Over the last two days I’ve introduced the concept of spatial diversity and discussed the role that it has played in the Supreme Court’s election law doctrine. Today I’ll explain how I quantified the concept, and I’ll then present some of my key empirical findings.
Since spatial diversity refers to the variability of a larger entity’s geographic subunits, I began by downloading newly available data for all Census tracts. Tracts are small spatial units with a population of about 4,000, and the data that I obtained covered vital areas such as race, ethnicity, age, income, education, profession, marital status, and housing. I then used a technique known as factor analysis to condense these raw variables into a much smaller number of composite factors. These factors capture much of the data’s original variance and reveal which of the raw variables, in which combinations, best explain the residential patterns of modern American life.
I then obtained scores for each tract along each of the factors. Next, I calculated the variances of these scores for the tracts within each congressional district. Finally, I created a weighted average of the variances, for each district, using as a weight the explanatory power of each factor. The end result was a single figure for each district that shows, with respect to a vast amount of information, how spatially diverse or non-diverse the district is.
The first thing that my scores allowed me to do was to identify the most spatially diverse districts in the country. Topping the charts is Illinois’s Seventh District, which combines Chicago’s prosperous Gold Coast with the poor and heavily black West Side as well as varied suburbs such as Oak Park and Maywood. Next on the list is my own district, New York’s Eighth, which merges the west side of Manhattan with Brooklyn neighborhoods like Borough Park and Brighton Beach. Districts of this sort are both troublesome from the standpoint of democratic theory and a potential sign of gerrymandering.
My scores also enabled me to rank states based on the average spatial diversity of their constituent districts. Among the larger states, California, New Jersey, and Texas have particularly high averages, while Ohio, Pennsylvania, and North Carolina have particularly low averages. This is prima facie (though not conclusive) evidence that the former states have worse district plans—more disruptive of communities, more suggestive of political mischief—than the latter.
Even more interestingly, I used my data to assess the claims (which I discussed on Monday) that spatial diversity is linked negatively to participation, representation, and competition. For participation, I examined the relationship between voter roll-off rate (i.e., the gap between districts’ presidential and congressional turnout rates) and spatial diversity. Even controlling for other variables that are known to affect turnout, I found that the most spatially diverse districts have roll-off rates about six points higher than the least spatially diverse districts. High levels of spatial diversity, that is, do indeed impair voter participation.
For representation, I looked at how well districts’ key characteristics and underlying partisan orientations explained their House members’ voting records. In the most spatially diverse districts, I found that key characteristics were a relatively poor predictor of voting records while partisan orientations were a relatively good predictor. But in the least spatially diverse districts I found the opposite relationship: Key attributes better explained House members’ votes than partisan leanings. The implication is that representation is better in the least spatially diverse districts—more responsive to constituents’ actual interests and less driven by sheer partisanship.
Finally, for competition, I examined the relationship between electoral responsiveness and the average level of spatial diversity in a state. (Responsiveness refers to the rate at which a party gains or loses seats given changes in its statewide share of votes.) I found that the more spatially diverse a state’s districts are, on average, the less responsive its elections are to changes in public opinion. The best way to increase competition, and to foil incumbent-protecting gerrymanders, is to draw more spatially homogeneous districts.
As astute readers may have noticed, none of the above discussion deals with the role of race in redistricting. Later this week I’ll present my empirical findings on that controversial topic.