With the 2016 presidential election a little more than one week away, predictions of whether Donald Trump or Hillary Clinton will win the White House look more important than ever, as the chance for a sudden turn of events that could change the course of the election shrinks as time runs out.
One such event took place on Friday, October 29, when FBI Director James Comey sent a vaguely worded letter to eight Republican congressional committee chairs, saying that the FBI had found emails “pertinent” to the investigation into Clinton’s use of a private email server when she was Secretary of State.
Whether the strange letter, which the Clinton campaign saw as an attempt to tip the scales in the election, will cut into Clinton’s polling lead will not be known for several days, when polls cover the time period after the Comey announcement come out.
But it may be worth remembering that during the FBI’s original email investigation, during which Clinton opponents confidently anticipated that she would be indicted, Clinton never trailed in the polling averages. For example, the following graphic from Huffington Post Pollster.com shows her lead prior to and one month following, Comey’s July 5 announcement that the FBI would not recommend charges against Clinton.
If her prior performance is an indicator of future performance, the Friday letter from Comey will likely have little if any negative impact on Clinton.
But rather than simply look at polls, several top election experts have created statistical models that predict the probability of each candidate winning election, generally expressed in terms of percentage probabilities.
All of those major election forecasting models agree that Clinton, nine days in advance of the November 8 nationwide balloting, remains a heavy favorite to win the presidency. But how favored is she? That remains a matter of debate among the various production specialists.
The best-known of the election prediction statistical models, the FiveThirtyEight.com model, created by statistician Nate Silver, is also the most “bullish” on Donald Trump. As of October 30, in Silver’s “polls only” model — that is, a prediction based only on polling data — Clinton had an 81.1 percent chance of winning on November 8. Donald Trump stood at 18.9 percent.
In other words, according to Silver and FiveThirtyEight.com, Trump has slightly less than one chance in five to win the election.
The FiveThirtyEight.com “polls plus” model, which also factors in other factors such as the state of the economy and historical trends, gives Trump an even better shot at becoming President of the United States. In the polls plus version, Trump’s chances stood at 21 percent, while Clinton’s chance dropped to 79 percent.
According to Silver, his model accounts for a larger-than-usual share of voters who say they are undecided. The Silver model also uses historical data going back to 1972 to “calibrate” the 2016 polling data. But as Silver himself explains, polling has become increasingly accurate over time.
“If we calibrated the model based on presidential elections since 2000 only — which have featured largely accurate polling — Clinton’s chances would rise to 95 percent, and Trump’s would fall to 5 percent,” Silver wrote.
But he added that “going back to 1972 takes advantage of all the data we have,” rather than relying on the smaller sample size of the last four presidential elections.
But other election prediction models are significantly more confident about Clinton’s chances, and more pessimistic about Trump’s.
The Benchmark Politics model, unlike the FiveThirtyEight.com version, includes “county-level demographic data” as well as polls and economic statistics to create its model — and based on those factors predicts Clinton with a much more imposing 91 percent chance of victory, to only nine percent — less than one in 10 — for Trump.
The New York Times Upshot model is based only on polling data, like Silver’s “polls only” model. But Upshot clearly employs a different statistical formula, because the site’s prediction is exactly in line with Benchmark Politics, giving Clinton a 91 percent chance of taking the White House, to nine percent for Trump.
Finally, data scientist Sam Wang of Princeton University has been making election predictions since 2004 through his Princeton Election Consortium. Using his statistical methods, Wang gives Clinton a 99 percent chance of winning — making her victory a near certainty.
[Featured Images by Chip Somodevilla and Justin Sullivan/Getty Images]