Predicting the Election by Anticipating Voter Behavior
A system is a group of interrelated entities interacting separately to form a unified whole. An election is a system which culminates on election day when voters turn out to place their vote. And being a system, the results of election day can be predicted based upon the events that take place throughout the election. Hence the basis for voter polls and projections.
But every once in a while, a system gets disrupted. A system behaves in unexpected ways through the introduction of unforeseen events, or inputs. For the 2020 election, that disruption would be COVID-19. And its impact on election day will influence the results of the election far more significantly than we anticipate – even as we are just days out.
The reason we will fail to adequately anticipate the election turnout comes from the way we perceive unforeseen events. More specifically, the thought patterns, or heuristics, we create to simplify an otherwise complex system. Heuristics dominate our everyday lives and are largely the autopilot that drives us throughout the day. They define our routines, our tendencies, our “why do I always do this” moments. And in aggregate, they will define the election – by defining the behavior of voters on election day.
We anticipate two major trends impacting voter behavior: (1) the overwhelming number of early votes, nearly two-thirds of which were casted in-person relative to by-mail; and (2) the perceived risk of COVID-19 among those who plan to vote on election day.
The significantly higher than normal number of people voting early may sway the polls by giving off the impression to those remaining voters that their election day vote may not be as critical or at least as relevant as when fewer people voted early. This skews to impact Biden voters more than Trump voters since we know more Democrat voters vote early than Republican voters. Which the available research confirms.
According to TargetSmart, a data research firm, and NPR, more Democrats have cast votes so far, by a wide margin, relative to Republican voters, with the difference being most pronounced among African American and Elderly voters. But conspicuously, we do not see this trend among Younger voters – at least not yet.
And the biggest variable in early voting is whether the increased number of votes casted early is simply a time-shift in the votes or represents a larger voter turnout. The media seems to believe early voting is a sign of a larger voter turnout, but an analysis of the heuristics underlying COVID-19 behavior seems to suggest otherwise.
COVID-19 has shown that we tend to substitute existing behavior patterns rather than create new behavior patterns when confronted with the need to change. If we are predisposed to work, then we will find new ways to work. If we are predisposed to vote, then we will find new ways to vote. And if we are not predisposed to vote, then we will not seek to vote, regardless of whatever is changing around us.
We substitute patterns of behavior and perceive it to be change. That itself is a heuristic. And largely why many pundits believe the early voting is more a time-shift. Which, honestly, is not hard to predict. What is harder to predict, and most critical, is the election day behavior of the remaining voters.
The perceived election day risk of COVID-19 disproportionately affects voters who are likely to vote for Biden than for voters who are likely to vote for Trump. Something proven multiple times through surveys conducted over the course of the year. Democratic voters are more likely to observe social distancing protocols and to follow restriction guidelines.
Which may make it more difficult for Democrats to vote on election day relative to Republicans. And while many polling locations have addressed voter difficulties related to COVID-19 restrictions, we still will not be able to predict its full impact until the actual day of the election.
That is how heuristics work, even when you know they exist, they are hard to predict, and for those will be affected, it is hard to acknowledge its influence. Which is why the media is understanding the impact of COVID-19 on election day turnout – the public itself is understating its perceived impact.
Given the poll numbers decidedly favoring Bide, there is a risk of voter fatigue on election day among Democrats. Many may assume that since their more politically motivated colleagues have voted, and Biden holds a commanding lead in many parts of the country, that their vote is unnecessary.
This phenomenon is not new. We know voter fatigue exists. And it is also a heuristic.
What we are not anticipating is the heightened effect of voter fatigue. I know this may seem sacrilegious to say, given all the heightened rhetoric of late, but many voters will find themselves simply lacking the desire to go out and vote on election day.
And the reason has less to do with the political animus, and more to do with COVID-19 fatigue.
We as individuals, as a society, have expended significant willpower to overcome the monumental challenges the pandemic has presented. But willpower is finite, and studies have shown that willpower, once expended, predisposes people to make subsequent decisions that lack willpower – demonstrating that willpower can be seen as an opportunity cost.
Translation: since people have expended significant willpower overcoming COVID-19, many will have less remaining willpower to overcome voter inertia. As a result, far fewer voters will turn out on election day than anticipated, and the overall number of voters will either be slightly below 2016 numbers or possibly significantly less.
So what will go through the minds of voters on election day?
Voter safety will be on the minds of everyone turning out to vote on election day. But the perceived safety will vary among voters and by location. Republican voters tend to be less concerned with COVID-19 risks than Democratic voters, and parts of the country with Republican leadership tend to have less restrictions set up at polling locations, making it relatively easier to place a vote than in locations with more stringent restrictions.
The impact will be marginal individually, but in parts of the country where the difference between the candidates is marginal, the impact should be significant in aggregate.
In states like Pennsylvania and Texas, we anticipate these marginal differences to sway the polls in favor of Trump. Both states have seen recent upticks in COVID-19, and cases have disproportionately affected minorities and urban demographics, which have traditionally gone Democratic.
The worsening COVID-19 will ironically make voting more difficult for regions disproportionately affected that have responded with more voter safety measures – which in the minds of individual voters are seen as voter restrictions – which are regions disproportionately Democratic. And the remaining Democratic voters who wait until election day to cast their votes will face both the perception risk and the logistical hurdles of voting in the time of COVID-19 – further inhibiting their turnout.
The headlines may say the worsening COVID-19 will hurt Trump on election day, but an analysis of individual voter behavior relative to COVID-19 seems to indicate that Trump will benefit from the behavioral changes – which if significant enough, may be enough to pull off another underdog victory.
Whether these factors have already been built into the polls is uncertain, but what is certain is that these factors will be more impactful that we currently believe – that is the nature of heuristics, even when you see them coming, you cannot adequately predict them.
Antibiotic Prescriptions Associated With COVID-19 Outpatient Visits Among Medicare Beneficiaries, April 2020 to April 2021
Outpatient Visits for COVID-19 and Associated Antibiotic Prescriptions Among Medicare Beneficiaries Aged 65 Years or Older, by Setting, US, April 2020 to April 2021. The volume of COVID-19 visits differed by setting: emergency department, 525 608 (45.8% of all visits); office, 295 983 (25.3%); telehealth, 260 261 (22.3%); and urgent care, 77 268 (6.6%).
Source: Journal of American Medical Association Network