CO: PPP says Gardner+3 (user search)
       |           

Welcome, Guest. Please login or register.
Did you miss your activation email?
March 29, 2024, 06:03:48 AM
News: Election Simulator 2.0 Released. Senate/Gubernatorial maps, proportional electoral votes, and more - Read more

  Talk Elections
  Election Archive
  Election Archive
  2014 Gubernatorial Election Polls
  2014 Senatorial Election Polls
  CO: PPP says Gardner+3 (search mode)
Pages: [1]
Author Topic: CO: PPP says Gardner+3  (Read 4590 times)
SPC
Chuck Hagel 08
Atlas Icon
*****
Posts: 10,004
Latvia


« on: October 21, 2014, 05:38:58 PM »
« edited: October 21, 2014, 05:48:01 PM by SPC »

Rasmussen aside, final PPP was Buck +1, final Marist was Buck +4, final CNN was Buck +1

Which is a +2 Buck margin, which is well-within the general margin of error for most polls.

The margin of error for one poll. Multiple polls being systematically a couple of points wrong, all in the direction of one candidate is something different.

I'm not saying the same thing is going to happen, because pollsters clean up their methods and improve over time, but it's foolish to say it DIDN'T happen in 2010.

Okay, let us suppose that we pooled the 2010 PPP, Marist, and CNN polls together. While we are at it, we should probably include the SUSA poll that showed that race tied (or are we excluding that on the basis that it does not fit the narrative of Colorado possessing some mysterious polling ether?). Those combined four polls would have a sample size of 3124, which corresponds to a margin of error of 2% (I would say 1.75% but muh decimuls!). So, even when taking into account that the polling average reduces the overall margin of error, the polling average predicting a Buck +1 victory (PPP gets weighted more since their sample size was bigger) when a Bennet +1 victory actually occurred is not some spooky phenomenon that can only be explained by systematic polling error.

Furthermore, an important thing to remember here is that we are not dealing with rocket science; political science is a social science and thus, as Mandenbrot has shown, may not be as applicable to the normal rules of a Gaussian distribution. If we were dealing with atoms placed on a knife's edge, then it would be fair to presuppose that they will split roughly evenly, but such a phenomenon does not necessarily apply to undecided voters. Indeed, when I examined polling averages vs final outcomes for competitive Senate elections between 2006 and 2012, I found that the histogram looked much closer to what a normal distribution would look like if one multiplied the expected standard deviation by three (in other words, the tails were fat). Colorado was not even particularly unusual (Bennet would have had 3:1 odds using this method, rather than the ~40:1 one would expect from the expected standard deviation). Indeed, the only unusual results were West Virginia 2010 (to be expected when PPP has only Rasmussen checking their work) and Nevada 2010 (a more plausible case for systematic polling bias if I have even seen one)
Logged
SPC
Chuck Hagel 08
Atlas Icon
*****
Posts: 10,004
Latvia


« Reply #1 on: October 21, 2014, 06:25:59 PM »

If it quacks like Gardner +3 generally across the board, it's Gardner's race to lose right now. Tossup/Slight Lean R.

Quote from: Restricted
You must be logged in to read this quote.
Logged
Pages: [1]  
Jump to:  


Login with username, password and session length

Terms of Service - DMCA Agent and Policy - Privacy Policy and Cookies

Powered by SMF 1.1.21 | SMF © 2015, Simple Machines

Page created in 0.02 seconds with 14 queries.