What the 2016 election polls tell us about the 2020 race.
After what happened in 2016, can we really trust Biden’s lead in the polls so far? And what do we have to keep in mind as we look at polls in the final days leading up to November 3rd?
Let’s find out.
The 2016 polls were mostly right
Despite many still saying the polls got it wrong in 2016, if you look at the data it turns out they were pointing in the right direction if read correctly.
National polls: For a period of time, Clinton maintained a head-to-head lead against Trump which narrowed down to 3.2% in the average of polls right before the election. With Clinton winning the popular vote on election night with a 2.1% margin, well within the margin of error, it’s safe to say the national polls were right.
State polls: If you compare the state poll averages before the election to the actual results, the polls correctly predicted the winner in 46 out of 50 states.
Of the remaining 4 states, 3 of the results deviated within the margin of error (<3.5%): Pennsylvania and Michigan flipped to Trump, and Nevada to Clinton. Only 1 state got it wrong: Wisconsin.
So overall, even the state polls were not really that bad.
Source: Data from RealClearPolitics for MI, NV, PA, WI.
The Trump ‘surprise’ factor
Those undermining Trump in light of recent polls, need to be reminded that while Pennsylvania, Michigan and Wisconsin were the upsets which landed Trump the electoral college win, he did pull off a few more surprises.
Among states where the candidates were separated by less than 10 points, Trump achieved a favourable swing beyond the polling margin of error in at least 6 states. Of these, Wisconsin was the only one he flipped — but he managed to widen his lead or reduce his loss in the remaining 5 states (listed below) beyond what the polls predicted. The Trump ‘surprise’ factor, as you may want to call it, averaged around 7% across these states — a number to keep in mind to avoid complacency in the 2020 battleground state polls.
Source: Data from RealClearPolitics for IA,MN, MO,OH,SC.
2020 is not 2016
Throughout the 2020 race, Trump has defended his plummeting poll numbers pointing them out as fake or inaccurate. While we shouldn’t discredit the argument given 2016, things are different this time:
(i) Biden’s lead is stronger than Hillary’s: In 2016, Clinton experienced a slow declining lead in polls (see chart below), coupled with a campaign smeared by controversies and a sense of complacency with outright dismissal for a possible Trump win, which created much of the “surprise” in the end. As shown above, the polls weren’t completely wrong. Even before election night, based purely on polling, Clinton was sitting comfortably on only 272 electoral votes, just 2 above the minimum required.
As of mid-July, Biden had a 9 point lead over Trump, more than double that of Clinton at the same time in 2016. Clinton never achieved the leads he has seen in the past couple of months. As recently evidenced, even if the polls are off and the 2016 adjustments are made to those numbers, Biden holds a comfortable lead against Trump in key battleground states for 2020 and maintains an electoral college advantage.
(ii) Trump is losing Republican states: Forget battleground states, a bigger worry for Trump is Biden’s strong numbers in Republican strongholds like Texas and Georgia. The two candidates are statistically tied in Texas as at July — a scary prospect for Republicans since losing Texas (with its 38 electoral votes) is pretty much game over for Trump.
As of July, Biden is also out polling Trump in Florida, North Carolina and Arizona which are only adding to his bonus tally and are not necessarily states he must win to get to a minimum 270 electoral votes.
(iii) Voter turnout and participation: 2016 saw a 20-year low voter turnout with estimates showing 43% of eligible voters, or up to 100 million Americans, did not vote. A significant statistic in an election where the winner was decided by around 80,000 votes in 3 states. However, it is clear that the political engagement in the US has shook up since then and if the 2018 mid-term elections and recent protests are any indication, we should expect to see a higher turnout in the 2020 election. This would more likely than not support the accuracy of the polling data.
Most importantly, with millions of votes to be cast early or by mail this year, the Trump campaign has a lot less time to turn his poll numbers around.
(iv) Biden is winning Trump’s demographics: Biden is holding strong leads in demographics which helped Trump win in 2016, particularly:
- Suburban voters: Biden maintains double-digit leads with suburbanites, a group Trump led by 4 points in 2016 according to exit polls; and
- Seniors: Biden is leading Trump among senior voters by up to 8 points in recent polls, a group which backed Trump in 2016 by the same margin.
How to read 2020 polls with a 2016 lens?
Now of course a lot can change in the final weeks leading up to the election, including either candidate’s fate in the polls. A few things to keep in mind when reading the flurry of polls in the coming days:
- Always refer to the average of recent polls, instead of any one particular poll being sensationalised in the news.
- Account for a 3–3.5% margin of error in poll results, as specified in most polls.
- Keep in mind Trump’s ‘surprise’ factor (approx. 7%), in competitive states, a metric which could represent the shy Trump voters, or the so-called silent majority, if there is one.
It is also important to remember that the added complexities of the public health crisis cannot be measured in these polls. The pandemic along with voter turnout as well as accessibility to vote (whether in person or by mail) remains to be the critical issue which could influence the election outcome.
