For new readers (of which there are many!), I make a brief introduction to our highest-value product.
The Princeton Election Consortium’s main contribution is to give you a high-quality “thermometer” of exactly where the Presidential race is, based on state polls – and therefore the Electoral College, which is how the Presidency is determined. We do not use econometric indicators or national polls because these are only indirect measures (“What political science models really tell us,” August 28).
The track record of our state-polls-only approach in past years is quite good. In 2004, on Election Eve, the median was Bush 286 EV, Kerry 252 EV – the exact outcome. In 2008, we predicted Obama 364 EV, McCain 174 EV – just 1 EV off from the final outcome of Obama 365 EV. (For true nerds: in the latter case, the error was effectively 0.1 sigma, which suggests that we might be overestimating the 95% confidence interval.)
In addition to this, our time resolution is excellent. Other aggregators (Pollster, FiveThirtyEight) use methods that typically must be integrated over at least a week to reveal conditions. In contrast, we give a highly precise statistical snapshot of the race based on polls only.
This means we can test the recent assertion that the Romney campaign has the momentum. In 2004, Joe Lieberman’s Presidential campaign claimed to have “Joementum” to indicate they were going places. Where they went was a “three-way tie for third place.”
So…does the Meta-Analysis show real momentum? Or “Ro-mentum”? (update: bonus InTrade takedown at the end!)
Here’s a history of the Electoral Vote estimator. It achieves high resolution because we use poll margins in all states to calculate a win probability – and then we evaluate all 2.3 quadrillion possible outcomes, using a math trick to do it quickly. All the swings of the race are made visible. This year, the largest effects came after the Ryan VP nomination, the Democratic convention, and the first debate.
What is apparent is that the large plunge after Debate #1 came to a stop last week, right around the time of the VP debate. After that and Debate #2, Obama made some recovery. Now we are at a plateau, in which Obama is slightly – but decisively – ahead.
How far ahead? That can be answered using the Popular Vote Meta-Margin:
The Meta-Margin shows how much the race would have to swing to create an electoral near-tie. It is precise to within 0.5% or less.
Today, the race is quite close. However, note this. In terms of the Electoral College, President Obama has been ahead on every single day of the campaign, without exception.
I would then give the following verdict: Indeed the race is close, but it seems stable. For the last week, there is no evidence that conditions have been moving toward Romney. There is always the chance that I may have to eat my words – but that will require movement that is not yet apparent in polls.
The popular vote is a different story. I estimate an approximately 25% chance that the popular vote and the electoral vote will go in opposite directions – a “Bush v. Gore scenario”. I regard this as a serious risk, since it would engender prolonged bitterness.
In summary: Ro-mentum!
Update: Via Marginal Revolution: for about $1,250 it is evidently possible to manipulate InTrade. This morning’s swing toward Romney was caused by one trader’s manipulation. Ro-mentum!!