A guest post by Stephen Russell:
In my previous posts I described how the polls were wrong in 2016; what can be said about the accuracy of polls in 2020; the possibility of “shy Trumpers” distorting poll results and the “enthusiasm gap”.
Despite winning the popular vote by a 2.1 percent margin in 2016, Hillary Clinton lost the election to Donald Trump through failure to win the individual states she needed in the Electoral College. Trump’s supporters are encouraged by this to hope that he can, once again, win the election, even if he gets fewer votes than Joe Biden (as polls are currently suggesting). There is good reason for their optimism in this regard.
But first, some background:
In 2016, Clinton won states worth 232 Electoral College votes. She needed 270 – 38 more than she actually got. The three closest states on Trump’s side were Michigan (16), Pennsylvania (20) and Wisconsin (10) – together worth 46 votes. Trump won Wisconsin by 0.76%. Ergo, with a popular vote margin of 2.1% already, Clinton needed a margin of 2.9% of the popular vote to carry those three sates and thus the Electoral College.
This was the fifth time in US history that a candidate has won the popular vote but lost the election (the others were in 1824, 1876, 1888, and 2000). 2016 was not the most extreme example. The loser on all five occasions was a Democrat* and it is not surprising that Democrats wish the College abolished, or failing that, circumvented.
However, the tilt in the College does not always go that direction. In five of the last ten presidential elections it has actually favoured the Democrat candidate – most recently in 2012. The degree of tilt has usually been small. In six of those ten elections the tilt has been less than one percent. Only twice has it been greater than two percent, and the 2.9 percent tilt in 2016 was the highest in at least 40 years.
Will there again be a Trump-favouring tilt? Probably yes. It has been a consistent feature of the polls that whatever the ups and downs of support for Trump or Biden, the polling margins in the key battleground states that Biden must win to take the election have been less favourable to him than in national polls.
The Economist’s election model gives Trump just a 4% chance of winning a plurality of the popular vote, but a 15% chance of winning the election, for precisely this reason.
For example, on July 8, fivethirtyeight.com had Biden ahead nationally by 9.6%. But ahead in Pennsylvania by only 7.6% (that being the tipping point state on that day). If you subtracted 9.6% from Biden’s position in each state so that the national vote were a dead heat, Trump would win four of Biden’s must-win states and romp home in the Electoral College. (Incidentally, that is not a sound method of prediction because of elasticity issues, but it is good enough to illustrate the point.)
With only a handful of polls in individual states, the averages there tend to bounce around a lot. So the size of the tilt has been erratic, and the identity of the tipping point state keeps changing. Sometimes it is Pennsylvania, sometimes Wisconsin, sometimes Florida, sometimes Arizona.
That said, five weeks later, on August 15, the result was effectively identical: Pennsylvania the tipping point state, with Biden holding a polling lead there exactly 2.0% less than his lead nationally.
It should be noted that fivethirtyeight.com claims that historical attempts to predict the degree and direction of tilt before each election have been notably unsuccessful. They were certainly unsuccessful in 2016. The final polls at that time picked New Hampshire to be the tipping point state, and picked that the vote there would exactly match that of the nation as a whole, so there would be no tilt in the Electoral College.
In 2000, polls gave George W Bush a handy margin, triggering speculation that he might win the popular vote but lose in the Electoral College. In the event, he did the opposite.
It is quite possible that the tilt in favour of Trump will be even bigger than it was in 2016. Political analyst Larry Sabato said in January he believes that Biden could get double the popular vote margin Clinton got, and still lose.
It is also possible it will reverse direction and favour Biden. But irregular observations of the state poll differentials so far are pretty consistent in suggesting a Trump tilt of about 2% – less than in 2016, but still among the highest in recent history.
* Neither the Democratic nor the Republican Party existed in 1824. The loser of that year – Andrew Jackson, who had won a plurality of the popular vote – helped found the Democratic Party in response to the “corrupt bargain” that cost him the election. He went on to win in 1828 and 1832.