In the wake of the 2012 House elections, it looks like Democrats won a slight majority of the major-party votes (roughly 50.5%) but only about 46% of the seats. A story has gradually developed that pins this gap on redistricting (for example, see here, here, and here), since Republicans controlled the line-drawing process more often than not this time around. Matthew Green pushes back a little on this narrative, arguing that, if anything, House seat share and vote share correspond more closely today than they once did. But he doesn’t necessarily deny that redistricting is to blame for the gap this year.
We have looked at this question several times before (here, here, and here) and concluded that redistricting is a wash. But we based this conclusion on a multi-year model with both incumbency and the partisanship of the constituency (as measured by the presidential vote in each district). What if our model missed something important about the national climate this year? Not crazy, since it was too bearish on Democrats. For that matter, what if all this talk about incumbency is nonsense and it’s all about the district?
Let’s work with those assumptions. We’ll drop our regular model and go bare bones. Two steps: 1) identify the relationship between this year’s actual election returns and the 2008 presidential vote in each district (calculated by Daily Kos), 2) use this relationship plus the 2008 presidential vote in the old districts to estimate what would have happened under the old lines. No incumbency, no assumptions about national climate. For the redistricting story to hold, this exercise must eliminate the discrepancy between Democratic vote share and seat share. Otherwise, something else is going on.
Results are in the graph below. Democrats do gain more seats under this simulation—seven more total—but fall far short of matching their predicted vote share. The point should be clear: even under the most generous assumptions, redistricting explains less than half the gap between vote share and seat share this election cycle.
And it’s worth noting just how generous these assumptions are. This bare-bones model misses more individual outcomes than any handicapper or other forecasting model. It ignores one of the most important ways that a gerrymandering party tries to stick it to the other side (i.e., by moving incumbents to more difficult territory), as well as one of the most important ways that a gerrymander doesn’t work (i.e., incumbents beat expectations based on district partisanship alone). It’s not really a bad model. But it’s not really a good one, either.
If redistricting doesn’t explain the discrepancy, what does? We have argued that incumbency is a likely culprit, but as Dan Hopkins recently pointed out, Democrats also do worse because they are more concentrated in urban areas. They “waste” votes on huge margins there, when the party could put many of those votes to better use in marginal seats. (See this paper by Jowei Chen and Jonathan Rodden for more evidence on this point.)
It matters to get this right. If it’s redistricting, then Republicans rigged themselves control of a key lever of federal policymaking. This casts fundamental doubt on the legitimacy of the Republican majority, and bolsters calls for redistricting reform as a way to fix the problem (for the record, I’m not exactly an opponent of such reform). But if the explanation is incumbency or political geography, all it means is that our single-member district majoritarian system distorts outcomes in ways that are difficult to eradicate without moving to a different system entirely.
*For the sake of this exercise, we estimate our model without uncontested races and then add them back in for our predictions. A gerrymandering party can’t anticipate which races will be uncontested, so such races are properly considered outside the realm of redistricting—especially for an exercise devoid of other candidate effects like this one. At any rate, this fact explains why the Democratic vote share is a little better in this exercise than in the actual result.
{ 39 comments… read them below or add one }
Interesting, thanks for the post. One bit does need expansion and clarification, I think- “A gerrymandering party can’t anticipate which races will be uncontested.” Of course, there is always a chance that a candidate will file in a district where, let’s say, 90% of voters register as supporters of the other party. In that sense, it’s true that a party can’t know in advance which of the districts it creates will be uncontested. But surely it can rank the districts in order of competitiveness and anticipate that the least closely divided districts are the likeliest to be uncontested.
Eric:
At the end, you write: “It matters to get this right. If it’s redistricting . . . But if the explanation is incumbency or political geography . . .”
But in your analysis you find that “it” is redistricting and incumbency and incumbency and political geography. So what’s all this “or” stuff? If reforming the redistricting process would reduce partisan bias, that’s a worthy goal, even if it doesn’t set bias to zero.
Andrew:
I wasn’t trying to diss on reform, and the wording could certainly have been better. But we need to be clear about what reform would accomplish. If political geography and incumbency are a big piece (probably most) of the story, reform has a much higher bar to clear.
I was also a bit skeptical about the gerrymandering effect, but when I created a relatively simple model accounting for partisanship, but not incumbency (it would be interesting to add), and found huge effects in terms of who controls redistricting. One problem with comparing just to the 2008 districts (as you did above) is that those districts themselves may have been drawn based on political concerns.
The analysis: http://blog.fellstat.com/?p=264
Ian, thanks for the link. I love your plot of vote share by redistricting authority.
Jowei and I have done a lot of additional work on this but haven’t written it up yet. We are trying to separate human geography from gerrymandering using a variety of different simulation approaches and contrasting with implemented plans.
The basic story is that in many (though certainly not all) states, Republicans would do very well with a non-partisan algorithm based on compactness and contiguity. In some cases, gerrymandering helps them do even better than this benchmark.
Mike McDonald is right: when Democrats are able to win and draw the districts, they can combat their geographic problem by drawing non-compact radial districts from the city center to the suburbs, as in Chicago. But for the Democrats, this “solution” (win every chamber and draw your own districts) is not very realistic in states like Florida, Indiana, etc.
Most reform proposals would do rather little for the Democrats. But something is better than nothing, right? Maybe in a couple of places. The danger for Democrats, however, is that reform can lock the Republican advantage into place, taking Mike’s “Chicago solution” off the table.
The best hope for Democrats would be 1) some form of PR, or 2) some kind of partisan symmetry standard. I wouldn’t hold my breath on either of those.
The old districts were still skewed toward Republicans, just not as much as the new map. What would the results be from truly independently drawn seats?
Good luck explaining this (sort-of) “null finding” to journalists. It is really hard for them to get their minds around something like “this is a well-known property of plurality, single-member district systems.” There’s just no story there… nothing under-handed, nothing partisan, no actors and subjects.
Of course we all know it IS a well-known property of FPTP systems.
Honestly, I think this is ONE of the reasons why a few years ago when Nate Silver tried to predict the UK parliamentary elections, he was so off so much…. over there they don’t have a hard and fast “Baker v. Carr”-type one man, one vote rule.
Kevin.
We have argued that incumbency is a likely culprit, but as Dan Hopkins recently pointed out, Democrats also do worse because they are more concentrated in urban areas.
Have you seen any of the Philadelphia area Congressional districts? Are more Democrats in urban areas? Yes. But again, gerrymandering helps the GOP throw a lot of Democrats in a district.
If the democrats got 50% of the vote and only 46% of the seats, then it’s a mathematical fact that the districts are unbalanced by party. What’s the point of arguing about whether or not the “re” part of redistricting is responsible for this? (That is, what’s the point in comparing districts in the 2000s to the 1990s districts?)
Geography can explain why districts are unbalanced. But how does one separate the effects of geography from the effects of the districting process?
Two things:
1) It’s not a mathematical fact that the *districts* are to blame. If a bunch of Republican incumbents win by narrow margins in districts that ought to be a win for the Democratic party, then incumbency is to blame, not districts.
2) We separate geography from redistricting by using the same 2008 presidential vote both pre and post redistricting. Geography is the same for both; only districting changes.
I think that last-name-less Eric has more of a point than you are acknowledging. If we decided to stop caring about compactness, respecting political boundaries, communities of interest, majority-minority district requirements or anything beyond partisanship, we *could* carve up cities. If we did that, we could make sure that neither party has its voters spread out more efficiently than the other, thereby reducing, or in principle, eliminating partisan bias. Geography only matters if we decide to let it matter. Phil Burton didn’t.
Fair enough–but not even the most radical redistricting reform proposes going that far.
I have proposed going exactly that far.
Explain Ohio. Democrats packed into 4 districts with a 70-30 partisan margin (2008 presidential margins from kos: 67-32; 82-17; 67-31; 62-36). Explain why this does not, by itself, produce remaining districts (12) that strongly favor the other side.
This is arithmetic, not abstract algebra. It insults our collective intelligence to contend that this can’t impact election outcomes.
Marc, let’s keep this conversation civil, please.
I never said there weren’t some states that gerrymandered. But the claim is for a *national* effect that favors the Republicans. And there is very little of that to be found.
Look at all those districts in Chicago where Obama got upwards of 100% of the vote. This isn’t a result of gerrymandering, if by gerrymandering you mean anything more than “Not designing districts to minimize the effects of partisan distribution.”
It’s a result of Democrats tending to live together in clumps where virtually everybody is a Democrat. You’ll search in vain for precincts where Republicans get 99% of the vote, let alone 100%. They just flat out don’t exist. Republicans don’t self-segregate themselves like that. Or maybe they don’t drive out any Democrats they find among themselves, once they reach a critical mass.
The resulting clumpy distribution is going to under-represent Democrats in any system except at large proportional representation. No gerrymandering is needed to cause that, you’d have to engage in serious gerrymandering to PREVENT it.
Sure, the Republicans do a bit of gerrymandering, the Voting Rights Act practically mandates it. But most of the ‘gerrymandering’ is a consequence of where Democrats chose to live.
If that were the case, then a look at the national numbers would reveil a bias toward republican control, adjusting for who controlled the redistricting process and partisanship of the state. Based on the data I’ve looked at, this does not appear to be true nationally. There is some evidence of it in the florida legislature though…
Dr. Fellows, I’m curious to know who, if anyone, is working on algorithms that can draw the most compact boundaries possible while still taking into account local political boundaries and major geographic barriers? I would think this isn’t easy and one might have to set up different algorithms for different states or regions of states, but it seems as though it should be doable. I would think the results of some such algorithms might need tweaking by hand. And then might it not be possible to establish an algorithmic method for determining the degree to which such “tweaking” is unnecessary or biased? Would be grateful for any insight or reference you can provide on this.
Have you looked at the vote shares in rural counties in the panhandle region of TX and Oklahoma? I haven’t done a precinct by precinct look at those counties, but any county that is voting 90% R is likely to have some precincts that approach 99%R.
My original point is that for the 50% vote/46% representation to occur, the districts must be unbalanced by party, as measured by the 2012 votes. This is a mathematical fact. I hope this can be accepted without saying that districts are to *blame* or that malicious gerrymandering must have occurred. If a Republican incumbent won, its because there were more Republican votes in his/her district than there were Democratic votes. Are you arguing that some of these votes came from people who would have voted for a Democrat if there had not been an incumbent running? I can see some logic in this, but I would still argue that the best measure of Republican versus Democratic votes is just to count the votes.
I see how you can separate geography from the “re” part of redistricting by comparing old to new districts. But geography affected the makeup of the old districts also.
Rather than using the 2008 districts, I’d like to see whether nationwide *non-partisan* redistricting would have changed the outcome of the 2012 House elections. I’m thinking here of comparing the actual results to what would have come out of some of the mathematical redistricting approaches (shortest split, etc.). You’d need really granular election data, like ward-by-ward, which I don’t know if that’s available yet. Also, I don’t personally have the patience or likely the math to produce a good conclusion.
But I think it’d be a really interesting comparison to see.
My intuition is that the results of using a mathematically algorithm would be heavily dependent on the exact algorithm used. Shortest split in particular has a tendency to bisect areas of high population density, which would work to the advantage of Democrats. But most other algorithms I have seen suggested, including the one used in the Chen/Rodden paper cited above, tend to keep high population density areas together, which advantages Republicans.
This is a case of selecting on the dependent variable. The correct hypothesis test is to compare redistricting under a unbiased plan. It is not to estimate non-gerrymandered effects from “new” and “old” gerrymandered districts — those “old” districts are also biased towards the Republicans, as I and others have extensively documented.
The advocacy efforts that Micah Altman and I enabled during the last round of redistricting through our DistrictBuilder software demonstrated alternative ways to drawing districts, i.e, setting up the proper counter-factual instead of using the gerrymandered districts to estimate effects, as is done here. Thousands of users created hundreds of legal plans that simultaneously respect the Voting Rights Act, have more partisan fairness, have more competitive districts, respect more existing political boundaries, and are more compact than those that created by redistricting authorities.
How is this possible, if Democrats are inefficiently distributed, as is the frequent Big Sort claim repeated here? Congressional districts are very large. It is not difficult to create pro-Democratic redistricting plans by unpacking Democrats from all but the most concentrated urban cores. This was the general strategy pursued in Illinois, where Democrats won 12 of the 18 seats — the districts around Chicago follow a spokes of the wheel approach compared to the previous layers of an onion approach enacted by Republicans. But while academics are fixated on places like San Francisco to tell the Big Sort story, selecting on the dependent variable in this way obscures the redistricting reality in many states. In the states with numerous mid-sized cities like Columbus, creating a district that respects Franklin County instead of splitting it multiple times results in a plan that is more fair, compact, and respects existing political boundaries. I wrote a report that documents this dynamic in five Midwestern states and Micah and I are now in the process of evaluating the data we collected via our public mapping efforts which will show the same (as do the reports generated by our advocacy partners).
So, for the sake of good political science, let’s please stop estimating gerrymander effects solely from gerrymandered districts.
Michael: Is your report the one found at this link?http://elections.gmu.edu/Midwest_Mapping_Project.pdf
Michael:
That’s all fine–I don’t begrudge you the right to ask any question you want. But there were dozens of districts drawn under Republican control this time around that were not in 2001. Shouldn’t that count for more than 7 seats?
Nuts. you are assuming that the districts were not gerrymandered in 2008 or whatever. You might hit the wayback machine and check out how the Reps rigged Texas in the middle of the decade or ask some questions about the representation in the PA delegation vs the way the state votes. Yes, the Dems can overcome this in wave elections (2006 and 2008) but the deck has been stacked for a couple of decades or more with the Reps concentrating on state legislatures.
You do realize that the 2002-2010 maps were drawn overwhelmingly by Republicans as well, right?
This, alone, will skew the analysis to show that redistricting doesn’t much matter.
Instead, a better method would be to look at turnover in state legislatures and compare the differences in maps produced by partisans on each side and maps produced by a commission or various other non-partisan or bi-partisan methods only for the new cycle. I.E. how well does each of these states’ maps capture the vote choice. If Republican maps and Democratic maps do not match well to the non-partisan or bi-partisan maps, then you can easily say that the maps themselves matter.
This isn’t to say that non-partisan or bi-partisan maps are, on average, going to match the results statewide. Obviously, because of the well-documented political geography where Democrats are super concentrated in urban areas, this won’t be the case. But its the difference between this and how maps produced by Democrats and Republicans perform that is the key. Any difference whatsoever indicates that the maps do matter.
Not to harp on a point, but this is exactly what I did. See the link in my previous comment.
Well then!
I was in a big hurry when I typed that up (sitting at the Apple genius bar), so I didn’t read any above comments.
Sorry for cramping your style.
There is another (partial, if not full) explanation: that the effects of redistricting are OVER MORE THAN ONE REDISTRICTING CYCLE. That is, twelve years ago Republicans were successful in skewing the districts, and two years ago they did more of the same.
It’s actually pretty simple: the US House playing field IS tilted toward the Republican Party simply by the fact that Democrats are more naturally packed in urban areas.
Sure, if Democrats controlled redistricting in every state and could draw Maryland -type outlandish gerrymanders, they could spread their urban vote more and at least get parity. But even then it’s tough. But when Republicans are in control, it’s so much easier for them — they can keep Democrats packed in cities (or majority-minority districts) and win supermajorities because of it.
Back in the 1990s, in plans largely drawn by Democrats, the overall bias in the partisanship of districts was nearly exactly the same as today. What was different is that there was more ticket-splitting and more chances for parties to win seats in the other party’s turf – -particularly conservative Democrats in the former slave states. That’s vanishing now.
Take this year. Democrats won all but one seat in the 177 districts in which Obama had gained his highest share of the 2008 vote — down to where he was 2% ahead of his national average. And the one exception was due to the screwy Top Two system that sent two Republicans to the top two in CA-31 in a 54% Democratic districts.
But on the flip side, the Democrats didn’t pick up a single new seat in a district in which Obama ran more than 5% behind his national average, and only three where he ran more than 2% behind his national average. And there just are a lot more districts on the GOP side of the 50-50 wall than Democratic side — about 47 in all. And those moderate-to-conservative southern Democrats? Their number is down to single digits.
So yes, the best answer is not trying to figure out convoluted redistricting schemes so Democrats can get fairness from gerrymandering, but instead to go to modest, American-style, fully constitutional forms of proportional representation. We did maps for the whole nation comparing districts as they are and “fair voting’ as it could be at http://www.fairvoting.us. Check it out. It does indeed tell a tale. Not only does it mean every voter in a state with at least three seats would have shared representation, but it pretty much removed the GOP bias.
The ENTIRE gerrymander systems in the U.S.A. are ALL ANTI-Democracy since 1776.
The most ENEMY voters are PACKED into the fewest gerrymander districts (political concentration camps) possible — using prior election results — i.e. the 2008 Prez stats in the 2011-2012 gerrymanders.
The rest of the area is CRACKED to get friendly majorities — at least 55 percent, if possible for a mere 10 percent SAFE seat margin.
1/2 votes x 1/2 gerrymander areas = 1/4 CONTROL indirectly.
The folks making the gerrymanders are ALL EVIL Stalin/Hitler types.
Too difficult for the SCOTUS math morons to understand since the 1964 gerrymander cases — a mere 48 LONG years.
Much worse control percentage in the gerrymander U.S.A. Senate and in plurality winner primaries.
Save Democracy – Total Votes / Total Seats = EQUAL votes needed for each seat winner in ALL legislative body elections.
NO caucuses, primaries and conventions are needed or wanted.
Fail. How much the old districts were gerrymandered was not even considered.
So rather than being caused by *intentional* gerrmymandering by Republicans, the outcome is caused by the *unintentional* self-gerrymandering of Democrats into urban areas. This assumes that districts must not partition the cities, but e.g. pie-shaped districts centered on the cities would probably be able to create more competition.
Eric,
Ignoring the fact that redistricting has been biased for decades (at least) in GOP favor, I’d like to point out that your model SIGNIFICANTLY reduces the margin of Democratic minority. The recent US Houses have been extremely partisan, so one might consider the balance of votes to be a strong indicator of the behavior of the house, or put another way (as GOP leaders are hoarsely shilling) a clear call for divided government.
Don’t lose sight of the politics of the situation. Given the clear movement among the forward-looking members of the GOP toward rethinking their approach, that reduction in the degree of Democratic minority could be QUITE significant politically. The number of highly contested votes the GOP win by less than 2%, may be telling.
Keith – redistricting has been biased towards the GOP for at most one decade. From 1972 to 1992, the Republicans were under-represented by between 3% and 11%. In 1994, the Rs were over-represented, but so were the Ds, due to the minor-party vote being entirely futile. From 1994 through 2010, the Rs were never over-represented by more than 4%, and were under-represented at some elections, while the Ds ranged from 1% under-represented to 5% over-represented.
2012 was different – the Rs are 5.5% over-represented and the Ds 3.5% under-represented. So the Ds are now not quite as under-represented as the Rs were for 20 years.
The key phrase is at the end of the article: “it means is that our single-member district majoritarian system distorts outcomes in ways that are difficult to eradicate without moving to a different system entirely.”
Other electoral systems avoid the problems with geography, political boundaries and gerrymandering. These systems are widely used in other countries, and have been used in the U.S. in the past.
At large elections in multi-seat voting districts, using one of the various forms of proportional representation, party open list, single transferable vote, partial block voting, etc., have many advantages. No electoral system is without problems. Given the dysfunction of recent Congresses, we need a real discussion about these alternatives.
A different system would encounter significant resistance, and would take decades to reach enough acceptance to implement. However, since each state may choose their own method of selecting representatives, the laboratories of democracy would yield significant improvement over our current system.
We need to start the discussion.
Absolutely agree, Eric – -see my comments above to that effect. Check out our very revealing maps of how this might really look at http://www.FairVoting.US – -we have plans for every state, with real partisan numbers and racial numbers.