Robert Draper in the latest Atlantic magazine says:
[R]edistricting today has become the most insidious practice in American politics—a way, as the opportunistic machinations following the 2010 census make evident, for our elected leaders to entrench themselves in 435 impregnable garrisons from which they can maintain political power while avoiding demographic realities.
Draper focuses special attention on redistricting in Texas and North Carolina as examples of partisan gerrymandering by Republicans, though he also mentions Illinois and Maryland as examples of gerrymandering on the Democratic side. (He also cites California as a gerrymander engineered by Democrats through the new Citizens Redistricting Commission. But this story, and the supposed Democratic seat gains it depends on, have been pretty thoroughly debunked.)
In a recent post in our series on the 2012 House elections, we concluded that the redistricting had no effect on the predicted partisan breakdown in the House this fall. But even if the overall effect is zero, what is the effect in individual states? Can we find evidence for the gerrymandering Draper mentions?
The graph below shows our predictions for individual states, organized by the size and direction of the Democratic seat gain or loss (h/t to Jonathan Kastellec and Eduardo Leoni for the R code that forms the basis of this graph):
We predict no change for 33 states. We do find evidence for the gerrymanders Draper mentions in North Carolina (-2 Democrats) and Illinois (+2 Democrats). But not much is happening in Maryland. And the largest Democratic gains were in Florida (3 seats) and Texas (2 seats), where Republicans were in control, while the party’s largest losses aside from North Carolina are in Massachusetts and New York, where the Democrats were in control (2 each). (h/t to Justin Levitt for his breakdown of party control.)* Why would a party allow itself to lose seats?
One factor we haven’t yet considered is reapportionment, which changed the total number of seats in each state, making it harder to maintain the status quo. To get a handle on this problem, we imagined how a bipartisan redistricting process would approach it. First, it would probably try to divide as many seats as possible evenly between the parties. A state with a two-seat gain would give one to each party, a state with a four-seat gain would give two to each party, and so on. We can treat these changes as “expected” and subtract them out. However, states with an odd number of seats lost or gained will always have one seat that can’t be split. We’ll treat this last seat as a pure redistricting effect and not correct for it. (The only exception is Massachusetts, where all the seats were held by Democrats so the state’s one-seat loss from reapportionment had to come from that side of the aisle.) The end result of all these adjustments looks like this:
New York and Massachusetts now look better for Democrats, but still a loss of one seat for each. Likewise, Texas and Florida look worse for Democrats, but they’re still holding their own or coming out ahead. The adjustment also makes Ohio look like a Democratic gain, since the state lost two seats but the Democrats are in shape to lose none.
Looking at these results, party control of the redistricting process has been a bust. Only three states with total one-party control produced plans with the expected partisan change: North Carolina (-2 Democrats), Oklahoma (-1 Democrats), and Illinois (+2 Democrats). At the same time, several states had one-party control that produced a gain for the other party: Massachusetts, New York, Arkansas, Ohio, and Florida. So we’re back to the original question: what gives?
Several possibilities come to mind. The simplest is just that our model is missing something. It only takes into account the presidential vote in the district and whether an incumbent is running, plus some national level factors. This captures a lot of what a gerrymandering party would have available to it, but not everything. Personal characteristics of the incumbent, detailed knowledge of the demographics and politics of the district, as well as an understanding of strong potential opponents who also live in the area would all facilitate drawing a more effective gerrymander. Not to mention the dynamics of the ultimate campaigns run in each district, which cannot be known at the time of the redistricting but might dictate a different outcome in specific cases.
But there are a couple substantive reasons to expect this result. First, a redistricting party doesn’t have free reign to do as it pleases. Districts must be equal in population (the allowed deviation between districts is vanishingly small), the Voting Rights Act constrains activity in many states, and some states also have specific rules that limit the options. (Along these lines, Florida voters recently imposed boundaries on what the legislature could draw, and Texas’s lines ended up being drawn by the courts.) Moreover, a party’s desire for more districts can often conflict with its own incumbents’ desire for safe reelection. And a redistricting party must work with the state it has, not the one it wants. Republicans and Democrats may not live in close enough quarters to permit the sort of seat-maximizing gerrymander a party would otherwise want.
Finally, these numbers may reflect the core reality of this election: Republicans are at historic highs and don’t have much room to grow. Ohio is a perfect example of this. The state is a toss-up and even leans Democratic in the presidential race, yet Democrats currently hold only 5 of 18 seats. The state lost two seats from reapportionment; what are the chances Republicans could bring the Democrats down to 3 out of 16? You can’t squeeze water from a stone.
None of this means we should ignore the signs that some states, like North Carolina or Illinois, have gerrymandered, or that we might not have more signs of gerrymandering by looking carefully at specific states. It’s just to say that gerrymandering, broadly considered, is not a big part of the picture.
*Because our model has error, the sum of all these predicted changes is a one-seat loss for Democrats rather than the aggregate one-seat gain we predicted before. But this is certainly within the expected range.