Taking the pulse of Nate Silver's numbers

(CBS News) After crunching the numbers, our CBS News polling unit currently shows President Obama with the narrowest of leads over challenger Mitt Romney. Subject to change over the next three days, of course. The promise and peril of high-tech vote prediction was first demonstrated live on our air back in 1952 . . . as Martha Teichner reminds us now in our "Sunday Morning" Cover Story:

It wasn't just the first coast-to-coast broadcast of a presidential election, and Walter Cronkite's first time anchoring on election night; 1952 marked the first time a computer was used to project the winner.

Laughable now, groundbreaking then.

"Univac, can you tell us a prediction now? I think that Univac is probably an honest machine . . . "

But Univac did have something to say. With not even 3 1/2 million votes counted, "he" predicted 100 to 1 odds of an Eisenhower victory, in a landslide so huge it seemed impossible, given what was then thought to be a close race. The results were withheld for several hours, fearing humiliation. But Univac was right.

The war between the stat heads and the pundits had begun.

Fast-forward 60 years, to 2012, and we have New York Times blogger Nate Silver's prediction of an Obama win on Tuesday, no matter what the popular vote:

"We have him as a 75 percent favorite roughly, almost an 80 percent favorite, in fact, to win in the Electoral College, less than that in the popular vote," said Silver. "Politics is full of people who are trained to manipulate the way that we view information, so when they see information that they don't like - based on what our computer program that we designed four years ago says every day - they're going to become very, very upset.

"They can't manipulate what my computer says," he laughed.

So is this another Univac moment - or a stat head headed for a fall?

Nate Silver has become the numbers geek pundits love to hate, particularly Republican pundits.

"Nate Silver says this is a 73.6 percent chance that the president is going to win? Nobody in that campaign thinks they have a 73 percent chance, they think they have a 50.1 percent chance of winning," former Republican Congressman Joe Scarborough said last week on MSNBC's "Morning Joe." "And you talk to the Romney people, it's the same thing."

"And anybody that thinks that this race is anything but a toss-up right now, they're jokes."

Silver argues that the race is tight, but not a toss-up. He calls his blog 538, because that's the total number of Electoral College votes; a presidential candidate has to get at least 270 Electoral Votes to win. No other number really matters, and Silver's computer model says President Obama is ahead in the states capable of delivering that magic 270 - namely, Ohio, Wisconsin, Nevada and Iowa.

So how does this model work?

"Every poll that people read about goes into the model and feeds the projection in some way," Silver told Teichner.

There might be 30 or 40 of them a day, in this year's poll-happy universe. Silver's model averages all those polls, then factors in how well they've performed in past elections, and comes up with probabilities, like gambling odds.

"So in Florida, for example, we had Romney with a 60 percent chance of winning. That's how often, when you have a one-point lead in the average of polls, you've wound up winning in the past. In Ohio, we have Obama with about a 75 percent chance of winning, because he has a larger lead, and so it's more likely to be enduring on Election Day itself."

"Now, is this any different from crowd sourcing?" Teichner asked.

"Well, in a way, that's what we're doing," he replied.

And think "Moneyball" - the book and film about how a geek used numbers to outdo the Oakland A's scouts, using their intuition to find undervalued baseball players who could win. In other words, the stat heads versus the pundits again.

Nate Silver started out as a baseball statistician before moving on to politics. His new book, "The Signal and the Noise," describes his theories.

In the 2008 election, Silver called 49 out of 50 states right.

Teichner asked Sarah Dutton, director of surveys for CBS News, to describe the difference between the kind of polling that CBS does and the sort of process that Nate Silver goes through.

"Well, I think those are very different animals," Dutton said.

The polls CBS News conducts along with its polling partner, The New York Times, are among those fed into Nate Silver's model.

"What we do is more of a snapshot in time than a prediction of an election outcome," Dutton said. "And so it's a very different kind of thing. We ask people, if the election were being held today, which candidate would you vote for? But really, most of our polls are about things that are not horse race-related but maybe give you really some insight into sort of the mind of the electorate."

Here are CBS' most recent poll results: "A very close race," Dutton said. "Forty-eight percent for President Obama, 47 percent for Mitt Romney - just a one-point lead for the president. And that is within the poll's margin of error."

Meaning, President Obama and Gov. Romney essentially neck-and-and at the end of that race.

What Nate Silver does is use his model to say President Obama is 70 to 80 percent more likely than Mitt Romney to cross the finish line first . . . and that's what makes him controversial.

"Nate has gotten very good reviews from people who have studied his methodology, but just like a clock, even the most inaccurate pollster gets it right from election to election," said GOP strategist Frank Luntz, a CBS News consultant. "And just like a clock, even the most accurate people will get it wrong from time to time."

"Mitt Romney's pollster has staked his reputation that Mitt Romney is going to win this election; they're going to win Ohio and many of these key states, and that the published polls are wrong," said Luntz.

Teichner asked Silver if he were absolutely convinced Mr. Obama was going to win. "Oh, no. I'm convinced that if offered even money I would be happy to bet on Obama. I would need a pretty good price to bet on Romney, the two-to-one wouldn't do for it for me. Three-to-one might."

So who or what to believe? In the 7-Eleven coffee cup vote - right in the last three elections - Obama is ahead 59 to 41 percent.

Obama Halloween masks outsold Romney masks 63 to 37 percent. The mask "poll" has been right since 1996.

If this is any gauge of the tightness of the race, in the Family Circle magazine "First Lady Cookie Contest," Michelle Obama's white and dark chocolate chip cookies squeaked by Ann Romney's M & M cookies by 287 votes out of more than 9,000.

To quote a horrible journalistic cliche, only time will tell.

For more info:

• "The Signal and the Noise: Why So Many Predictions Fail - But Some Don't" by Nate Silver (Penguin)