People Believe If 90% Prefer A over B, A Must Be Much Better than B. Are They Wrong?
. 
Graham Overton Ioannis Evangelidis Joachim Vosgerau
In a 2024 blind taste test conducted in the Republic of Ireland, 72% of the nearly 12,000 respondents chose Pepsi Max over Regular Coca-Cola. It might feel as though this means that people like the taste of Pepsi Max much more than that of Coca-Cola, but that’s not necessarily correct. Yes, many more people preferred Pepsi Max over Coca-Cola, but it may be surprising that people only tended to like Pepsi Max a little bit more than Coke.
How People Misinterpret Poll Results
Polls are a popular tool for measuring public opinion because they can easily convey what the masses prefer. But when a poll asks people to choose their favorite among a set of options, it can effectively reveal what people prefer without clearly establishing how much they like it. Nevertheless, people tend to interpret polls that way. When people learn that a large percentage of people chose one option over another, they believe that option is liked much more than the other. But that’s not necessarily the case.
In fact, it’s statistically most likely that people like one option only a little more than another, regardless of the percentage of people preferring one option. Our own analysis of large data sets—including ratings for 600,000 jokes, 1.5 million beers, and 25 million movies—supports this observation. Across various pairs of jokes, beers, and movies, the differences in liking are generally small, even when a substantial majority favors one option.
Testing this Mistake for Sports Predictions
Every year, ESPN football commentators and analysts predict the outcome of the Super Bowl. ESPN.com provides readers with two types of predictions: (1) the percentage of experts predicting each team to win, and (2) experts’ predictions about each team’s score—that is, their point spread. For the Super Bowl LI, for example, 72% of experts predicted the New England Patriots to win over the Atlanta Falcons. But these experts thought the Patriots would only win by a slim margin, predicting a point spread of just 1.87 points, on average.
To see whether people truly misinterpret polling results, we surveyed 192 American football fans. We told them that 72% of the ESPN experts had predicted “Team A” to win the Super Bowl. We didn’t tell them specific details so they couldn’t look up the results online. Given that polling information, we asked everyone to estimate what these experts thought the final point spread would be.
These football fans overwhelmingly guessed the spread to be much larger—around 13.6 points—than the ESPN experts’ average prediction of 1.87 points. As it turns out, the Patriots won by 6 points, which even though it was more than six times greater than the expert’s prediction, it was still less than what participants had predicted.
Beware of Poll Results
We didn’t just test people’s beliefs about polls for Super Bowl predictions. In other studies, we found that people had the same bias for very different polls. For instance, when we asked people to estimate how much better one wine was over another when they learned that 80% of people prefer it, we found that people overestimated how much better it is. Similarly, people overestimated how much funnier one joke was compared to another when they learned how many people had picked it as the funnier one.
Poll results are ubiquitous and can be very persuasive. However, be careful about the conclusions you might draw beyond the raw percentages. The leading option in a poll may not be as superior as the numbers suggest.
Read full paper:
Graham Overton, Ioannis Evangelidis, Joachim Vosgerau, People Believe If 90% Prefer A over B, A Must Be Much Better than B. Are They Wrong?, Journal of Consumer Research, Volume 52, Issue 1, June 2025, Pages 135–156.
