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

Charlie Cook skeptical about a House switch

September 29th, 2012, 9:00am by Sam Wang

Via Andrew Sullivan, veteran election-watcher Charlie Cook says the House is currently not looking like it will switch. Of the Congress-watchers, I think Cook and his team are very much worth listening to for precise snapshots at any given moment.

However, their snapshots are subject to change over time. In September 2010 he forecast GOP gains of over 40 seats, which grew to 50-60 seats by Election Eve. The outcome was a gain of 63 seats. That was a wild midterm year, and it says nothing bad that he was a little bit off at the end. But more to the point, that year his September report was about 20 seats off base.

It didn’t matter, in the sense that a 40-seat gain was already enough for the GOP to take control of the chamber. But this year, if the outcome were to go toward the Democrats from Cook’s current prediction by a similar number, it would reverse control of the chamber. So I think it is a bit early to reduce House control into a categorical yes/no judgment. There is general agreement that the GOP will lose seats – we’re just haggling over the number.

One thing Cook does do is give an indication of which way he thinks the wind is blowing. He writes: “the next week or 10 days are…critical for Romney and the GOP. If things don’t turn around, a stampede could ensue.”

But perhaps because of his (and his staff’s) focus on individual races, he seems to have made a second error: “For the most part, the deterioration of the Senate outlook is unrelated to Romney’s problems at the top of the ticket.” As Rick in Miami found the other day, the presidential and Senate fortunes track the Presidential race rather closely.

This is clearly indicative of downticket coattails. The reason this matters is that enthusiasm for the top of the ticket has loud echoes in Senate and House races. If the Meta-Analysis is above Obama 330 EV on Election Eve, the Democrats should have quite a good shot at re-taking the House. As stated before, I regard them as having a very slight advantage. I will be interested in Cook’s take on things, 3 weeks from now.

Tags: 2012 Election · House

25 Comments so far ↓

  • Ottovbvs

    There’s no doubt if Obama wins then the Dems are going to make large gains in the house. In strictly political terms if the Republicans hold onto the house but with a very narrow majority say single figures this could be the best outcome for the Dems because it locks the Republicans into responsibility for governing. There’s much legislation that has to be passed and Boehner will need Democratic votes to pass it and will be forced to compromise on a lot of Republican ideology. This is going to cause mayhem in the GOP.

  • Olav Grinde

    Good point.
    Let’s not forget that the Democrats are a divided party. There are Blue Dogs that are extremely opposed to socially progressive legislation. Moreover, it’s literally impossible to pass legislation that really bites when donors/lobbyists have bought votes on both sides of the aisle.

    Republican senators and congressmen, on the other hand, have to an impressive (depressing?) degree been marching lock-step. They might do so even more if the Democrats ostensibly have “control” of both chambers of Congress.

    • Brian


      I don’t think the Blue Dogs are a good example if you want to prove that the Democrats are a divided party. The Blue Dog caucus, at its height, consisted of 54 members after the 2008 election. The 2010 election cut their membership in half, to 26 (10% of the 243 elected Democrats in Congress). Compare with the Progressive Caucus which currently has 79 members (32%), and the Black Caucus which has 42 (17%). Seems like most Democrats are pretty much rank and file. (Note: I included Independents in my count of Democrats in the Senate).

      Similarly, there are 66 Republicans in the Tea Party Caucus, plus an additional 4 who are essentially Tea Partiers but didn’t join because they reject the formality of a Caucus. Those 70 represent 24% of the 287 Republicans in Congress. The Tea Party caucus is 2.7 times as large as the Blue Dog caucus, and it’s ability to disrupt the GOP’s agenda shows. The large Blue Dog caucus in its heyday was similarly disruptive to the more progressive Obama/Pelosi/Reid agenda in 2009/2010.

    • OwlofMinerva

      Brian – the Progressive Caucus and Black Caucus are largely overlapping. And I do not know what you mean by “rank and file”.

      The Democrats are not by any meaningful way “divided”. They are strongly behind the Obama administration’s agenda. They would only be divided if the progressives refused to go along with it, but that has not been the case. Obama has had remarkably little trouble from his left flank as he carries out the Washington consensus.

    • Ottovbvs

      The Democratic blue dogs are a relatively small part of the Democratic caucus (10-25 seats I believe and some of these might disappear) whereas the hard right in the GOP caucus is in the 90-100 range judging by voting tallies on highly ideological subjects like raising the debt ceiling or passing the farm bill. If the Republican majority were cut to single or low double figures they are going to represent half the Republican house caucus. Boehner could barely control his caucus in the present congress so if half of it became these nihilists he’s going to lose control completely. Assuming Obama wins and the Dems narrowly hold the senate Obama is going to have the whip hand in the Bush tax cut and sequestration negotiations. I don’t expect the Republicans to concede a compromise on the tax cut expiry (although they might if it was part of a larger deal on sequestrations) and I suspect Obama wouldn’t be unhappy to see them all expire after a lot of kabuki theater because he gets deficit reduction and is able to blame the Republicans for blocking the extensions for the under 250k only group.

    • OwlofMinerva

      By “Washington consensus”, I mean Democratic Party ex-DLC centrist policy consensus – not meaning to imply the Republicans are in agreement.

      Also, Brian, the 32% progressive caucus gets smaller when the Dems pick up more centrist seats. I believe it was 25% in the 111th congress. And of those, only a single member is in the Senate (and he’s not even a Democrat).

      In short, let’s not exaggerate the influence of the congressional progressive caucus

  • Amitabh Lath

    I am having trouble with two opposing concepts.
    a) We are told that the undecided fraction is quite small this year. b) The D-R is shooting upwards really fast. Look at just the last week!

    One would suppose b) happens because undecided voters (finally) make up their minds. But a) says no, there aren’t that many. So R’s are becoming D’s?

    If so, then all sorts of narrative goodies like gerrymandering congressional districts etc have to be rethought. If a chunk of reliable R voters (say, women) are no longer reliable?

  • John Jacobs

    If you use ‘less smoothing’ it shows that republicans are actually slightly ahead of democrats in the latest polls . Are these wild daily fluctuations in any way reflective of the state of the race?

    • Sam Wang

      If you mean the HuffPost software, I think the less-smoothing option forces the line to terminate with the most recent poll. There might be movement back toward House Republicans, but these are such noisy polls. Better to wait and see. Oh, my knee…

  • Paul K2

    Amitabh, I think you just hit the nail on the head. The question is whether Republican women that cross to the Democrats for Obama (and some clearly are, check the gender gap), will also vote D down ballot. If Obama wins by 5-6% because the women crossed over, then its hard not to believe that there won’t be down ballot impact.

    OTOH, I think Sam Wang is underestimating the power of gerrymandering this cycle. Clearly it has made a huge impact in some states… For example, in Pennsylvania, the Democratic congressional vote should swamp the Republican votes, yet the Rs currently hold 12 of the 19 seats, and due to redistricting, two western D seats have been combined eliminating one incumbent. Right now, the likely outcome has the Rs getting 12 of the 18 seats, even the Democrats could get larger vote totals.

    The gerrymandering software used in PA was developed by Carnegie-Mellon University, and this is the second time the Rs have used it to redistrict the state. They have gotten really good at this game, using it to overcome one person- one vote principles.

    • Sam Wang

      Paul K2, your argument is the same kind of qualitative argument for which I found Weigel at fault.

      A quantitative arugment would go like this: the difference between 12R/6D and 9R/9D is a swing of R+6, which corresponds to approximately 1.0% of national vote swing. If you added up all such effects of redistricting nationally, this would give a net change that would address my question. Your cherrypicked example is probably among the most favorable to the Weigel argument, so there is some question of how large the effect is. Indeed, the fact that four cycles of redistricting have failed to lock in a Democratic advantage suggests that the total effect could be fairly small.

      Such a sum is also complicated by the fact that different-sized swings can have very different effects. For example, in case of a large swing, redistricting can backfire, as I pointed out. That R+1.0% advantage could disappear or even reverse with a 10-point swing, which is about what we have at the moment. I don’t doubt that the Carnegie-Mellon people are good at this, but surely there are limits.

      I recognize that there is a point to be made here. However, an argument like yours is not sufficient until it is extended to all districts, and for a variety of swing amplitudes.

  • Nancy

    Hi, Sam,

    This is a bit off-topic, but I’m wondering, having just seen a video about how the Electoral College allows a candidate to win with a startling minority of the popular vote (I’d known it was possible but not just how little it would take!)–would all states splitting their EVs proportionally as Maine and Nebraska do be more fair, or would the power of a vote still be dependent on geography? I’m very curious to hear your thoughts. Thanks!

    • Zach C

      Nancy, It would do very little to ensure that a popular vote winner didn’t lost the election. According to Fairvote, “In 2000, again, George Bush lost the popular vote to Al Gore by 0.52 percent, but led in the Electoral College by 0.93 percent. With the congressional district method of allocation, Bush’s share would have exceeded Gore’s by as much as 7.06 percent.” Because each candidate would get 2 electoral votes for winning a state plus the number of districts won, this would give a lot of undo weight to small states. The Republican party traditionally wins the majority of states because its base of support is more rural. Hence why Bush would had won in 2000 by a greater margin under the Nebraska/Maine system.

    • Matt McIrvin

      One way to look at it is that in the current system, the electoral advantage given to small, usually rural states is roughly balanced out by the underrepresentation of the similar voters in rural areas of large states containing major urban areas.

      If all states used the Maine/Nebraska system, the electoral small-state advantage would still be there, but rural areas of large states would not be so underrepresented. The likely outcome would be a large bias toward Republicans (unlike going to a true national popular vote, which would be more of a wash).

  • William Ockham

    You folks should take a close look at the methodology that Rand is using to calculate the shift between candidates. It is not at all equivalent to someone changing their mind.

    Here’s a different way of thinking about the presidential election. Assume that by Jan. 1 that everyone has already committed to either the Dems or the Reps. There are no undecided voters. No one ever really changes their mind. Even the people who think they are undecided have a presidential preference that is fixed (and at an aggregate level, very predictable). The only decision that a voter makes is whether or not to vote.

    The movement you see in the polls is made up of random noise and changes in peoples’ feelings about the candidates/parties. Those changes in feelings represent nothing much important until around Labor Day when normal people (not the folks who come to this website) start to take politics seriously and begin to decide whether or not they will act on their feelings. People who decide that they aren’t going to vote begin to tell pollsters that and get filtered out of the poll results.

    The reason this matters is that it means that the movement of the polls isn’t a random walk between here and the election. People are making up their mind whether or not they will vote. It will become harder and harder for any change to occur. Don’t look for any dramatic changes, certainly nothing that will move the meta-margin.

    The more interesting development is that the Senate races are becoming nationalized. The House has been for the last three presidential election. I’m going to disagree with Prof. Wang and say that it’s fairly certain that Nancy Pelosi will be the Speaker of the House for the next session. The presidential polls are a better indicator of the House vote than the generic House polls.

    • B

      William – wonder if you could elaborate on your first paragraph and explain what the actual RAND methodology for “shifts between candidates” is – if not what it sounds like.

      I took a look at their methods PDF and it seems #9 on pg 7 is the relevant section – but I’m not a math person so I’m not following.

    • Joseph

      The Rand report also indicates that about 8% has consistently not identified their preference. We of course don’t know whether that’s because they are undecided or because they just don’t like revealing their preference. Anecdotal evidence from my wife, who has been calling for the Obama campaign into registered voters, suggests there are very, very few people who have not made up their minds, but a very large number who refuse to say what they’ve decided! She estimates it may be as high as 20% of the people she’s talked to.

      Related to your comment, that may very well mean that we have people who would vote Obama IF they voted and people who would vote Romney IF they voted, and that those that are actually motivated to vote are decreasing on the R side and increasing on the D side.

  • Amitabh Lath

    Partha asked “That is an interesting observation. Could you provide a link? Thanks.” I referred to D-R but I was actually looking at 2012 “National House Race” at

    I zoomed in from 40% to 50%. I kept all polls, but you can filter robopolls or internet polls depending on your theological beliefs.

    The slope on the blue line is +5%/ couple of months. The red line looks like a low amplitude sine wave, going nowhere.

    This is not a plot that moves much, so I was drawn to this monotonic rise feature.

    I confess I do not know what could cause this, absent undecided voters turning blue. In this plot, R seems pretty flat.

    • Partha Neogy

      Thank you. I think I see what you mean. The blue line has moved up by 2 to 3% since August, while the red line has remained essentially unchanged. I suppose one could arrive at the conclusion that you did, but the scatter in the data gives me pause.

  • William Ockham

    Well, the first problem with the Rand methodology is that they allow people to split (give 70% to Romney, 30% to Obama) which is interesting information, but you don’t get to vote that way…

    They also give you the option of saying how likely you are to vote at all. Again, it is good information, but in the end, it is going to be all or nothing.

    But the biggest problem with the calculation of the shifts (from Obama to Romney and Romney to Obama) is in this sentence:

    For the computation of these shifts, we first combine the probability of nonvoting and the
    probability of voting for an “other” candidate.

    The probability of nonvoting comes from an answer to a completely different question than the probability of voting for other. This leads to a systematic overestimation of the shifts.

    Here’s a simple example. Let’s say someone’s last week’s answers were 80% chance of voting with a 75% preference for Obama and a 25% preference for Romney. This week they report there is 90% chance they will vote and 80% preference for Obama and 10% preference for Romney.

    According to Rand’s formula, because the chance of not voting and the chance of voting for Romney are going down, Obama’s shift is +15% (+10% for the increased chance of voting and +5% for the increased preference for Obama). Romney loses 5%. I think this is clearly wrong.

    If you are going use numbers like this, I would say that in week 1 of our example there was a 60% chance of voting for Obama (.80 * .75, the chance of voting multiplied by the preference for Obama) and a 20% chance of voting for Romney (.80 * .25). In the second week, there was a 72% chance of voting for Obama (.90 * .80) and a 18% chance of voting for Romney (.90 * .20). Therefore, the shift to Obama was +.12 and the shift from Romney was .02. In this case, they have overestimated the shift to Obama and away from Romney.

    If you run their numbers, they show a higher total shift to Obama (after subtracting all the shifts to Romney) than the gain that Obama has made.

    More fundamentally, these numbers don’t generally reflect one person flipping from Obama to Romney or vice versa. These numbers represent changes in people’s perceptions of the strength of their support for their more preferred candidate. That’s useful information, but it isn’t a reflection of a changed vote from one candidate to the other.

    If I had access to Rand’s raw data, I would be more likely to model the data this way. Every respondent would start with a vote of 1. That would get discounted by the chance of not voting, i.e. if you are 80% likely to vote you would be down to .8 vote. Then, I would subtract the reported chances of your less preferred alternatives from your highest preference, bounded by zero. So, if you didn’t report having more than a 50% chance of voting for either of the two major party candidates, you wouldn’t get counted at all. If your preference was D 70%, R 20%, Other 10%, I would count you as a 40% of a D vote multiplied by your chance of voting. Why? Because elections are decided by the people who are certain.

    • wheelers cat

      true, but like the RAND survey explains, values outside the MOE represent a true change and not noise.
      Have you read the PDF?
      The designers state that the RAND survey is a measure of change in opinion. The likely vote weight is a part of their design, and its a feature, not a bug. I think, because of the design of the panel, RAND also measures the penetration of internet memes.
      Although the panelists were initially randomly selected, all were offered internet access if they didnt initially have it.

      “If I had access to Rand’s raw data, I would be more likely to model the data this way.”
      Knock yourself out. You seem to want to measure certainty of voting. That just isnt what RAND is attempting to measure.

  • Michael K

    I am largely persuaded by the scatter-plot of the last 33 elections that Cook is over-rating the GOP’s re-districting advantage.

    That said, there is good reason to believe that Republicans should have an inherent structural advantage over Democrats when it comes to gerrymandering.

    As a general rule, while densely populated urban areas overwhelmingly skew Democratic, rural areas skew Republican but typically by a much lesser amount. The most Republican-skewed rural districts in the country (AL-6 and TX-13) have PVI’s of R+29. Meanwhile there are (by my manual count) 20 urban districts that are D+30 or more (led by NY-15 and NY-16 at D+41).

    Not only that, but those densely populated blue regions are typically surrounded by competitive or blue-leaning suburbs. So it’s extremely tough and perhaps implausible for Democrats to gerrymander that geographically-tiny blue core into many D+10 or D+15 districts stretching into deep-red rural areas. Whereas it ought to be much easier for Republicans to gerrymander that deep blue core into a few super-concentrated blue districts, and then combine small cities and suburban/exurban areas with their rural strongholds into R+10 or R+15 districts.

Leave a Reply to Michael K (Cancel)