Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Bush v. Gore times five

October 16th, 2012, 8:00am by Sam Wang

Are you upset about the fact that the Presidential race is set up to allow the possibility that the candidate with fewer popular votes might still win? For example, Gore won by about 0.5% more votes – yet Bush became President. Depending on specific conditions, this “lumpiness” can help one Presidential candidate or the other. As it turns out, the structural change in House districts creates a problem similar to these – but larger, and in one direction only.

Previously I gave a detailed calculation regarding the state of House races. Today’s essay is shorter, though still geeky. You may have an epiphany about our democracy…or decide to see what is happening at The Onion.

The entire story is contained in this one graph.

The black line is my estimate of where the playing field is. The red line shows the range where I thought the race would end up – though note that events are overtaking this somewhat. More on that in a bit.

Each point represents one Congressional election since 1946, with an arrowhead (^) pointing at the most recent one. The gray shading shows the area covered by the data, and gives an idea of what used to be considered likely or possible. The gray region crosses diagonally through (0,0), which means that in the past, winning the popular vote usually meant winning control of the House. Until now.

Here is what I did, skipping most of the math.

(1) First, I had to figure out how votes would translate into seats in 2012. That is shown by the black diagonal line which skirts the bottom of the gray region. Notably, it does not go through (0,0). Compared with history, Team GOP now has an advantage about equivalent to getting a 2.5% extra popular vote share across the board. This advantage comes from partisan redistricting plus incumbency, a powerful combination. The low slope of the diagonal is another likely consequence of redistricting – seat count is less sensitive to vote outcomes.

The difference between 2.5% and zero is substantial. Many fans of U.S. politics like to go on about the unfairness of the Electoral College. However, that problem is less than 1%, and averages out to zero. Redistricters cannot take Philadelphia out of Pennsylvania and stick it into New Jersey. But House districts are a different matter.

An advantage of this size is unprecedented…almost. The one data point below the diagonal is 1996. Analysts often cite 1996 as a strange year: Democrats won the House popular vote, but Republicans retained control. Well, from now on, every election for the next ten years is potentially like that. It’s like global warming: previously extreme events will now become routine.

If anyone thinks I am underestimating this effect, do tell. I am all ears. But please read “The Very Hungry Gerrymander” (Oct. 4) first.

Now to the popular vote.

(2) To estimate the likely popular vote on November 6th, I used the generic Congressional ballot to predict a D+2.5+/-3.0% popular vote win. That is shown as a red line segment along the diagonal. This is the weak point of the argument, because the popular vote is a hard thing to predict.

I do not think there will be a “wave election” – a mass eviction of incumbents. The real wave election was in 2010, when the GOP won the national popular vote by 7%. Compared with that, the pendulum is certain to swing back a bit. But by how much?


The estimable Charlie Cook thinks the House is unlikely to flip, largely based on current district-level data. Keep an eye on him to see if his estimate changes.

Where will conditions end up on November 6th? As you can see in the Meta-Margin, the Obama campaign took a 5-point hit last week after the debate, and is now below Obama+1%. Re-elected Presidents have coattails, and the House popular vote is said to follow the Presidential vote closely. The tightening of the Presidential race suggests that the result will be closer to the left end of the red line segment. Or it could come back. Suspenseful!

Tags: 2012 Election · House

46 Comments so far ↓

  • Reason

    Interesting. BTW, what is your take on the recent Gallup poll about the swing states, Dr. Wang.

    • Sam Wang

      I haven’t looked at it. Examining the details of a single poll takes the “meta” out of “meta-analysis. Plus I’m traveling. See my general take here.

    • mnpundit

      Ha, let alone the Kos/PPP poll!

      Though contrary to my comment yesterday I just want to throw up my hands and shout “Oh for God’s sake!” rather than weep in a corner.

  • Anbruch

    As to how far the the Meta-Margin might swing back pulling the House with it, it all depends on why we believe Obama lost those 5 points in the MM. I’m personally of the view that the debate itself was only a modest factor, that some of it was a wearing off of Romney’s 47% comment and most of it the poor handling of the post-debate by Obama supporters in the media and on the internet. But we shall see. If the MM continues up, it will almost certainly be attributed to the debate even though it is already bottomed out and seems to be naturally headed that direction in any case. We’ll see.

  • wheelers cat

    I’m going to argue the “new normal” point.
    2012 is the last year the GOP has a chance at the presidential election because the demographic timer has already started going off.
    The white voter share of the electorate is in steady gradual decline and the GOP remains the all white party, particularily with Romney’s embrace of the Sailer Strategy.
    IPOF, many have postulated that Romney needs 63-65% of the white vote to beat Obama.
    I think that is mathematically impossible.
    eg, Kerry would beaten Romney this year.
    He got 37% of white males and 41% of white females.

    • kel

      The problem is that there are actually some very conservative view points among minority voters, Hispanics in particular, to be pandered to. If democrats fail to follow up with true organization and representation of minority groups. If they fail to convey democratic core values both rationally and with conviction, this demographic advantage will be short lived.

    • wheelers cat

      Right now hispanics go Obama by 70%, up from 67% in 2008.
      And there are relatively more of them.
      Its not about “if democrats fail”, its about the GOP extending opportunity to hispanics and abandoning the Sailer Strategy.
      They have had zero luck with that so far, because the rabidly anti-immigrant base wont allow them to reach out to hispanics.
      That is why I think Julian Castro will be on the ticket in 2016. To counter Marco Rubio.

    • Ms. Jay Sheckley

      Warren Castro 2012
      [Romney Ryan 1040]

  • wheelers cat

    AND…if Obama is re-elected (and I think he will be, based on empirical data) he will have the opportunity to appoint at least two supreme court justices.
    It may be that radical redistricting will come under the scrutiny of a liberal court.

    • Anbruch

      The problem is not so much that the redistricting is all that more radical in principle than it has ever been, it is just much more precise.

    • wheelers cat

      Redistricting is still profoundly anti-democratic. Look at how the justice system has treated voter suppression laws this cycle. Only Kansas and Tennesee have voter id laws.
      Redistricting flew under the radar. I didnt know anything about the effects until Dr. Wang’s excellent analysis.

  • Olav Grinde

    Wheeler’s Cat, I much appreciate your optimism.

    Yes, I think you may be right, with one proviso: if the Democrats are able to get out the vote. That will depend on firing up enthusiasm, which currently is at a dismal level — which in turn depends on tonight’s debate.

    I must confess that I feel a great sense of trepidation about tonight’s debate. First and foremost, Obama must appear presidential. If he once again fails in that endeavour, I think it’s game over.

    Second, he must be engaged and empathetic, expressing that he is genuinely concerned and stands alongside the person asking the question — and by extension cares about the voters sitting at home in their living room.

    Thirdly, President Obama cannot let Romney get away with a stream of lies and new stances on the issues, allowing him to pretend that has been the view of the GOP candidate all along.

    If he can do all this — and I am convinced Obama has the potential — then he will be able to restore a lot of the confidence he lost.

    PS. Expect Mitt Romney to deliver some of the zingers that he didn’t use in the first debate.

    • Reason

      I have to agree. If Obama does not do these things tonight, it is over. The momentum will most likely be with Romney. I also wanted to say this site is an oasis of sanity in this MSM, polling spin.

    • wheelers cat

      I am going to take the heretic pose here (props Nassim Nicholas Taleb) and question the conventional (Bayesian) wisdom that enthusiasm correlates absolutely with likelihood of voting in 2012.
      I just dont think R^2 = 1 for enthusiasm and likelihood of voting ANYMORE. Maybe it did in the past.
      I know a LOT of depressed democrats and they are still going to vote. Even Sullivan for all his mega-histrionics is still going to vote for Obama.
      And look at RAND, which corrects for the enthusiasm/response no-enthusiasm/no-response problem.

    • Sam Wang

      Mood would only need to alter the probability of voting by 5% to lead to the swing that was seen post-debate. Resolving that by interviewing your friends would require hundreds of documented conversations.

    • wheelers cat

      Alternatively….maybe R^2=1 for “conservatives.”
      And not for liberals.
      Because we are not the same.

    • Anbruch

      I also don’t think Democratic enthusiasm is really down to an exceptional extent at the moment—at least if I’m to judge by Twitter and the blogs I visit. It bottomed out a week ago Monday and has been increasing since. A good debate performance tonight will obviously help that, a poor one will hurt it, and something in the middle will probably continue to ride the natural wave up.

    • Matt McIrvin

      Actually, Andrew Sullivan can’t vote at all. Until recently (policy lifted by Obama), HIV-positive immigrants were forbidden from naturalizing, and actually getting citizenship takes some time.

      I suspect the fact that he can’t personally vote is one of the things that’s frustrating him, and I can understand that.

    • wheelers cat

      wow, ty Matt. you are likely right.
      Poor Andrew probably sees Obama as his hope for a vote.
      He’s got skin in the game.

      I was worried in the middle of the night about Obama and teh wimmenz (my night terror du nuit). Turns out its just republican spin.

  • Some Body

    I think (if you have time to do that at any point) it could be a good idea to give this analysis sharper historical perspective. The grey cloud passes through (0,0), but, just from looking at the points on the graph, there seem to be clusters above and below the line passing through (0,0). It is quite possible that there were periods in the past when there was similar bias in the Republican or the Democratic direction. In other words, the situation we have now may be more normal than it appears.

    Apart from that – Republicans had some apparent redistricting following the 2000 election, which eventually backfired. This may happen again this decade.

    • Marc

      Yes, I was trying to do the same thing to see if historically there has been a structural advantage for one side or the other. Don’t just draw a 45º line, though.

      Draw a line through (0, 0) and (10, 43.5). Points above the line represent elections where the Democrats benefited from a structural advantage. Points below the line represent the same for Republicans.

      So, it looks like historically, Democrats have been better at gerrymandering. Would be interesting to see the structural advantage plotted over time.

    • Some Body

      Marc – there are many possible explanations for this effect.That’s why it’s not enough just to do what we two did, and it’s better to actually examine the raw data (though that’s not something I can volunteer to do myself; but maybe someone else can…)

      First of all, we should have the right baseline (the point (0, 0) must be there, but (10, 43.5) is debatable). Sam claimed in the past that you’d have approximately 6 seats for each percentage point of the popular vote, based on history. That would give you (10, 60), which, I must say, also looks more correct on the graph; you still have most points above that line.

      Then we should check if the points above the line and those below the line tend to be from different decades, or just plot the lines for each decade to see if there is a systematic tendency for a decade’s line to be above the overall line or below it, but not to cross it. It would also be good to see who got to do most of the gerrymandering every decade and see if the decade’s results indeed give an advantage to the gerrymandering party or not.

      Then we should separate out incumbency effects. the Democrats controlled the House since FDR’s time and up to 1994. That can be enough to explain why most dots are above the line. We should calculate to what extend the distance of election n from the line depends on the y axis value of election n-1.

      Now, if after all that, we have a noticeable effect favouring one party over another – this would be strong evidence for effective gerrymandering.

    • Some Body

      PS: I see where you got the 43.5 number (100% of the electorate = 435 members of the House), but that assumes the function there is strictly linear (i.e. the distribution of partisan leaning among congressional districts is flat, covering the whole field from 100% R/ 0% D to 100%D/0% R), which doesn’t seem to be the case, and doesn’t fit reasonable expectations either. It stands to reason that there are far more districts with a partisan composition in the vicinity of 50/50 than at the far edges of the spectrum.

    • Marc

      I agree that there are several reasons why the data points would not be evenly distributed about the ideal (0, 0) (10, 43.5) line. But, that’s what interested me in the first place. Can we tease apart those reasons for deviation from the ideal?

      Gerrymandering would be one (but not the only) reason. Unequal populations of congressional districts might be another. Average is around 650,000, but the range is about 500,000 to 950,000 per Wikipedia. Unequal margins of victory per district. Incumbent effects. Etc.

      If Sam has found a ratio of 6 seats to one percentage point, why is it six rather than 4.35? If it favors one side over the other, why? Does that advantage change over time? I don’t propose that we can answer these here, but they’re interesting questions.

      By the way, the magnitude of these sort of deviations from ideal are much larger in the UK House of Commons. In that case, the difference favors the Conservatives and Labour significantly over the Liberal Democrats. in the 2010 election, actual seats were (306, 258, 57) compared to ideal seats (235, 188, 149) respectively out of a total of 650 seats. See this link:

  • Tracy Lightcap

    I don’t want to rain on the parade here, but I’m not sure that the redistricting problem Dr. Wang cites has quite the effect he says it does. See:

    I think McGhee is probably right about this; redistricting doesn’t seem to have much favored one party or the other, despite the appearances.

    The Cat is right too. We know that more then half of the “unenthusiastic” voters who say they aren’t likely to vote in November end up doing so anyhow. Enthusiasm can move likely voter polls, but, unless one side or the other sees they don’t have a snowball’s chance, it doesn’t seem to make much difference in terms of turnout.

    Well, we’ll see who’s right on this soon enough.

    • Sam Wang

      Incorrect. The idea that there is no effect is contradicted by facts and quantitative analysis. That work you cite has error bars so large as to be impossible to derive your conclusion. In fact, any outcome in the last 20 years would fit within their prediction range. I have written on this extensively.

      My firm prediction is that the outcome will be near the black diagonal, a Republican-favoring parade route. Probably within the red segment.

  • Reason

    I think it is just safe to day that, we really will not know for sure until Nov. 7th. My concern, is not data driven polls like this but the hyper partisan polls that are affecting swing states like Ohio and Pa.

  • Ms. Jay Sheckley

    Am I unhappy that the popular vote ≠ EV? Sometimes. More in the past, and on a partisan basis. There’s a referendum to retire the Electoral College, passed by New Jersey, for instance, but maybe it’d be used to divide and conquer in some new, slick way.
    On the other paw, some days it seems these guys are running for president of Ohio. On the right rear paw, the guy who’d make Youngstown flourish could do a lot for our nation. It’s possible that Youngstown St Vincent de Paul Soup Kitchen photo op will become a 47% moment.
    And I wonder if the weird combo of a blizzard of lies in debate and in conspiracy theories in “2016”, just the quality of their anger instead of honest debate, turns this series of encounters into a black swan?
    RE voting if there’s “not a snowball’s chance”, One Thanksgiving we may wake up to find that enough polls had been tweaked and over-reported enough to effect the outcome.
    It’s silly to trust the polls so much that you think no one has to vote. _Ceçi pas une pipe_ . The rooster is not the sunrise! Am I the only one who, faced with defeat, thinks- let it not be _my_ fault? If you can vote and don’t, however poor the prospects, why should even you believe you cared?
    On the subject of tabasco, I’d be more interested if you’d offer to drink a teaspoon to tablespoon of tabasco if the result were out of your RED prediction zone. Isn’t red more suited to tabasco anyway?? As for that _cup_ of tabasco, the net reports different results: (1) colorful projectile vomiting, (2) great pain, (3) hospitalization and (4) winning $130. Eat a lot of bread or rice first.

    • Albert Ericson

      Isn’t it “Ceci non pas une pipe”?

    • Alan Cobo-Lewis

      Really want to reply to Albert Ericson, but there’s no link to do so.

      Ceci n’est pas une pipe.

      But since I’m replying to Ms. Jay Sheckley, let me also say “St Vincent de Paul Ryan”, a la Paul Ryan Gosling

    • Albert Ericson

      Oops — you are right! But the point is it is not a pipe, only the picture of a pipe. And the current polls are not the election, only an (early and perhaps inaccurate) picture of the election.

  • Ms. Jay Sheckley

    More interesting poll than it seems at first- it’s believable that Florida has enough elderly voters to skew the result. Look at New Hampshire, on PEC whire; on 7-election in stripes.

  • Slovo

    “Obama extends slim lead over Romney in White House race”

  • Ralph

    wheeler’s cat: I agree with that guy’s conjecture on the Gallop poll. Their ‘likely voter’ questions are biased against likely Obama voters. I would think that they’d know that.

    I know when I was called by Gallop, they asked no questions that I could construe as relating to likely voters. Yet, they were reporting likely voters.

    I had some personal experience with a Gallop poll at my work place. The group I was in had 12 people. The margin of error was ±29%, yet they were making recommendations of remedial action based on 12% variations.

    In other words, as an industrial tool, Gallop polls are worthless. One can only assume that it is part of their core business strategy.

  • Patrick Draut

    The possibility of an EV tie is increasing (> 1%), which may cause the Interweb to implode if it came to fruition.
    In a tie scenario, BHO holds onto Ohio, NH, and Wisconsin… while losing Iowa, Colorado, Florida, Virginia, and Nevada.
    I know Dr. Wang does not like to look at specific combinations as there are quadrillions; however, this map is certainly plausible based on polling. If Romney makes up ground in Iowa and Nevada it has potential to become the most likely outcome.

  • Ohio Voter

    Also some news today coming out of Ohio. The Supreme Court is declining to block early voting efforts in Ohio. It’s estimated that about 100k Ohioans will vote in the three days leading up to the election.

    In Franklin County (home of Columbus), about 17% of voters cast early ballots in 2008 with a 4 to 1 Democratic advantage

  • LDK

    Why the gap between Dem. senate votes and Obama? Hard to imagine that many people split ticket vote. Most people would not seem to be that strategic.

  • Reason

    Question about the Gallup LV poll. It now as Romney up 4 from just yesterday. Are they sampling in red states that would have gone to Romney anyway?

  • Mark F.

    An EV tie means the (probable) GOP House decides the election in favor of Romney.

    • Reason

      Actually, not really. About half the states have in their voting laws, that any congress person has to cast the vote for the majority winner in their state. So if a Republican congress person is in an Obama winning state, they have to vote for him. But I still think the numbers are on the Repubs side, even after the election.

    • Pat

      That’s an interesting point. Has anyone looked at this more carefully? Considering which states have the rules that Reason mentions and considering the current delegations of each state (as a benchmark) could provide an interesting estimate of what might happen in case of 269-269 tie.

  • Olav Grinde

    Incidentally, Nate Silver makes an interesting observation in today’s entry at

    There is no evidence… that the second presidential debate is any less important than the first one. On average, it has moved the polls by 2.3 percentage points in one direction or another — almost exactly the same as after the first debate, which moved them by 2.4 percentage points on average.

    That would suggest President Obama has a significant chance to recover and “erase” the impression of his dismal effort in the first debate.

    Yes, the Supreme Court decision is good news for the Democrats — but more importantly, it’s good news for democracy. When secretaries-of-state see their main job as repressing the vote of certain demographics, rather than enabling as many people as possible to vote, then something is terribly wrong!

  • Michael

    “About half the states have in their voting laws, that any congress person has to cast the vote for the majority winner in their state.”

    The House doesn’t vote individually, they vote by state delegation. One vote per state. Romney will win many more states even if he loses the election.

    • Michael

      Oops. Didn’t say that quite right. Romney will have to win more states than Obama in order to get to a 269-269 scenario.

  • Ms. Jay Sheckley

    There was a new paragraph by Sam that just winked in and out of existence like a metaphysical bookstore. Something about contradictory polls. Thanks, Sam, for all your efforts.

Leave a Comment