Saturday, January 12, 2008

The New Hampshire Primary and the Exit Polls

I have not had time to analyze the New Hampshire primary results so what follows is not based on a thorough investigation, but I cannot resist posting the following article with a brief critical commentary. As everyone knows by now, the polls on the eve of the primary were showing Clinton trailing Obama by double-digits. The exit polls, according to the article below, also showed during the evening, before being "adjusted" to correspond to the official tallies, that Obama was ahead by a big margin - I haven't, as I mentioned, investigated this myself to confirm this, but allow me to assume for now that what is said below is true.

As anyone knows who has been reading my blog or hearing me speak, I am not a fan of Obama's, nor do I think that the elections are anything but a dangerous, terrible and cruel deception for the people. The only way to change things is through the people acting independently and altering the political atmosphere. People for the last several election cycles have put a lot of time, energy and hope into the process in the desperate belief that they can see some genuine change through voting and electoral politics. The fact that elections have demonstrably been stolen, and perhaps this primary too, should be looked at by people and the proper conclusions drawn.

It does not make sense that a candidate can have a double-digit lead on election eve and then lose by several points in the official tallies. If the pre-election polls are close enough to the election and, moreover, the exit polls are consistent with those pre-election eve polls, then it isn't statistically reasonable for the official tallies to come out as they did. Commentators have been coming up a with a lot of speculation to try to explain why.

Perhaps there is an explanation that will hold water, but now that I hear about the exit polls, those explanations must explain one ineluctable fact: exit polls are the gold standard of voting. They are the indicator of fraud. People don't forget who they just voted for as they just got through voting. We don't have to estimate who will turn out and in what numbers because the exit pollsters are sampling the people who actually came out to vote. The idea that Clinton voters lied to the exit pollsters and said they'd voted for Obama because they were ashamed to admit that they'd voted for the woman in the field doesn't pass the smell test.

Finkelstein evidently does not understand the nature of exit polls. Why would voters in overwhelmingly white Iowa vote for Obama but very white New Hampshire have a hard time voting for the black candidate after already telling pre-election pollsters and exit pollsters that they were voting for Obama? Do you think that Hillary's much touted moment of "vulnerability" in a cafe is enough to swing the vote by double digits at the last instant? Granted, a number of people said they were unsure until the last minute, but these numbers and results on their face are highly suspect and inconsistent with polling and statistical experience.

Case closed: Why Hillary won

by Daniel Finkelstein

More on why Hillary won when the polls said that she wouldn't.

My two leading theories were the Bradley Effect (people say they are going to support a black candidate then don't) and the Spiral of Silence (people are embarrassed to tell a pollster they really supported Hillary not Obama).

Both of these had arguments in their favour, but both had this problem - the exit polls. Surely if either of these effects was important they would have made the exit polls wrong. And yet the exit polls seemed, looking at them the morning after the night before, to be right.

Now I've spoken to one of the best polling gurus in the business, Andrew Cooper of Populus, and I think I understand.

The correctness of the exit polls is an illusion.

When I first saw the exit polls (at 1AM Wednesday morning UK time) they showed a 39-34 per cent advantage for Obama. When I woke up at 7AM they seemed to be totally in line with the result.

The reason? The exit polls are reweighted as the night goes on to incorporate the results as they are counted. And the original polls disappear from the website.

This is very important indeed in gaining an understanding of the Hillary victory.

It means that any explanation of her victory must explain that voters told exit pollsters after they had voted that they were not for Hillary even though they had just voted for her. This means that the polls were wrong because people weren't telling the truth to pollsters and not because of a last minute change of heart.

So you can dismiss, for instance, the crying as an explanation because even if it didn't turn up in last minute opinion polls it surely would have done in an exit poll.

We really are just left with the Bradley Effect and the Spiral of Silence.

So which is it? The Spiral of Silence. How can I be so sure? Because of this graph from Matt Yglesias's site.

As you can see Obama's vote came in on the money. Hillary's didn't, not by miles. People voted for her who didn't tell pollsters that they would. And they kept it to themselves even after they had done so.

The wikipedia entry is particularly good at explaining how something like Obama's incredible media coverage post-Iowa might make people unwilling to admit they were actually for Hillary.

Case solved.

No comments: