Skip to main content

View Diary: Senate Snapshot, October 1st: The Ultimate Election Forecast (78 comments)

Comment Preferences

  •  You have just identified a real weakness (4+ / 0-)
    Recommended by:
    Odysseus, joanneleon, Scientician, elwior
    in these models: it only goes back 12 years.

    There is no detailed data on an election taking place during a significant recession.

    These probabilities all have enourmous selection bias.

    But as Chris said, getting the data is hell.  It took me months to put together the data on the relationship between Iowa and New Hampshire.

    The bitter truth of deep inequality has been disguised by an era of cheap imported goods and the anyone-can-make-it celebrity myth - Polly Toynbee

    by fladem on Fri Oct 01, 2010 at 05:15:17 PM PDT

    [ Parent ]

    •  Sample size (2+ / 0-)
      Recommended by:
      Odysseus, elwior

      Yes, I read Nate's article when it came out and the sample size for the 6-9 pt range (a sample size of 7) just seems absurdly low.  The fact that we're looking at a 12-yr period, which doesn't allow for a lot of variation in business cycles, and other things, adds to the absurdity.  The fact that Nate now has a very large audience of likely voters (NY Times readers) makes what he is doing rather irresponsible too, IMHO, especially when he keeps using one race, PA Senate, as his example.  It seems like he desperately wants to be right on this one, and has not a care in the world about what effect he might be having on the race at a time when voter turnout is going to be especially important.

      I'm not a statistics expert, but I have a little background, having taken two courses in it, and having had a lot of math courses in college (engineering degree.)

      Also, the margins of error and the lack of consistency among the polls are a problem.

      Now, I'm very biased on the Sestak race, as I've been a supporter of his for years.  So I'm also thoroughly pissed off that Nate keeps using the Sestak/Toomey race as an example, which in itself, since he is now published in the NY Times, could skew the results of the race or at least depress voters who favor Sestak, IMHO.  I'm pretty sure the NY Times is widely read in Pennsylvania, particularly in the region where Sestak is most likely to draw votes.  

      From my perspective, Nate is like a short-seller right now.  And I'm beginning to question the value of these predictive models.  It's not like the weather, which isn't going to change its behavior after it reads Nate's (and/or Chris') predictions.  It's more like the stock market.  I'm inclined to tell both Nate and Chris to knock it off, and let the voters decide based on the merits of the candidates, and knock it off with the largely ego-based modeling on this election, and quit talking about their "Ultimate Election Forecasts" and "Uncanny Accuracy of Polling Averages" and "The World's Greatest Forecast Election Forecast, In the World."  (For anyone who might be wondering, no, I didn't make those up, those are actually phrases used by Nate and Chris.)

      Uncanny, and World's Greatest, et al, are not claims that should be made by guys with 12 years worth of data, and Nate, in particular, should not be predicting with such Grand Poobahness with a sample size of freaking 7.

      Disclosure: I'm working as an unpaid citizen journalist covering the Sestak campaign/ PA Sen. race for Huffington Post's "Eyes and Ears 2010" project

      by joanneleon on Fri Oct 01, 2010 at 07:26:45 PM PDT

      [ Parent ]

      •  I agree with you joanne, and I hope you do (0+ / 0-)

        tell these guys to knock it off, particularly Nate's use of the Pa. race.

        "We the People of the United States...." -U.S. Constitution

        by elwior on Fri Oct 01, 2010 at 08:58:15 PM PDT

        [ Parent ]

      •  In finance (1+ / 0-)
        Recommended by:
        joanneleon

        the quants (I am one) built models of stock price movements.  Nassim Taleb criticised them for the same reason you and I have concerns about the numbers Nate generates.  Taleb's nickname is the "Black Swan".  The name comes from his observation that quantitative models systematically underestimate the chance of a "black swan event".

        The theory is based on the knowledge that there are more black swans than genetic theory would suggest should exist.  Taleb's theory was designed to warn hedge funds that their models would fail because of:

        1. the disproportionate role of high-impact, hard to predict, and rare events that are beyond the realm of normal expectations in history, science, finance and technology, 2) the non-computability of the probability of the consequential rare events using scientific methods (owing to their very nature of small probabilities) and 3) the psychological biases that make people individually and collectively blind to uncertainty and unaware of the massive role of the rare event in historical affairs

        Nate's model is far less built on solid statistical theory than the quantitative models that came unraveled during the financial crisis were.

        In particular, if you look at national polling, there is about a 33% of a 5 point national swing in the last 5 days of a national election since 1968.  But Nate's model ignores this possibility.

        In any given election Nate's model will probably work.

        Until, like the proverbial black swans, it doesn't.

        The bitter truth of deep inequality has been disguised by an era of cheap imported goods and the anyone-can-make-it celebrity myth - Polly Toynbee

        by fladem on Sat Oct 02, 2010 at 03:00:52 AM PDT

        [ Parent ]

    •  Uh, 2008? (0+ / 0-)

      That was right in the middle of a significant recession.

      •  What was unemployment (0+ / 0-)

        in 2008?  The last time unemployment was even close to where it is now was 1982.

        The bitter truth of deep inequality has been disguised by an era of cheap imported goods and the anyone-can-make-it celebrity myth - Polly Toynbee

        by fladem on Sat Oct 02, 2010 at 02:49:27 AM PDT

        [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site