Thursday, September 17, 2020

Kick the Tires and Light the Fires: Forecasts

 Campaigns are now in their final run. Conventions are over and polls continue to accumulate. And, of course, the forecasts are being published. In the table below are the six forecasts I’m following, providing a range of views on the race as of September 15.
Presidential Election Forecasts
Larry Sabato’s Crystal Ball, shown in the map below, lists the toss-up states’ electoral votes, which corresponds to their view of what’s in doubt. They have Joe Biden with 269 electoral votes and 65 toss-up states. They forecast a close race in popular and electoral votes. The Crystal Ball also offers a best Republican map and best Democratic map (243 Biden, 295 Trump/350 Biden, 188 Trump). All the forecasters emphasize that this is a state-by-state race and generally list similar states in play: Arizona, Florida, Georgia, Iowa, Michigan, Minnesota, Nevada, New Hampshire, North Carolina, Ohio, Pennsylvania, Texas and Wisconsin. Two other well-known forecasters who use polling – current and historic – plus information on specific state campaigns, Cook and Inside Elections, have Biden with more than 270 electoral voters and a small number of toss-up states. Cook does expect a close electoral contest based on recent elections.
RealClearPolitics only uses reported polling and believes there are 211 electoral votes in play, partially because of inadequate current polling. But when forced to distribute all electoral votes by current available polling, they give Biden 352 votes to 186 for Donald Trump.
Finally, two forecasters distribute all their electoral votes and publish a confidence factor. The best known is Nate Silver’s 538, which uses a sophisticated formula to weigh the quality of polls, add historic data and consider factors, like undecided voters. They have Biden with 321 electoral votes, but with a risk factor of 28 percent, the same amount as in 2016 when the forecasts were wrong and many with 98 and 99 percent confidence levels.
Crystal Ball
Electoral College Rating
September 10, 2020
Finally, a new forecast is from the weekly, The Economist. It has Biden with 334 electoral votes with a risk factor of only 16 percent.
It should be restated that the forecasters were dramatically off in 2016. They are more cautious today, and much of the current polling, especially state-level, has improved. The national polls were accurate, but overstated Hillary Clinton’s position, which contributed to the models under-considering risk factors and commentators (and Clinton) considering the win a certainty (i.e., no concession speech). Forecasts are still useful gathering places for all the data available and the latest thinking about the political environment. Just remember the risk. These are tools, not certainties.

No comments: