Poll Aggregators Roundup – Saturday, 11/3/12

Poll aggregators agree: it's Obama's election to lose. To Romney supporters they say: beware of pundits' blabber about a toss-up. To Obama supporters they caution: don't begin to celebrate just yet.

The eight poll aggregators on the table below have different methodologies, from statistical models all of their own (like Nate Silver's FiveThirtyEight) to simple poll averages (like RealClearPolitics). They all include hundreds or thousands of polls and do frequent updates, some of them so fast that when I finish this post their numbers might already have changed. Some are run by liberal statisticians; there's one libertarian who leans Democrat (Electoral Vote) and one self-professed Republican (Election Projection). They all try to crunch the numbers in an honest way, because their reputation depends on getting their predictions right and close to the results.

And they all have the same prediction: Obama is going to win the electoral college and is also ahead on the popular vote. Their estimate of electoral votes varies; the highest is 323 (Princeton Election Consortium) and the lowest is 281 (some of them –like Pollster– do not include tossup states in their estimate). And except for the Republican leaning Election Projection, they predict that Obama will also win the national popular vote.

Blue indicates Obama is ahead; red, that Romney is ahead. I included only the states where, according to FiveThirtyEight, either candidate's chance of winning is less than 90%. The consensus is that the states on the list are closest ones. 

Take a look at the numbers. You will see why pundits are wrong when they say the election is a toss-up. But Obama hasn't locked the election yet. For a comment on that, read after the table.

	538	Pollster	Poll Tracker	Electoral Vote (1)	Princeton Election Consortium	Election Projection	Politics by the numbers (2)	Real Clear Politics
Popular Vote	50.6 – 48.4	47.4 – 47.2	47.9 – 47.5	NA	+2.98	49 - 49.4	51.44 - 48.56	47.4 - 47.3
Electoral College (270 to win)	305.3	281	303	281	323	290	281	290
Colorado	+1.5	+1.5	+2.1	Tie	+1.0	+0.91	NA	+1.0
Florida	+0.4	+0.3	+1.2	Tie	+1.0	+0.9	NA	+1.4
Iowa	+3.1	+3.4	+3.1	+3.0	+3.0	+2.1	NA	+2.0
Nevada	+3.9	+3.7	+4.8	+3.0	+4.0	+3.0	NA	+2.7
New Hampshire	+3.3	+2.7	+4.3	+3.0	+2.0	+1.4	NA	+1.8
North Carolina	+2.5	+2.3	+1.5	+2.0	+0.5	+4.0	NA	+3.8
Ohio	+2.9	+3.0	+3.0	+3.0	+3.0	+0.4	NA	+2.9
Virginia	+1.2	+0.9	+2.1	Tie	+2.0	Tie	NA	+0.3

(1) Site does not include popular vote prediction
(2) Site does not include state estimates

Democrats hope for a non rainy day

Some sites calculate Obama's chance of winning.  FiveThirtyEight gives him 83.7%. Princeton Election Consortium gives him 98%. Politics by the numbers, 86.64. What does this mean? That Obama can still lose –but he has less ways to lose than Romney.

Let's assume that his odds are 80% (in 4 out of 5 cases, he wins the election). Sites like 538 run millions of simulations of voting scenarios on election day and make their prediction. 80% chance of Obama winning mean that 20% of the simulations indicate voting scenarios in which Romney wins.

For those who object, "But an election is a one-off event," think of the weather prediction. I live in Los Angeles. Two weeks ago, the weather forecast predicted a 20% chance of rain. The prediction implied an 80% chance of no rain, but it rained that night. How good is the prediction? If it rains 1 out of 5 times when the forecast predicts 20% chance of rain, then the prediction is quite good.

Romney can still win, but the voting scenarios for him to do so are greatly diminished. What Democrats can hope for is that it doesn't rain on their parade.