Every two years the House elections don’t go the way partisans think they should. Every two years partisans on the losing side place the blame on gerrymandering, the purposeful drawing of district lines to achieve a specific outcome. Every two years there is a new study showing that gerrymandering is not the problem. So here we go again.
Jowie Chen and David Cottrell, of University of Michigan and Dartmouth respectively, published a study (PDF) just before the election last year showing that gerrymandering has almost no effect on the overall makeup of the US House of Representatives. Previous studies have shown a significant or minor one, but Chen and Cottrell easily point out the flaws in most of those, especially in how they create a control group. This new study eliminates those inherent biases by using computer models to create a number of different nonpartisan district maps.
As a baseline, they used the 2008 election results at the precinct level.
We map the votes of these precincts to Florida’s 484,481 Census blocks according to population and then aggregate the votes into a set of 15,640 similarly-populated square polygons so as to produce a geographically-precise spatial grid of the state. These 15,640 “squares” of the grid are then used as the building blocks for the districting simulations.
The computer then chooses precincts at random to begin each district and adds more contiguous precincts until a district-size population is reached.
Without knowing the demographic makeup of each precinct the computer models avoid the various biases present in any manmade districting effort. This necessarily leads to possible outcomes where racial representation concerns are not considered, but the point of the study is to determine if that changes the outcome. It doesn’t, overall.
The results of the simulations were staggering…in that almost nothing changed.
After conducting 200 tests, each using a different computer-generated map of every state where data was available, the number of Republican seats in the House only changed by three at most. The vast majority of simulations fully devoid of partisan districting showed Republicans losing only two seats. Not exactly the overwhelming partisan advantage that gerrymandering is alleged to provide.
At the state level, however, the results are very different.
Many states produced a significant difference in the share of votes each party had compared to the actual 2008 outcome. Several states had more Democratic seats than they should have, and several had more Republican seats than they should have, though most states had an actual ratio within the simulated range, and almost none of them were off by a whole seat or more.
The red X is the actual 2008 election result, and each tiny grey dot represents one computer simulation result for that state. Most actual results fell within the simulation totals, meaning using nonpartisan district maps would produce the same share of Republican and Democratic seats in that state’s House delegation.
But in a few states, like California, Wisconsin, Arizona, and Louisiana, the actual election results were completely outside what the simulations produced. It’s worth noting that in all but one case, even these skewed state results only produced one more seat for the advantaged party than otherwise would have. California, on the other hand, has at least two more Democratic representatives than computer models suggest it should.
Even with these state-level advantages taken into account, there is no significant overall advantage in the House. Skewed state delegations don’t affect congressional control if they’re balanced by skewed state delegations on the other side.
It is simply the case that the effect of gerrymandering is small and that removing bias from districting process - whether it is racial or partisan - is not likely to change the partisan composition of Congress.