About Me

My photo

Did Socrates say "Know Thyself", or was he misunderstood, as all are. Show Thyself is all we can do. The knowing is unknowable.  

I am filled with joy.  It can't be helped.  

Became a Farmer, Builder, Musician, Tank Commander, Librarian, Lawyer and Minister. I have failed at many things. And now retired.  Filled, just filled, with Joy. 

Thursday, November 06, 2008

The Pollsters did a very good job of predicting the outcome...

My favorite Poll-master - at Electoral-vote.com, has analyzed the relative success of polsters against actual voting results. Looks like ARG, Insider Advantage, and
Quinnipiac U. were on the money at 100% accuracy/correspondence to reality.
Mason-Dixon and Rasmussen were relatively inferior, missing the close calls, although even they came in favoring Obama in overall outcome.

"In Colorado, for example, the scaled poll predicted an Obama vote of 52.7% and a McCain vote of 45.9%. The candidates got 52.5% and 45.9%, respectively. Thus the polls were remarkably close to the actual vote and there is not a hint of Obama polling well but doing badly on election day.

Now look at the eight other swing states. In Florida the pollsters were right on the money. In Indiana, Missouri, North Carolina, Ohio, and Virginia they were off by a maximum of 1.2% for Obama. In the other two states, Nevada and New Mexico, the polls for Obama were off by 4.1%, which is within the margin of error. In both cases Obama did better than predicted, so clearly there is no Bradley effect here. If nothing else, any future discussions of the "Bradley effect" can be quickly ended with the statement: "It didn't exist in 2008."
Comparison of the Pollsters

There are various ways to compare the pollsters. Professional statisticians look to see if the actual results were within the bracketed range predicted by the pollster. For example, if a pollster predicted Obama 51% and McCain 49% with a 3% margin of error and the final result was McCain 51% and Obama 49%, the pollster would claim to have gotten it right because Obama fell within the predicted range of 48% to 54% and McCain fell into the predicted range of 46% to 52%. Laymen would probably say the pollster got it wrong--because they don't understand what the prediction really is. To keep it simple, let us use the second approach and see which pollsters predicted which states right. Here are the data.
Pollster CO FL IN MO NV NM NC OH VA Rt/Tot Pct
ARG Yes Yes Tie Tie - - Yes Yes Yes 5/5 100%
Insider Advantage Yes Yes - Yes Tie - Tie Yes - 4/4 100%
Quinnipiac U. Yes Yes - - - - - Yes - 3/3 100%
Opinion Research Yes Yes No Yes Yes - Yes Yes Yes 7/8 88%
SurveyUSA - Yes Tie - - Yes No Yes Yes 4/5 80%
Research 2000 - Yes Yes No Yes - Yes - - 4/5 80%
Mason-Dixon Yes Yes - - Yes - No No Yes 4/6 67%
Rasmussen Yes No No - Yes Yes No - Yes 4/7 43%

In some cases the pollster predicted a tie, that is, said it was too close to call. Let us remove those from the sample. In that case the scores for these pollsters are given in the last two columns. If you want to play with the data yourself, here they are .csv format."