Thursday, October 23, 2008

What to watch for in the polls--forget precision


With the national polls running from Obama+1 to Obama +14, what the hell are we to make of it?--If you are profoundly, and in my case, pathetically, addicted to the polls, you might be confused by all of this. Me too. But here are some tips on what you can expect in the next week-and-a-half.

1. In most--not all--presidential elections, the polls showed a tightening of the race in the last few weeks or days. People change their minds or chicken out or drop out. This has not happened yet. John McCain seemed to be making some gains last week, but that has been offset by a reversal this week. The polling aggregators, the guys who take all the polls and do things with the numbers, all seem to agree that Obama leads by about 6 points and that hasn't changed much in the last few weeks. The lead is pretty steady. If you look at the charts, particularly from politico.com, and turn the sensitivity down (use the tool button), you will see the true vector--which is fairly flat.

2. The lagtime between the national popularity polls and the state polls shrinks. Now there is about a two day delay: if one of the candidates moves in the national polls, it is reflected in state polls two days or so later. As we get closer to the election, that gap closes. The state polls are still reflecting McCain's uptick a few days ago. And of course, as I mentioned before, ignore the day-to-day changes. They are likely statistical noise.

3. In the days before the election, the polls will begin to converge toward one number. The outliers will move toward that number, meeting the other polls at whatever number it is going to be on election day. Add a grain of salt: Polling ends a day or so before the election and will miss very late shifts in the electorate. That's what happened in the Democratic primary in New Hampshire; the polls missed a late move for Hillary Clinton. That could happen in the national race but is not likely. If Obama really is 6 points ahead, that won't change between Sunday and Tuesday barring a terrorist attack or something.

4. All the statisticians who are projecting the winner (including those who claim they are not making predictions) are predicting an Obama blow-out, again by 6 or 7 points, fairly historic.

5. They could all be full of shit. The ghost of Thomas Dewey will not be exorcised for a long time.

1 comment:

Daro said...

Good points. Good enough to convince me to stop even bothering about polls!

Similar topic: Glenn Greenwald has a very apt piece on punditry of which I quote the first few paras. Check it out;

http://www.salon.com/opinion/greenwald/

Another myth fallen: Obama's "Jewish problem"

Salon's Mike Madden and Walter Shapiro undertook a very difficult task today: mining through the virtually endless possibilities in order to identify "The punditocracy's Seven Biggest Blunders of the 2008 election." As I wrote earlier this year in The National Interest: "the record of the American pundit class with regard to the 2008 presidential election can be summarized in one word: wrong." So I don't envy Madden and Shapiro's job of having to choose the top seven from that lengthy list.

The seven "blunders" they describe are certainly large and embarrassing, but there is a new one worth adding to the list: namely, Obama has a "Jewish problem," and Republicans are thus likely to attract a much higher percentage of the Jewish vote than even before. That theory was propounded all over the place, including by then-Weekly-Standard blogger and now McCain campaign spokesman Michael Goldfarb, who claimed earlier this year:

Obama has a Jewish problem, whether or not it's merely guilt by association is irrelevant. Politics is about perception, and the perception is that Obama's one step removed from the Nation of Islam.