Big polling experts in action: Credit where due! Among cable stars, Lawrence O'Donnell took the cake when it came to crazily bogus predictions.
In this earlier post, we recalled the ridiculous way he assured the world that Candidate Clinton was running away with Florida. A few nights later—on Friday, November 4—he opened his show with the number-crunching prediction savant, Princeton professor Sam Wang.
Professor Wang said the Senate could go either way. No so with the White House!
O'DONNELL (11/4/16): Your consortium has put out a projection of a 99 percent chance of [Clinton] winning, is that correct?Professor Wang had even worked in a joke, so sure was he of the outcome. At any rate, he was reporting a probability of more than 99 percent.
WANG: Yes. We should be careful about probabilities this high because once the probability gets that high, it's not so clear that it matters whether it's 91 percent, 95 percent or 99 percent. But yes, that's right. The probability is high enough that my side is reporting it as currently above 99 percent.
O'DONNELL: Just— Is it possible to get a hundred percent probability on one of these things?
WANG: It is never possible to get exactly a hundred percent probability. If my site says 101 percent, then you should worry.
O'DONNELL: OK, so she got the best score she can get. That's the highest score on the SATs here, as it were, of this thing.
Next, O'Donnell asked Wang to detail his methodology. The skillful professor explained that he only uses the most accurate polls. Along the way, O'Donnell displayed the mawkish behavior TV performers tend to direct at mental giants from Princeton:
WANG: The Princeton Election Consortium and election.princeton.edu uses entirely state polls. State polls are the most accurate of the polls that are available to polling analysts in the U.S.If a hurricane hit the East Coast, Wang's prediction might be affected, he said. Four nights later, in beautiful weather, Dewey beat Truman again!
And so what we do is we take each state, take the available polls for the last two weeks, actually a little bit shorter period now, calculate a median, so arrange them in order and calculate the median of those polls, figure out the possibility from that of who's ahead in either state, in each of those states, and then take that and compound that across all 56 races.
There are 56 races that determine the Electoral College and that's over 72 quadrillion possibilities and so I have a little script that I've written on my computer that does that. It figures out what the median is, or the midpoint of all those possibilities. And the median currently is at 312 electoral votes, and so the general idea is that is our sharpest snapshot of where the race appears to be, based on today's state polling.
O'DONNELL: That is the first time I've heard "72 quadrillion" used in a sentence, but that's because I'm not in your class at Princeton. And so between now and poll closing Tuesday night, what could change this?
WANG: It would be really unusual for anything to change. I would say that opinion doesn't move very quickly. In the last two decades of polarized politics, voters have gotten more and more entrenched and the only thing that can change is small variations in turnout.
So for instance, if there's a really bad storm on the East Coast then that could depress turnout maybe one or two points, same on the West Coast. So I think given the way things stand right now, it would take a very large weather event to change things and, you know, the forecasters in that domain are not expecting a giant weather event.
It's amazing how little is currently being said about the three interventions in this campaign performed by Comey the God. Also, how little is being said about the failure of the predictive giants who sailed the ship of state into an iceberg, as was done with the unsinkable Titanic of old.
In this morning's New York Times, Julie Turkewitz filed a fascinating report from the conference of the International Association of Political Consultants. She described a remarkable panel discussion involving three polling experts.
Who knows? Maybe Turkewitz is misreporting what these experts said. But if Turkewitz is reporting it straight, they made some remarkable statements.
The experts seem to have described themselves as complete incompetents. Presumably, the word "some" really means "two" in this three-panelist context:
TURKEWITZ: In one session dedicated to polling, three panelists who had predicted Mrs. Clinton would win took to the stage, framed by a royal blue backdrop. Instead of PowerPoint presentations and state-by-state voter analyses, there was morose self-flagellation, as some admitted they had spent the election seduced by “magical thinking,” unable to envision a Trump presidency and therefore blind to the story in front of them.If Turkewitz is reporting this straight, two of the pollsters confessed to "magical thinking." They'd been unable to envision a win by Trump. For that reason, they were blind to what was occurring.
Margie Omero of PSB Research theorized that pollsters had held back Trump-leaning data, unwilling to release something that looked like an outlier. Or that Trump supporters had simply not told pollsters the truth, either embarrassed by their choice or angry at callers whom they perceived as part of a conspiracy against him.
“It was impossible to conceive of an incoming President Trump,” said Ms. Omero, whose firm has worked for both former President Bill Clinton and Hillary Clinton over the years. “A couple-point advantage seemed comforting to prevent something so catastrophic.”
Mr. Anderson, the Democratic pollster, said he should have seen the win all along.
“The story that played out was right in front of me,” he said. (Mrs. Clinton never went above 45 percent in his surveys.) “But my basic assumption, and why I think why I was wrong was I continued to say to myself: ‘This segment of voters, about 15 percent, who view both of them unfavorably, they think Clinton has the judgment, temperament and qualifications to be president. They do not think the same of Trump.’ So I believed that at the end of the day they would vote for Clinton.”
Omero theorized that pollsters simply withheld data which favored Trump. That's pretty much what Trump said all along.
Omero further said it had been impossible to conceive of Trump winning. Anderson seemed to say that he simply refused to believe that a key segment of voters could possibly break for Trump. He'd been reading the flight of birds.
We've suggested for years that you shouldn't put your faith in our professors and experts.
We've long been amazed by the general incompetence of the nation's education experts. If Turkewitz is reporting this straight, it may be time for a pair of top polling experts to go.