Now that we've had two eerily accurate presidential forecasts from Nate Silver, I wanted to document just how accurate he was. We all know that he was 50 for 50 states and though the votes aren't all in yet, he also came really close in the popular vote. Below the fold, I have a chart that shows his final Nowcast compared to the actual results and the differences for 13 of the most competitive states...my thinking being that these are the states which had representative samples of polls...though I could be wrong about that. By the way, this is my first diary so be gentle.
Obama Rmoney Other
Colorado 11/6 Nowcast 50.8% 48.3% 0.9%
Actual 51.2% 46.5% 2.3%
Difference 0.4% -1.8% 1.4%
Obama Rmoney Other
Florida 11/6 Nowcast 49.8% 49.8% 0.4%
Actual 49.9% 49.3% 0.9%
Difference 0.1% -0.5% 0.5%
Obama Rmoney Other
Iowa 11/6 Nowcast 51.1% 47.9% 1.0%
Actual 52.1% 46.5% 1.4%
Difference 1.0% -1.4% 0.4%
Obama Rmoney Other
Michigan 11/6 Nowcast 53.0% 45.9% 1.1%
Actual 53.8% 45.3% 0.9%
Difference 0.8% -0.6% -0.2%
Obama Rmoney Other
Minnesota 11/6 Nowcast 53.7% 45.1% 1.2%
Actual 52.8% 45.2% 2.0%
Difference -0.9% 0.1% 0.8%
Obama Rmoney Other
N Carolina 11/6 Nowcast 48.9% 50.6% 0.5%
Actual 48.4% 50.6% 1.0%
Difference -0.5% 0% 0.5%
Obama Rmoney Other
New Hampshire 11/6 Nowcast 51.4% 47.9% 0.7%
Actual 52.2% 46.4% 1.3%
Difference 0.8% -1.5% 0.5%
Obama Rmoney Other
Nevada 11/6 Nowcast 51.8% 47.3% 0.9%
Actual 52.3% 45.7% 2.0%
Difference 0.5% -1.6% 1.1%
Obama Rmoney Other
New Mexico 11/6 Nowcast 54.1% 44.7% 1.2%
Actual 52.9% 43.0% 4.2%
Difference -1.2% -1.7% 3.0%
Obama Rmoney Other
Ohio 11/6 Nowcast 51.3% 47.7% 1.0%
Actual 50.1% 48.2% 1.6%
Difference -1.2% 0.5% 0.6%
Obama Rmoney Other
Pennsylvania 11/6 Nowcast 52.5% 46.6% 0.9%
Actual 52.0% 46.8% 1.2%
Difference -0.5% 0.2% 0.3%
Obama Rmoney Other
Virginia 11/6 Nowcast 50.7% 48.7% 0.6%
Actual 50.8% 47.8% 1.4%
Difference 0.1% -0.9% 0.8%
Obama Rmoney Other
Wisconsin 11/6 Nowcast 52.4% 46.9% 0.7%
Actual 52.8% 46.1% 1.1%
Difference 0.4% -0.8% 0.4%
Some Observations:
> Within 1 percentage point on Obama forecasts in 10 of these 13 states
> Over estimated Obama in 5 states
> Under estimated Obama in 8 states
> Over estimated Romney in 9 states
> Under estimated Romney in 3 states
> The "other" category was consistently underestimated (except MI). ('Other' on the Nowcast was calculated by me based on the difference between the two forecast total and 100)
> New Mexico's 'other' was very under-represented...3.0% difference. I'm guessing that had to do with the quality of the polls as percentages were drawn from both candidates...but again...I could be wrong.
> The least accurate forecast was Ohio, where his forecast fell short for Obama by 1.2%.
I will leave it to you to interpret why we might see these variations...but as for the Ohio variation...it leaves me to want to put on my tin foil hat and look more closely at those Tagg Romney-owned voting machines...and the voter suppression efforts there.
In any event, I would think that it's safe to say that if a future Nate Silver forecast has any given state by 2.0% or more (even less actually)...you can be pretty damn sure it's in the bank.
Finally, let me just share that after the 2008 election, I sent Nate Silver a fan letter...really! My son is in the advanced math program at school and I wanted him to know that I appreciate people who call attention to the benefits of being good at math/statistics (or in his case, great at it). This year, when it was clear that the media was determined to carry on with this BS mitt-mentum meme and the false horse race narrative, I instituted a self-blackout and only allowed myself access to Daily Kos and fivethirtyeight.com. Thank you all for getting me through, and thanks again Nate!