Idiot Boxed

There was plenty of humiliation to go around in the aftermath of the 2000 elections. Vote counters and ballot designers, election boards and state legislatures all came in for heavy criticism. But special ignominy was reserved for the five major broadcast and cable networks and their news operations. The networks that night broadcast multiple incorrect reports, including the premature and still-disputed claim -- initiated by FOX at 2:15 a.m. -- that George W. Bush had won the state of Florida. “It was the most embarrassing evening in the history of network TV, politically,” says Tom Rosenstiel, director of the Project for Excellence in Journalism. “They just embarrassed themselves and failed in the basic public-interest obligation they have come to have over the years.”

The great sin of election night 2000, in the view of Rosenstiel and others, was that the television networks had broken one of the cardinal rules of journalism: Never rely on a single source of information when a second is available. That night, two organizations, The Associated Press (AP) and the Voter News Service (VNS), were tabulating vote results across the country. The AP was producing generally accurate numbers, but the network decision desks -- the people who decide when to declare winners in different races -- were looking only at the VNS data. After all, the VNS was theirs: It was the consortium into which the networks had all pooled money after they abandoned their own separate vote-tabulation projects as too expensive the decade before. Unfortunately, the VNS data were badly flawed. It was the VNS numbers that led FOX -- whose data-analysis division that night was headed, incredibly (or incredibly for any place but FOX), by a first cousin of George W. Bush named John Ellis -- to call Florida, and thus the presidency, for Bush. The other networks quickly followed suit. The call was retracted hours later, but not until long after it had registered its impact: For the next 35 tortuous days, Bush was the presumed president-elect, and Al Gore the presumed sore loser.

The networks engaged in much public hand-wringing in the weeks that followed the elections. Anyone who remembers a rueful Dan Rather saying, “If you're disgusted with us, frankly I don't blame you,” or who saw network executives go to Capitol Hill to explain themselves to Congress could reasonably have assumed that steps would be taken to ensure that such a debacle would not happen again. Network executives later acknowledged that cross-checking the VNS data with the AP's numbers would probably have prevented the incorrect call of Florida for Bush, and they promised to make certain that a single source with bad information could never again create a media-wide failure like election night 2000.

So here we are, just weeks away from November 2, 2004, and the situation is this: On election night, once again, the networks will be relying on a single source for vote tallies.

Instead of making certain that their decision desks will have two independent sources of information for vote tallies this year, the networks, in an effort to control costs, have completely eliminated one of the two vote-counting organizations that was operating in 2000. The result is that rather than fixing the system, the networks have hardwired it to produce a single source of vote tallies, ensuring that they will face the same situation that caused the largest of the election-night errors -- the “most embarrassing evening in the history of network TV, politically” -- four years ago.

* * *

On election night 2000, the VNS ran a vote-tabulation operation that focused on “top-of-the-ticket” races: the presidency, governorships, and House and Senate contests. The AP ran a parallel vote count that included all the races covered by the VNS as well as state legislative races and some local elections.

In the hour leading up to the networks' unanimous declaration that Bush had won Florida and the general election, the VNS and AP vote counts began to disagree wildly. As late as 1:02 a.m., the AP's data had indicated an almost 113,000-vote lead for Bush. But from that point on, the margin began to narrow quickly and dramatically. By 1:12 a.m., the Bush lead had fallen to 82,000 votes. Ten minutes later, it stood at 63,000. By 2:02 a.m., it had dropped to 56,000.

The VNS data, meanwhile, were telling a very different story. At 2 a.m., VNS data indicated that, with 96 percent of precincts reporting, Bush led by about 29,000 votes. Five minutes later, the Bush lead jumped to 51,000. The 22,000-vote swing would later be found to have resulted from a data-input error in Volusia County, which actually subtracted votes from Gore's total and added votes to Bush's.

At 2:12 a.m., the AP count dropped the Bush margin to 48,000 votes, many of them, according to the AP, in the Democratic southern part of the state; and with more than 200,000 votes still uncounted, a Gore victory was still possible. But the VNS held steady, now reporting a 51,433-vote edge for Bush and estimating that fewer than 180,000 votes remained uncounted. So at the exact moment that Bush's lead was dwindling according to the AP's tabulations, the networks were preparing to call the state for Bush. At 2:15 a.m., FOX News went on the air and declared that Texas Governor George W. Bush had defeated Vice President Al Gore by a slim margin in Florida, winning the state's coveted electoral votes and thus the presidency.

One minute later, the AP discovered the error in Volusia County and corrected it, resulting in the Bush lead declining further still, to 30,513 votes. This difference was now within the margin of error of the networks' statistical models, making any projection very dangerous. But in a headlong rush to follow FOX, the other TV networks weren't looking at the AP data, and they fell in line. With the VNS data still uncorrected, NBC called the race for Bush at 2:16, CBS and CNN (which were sharing a decision desk) called it at 2:18, and ABC called it at 2:20.

At 2:22 the AP again revised its numbers, and the Bush lead fell to 15,359 votes. The VNS continued to show Bush with a much larger lead, failing to correct for the Volusia County error until 2:51 a.m.

The concern that pressure not to fall behind their competition was responsible for the other networks calling the race immediately after FOX did was brushed aside by the networks, most of which claim to this day that when FOX made the call, they were within minutes of doing the same themselves. “You were keeping track of what other [networks] were doing,” says Dan Merkle, who worked at the ABC decision desk in 2000 and will be its director this November. “But looking at the numbers, it was pretty close. If FOX hadn't done it somebody else would have, maybe a few minutes later.”

Indeed, to many network officials, the question of competitive pressure on election night 2000 seems overblown. Linda Mason, CBS vice president for public affairs, believes the fact that nobody on the CBS/CNN decision desk checked the AP's numbers is far more bothersome. “I am not sure why they didn't use [them],” Mason told me. “They were using the tried and true numbers that they had used in the past, and they thought they didn't need anything else.

* * *

Network executives expressed contrition during congressional hearings that followed the election. In February 2001, CBS News President Andrew Heyward sat beside his counterparts at hearings before the House Energy and Commerce Committee and told lawmakers, “CBS News and the other network news operations made very, very serious mistakes that night, and they are mistakes that all of us at the table, and certainly I, deeply regret.” Roger Ailes, chairman and ceo of FOX News, blamed incorrect data from the VNS. “As FOX relied on those numbers,” he testified, “we gave our audience bad information. Our lengthy and critical self-examination shows that we let our viewers down. I apologize for making those bad projections that night.” He concluded, “It will not happen again.”

The networks all promised to institute rules for their election coverage that would reduce the possibility of bad calls in the future. They pledged to take steps to insulate their decision desks from the pressure to call races sooner than they want to just because another network has already done so. And one by one, the network chiefs also assured Congress, either explicitly or implicitly, that in the future they would use more than one source to gather vote-tabulation data. “We also need to make certain that we use supplementary sources of data as a way to verify the accuracy of the information we get,” said NBC President Andrew Lack.

They may have meant it at the time, but after another spectacular failure by the VNS in the 2002 election, when a computer crash rendered all of its exit-poll data unusable, the equation changed. The networks collectively abandoned the VNS and created a new entity: the National Election Pool (NEP). Made up of the five major networks and The Associated Press, the NEP provides its members, as well as other news organizations that subscribe to its service, with exit-poll data and vote counts on election day via a secure Internet connection. The NEP is, in effect, a sort of shell of the old VNS into which new exit-polling and vote-counting operations have been inserted. Rather than try to reconstruct the VNS exit-polling operation, the NEP hired veteran exit-polling specialists Warren Mitofsky, of Mitofsky International, and Joe Lenski, of Edison Media Services, to collect the data for the coalition. The NEP also chose to abandon the old VNS vote-counting structure in favor of contracting with the AP to provide tabulations, throwing the possibility of a two-source system out the window. The decision, many network officials said, was largely financial.

“I think we made a choice between trying to finance two discrete systems completely independent of one another and investing heavily in a single system with redundancies, quality controls, and built-in system checks,” says CNN political director Tom Hannon. “We came to the conclusion that we got the more reliable product, instead of investing in two lesser models, by combining them into a single system.”

The explanations have not satisfied many longtime journalists, who disdain the networks' decision to go with a single source. “It must be recognized that networks will not spend the money necessary to follow one of the most basic rules of reporting: two sources. That's a business decision, not a journalistic one,” says Joan Konner, professor and dean emerita of Columbia University's journalism school and one of the co-authors of a post-election review of coverage commissioned by CNN. “The networks say it's too expensive, but when you look at the size of their parent corporations, it's chump change,” adds Philip Meyer, Knight Professor of Journalism at the University of North Carolina at Chapel Hill and a Pulitzer Prize–winning reporter. “When the networks cover election night, having different organizations doing it with different methods, we always had some reason to be suspicious if they came out with different results. There was always a check on everyone—they checked each other. Having a single source for information is not good in a democracy.”

* * *

At first, says Linda Mason of CBS, now a member of the NEP's board of directors, the AP was “a little reluctant” to be the only source of vote counts. The wire service considered creating an entirely separate system for collecting vote tallies, in order to provide a check on the possibility of input errors, but found it was too costly. “It's extremely expensive,” said Tom Jory, the AP's director of election tabulations. “There are 4,600 counties, which means 4,600 people out there on election night. To double that, not only is it extremely expensive but it really doesn't buy you that much.”

Instead, the AP has increased its investment in its existing vote-tabulation infrastructure, creating redundant systems centered in New Jersey and Missouri. It has built-in, sophisticated statistical-analysis processes designed to detect and flag possible errors. In addition, editors familiar with each state's history and politics will be assigned to monitor the vote-tabulation results, and to be on the lookout for anomalies that aren't caught by the system. Jory said the AP is well aware that it will be operating “without a safety net” on election night. “Obviously, anytime you are in a position like this you are apprehensive, but we are confident that we can do it,” he insists.

It's true, of course, that it was the AP that had it right about Florida last time. But there are reasons why that fact should not translate into complete confidence in the system that the wire service has set up for this November. The AP's two redundant systems will still be working from the same inputs, meaning that if bad data should somehow get into the tabulations, that redundancy won't be much help -- and might even hurt by appearing to validate incorrect information.

Consider the not-so-remote possibility that the 2004 presidential election hinges on the result in a single state, with an electorate closely divided between Democrats and Republicans. Say, for the sake of argu- ment, that it's Florida.

The most favorable re-count scenario for Bush in the aftermath of the 2000 election found that his final margin of victory over Gore in Florida was 537 votes. Multiple other re-counts found smaller margins or determined that Gore, in fact, received more votes than Bush. In any case, the difference between the candidates in Florida, as a percentage of the 5.96 million votes cast, was so small as to be statistically insignificant.

The safeguards put in place by the AP are statistics-based systems. A vote total from a particular precinct that is significantly different, in a statistical sense, from totals in previous elections or from results predicted will raise red flags. For instance, the Volusia County error that caused confusion on election night would be flagged immediately by such a system. But errors of smaller magnitude would not. And in 2004, as in 2000, errors of small magnitude may be more than enough to create the temporary impression that the wrong candidate has won.

The temporary impression of victory in 2000 proved invaluable to George W. Bush, constituting the greatest windfall for a candidate from a few moments of television in American political history. The short-term anointing of Bush created what political consultants call a “frame” -- basically a starting set of assumptions for the public debate that would follow. An independent report commissioned by CNN in the weeks after the election described it this way: “The unanimous network calls for Bush created a premature impression that Bush was the winner in Florida. That characterization carried through the post-election challenge. Gore was perceived as the challenger and labeled a ‘sore loser' for trying to steal the election.”

Political campaigns spend millions of dollars on advertising, and an equally impressive amount on political operatives, to try to frame issues to their advantage. But in the 2000 presidential election, the Bush campaign got the frame of its dreams, and it got it for free. Right down to Dan Rather advising CBS viewers, “Hang it on the wall. George W. Bush is the next president of the United States,” the Bush campaign could not have asked for anything more.

* * *

There do exist reasons to be hopeful that the networks will be more cautious in November than they were four years ago, and that an early call by FOX won't set the dominoes tumbling the way it did in 2000. While none of the network officials interviewed for this story would specifically point to FOX as being a weak link in the chain, it seems clear that they all view its decisions as suspect. Several noted that during the 2004 Democratic primaries, FOX seemed most eager to push the envelope in terms of predicting winners early, often leaving the other networks shaking their heads at the willingness to risk another wrong call so soon after 2000. (FOX officials refused multiple requests to provide comments for this story.)

Second, during the Democratic primaries, there were often significant lags between the various networks' projections of winners in some states. One of the most significant was on Super Tuesday, when FOX made an early projection that John Kerry would carry Georgia. “The numbers seemed to indicate that ultimately Kerry [would] be the victor, but that there was a reasonable chance that [John] Edwards would win,” recalls CNN political director Tom Hannon. “So, under the circumstances, we were not going to make that call when there was a likelihood that it would be wrong.” So, the networks vow, they've learned their lesson. “I think it is fair to say that the primaries indicated, and the general election will indicate, that everyone -- or most everyone, at least -- is going to be extra cautious in states where the vote is close,” says NBC Vice President Bill Wheatley.

This sounds like good news to many who criticized the networks' performance in the aftermath of the 2000 election. But the primaries aren't the presidential election, and it's fair to ask if, when the pressure is truly on the decision desks, the networks will really be able to resist making a call when one of their competitors has already done so -- using bad information, hasty analysis, or their own ideological preference.

This has got to be avoided, not only because the networks played a role in easing the current occupant of the White House into office four years ago but because what remains of the public trust of the news media could be irreparably shattered by a recurrence. The networks need to recognize that the “disgust” that Dan Rather worried about in 2000 would be transformed into something of a completely different magnitude if they inject themselves into the electoral process again this November.

Rob Garver is a journalist who lives in Springfield, Virginia.