Fit Cities vs Fat Cities

City Differences in Obesity and Physical Activity

To test the hypothesis that more sedentary cities are also more obese, I used 2010 CDC data on BMI and self-reported sedentary behavior in nearly 200 US cities, which I found here.

I also wanted to look at what makes certain cities more physically active than others. To test the possibility that sedentary behavior is influenced by walkability, I looked up the walkability score for each city in the data set. I also used 2010 census data as compiled on statistical atlas to look at demographics and median income in each city, to test the possibilities that race and class influence sedentary behavior. I also recorded the latitude and average annual precipitation in each city to see if hot and/or wet climates promote sedentary behavior.

After removing cities for which high-quality data was not available, and also removing cities outside of the contiguous 48 states to keep Hawaii and Alaska from messing up the latitude comparisons, 186 cities remained in the final data set, which you can download here. Here they are, superimposed on a map of the USA:

Sedentary Behavior and Obesity

Within these 186 cities, the percentage of the population self-reporting zero physical activity is strongly correlated to the percentage of the population with a BMI>30. Differences in obesity rate between cities can largely be explained by differences in physical activity, and there’s a less than 0.001% chance of this correlation occurring due to random variations in the data set.

Based on this data, it’s plausible that virtually everybody who gets zero physical activity is obese. With a correlation this strong between obesity and physical activity, it’s interesting to know what attributes of a city might promote more or less physically active lifestyles. In the meantime, it’s also a good idea to make sure you’re not one of the people reporting zero physical activity.

City Planning

The first possibility I considered is that some cities are much more walkable than others, and that this may explain city-city differences in sedentary behavior. To test this hypothesis, I used walkscore.com’s walkability score, which ranges from 0 (can’t go for a walk at all) to 100 (never need to use a car or public transit). Here’s what I found:

The relationship I expected does show up, but it appears to be specific to low-walkability cities. Moreover, the correlation is fairly weak. It looks like there’s much more difference made by people’s choice to use or not use pedestrian infrastructure than difference made by the infrastructure being there to use. For cities with less than 50% walkability, there’s an inverse relationship between walkability and sedentary behavior (0.03% chance of correlation occurring due to noise).

For cities with greater than 50% walkability, there’s no detectable relationship between walkability and sedentary behavior. Notably, there were more than twice as many cities in the low-walkability group (n=136) than in the high-walkability group (n=50). This data seems to suggest that there’s such a thing as walkable enough, but that most US cities aren’t there.

Climate

I’ve personally found it much easier to get my daily run in since moving from Georgia (very hot and very humid) to Colorado (dry and temperate). I therefore hypothesized that southern latitude and high annual precipitation would correlate to more sedentary behavior. It turns out I was correct on both counts, but that the relationships are very tenuous:

You may note that there are some very active cities at very southern latitudes in the first plot. Those are mostly small resort towns in Florida, like Key West (latitude 24.6, 17% sedentary). In addition, although the precipitation vs sedentary behavior correlation is statistically significant, this relationship is highly variable and probably not noticeable in day-to-day life.

Sociological Factors

Poor people tend to have worse health outcomes than do rich people, so I hypothesized that median income in each city might correlate to the percentage of the population reporting zero physical activity.

The distribution for median income was highly skewed, so I log-normalized it. Bethesda MD was still so rich as to be a major outlier, so I removed it from this analysis. As hypothesized, there was a general trend for richer cities to be more physically active than poorer cities. I originally suspected that this could be confounded by poor cities being less walkable, but it turned out that there’s no detectable relationship between walkability and median income (data not shown).

People of color tend also to have worse health outcomes than do white people. The CDC has compiled extensive data sets on this disparity, and it’s very possible that some city-city differences are explained by demographic factors. I therefore also tested the hypothesis that the percentage of the population reporting zero physical activity might be correlated the percentage of the population comprised by non-Hispanic whites.

There was also a consistent, albeit weak, inverse correlation between percentage of the population comprised by non-Hispanic whites and percentage of the population reporting zero physical activity. These data suggest that white people may be more physically active on average than are people of color.

Multivariate Analysis

It’s very possible that the single-variable correlations I looked at above are confounded by mutual relationships between the predictor variables. For example, it’s possible that the percentage of non-Hispanic whites in a city only correlates to physical activity because both are correlated to median income. To correct for these sorts of possible confounders, I used a technique called multiple linear regression.

In this multivariate analysis, the three predictors of sedentarism that remained statistically significant were 1) median income (people in wealthier cities tend to be more active), 2) walkability (people in more walkable cities tend to be more active), and 3) racial and ethnic diversity (people in more diverse cities tend to be more sedentary). Collectively, these parameters explain 27% of the variance in sedentarism by city (R²=0.27).

After correcting for wealth, walkability, and demographics, southern latitude and high annual precipitation were no longer significantly correlated with sedentary behavior. Wetter southern cities also tended to also be the poorest, least walkable, and most diverse cities in this data set, and these relationships were confounders in the single-variable analyses discussed above.