Toluu

WSJ.com: The Numbers Guy

Carl Bialik examines the way numbers are used, and abused.

In Defense of the BCS ComputersDecember 1

For several weeks it’s been clear that college football’s Big 12 South Division could need outside help to sort out the top of its standings. With so little separating the teams on the field, some complex number-crunching provided the determinate factor.

Getty Images

Vondrell McGee and his Texas Longhorns defeated Oklahoma State in October, but Oklahoma passed Texas by beating Oklahoma State on the regular season’s final day.

Three teams — Oklahoma, Texas and Texas Tech — share identical conference and overall records after all winning this past week, and each won once and lost once when facing each other this season. The trio tied on the first four tiebreakers. The fifth, their standings in the Bowl Championship Series rankings, would determine the division’s representative in the Big 12 championship game, where any of the three would be favored to beat Missouri. If that team did, it would almost certainly qualify for college football’s national championship game.

The BCS rankings are an average of three components: two polls and a composite computer ranking based on six separate computer rankings. Texas was ah

Zogby’s Misleading Poll of Obama VotersNovember 20

During a campaign, pollsters can build credibility by forecasting election results accurately. Afterward, they can build revenue by using that credibility to attract private clients. These private surveys often have an agenda, and their numbers can’t be tested against an objective standard, such as votes. Such surveys can test pollsters’ standards of conduct.

Zogby International recently conducted a survey for a critic of president-elect Barack Obama and then, together with the sponsor, interpreted the numbers from the survey in a misleading fashion.

John Zogby, the president and chief executive of Zogby International, released a statement on his Web site Tuesday that defended the poll in the face of heavy criticism from bloggers and some media organizations, saying, “We stand by the results [of] our survey work on behalf of [author and former talk show host] John Ziegler, as we stand by all of our work. … We were hired to test public opinion on a particular subject and with no ax to grind, that’s exactly what we did.”

But on Wednesday, Zogby told me he was on vacation when the contract was reached and when the survey was conducted, and wouldn’t have approved the poll in the form it took, or

It Ain’t Over…November 10

While I’m on vacation, I’m providing links here to some other writing about numbers from around the Web.

In the rush to report Tuesday’s election results , it’s easy to forget that the numbers aren’t final. Paradoxically, early votes sometimes are counted last. Consider this passage from the Omaha World-Record: “Obama won 8,434 of 15,039 mail-in ballots counted Friday by Douglas County election officials. These early ballots arrived in the election commissioner’s office too late to be included in Tuesday’s election results.” Also in the article is a chart showing the impact of those late votes, turning a 569-vote deficit into a 1,260-vote lead for Obama in the Nebraska congressional district that includes Omaha, and producing the first split of electoral votes in Maine or Nebraska since the states starting awarding an elector for the winner of each congressional district.

Late votes and uncounted votes will help determine two Senate races not yet called: Those in Alaska and Minnesota. The Democratic candidate in each race trails. But Nate Silver, an Obama supporter and election number-cruncher, writes that in both states the numbers may favor the Democrat. In Minne

Breaking Down Voter-Turnout NumbersNovember 7

High voter turnout was predicted on Tuesday. And the number of votes counted has already shattered all records. But so, too, does the population of the U.S., regularly. The turnout rate appears to be shy of a new record.

Focusing on the number of Americans who voted for president is a surefire way to generate headlines touting records, such as those that ran in the Financial Times, the Atlanta Journal-Constitution and Politico. Just two of the last 15 presidential elections — in 1988 and 1996 — didn’t set a new record by that count, according to data compiled by Michael McDonald, a political scientist at George Mason University. Focusing on the total number of voters amid steady population growth is akin to celebrating new box-office records, except that turnout records are easier to come by because there aren’t other elections competing for voters’ attention.

So just how many people did vote this time? We don’t have final numbers yet, as usual. Shortly afte

Obama Won. So Did Election Forecasters.November 5

My print column this week notes the impressive accuracy of pre-election state polls in the presidential election, and the equally impressive performance of poll aggregators that combine all the disparate polling numbers into aggregate forecasts.

Fivethirtyeight.com, the forecasting site created by baseball researcher Nate Silver, gained special prominence thanks to its cool-headed commentary, impressive graphics and rich data source. And Silver’s forecasts beat just about every pollster active in many states on Tuesday — the notable exception was Zogby International — while also stacking up well against competitors that were also aggregating polls back in the 2004 election, Real Clear Politics and Pollster.com, which provides polling data to WSJ.com.

Nate Silver (Photo by Robert Gauldin)