John, Ben Highton, and I predicted that Democrats will get roughly half the vote but only about 45% of the seats. To what extent is this a product of the redistricting? If not for the redistricting, would Democrats be doing better?
We approach this question by asking what our model would predict using the old districts, as well as the incumbents sitting in those districts at the beginning of 2012. In other words, assume that the current crop of candidates was running but the 2011 redistricting had never happened. Comparing our original prediction to this counterfactual incorporates both forms of partisan redistricting effects: new district lines and the “displacement” effects of some incumbents forced to retire or run in uncomfortable territory.
Given these assumptions, our median seat prediction using the new districts was 194 Democratic seats, and the median prediction assuming the redistricting never happened is…194 seats. That’s it. The redistricting has no impact on our predicted seat share.
There are certainly reasons to have expected a different outcome. Republicans took control of the redistricting process in a number of states as a result of the 2010 election. The initial assumption seemed to be that Republicans would use this control to expand their majority (see here, here, or here). Yet since Republicans now hold more seats than in any Congress but one since before the Great Depression, it probably makes more sense to think they would protect the gains they have already made (see here and here for concurring views). That in turn would suggest the Republican seat share would remain the same even as support for Democrats grew—exactly what we’re showing.
But the changes to the districts, on balance, have been pretty small. The graph below has the distribution of the 2008 presidential vote in the 435 House districts in 2008 and 2012. (We once again rely on the Daily Kos presidential vote numbers for the new districts.) “Strong” districts have a presidential vote more than 10 points off the national average, “Lean” is between 5 and 10 points off the average, and “Tossup” is within 5 points. There is not a lot of change here: the number of seats in each category is roughly the same before and after the redistricting.
If there is a factor that explains the gap between vote share and seat share, it is more likely to be incumbency than redistricting. Congressional elections are not written on a clean slate. Instead, a substantial number of voters give their incumbent the benefit of the doubt unless offered ample reason to do otherwise. That means the party with more incumbents is going to do better, especially in a status quo election with no signs of a broader partisan tide.
There are a whole lot more Republican incumbents now than there were two years ago. That gives the party an important advantage that helps them keep seats they might otherwise lose. How important? If we take the new redistricting as given but pretend that no incumbents ran for reelection (an admittedly unrealistic scenario), our model predicts a Democratic vote share of 52% and a seat share of 51%, with a 55% chance that Democrats will take back the House. The legacy of past elections matters.
All appropriate caveats still apply. The uncertainty around the new districts is large, we haven’t taken into account any local campaign dynamics, etc. However, future campaign dynamics are unknown when the districts are drawn, so expectations are built on the factors we have already considered: the partisanship of each district and which incumbents, if any, are running for that seat. Those factors suggest the redistricting has had little net impact.