In the early 2000s, a wave of overconfidence spread through the US housing market and the global financial industry. Home buyers were confident that prices would increase, investors were confident they were making safe bets, and banks were confident that they weren’t overextended. Their errors sparked a global financial crisis and a series of prolonged recessions that scarred the lives of hundreds of millions.
Of all the errors in judgment that humanity is prone to, overconfidence may be the most damaging. Psychologists Don Moore, of the University of California, Berkeley, and Max Bazerman of Harvard, call it “the mother of all biases” in their textbook on decision making. The Nobel prize-winning psychologist Dan Kahneman has said it’s the “most significant of the cognitive biases” and the first thing he’d change about humans if he had a magic wand. When asked about the most important biases, psychologist Jenn Logg from Georgetown University, a psychologist of Georgetown, told that overconfidence belonged on that list all on its own.
Overconfidence comes in many forms, but “it is overprecision that I think is the most consistent and pernicious,” says Moore, who is also the author of Perfectly Confident: How to Calibrate Your Decisions Wisely. Overprecision involves being too confident about the accuracy of your own beliefs, and there’s an easy way to test yourself for it. In his book, Moore provides such a test, which inspired the one you see below.
Here’s how it works: for each of the questions, provide a numerical range with the bottom end below your best guess and the high end above your best guess. Your goal isn’t accuracy, it’s to set ranges that contain the answer four out of five times—no more, no less.
If you wanted to have the truth be in your range every time, you could just set really, really wide ranges. To benefit from the tool, aim to set the ranges wide enough that the correct value falls into your range four out of five times—but not so wide that all five answers are within your ranges. In other words, you want to set your ranges so the truth falls within them 80% of the time.
How did you do?
Most people get far fewer answers within their ranges than they were intending, evidence that they’re overconfident about their knowledge of the world. When people are told to set ranges that encompass the truth 90% of the time, for example, the average person sets ranges that include the true value just 50% of the time—evidence that they think they know more than they do. (Moore and Logg both said they’ve not found any differences by gender on this form of overconfidence.)
If you want to practice, we’ve built a separate test for members—you can try again with different questions and see if your confidence becomes more calibrated over time. (If you’re not a member, we’re 100% confident you should become one.)
Overconfidence in our own beliefs helps us feel better about ourselves (no one likes to think they don’t know what they’re doing). In some cases it can make other people think better of us as well. But it’s dangerous, for individuals and for groups.
In a recent study, not yet published, Barbara Mellers, a psychologist at the University of Pennsylvania, asked participants to answer numerical trivia questions like the ones above. Then she had the participants discuss their estimates in groups, and gave everyone a chance to revise them after the discussion. Individuals’ assessments improved only about half the time; the groups’ averages were hardly any better.
The problem stemmed from overconfidence. In some groups, the members with the most accurate initial estimates were also the most confident. In these cases, group deliberation tended to improve average accuracy. But in groups where the least accurate group members were also the most confident, group members still listened to them, driving down everyone’s accuracy.
Mellers’ research shows that overconfidence doesn’t just hurt individuals—it can drag down entire teams. Everyone has been in a meeting where the most vocal and confident person in the room clearly doesn’t know what they’re talking about. The most harmful meetings are the ones where that’s happening and the group can’t even recognize it.
It’s tempting to think just knowing more is the answer: surely you’d do better on the questions above if you had more knowledge about global business and world affairs. But that’s not so, says Moore. Instead, he says, self-assurance grows proportionally with expertise.
Still, it is possible to improve, and to bring your level of knowledge and your confidence more in line.
The first thing to do is to push yourself and your colleagues to be as specific as possible. “If you say I’m a great driver, everyone can claim that,” says Logg. “But if you can define what makes a great driver and get to specifics that’s going to help people become more calibrated.” For instance, people will be more accurate if you ask them “How many times have you rolled through stop signs?” (This is an example of “placement” overconfidence, where we overstate how we compare with others rather than overstating our views about the world.)
Second, think numerically and probabilistically. “When you can quantify things as much as possible, that helps,” says Logg. So don’t ask your team how they’re feeling about next quarter’s sales. And don’t just ask them for a number. Divide up the possible sales outcomes into buckets and ask about the chance that next quarter’s sales fall into each of those buckets.
Finally, it can also help to just literally think again. In one study, researchers simply asked participants to estimate the same quantity two separate times. The average of a person’s two estimates tended to be more accurate than either on its own. It’s essentially a way of creating a “wisdom of the crowd” effect in your own head.
If just doing the same thing twice seems too weird, you can make an estimate and then think “How might this be wrong?” Write down your reasons, and then consider changing your estimate based on what you’ve come up with.
All these practices can help, but don’t expect to overcome overconfidence any time soon. “You do get better,” Moore told , but he cautioned that good calibration in one domain might not translate to another.
“It is a noble aspiration that will take a lifetime of intentional effort,” he says. “My wife and family will still tell you that I am overconfident in all sorts of ways.”