By Philip Meyer 

It's time to break up the exit-poll cartel. 

Only three presidential elections ago, in 1988, each of the older networks - ABC, CBS and NBC - had its independent exit-poll operation. Sometimes they produced different results. 

For the networks, those differences, when they occurred, were embarrassments. For the public, they were a healthy reminder that polling, especially under time pressure, can be an inexact science. Seeing lots of polls with different outcomes kept us from trusting any one poll too much. 

Tuesday night, we trusted too much. The networks called the key state of Florida for Al Gore all at about the same time and had to turn around and uncall it two hours later. All were looking at the same set of data from Voter News Service (VNS). 

If we were required to have only one exit-polling operation, we couldn't do better than the one headed by Murray Edelman, who is also the president of the American Association for Public Opinion Research. He runs a taut, buttoned-down shop. 

Edelman describes in great detail his work in the previous presidential election in a chapter of Election Polls, the News Media, and Democracy, edited by Paul J. Lavrakas and Michael W. Traugott. He recalls how four years ago, when Bill Clinton ran against Bob Dole, VNS spun out national election projections starting at 1:26 p.m. on Election Day, with eight updates until 2:40 a.m. Not one of those estimates was more than two percentage points from the final national result. Its 8 p.m. call was exact. 

But something happened Tuesday night that made everyone pull back in Florida. One of exit polling's strengths is that the results can be checked against the official count in the precincts covered by the interviewers. In Duval County, where Jacksonville is located, and in the Tampa area, numbers looked wrong. When the problem is dissected - and it will be - there will be plenty of blame to go around. Both the components of VNS' projection model and the networks' competitive haste are suspect. 

In some countries, a discrepancy between a poll and an election result makes citizens suspect the election results. In 1988, when Miguel Basáñez conducted the first presidential election poll in Mexico and published the results, the ruling party announced a failure in the vote-counting system on election night. 

The official results were not released until a week later, and the Partido Revolucionario Institucional (PRI) fell to 50% instead of the 80%-90% it had reported in past elections. The poll, Basáñez believes, forced an honest count. That was the beginning of the end for the PRI's 71-year rule in Mexico, which finally came to a close in a peaceful election last July. 

Instead of detection of fraud, the primary reason for the creation of VNS was to save money and use some of that savings for a higher-quality product. It was founded in 1990 by ABC, CBS, NBC and CNN as Voter Research and Surveys (VRS). Three years later, VRS merged with the News Election Service, whose specialty was high-speed collection of official election returns. The result was VNS, and it is the pooled effort of the original four networks plus Fox and The Associated Press. 

It is the combination of fast relays of official returns and exit-poll data that make the broadcasters' election-night projections so spookily fast and accurate. But these advantages come at the cost of competition and the safeguards that it brings. Consider what Tuesday night might have been like if the six news operations each had its own exit-poll operation. 

With competition, we might not have needed to wait for the Duval County numbers to turn bad on Tuesday night and put Florida back in the undecided column. If the state were really too close to call, the networks using the other exit-poll services would have been saying so. The one with the questionable numbers might have been motivated to start searching for them earlier. 

And if there were systematic fraud at the polls, the presence of multiple exit polls, using different precincts and different methods, would detect it in a more convincing way than a single poll would. 

Even with the cartel, there is competition at the level of analysis and news judgment. As he prepared his audience for the first poll closing on Tuesday night, Dan Rather warned that CBS might not be the first to call a winner in each state, but that it would strive to be the most accurate. Each of the six news organization has to make an independent judgment about the tradeoff between speed and accuracy. 

But competition at that level is meaningless if all of the news organizations are working from the same bad data. In time, Edelman and his crew will have to explain what went wrong - whether it was clerical error, mathematical miscalculation or something that went wrong with the official vote count itself. 

It's true that if we had had six data-collection agencies in the field Tuesday instead of one, we might be even more confused than we are now. But we'd also be more careful about where we put our trust. The lesson of the 2000 election is that information monopolies are inherently unhealthy. VNS should disband voluntarily. 

Philip Meyer, who holds the Knight Chair in Journalism at the University of North Carolina, Chapel Hill, also is a consultant for USA TODAY and member of the newspaper's board of contributors. He is the co-author of a chapter in Election Polls, the News Media, and Democracy.