AJR  Features
From AJR,   January/February 2001

How They Blew It   

A behind-the-scenes look at the television networks' dismal performance on election night.

By Alicia C. Shepard
Alicia C. Shepard is a former AJR senior writer and NPR ombudsman.     

AS VOTES BEGAN STREAMING into Voter News Service's headquarters after Florida's 7 p.m. poll closing, it seemed clear to many network prognosticators that Al Gore was going to clobber George W. Bush in the Sunshine State. What a story. Florida's governor could not deliver the votes for his older brother.
But not all experts hired to help the TV networks on election night thought Florida was a done deal for the vice president 50 minutes after the polls closed. Some didn't trust the accuracy of Voter News Service projection models, models which ABC, CBS, NBC and CNN have relied on for 10 years and which only once had incorrectly projected a major race.
In the end, the doubters were right. At least three ABC analysts warned against calling Florida for Gore, but their advice went unheeded.
It didn't match the VNS models. Besides, raw votes from 120 key sample precincts and votes trickling in from counties were tracking with exit poll data collected that day at 45 sample precincts. Exit poll data and sample precinct votes churned through models that analyzed past voting patterns, factored in exit poll biases, and correlated how candidates stacked up against previous contenders. Around 7:45 p.m., exit poll data, which began coming in at lunchtime, showed a 6.6 percent lead for Gore over Bush. But election analysts knew only a fool would call Florida for Gore based on exit poll information alone.
As votes arrived from sample precincts carefully chosen to represent voters across the state, the model predicted a 5.4 percent lead for Gore. It indicated Gore needed a "critical value"--a statistical degree of certainty--of 2.6 or higher before any network could comfortably hand the vice president Florida. At 7:50 p.m., the "critical value" showed 3.2 for Gore. The Voter News Service model was more than 99.5 percent sure Gore would carry the state.
Under "status," at 7:50 p.m. the VNS screen said: "Call."
In the race to be first, NBC "won," jumping even before VNS at 7:49 p.m.
CBS waited one minute. Warren Mitofsky, who invented exit polls in 1967, has been in the race-calling business for 33 years. He ran CBS' election unit from 1972 until 1990 and is known for his caution. Mitofsky, working for CBS and CNN, had vote totals from 12 of 120 sample precincts and data from 38 exit poll precincts. Gore was doing so well that he concluded exit polls had been overstating Bush's numbers. "The real votes were telling us Gore was ahead," says Mitofsky. "The exit poll data gave us a slight lead for Gore, and the overlap of the two was telling us that the exit poll data should have given Gore more support."
There are three sources of data that VNS uses for its projections. Exit poll results, the least accurate of the three, come in three times during the day. They are only used to project winners. Once the polls close, raw votes from sample precincts are phoned in and measured against exit poll data. The tally that counts--the actual vote total--comes in throughout the evening.
At 7:50 p.m., Mitofsky and his partner, Joe Lenski, confidently instructed CBS and CNN to call Florida for Gore. Fox News Channel, in the presidential projection business for only the second time, followed suit at 7:52, joined by the Associated Press at 7:53 and CNN at 7:55.
"The sad fact is that was a straightforward call," says Jonathan P. Wolman, AP's executive editor. "VNS' projection material provides a guidepost that warns you statistically if there's a bias in the material that might skew the results. In this case, that bias indicator said it might be underestimating Gore's advantage."
But not everyone saw things that way.
At VNS' temporary quarters on the 93rd floor of Manhattan's World Trade Center, two political scientists working for ABC, each with a strong statistical background, didn't think the Florida result was clear-cut. Nor did they completely trust the VNS model. When the decision desk telephoned the two analysts asking, "Can we make the call?" both men advised against awarding Florida to Gore.
Polling places in Florida's Panhandle in the Central Time Zone wouldn't close for 10 more minutes. Only 237,115 actual votes had been tabulated in a state with 8.8 million registered voters. But other factors involving statistical probabilities and VNS models troubled Kenneth Goldstein, of the University of Wisconsin, and Christopher Achen, of the University of Michigan, where election surveys were pioneered in the 1940s.
Achen had flown to New York City six days before the election to prepare for the big night. He spent four days studying VNS models, trying to pinpoint why they had screwed up in picking the winner of the 1996 New Hampshire Senate race.
The networks--ABC, CBS, CNN and NBC--and the AP created VNS in 1993. The idea was to save money by pooling resources and receiving data amassed by a single source. Fox joined the consortium in 1996.
For the most part, the setup has worked well. VNS projections have largely been accurate, to the point that they have virtually been treated as facts--a state called for one candidate moments after polls close is seen as decided. But four years ago, the system failed: Each network, relying on exit polls, prematurely called New Hampshire for Democratic Senate challenger Dick Swett. When the votes were counted, it turned out Republican Sen. Robert C. Smith had been re-elected.
"I dug in and went over that as extensively as 30 years of experience would allow me," says Achen. "I worked out a rule of thumb to protect myself on election night from it happening again." Achen was so determined to prevent a similar error that ABC colleagues began referring to him as "Mr. New Hampshire."
Looking at the Florida data around 7:30 p.m., Achen noticed the Gore exit poll numbers were higher than VNS had predicted. This bothered him because exit polls tend to have a Democratic bias. Plus, not all exit poll data was in before 8 p.m. And what about absentee votes? Analysts expected 10 percent of Florida votes to be absentee. With about 6 million people expected to vote, that's 600,000 votes that exit polls know nothing about.
"But first and foremost," says Achen, "when I applied my rule of thumb to protect ABC against mistakes, it indicated it was too soon to call."
And so, while other networks were falling all over one another to declare Gore the victor in Florida, Achen held back. While the VNS model indicated more than 99 percent certainty for Gore, Achen saw it as more like 85 percent.
Neither he nor Goldstein advocated making the call. There was too much at stake to move precipitously. "It wasn't like we were calling an off-year dog catcher race in North Dakota," Goldstein says. "Besides, what's the hurry?"
At ABC headquarters on Manhattan's West Side, Paul Freedman was part of the six-person decision team formed to call the Senate, gubernatorial and presidential races. Freedman, a University of Virginia political scientist, also thought it was too early. "It's fair to say the three of us wanted to be more confident before making a call, because we thought there was too much uncertainty in the estimate," says Freedman. Others, he says, were also endorsing caution.
But they weren't advising in a vacuum.
By 8 p.m., the other networks were flashing Florida for Gore. Pundits were proclaiming that a Gore win there just might put him in the White House before the 11 o'clock news. Despite the misgivings of its experts, ABC's team couldn't resist the competitive pressure, and ABC decision desk chief Carolyn Smith made the call.
None of the advisers claims to be a white knight. They could have argued their case more forcefully, but they didn't. They are academics hired to share their wisdom, not adrenaline-charged journalists impatient to make a decision.
So, at 8:02 p.m., anchor Peter Jennings joined the pack. "ABC News projects that Al Gore wins the state of Florida and its 25 electoral votes," said Jennings. "Give him the first big state momentum of the evening. This is the biggest state where the race has been close, the fourth biggest electoral prize."
As Jennings spoke, Goldstein turned to Achen. "I think they may have fallen into the New Hampshire trap," he said.
Gore's lead began to shrink within 10 minutes of ABC's call.
"Think how we would have felt if we had really had the courage of our convictions and if ABC hadn't called it," Achen says. "But we didn't. The team was wrong, and I'm part of the team. I don't want to say it wasn't my fault."

ELECTION NIGHT IS SHOWTIME for the networks. The story is huge and fast-unfolding, and competition is fierce. With so much on the line, each network prepares extensively, beginning years before the presidential vote. They hire experts, spend lavishly on dazzling graphics, design eye-catching sets, and do more research than a Ph.D. requires.
The mission is simple: Get it right. During a November rehearsal, Smith lectured ABC election night personnel for 15 minutes, stressing that "we weren't in a race to be first," recalls Craig Ammerman, a former executive editor of the old Philadelphia Bulletin who has worked every national election for ABC since 1982. "We were there to get it right. Especially since the presidential race was so close."
But the drive to be first is powerful. This is a tricky dance when all of the networks are getting the same information at the same time and in the same way. It's particularly dicey in the case of an exceptionally tight election.
Sharing election data among networks began in 1964. That's when ABC, NBC and CBS banded together to form News Election Service to collect poll data and enter it into computers. While they shared vote counts, each news division ran its own election unit and made its own calls. Exit polls were used in the 1970s, but only to flesh out voting patterns and trends. That is, until NBC scooped everyone in 1980 by calling 11 states based on exit polls. That enabled the network to declare Ronald Reagan president at 8:15 p.m., while ABC and CBS waited for more votes to be counted.
"ABC didn't call the election until 10 minutes to 10, right before Jimmy Carter made his concession speech," says Mitofsky, who was then running CBS' election shop. "CBS didn't call it until 10:20 p.m., after Carter made his concession speech. That was my doing. Yeah, it was hard. I didn't need any lessons in exit polling from NBC. It bothered me that they were using them and we weren't."
After 1980, each network conducted its own exit polls on a massive scale. But costs mounted and network news executives, far more focused on the bottom line than in the past, scrambled to save money. In 1990 they created Voter Research & Surveys to conduct exit polls, offer analysis and make projections for all networks. Mitofsky was named to head the new operation. By joining forces, each network would save $9 million over a four-year period, according to a report cited in David W. Moore's book, "The Superpollsters." It also made VRS research affordable for CNN.
Philip Meyer, a pioneer in what is sometimes called "precision journalism," spoke out against a network consortium at a 1991 meeting of the American Association for Public Opinion Research, calling it "a bad idea." If VRS makes a mistake, Meyer said then, there could be "terrible consequences" because there would be no other exit polls to serve as a counterweight. If VRS was wrong, he reasoned, everyone would be wrong.
"My concern then and now is when you share responsibility like that across news organizations that normally compete, if everybody's responsible, then nobody's responsible," says Meyer, who teaches journalism at the University of North Carolina. "When they were competing organizations in 1988, they gave different answers, and that was embarrassing to the networks. The visible conflict is good because it reminds everybody how delicate these instruments are. Competition produces better results."
In 1993, VRS and NES merged to create Voter News Service, an effort to save even more money. Mitofsky left and Murray Edelman became editorial director. The terrain shifted that November when ABC broke from the pack and surprised everyone by making its own calls ahead of VNS--albeit using VNS data.
Thus began the age where networks artificially compete, using identical information supplied simultaneously but reaching their own conclusions on their own timetable.
On Election Day, VNS' payroll, which includes about 30 permanent employees, swells to about 45,000, including election year staff, data input operators in the Cincinnati area and New York City, exit poll interviewers and people who collect county and precinct votes in each state and the District of Columbia.
In Florida, VNS-trained interviewers conducted scientific exit polls in 45 precincts with 4,356 people after they voted and staffed 120 sample precincts and 67 county election offices.
VNS "reporters" collect vote totals and phone them in to operators. Statisticians at the World Trade Center crunch the numbers through a variety of statistical models based on historical and geographical voting patterns. Then the computer comes up with a probability that statisticians use to project a winner.
Some critics say VNS models are outdated and not statistically sound. "I do think we need a software update on giving estimates of probability that takes into account changes in the country, especially on absentees," says ABC analyst Achen. "What we saw this year is a sign that there's work to do." CBS analyst Lenski agrees VNS models "need to be adjusted to fully account for the intricacies of absentees and early voters."
"I'm one of the people who understands statistical models," says analyst Achen. "On election night, I can sit there as could other people on the team, and say, The computer is saying this, but the way the model works, that's not very trustworthy,' and that's what they hired us to do. The model makes certain assumptions that are true most of the time, but not always. The machine can say it's 99 percent likely and it really is, or it says it's 99 percent likely and that's not true. That's what happened at 8 with the Gore call."

ONCE A RACE IS CALLED, network election analysts put it aside and turn to other contests. Too much is happening to double-check a called race. Yet that's just what Goldstein did 15 minutes after ABC awarded Florida to Gore. Raw votes from Florida's 67 counties were pushing Bush ahead.
Goldstein happened to be watching the Florida Senate race screen when results from Duval County in the Jacksonville area came in. The data was odd. It showed Republican Rep. Bill McCollum gaining on Democrat Bill Nelson in the race for the Senate. Goldstein switched to the presidential screen. Gore was surging.
"That didn't make sense," says Goldstein. "A Republican closed the gap in the Senate and a Democrat widens in the presidential race. That tells you there's a serious problem. People don't vote like that."
CBS and CNN analysts Mitofsky and Lenski lost confidence in their Florida projection at about 9:15 p.m. "After we made the call, I was fine with it for the next hour," recalls Mitofsky. "Then it started getting suspicious."
Their decision screens showed numbers that didn't jibe for the northern Florida region. "We looked at all the counties in north Florida and saw Gore was getting 98 percent in Duval and Bush only getting 2 percent," says Lenski, executive vice president of Edison Media Research. "We called Murray."
At 9:07 p.m., a VNS operator had accidentally added an extra digit, pushing Gore's total in Duval County to 43,023 instead of the actual number, 4,302. The mistake was corrected by 9:38 p.m.
At 9:54 p.m., after watching the suspicious numbers for almost 45 minutes, CBS stripped Gore of his win and sent the race into the undecided column.
At 10:16 p.m., VNS sent a message to its members and 100 TV, radio and newspaper subscribers: "WE'RE RETRACTING OUR CALL IN FL BECAUSE WE DON'T HAVE OUR PREVIOUS CONFIDENCE." By then, everyone but NBC had already pulled Florida back.
"After VNS deleted the bad data," says Lenski, who's worked every election for CBS or VNS since 1988, "we realized it was dead even." They advised CBS polling chief Kathleen A. Frankovic to take Florida from Gore.
Word traveled from CBS News Executive Producer Al Ortiz into Dan Rather's earpiece. "Bulletin," sputtered Rather. "Florida pulled back into the undecided column. Computer and data problem. One of the CBS News election night headlines of the hour. This knock-down-drag-out battle drags on into the night, and turn the lights down, the party just got wilder."
Mitofsky hasn't made many wrong calls in 33 years. "They are embarrassing," he admits. "I'm chagrined by them. But if you're wrong, you're wrong."
But he had company: Everyone using VNS data had jumped the gun. It may have been a tad more embarrassing, though, for Mitofsky, since Rather had earlier assured viewers that they had settled on the reliable channel.
"Let's get one thing straight right from the get-go," said Rather at 7 p.m. "We would rather be last in reporting a return than to be wrong. And again, our record demonstrates that to be true. If you hear someplace else that somebody's carried a state--and if you are off, as you shouldn't be, watching them, then come back here because if we say somebody's carried a state, you can pretty much take it to the bank, book it, that that's true."
Fifty minutes later, the promise wasn't worth much.

SO WHAT, EXACTLY, had gone wrong?
The bad Gore call was not because of flaws in the exit polls or a data entry error, despite dozens of inaccurate media reports to that effect. Experts agree there was no bias in the exit polls, as there had been in New Hampshire. And the Duval County mistake, made an hour after the Gore call, played no part in the blunder, VNS Editorial Director Murray Edelman explained in a confidential November 14 memo to members.
Nor were other errors to blame, such as one made by a VNS staffer who inaccurately recorded figures for Lake County at 9:01 p.m. and again at 10:47 p.m., coming up with totals larger than was possible. By 11:59 p.m., the errors were corrected.
Nor was it due to a VNS operator shortchanging Gore by 4,000 votes in Brevard County, punching in 93,318 instead of 97,318 at
10:13 p.m., though that error may have played a role later in the evening since it wasn't corrected until 3:51 a.m, according to a VNS memo.
"I still believe the biggest problem in the model is that we did not correctly anticipate the impact of the absentee vote," Edelman wrote in the memo.
Edelman declined to be interviewed by AJR for this article. The network consortium will not allow any VNS employees to speak to the media about what happened on election night. And network officials also will not discuss the situation, beyond saying that they are investigating what went wrong.
University of Wisconsin political scientist Ken Goldstein is not surprised that absentee ballots played a key role in a bum call. Exit polls, while painting a portrait of Florida voters who cast ballots at precincts, tell analysts nothing about absentee voters. "One of the things you are looking for in a close race is you want to be sure you know what's been going on with the absentee ballots and that you are counting them properly," says ABC election analyst Paul Freedman.
But Goldstein and others say Edelman was cognizant of the absentee factor going in.
"If you don't include absentee votes in your model, you are going to be off," says Goldstein. "Everyone knew that Florida is 10 percent absentee and has always been. We got memos ahead of time that said, Don't forget about absentee.' Murray produced a lot of paper before the election on absentee voters. We had Excel spreadsheets on absentees, plus we had reports for each state on absentee history. Yeah, the absentees could be the explanation, but you knew about that."

AACADEMICS, STATISTICIANS AND news people working election night know uncertainty goes hand-in-hand with predictions. They expect and try to compensate for bad data or human error. Making projections is a science. But it is not foolproof.
What does ABC do when everyone else has picked a winner and ABC hasn't? "There's got to be social pressure," says the University of North Carolina's Meyer. "Who wants to watch an anchor who doesn't know who won when everybody else says they know?"
Says Freedman, "As long as the context is calling the winner first, you are going to see built-in incentives to be a little risky." Given the intensely competitive environment, it's reasonable "to expect that people will make decisions with competition in mind and not simply with the data in mind," he adds. "It's inevitable. It's built into the way the networks structure election night."
But why are the networks in the projection business at all? Who cares which network is first? Most people merely want accurate results. Many critics--among them election expert Curtis Gans--argue the networks do democracy a disservice by declaring a winner before the final results are in. "In almost every election there's this rush to judgment," says Gans, vice president and director of the Committee for the Study of the American Electorate, "and there's something inaccurate reported. Networks are creating the news by projecting winners, not reporting it. No data is as accurate as tabulated results."
A poll by the Pew Research Center For The People & The Press reported nine days after the election that 87 percent of the American people want the networks to stop predicting winners before the votes have been tallied. "There are a lot of people out there who are fed up with the networks," Gans says.

WHAT HAPPENED IN THE wee hours of the morning after changed--perhaps forever, but certainly for many years to come--how much Americans trust television networks. What will they believe in 2004 when a network projects a winner in Florida?
"What everybody is going to remember is not the campaign coverage but the election night coverage," says S. Robert Lichter, president of the Center for Media and Public Affairs. "So even if we could give [the networks] a brilliant grade, I think it would be a little like saying that the Titanic was doing just fine except for the iceberg."
Almost 41 million households were tuned to the four major broadcast networks and three cable news channels on election night, according to Nielsen Media Research. No waiting for tomorrow's paper. No competing to get on a crowded Web site. And TV graphics beat radio. Turn on the TV and Jennings and Rather and Tom Brokaw and the cable guys are there for reliable updates.
By 1 a.m., it was clear that whoever carried Florida would move into the White House in January 2001. As predicted, election night 2000 had been a wild ride. Now, all that had to happen was for someone to win Florida. Then all of those viewers struggling to keep their eyes open could finally go to bed.
At 1:30 a.m., everyone staring at election data screens--at the networks, at the AP, at VNS--was jumpy. Ninety-five percent of Florida's precincts had been tabulated, according to VNS, and there was still no clear winner. At 1:30 a.m., Bush had about a 60,000-vote lead, says AP Florida Bureau Chief Kevin Walsh. But how many votes were out? Where were they concentrated?
At 1:52 a.m., Walsh's figures showed the lead shrinking to 56,000 Bush. "It was very intense," recalls Walsh. "We were getting calls here and in Tallahassee from editors all over the country wondering if we were going to call the race, or looking for guidance."
At 2:10 a.m. in Fox's New York studio, election analyst John Ellis, George W. Bush's first cousin, saw the same VNS data as those at the other networks. It indicated Bush held a 51,433-vote edge. Only 179,713 votes were outstanding, according to VNS. (It would turn out that VNS had grossly underestimated that number. Instead, twice as many votes remained uncounted, according to a VNS post-election memo.) To close the gap, Gore needed to win 63 percent of them. It seemed an impossible task. Ellis advised the network to call the state for his relative.
NBC decision desk head Sheldon Gawiser was worried, too. On the phone with Edelman, Gawiser repeatedly asked why NBC couldn't make the call. "When he couldn't give me any reason not to call the election," says Gawiser, "then I told him I was going to go ahead and take a look at it myself. And we then broke the connection."
Fox anchor Brit Hume declared George W. Bush winner of Florida's 25 electoral votes and crowned him the next president at 2:16 a.m. One after the other, the networks tumbled, like so many dominoes--CBS and NBC at 2:17 a.m., CNN at 2:18 a.m.
Before 2:00 a.m., the ABC decision desk asked Achen about making the call. Again, he said it was too soon. The VNS model couldn't assure him the absentee votes had been counted, and he knew that at the end of a long night, errors in totals were likely. But Achen says he didn't lobby strenuously against the decision.
At 2:20 a.m.--again last--ABC gave Florida and the election to Bush.
Ironically, VNS itself did not award the state. But CBS' and CNN's Mitofsky says with 97 percent of the vote reported, Bush leading by 50,000 and about 180,000 votes not yet tallied, he thought it was an easy call.

BUT THE AP HESITATED. At 2:16 a.m., with 99 percent of the state's 5,884 precincts in, Walsh's count showed the Bush margin narrowing to 30,000 votes. "We didn't feel there was any way we could make that call," says Sandy Johnson, AP's Washington bureau chief and the one making final decisions on all of the races. "It just wasn't there."
As CNN anchor Bernard Shaw declared Bush the winner at 2:18 a.m., Republican operatives projected Shaw on a drive-in movie-size screen outside of the Texas State Capitol. The Republican crowd in Austin erupted in cheers. In Nashville, Gore supporters standing in the rain had another reason to feel glum. Soon, they were told, their man would arrive to give a concession speech.
CBS' Rather was awash in Ratherisms. "Let's give a tip of the Stetson to the loser, Vice President Al Gore, and at the same time, a big tip and a hip, hip, hurrah and a great big Texas howdy to the new president of the United States. Sip it. Savor it. Cup it. Photostat it. Underline it in red. Press it in a book. Put it in an album. Hang it on the wall. George Bush is the next president of the United States."
CBS' screen flashed: "Bush elected president."
Rather and CBS stars Ed Bradley, Lesley Stahl and Bob Schieffer began speculating on why Gore lost, what he would do next, whether he could have run a better race.
Still the AP remained silent.
Across the country, newspapers were bumping up against their final deadlines. Editors were edgy. Dozens of the AP's 1,500 daily newspaper members began calling New York, Washington, Miami. Why hasn't the AP called the race? In New York City, the AP's president and CEO, Louis D. Boccardi, called his D.C. team at 2:30 a.m. He, too, wanted to know what was going on. So did Executive Editor Wolman, who was in D.C., although Bureau Chief Johnson was honchoing election coverage. "I had to wonder if there was an overabundance of caution because of the bad Gore call," Wolman says.
Yes, the AP team was wary. It couldn't afford to blow it, not with all of those newspapers depending on it. But Johnson had confidence in an invaluable tool the AP had that the networks didn't: a backup system. Long before the AP joined VNS in 1990, it had cobbled together its own network for gathering election results.
"We count the vote in every state," says Tom Jory, the AP's liaison to VNS. "We count the same national races VNS counts as an independent backup to VNS and use VNS to edit-check our report. We often use VNS as a primary source, but we weren't using VNS in Florida."
Like VNS, the AP hires freelancers to gather vote counts. The freelancers are strongly encouraged during dress rehearsals the weekend before election night to call bureaus early and often. "As a result," says Walsh, "we were consistently out front of VNS and out front of the Florida secretary of state's Web site."
When VNS models showed Bush's lead jump from 29,386 at 2:05 a.m. to 51,433 five minutes later, network analysts thought the race was over. But AP's count was radically different. It showed that Bush's lead at 1:47 a.m. was 56,486. By 2:16, it had plummeted to 30,513.
The AP decided it wasn't ready to call the race. At about 2:30 a.m., seeking reassurance, AP decision editors in D.C. summoned Will Lester, who for 11 years had participated in the wire service's coverage of Florida elections as a writer and an editor.
"You didn't have to sit down and do fancy math to figure out that if Gore could win a substantial share of the outstanding votes, it was clear that the margin Bush had could evaporate," says Lester, the AP's poll writer. "Broward and Palm Beach are big metropolitan counties. Those counties coming in late definitely could wipe out a Republican lead."
"Can we call it?" Johnson asked Lester. "We need to know when we can call it." Lester eyed the data again. "You can't do it," he responded. At 2:30 a.m., as Rather deconstructed Gore's downfall, the AP's numbers showed Bush's edge dropping to 19,000.
Wolman was feeling enormous pressure. "After the celebration started in Austin," says the AP executive, "we spent all our time crashing numbers on calculators, trying to understand whether the race was over or was it possible for the Democrats to catch up?"
At 2:37 a.m., Johnson and others concluded Gore could pull it out. The wire service issued an urgent update, cautioning that while the networks were calling the race, there was "the narrowest of margins" between the candidates with votes still being tallied.
But everyone at the AP's Washington bureau was extremely nervous. "You have the weight of the other VNS members making the call," says Wolman. "By that time newspaper Web sites were printing 'Bush Wins.' Gore had conceded to Bush. There was doom and gloom in Nashville. We felt extremely lonely. We were thinking, 'Florida's the whole ball game. Don't blow it.' "
Many were thinking the same thing. "That would be something if the networks blow it twice in one night," said NBC's Tom Brokaw less than an hour before the network took Florida away from Bush at 4:00 a.m.
They did.
Could the debacle have been avoided if the networks had subscribed to the AP's special election-night wire, which flashed vote totals up to 12 times an hour, as much as four times as frequently as the main wire? None of them did.
"I wish we had," Lenski says. "At 3 a.m., we went into Yahoo! to find AP vote totals."

SOME NETWORK ANALYSTS explain the second blunder by blaming a computer glitch in one machine in one tiny precinct in Volusia County, which includes Daytona Beach.
In precinct 216 in DeLand, election workers began having trouble with one voting machine shortly before the polls closed. After 7 p.m., a poll worker drove the malfunctioning laptop-size machine to the Department of Elections to see if someone could get it to work.
"We would have had to remove the memory card and upload it ourselves, which is what we did at 10:02 p.m.," recalls Denise Hansen, the department's assistant supervisor.
Within 15 minutes, she says, the county attorney called. He'd been watching results stream in on a screen in the county's council chambers. Gore's vote was going backwards!
By 10:30 p.m., the county knew there was an error. But where? Officials tried for hours to track down the AccuVote machine vendor in Texas. After running a precinct-by-precinct printout at 1:24 a.m., the error jumped out. "You don't see a negative 16,000 votes for anybody," says Hansen. Not to mention a five-digit vote in a precinct with 585 registered voters--only 219 of whom had voted.
When VNS entered the Volusia glitch at 2:08 a.m., Gore's count in the county dropped from 82,619 to 72,152. The problem should have been caught by a VNS operator. Its computers flag anomalies, alerting operators to potential data problems that need to be checked.
"When there's a big decrease in a candidate's vote of 10,000, it has to be approved by a manager at VNS," says Lenski. He adds that if they'd received "weird data" from Volusia, they would have taken a close look at the county's totals.
Says the AP's Walsh, "I think everyone was affected by that erroneous data. However, our vote totals in all the counties were far ahead enough of VNS' that it affected us less. We were recording votes so fast in the Miami bureau that the Volusia problem did not have as dramatic an impact on the overall Bush margin for us as it did for VNS."
AP didn't notify VNS because while it could tell something was wrong in Volusia, "we had no idea what the specific vote drop was," Walsh says.
What looked like a 56,000-vote Bush lead according to VNS was probably more like 30,000, says CBS' Mitofsky. "Now we always expected his lead to close up somewhat, because the missing votes were in Democratic areas. We certainly didn't think there were more than 180,000 votes out. If we had known that there were as many as 360,000 votes out, as we do now, we never would have made a projection." It's not clear what caused that miscalculation.
At 2:30 a.m., the AP's count gave Bush a 19,000-vote margin while VNS' margin was almost twice that. To the AP, with votes in Democratic strongholds outstanding, it appeared Gore could pull ahead. To those eyeing VNS data, Gore didn't stand a chance.
At 3:11 a.m., the AP sent out a cautionary advisory. Bush's lead had dwindled to 6,000.
On CBS, Ed Bradley was holding up the AP report. Rather turned and Bradley said slowly, "The Associated Press believes that the uncounted votes in Broward and Palm Beach counties could allow a change of the lead in the Florida vote." There was laughter, groaning, amazement on the set. "Hello, 911? Cardiac arrest unit please," said Rather, pretending to make a phone call. Bob Schieffer covered his face in disbelief.
Rather recovered. "Let's not joke about it folks," he advised. "You have known all night long and we've said to you all night long that these estimates of who wins and who loses are based on the best available information we have. CBS News has the best track record in the business, over a half century plus, for accuracy on election night. But nobody's perfect."

CONGRESS IS HOLDING HEARINGS on the election night fiasco. Networks are hiring outside consultants to study the VNS system. Already, ABC, NBC and Fox have promised they won't call a state until all of its polls have closed. The networks are once again lobbying for a uniform poll-closing time. ABC has announced it will no longer allow television sets in decision desk rooms. Critics are beseeching the networks to stop projecting winners based on exit polls or sample precincts.
After all of the investigations are over, and Congress has studied the misfire into oblivion, and network executives have pounded their chests with mea culpas, and anchors have promised to never, ever project winners in tight races, and television sets are forever banned from election night decision rooms, the best solution may be simply to force everyone involved to repeat this mantra 100 times:
"It's not an off-year dog catcher race in North Dakota."