Friday, November 7, 2008
So, how did this site do on projections for the 2008 Presidential Election? Fairly well, actually, although there is always room for improvement.
Note: For purposes of this analysis, I am assuming that John McCain won Missouri, although most of the networks have not yet called it. He currently leads by 5,800 votes barring a recount.
Here is the top line rundown:
(1) Electoral Votes
Projected Actual Difference
356-182 364-174 Obama won 8 more than predicted
States Correctly Projected: 48 out of 50
States Missed: North Dakota predicted for Obama, went to McCain, Indiana predicted for McCain went to Obama
Electoral vote percentage correctly projected: 97.4% (524 out of 538)
(2) Popular Vote
Projected Actual Difference
Obama by 7.2% Obama by 6.5% McCain won 0.7% more than predicted
In other words, we predicted 99.3% of the popular vote correctly.
(3) State Averages
Average Miss in States Designated Battleground: 1.6%
Average Miss in States Not Designated Battleground: 2.8%
Total Average Miss: 2.5%
Note: These averages have been weighted by the number of electoral votes in a state
Clearly, I was a lot more accurate in battleground states, which shouldn't be a surprise since there is much more polling data available on close states and I spent much more time on the analysis of these states.
State Projection Details
a. Projected Non-Battlegrounds
There were 38 states that I designated as non-battlegrounds in the last projection. Of these 38 states, all 38 proved to be non-battleground with the projected winner in fact winning by more than 8%
b. Projected Battlegrounds
Of our twelve projected battlegrounds, I break my projection accuracy into a few categories:
i. Almost Exactly Right (Right Winner, within 2% of the vote) -- 7 states
North Carolina -- Projected Obama win by 0.4%, Actual Obama win by 0.3% -- 0.1% error
Virginia -- Projected Obama win by 5.3%, Actual Obama win by 5.6% -- 0.3% error
Missouri -- Projected McCain win by 0.6%, Actual McCain win by 0.2% -- 0.4% error
Ohio -- projected Obama win by 3.2%, Actual Obama win by 4.0% -- 0.8% error
Florida -- Projected Obama win by 1.6%, Actual Obama win by 2.6% -- 1.0% error
Georgia -- Projected McCain win by 4.3%, Actual McCain win by 5.5% -- 1.2% error
Montana -- Projected McCain win by 3.7%, Actual McCain win by 2.5% -- 1.2% error
ii. Right Winner, Wrong Margin (Right Winner, >2% error) -- 3 states
Colorado -- projected Obama win by 5.8%, Actual Obama win by 8.8% -- 3.0% error
Arizona -- projected McCain win by 4.2%, actual McCain win by 8.4% -- 4.2% error
Nevada -- projected Obama win by 5.7%, Actual Obama win by 12.6% -- 6.9% error
Arizona is fairly easy to explain -- McCain clearly got some late "home state" benefit that hadn't showed up in the polling.
Colorado and Nevada are more problematic for me to explain. Colorado was still relatively close to correct (within 3%), Nevada we were way off. Both states had massive early voting and late campaigning by Obama. This may have driven people to break late for Obama, but I'm not really sure. We just missed here, although we projected the right winner.
iii. Wrong Winner -- 2 states
Indiana -- projected McCain win by 1.5%, actual Obama win by 1.0% -- 2.5% error
North Dakota -- projected Obama win by 0.2%, actual McCain win by 8.8% -- 9.0% error
In the case of North Dakota, I'd been screaming for two weeks that we needed new polling. We didn't get it and clearly it might have shown that McCain had pulled back ahead, but we will never know. Suffice it to say, error will be a lot higher if the input data is old.
Indiana was always projected close, but 2.5% error is well above average for battleground states. I'd say this was the most surprising miss of the election year.
So, with all that, here's the scorecard:
National Vote Accuracy: 99.3%
Electoral Vote Accuracy (overall): 98.5%
Electoral Vote Accuracy (state by state): 97.4%
Battleground State-Level Accuracy: 98.4%
Overall State Level Accuracy: 97.5%
Percent of States Correctly Called: 96%
Percent of battlegrounds called within 1 point: 42%
within 2 points: 58%
within 3 points: 75%
Overall, I'm extremely proud of these results -- I got all of the major things about the election right both nationally and at a state-by state level.
Next week, I'll compare this site to other major projection sites so that you will get a feel for the comparative scorecard.
One Last Word on the Bradley Effect
Of the many things positive things that happened Tuesday night, one that may have been lost in the shuffle is the death of anyone talking about the Bradley effect again. There is not a single state where I could cite a Bradley effect based on our projections and the actual results. On election night, when the early returns came in from Georgia, I incorrectly states that it appeared there might be an effect in that state. I was wrong, as the final returns came very much in line with our pre-election projections.
It's a great day for America when race is not a key determining factor in our elections.
Quick Take on the First Week Post-Election
Obama has named Rahm Emanuel as his chief of staff. It is his choice and he should choose someone he is comfortable with, but this is clearly not a pick in the spirit of bi-partisan reconciliation. I think this is fine, provided he puts moderates and Republicans in some of his cabinet seats (Chuck Hagel, Colin Powell and Dick Lugar would all be obvious choices.) If he does not, it will become a problem. Despite a decisive victory (the highest percentage of the popular vote of any candidates since 1988 and the most total votes ever), Obama needs to be as bi-partisan a president as he was a candidate. The risk will be with control of both houses of congress to just ram through an agenda -- that temptation should be resisted.
Obama's first news conference was impressive. He was poised, funny and professional. Isn't it great to have a new President who can actually answer complex questions coherently? Having said that, not a lot specific was really said, so we'll have to wait and see. Hopefully the transition of power will go as smoothly as both Bush and Obama are promising.
Now that the election is lost, McCain aides are laying into Sarah Palin, accusing her of not only being a diva, but an idiot who doesn't know Africa is a continent, doesn't understand the 3 branches of government, doesn't know which countries are in North America, etc. This may all be true, but for once, I will come to Palin's defense. If this is all true, then it means Sarah Palin is ignorant, not stupid. McCain picked her, she didn't pick herself -- he was the one who made the judgment she was ready. She undoubtedly did the best she could for the campaign and if that wasn't good enough, it's on him for asking her to run -- it was unreasonable to expect her to say no. Palin I believe possess the intelligence to learn the issues and be a threat down the road. I don't care for her divisive style of politics, but I recognize political skill when I see it. McCain's campaign should be ashamed of itself for trying to lay the blame for the loss at her feet after picking her and doing so many things wrong: suspending the campaign, horrible performances in the first two debates, losing his message. George H.W. Bush won with Dan Quayle and surely Palin is no worse than Quayle.