A Note on 'Projected' Winners

NPR's Ron Elving, senior Washington editor, explains the process by which NPR and other news organizations will "project" winners of the 2004 election.

While America votes, thousands of poll workers are preparing to count the votes. If all goes as planned, or if even most goes as planned, the vote will be counted faster than ever. But there are also thousands of people working to report the results even faster -- and to report it hours or even days before the official count. These are the exit pollsters and the stringers for the Associated Press who will relay information from the polls and tally points around the country as fast as possible so that analysts at the nation's major national news organizations can project results by state.

Exit polls are done at selected precincts that have a history of predicting outcomes in key locations in each state. The pollsters ask people leaving the polls to share information about themselves and their vote anonymously. This information is collected in waves throughout the voting day. It is fed into computers and transmitted to the broadcast and cable networks and the Associated Press (all private news organizations that are partners in the enterprise). Each news organization then analyzes the data, using its own computer models and experts. The experts include both their regular political journalists and election professionals on contract.

As the evening goes on, the exit polls are supplemented by early returns of actual votes tabulated by the Associated Press, which employs 5,000 people to gather this data on this one night.

These sources of information are aggregated by the people and machines working at each national news organization. Each of these operations then attempts to project winners in each of the 50 states, to be announced as soon as the polls in that state close. Where the winner is not clear, each individual news organization will choose when to project a winner in that state. When each news organization has called enough states to give one candidate 270 electoral votes, it will project that candidate the overall winner.

In the past, breakneck competition between broadcasters has tempted the prognosticators to guess as soon as they feel reasonably sure. This has led to very early projected winners in the elections of 1980 and 1984 and 1996 -- when the suspense was over early even on the East Coast. These state projections have almost never been wrong. But the potential for error was always present.

In 1996, New Hampshire was projected for President Clinton by mistake (vote totals from one precinct were misreported) and then corrected. That did not matter in the overall result, but a far more significant error occurred in 2000. In Florida, exit polls pointed to a win for Democrat Al Gore but were not sufficient to project him the winner. Nonetheless, overeagerness and overconfidence led to a premature call for Democrat Al Gore in that state by several news organizations (including NPR). This was reversed and became a premature call for George Bush, which led in turn to the premature judgment that Bush had won overall. In fact, Florida's electoral votes were not finally awarded to Bush until five weeks later. As a result, this year, all news organizations involved have promised to be more cautious.