Author Archive | Eric McGhee

Are the Democrats still at a disadvantage in redistricting?

To what extent are Democrats inherently disadvantaged in congressional elections because of where they live?  This was a hot topic of conversation just after the 2012 congressional elections, when the Democrats won a majority of the popular vote but a distinct minority of the seats in the House of Representatives.  The subject has recently resurfaced in public debate, and the issues are important enough that they deserve some consideration.

First, some background.  Jowei Chen and Jonathan Rodden (hereafter, C&R) have developed a computer program that draws thousands of simulated districts using precinct-level data and a handful of systematic criteria. They then compare the actual redistricting plan to the simulated ones.  If they don’t differ substantially, then any bias in the district lines cannot be blamed on gerrymandering.  Or so goes the argument.

This method has attracted considerable attention because it appears to show that Democrats suffer because they are concentrated in urban areas and “waste” votes on big wins, not because Republicans have drawn the lines that way.  In other words, Democrats are “unintentionally gerrymandered.”  C&R are not the only ones to make this claim (for a recent example, see Nate Silver’s post here), but those who make it commonly cite this research.

Michael McDonald, a respected expert on redistricting, recently offered a thoughtful critique of the C&R method, calling it unreliable and divorced from the reality of redistricting.  He has pointed to the many fair and apparently usable maps drawn by the public in the last redistricting cycle—including in key states where C&R had claimed a strong bias against the Democrats—as evidence against the C&R conclusions.

It does not help C&R’s case that they recently updated their analysis for Florida—the first and most famous application of their method—with more recent presidential election data and found a very different result.  Instead of an inherent bias against Democrats and no clear sign of gerrymandering, they found no inherent bias and a strongly gerrymandered outcome.

So what does this mean?  Is the C&R method reliable, and is it telling us something useful about redistricting? McDonald argues that algorithms will often produce “nonsensical, spaghetti-like districts” that would never be part of a real plan.  Worse, he argues we will never know how often an algorithm like C&R’s will make this mistake, so the product of the algorithm can’t be trusted.

I don’t doubt McDonald’s experience with these algorithms, but I think his criticism misses the point of the C&R approach.  Unlike many past efforts at automated redistricting, this algorithm isn’t meant to replace human line-drawers by generating maps that would clear every legal and political hoop. Instead, it offers some sense of the range of unbiased possibilities given the realities of political geography. It asks, “What if we drew districts at random according to a few simple rules?  Would the results be any different than the actual plan?”  If the answer is “no,” then the humans drew lines as if they were unbiased computers, regardless of any other factors that led them to choose one set of lines in particular.

As for whether the algorithm produces weird districts, all of C&R’s simulated maps for Florida are posted on the web, and to the naked eye they don’t look much worse than many legally adopted plans in other states.  And that’s without directing the algorithm to favor compactness.  When C&R take that extra step, the results are downright boxy.

To be fair, C&R sometimes imply that in certain states it’s impossible to draw compact districts that are fair to both parties.  I don’t think that’s quite right.  Their method does not tell us what is possible—only what is likely in the absence of any partisan intent.  By contrast, the authors of the publicly-drawn districting plans McDonald points to may have had a partisan intent—specifically, the intent to undo the underlying geographic bias.  That suggests something important about redistricting reform: the map-makers may need to take explicit account of partisanship to avoid reproducing the inherent bias in the underlying political geography.  That’s a little ironic, since redistricting reformers often propose that map-makers ignore partisanship completely.  But it may be true all the same.

What about Florida?  Doesn’t the complete reversal of C&R’s original finding suggest serious flaws in their method?  Not necessarily. In fact, the same shift in political geography that C&R uncover is visible in the raw data. Democrats won 50% of the presidential vote but only 47% of the precincts in 2000, while in 2012 they won 51% of both.  In other words, measured this way the bias disappeared.  There are problems with using only these raw data, to be sure.  A better method would correct for large differences in population between precincts, and would take into account how the precincts are related to each other in geographic space.*  In other words, it would do exactly the things the C&R method already does.  Nonetheless, the fact that multiple methods all point to a similar result should give us some confidence that the C&R method isn’t doing something screwy.

But if the C&R method is valid and useful, doesn’t the result in Florida then prove that “unintentional gerrymandering” isn’t a factor anymore?  It certainly might, though it’s too soon to say.  We need to apply the C&R algorithm to more recent presidential results in other states.  There may have been similar change outside Florida, but given the continued concentration of Democrats in urban areas I’d be skeptical it has been too widespread.

What the Florida case does highlight, however, is a point that John and I have made before: redistricting doesn’t guarantee results.  Consider how quickly the reality in Florida has shifted.  In 2001, Republicans in Florida drew districts that reflected the state’s anti-Democratic geographic bias, but by 2008 (the year of the updated C&R numbers) this bias had disappeared and in 2011 Republicans were shoring up an eroding position.  Could their efforts hold back the tide for the next 10 years?  Maybe.  But based the last 10, I wouldn’t be terribly confident.

Moreover, none of this considers the imperfect link between the presidential vote and the vote for Congress.  While the two are more closely linked than ever these days, they are still different.  Incumbents routinely outperform their party’s presidential candidate, and strong challengers or uniform partisan swings can overwhelm the most carefully drawn redistricting plan.  John and I estimated that incumbency alone could account for the entire discrepancy between the national seat share and vote share this cycle.

Gerrymandering is certainly real, and it was egregious in a number of large states this last election cycle.  Since even a single unfair outcome is one too many, independent commissions ought to be the preferred way to ameliorate the problem.  But we should be cautious about expecting too much from such reforms.  A neutral process will always carry the risk of a biased outcome, so we should either require plans to eliminate bias or accept that sometimes bias will happen without intent.

(Thanks to Jowei Chen for sharing precinct-level data from Florida.  Naturally, he is not to blame for any errors in the manner they are used.)

*However, size was largely uncorrelated with Democratic vote share in the precincts, so its impact on the raw numbers presented here is probably modest.  If anything, it may understate the degree of bias against Democrats since larger precincts were very slightly more Republican than smaller ones.

Continue Reading

Gerrymandering still isn’t a very big deal

Noam Scheiber has a recent piece about the Republican party that makes two points.  The first is that Republicans in Congress—especially the House of Representatives—are more out of touch with mainstream political opinion than their Democratic counterparts, and have grown more so recently.  There are a lot of reasons to think this point is sound.  See here, here, and here (though see also here for a different perspective).  At the very least, it’s debatable.  Personally, I tend to agree with Scheiber.

However, I disagree about what Scheiber believes is the cause:

What explains the PR pileup that GOP elders can’t seem to clear to the side of the road? Partly it’s the structural forces at work in American politics…But the more direct and mundane explanation is gerrymandering. Thanks to the way Republican legislatures drew congressional districts in 2000, the median House district leaned Republican by two points over the next decade—a big edge given the tiny margins that frequently decide competitive races. Since 2010, the built-in advantage has grown to three points. The result of all this gerrymandering is to give the Republicans a death grip on the House. In 2012, they won 1.4 million fewer votes than Democrats in all the House districts combined, but still managed a 33-seat majority.

As John and I have argued, the evidence that gerrymandering has given the GOP any sort of grip on the House—death, Vulcan, or whatever—is  weak.  And the connection between the districts (however they ended up that way) and polarization is also pretty feeble.

But what about the median district argument?  The median was +2 Republican before the redistricting, which means it was 2 points more Republican than the mean (given the way the Cook Political Report measures these things).  Then it became +3!  That’s a lot, right?

Maybe.  But let’s put it in context.  The median district has been more Republican than the mean in virtually every House election since 1952.  See the graph below.  In other words, the Republican party had the same advantage in 1952, but was a very different party.  So something else is going on.

Median District 1952-2012

Why is gerrymandering is such a popular explanation for things?  I’d guess it’s because it reeks of corruption and manipulation, so it taints everything it touches.  It’s easier to get mad about something if you can blame it on politicians rigging the game to thwart the will of the people.  If more mundane forces are at work, it becomes harder to get mad, and harder to think of the solution.

I’m not trying to pick on Scheiber here, since I think everyone has a natural tendency to lurch toward these kinds of explanations.  It’s just that sometimes there isn’t much evidence in support.  This is one of those times.

Continue Reading

Details on the 2012 House predictions

We created and refined a prediction model for House elections and recently evaluated the model’s performance against the actual results.  In the interest of evaluation and transparency, we have uploaded details and data as Google docs for anyone to access and download.  The model coefficients and prediction process can be found here.  A spreadsheet of predicted values and standard errors can be found here.

Continue Reading

The history of the House votes-seats discrepancy, in two graphs

My recent claim that the gap between Democratic vote share and seat share this election was not caused by redistricting has generated a lot of conversation.  The most common complaint is that the old districts were gerrymandered to favor Republicans, too.  If the new ones left that gerrymander in place…well, gerrymandering might still be a factor, and reform might still make things better.

This is a fair point:  while the impact of redistricting this year was almost certainly small, my claim that reform could only have a minor impact overstepped the data I had in front of me (see here and here for evidence that I was irrationally exuberant).  Nonetheless, I thought it would be useful to provide some broader historical context for this discussion.  How long has it been true that Republicans would likely win a majority of seats even when they don’t win a majority of the vote?  Do changes in this bias coincide with redistricting?

Toward that end, below is a graph of partisan “symmetry” bias from 1952 to 2012, with dotted vertical lines to mark the point when most (though not always all) districts have been redrawn.  Symmetry is a hypothetical:  What if the two parties both received the same share of the vote?  Would they both get the same share of seats?  If not, which party would have an advantage?  I’ve had my concerns with this measure, but it seems perfectly suited to the conversation we’re having right now.

(NB:  To calculate these numbers, I used a program called JudgeIt for R that was developed by the Monkey Cage’s very own Andrew Gelman, along with Gary King and Andrew Thomas.  The program bases its numbers on a statistical model that reduces some of the noise and allows us to explore counterfactuals.)

I take away four observations from this graph:

  1. The bias in favor of Republicans has been around since the early 1990s, but it did not directly coincide with the 1991 redistricting.  Instead, it emerged after Republicans took control in 1994.

  2. Parties generally, though not always, have a bias in their favor when they have more incumbents (i.e., when they hold a majority).

  3. Bias varies a lot from one year to the next, and is not a fixed feature of a redistricting cycle.

  4. The bias in favor of Republicans this cycle is the largest since before 1952.  The party also had more incumbents running than at any time since 1948.

In other words, as we’ve been arguing, incumbency matters to discussions of partisan advantage.  This is very much in keeping with the conclusions Gary King and Andrew came to when they examined this question 20 years ago (though the magnitude of the bias I’ve measured here is larger than they found in their paper).

What happens if we statistically remove incumbents and pretend that every district is open every year?  We get this graph:

This suggests the Republican bias has been around since the 1950s, a conclusion that is also consistent with King and Gelman’s analysis.  Moreover, shifts in the partisan direction of bias almost never coincide with redistricting.  The one exception is 1992, when a brief advantage for Democrats (that itself did not coincide with redistricting) became an advantage for Republicans.  (However, see this book for an argument why redistricting accounted for the disappearance of a pro-Republican bias that existed outside the South in the 1950s.)

This doesn’t prove that gerrymandering is irrelevant.  Perhaps we’re just living with the legacy of the lines drawn in 1991.  And it’s still possible that the bias could be alleviated with aggressive enough reform.  But to me this says there is a lot more to the question than redistricting, and that the Republican advantage has been with us for some time.

Continue Reading

Redistricting does not explain why House Democrats got a majority of the vote and a minority of the seats

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.

Continue Reading

How did our House prediction do?

We offered some predictions about House elections in earlier posts (see here, here, and here).  We based our predictions on a model that included some national factors like presidential approval and the state of the economy, plus some district variables like the district presidential vote and incumbency.  Now that we have something close to the final House results, how did we do?

The bottom line:  our model performed quite well, predicting only 7 fewer seats for the Democrats than they actually won, and miscalling only 21 races out of 435.

Now for the details.

Before we can get any further, we have to decide which model to evaluate.  Our first prediction was a purely “fundamentals” model that used only incumbency and the district presidential vote to distinguish one district from another.  Nothing about the relative strength of the candidates was included.  This model proved to have a large amount of error, suggesting that candidate strength does make a difference.  Adding campaign spending to the model brought down this error a lot, though for our forecast it forced us to use fundraising in the summer as an indicator of likely spending in the fall.  We couldn’t be certain how well that would work.  But since summer fundraising still falls far before the election, this approach stuck to information that was publicly available before the most intense period of the campaign season.  That comes close enough to a “fundamentals” model for us, so we’re going to use it as our final prediction.*

There is more than one way to evaluate the model’s performance.  The first is to see how close it came to the topline vote and seat share.  In this respect, we were a little too hard on the Democrats.  Based on current results at the New York Times’s “big board,” the Democrats have won 195 seats for sure, and six of the remaining seven are leaning their way.  That’s a total of 201 seats.  By contrast, our model predicts 194 Democratic wins,** missing the actual result by 7 seats.  Our model also predicted a Democratic two-party vote share of 48.9%, or about 1.9% below the actual result of 50.7%.  The model expected a vote share at least as high as the actual one about 36% of the time, and it expected a seat share that high about 33% of the time.  So both fall in a comfortable range for the model’s error.

The second way to look at our model is to see how well it predicted each individual race.  Below is a scatter plot of the predicted vote share against the actual vote share for all 435 races.  The diagonal line is equivalence:  if the predictions were exactly accurate all the points would fall along that line.  The red data points are cases where the model missed the winner:  there were 21 such cases overall.

Perhaps the most striking aspect of this graph is the curved relationship between our prediction and the outcome.  The model is too hard on Democrats at the low end and a little too easy on them at the high end, producing a sort of s-curve.  This curvature is entirely a function of using early fundraising as one of our predictors.  The model without campaign money has a more linear relationship, but it also gets the actual outcome wrong more often.  In other words, early money misses some of the dynamics of the race, but it does a good job of discriminating between winners and losers.

So our model was bearish on Democrats, was somewhat off on district vote shares, and missed the actual winner in 21 cases.  How does this compare with other predictions?

In terms of total seats, our prediction was closer to the final number than Charlie Cook, and two seats worse than Larry Sabato.  It was also two seats better than than Sam Wang’s generic ballot prediction (which considered 201 to be a highly unlikely outcome), and three seats worse than his Bayesian combination of the generic ballot and the handicappers.  So on this score, Wang’s hybrid beats all other forecasts by a nose.

What about predictions for individual seats?  Here we have to drop Wang, since he didn’t offer such predictions.  But we can still look at Sabato and Cook.  Compared to these two handicappers, our 21 misses were the highest of the bunch.  Sabato was the best, miscalling only 13 races, while Cook fell in between at 17.


However, one can think about this a different way.  Our model only used information publicly available by the end of the summer (when the last primaries were decided).  The handicappers had lots of information (some of it proprietary to the campaigns) up to election day.  Yet our miss rate wasn’t that much higher.  At most, all that extra information amounted to 8 correctly called seats.  The outcome of the rest could have been known far earlier.

To be fair, the races that distinguish between these forecasts were incredibly close.  None of Sabato’s missed predictions were decided by more than 10 percentage points.  Only two of our missed predictions fit this description, and one of Cook’s.  Most of the misses were far closer.  Many of these races could have gone the other way, meaning the differences between the success rates might be due to chance alone.  There should be some credit all around for correctly identifying the races that were likely to be close in the first place.

On balance, there is much that we could improve about our model, and we hope to keep working with it in the future.  In the first time out of the gate it got the lay of the land quite well, and actually came within spitting distance of forecasters who had a lot more information at their disposal.  We think that’s pretty good.

*We also did a prediction with Super PAC money, but that was more for explanatory purposes than anything else.  It certainly violated our rule of using only the fundamentals.

**Our original forecast was 192 seats, but in conducting this post-mortem, we discovered an error in our code that mis-classified some uncontested races.  When the error was fixed, the prediction bumped to 194.  Other aspects of the prediction were basically the same.

Continue Reading

Which House races have the most at stake?

This post is jointly authored with Boris Shor of the University of Chicago’s Harris School (currently Robert Wood Johnson Scholar at UC Berkeley).


In a recent post, we showed that party trumps constituency in House elections:  on average, Democratic and Republican House candidates are ideologically distant from each other in both competitive and uncompetitive districts.  Keith Smith reinforced this basic finding in a subsequent post.  (As before, see here for details about the data and methodology, and here for the actual ideology scores.)

However, there is still a lot of variation from one race to the next.  Some contests offer a stark choice between two candidates with very different views on the issues, while others offer much less stark differences.  This is important because it makes sense to focus on the races which have the most at stake.

Below we have listed the competitive races that feature the clearest ideological choice.  “Competitive” means those rated “leans” or “toss-up” by the Cook Political Report.  “Clearest choice” means ideological differences greater than the average for all the races.  Incumbents are indicated with an asterisk.


Continue Reading

For Congressional Candidates, Party Trumps Constituency

This post is jointly authored with Boris Shor of the University of Chicago’s Harris School (currently Robert Wood Johnson Scholar at UC Berkeley).


As election day draws near, there has been growing interest in just what sort of 113th Congress we might expect to emerge.  Will it be more polarized and ideological than the 112th or less?  The Washington Post’s Aaron Blake has speculated that redistricting will lead to a more polarized House by drawing fewer competitive seats.  This idea received some pushback from two of Monkey Cage’s finest:  first from John Sides, then from Nolan McCarty, both of them citing McCarty’s own research with Keith Poole and Howard Rosenthal pointing to only a modest effect from redistricting.

Boris recently offered a way to assess these claims for the current election cycle.  He has calculated the estimated ideological position on the liberal-conservative spectrum (“ideal point” is the jargon) for at least one candidate in almost every House and Senate race this cycle.  (See here for details about the data and his methodology, and here for the scores.)  This includes estimates for 722 House candidates from 419 districts and 64 Senate candidates from 33 states with elections this year.

We have combined his data with the partisanship of the congressional districts, measured with the 2008 presidential vote in these districts (the numbers are from Daily Kos).  We wanted to know whether the partisanship of the district is at least correlated with candidate ideology, and how a candidate’s party label plays into all that.

The main take-away: party trumps constituency. The ideological differences between Republicans and Democrats are much larger than the differences between candidates running in Republican and Democratic districts.

As one would expect, candidates get a bit more liberal as Obama’s share of the 2008 vote increases.  But the difference is surprisingly modest.  Moreover, the gap between the parties is vast at every point along the way, and overwhelms this modest relationship.

There are also some interesting nuances.  First, extremists and moderates can be found almost everywhere.  There are some relatively moderate Democrats in heavy Obama districts, and some moderate Republicans in districts that went strong for McCain.  Keep in mind that these ideologies are based on each candidate’s positioning in the campaign.  Measures of elected legislators’ voting behavior in office generally show far fewer moderates.  This suggests that, either by design or due to pressures beyond their control, candidates often campaign as moderates and govern as partisans.  We’re not the first to make this point—see here (gated)—but it’s interesting to see it confirmed.

Second, just because the partisan gap is large doesn’t mean the parties are completely unified.  In fact, within each party caucus there are two distinct camps, one much more extreme than the other.  At his own blog Boris has offered more detail on this pattern.  This suggests more division within the parties—both its winners and its losers—than is often recognized.

Third, eyeballing the chart makes it look like both factions of Democrats and the moderate faction of Republicans are at least somewhat responsive to district partisanship.  But the conservative Republican faction at the top of the graph is flat. So some Republicans appear to be immune to the even modest pressure imposed by district realities.

Despite these nuances, however, the overall point is that the parties differ drastically.  McCarty offers perhaps the best summary:  “polarization has grown because Democrats and Republicans are representing moderate districts in increasingly extreme ways.”  We can add that this appears to be true in all the other districts, too.

(Note:  The original post incorrectly referred to the current Congress as the 113th, and the next one as the 114th.)

Continue Reading

The Effect of Redistricting on House Elections, Revisited

The Cook Political Report has come out with its “PVI” numbers of each of the new districts, which are measures of district partisanship based on the presidential vote.  They use this measure to argue for an important redistricting effect this cycle in favor of Republicans.  David Wasserman has the details:

Unfortunately for Democrats, this year’s index tells a dire story of what can happen when a party suffers an ugly election cycle right before redistricting. Democrats…lost…more than 680 state legislative seats – empowering Republicans to draw ten-year maps in four times as many districts as Democrats. As a result, thanks to effective GOP cartography, the number of “strong” Republican seats has jumped from 182 to 190 and the number of “strong” Democratic seats has fallen from 150 to 146. Meanwhile, the number of “swing” seats has fallen below 100 for the first time, from 104 to 99. (emphasis in original)

The Cook report is a venerable organization whose analysis is rightly respected, and I have no doubt that their presidential vote numbers by district are the most accurate available.  But the notion of big redistricting effects this year is Zombie Politics, and the poor corpse should be put to rest.

First, contrary to the quote above, a party that aims to gain seats through redistricting wants fewer “strong” seats, not more.  Since one only needs 50%+1 votes to win (in a two-party race), every vote beyond that threshold is “wasted” in the sense that it could be used to bolster the party’s chances in some other district.  More strong seats means bigger winning margins, which means more wasted votes.

Second, the idea that the 2010 elections dramatically increased Republican opportunities to gerrymander presumes they can actually control the process in the states they won.  This is not necessarily the case.  Here’s what we said on this subject a couple weeks ago:

[A] redistricting party doesn’t have free rein 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, counting districts as Democratic or Republican using only the presidential vote assumes everyone who votes for Obama will vote Democratic for Congress, and ditto for Romney and the Republicans.  But even in today’s polarized environment, incumbency matters.  A moderately Democratic PVI with a Republican incumbent is probably a Republican win.  What happens when we pair redistricting with incumbency?  A close look at the Cook Report’s 25 incumbents most hurt by redistricting reveals that almost half of them (11) are Republicans.  And the Cook predictions for those races split evenly between the parties at 9 a piece, with the rest as toss-ups.

That’s why we feel the best approach incorporates both district partisanship and incumbency, and uses past election results as a guide for how important each is likely to be.  When we ran those numbers, we found redistricting was a wash.  We also found in the same analysis that Democrats would get a much more substantial seat gain if there were no incumbents running and all seats were open.  The structural advantage for Republicans this year stems from incumbency, not redistricting.


Continue Reading

How Money Might Affect House Races (Part 2): the Super PACs

In the wake of the U.S. Supreme Court’s Citizens United decision, Super PACs have become incredibly active.  In case you have been trapped in a dark cave somewhere, Super PACs are independent expenditure committees that can receive and spend money in unlimited amounts so long as they do not coordinate with any candidate’s campaign. According to the Sunlight Foundation, all independent expenditures (including money from the parties) totaled $589 million this cycle.  That’s a lot of money.

Attention is starting to turn toward the specific activity in House races.  In fact, some experts have been worried for a while that the real impact of Super PACs will fall in these lower-profile races, not in the big-time world of presidential politics.  Here’s Rick Hasen, a widely and rightly respected expert in election law, in January:

Given the expected vast spending by presidential candidates and parties in the general election, I am not very concerned that Super PAC spending will influence the outcome of the presidential election, though it might…But I am concerned that Super PAC spending will influence the outcome of close Senate and congressional races.

How much money are we talking about?  As of October 3, $122 million in outside money has been spent in House races.  About three-quarters of that money ($95 million) has been spent in the general election, and almost half of that ($43 million) has been spent by parties, who follow different rules and have different incentives.*  That leaves $52 million spent by Super PACs.  This is still a large amount of money.  But to put it in perspective, the fall candidates had raised over 12 times as much by June 30.

What we want to know is how this money is going to affect the races for the House.  There are three questions here.  First, how much is being spent by each side?  Second, where is it being spent—that is, on which races?  And third, does it make a difference?

The answer to the first question is that there has been a surprising amount of balance in House Super PAC spending so far.  If we count money spent against a candidate as money favoring the other side, then Democrats have received $26.5 million of Super PAC money compared to $24.7 million for Republicans.

The second question is arguably more important.  In the races where it is spent, is it spent evenly, to support candidates who are struggling, or to wipe out potential competition?

The money is certainly concentrated.  The top five percent of races account for 88% of the Democratic Super PAC money and 73% of the Republican Super PAC money.  Moreover, Super PAC spending tends to go where the action is already intense:  Super PAC spending by one side tends to be higher where the candidates have raised roughly equal amounts of money and where there is a lot of Super PAC spending by the other side.  (Two potentially important factors—the presence of an incumbent, and the partisan composition of the district—don’t seem to matter.)

But these relationships are general tendencies, and weak ones at that.  The reality is that Super PAC spending tends to have a strong partisan skew in any particular race.  The graph below shows the relationship between Super PAC spending on each side.  The labels identify the highest spending races, and the color coding indicates races where the Democrat (blue) or the Republican (red) has more Super PAC support.  Equal spending by both sides would fall along the dotted line, yet almost no races land there.  Instead, heavy Super PAC spending on one side is often matched with little or nothing on the other.

When this money is added to the fundraising totals for each candidate as if it were their own, it doesn’t generally tip the balance much.  Candidates who have good fundraising numbers are still in the game with Super PAC spending.  Those who do not, are not.  There’s just too much money in these races already, and the marginal returns of additional spending are too small.

It’s worth noting that the story is very different for party spending.  Party independent expenditures, which account for a huge portion of outside spending, tend to temper the effect of candidate fundraising and Super PAC activity.  When party outside spending is added in as well, it strengthens candidates who are competitive but slightly behind, rather than strictly reinforcing the status quo.

The last question is:  what does it mean for our prediction?  Consistent with all our other analysis, the answer is “not much.”  With Super PAC money we predict a three-seat loss for Democrats, compared to a two-seat loss with regular fundraising, and a one-seat gain without fundraising factored in.  The result is identical if party money is included too.  This is close enough to be a wash.


*Party expenditures are different for a couple of reasons.  First, parties have different incentives than almost any other political actor, because they actually want to control government rather than promote one particular issue.  Second, they are subject to donation caps that make it harder for them them to raise money, though they can still spend it independent of the campaigns in unlimited amounts.

Continue Reading