Monday, November 03, 2008

Free at last! Free at last! Almost--UPDATED


[UPDATE: The final polls have merged at something of a consensus. Obama by 6]

If anyone asks, I've been a shoe salesman all this time--
Either Barack Obama will win the presidency tomorrow or American journalism and public opinion research (political division) will awake Wednesday with all their credibility destroyed. If that happens, both deserve it.

If John McCain actually wins, every political pollster would have been wrong, and every political reporter who based his or her stories on those polls (and they shouldn't for other reasons we can go into later) would have been wrong and and there is no reason to believe those idiots again.

I don't think that is going to happen. With voting one day away (except for the one-third of American voters who have already voted) here is how things stand--if the polls are right:

Every poll, tracking and daily, has Obama ahead by margins ranging from 4 to 13 points. How could that be? All use different techniques to draw their samples and do the actual polling (humans on the telephone vs. computers on the telephone), but more importantly, all use different methods of deciding who is going to vote and who has already voted. Some call cell phones, some don't. So, some of the polls will be wrong, we know that. Polling in these matters usually come within 2 points of reality. If your poll predicts a guy will win by 4 and he wins by 2.6, most pollsters would think it was a normal year. If you come in with 13 and the margin is 4, you have screwed up. As usual, the tracking polls and the daily polls have converged and agree, and some of the polls coming in Monday morning seem to be heading toward Obama +6, which is on the edge of a landslide. These include a few surveys that tend to have a Republican tilt.

I should point out that most experts believe that it doesn't matter which polls are correct: if Obama has anything between a 3 and a 13 point lead, he wins. That plurality will bring the state votes along with it.

The internet sites that average and aggregate the polling, often using statistical manipulations, all seem to agree that Obama will win by 5-6 points and with better than 300 electoral college votes.

(One interesting note, Pollster.com has a story on how Obama gets around a 4 point jump if pollsters get to cells phones).

At this stage, the national poll numbers and the state-by-state poll numbers are in tune. If one goes up, the other does as well. There is no sign whatsover that McCain has tightened either the national or state polling in any substantial manner, certainly not enough to overcome a 5-6 deficit in one day. That doesn't happen. This campaign was essentially over after the second debate and the collapse of the economy. The numbers haven't changed materially since.

We all know how Harry Truman beat Thomas Dewey in 1948 despite predictions from pollsters he would not. The fault then (and in New Hampshire this year) was that they stopped polling too early and missed a late surge. That is not happening this time; pollsters are out in the field as you read this and I'll update it later today when those polls come in. And it is hard to imagine a late surge that would overcome a 5 point deficit. These are not predictions, but with 24 hours or so left, they might as well be.

Some folks, often in economics departments, will stick their necks out and make predictions and most of those sites think Obama will win by 6-7 points, with electoral college votes around 350.

What to watch for? If Obama wins Virginia and/or North Carolina by anything resembling a healthy margin (and those polls close early eastern time), the cake is baked. Go to sleep. If he loses those, bring out your sleeping bags and switch from decaf--it will be, in the words of Bette Davis, "a bumpy ride."

Unless all these polls, everything we know about public opinion surveys, all the bright pundits and reporters are flat wrong. Then, remember, I was always a shoe salesman.

j

1 comment:

Daro said...

Diebold hacking, voter-caging, ballot dumping... I hope they all cancel each other out but I doubt it. Here's one prediction you don't have to poll - if McCain "wins" there'll be revolution in the streets.