Why Most People Overestimate Trump’s Chances of Winning Today’s Presidential Election

Though Biden is far ahead in the polls, many think Trump has a good shot today. He doesn’t. Here’s why so many people are probably wrong.

Image by Kari Sullivan, courtesy of Unsplash

Poll after poll suggests that Donald Trump is about to lose the 2020 U.S. Presidential Election. Cutting right to the chase, polls suggest that Trump is going to lose in most of the swing states that he narrowly won in 2016. For example, polls show that Trump has only the teensiest chance of winning Michigan. He is also way behind in Wisconsin. As of October 29, just five days before the election — even Georgia and Texas were toss-ups. But if Trump is so far behind, why are so many Americans, including many Biden supporters, convinced that Trump has a decent chance to be re-elected?

The answer is that when we predict the future, we rely on a handful of intuitive rules of thumb for making snap judgments. These rules tend to overweight the past rather than the present. They also overweight what is obvious to the naked eye rather than what is clear only to pollsters and statisticians. With this in mind, consider four rules of thumb and how each rule nudges many people into believing Trump will pull off another victory.

Availability. Judgment by availability means judgment based on that which sticks in memory. Most people overestimate murder rates in the United States because murder gets a lot of media attention. Likewise, most people overestimate their risk of dying in shark attacks, plane crashes, and train wrecks — all of which are much more memorable than probable. Whether you think of Trump as a shark or a train wreck, he is definitely memorable. He is unpredictable, and he makes a lot of highly memorable claims. As we hear these claims over and over, and as we recall that Trump won in 2016 — when he was also way behind in the polls — it just feels like he could do it again. Once bitten, twice shy.

Representativeness. We also predict future events based on how well these future events resemble past events with which we are already familiar. I just noted that Trump won in 2016 when he was behind in the polls. This election has many of the specific features of that past election. It resembles it. There were debates, for example. There were concerns about the economy and social justice. There was no COVID-19, of course, but COVID-19 has exaggerated the political divide that was a defining feature of the 2016 election. There was even a controversy about a Supreme Court nominee in 2016, and Republicans got their way about it. To many Americans, this election strongly resembles the last one. And this gives many the gut feeling that Trump will win again.

Anchoring and Adjustment. When making difficult judgments, we often update them as new information comes in. However, when we do so, we tend to stick too close to our judgmental starting points — our anchors. Where we start a judgment has a disproportionate effect on where we end up. Amos Tversky and Daniel Kahneman first showed this when they gave high schoolers five seconds to estimate the answer to a math problem. For half the students, the problem was 8 x 7 x 6 x 5 x 4 x 3 x 2 x 1. For the others, the problem was 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8. Students who got the first problem gave much bigger estimates. This is because they began with anchors like 56 (8 x 7) or 336 (8 x 7 x 6) and then adjusted upwards from there. The anchor for the other students was presumably either 6 (1 x 2 x 3) or 24 (1 x 2 x 3 x 4). Arbitrary starting points matter because people stick too close to them. Many Americans begin their projections of the 2020 election with the fact that Trump won the last election. And where the electoral college is concerned, he won handily. He got 306 votes to Clinton’s 232. Many Americans simply cannot get this outcome out of their heads when predicting what will happen next.

Numerosity. Another seductive judgmental rule of thumb is numerosity. We count. Counting can be highly reliable. Five apples usually make more applesauce than three. But pretty often, the things people count (e.g., pizza slices, electoral college votes) are not identical. In 2020, Trump won 30 states, and Clinton won 20. Further, many of the states Trump won were physically large. Alaska and Texas alone account for almost 10% of the U.S. land mass. So, if you were to look at a 2016 electoral college map, you’d see a lot of red. But most large red states have more prairie than people. Trump won more land, but Clinton won more votes. Of course, the controversial electoral college system is still in force this election. But this time around, Biden’s election team seems to have taken this into account, wooing voters in key swing states. Or perhaps Biden just has more centrist appeal. Or maybe it’s that Biden pays more in taxes in the United States than he does in China. Whatever the reason for Biden’s large lead in the polls, it’s highly unlikely Trump can repeat what he did in 2016.

If you’d like me to attach a number to this blog, my own estimates give Trump just under a 10% chance to win. The last time I checked, Nate Silver’s expert team at 538 offered a slightly more favorable likelihood of 12%. This is a far cry from zero, of course, but it’s a much further cry from 50%. Which forecast will prove correct, then? Will it be that made by polling science? Or will it be the gut feelings of millions of Americans. I’m putting my money on science. And I can always move to Uruguay or Sweden if I’m wrong.

For Further Reading

Kahneman, D. (2011). Thinking: Fast and slow. Farrar, Straus, and Giroux.

Pelham, B. W., Sumarta, T. T., & Myaskovsky, L. (1994). The easy path from many to much: The numerosity heuristic. Cognitive Psychology, 26, 103–133.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store