Sometimes I feel a little like Captain Ahab, forever in search of an elusive white whale. In my case, though, the whale is profoundly geeky: I’m in search of a definitive study, or set of studies, showing the relationship between urban design and transportation habits—particularly, how neighborhood design affects fuel use.
So far, that particular white whale remains elusive—but searching for it turns up all sorts of interesting tidbits. Like this one: University of California researchers David Brownstone and Thomas Golob have looked at the relationship between residential density and driving habits, and concluded that:
Comparing two California households that are similar in all respects except residential density, a lower density of 1,000 housing units per square mile…implies an increase of 1,200 miles driven per year…and 65 more gallons of fuel used per household.
Thar she blows!!
Let’s look at those numbers a bit.
A thousands households per square mile translates into about one and a half households per acre. So going from a neighborhood designed on the post-war, upper middle class ideal—your own home on 2 private acres—to the reality in many of the Northwest’s more compact urban areas—a mixture of single family homes with small yards, together with some multifamily housing, with an average of around 10 housing units per acre—you increase density by just over 6,000 housing units per acre.
And, according to the numbers that these authors have crunched, living in a compact neighborhood rather than a sprawling exurb would lead to a decline in gasoline consumption of…wait for it…395 gallons of gasoline per household per year!
That’s a lot of gas. By comparison, the average resident of the Northwest states consumes about 390 gallons per year; so living in a denser neighborhood does as much to reduce your driving as having one fewer person in your household.
So if Brownstone and Golob are even close to being correct, the kind of neighborhood you choose has a tremendous influence on your total gas consumption, and the overall impact of your daily driving on the climate. And their findings argue that one way to reduce fuel consumption is to encourage new development in compact neighborhoods.
This, of course, is just one study among many. There are some researchers who find that the relationship between density and driving is far more tenuous. Still, these sorts of results are what keep me searching for that white whale. (It’s there…I know it…)
Photo courtesy of Flickr user Skelly B under a Creative Commons license.
Jessica
Hold on a minute, Clark. You are comparing a per household reduction with per person total use. I think an apples-to-apples comparison will still show a large improvement but be a little more accurate. Also, what about running your calculations comparing the density of individual Northwest neighborhoods? Finally, did the authors take into account the effects of pure residential vs. mixed residential and commercial?
Branden
If you look up Robert Cervero, you’ll find a lot of studies that he’s conducted on the issue of density and VMT.
morgan
Jessica – I believe Clark is comparing apples to apples by comparing gallons to gallons—as in different things a 390 gallon reduction is equal or similar to. Also, density is a common proxy for development types, because it has been so widely measured for a while, where as specific types of built environments have not.Clark – I think you cherry picked from the results. This study explores primarily how different people/families choose to live in different densities, and I think they are mainly trying to say that it’s the differences in the people (income, family size, etc) that are driving fuel consumption patterns rather than the density itself:”Aggregate studies examining the bivariate relationship between vehicle miles traveled and density find a large significant inverse effect (see Newman and Kenworthy, 1999). These studies are flawed because they do not account for the possibility of residential self-selection, which is the tendency for those households that prefer non-private vehicle travel to locate in dense areas with more transit and shorter trip distances.””Once we included a complete set of socio-demographic control variables, we could not reject the hypothesis that there are not significant self-selection effects.” The authors also make some disappointing assertions:”The final section concludes and argues that the impacts of increased residential density are too small to make increasing density a relevant policy tool for trying to reduce VMT or greenhouse gas emissions from residential vehicles.” ouch!Adding salt to the wound, they continue:”… One of the best of these is Bento et al. (2005), which used the 1990 National Personal Transportation Study to build disaggregate models of number of vehicles per household and vehicle miles traveled (VMT) per vehicle. They supplemented the density measures in the data with road density, rail and bus transit supply, population centrality, city shape, jobs-housing balance, population density, land area, and climate. Bento et al. (2005) found that the magnitudes of the impact of any of their built environment measures were frequently statistically insignificant and small in magnitude.” ouch!!!
Michelle
Hmmmm… Maybe there’s no white whale in the line of sight, yet. But with the cherries and salt, we could probably make some darn good chutney! 😉
Clark Williams-Derry
Morgan – First, the Bento study is (a) one of the studies I mention at the end of the post that I allude to—ones that find very weak relationships between neighborhood characteristics and driving (b) part of Golob & Brownstone’s lit review; it’s the study THEY WERE REACTING TO when analyzing the same data set in new ways. Basically, Bento’s analysis of NPTS data found no strong relationship; but G’s analysis, using different methods and models, did find a strong relationship. That tension is what keeps me searching for the white whale…Second, I think that B’s abstract is simply wrong, and not supported by the text later in the document. In their conclusion section, they emphasize increasing residential density is *difficult*—as evidenced by a (rather cursory) review of US urban areas. Their quote: “Unfortunately for those wishing to use land use planning to control residential vehicle use, it is very difficult to increase the density of an established urban area by 40%.” So in their view, it’s the high difficulty of increasing density, compared with the medium-sized magnitude of the impacts, that makes land use a weak tool to deal with GHGs.They may be right—yet Sightline’s analysis of census data in Metro Vancouver, BC shows that the average density of residential areas increased by 29% in just 15 years (between 1986 and 2001). [I can’t put my hands on comparable 2006 data—we had a server glitch a few months back that scrambled our files :(. But I have little doubt that average density increased still more, though the pace may have slackened a bit.] So clearly, difficult is not the same as impossible; and the difficulty depends to a large degree on politics, which are hard to predict. And besides, it seems to me that their research leads to the OPPOSITE conclusion that they reach—namely, that the clear energy benefits of compact development ought to inform planning decisions about where to put NEW development—low-density exurbs will increase fuel use, oil dependence, and vulnerability to oil shocks, compared to compact development and infill.
Clark Williams-Derry
Jessica -AS far as I can tell, they didn’t include any land-use metrics in their analysis other than density. Residential density likely winds up as an imperfect proxy for other neighborhood characteristics, such as parking, block size & connectivity, mixed use development, etc.Second, they did control for household size, along with income, education and race; they found that for “two California households that are similar in all respects except residential density, a lower density of 1,000 housing units per square mile (roughly 40% of the weighted sample average) implies an increase of 1,200 miles driven per year (4.8%) and 65 more gallons of fuel used per household (5.5%).Lastly, I’d be interested in running those estimates—if only I could find the time!! (Should be simple enough, in theory, but this sort of thing always takes longer than you expect…)
Michele
what about the overall size of the urban area—what would it look like if you compared two households that are similar in all respects including residential density but one was in San Jose, California and one was in Corvallis, OR?
Todd Litman
There is considerable research on how various land use factors (density, mix, walkability, transit accessibility, road and path network connectivity, roadway design, parking supply, etc.) affect travel behavior, as discussed in our report, “Land Use Impacts On Transport: How Land Use Factors Affect Travel Behavior” (www.vtpi.org/landtravel.pdf). I agree that density is often a surrogate for other factors, such as land use mix and walkability, and so may have less impact on per capita vehicle travel by itself than Brownstone and Golob’s analysis may indicate. However, there is very good evidence from many different sources indicating that a combintation of land use policies (increased density, mix, connectivity, walkabilty, parking pricing, etc.) can reduce per capita vehicle travel 20-40%, and even more if implemented with regional transportation policy reforms, such as major investments in high quality public transportation.Yes, a portion of the differences in travel activity between multi-modal and automobile-oriented locations results from self-selection (people who, due to necessity or preference, use alternative modes more than automobile travel tend to locate in transit-oriented areas), a number of studies have examined this affect and conclude that it accounts for just 20-40% of the differences in travel. See:Xinyu Cao, Patricia L. Mokhtarian and Susan L. Handy (2008), Examining The Impacts of Residential Self-Selection on Travel Behavior: Methodologies and Empirical Findings, Institute of Transportation Studies, University of California, Davis, Research Report UCD-ITS-RR-08-25; at http://pubs.its.ucdavis.edu/publication_detail.php?id=1194; also Report CTS 08-24, Center for Transportation Studies, University of Minnesota (www.cts.umn.edu); at http://www.cts.umn.edu/Publications/ResearchReports/reportdetail.html?id=1684. Robert Cervero (2007), “Transit Oriented Development’s Ridership Bonus: A Product Of Self Selection And Public Policies,” Environment and Planning, Vol. A, No. 39, pp. 2068-2085; at http://www.uctc.net/papers/765.pdf.