About Those Election Polls…

Some Lessons from the 2016 Election

Think the polls got the 2016 election terribly wrong? Think again. The national polling average shortly before the November 8th election had Hillary Clinton up by 3.2 percentage points, which was very close to the final difference in the popular vote. Remember, HRC won the popular vote by nearly 2.9 million, which works out to about a 2.2 percent victory margin, certainly within the polling margin of error. If not for the archaic and undemocratic Electoral College system, Trump would have never become president.

Right now, Joe Biden is polling better against Trump than Hillary did that fateful November. Real Clear Politics currently has Biden with a 7.4 percentage point lead in their polling average, and their “No Toss-Up” map has Biden winning the Electoral College vote 337 to 201. In a normal election year these would be very reassuring numbers for Democrats. But nothing is normal in 2020.

The 2016 election marked the second time, out of the last five presidential elections, that the loser of the popular vote won the Electoral College vote, and both of those times the Republicans were the beneficiaries of a system that has a strong bias in favor of less-populous states (among other problems).

If a few states had gone the other way the end result would have been different and the outcome in three states was decided by less than 80,000 votes. Trump won the decisive EC votes in Pennsylvania, Michigan, and Wisconsin, by a margin of roughly 78,000 votes, which provided him with 46 EC votes, enough to give him a total of 306. So, winning the popular vote by 2.9 million wasn’t enough, Clinton needed 78,000 votes in three key states to secure the nation’s highest office. It’s a bit like having a city the size of Scranton, PA, or Farmington Hills, MI, decide who would govern the nation.

I’ve looked at the polling data for those three states this time around and it’s generally looking good for Biden and the Democrats. Here is the picture for the August 2020 polling average from Real Clear Politics:

Pennsylvania — Biden up 6.4 percentage points.
Michigan — Biden up 6.7 percentage points.
Wisconsin — Biden up 6.5 percentage points.

This is ostensibly good news for Democrats but not if we look at how these numbers compare to August, 2016. It’s on the state level where the 2016 polls were off the mark. In all three states, Clinton had a significant lead in the polls. Something happened, and that something boils down to three possibilities: 1) the polling samples were not representative, 2) Trump supporters were less likely to state their preference to pollsters (the “shy voter hypothesis”) or 3) something happened between August and November that swayed many voters in these three states (and, most likely, elsewhere).

The polling in three key states, 2020 v. 2016

“Non-response among a major core of Trump voters” is the key reason pollster Patrick Murray gave for why Monmouth University Polling got it wrong in Pennsylvania. Republican voters are also less trusting of the news media, and that distrust likely extends to polling as well. However, a major post-election study by the American Association for Public Opinion Research found little evidence of a “nonresponse bias” or a “shy voter hypothesis.”

On the other hand, there is some evidence that pollsters over-sampled college-educated voters (and education was fairly well correlated with voter preference). Pollsters could have adjusted for this by weighting different categories of voters differently after the polls were taken but many didn’t, particularly in the state-level data.

Be that as it may, the key variable seems to be a late change in the choice of many voters. Things happened between August 2016 and Election Day. Possibly the most significant was FBI Director James Comey’s letter to Congress on October 28th, informing Congress that new emails had been discovered related to the investigation of Hillary Clinton’s possible mishandling of classified information. In other words, the polls were probably quite accurate but they can’t predict the future with certainty in part because things can happen between the time a poll is taken and the day of the election.

So, what are the lessons in all this? There are many. One is that a lead is never safe and Democrats can’t take for granted their support in close-margin blue states. Democrats have already turned their backs on younger more progressive voters. The DNC is placing its bet on winning over defecting Republicans rather than the millions of potential voters who support progressive policies such as Medicare for All and the Green New Deal. In fact, when convention organizers unveiled the speaker list, a group representing more than 200 convention delegates under age 35 felt compelled to write an open letter calling for more generational diversity in the program because the list seemed more like a nostalgia trip than a hopeful view of the future. And, even with outdoor activities curtailed due to the pandemic, TV viewership of the convention was down 6 million compared to 2016 and 2012.

Another October surprise could wipe out the current Democratic lead. What form could that take? Maybe an announcement of a vaccine for the Coronavirus? Maybe an outrageous smear of Biden/Harris? Who knows? Democrats face a number of obstacles including systematic voter suppression consisting of relentless attacks on the legitimacy of voting by mail, the undermining of the U.S. Postal Service, and Trump’s near-daily pre-emptive attack on the legitimacy of the election results.

Further darkening the picture is the fact that new voter registration is favoring the Republicans. According to Politico, the Democratic data firm TargetSmart found that “while new voter registrations had plummeted amid the coronavirus pandemic, those who were registering in competitive states tended to be whiter, older and less Democratic than before.”

The report confirmed what state elections officials and voter registration groups had been seeing in the field: neither Democrats nor Republicans had been registering many voters during the pandemic but Democrats were suffering disproportionately from the slowdown.

There is some good news though, the University of Southern California recently conducted a poll that found that 52 percent of nonvoters in 2016 say they plan to vote for Biden in November. This compares to just 32 percent who say they will vote for Trump. Biden also has an 11-point lead among voters who backed a third-party candidate in 2016.

Of course, there’s no way of determining how many of those nonvoters will show up on election day but with feelings about Trump so strong and his disapproval rating at 54 percent, it could be a deciding factor.

For democracy to be saved, Trump must be defeated but for democracy to flourish many things have to change. More on that in an upcoming article. [See: Democracy Needs a Reboot]