Press Box

Defending the Projectionists

The TV news networks deserved their trip to the pillory for botching the Florida presidential call, with scores of newspapers ridiculing them for their Dewey Defeats Truman-sized error. But I’m not so sure about the sequel beatings. Rep. William Thomas, R-Calif., has announced plans to hold congressional hearings on the networks’ debacle before the end of the year. Look for him to interrogate the subpoenaed network chiefs as if they’re tobacco CEOs. Louisiana demagogue Billy Tauzin, who works a day job as a Republican member of the House, wants hearings of his own next year, but I doubt there will be much in the way of flailable meat for him to whack by the time Thomas is through. Or after the Annenberg, Pew, and Shorenstein fellows pick the network and VNS bones clean.

Or by the time Steve Brill finishes. Brill, whose idea of persuasion is shouting into your ear, handed the networks a Six-Point Diktat this week about how they should conduct themselves in covering future elections: 1) No projections in any voting jurisdiction before all its polls close. 2) No declaration of winners that isn’t based on actual vote counts. (All projections must be labeled as such.) 3) Similar qualifying language in all voice-overs, headlines, etc., making it clear that the projections are projections. 4) Disband the Voter News Service–the information cartel run by CBS, NBC, ABC, CNN, Fox, and The Associated Press–which produces exit polls and other data used to project elections. Also, increase competition by promising that each network will do its own exit polling and tabulations. 5) Ban the pundits from pontificating on a prediction before it’s mathematically certain. 6) Vow to check with Steve Brill before predicting any winner. If Steve isn’t available, call David Kuhn. If David is out stoking the buzz, dial Tom Rosenstiel. If Rosenstiel can’t come to the phone, send e-mail to James Fallows.

OK, I made Diktat 6 up. But you get the idea: The networks shall relinquish the right to report any news that is not 150 percent accurate!

At first blush, Brill’s zero tolerance approach to election coverage sounds swell. Ban all errors! Call no election before its time! And best of all, fill the streets of New York with VNS blood! But you can no more fail-safe engineer Election Night journalism with Brillian diktats than you can fail-safe engineer nuclear reactors or space shuttles. Nothing can be made 100 percent reliable. Things melt down. Things blow up. Counts don’t add up.

But if there is any journalistic enterprise that comes close to mission-critical perfection, it’s Election Night projections. Warren Mitofsky, a projection godfather and the man behind the curtains at both CBS News and CNN on Election Night, boasts that in his 33 years of calling more than 3,000 elections around the world, he’s blown only five contests, not counting Tuesday’s election. Five! Maybe instead of investigating why the networks got it wrong Tuesday, Reps. Tauzin and Thomas should figure out how they get it right so often. Like pure drinking water from the tap and steady wattage from the wall socket, precise election projections are something that we spoiled news consumers have come to take for granted.

But Three Mile Island did go gooey, the Challenger did turn into a roman candle, and the networks, backed by VNS, did blow Florida. At 7:50 p.m., NBC used the VNS data to project Gore the winner. Mitofsky directed CBS and CNN to do the same minutes later, and a couple of minutes after that, VNS made the same recommendation for its members and subscribers. With the sigh of a man weighted down by a very large hair shirt, Mitofsky accepts blame for making the mistake. The data collected by VNS, which included the now infamous exit polls, wasn’t as good as it should have been, he surmises. The statistical model was lacking, he concedes. “The system isn’t perfect,” he says.

According to Mitofsky, the maximum margin of error for the VNS statistical models used to project Election Night winners is about 1 in 200. (In practice, the projections are a lot more accurate, as Mitofsky’s batting average above attests.) Mitofsky points out that the Florida mistake is only the second VNS-derived projection error since 1990, when a New Hampshire senatorial contest confounded the model. Is Brill’s Content as accurate as Warren Mitofsky? I say “No,” and demand a congressional investigation!

Although Mitofsky and the other network seers depended somewhat on VNS exit poll data when they gave Florida to Gore at 7:50 p.m., no exit polls were used when they gave Florida to Bush at about 2:15 a.m., just the hard votes tallies from VNS and the state. At this point, Bush was leading by 56,000 votes with 180,000 outstanding ballots, meaning that the uncounted would have had to go to Gore 2-to-1 in order for the vice president to win–a preposterous notion. “Anybody with an envelope and a pencil could have done what we did,” says Mitofsky, who was clicking on the Web site of the Florida secretary of state over and over for the latest vote returns.

(Against this statistical backdrop, it’s somewhat hilarious that the press ethicists are calling for the head of John Ellis, the Fox News prognosticator and Bush cousin who gave Florida to Bush contemporaneously with the networks. Although Ellis has been fairly upfront about rooting for his cousin, we need to remember that he was looking at the same data as Mitofsky and the other network seers who made the same call. What incentive did he have to project Bush the Florida winner if he knew the numbers wouldn’t ultimately support it? It’s not like projecting Bush the winner could make it so. If the press ethicists want to banish Ellis from the prognostication bar because he’s partisan Republican or because he’s phoning high-level Republicans who are blood relatives for inside dope on Election Night, let’s apply similar litmus tests to the prognosticators who canoodle with Democrats. As a journalist, Ellis would be derelict not to draw on every scrap of information he could find.)

By 2:40 a.m., the Bush lead was still 56,000 with 68,000 votes outstanding. But by 3:10 a.m., a call came in putting the Bush margin down to 10,000. “It was as if a spigot had been opened, and the Bush votes were pouring out,” Mitofsky says. The VNS data collection operation, it turns out, had grossly underestimated the size of the outstanding vote. When the lead narrowed to 2,000 at 3:57 a.m., Mitofsky had CBS and CNN retract the Bush projection. So did the other networks.

For a disgraced wizard, Mitofsky is fairly open. “It’s an embarrassment, but I don’t know how anybody is going to improve their situation by hiding,” he says. As the congressional investigations convene, Mitofsky and company’s embarrassment reddens, and Brill orders a gross of thumbscrews, let’s corral our criticism of VNS and the networks just this side of the First Amendment. The freedom of the press means nothing if diligent journalists can’t make occasional mistakes without prompting inquisitions, especially if they’re willing to issue retractions as promptly as the networks did. Any laws that would proscribe exit polls or projections or would prevent news operations from pooling their coverage are assaults on the First Amendment and should be treated as such.

Having lost their reputation for infallibility, the network seers should find the silver lining. Be more candid with your audience about your Election Night methodology. Educate them about the margins of error that go into every projection. Don’t blame the data when you screw up, blame yourselves.

And show us what goes on behind that curtain.