|
The Florida Flip-Flop and Weather ForecastsRoger A. Pielke, Jr. As the spectacle of naming the nation's 43rd president comes to its conclusion, it is time to consider some of the less-discussed lessons of the election. Specifically, the broadcast media's action on election night in first calling Florida for Gore, then for Bush, then for neither – the Florida Flip-Flop – provides two important lessons about election-outcome forecasts and our nation's democratic process. It also provides an opportunity for the weather community to share with those in the media responsible for covering politics some of the lessons that it has learned. First, the media need to more be more rigorous in their use of election-outcome forecasts. The Florida Flip-Flop is the media's most embarrassing moment in U.S. presidential election history since "Dewey defeats Truman," immortalized as a headline in the November 2, 1948, photograph of jubilant president-elect Harry S. Truman holding a copy of the Chicago Daily Tribune. But good came from the "Dewey defeats Truman" incident as it stimulated needed advances in polling and election-outcome forecasts. Indeed, academic political science developed a sub-field of polling and predicting elections using rigorous statistical methods. Based on this work, extrapolations from systematic surveys to election outcomes underlie the broadcast media's calls of state presidential election winners. Historically, such election-outcome forecasts are very accurate. While some might say that this track record is pretty good, this year's election clearly shows that it is not good enough. To ensure that these errors don't happen in the future, the media should use the tools at their disposal more appropriately – by subsuming the competitive incentives to be first in calling state election outcomes to a shared agreement on when such calls should be made. To get some help in establishing thresholds and implementing methods for assessing when those thresholds have been exceeded, the media should look to the experiences of the weather forecasting community. Weather forecasters have an impressive track record of assessing the skill of forecasts and working with the media to establish thresholds for action. The broadcast media would do well to apply these same methods and procedures to the forecasts they use on election night. Specifically, the tools of forecast verification developed by Allan Murphy and others provide a powerful set of tools for evaluating forecast "goodness." In an election context, the costs of a forecast "bust" are very high. Care should be taken such that the odds of a forecast error are miniscule. The weather community has tools that can help. Second, because election-outcome forecasts can affect politics, the media and/or politics must change. Unlike forecasts of tomorrow's weather, election-outcome forecasts can directly affect the processes being predicted. The major media wait – or are supposed to wait – until polls close in each state before calling the outcome, because if people knew who is likely to win in their state, they may be less likely to vote. In some circumstances this conceivably could change the election outcome. Imagine a nightmare not too different from this year's reality: What if the networks had waited several hours to reverse their call that Florida had gone to Gore and, as a consequence, voters in closely fought western states changed their behavior? This scenario could have led to a Gore victory, independent of the outcome in Florida, if a greater number of Bush voters – disillusioned by the apparent inevitability of a Gore presidency – chose not to vote, shifting several states to Gore that should have gone to Bush. Even though this did not occur, the current electoral mess was arguably exacerbated by the media's actions. There are two ways to deal with this situation. First option: change the media. If the media can agree on guidelines for issuing and conveying severe weather warnings in a manner that enhances rather than endangers public safety, surely they can also agree on ways to issue election-outcome forecasts that enhance rather than endanger democracy. In the case of weather forecasts, self-regulation is enhanced by the specter of potential legal liability for erroneous forecasts and also more-or-less generally accepted government policy for the provision of weather warnings. It is not obvious what specific rules make sense for media coverage of elections, particularly given the growth of the Internet, but collective attention to the process of election-outcome forecasts is now clearly needed. Again, the weather community's experiences, particularly in issuing hurricane and tornado warnings, can provide guidance. Second option: change politics. For example, states could agree to coordinate the electoral process by establishing a national 24-hour period to vote that would begin and end at the exact same time for each state. This might change the incentives for the media to call individual states as polls close one after another. There will be no shortage of proposals for election reform in the aftermath of this election as governments and academics will focus considerable attention on the Electoral College, voting mechanisms and procedures. In such retrospectives it will be important to ensure that election-outcome forecasts are in concert with the political process. Like a hurricane that approaches the coast, revealing massive snafus in evacuation procedures, the 2000 election is a warning that our democracy does not benefit automatically by advances in political science and related technologies; it also depends on how we use the resulting knowledge and tools. Just as "Dewey defeats Truman" stimulated great improvements in the quality of political predictions, the Florida Flip-Flop should motivate improved use of election-outcome forecasts in our democratic process. Fortunately, there is a valuable body of expertise in the use (and misuse) of forecasts right here in the weather community. — Roger A. Pielke, Jr. Comments? thunder@ucar.edu WeatherZine #25 Home Page
| Comments and Feedback
| Site Map |
