|
<advertisement>
|
 |
 
|
|
media search: |
 |
January/February 2001 | Contents
THE BIG MISTAKE
BY NEIL HICKEY
Don't blame just the exit polls.
That's the beginning of wisdom in puzzling out how and why television's elections
analysts made such a stupefying hash -- the worst in their history -- of "calling"
the winner in the 2000 presidential contest. Florida's twenty-five electoral
votes had become the Grail. Assigning them conclusively to George W. Bush
required thirty-five days, rather than mere minutes or hours -- the usual
elapsed time for network voting wizards to proclaim a victor. Pinpointing
what went wrong troubled the hearts and minds of the general public as well
as those at Voter News Service and their members: the four major broadcast
networks, CNN, and The Associated Press. In one night, television's quick-draw
projections raised more questions about their accuracy and credibility than
they had in four decades of (mostly error-free) vote-calling.
First: a thumbnail timeline of November 7-8 (times approximate):
7 p.m. EST -- Polls close in most of Florida.
7:49-8 p.m. -- All six VNS members hand the presidency to Gore.
9:55 p.m. -- Networks begin retracting the Gore "call."
2:16 a.m. -- Fox News Channel projects Bush as the winner in Florida. Within minutes, CBS, ABC, NBC, and CNN follow suit. The Associated Press, the sixth member, holds back.
Near 4 a.m. -- Networks start retracting the Bush "call," declaring the presidential race in limbo.
Dan Rather's dramatic, middle-of-the-night announcement of the Bush "win" is now part of election night legend: ". . . a hip-hip hooray and a big Texas howdy to the new president . . . . 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 W. Bush is the next president of the United States." Upon recanting that grandiloquence, Rather told viewers: "If you're disgusted with us, frankly I don't blame you." Five weeks later, Rather's anointment of Bush proved true, but by that time he and his counterparts at the other networks were seriously overdrawn at the credibility bank they'd carefully constructed over many years of vote projecting.
Among the nation's newspapers, television's declaration of a Bush victory -- plus a concession phone call (quickly retracted) to Bush by Gore -- triggered an unprecedented chain reaction: BUSH APPEARS TO DEFEAT GORE, declared The New York Times in about 115,000 copies that reached the streets. The St. Louis Post-Dispatch: BUSH WINS A THRILLER. The Boston Globe: IT'S BUSH IN A TIGHT ONE. The Miami Herald: BUSH WINS IT. Others dodged that bullet, among them the Los Angeles Times, Chicago Tribune, The Washington Post, Newsday. The Star Ledger in Newark produced a BUSH WINS headline, but then destroyed the evidence. "We went for the head fake," jokes Jim Willse, the editor. None of the papers left the building. "We were very meticulous. We unloaded the trucks, to the great amusement of the drivers, and trashed all the copies."
Never before had newspapers' near-total dependence on television's vote projections been laid so bare. TV's elections analysts with their sample precincts and illuminating exit polls had projected winners in thousands of races over the decades -- congressional, gubernatorial, presidential -- and been wrong in only a tiny fraction of them. Suddenly, on November 7, those systems melted down, leaving party activists, newspaper journalists, and the public demanding to know how the TV people had initiated such chaos.
Immediately after the election, however, VNS's board of managers -- composed of representatives of its six members -- hauled up the drawbridge, declining to answer all inquiries about what had gone wrong.
NBC and Fox threatened to quit VNS unless the service fully explained its massive election night errors and promised they wouldn't happen again.
VNS has sent a preliminary report to its members indicating that the mistaken Gore call was "a combination of many factors" -- among them, an unexpectedly large rise in the number of absentee Florida voters, which increased the potential for inaccurate projections. Also: VNS may have overestimated the size of the black vote and underestimated the size of the Cuban vote, both of which mistakes could have made things look better for Gore. The erroneous Bush call, the report shows, was based mostly on incorrect results coming out of Volusia County, as well as on an understatement of Palm Beach County's outstanding votes.
As is customary on election days, VNS fed its members three sources of information: exit polls, actual vote results from a sample of precincts, and unofficial county returns. VNS also supplies projections to the member organizations, each of which maintains its own independent decision desk and is individually responsible for the calls it makes. Under an agreement with Congress in 1985, networks call no winners in a state until the majority of its polls are closed.
Warren Mitofsky, a vote projection pioneer and an analyst on November 7 at the CBS/CNN joint decision desk, told CJR that, presented with the same early-evening data from VNS, he'd probably call the race for Gore again. "I don't know how I'd be smart enough not to. Every bit of information we had was clearly pointing to a Gore call." Only the exit poll data was available as the voting places closed, and that showed no clear winner. When the actual vote totals began coming in, all six VNS members gave Gore the nod.
So don't blame the exit polls, Mitofsky insists. "Everybody is dumping on them and that wasn't the source of the problem."
NBC's elections director, Sheldon Gawiser, agrees with that. "Nobody called it based on exit polls," he says. "The actual vote from the sample precincts showed an even bigger Gore lead than the exit polls did," so the call seemed like a "no brainer," but the data coming in from VNS was seriously wrong. "We don't know why. That's what they're trying to figure out."
Meanwhile, at AP's Washington election headquarters, executive editor Jonathan Wolman and his team were scrutinizing the same VNS information. "We were seeing on the projection screen a 7 percent lead for the vice president," Wolman says, as well as an indication from VNS that if the figure were wrong, it was likely to be slightly wrong in favor of Gore. "As actual votes were tabulated and weighted against other data, it was clear that Gore didn't enjoy any such margin." That insight came too late. AP joined the other VNS constituents in calling Gore the winner.
Exit polls were even less a source of the problem in the egregious projection of Bush as victor just after 2 a.m. With 97 percent of the precincts reporting, Bush enjoyed an apparent 51,000 vote lead with (supposedly) about 180,000 votes yet to be counted, Mitofsky recalls. "We were well aware of the absentee votes, and well aware of where votes were missing. We knew that much of the outstanding vote was going to come from Democratic areas in Broward and Palm Beach counties. At the time we made the call, we expected Bush's lead to drop to about thirty to 35,000. We found out later that there were close to 400,000 votes outstanding, not 180,000. There were errors in Volusia County and elsewhere that nobody told us about. You can't make correct calls if you're looking at wrong data." Volusia County was a morass of problems on election night, NBC's Gawiser remembers. At one point, county officials ran totals through their computer and concluded that the Socialist Workers party got 9,000 votes -- a virtual impossibility.
Over at AP, Wolman's analysts were trying to decide if Bush's apparent lead was impregnable. When the networks made Bush the winner, AP's team hurriedly began poring over county returns. "Our decision desk concluded that there were so many Democratic votes out there that Gore might be able to catch up," Wolman recalls.
Reuters moved a story at 2:31 that Bush was the new president. At 2:37, AP advised its clients in an urgent update that the race was not over. Then at 3:11, it moved a "cautionary advisory" that Bush's lead in Florida had shrunk to about 6,000 votes and that uncounted returns in two mostly Democratic counties could determine the outcome. Subsequently, in a "special message" to AP's newspaper and broadcast members, AP president Louis Boccardi built some space between his agency and the other five VNS partners. While acknowledging that AP did hand Florida to Gore early in the evening, he added that, on the Bush tally, "the pressure to join the parade was enormous, but AP people . . . held firm, to their great credit."
|
SOME
THAT CALLED THE
ELECTION PREMATURELY |
|
New
York Post |
A major irony of the election night debacle is that if the television networks had depended on exit polling alone to project a Gore victory in Florida, they never would have made that all-too conspicuous 8 p.m. goof. The reason: experts know that voting estimates based on exit surveys alone are notoriously undependable and are never used to make projections except when one candidate is clearly way ahead. But on November 7, exit polls were showing Gore with only a slight lead -- not nearly enough to make a call. If the TV people had stuck with that guidance -- rather than incorporating in their decision faulty vote totals from counties and sample precincts -- they'd never have risked making a projection. "In the food chain of culprits, I'd put exit polls at the bottom," says Martin Plissner, former executive political director of CBS News and for seven years a member of the VNS board of managers.
More fundamental, clearly, is the fact that VNS is the sole supplier of voting results to its members, the single source of shared information upon which projections are made. That offends a basic principle of good journalism that requires multiple sources before publishing or broadcasting a story.
A bit of history: before 1990, each of the TV networks had its own sample precincts and conducted its own polling on which it based projections. As costs rose, that was seen as an expensive, duplicative way to do business. The urge to pool election coverage as a money-saving tactic grew, and VNS emerged to handle the whole job. Suddenly, the decades-old, furious competition between networks to crush each other in calling races seemed over. The new structure worked well in the 1990 and 1992 elections, with VNS (It was then called VRS -- Voter Research & Surveys) making projections for everybody.
Then in 1994, ABC News -- much to the chagrin of the other VNS partners -- unilaterally fielded a team of statisticians and vote analysts, cherry-picked a number of hot races, and scooped the opposition with some eye-catching early projections. Feeling blind-sided by a sucker punch, the other networks vowed not to let ABC get away with it, and by 1996, everybody was back hustling to be first with individual projections based on VNS-supplied data.
The question then became, and remains: was VNS a sensible economy in the first place? Four or five networks reporting, polling, analyzing, and projecting elections give the public a richer universe of election night news than a one-source megalith. And, no matter the hot competition, networks traditionally, to their credit, have given higher priority to being right than being first. The big problem with VNS on this election night is that the VNS system had never before been tested in a close election. Previous presidential races in the VNS era were comparatively one-sided. "In this one," says Hal Bruno, a veteran elections analyst and former political director of ABC News, "the exit poll warned you during the day: watch out! And then the key precincts came in and there were more warning signs. What happened is that a lot of people drove through those warning signs that said 'dangerous curves ahead.'"
One idiosyncratic aspect of the exit polling in Florida probably distorted the survey results. Some interviewees, it appears, trustingly told exit pollsters they'd voted for Gore but actually had mistakenly voted for Pat Buchanan, or had marked ballots for both Gore and Buchanan, rendering them invalid. Kathy Frankovic, a CBS News producer and director of surveys, says: "There's clearly no correction for that. We have no idea of the magnitude of that error." Mitofsky and others feel that its effects were relatively minor.
According to a survey commissioned by The Miami Herald and published on December 2, the networks and AP were correct in calling Gore the winner -- even though they did so for invalid reasons. Gore would have won Florida by 23,000 votes, the study concluded, if 185,000 ballots -- discarded for their ambiguity -- had been analyzed for the intent of those voters. A number of media organizations are conducting ex post facto vote counts in Florida that may tell who really won the state.
The nation's newspapers are busily reassessing their reliance on television's reporting of presidential races. Editors standing in their newsrooms at 2 a.m. are straddling two important interests, says Rich Oppel, editor of the Austin American-Statesman and president of the American Society of Newspaper Editors: "The most important interest is being right. The second is not looking like you're yesterday's news." He doesn't plan to turn off the television sets in his newsroom on election nights. "Our colleagues in television are smart people, and worth watching; but a newspaper editor's operating principles need to be caution and restraint." John Walter, managing editor of the Atlanta Journal-Constitution (Its headline: bush is declared winner: florida vote turns the tide) agrees that TV is a necessary tool on election nights. The great lesson for Chris Peck, editor of the Spokane Spokesman-Review, is that newspapers need to build in systems to help insulate the print press from so heavy a reliance on television. To resist what he calls "the incredible bandwagon effect" of TV's hair-trigger projections, it's useful, he suggests, to have a contrarian in a corner of the newsroom, an old hand with institutional memory of local and national races who hoists red flags to stave off error.
Two TV-related events conjoined to lead newspapers over the cliff. First: the 2:16 projection for Bush by Fox News, echoed within four minutes by NBC, CBS, CNN, and ABC. (Fox consultant John Ellis, the governor's cousin, got credit for starting the stampede, a gesture that raised prickly questions about the journalistic propriety of a candidate's relative occupying so sensitive a post.) Second: TV reports that Gore had conceded in a phone call to Bush and was heading for the War Memorial in Nashville to make it official. Says Rich Oppel: "I should have relied on AP. But a seasoned editor, looking at those two factors together, has got to put a lot of weight on them. They pretty much ended the contest for most newspapers."
One big mystery for many outside observers was why the networks, having blown the projection massively at 8 p.m., felt compelled to risk blowing it again around 2 a.m. -- 11 p.m. in the west -- when much of the country was asleep. Assessing the cost versus the benefits, they had little to gain and a trove of credibility to lose. So why bother, except for vain bragging rights?
One such outside observer, Robert Lichter, president of the Center for Media and Public Affairs in Washington, says that TV's performance reminded him of a scene in Oscar Wilde's celebrated comedy The Importance of Being Earnest, in which a dowager, Lady Bracknell, is asked for her ward's hand in marriage by a young man who confesses he's an orphan. "To lose one parent may be regarded as a misfortune," she tells the young man haughtily. "To lose both seems like carelessness."
VNS and its constituents were both careless and unlucky in two unjustified projections. But as AP's Jonathan Wolman says, "Election work is an art, not a science; it's important to keep that in mind." Projection from samples is an exercise in probability. As precinct numbers, exit polls, and county totals are simmered in the same stew and then tasted, they produce increasing levels of probability -- never certainty -- that one candidate or the other has won.
"When you handle election data," says Warren Mitofsky, "one of the things you learn is that it's subject to mistakes." Veteran vote projectors know in their bones that when such mountainous volumes of numbers are processed on election night -- by county officials, VNS operators, network staffers -- errors inevitably creep in. People get weary. "It's just human," he says. "So you build in checks and balances. This system needs more checks and balances in the vote counting end."
It also needs an updating of VNS's software, according to Christopher Achen, professor of political science at the University of Michigan and a consultant to ABC News who was stationed at VNS's election headquarters in New York's World Trade Center on election night. One big problem for the networks, he claims, is that VNS's probability forecasts are imperfect because of shortcomings in the software -- producing occasional wrong conclusions in the exit polls that a human being would catch but a computer won't. The system, for example, takes exit polls at face value without correcting for a demographic bias that usually overrepresents Democrats. Why? No one is quite sure, Achen says. "Maybe because people with more education are more likely to fill out questionnaires, and those tend to be Democrats."
What's sure is that TV's election night practices are in for significant reupholstery well before the 2002 races. Several networks promise they'll project winners in the future only when all polls have closed in a state, not just a majority of them. ABC intends to advise viewers that projections are "informed, statistically based estimates" of the probable outcome of elections, not definitive declarations. They'll also remove television sets from the proximity of their decision desks so that analysts feel less pressured to make hasty calls.
Beyond that, legislators -- mostly in the person of congressman Billy Tauzin, Republican of Louisiana -- have been scrutinizing TV's election night performance. Tauzin says he won't sponsor any bill aimed at preventing exit polls or limiting vote projections -- legislation which, in any case, would clearly affront the First Amendment. He and a Democratic congressman, Ed Markey of Massachusetts, are introducing legislation to require the fifty states to close their polls at the same moment -- an often-proposed idea that would force drastic changes in the way TV news handles projections.
Despite the mistakes, gaffes, and embarrassments, or perhaps because of them, election night 2000 attracted the most households and viewers to TV screens since Nielsen began keeping such records with the 1960 Kennedy-Nixon cliffhanger. The late-night host Conan O'Brien joked that the networks were so thrilled with the ratings that they plan to call all elections incorrectly from now on.
The public's loss of trust in television news, however, was no laughing matter. In a CNN poll 79 percent of Americans said the networks did not act "responsibly" on election night. In future close elections, will most viewers believe what the networks tell them? How long will it take to regain their confidence? Why serve up quick-draw projections at all, since the public isn't clamoring for them? Is it really worth each network's paltry saving of $5-$10 million per election cycle to cede to a single entity so much influence and discretion? Or, contrarily, should the networks dismantle their individual decision desks and delegate a reconstituted, better funded VNS to make all projections, but in a more cautious, unhurried, less frenzied, and non-competitive mode?
VNS's preliminary report claims that budget limitations have prevented it from updating its systems and adding features that would make it more foolproof. And it hints that things worked better when VNS was making all the projections and the members weren't competing hotly against each other to be first with calls. During that time, the consortium's performance was error-free. "We need to revisit and rewrite many of our procedures," says the report, and promises that the internal investigation will continue.
A lot went wrong on November 7. "We were as good as the information we were getting from sources we trusted," CBS News president Andrew Heyward wrote to congressman Tauzin. "In this case, that information was not good, and neither were we."
That, too, is the beginning of wisdom.
Neil Hickey is CJR's editor at large.
