“Senator, it’s Dan Rather. How are you doing?”
It was 8:53 p.m. on Tuesday, March 2, just 46 minutes after the Associated Press had reported that Senator John Edwards would drop out of the race to win the Democratic Presidential nomination. Senator John Kerry had won most of the key states, including Mr. Edwards’ last Southern stand, Georgia. Mr. Rather suspended his telephone receiver in mid-air, NYTV on the line, as Mr. Edwards called the anchor of CBS News personally on another to confirm the reports-and, from what was audible across the synapse of Mr. Rather’s desk, a little consolation.
“I know it’s been a long, hard slog,” Mr. Rather could be heard saying to Mr. Edwards, “and I hope not to lose touch with you. If you need a place to stay, let me know.”
Behind Mr. Rather, Freddy Mercury could be heard belting out “We Are the Champions” on a CBS commercial.
With that, Mr. Rather completed the night’s biggest headline for CBS News: The news had called him. And he’d done it, as usual, with a kind of galloping grace. Even as Georgia had looked all but lost to Mr. Kerry-Fox News had been first to project him the winner in that state-Mr. Rather and CBS News were playing it cautious. After 8 p.m., the network still hadn’t given Georgia to Mr. Edwards, even though the A.P. was reporting that Mr. Edwards was ready to throw in the towel. At 8:25 p.m., Mr. Rather went on the air for an update and finally said it: “John Edwards is reported ready to quit.”
“We would rather be last than be wrong,” Mr. Rather later said. “If you start from that premise, then we’re going to be conservative and careful. And we were tonight and I hope that we will be in the future.”
After all, he said, “there’s no such thing as half pregnant.”
As of 9:45 p.m., Fox News remained the only network to have called Georgia for Mr. Kerry, with the rest of the networks and CNN remaining as cautious as Mr. Rather. Clearly, the incoming exit polls meant different things to different people.
Just as in every election since 1968, CBS News was using exit polls again to pick a winner. And for that, they relied on the same man they had relied on for 36 years: a tall, white-haired, 69-year-old gentleman, gravel-throated, his grin still impish: Warren Mitofsky.
On Tuesday, Mr. Mitofsky-whom pollster John Zogby had deemed a “grumpy old man” in The New York Times on Feb. 19-sat in his office in Somerville, N.J., overseeing the mind-numbing computer models that sliced and distributed data as it poured in from all 10 states, eventually piping the results to five TV networks-ABC, NBC, CBS, CNN and Fox-which would then be eyeballed by their analysts as they decided when to make a call.
Mr. Mitofsky, who created the gold-standard CBS News/ New York Times poll in 1975 for CBS, pioneered the art of projecting elections for network TV when he became the first to use exit polls in 1967 for a governor’s race in Kentucky. Although his legend has rarely reached beyond the confines of TV news, Mr. Mitofsky has been the elder statesman of TV’s crystal-ball business for 37 years.
“He is the best in the business, period,” said Mr. Rather, who has known Mr. Mitofsky for 35 years, “because he has been-through thick and thin, through sunshine and storms-the most accurate. Not only the most accurate, but the most accurate by far .”
Mr. Mitofsky says he has projected about 2,500 elections and made only six mistakes.
“Statistically,” said Mr. Rather, “nobody in the business can come within five area codes of those kind of figures.”
Nevertheless, however, the last major projection using Mr. Mitofsky’s data led to the biggest error of all time: calling Al Gore the President of the United States at 7:50 p.m. on Nov. 7, 2000. While the networks insist that they’ve completely overhauled the system since then, disbanding the Voter News Service and regrouping as the National Election Pool, or N.E.P., Mr. Mitofsky has his own take: The events of 2000 were a rare confluence of minor errors-”a fluke,” he has written-and the new system won’t be radically different from the old one. On Jan. 27, Mr. Mitofsky told the Chicago Tribune , “These are the same models. I would say the changes are subtle …. They’ve all agreed to use the same criteria.”
In other words, on Nov. 2, 2004, there will still remain the infinitesimal chance that things could go wrong again. And as a result of his not being perfect-of his being human in an inhuman age-the networks have come to consider Mr. Mitofsky a public-relations liability. Linda Mason, the vice president of public affairs at CBS News and an N.E.P. representative, said that Mr. Mitofsky didn’t really mean what he said to the Chicago Tribune .
The new projection system, she said, has completely overhauled. “It’s been changed dramatically,” Ms. Mason maintained. “What he said is not what he meant to say, I understand.”
Well, she would know.
Of course, there was no way of knowing what Mr. Mitofsky meant to say, because, like some romanticized duffer in a Rod Serling or Paddy Chayefsky script, Mr. Mitofsky’s discursive intelligence had run up against corporate needs, and he was muzzled; the consortium insisted that he not talk on the record for a profile.
“The overall big plan, the schematics, remain the same,” Ms. Mason continued. “And he was the originator of those schematics back in the 70′s, but the utilization is different, the methods of collection are different, and technology has improved so much. So it’s not the same.”
In the Serling script, Mr. Mitofsky would be replaced by DEMO 2100, the network-pandering computer.
But this was not Serling or Chayefsky. And according to multiple sources, the networks let Mr. Mitofsky know in no uncertain terms that he should keep his trap shut after his statements in the Tribune . In fact, they had already tried to put it into writing early on: According to a source familiar with the situation, the networks stipulated in Mr. Mitofsky’s contract that he not speak to the press unless he first consulted with them. Despite that, the networks have had a tough time keeping Mr. Mitofsky’s spiky opinions at bay. As recently as Feb. 19, he criticized the polling firm Zogby International for predicting that Senator Kerry was ahead of Senator Edwards in the Wisconsin primary by 27 points-a measure that proved to be off by 21 points.
The polling, he told The New York Times , made Mr. Edwards “look like a hero.”
“A lot of people just keep their mouth shut, but Warren, he passionately believes in what he’s doing,” said Mr. Rather. “In the world where being outspoken is considered not exactly clubbable, Warren takes his knocks because he’s outspoken. You don’t want to ask Warren what he thinks unless you really want to hear it.
“Some people don’t like it, but I love it,” he added.
Mr. Mitofsky, of course, is one of those barnacled, cantankerous TV executives of the old stripe, the kind of guy CBS News had a great history of producing and a not-so-great history of silencing: indispensable pros with too much talent to scrub clean. A certain Don Hewitt comes to mind; Burton Benjamin, who was Walter Cronkite’s old producer; Mr. Cronkite himself; Fred Friendly, of course; Mr. Rather. But in this case, Mr. Mitofsky was effectively convinced that his job security depended on staying true to his contract. On Friday, Feb. 27, a network spokeswoman serving as the publicist for the N.E.P. said they realized that Mr. Mitofsky had been in the business for “80 years and everything,” but that Mr. Mitofsky did not “represent” the consortium and therefore would not be able to discuss anything pertaining to the N.E.P.-and especially not the inner workings of the top-secret process.
Speaking of Mr. Cronkite, when reached by NYTV, the former CBS News anchor cited “the great political and sociological benefits of exit polling …. It’s a shame that it got turned into the business of forecasting returns rather than appreciating the historical information that was accumulated.” As for the 2000 election, Mr. Cronkite said: “Who could figure that the voting machines and the voting technique would be so out of whack, as it was in Florida, that one state would determine the outcome? Which, of course, amplified-magnified terribly-the system that was used. But the system didn’t deserve the criticism it got. It wasn’t the system’s fault.”
A source close to CBS News told NYTV that network executives considered Mr. Mitofsky “a very, very difficult, ornery man” on whose data expertise they reluctantly relied. They now saw Mr. Mitofsky not as a legendary projections guru, but as an annoying cog in a newer, decentralized machine that they’d built to keep another debacle from occurring. They characterized the N.E.P. as made up of multiple partners, with Mr. Mitofsky’s company, Mitofsky International, merely a participant in that process. The network spokeswoman pointed out that Mr. Mitofsky also had an active and equal co-partnership with Edison Research, whom insiders said the networks preferred to deal with over Mr. Mitofsky.
“We think the process speaks for itself,” said Ms. Mason. “We’re not interested in showing how the sausage is made if it works well. No one person is totally responsible for this system.”
The friction over how to characterize the changes in the new process came after four years of internal hand-wringing and conflicting reviews of what went wrong and who was to blame. But the networks were especially motivated by a Congressional hearing before the House Energy and Commerce Committee, in which Republican Congressman W.J. (Billy) Tauzin of Louisiana sought to find out just why every major news anchor had told the public the wrong thing twice on the evening of Nov. 7, 2000. By all accounts, the probe was highly confrontational, and network executives complained that it was unfair. Subsequently, the networks had to pump money into revising the system and disbanding V.N.S., which they renamed the National Election Pool.
The message: All better now.
Mr. Mitofsky has agreed in print that this year’s Presidential election would feature an improved projection system. Mr. Mitofsky believed the old system was pretty good, and the new system had improved things by a few degrees-it was also pretty good. But none of his public statements in the last two years support the network’s image of a “dramatically” overhauled system, especially his comments in a public-opinion article published in 2002, in which Mr. Mitofsky concluded: “In my opinion, nothing can keep the networks from someday calling another race incorrectly.”
For now, if you want the networks to give you insight into this new system, forget it. Dan Merkle, the director of the ABC News decision desk, wouldn’t even tell NYTV where the analysts physically sat in relation to Peter Jennings on election night as they pored over Mr. Mitofsky’s data. “I don’t want to get into that,” he said. “When it gets down to what we do on election night and how we do it, we’re obviously in a competitive business-it’s not the kind of thing we typically talk about. It’s more from the competitive angle: We have a way of doing things that works well.”
Actually, he said, it’s not just about competition.
“It’s a news business, and obviously you need to be as timely as possible,” he said next. “But the most important thing is accuracy.”
Accuracy, competition, avoidance of government penalty-they’ve plagued TV election projections since networks started making them in the 1950′s. In a speech to the Washington Statistical Society, written by Mr. Mitofsky and edited by former CBS News election consultant Murray Edelman, Mr. Mitofsky recalled those first attempts: “At CBS, UNIVAC and Charles Collingwood gave the ‘odds’ on a winner,” he wrote. “The odds that John Kennedy would beat Richard Nixon were 3 to 2. That sort of thing. That was embarrassing.”
But that changed when CBS discovered Lou Harris, the founder of Harris Research. In a phone interview, David Blank, the CBS News chief economist from 1955 to 1982, who worked with Mr. Mitofsky, recalled that when CBS first learned of Mr. Harris during the Presidential race of 1960, Senator John F. Kennedy was personally using Mr. Harris’ projection system to get the early results. “Lou Harris was sitting in the bathroom in the second floor of Kennedy’s Hyannisport house, with a telephone calling down to Kennedy, and he was so ahead of the networks, everybody was overwhelmed,” said Mr. Blank, now 83. A CBS News executive at the time, Richard Salant-who became president of CBS News in 1961-”called Harris and said, ‘I don’t know what kind of magic you have, but will you come work for us?’”
Mr. Harris’ polling group became one of those three-word monikers that networks have favored ever since: Voter Profile Analysis. “They thought they had a contractual agreement with God,” said Mr. Blank. “They thought they had the inside of the most perfect forecasting system in the world.”
But Harris’ system-based on quotas and using sample precincts within states-ran into problems in 1966, during two races for governor. The problem: If it failed to eliminate deviant precincts from its samples-for instance, when African-Americans voted in large numbers for Republicans and third-party candidates-it spat out projections based on bad assumptions. In Georgia, segregationist Lester Maddox was the Democrat; in Maryland, Spiro Agnew was the Republican. CBS got both races wrong.
A lifelong Democrat who grew up in the Bronx, Mr. Mitofsky had studied statistics at Guilford College in Greensboro, N.C., the birthplace of Edward R. Murrow, and after graduate school, worked for the Census Bureau in 1960. In the mid-60′s, Mr. Mitofsky heard a story from a market-research executive named George Fine, who told him that he was interviewing moviegoers leaving theaters to collect feedback for a film company so they could revise the films.
In 1966, at 32, Mr. Mitofsky was hired at CBS, where he eventually refined Fine’s movie-house concept for voting, flagging down voters outside of representative precincts and collecting the data by phone to create a close approximation of how an election would likely turn out. In 1967, Mr. Mitofsky used exit polls to predict the governor’s race in Kentucky. With that, CBS formed a partnership with I.B.M. to use their room-sized computers to tabulate the data, and in CBS News used the system in 21 states to predict the outcome of the Presidential race in 1968. It was so successful that it has remained largely the same ever since.
Later, Mr. Mitofsky invented the CBS News/ New York Times poll, which is still widely used today. As of this writing, the latest one had Senator Kerry with 57 percent of primary voters to Senator Edwards’ 18 percent.
For 20 years, CBS largely dominated with their projections, but that reign ended in the late 80′s, when the high cost of projecting results-close to $20 million for a four-year election cycle-forced them to pool their resources with the other networks, which were thrilled to stop getting beaten on election night.
“Warren perfected exit polling so well that our major competition just gave up trying to match it, and came and said, ‘Why don’t we pool it?’” recalled Mr. Rather. “It was a great mistake for CBS News, in my opinion. But it wasn’t my decision to make, nor should it have been. In 1984, we were so much better than either one of the two, and they knew it. They said, ‘Why don’t we just save money and all of us just pool?’
“It was frankly a foolish thing to do,” Mr. Rather continued, “but the powers that be at that time thought it would save money … we were so far ahead of the others, it was a laugher. And they knew it …. Warren at the time said, ‘Look, I’m a professional-if this is what you want to do, O.K., but we really do have a hammerlock on this.’”
In 1990, the Voter Research and Surveys was invented, and Mr. Mitofsky called all the 1990 and 1992 races for the networks. In 1994, V.R.S. merged with the vote-counting arm, News Election Service, to create Voter News Service. That year, ABC News began aggressively trying to project a winner ahead of the other networks using V.N.S. data, which renewed the competition again, with each network making its own projections.
Then came the 2000 election.
The consensus conclusion on what went wrong when the system projected Al Gore as the winner at 7:50 p.m. on Nov. 7, 2000, was that V.N.S. had: a) underestimated the absentee vote, b) ran into a processing error in a crucial county, and c) measured the projection data against the best historical comparison instead of the last three historical votes. The so-called best historic reference proved not to be the best, thus creating a false sense of comfort with Mr. Gore’s lead at that time.
In combination, the confluence of errors, in Mr. Mitofsky’s mind, was anomalous. And even though there is now more data by which to measure the quality of the exit polls, the polls themselves remain the crux of the projection system. In articles and lectures, Mr. Mitofsky has gone into great detail about how the system has been changed: The computers were upgraded and more secondary measures were put in place to catch errors or anomalies, which had always been the underlying principle of projection-cutting risk based on the exit-poll data.
“As it was, the Florida problems have been taken very seriously and were addressed in the new system,” wrote Mr. Mitofsky in the public-opinion article. “That projection system retains most of the basic design of the old system. The new system will do a better job of handling the problems of 2000. Absentee vote and the end of night calculation will be improved, but they will continue to be problems as all proposed solutions are less than perfect. Quality control of the county vote will be better, but will still need improvement. Multiple past races will be used in the estimators to good effect.”
Mr. Mitofsky said that the Gore “projection stained the solid journalism in the reporting of elections for the last 38 years,” but he also said that the system’s essential integrity remained. “From 1967 through 1988, which was the year we worked only for CBS,” he wrote, “we had five mistakes out of the 1,500 or so state elections we made projections for in that period. During the 1990s, the V.N.S. pool had only one error in about 700 projections. However, we have since learned that 99.7 percent is not good enough when that rare event is highly visible.”
Warren Mitofsky’s message to the networks is evident: Nov. 7, 2000, may have been an anomaly, but it could also be a harbinger. A word to the wise: Those who ignore the past will be forced to miscall the future.
Follow Joe Hagan via RSS.