I just posted a new article on SSRN, The Causes and Consequences of Gerrymandering, that will soon be appearing in a terrific William & Mary Law Review symposium. The piece uses nearly fifty years of data at both the congressional and state house levels to analyze (1) the factors that are responsible for district plans’ partisan skews; and (2) the implications of these skews for legislative representation. I’ll talk about causes in this post and about consequences in another post tomorrow.
To start off, one common hypothesis is that institutional control of the redistricting process is related to plans’ partisan tilts. In other words, we might expect Democrats to design pro-Democratic plans, Republicans to craft pro-Republican maps, and other institutions to be more neutral. In my article, I find very strong support for this hypothesis. Controlling for other variables, unified Democratic government is associated with a significant pro-Democratic shift in the efficiency gap (and partisan bias), and unified Republican government is linked to a significant pro-Republican shift. But when commissions, courts, or divided governments are responsible for redistricting, there is no consistent movement in either party’s direction.
Next, one often hears the argument that greater minority representation benefits Republicans. The logic is that the districts that elect minority-preferred candidates tend to be heavily Democratic. So creating more of these districts inefficiently “packs” Democrats and assists Republicans in the rest of the state. I find little evidence for this claim in my piece. There is no relationship at all between Latino representation and the efficiency gap (or partisan bias). Greater black representation does yield a Republican advantage, but only a very small one: on the order of a percentage point as black seat share rises by a standard deviation. And even this effect evaporates when Democrats are in charge of redistricting (and avoid packing their supporters).
Still another explanation for plans’ partisan skews is political geography. Here the idea is that Democrats are heavily concentrated in cities while Republicans are more evenly dispersed in exurban and rural areas. Maps that reflect these spatial patterns will allegedly tilt in a Republican direction. I also find little support for this hypothesis in my article. The best available proxy for political geography, a state’s level of urbanization, is unrelated to the efficiency gap (or partisan bias) at the congressional level. And while there is a statistically significant link at the state house level, it is substantively small and mitigated by party control of the redistricting process.
Two last points: First, more academics who study redistricting should be doing this kind of work. For years, scholars have been obsessed with measuring partisan gerrymandering. As a result, there hasn’t been enough progress in understanding the drivers and implications of this activity. Second (and relatedly), my results did not change significantly when I substituted partisan bias for the efficiency gap in my models. I suspect they also would not change much if I used the mean-median difference, the difference between the parties’ average margins of victory, or any other gerrymandering metric. This is all the more reason for the academic debate to turn from measurement to other issues.