Princeton Election Consortium

A first draft of electoral history. Since 2004

Two changes in Senate knife-edge races

August 26th, 2012, 3:53pm by Sam Wang


Thank you for supporting candidates at my ActBlue page. Senate polls have shifted, and I am adding two Democratic candidates who have entered the knife-edge range (CT-Murphy and MT-Tester). Four more races are still of prime importance in Senate control: ND-Heitkamp, IN-Donnelly, VA-Kaine, and MA-Warren. For example, this article outlines the surprisingly strong candidacy of Heidi Heitkamp.

Republican readers are welcome to visit the National Republican Senatorial Committee site. I recommend their candidates in the same races.

At present, I do not recommend giving to two candidates who have left knife-edge territory: FL-Nelson and WI-Baldwin). For now, donations will not be effective at shifting those races. MO-McCaskill is off the table as well. Rep. Todd Akin has seen to that.

I estimate the current probability of Democratic retention of Senate control as 65%. This is a very rough estimate. It is an increase from the previous estimate of 52%, mainly because of Missouri. Anything between 20% and 80% should be considered a knife edge.

Senate control 65-35 D-R

Tags: 2012 Election · Senate

13 Comments so far ↓

  • BillSct

    The ct in BillSct stands for Connecticut so let me comment on Murphy vs. McMahon (disclosure: I’m a democrat and was a delegate to the State Convention that nominated Murphy.)

    When McMahon ran against Blumenthal two years ago she annoyed a lot of people with her incessant TV commercials. I don’t think that effect has completely worn off yet. And she demonstrated her political amateur status when she took credit for the attack on Blumenthal about his mis-speaking about Vietnam. What should have been an October surprise surfaced during the summer and Blumenthal had plenty of time to recover.

    Murphy is an experienced candidate having been elected to the Connecticut House of Representatives in 1998 at the age of twenty-five. He’s a Democrat career politician in a very blue state running against a woman who is regarded as trying to buy herself a Senate seat.

    To be sure, Connecticut is capable of electing moderate Republicans such as Chris Shays or Rob Simmons, both career public servants; Shays as a politician, Simmons as a military officer later turned politician. Simmons could probably have beaten Blumenthal and Shays would, I think, have been a tougher opponent than McMahon. Also, Connecticut has gotten bluer since the recession when a lot of Republican financial geniuses in Fairfield County lost our shirts and their jobs when they crashed the markets. A lot of those folks have left the state.

    Now if Quinnipiac showed Murphy behind than I’d start getting concerned but Rasmussen…we already know he’s had some other weird results this week. But don’t let any of this interfere with following Sam’s advice.

  • PaulW

    Sam, Thanks so much for this site and your ActBlue recommendations. I just went there to make a donation and noticed a likely typo- in the header paragraph you say, “I especially recommend Heidi Heitkamp and Chris Donnelly…”

    Did you mean Chris Murphy or Joe Donnelly? I assumed at first it was Chris Murphy, then I went looking for polling in the Indiana race, didn’t find much, but it was much closer than I would have thought. Having lived a few very long years in Indiana I just assumed the Teapublican would be running away with the race.

    • Sam Wang

      I meant Joe Donnelly. Thank you! You can give to any candidate if you like. I find the ActBlue interface to be rather trying.

  • BillSct

    One other thought. The Rasmussen Presidential poll in Connecticut from 8/21/2012 has Obama leading Romney 51% to 43%. Are Obama supporters in Connecticut really going to split their vote to elect a self-funded millionaire over a Democratic Congressman who comes from one of the most Republican Congressional districts in the state?

  • Pat

    Sam,
    Interesting histogram. I assume that this is simply calculated by assigning a rough win probability for each senate race. Could you disclose which win probabilities you have used for each seat?
    Even if these are rough estimates, if would be useful to know where the histogram comes from.
    Thanks!

    • Sam Wang

      Not exactly. Precise win probability can be calculated for polls shifted by a given amount diff from current conditions. Then allow diff to vary from -3.0% to +3.0% in steps of 0.1%, while at each step calculating an outcome histogram in terms of number of seats. Finally, average all the outcome histograms.

      The November win probabilities can also be averaged across [-3, +3]%. Expressed in terms of the Democratic candidate: CT Murphy 76%, IN Donnelly 36%, MA Warren 38%, MO McCaskill 94%, MT Tester 24%, ND Heitkamp 28%, VA Kaine 50%, WI Baldwin 1%. Plus 47 D/I, 45 R. Italics indicate an incumbent.

      I really must emphasize that these numbers are rough and subject to change. In particular I think Heitkamp’s chances are better than indicated.

  • Pat

    I was thinking that the [-3,+3]% variations might be useful for assigning the win probabilities for each seat, but that one you have those numbers (i.e. CT Murphy 76%, IN Donnelly 36%, etc.) it is easy to calculate 1 histogram.

    For example, in a situation where there are 2 races, with Dem probability win of 80% for the first seat and 30% for the 2nd seat, I would simply calculate the probabilities:
    - Dems win 2 seats = 0.8*0.3=24%
    - Dems win 1 seat = 0.8*0.7+0.2*0.3=62%
    - Dems win 0 seat = 0.2*0.7=14%
    It seems it would be easy to do the same thing with a larger number of seat, just taking into account all combinations giving a certain number of total seats.
    Why need to calculate a histogram for each poll variation [-3,+3] and average the histograms?

  • Pat

    I see.
    I always liked your idea that simpler is better. It would interesting to see how the trivial histogram obtained from that simple direct calculation differs from your more sophisticated approach. Any significant difference?

    • Sam Wang

      The problem with a direct calculation using probabilities without the diff variable is that for any given diff, the averaging effect of N binomial trials creates a narrow distribution of outcomes, of order sqrt(N) in width.

      Re simplification: the calculation should be made as simple as possible…but no simpler.

  • sanford braver

    your Act Blue contribution page is not operating properly. Stays on Please Wait interminably. Tried several times over several days.

  • Doug Schepers

    Thanks for the analysis and tips, Sam. I’ve donated to everyone you recommended on ActBlue, except Warren, whom I’ve already donated to.

    I also donated to McCaskill elsewhere, with due respect to your recommendation, since I’m originally from Missouri… and I’m not confident that Akin is going to lose, despite the trouble he’s cause the GOP in general.

Leave a Comment