Skip to main content

View Diary: Rating the 2010 election forecasters (54 comments)

Comment Preferences

  •  I chose it over a year ago (2+ / 0-)
    Recommended by:
    Buckeye Hamburger, Scott in WI

    When I began my research on this stuff. published that I had done so before I began posting here at Daily Kos. It's not a recent decision.

    My basic reasons:

    1--I've only done research on election forecasting for campaigns projected within 18.5% or less. I'm not posting a forecast that's not based on research. I cut my research off at 18.5% because there is only so much time in life, and resarch on even the 18.5% campaigns takes several weeks a year.

    2--I'm only interested in forecasting competitive elections. I honestly don't care if I have a methodology that more accurately forecasts blowouts. I just don't. My goal is to forecast competitive ones.

    3--Going further out creates a problem of time in terms of putting the forecasts together, spent on elections where the outcome is a foregone conclusion. Feels utterly pointless.

    4--It also creates a problem of presentation. The more numbers you put in your projection, the harder it is for the audience to easily it. Less is more in election forecast presentation.

    And, perhaps most importantly...

    5--Elections that are further out are harder to poll, usually because people are paying less attention. They also have fewer polls. This means that poll-based election forecasts will all do pretty terribly in blowout elections. It can create white noise that makes a forecast look inaccurate, whereas those forecasts are actually very accurate for the campaigns people are actually following.

    ***

    Even with all that said, 18.5% still seems pretty arbitrary. I'd be happy to use a different measure in the future, one that people can agree on ahead of time. However, I am simply not going to forecast every election. Even if I did, I refuse to have my ability in this area judged by my ability to determine if a lightly contested Senate incumbent will win by either 40% or 35%.

    I want the accurate forecast for the competitive campaigns, and I want to present it in the most transparent, easily understandable format possible. That's my goal, and going further out than 18.5% would hurt it. If anything, I should restrict further, perhaps to 15.0%.

    •  Ahh (0+ / 0-)

      Thanks for explaining!

    •  OK, but what led to the figure? (0+ / 0-)

      Thanks for the response. The need for a cutoff, your reasons 1-5, is clear, and it's important to note that it's not a "fudge factor" chosen for the purposes of this analysis if it was picked in advance. But what made you decide on that particular figure? There must have been a rationale for 18.5% rather than, say, 17.5% or 19.5%.

      Absolute power corrupts absolutely.

      by Buckeye Hamburger on Sun Dec 19, 2010 at 11:12:40 PM PST

      [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site