Monday, November 03, 2008

The US Elections: the Polls

In recent weeks virtually every single poll has put Barack Obama in the lead (there was a poll that said otherwise but that turned out to be Zogby playing with himself the numbers). Yet somehow I can't help feeling that there's something seriously wrong about the opinion poll results. I'm not doing this out of a deluded hope that John McCain will win the election - to prepare myself for his loss I've engaged in certain exercises (fetal crouches and thumb suckings).

The source of my discontent is that in 1992, Bill Clinton rode as high as a 20% poll lead over the Elder Bush before falling back to a respectable 6%. He did this by running on a platform of "It's the economy, stupid" in a recession. Barack Obama has been the beneficiary of a meltdown on Wall Street. Things are so bad that if a lynch mob were given the choice between Barack and a Wall Street Banker, they would hang the Banker without hesitation. Yet Barack has never gotten anything like Clinton's lead in the opinion polls. Coupled with other factoids such as the Vice-Presidential debate between Sarah Palin and Joe Biden pulled 75 million viewers whereas Barack Obama could only pull a third of that for his informercial and my feeling is that the pollsters are getting a distorted impression of the mood due to the recent Wall Street calamity.

I'm not going to engage in an detailed analysis of the polls as it will quickly degenerate into something less informative than a gaggle of Haruspices arguing about the state of a liver. Nor can I plausibly argue that all the polling companies are shilling for Barack because even the Republican polls is saying pretty much the same thing.

If I were to hazard a guess, I would say that the polling is capturing Barack's support accurately but underreporting McCain's support in the countryside. This is because polling companies in order to cut costs produce a representative sample by conducting polls in the easy-to-reach big cities and adjust the results using the smaller sample of the harder-to-reach areas such as the countryside to arrive at a reasonable approximation of the sampling population as a whole. This corner cutting is legitimate and it works and the polling companies would go out of business if they didn't.

However this corner cutting can produce misleading results from time to time. For example Tom Bradley failed to become Governor of California not because moderate democrats were lying to the pollsters about their willingness to vote for a black but because there was an anti-gun proposition on the ballot. The NRA sounded the duelling banjos to defeat it and the unanticipated extra voters voted against Bradley for good measure because he had supported the proposition.

Does this effect significantly distort the poll results? I don't know. If it did, will it be large enough for McCain to win the election? I really don't know. All I can do is wait...