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

Senate update: downticket cutting loose…or still catching up?

October 13th, 2012, 9:00am by Sam Wang

Up to now, Senate fortunes have tracked the ups and downs of the Presidential race. What about now, after last week’s downturn?

One measure is the likelihood of getting to 50 or more Democratic/Independent seats. In today’s update, the probability* of Democratic retention is still high, 95% (calculated by splitting the 50-seats scenario in half). The median outcome is 54 D/I, 46 R. The 68% confidence interval is 53-55 seats.

Last week’s Presidential crash is not yet apparent in Senate races – though in most cases, fresh information is not yet available. Where we do have post-debate Senate polls some races have not moved much (MA, WI, VA), or have followed the Presidential ticket down – and perhaps back (OH). I am a bit puzzled as to why downticket races didn’t follow the President downhill. Given the infrequency of Senate polling, I may never find out the answer.

By the way, Indiana, North Dakota, and Montana still appear to be on the edge. This is a place where you can make a difference for your side. Visit ActBlue (D) or Crossroads GPS (R).

Thanks to Rick in Miami.

*The histogram and graph above assumes that 50 D/I lead to Democratic control,  in which case the probability is 97%.

Tags: 2012 Election · Senate

47 Comments so far ↓

  • Anbruch

    If I’m reading the graph of the National House Race numbers correctly, they have followed the presidential race, though not as steeply. If that’s the case, it’s really only the Senate races that have not followed presidential race. That seems extremely curious and it suggests to me either a systemic problem with the polls, which seems unlikely, or that a certain portion of voters now supporting Romney do not trust him. At the same time, all things being equal (which I know they are not) the fact that the Senate races are this far out of sync with presidential and House races is evidence against the shift from Obama to Romney being all (or even mostly) driven by enthusiasm.

    • Sam Wang

      Or we give it time.

    • Anbruch

      Sam, I gather from your comment that you believe the races remain locked and the difference is a product of available polling on the various races. From a statistical point of view and given the inevitable uneven polling of races, what would we need to see in the meta-analyses in order to consider the gap between the Senate races and presidential race as likely real?

    • Sam Wang

      I think if the race persists at <O+1%, Senate races will narrow. It will take a week to find out.

  • kel

    State level issues more important? Reluctance by swing voters to invest control in one party?

  • Philip Diehl

    For you Dems thinking of making contributions to Senate races where you can make a difference, here are recommendations based on last night’s polling and probabilities from Five Thirty-Eight:

    Number 1: Indiana where the D is running against a Right Wingnut. The R is highly vulnerable and if Donnelly wins, it’ll be a big pickup. Here’s the link to contribute:

    Number 2: Montana where the D incumbent is behind but has a shot at winning. Your $ goes a long way in the Montana media markets.

    Number 3: Connecticut where the D has a narrow lead over a WWF magnate who’s spending boatloads of her own money–the only reason she’s competitive.

    Number 4 (tie): Claire McCaskill in MO and Elizabeth Warren in MA who have opened small leads but have not nailed down their races. Same for Kaine in VA.

    Give till it hurts.

    • MarkS

      Or, you could just go to Sam’s Act Blue page (link at left), and contribute to any combination of these three in one easy step.

  • wufwugy

    The Senate races didn’t follow the Presidential race because the shift is response enthusiasm for Romney. No partisans switched sides and true non-partisans didn’t watch the debate; this is really just a lot of people decided on Romney finally seeing that he’s not a complete looney, thus they’re comfortable expressing their opinion

  • Ms. Jay Sheckley

    As a moderate, I just give to the DCCC and let them decide.

  • Ms. Jay Sheckley

    My household is confused: Before this is over, you expect the metamargin to fully close and flip?

    • wheelers cat

      I do not Ms. Jay.
      The temporal window has closed for the curves to flip.

    • Matt McIrvin

      I predicted it would flip a couple of days ago, but now I’m just waiting and seeing. We’re in a period in which events occur faster than polling windows can open and close; the vice-presidential debate is a new variable in the mix, and the second presidential debate will happen before the polls have properly responded to that.

  • Obama 2012

    538 bounced off a bottom today… seems like a very good sign. (along with some movement towards Obama in a few trackers)

    • Froggy

      Not much released today — “Saturday, not Sunday, … is the slowest day of the week for polling.”

      Hey, did anyone notice that Nate Silver is now providing error bars on his EV and Senate forecasts? Obama at 285.4 EV +/- 72 EV doesn’t exactly give me a warm feeling.

    • Sam Wang

      That is an amazingly large error bar. It includes every single forecast of the season from anyone.

    • CobaltinSF

      …yes and the IBD/TIPP poll has the race at a tie today, stating on its web site that Romney’s bounce has receded. More surprisingly, the independent, non-affiliated Behavior Research Rocky Mountain poll just out within the hour has OBAMA 2 POINTS AHEAD IN ARIZONA. How can this be? No one has focused on Arizona, assuming it was deep red. Perhaps this is a case of the democratic Senate candidate, Carmona, actually helping Obama up-ticket??

    • Rick

      The bounce off the bottom may be because of a dearth of weekend polling and a weird poll of Arizona. Friday was a horrible day for Obama. I’m going to be on pins and needles to see what the newspaper polls tomorrow look like. The horse race has been relatively stable for the past few days, and RAND ticked up today, so maybe the bleeding has stopped. But I think we’ve all said that every day since the day after the debate.

    • Olav Grinde

      Obama +2% in Arizona? Now that is interesting!

    • Olav Grinde

      One more point: I experimented with PEC’s Electoral College Map…

      Giving Romney all of the South (including Virginia and North Carolina) + Florida + Ohio + Colorado + New Hampshire.

      And with Obama taking Nevada + New Mexico + Arizona.


      Romney: 268
      Obama: 270

    • Froggy

      I wouldn’t get too excited about the Arizona poll. There have been exactly three Arizona polls in the last two years that have shown Obama beating Romney. This pollster accounts for all three.

    • Michael Worley

      It seems like the error will narrow. For a 95% confidence interval, it’s not that broad… it’s not like we know exactly which way things will fall.

  • CobaltinSF

    The bounce off the bottom is real. PPP Ohio poll just out. Obama up 51% to 46% Romney. 19% of Ohioans have voted and of those 76% voted Obama/24% Romney. Check it out at Politico….there seems to be a skew between high numbers of people who have voted and say they’ve voted for Obama, versus likely voter numbers…wouldn’t the actual voter results be more accurate???

    • Rick

      This is an honest question from a hardcore Obama supporter. Are polls from PPP any more credible than polls from R pollsters like McLaughlin? Or quasi-R pollsters like Gravis? I want to believe, but I’m a Democrat, which means I always think we’re losing. ;-)

    • Billy has some great statistics on early voting. Obama has a +6 ballot request advantage in Ohio. Granted, there is some discrepancy between requesting and actually casting the ballot, but there’s some data floating around on that site and elsewhere on the web showing that it shouldn’t add too much variability.

    • Obama 2012

      I really think the pollsters are underestimating the Obama turnout machine with their LV models. They have invested a huge amount of money/energy into their GOTV efforts. They have three times as many field offices as Romney.

      And I really think they’re just underestimating enthusiasm for Obama in general – 4 million people have donated to his campaign in 2012 (3 million in 2008.)

    • Matt McIrvin

      PPP has a Democratic house effect but it’s not +5.

      In general, these house effects are smaller than people think they are when they complain about the polls… even for Rasmussen. Any single poll could be an outlier, of course.

  • CobaltinSF

    PPP is the third Ohio poll this week that has Obama over 50% there, the other two were CNN and NBC/Marist…and the AZ poll isn’t weird, Behavior Research is a very respected pollster…they, unlike many pollsters, actually conduct polls in Spanish if the voter speaks Spanish. So this poll may be evidencing that the Latino vote is being grossly undercounted by pollsters who will only conduct their polls in English, i.e. CO, NV and FL may be skewed toward Romney because I believe most Spanish speakers vote dem…

    • Olav Grinde

      Now this is interesting, and it’s something that I haven’t seen discussed here. Is it true that some “professional” pollsters are basing their numbers on English-only robocalls conducted in heavily-Latino states?

      I would think that has the potential to skew the results so much that it renders the entire poll worthless!

    • Arbitol

      The “latino voters who prefer to speak Spanish” hypothesis was floated in connection with the blown 2010 polling in Colorado, Nevada, and California. Makes some sense to be – but then, I want it to be right.

    • Obama 2012

      It also makes sense that Carmona would have up-ticket coat-tails because if Latino voters are coming out in big numbers for him, they’re almost certainly also voting for Obama at the top of the ticket. If Arizona has uncharacteristically huge Latino turnout… it could be the surprise of 2012.

    • Matt McIrvin

      Was polling of Arizona way off in 2004 and 2008? If not, I wouldn’t hold my breath.

  • Ajay

    I am having trouble understanding one thing – if Obama is losing ground in the polls but Democrats are still going to keep the Senate, doesn’t this suggest that a lot of people could be splitting their ballots? Are a lot of people in Virginia going to vote for Romney and Kaine? Seems unlikely, no?

  • Leo McNellis

    I haven’t heard anything about the English-only polling issue you mention, Olav Grinde — but it is certainly a point worth considering if true.

    I’m starting to think that the recent vicissitudes in polling data might be the result of poor sampling pools. Being a strong Obama supporter, the last thing I wanted to do after Obama’s poor debate performance was to talk to a pollster. On the other hand, I might have considered talking to a pollster the day after the Democratic convention.

    If both pollsters had called me right after these two events, my voting intent would have been recorded in one poll but not the other. One thing that has not changed, however, is how I intend to vote on November 6.

    • Obama 2012

      Yup, this is why pollsters often call these “bounces” because it’s just more enthusiastic people answering the poll call… but often there’s not much of a real shift going on here. Just a short term enthusiasm boost.

      This is not to say that the debate didn’t have some real impact, I think it did – but I also think that it’s impact isn’t nearly as big as you would think from some of the polling.

      I also think it’s worth noting the huge amount of right wing pollsters in the field post-debate. Narrative setting seems to be the goal. If you take out all of the right wing pollsters – the swing state polls haven’t really been all that bad.

    • Matt McIrvin

      The media narrative created by the debate and the initial polls absolutely killed Democratic enthusiasm in the short term, and every time there’s a bit of good news for Ds (the jobs report, the Biden/Ryan debate, the downtick in unemployment), it gets counterbalanced by another horrible-sounding poll result, in a sort of feedback loop.

      I was starting to worry for a while that the gain in the process was >1, which could create runaway feedback and the kind of Reaganesque landslide for Romney that Prof. Wang says is not possible. But the Ohio results suggest that this kind of runaway is probably not what’s happening.

      It’ll be interesting to see how things shake out after the second presidential debate.

    • MAT

      A RAND style poll interviewing the same people over and over setup in all of the swing states would be pretty interesting. I think I read that RAND pays $2 a week to each of their 3500 audience, i.e. $7000 a week. Say, 8 states – $56,000 x number of weeks. That doesn’t seem like an outlandish amount of money compared to what is getting thrown around for TV ads.

    • wheelers cat

      Good commentary.
      RAND continues on the downward trend that Dr. Wang predicted today.
      inflection point 7 days after Debate1.
      @the Matts & Obama2012
      That is what I think.
      Every thing we “know” about polling and forecasting is based on Gauss and Bayes…..what if that is no longer true?
      What if the evolution of technology (cell-onlies and smartphones) and the evolution of the electorate (majority-minority markets) have changed the shape of the information so radically that we are actually only sampling the subset of the [enthused]electorate that will consent to take a poll?
      And what if (as Taleb postulates) the actual structure of reality is Mandelbrotian and not Gaussian? What if Bayes is playing us false because technological advances and demographic evolution have changed the “freed market” in information? What if we cant use priors to predict because so many events have never happened before?

    • wheelers cat

      Some of that is already available– they do Ohio, Pennsylvania and Florida. And you could ask for any state I presume.

      The fifth graph shows one of the first three graphs explained above, but now broken out by respondent characteristics. Each day we present the data broken out by a different characteristic, such as age, education, income, position in the labor market, race, sex, and state (Pennsylvania, Florida, Ohio). Anyone interested to know more about the results broken down by respondent characteristics, or about other ways data from the survey can be accessed and used, is invited to contact Arie Kapteyn, director of RAND Labor and Population, at

  • wheelers cat

    And heres my neuroscience question for Dr. Wang.
    If the only underlying structure of a given bounce is raw enthusiasm (example, Palin 2008) is the bounce necessarily “soft” given the biological basis of behavior and that rabid enthusiasm is unsustainable?
    I think information erodes enthusiasm, as in Palin’s case when subsequent interviews proved her to be a wildly unsuitable candidate, and pop culture mockery cemented those perceptions.

  • Olav Grinde

    If Romney succeeds in taking the White House, and the Democrats retain the Senate, then Harry Reid should drop any thoughts of filibuster reform.

    In this scenario, however, I do wonder whether Democratic senators would be as capable of keeping a united front as the Republicans have been in the last two years.

  • A New Jersey Farmer

    It seems to me that enthusiasm for Romney could have been the impetus for such a spike in his RAND direction, but curiously, that turn did not occur decisively until October 9 and 10. Obama even improved his standing at the same time that other polls were showing a Romney bounce. Now that the data are moving more sharply in Obama’s direction, I think that other polls will follow. Even with all of the press and right leaning pollsters, Romney is still only either ahead by 2, 1 or tied in national polls. Or this could be wishful thinking. It’s been that kind of election.

    • NC Obama Guy

      Couldn’t the change in RAND be delayed purely because they only poll 500 of the 3500 per day? Wouldn’t that also explain why the inflection point would happen at approximately 7 days after the altering event?

    • Matt McIrvin

      NC Obama Guy: That should have the effect of spreading out any sudden jump into a week-long slope. But if it’s sufficiently large, you’d be able to see one-seventh of the effect even on the first day.

      That was what confused me about RAND: some of the early poll results were implying a *colossal*, landslide-sized spike, maybe even a 12-point overnight jump going by the most Romney-sympathetic interpretation of Pew. If that kind of thing had happened in the RAND sample it’d have been immediately apparent.

  • A New Jersey Farmer

    Good point. I’m interested in hearing what Sam has to say about that. I was just looking at the Intention to Vote graph. Why would that move so much lower for Romney today?

  • NickT

    I have to point out that McCain did relatively well in Arizona because of:

    a) Home state advantage
    b) Quite a good performance among Latinos

    I doubt Romney can match him, so it would make sense for the race there to be quite a bit closer than it was in 2008, when Obama lost by 9 points. I don’t think this is the year Arizona flips, but I could be wrong. My guess would be Obama 47 Romney 52 Other 1.

  • Rick in Miami

    Monday update: I’ve added Pennsylvania (Smith vs. Casey) since that’s moved from solid D to leans D, and added all polls released over the weekend. The probability of D/I control now rests at 89%, and still shows only a very moderate drop from its high of 97% on 3 October.

    It’s now been nearly two weeks since the first debate. I am beginning to believe that the difference between this curve and the presidential debate meta-margin is real, not just a function of delayed polls for the senate races. For example, I withheld polls released after date T. (The release date is sometimes many days after the mean date of the poll.) Then I examined how long after the DNC that bounce (which was very dramatic in the Senate curve) became evident. The answer: T+7 to 8 days. In other words, if the Senate race polls are about as timely now as they were then, we have seen the effect of the post-first debate bounce in this curve.

  • HWAV

    I’m a little late to the game on this, but I don’t see how Akin can win if he loses the female demographic by 15%. Women traditionally vote in greater number than men.,5770&demographic=females

Leave a Comment