Early this morning, Donald Trump surpassed 270 electoral votes to become the presumable 45th president of the United States of America. Hillary Clinton, his opponent, was a heavy favorite by almost every measurement, but lost by what currently appears to be a significant margin.
Below is a table with the websites that predict who will win the presidency, along with the number of electoral votes they think a certain candidate will win vs. their opponent. “C+10” means that the website predicted Hillary Clinton would win 10 more electoral votes than Donald Trump.
|New York Times Upshot||C+106|
|Princeton Election Consortium||C+76|
In the end, however, Trump won handily. While not all states have been called, it looks like he could win anywhere from 50 to 80 electoral votes more votes than Clinton — far than 100 electoral votes different from most predictions. Even more interesting will be the final vote percentages from each state vs. each predictor’s prediction for each state, but we will not have the final vote information for possibly days.
Some predictors were extremely confident in their results. Sam Wang, who works on Princeton’s Election Consortium, tweeted on October 18th that he would “eat a bug” if Trump won more than 240 electoral votes. Well, Mr. Wang, Trump has far more than delivered.
It is totally over. If Trump wins more than 240 electoral votes, I will eat a bug. https://t.co/3eefhWzI3y
— Sam Wang (@SamWangPhD) October 19, 2016
Nate Silver from FiveThirtyEight was much less sure of a Clinton victory. While he gave Clinton a solid advantage, he was accused of “putting his thumb on the scales” in an article by Ryan Grim on The Huffington Post. The article claimed that Silver was “wrong” for “unskewing” polls (which is his attempt to adjust polls to be more accurate). Silver responded to this article on Twitter defending his poll adjustments.
In the end, while essentially every prediction website was way off, FiveThirtyEight proved to be the most accurate at giving Trump a realistic chance. They gave Clinton “only” a 71.8% chance of winning (in their polls-plus model), compared to 85% from the New York Times Upshot, 89% from PredictWise, 92% from DailyKos, 98% from the Huffington Post, and >99% from the Princeton Election Consortium. FiveThirtyEight understood the degree of uncertainly in this election and accounted for that in their calculations.
By no means was this the first time that Trump outperformed expectations. He won the Republican primary election too, when he was predicted time and time again to lose. In that election too, hardly anyone predicted Trump’s eventual win, as Nate Silver explained at the time.
In the coming days and weeks we will learn more about what exactly caused polls and predictions to be so off. All that we know for certain now is that the polls failed dramatically. Clinton’s campaign had a huge data team, but clearly, they missed something somewhere.
This was a fascinating election to watch play out. Trump far outperformed expectations and it will be interesting to see what exactly went wrong with the polls.