In our ongoing quest to discover how land use and urban form links to human health effects, I recently stumbled across something odd. It’s a 2000 study of vehicle emissions per household in Puget Sound, authored by Larry Frank. I wanted to find out if there is a connection between air pollution and urban density. According to this study, there is, but in a way I didn’t expect.

It turns out that the strongest land-use correlate to low household emissions is not residential density, but job-site employment density. That is, from a statistical standpoint, it matters less whether you live on Capital Hill or the Sammamish Plateau than whether you work in downtown Seattle or Bothell. The difference, I suppose, is that downtown Seattle and other places with high employment density are well-served by transit and are generally easier to get to with lower vehicle emissions than more far-flung workplaces.

Interestingly–this is only interesting if you’re a geek; otherwise skip to the next paragraph–the drop in household emissions does not observe a linear relationship with employment density. For the lowest three quartiles of employment density household emissions are about the same (they’re a little higher in the lowest density quartile), but then they drop off sharply at the beginning of the highest density quartile. This suggests that there’s a threshold of employment density—perhaps the density at which transit, carpooling, etc become viable—after which emissions drop quickly.

  • It’s also interesting, I think, that in this study residential density is less strongly correlated with lower household emissions. There is still a correlation—higher residential densities meant less vehicle emissions—but the difference, while significant, was relatively minor.

    One reason perhaps emerges in another set of correlations. This study found that households located in census tracts with high employment density, greater mixes of land-use, and greater street network density—in other words, places with many characteristics of city living—actually generate more vehicle trips and more vehicle trips with a cold engine (which produces a disproportionate share of tailpipe emissions). Probably, this is because there are more services and amenities nearby and there’s less incentive to "chain" trips together as a typical suburban commuter might on the way to or from work. Even so, the higher density households produce fewer emissions simply because the trips are not as long as for households in lower densities.

    There’s a lesson here, maybe, for those of us interested in urban form as well as everyone who’s interested in improving air quality. From a public health perspective, it may make more sense to concentrate jobs in dense nodes with good transit access than to worry about other land-use features. Maybe the best reform to reduce vehicle emissions is more office space downtown.

    About the study: The study uses an exhaustive (heh, heh) methodology that calculates three types of emissions (NOx, CO, and VOC) that accounts not only for driving distance, but also for speed, travel time, and emissions from starting the car (adjusted for estimated engine temperature at start). Its findings are based on data from the Puget Sound Transportation Panel Travel Survey, which records travel for 1,700 households over a two-day period by giving each member of the household over 15 a diary for recording trips and their characteristics.

    UPDATE 8/10/05: Here’s a link to an abstract of the study. As far as I know, the full version is not online.