prize in Economics (technically the Bank of Sweden Prize in Economic Sciences in Memory of Alfred Nobel) in 2000 for his contributions to the development of discrete choice econometrics, but the computational complexity of the models and the sparseness of individual travel data (both especially daunting in the late 1960s and 1970s when discrete choice models were first developed) reduced the use of choice-based, micro-data models in practice. The modern land use–travel behavior literature has moved beyond the four-step model's decidedly nonbehavioral approach by focusing on individual data, but as is discussed in the section “Efforts to Move from Reduced Form to Structural Models of Land Use and Travel Behavior” of this article, the typical land use–travel behavior research approach does not incorporate the full range of behavioral choice suggested by McFadden's work; nor has the literature agreed on how best to address the full range of behavioral aspects of transportation.

5. Here we use the term “land use” to refer to all measures of the built environment, not just those measures that might strictly relate to zoning and land use arrangements, but also measures of infrastructure, development patterns, spatial arrangement of all aspects of the built environment, and design treatments and aesthetics.

6. Entropy indices are calculated as shown below, where P_j 5 the proportion of land in the j^th land use type, and J is the number of land uses. The dissimilarity index can be calculated as where K 5 number of developed grid cells in the larger geographic area, j indexes grid cells, and i indexes the eight grid cells that abut (or border) a grid cell when units are divided into a rectangular grid, with X_i 5 1 if abutting grid cells have differing land uses.

7. As an example, approximately 24% of the households in the subsample of the 2000–2001 California Statewide Household Travel Survey analyzed in Heres-Del-Valle and Niemeier (2011) took no car trips on the survey day. Approximately 19% of the households in the Southern California Association of Governments 2001–2002 travel diary for the five-county metropolitan Los Angeles area reported no car travel during the diary period (Boarnet, Houston, Ferguson, & Spears, 2011).

8. The distinction between structural and reduced form models is common in economic analysis. Structural models allow statistical estimation of the underlying behavioral parameters, while reduced form models statistically estimate relationships that do not reflect the underlying economic structure but which, nevertheless, may be useful to analyze particular policy questions. For a discussion, see econometrics textbooks such as Davidson and McKinnon (1993, chap. 18), Gujarati (2003, chap. 19), and J. Johnston and DiNardo (1997, chap. 9).

9. For a recent discussion of demand theory applied to land use and travel behavior, and a suggested method to use path choice to identify land use effects, see Crane and Guo (2011).

10. As before, the coefficients are scalars or vectors as appropriate, with coefficient vectors shown in bold.

11. McFadden's (2007) example suggested that votes on transportation projects might be shaped by imitative behavior, with persons being influenced by, for example, a small group of vocal opponents of congestion pricing. Yet, the generalization to travel behavior is straightforward. If most of an individual's acquaintances drive three blocks rather than walk, perhaps persons then view driving such a short distance as a sensible choice.

12. GHG emissions are not directly linked to VMT. In general, GHG emissions measurement would require information about tailpipe emission factors and driving behavior, including both VMT and characteristics that influence emissions, such as speed and acceleration/deceleration, and vehicle fleet composition. Such data are sometimes sketchy, and in some cases (e.g., real-time travel behavior) only rarely available. For that reason, VMT is often currently used as a proxy variable.

13. Handy (1993) popularized the distinction between neighborhood and regional land use measures applied to travel behavior.

14. See Naess (2011) for a similar result, namely that metropolitan-level land use variables are more important than neighborhood-level variables for car travel, based on data from Copenhagen.

15. For a discussion of this question and the debate on the causal role of highway infrastructure in U.S. urban development patterns, see Boarnet and Haughwout (2000).

16. See, for example, Card and Krueger (1994) and Dehejia and Wahba (1999, 2002), for general discussions of quasi-experimental control group matching methods.

17. For an overview of the potential for locationally aware technologies and other data sources to transform social scientific analyses of behavior in space, see Miller (2010).

18. The before-after, treatment-control group studies that have appeared in the land use–transportation literature have more often evaluated the effect of transportation infrastructure changes on land use patterns. See, for example, Cervero, Kang, and Shively (2009); Chalermpong (2004); or Funderburg, Nixon, Boarnet, and Ferguson (2010). Some studies of travel behavior have examined travel changes before and after residential moves (e.g., Handy, Cao, & Mokhtarian, 2005; Krizek, 2000, 2003b) but such studies are not evaluations of transportation infrastructure changes, as is suggested here, but rather studies of travel changes coincident with residential relocations. The program evaluation approach advocated here is better exemplified by the Boarnet, Anderson, et al. (2005) before-after study of travel changes associated with the California Safe Routes to School program or the Krizek, Barnes, and Thompson (2009) study of travel changes associated with new bicycle infrastructure.

19. See, for example, the studies summarized in Ewing and Cervero (2010); Boarnet and Handy (2010); Spears et al. (2010); and Handy, Tal, and Boarnet (2010a, 2010b).

20. On this last point, Boarnet, Houston, Ferguson, and Spears (2011) give evidence that the magnitude of land use–travel associations varies across different land use characteristics and hence locations in the greater Los Angeles region.

21. Integrated land use–transportation modelers are already working hard on this question. See, for example, the discussions of UrbanSim and the PECAS model in, respectively, Waddell (2002), and Hunt and Abraham (2003).