11/05/2012

Who's going to win?

THE U.S. PRESIDENTIAL ELECTION on Tuesday November 6th will be won with a convincing electoral college majority. At least 70 votes.

That's my conclusion of poring over the polls, listening to what people are enthusiastic/despondent about, looking at where the money and candidates are being used.

The problem is, I'm not sure WHO is going to win by at least 70 electoral college votes.

The reason for this is that the polling data looks open to very different interpretation, depending on how one sees the last presidential election (which was won in convincing fashion by the then Senator for Illinois Barack Obama) in 2008.



Was the 365 to 173 thrashing the result of deep and permanent demographic changes in U.S. politics which favours the Democratic Party? If so, Nate Silver's 538.com model would tend to be right.

This predicts an improvement for Mitt Romney, the Republican candidate, of two or three states and the 2nd congressional district of Nebraska: Indiana, North Carolina and possibly Florida. With the changes to the number of electoral votes per state from the 2000 census, we would get: Obama (303), Romney (235).

The 538.com model is based on the notion that the opinion polls have corrected sampled the U.S. electorate and that Nate Silver has corrected rated their reliability. He did a good job of this in 2008 and a lousy one in 2010 (but for congressional elections, so he might simply be better at presidential forecasting). He also got the 2010 U.K. election wrong, even when he had the BBC exit poll data...

Intrade is almost matching 538.com, but as the Wisdom of Crowds warns, a problem is "groupthink." What if 538.com is influenced by Intrade and Intrade's punters are influenced by 538.com? The number of trades is in the hundreds or a few thousands, which means that if there is a herd effect, both Intrade and 538.com could be wrong.

Simply: if the opinion polls are adjusting their demographic samples to make them look like the turnout of 2008, then 538.com is wrong, and if they are wrong, it is likely that Intrade punters are misled.

Was 2008 a one-off?

If so, the Gallup electorate polls are correct and we shall see a landslide to the former Governor of Massachusetts, Mr Romney. Bob Krumm outlines the case here. The difference is that Gallup has estimated a big drop in enthusiasm among demographic groups that elected President Obama. The problem is that there isn't a lot of precedent: 1996 and 1980 were the last two presidential elections where Republicans faced a Democratic incumbent. On both occasions, the "normal" opinion polls greatly underestimated the Republican vote. If they're doing it again, we could see something like this: Romney (377), Obama (161).

Known unknowns?

I still have trouble believing that the Democrats won Indiana without a significant amount of cheating in 2008 and the shambolic organisation of elections at local level makes it awfully tempting, I'm sure, for abuses to occur. Whether it's "losing" some postal votes, "accidentally" placing provisional ballots in the "checked and OK to count" column, the cemetery vote or intimidation, I have little doubt that the losers this time will be sure the other side cheated.

So I don't know if Minnesota will be won by Romney, or if Obama will pick up Georgia.

But I will say that if Romney wins Pennsylvania (polls close in that state at 8pm Eastern Standard Time [1am GMT]) he's going to be President. If Obama wins Florida (polls close at the same time, though the media will try to claim the result an hour earlier, but this should be ignored as the late voters will be in Republican districts), he is re-elected comfortably.

I think it will be something like this: Romney (315), Obama (223).

No comments: