Obesity and the Built Environment

Planners have asserted for years that the built environment affects health, claiming statistical correlation between mixed land uses and lower obesity rates. This finding is important, because it casts the blame for American obesity on the suburban landscape. Planners now have another tool in their arsenal arguing for better land use mix (a cause that I support). As the thinking goes, people who are able to walk to activities will do so, and thereby have lower obesity rates than the general population.

I just finished reading a paper (by my advisor and her former doctoral student in Transportation) that critiques the methodology in at least one of these studies, a 2004 study by Frank, Andersen, and Schmid (American Journal of Preventative Medicine) relating obesity to travel patterns in Atlanta. The authors constructed a linear regression model based on a household travel survey of Atlanta residents. They showed statistically significant positive correlation with age (older people are more obese), in-vehicle travel time (obese people have longer commutes), and being black. Statistically significant negative correlation was seen in education (people with higher incomes are less likely to be obese), income, daily walking distance, uniform land use patterns, and being a woman.

My advisor's analysis examined the AJPM model and improved on it by introducing non-linear relationships. For example, people tend to gain weight until they reach about 60, and then lose it again. So using age as a strictly linear variable misses the impact that late age weight loss may have on obesity. This and other adjustments increase the ability of the model to predict obesity far more than improving the prediction of the land use field. So while people who live in mixed-use areas are obese less frequently than the general population, land use by itself is a poor predictor of obesity.

What are needed are studies that carefully measure both travel behavior and Body Mass Index. But the studies have not indicated that health records should be a major focus of travel survey collection. Indeed, the Transportation paper indicates that shifting the focus of a travel survey to health information can be a very risky proposition, because you may end up under-sampling more important variables, like actual travel behavior.

Pac-12 Graphics

I'll be updating this more regularly. I'm also expanding the blog to cover my thoughts on graphical communication.

With the University of Utah and the University of Colorado joining the PAC-10 athletic conference next summer[1], the conference needed to split into divisions. Numerous alignments were proposed, but the university presidents and athletic directors eventually settled on a “North/South” alignment. To advertise the split, the Deseret News (Salt Lake City) produced the following graphic:

But the graphic distorts many items to give the illusion that this is a clean geographic split. First, the logos of several schools are deceptively placed. Utah is in Salt Lake City, and Colorado is in Boulder, both cities which are much closer to Wyoming than Arizona. In addition, Stanford and California are each shown further north than Palo Alto and Berkeley really are. Finally, the conical projection used for the state lines tends to cause California, Oregon, and Washington to bend away to the northwest.

As a UCLA fan blog noticed,

---

[1] As a fan and alumnus of Brigham Young University, I am of course jealous that BYU was not invited into the PAC-12. This disenchantment has been set aside in the current defense of graphical integrity.

A correction...

So, my analysis yesterday had some error in it. The database software I was using (ArcGIS) had difficulty calculating negative changes in density. The histograms above are right, and the ones below are wrong.

The "Marta Stations" histogram now includes all census tracts within 1/2 mile of a Marta station, not just tracts containing a station. This didn't actually change a whole lot. The trend is still mildly negative, but statistically insignificant.

A potential explanation: "density" in this field is purely based on population. Thus if a neighborhood of old, one-story homes is demolished to make way for a bigger development (Edgewood or Atlantic Station, for instance) the density will drop, while the actual activity will increase. Really what I need to do is sum households and jobs in a district, and compute density from that.

Impact of Transit Stations

It is widely suggested that population and density increase in neighborhoods close to transit, and this is one of the most attractive arguments for improved transit service. After all, if building a new rail line can increase property value and promote sustainable development patterns, why not do it?

But a bigger question might be how long this trend continues. After all, if policies are not changed to discourage sprawl, the long-term investment in rail transit might be overstated. The red histogram below shows the change in density over all census tracts in Atlanta from 2000 to 2009 (see thelast post). The blue histogram is the change in density for census tracts with a MARTA station.

So, is there a difference between the two? Not really. A statistical analysis reveals that the probability of the means being equal is about 30%. This fails the significance test. But this should be a relief for the transit supporters, because the mean, median, and skew of the MARTA station data all indicate that rail stations actually had a negative impact on density from 2000 to 2009.

Is this the end of the story? Not really. I will do this again using a more robust method for selecting the census tracts where MARTA stations lie. As it is, sometimes the tracts have a considerable amount of activity away from the station, whereas the most natural activity centers from the stations are actually in a different tract.

But still, the preliminary data seem to indicate that Atlanta has not pursued strategies to take advantage of the benefits of MARTA.

Data in this analysis came from the Atlanta Regional Council website.

Transit Service

In my previous job at the Utah Transit Authority, I was asked to look at travel times on transit compared with travel times by car under different service scenarios. The idea was, "Will changing the bus network substantially change the travel time by transit, and make transit more or less appealing?"

The images show competitive transit trip origins in Provo (most are headed to Brigham Young University). A trip is "competitive" if the total door-to-door trip time is less than 150% of the automobile time. That is, a transit time of 45 minutes would take 30 minutes or more in a car to be considered competitive.

The blue lines are under the existing UTA bus network, and the red lines are a proposed service network change, with essentially no increase in revenue miles. We found we could increase the competitive trips by about 15% with no substantial cost outlay. While this didn't always transfer directly into more transit riders in the model, there is clearly a potential to do this. (Mode choice models incorporate much more than travel time into the decision parameters, but perhaps a better service could begin to change behavior.)

This post is in partial response to the DC Metro's travel speed map on Streetsblog yesterday morning.

Emptying the City

This map shows population change in the Atlanta region between 2000 and 2009. Red, orange, and yellow regions represent places where the change in population was negative, and green-blue represent population increases.

According to the 2000 Census and a 2009 estimate update, just over 1 million people moved to Atlanta in that nine-year span. The map at the left indicates that almost all of them moved into the suburbs. The fact that several dozen zones actually lost population indicates that people move to the suburbs regardless of available housing in the urban core.

Certainly many areas inside the perimeter increased population. But the trend is clearly to the outside. Why is this a bad thing? Most jobs are still in central areas, meaning that suburban residents have to drive further to reach jobs and entertainment than urban residents. This complicates traffic congestion, parking, and city services.

This could represent some good things, though. The data could also show urban families moving to the suburbs in pursuit of better education for their children and to escape more intractable social ills. But does the increased commute time and cost negate these other benefits? More on this later.

Data in this post came from data available on the Atlanta Regional Commission website, http://www.atlantaregional.com/info-center/gis-data-maps

Ground Rules

So, here it goes. But I'm going to lay out some ground rules.

First, this blog is going to be about Atlanta, numbers, policy, and the space between all three. This is not going to be about me. I'm tired of narcissistic blogging.

Second, this blog will chronicle stuff as I learn them. I write this for me, not for the world. That said, I hope to have fellow grad students post here as guests with some frequency.