CONFIDENCE . . . AND OVERCONFIDENCE
ALL YOU NEED IN THIS LIFE IS IGNORANCE AND CONFIDENCE; THEN SUCCESS IS SURE.
MARK TWAIN, LETTER TO MRS. FOOTE, 1887
Of all the errors and biases that undermine our judgment, the most frequently cited is overconfidence. Already in 1995, summarizing research from the previous two decades, behavioral economist Richard Thaler observed, “Perhaps the most robust finding in the psychology of judgment and choice is that people are overconfident.”1 Around the same time, psychologist Scott Plous reached a similar conclusion: “No problem in judgment and decision making is more prevalent and more potentially catastrophic than overconfidence.”2
Since then the notion that people suffer from overconfidence has been repeated over and over, to the point where it has become accepted as fact. A few quotes from recent years are representative:
• Joseph Hallinan, Pulitzer Prize–winning business journalist, in Why We Make Mistakes: “[M]ost of us tend to be overconfident, and overconfidence is a leading cause of human error.”3
• David Brooks, New York Times columnist, in The Social Animal: “The human mind is an overconfidence machine.”4
• Ken Fisher, president of Fisher Investments: Investors make mistakes due to an “innate tendency toward overconfidence.”5
• Nate Silver, in The Signal and the Noise: “[O]f the various cognitive biases that investors suffer from, overconfidence is the most pernicious. Perhaps the central finding of behavioral economics is that most of us are overconfident when we make predictions.”6
And so, given this widespread problem, we are advised to beware. We’re urged to acknowledge our natural tendency to be overconfident and to guard against it.
At first glance all this makes good sense. Look in any dictionary, and overconfidence is defined as “excessive confidence” or “greater confidence than circumstances warrant.” The adjective overconfident means “too confident” or “excessively confident.” These definitions are reasonable, because the prefix over denotes excess. Naturally we should want to avoid excessive confidence. Who wouldn’t?
Yet in Chapter Two we also saw that positive illusions can improve performance. Shouldn’t it follow that a very high degree of confidence is a good thing? Why must it be “potentially catastrophic”? Furthermore, as we saw in Chapter Three, when performance is relative and we need to do better than rivals, a very high level of confidence can be more than useful and even necessary. We need to ask: If overconfident means too confident, too confident compared to what? If overconfidence means greater confidence than circumstances warrant, which circumstances are we talking about? Very soon, what seems like a simple idea becomes much more complicated.
In an essay published in 1947, “Politics and the English Language,” George Orwell wrote: “A man may take to drink because he feels himself to be a failure, and then fail all the more completely because he drinks. It is rather the same thing that is happening to the English language. It becomes ugly and inaccurate because our thoughts are foolish, but the slovenliness of our language makes it easier for us to have foolish thoughts.”7 That’s a good summary of the current state of affairs regarding overconfidence. Foolish thinking has led to inaccurate language, and slovenly language has made it easy to have foolish thoughts. To help make great decisions, we need to take a new look at what we mean by confidence . . . and what we mean by overconfidence.
OVERCONFIDENCE IN OUR DAILY CONVERSATION
In everyday speech, overconfidence is typically applied after the fact to explain why something went wrong. Examples abound in all walks of life.
In November 2012, after a long and closely fought campaign, President Barack Obama was reelected by a small margin of the popular vote and a decisive majority of electoral votes. Soon reporters declared that his opponent, Mitt Romney, had suffered from overconfidence. Their evidence? Rather than working to get out the vote on election day, Romney spent his time drawing up a list of White House appointments, while $25,000 worth of fireworks sat nearby for a victory party that never took place. A senior aide explained: “It was overconfidence based on inaccurate assumptions and flawed data.”8 Curiously, no one had accused Romney of overconfidence during the campaign, when he worked long hours, going from state to state and speaking at rallies from morning to night. But when the votes were counted and he fell short, it was irresistible to claim he had been overconfident.
Four years earlier, the same label had been pinned on a different losing candidate. In June 2008, when Hillary Clinton ended her long quest for the Democratic nomination, the New York Times explained that her campaign had been “suffused in overconfidence, riven by acrimony and weighted by emotional baggage.”9 By overconfidence the article meant that the Clinton campaign had been complacent, perhaps imagining the nomination was in the bag, which led to serious errors and eventually to defeat. In fact, a closer look showed that Clinton had worked tirelessly and campaigned relentlessly, which is hardly the stuff of complacency. What had gone wrong? The race was decided during a few weeks in February, when Barack Obama sprinted ahead in delegate count thanks to a shrewd focus on small states, gaining a lead that he never lost.10 True, the Clinton campaign made a crucial strategic error by overlooking small states, but that by itself doesn’t justify a charge of overconfidence (unless of course we attribute any bad outcome, after the fact, to overconfidence, which is often what happens). As for Barack Obama, he was described as having gaudy confidence and outsize confidence, but never overconfidence, and for an obvious reason: he was the eventual winner. His confidence, no matter how great it may have been, turned out to be justified. But on reflection, who had displayed unwarranted confidence: a prominent senator backed by an impressive political machine, or a first-term senator with little national experience?
These examples are typical. Overconfidence is a common explanation any time something turns out badly, and not just in politics. When the Fukushima nuclear reactor was damaged in March 2011 by a tsunami, spilling radioactivity into the air and sea along the Japanese coast, a civil engineer pointed to three separate instances of overconfidence: a poor understanding of earthquakes when the reactor was designed, reliance on simplistic failure models regarding plant reliability, and an emphasis on the reactor vessel rather than on spent fuel storage after the tsunami struck. Similarly, when Hurricane Katrina devastated New Orleans in 2005, FEMA director Michael Brown charged that the Bush administration had failed to take precautionary measures because it had been “overconfident” that it could handle the crisis.11
We know that athletes understand the importance of confidence. Quite naturally, when they’re victorious they give credit to their high confidence, and when they’re defeated are quick to blame overconfidence. In December 2012, before his fight with José Manuel Marquez, a smiling boxer Manny Pacquiao was described as relaxed and composed, “his face filled with confidence.”12 Hours later, after a stunning right cross had knocked him cold, Pacquiao saw it differently: “I just got overconfident in this fight.”13 Well, at least that’s how it seemed after the fact. And by that logic, Pacquiao may believe that if he just watches out for overconfidence, he should be able to win the next time they meet.
Perhaps nowhere is failure blamed on overconfidence as often as in the business world. In July 2011 Netflix, the enormously successful movie rental company, announced that it would split its rental business from a new video streaming business and charge customers extra if they wanted both. Customers were outraged, and after three weeks, chief executive Reed Hastings was forced to apologize and reverse his decision. The damage was severe: Netflix lost 800,000 subscribers, and its stock price fell by more than 25 percent. In October Hastings showed contrition and admitted that he had been guilty of overconfidence.14 As he explained, he had been too sure of himself and failed to take note of customer concerns. In the future, he promised, Netflix would slow its decision making to make sure there was room for debate. Interestingly, this comment came from a man who had often been praised for bold and courageous decisions. Companies don’t fail because they move too swiftly, Hastings had said, but because they move too slowly. That made good sense as long as his swift moves were successful. But when a bold move turned out badly, he blamed overconfidence.
A few months later, in May 2012, when JPMorgan Chase lost more than $2 billion because its risk-monitoring systems failed to spot the dangers in its derivatives portfolio, chief executive Jamie Dimon pointed the finger at—you guessed it—overconfidence. Overconfidence gave rise to complacency, Dimon explained, which led to the massive error.15
Recently I conducted a search for the words overconfident and overconfidence in the business press. By far the most common use was retrospective, to explain why something had gone wrong. Examples ranged from KFC’s initial failure in India (“‘They came into the country overconfident,’ said a local customer, before biting into a Chicken Tikka Wrap n’ Roll”16), to Bob Nardelli’s bumpy tenure at Home Depot (one investor complained that Nardelli “comes across as arrogant and overconfident”17), to why Airbus lost market share when the A380 was delayed (“Airbus managers also quietly concede that they became overconfident after several years in which Boeing gave up market share to a more aggressive Airbus”18).
In all of these examples, from politics to natural disasters to sports to business, overconfidence offers a satisfying narrative device. Take any success, and we can find reasons to explain it as the result of healthy confidence. We nod approvingly: They were confident—that’s why they did so well. Take any failure, and we shake our heads: They were overconfident—that’s why they did poorly. If they hadn’t been so sure of themselves, they might have done better.
There’s an appealing syllogism at work:
• Things turned out badly, so someone must have erred.
• Errors are due to overconfidence.
• Therefore, bad outcomes are due to overconfidence.
Unfortunately each of these statements is flawed. First, not everything that turns out badly is due to an error. We live in a world of uncertainty, in which there’s an imperfect link between actions and outcomes. Even good decisions sometimes turn out badly, but that doesn’t necessarily mean anyone made an error. Second, not every error is the result of overconfidence. There are many kinds of error: errors of calculation, errors of memory, simple motor errors, tactical errors, and so forth. They’re not all due to overconfidence.
Those first two flaws lead to the third. It might be convenient to blame bad outcomes on overconfidence, but the logic doesn’t add up. Even worse, the effect is to dilute overconfidence to the point where it’s all but meaningless. When any failure can be attributed to overconfidence, the term means nothing at all.
In Decisive: How to Make Better Choices in Life and Work, Chip Heath and Dan Heath cite the famous comment by a Decca record executive in 1962, who had just listened to an audition by an aspiring quartet called The Beatles, but declined to offer them a contract because he felt guitar groups were a thing of the past.19 He turned out to be wrong, of course, but is that really evidence of overconfidence? We might just as easily say he was underconfident when it came to guitar groups. (We could even claim he was overconfident in expressing a view of underconfidence, a hair-splitting argument that illustrates the deeper problem—overconfidence can be used to mean almost anything.)
There’s another danger when we attribute failures to overconfidence, and it’s potentially the most serious of all. Accusations of overconfidence carry a moral overtone. When we charge people with overconfidence, we suggest that they contributed to their own demise. We imply that they at least partially deserved their fate. They committed one of the seven deadly sins, sometimes called pride or vainglory. We think: They were too sure of themselves. They should have known better. They got what they had coming.
From there, it’s a small step to imagine that if we can simply avoid the sin of overconfidence, we won’t meet the fate that befell others. After all, most people don’t think that they suffer from excessive pride or vanity. Others might exhibit overconfidence, but we don’t imagine the term applies to us. And so we comfort ourselves with the notion that we have little to worry about. Did Netflix blunder because Reed Hastings was overconfident? Too bad for him, but I won’t make that mistake, because I’m not overconfident. Those disasters that befell KFC and Home Depot? If they were due to overconfidence, I can rest assured they won’t happen to me. We may even experience a bit of schadenfreude, taking pleasure at the misfortunes of others.
The irony, of course, is that those people didn’t see themselves as overconfident, either. At the time, they believed they were being appropriately confident, and that their actions were bold and decisive but surely not excessive. It’s only after things turn out badly that we hear the term. We’ll never learn from the errors of others if we attribute them to overconfidence. We end up fooling ourselves.
OVERCONFIDENCE IN THE HERE AND NOW: NOT ONE THING BUT THREE
Fortunately, overconfidence isn’t only inferred after the fact. It can also be studied in the here and now, defined as confidence that exceeds what’s objectively warranted. That’s the way it has been studied in decision research, and here the evidence seems clear. Decades of research have produced very consistent findings that people really do suffer from a bias of overconfidence.
Some examples have been cited over and over, to the point of folklore. In 1981, Swedish psychologist Ola Svenson found that 93 percent of American drivers rated themselves better than average. Swedish drivers were somewhat less extreme, with only 69 percent claiming to be better than average.20 Surely this can’t be correct. The obvious explanation is that these drivers were overconfident. Another study found that 25 percent of high school seniors believed they were in the top 1 percent in terms of their ability to get along with others,21 and still another found that 37 percent of engineers rated themselves in the top 5 percent of all professionals.22 University professors aren’t immune either, as a large majority rate themselves above average when it comes to teaching ability.23 Surely that’s incorrect. They must be overconfident.
Other evidence comes from experiments discussed in Chapter One, in which people are presented with questions of general knowledge—the length of the Nile, the year Mozart was born, and so on—and asked to estimate a range that they are 90 percent confident contains the correct answer. The original study, conducted by Marc Alpert and Howard Raiffa in 1969, found that the “90 percent confidence range” contained the correct answer less than 50 percent of the time.24 Variants of this study have been conducted countless times and produced very consistent results. Time and again, people provide ranges that are far too narrow. The inevitable conclusion: people are overconfident.
Given all of these examples, the evidence seems overwhelming that people suffer from overconfidence. But when we look more closely, however, it’s not clear at all. As Don Moore and Paul J. Healy described in a 2008 article, “The Trouble with Overconfidence,” the single word—overconfidence—has been used to mean three very different things, which they call overprecision, overestimation, and overplacement. They explain: “Researchers routinely assume, either explicitly or implicitly, that the different types of overconfidence are the same.”25 But they’re not the same, and when we take them one at a time, the notion that people suffer from a widespread tendency to be overconfident begins to unravel.
Overprecision is the tendency to be too certain that our judgment is correct. Those studies that asked for 90 percent confidence ranges? That’s an example of overprecision. In The Signal and the Noise, Nate Silver mentions overconfidence as a serious problem when making predictions. He’s referring to overprecision: the tendency to believe a prediction is more accurate than it turns out to be.26
Overestimation, the second kind of overconfidence, is a belief that we can perform at a level beyond what is objectively warranted. When golfers believe they can sink 90 percent of their six-foot putts, that’s overestimation. When we believe we can complete a task in a shorter period of time than we can, that’s overestimation. Overestimation is an absolute evaluation; it depends on an assessment of ourselves and no one else.
The evidence for overestimation isn’t nearly as strong as the evidence for overprecision. For many ordinary tasks, there’s good evidence that people believe they can do better than they really can. As Tali Sharot writes in The Optimism Bias: A Tour of Our Irrationally Positive Brain, most people also believe the future will be better than the present. But there are limits. When it comes to difficult tasks, many people believe they won’t do very well and at times even underestimate how well they can do. Overall, it’s a stretch to claim that people have a general tendency to overestimate.
Overplacement, the third kind of overconfidence, is a belief that we can perform better than others. It’s not an absolute judgment, but a relative one. When 90 percent of American drivers believe they’re better than average, that’s overplacement. When 80 percent of students believe that they’ll finish in the top 20 percent of their class, that’s overplacement, too. A well-known example of overplacement is Garrison Keillor’s fictional town of Lake Wobegon, where “all the children are above average.” Of course it’s impossible for a majority to fit in the top half of the distribution, but many studies suggest that we think we do. Sharot writes, “most people perceive themselves as being superior to the average human being.” She calls it the superiority bias and says it’s a pervasive error.27
When it comes to overplacement, much of what we have come to believe isn’t just exaggerated but actually incorrect. When I teach executives about decision making, I often ask them to complete a short questionnaire with many questions that call for judgments and choices. First, with a nod to Svenson’s study, I ask them to evaluate themselves as drivers, then to compare themselves to their peers. The great majority—71 percent of more than four hundred people I asked over a period of several months—rated themselves above average, a finding that’s very consistent with Svenson’s results all those years ago.
If I had stopped there, I too might have concluded that people are overconfident. But my questionnaire went on to ask about a very different skill—drawing, as in the ability to draw a good portrait. This time, not only did most people think they were not very good, but they also believed that they were worse than their peers. The great majority—59 percent of the same population—ranked themselves as below average. That’s not what we’d expect if people truly suffered from a pervasive bias of overconfidence.
What should we conclude from these replies? In fact, they’re not wrong. They make sense when we consider what people know about themselves and about others. Let’s start with driving. What driver do you know best? Yourself, very likely. Unless you have a personal chauffeur, you probably know more about yourself than about any other driver. And what do you know about yourself? Most likely that you’re a very good driver. You’ve never been in a major accident, and rarely (if ever) have you been stopped for speeding or any other serious violation. Dozens of times every week, you buckle up, turn on the ignition, put your car in gear, and drive safely to your destination. By any objective measure, you really are a very good driver. You might even have a safe driver discount from your insurance company to prove it.
Furthermore, you know there are many bad drivers on the road. Every week you hear about serious traffic accidents and reckless behavior behind the wheel. The US National Traffic Safety Administration reported 32,367 traffic fatalities in 2011, of which 9,878, or 30 percent, were caused by alcohol-impaired drivers. That’s an alcohol-related death every fifty-three minutes, and you were (knock on wood) nowhere near any of them.28 With so much evidence of bad driving, it’s entirely reasonable to infer that you’re above average. In fact, if you’ve never been in an accident and have had only a few infractions, you might well conclude that you’re among the very best. There’s really no obvious reason to believe that anyone is much better! So if you place yourself in the top 20 percent of all drivers, is that an excessive evaluation? Not at all. It’s entirely reasonable based on the information you have about yourself and about others.
Now consider drawing. Drawing isn’t a routine task that most people come to master. Most of us have never learned to draw very well, and we know it. When we were in school, we found it awkward and even embarrassing to try to draw someone’s likeness, and we probably stopped trying long ago. We don’t know how well most other people draw, but we do know there are many good artists in the world, and infer that others probably are, on average, somewhat better than we are. And that’s exactly what we find. Most people think they’re below average. Little do they realize that almost everyone else has the same view.29
The same holds for other tasks that we find difficult, like juggling.30 We don’t really know that others are better, because most of us have never tried to juggle in a group setting, where we could see how frustrating it is for just about everybody. But we’re aware there are some very good jugglers out there, and we know that we’re not one of them, which makes us think we’re worse than average. One study asked US college students to estimate whether they would outperform other students on a quiz about indigenous vegetation in the Amazon basin. Only 6 percent believed they would finish in the top half; the other 94 percent thought they would score below average.31 That’s not what we’d expect if people really had a persistent tendency for overplacement. A simpler explanation is that the way we place ourselves depends on the difficulty of the task and the information we have.32
Once we break down overconfidence into its different parts and take a close look at each one, we shouldn’t conclude that people see themselves as superior to others, period. We’re not overconfidence machines at all. Responses depend on the specific skill in question and on the information we have.33 Rather than claim people are biased, it might be more accurate to say they’re myopic. They see themselves clearly, but have less information about others, and generally make sensible inferences accordingly.
Lumping together three different kinds of overconfidence is convenient, but as Orwell warned us, it can lead to foolish thoughts. As an example, consider a column by David Brooks in the New York Times about health-care legislation. Brooks began by making a simple assertion: “Humans are overconfident creatures.” For evidence, he mentioned some of the studies we have seen again and again: “Ninety-four percent of college professors believe they are above average teachers, and 90 percent of drivers believe they are above average behind the wheel. Researchers Paul J.H. Schoemaker and J. Edward Russo gave computer executives quizzes on their industry. Afterward, the executives estimated that they had gotten 5 percent of the answers wrong. In fact, they had gotten 80 percent of the answers wrong.”34
From there, Brooks claimed that the 2008 financial crisis was the result of overconfidence. Then, after declaring that “the bonfire of overconfidence has shifted to Washington,” he wrote that the Obama administration’s health-care reform was bound to be deeply flawed, because it had been shaped by humans who—as evidence has shown—suffer from overconfidence.
All of this sounds reasonable until we understand that overconfidence is not one thing but several. The studies showing that teachers and drivers rate themselves too highly are examples of overplacement, typical for relatively easy tasks, but not for difficult tasks. As for the computer executives who often gave wrong answers, that’s overprecision. These are very different examples and don’t add up to the sweeping claim that “humans are overconfident,” nor do they justify the assertion that the 2008 financial crisis was caused by overconfidence, unless of course (as so often happens) we’re inclined to blame any failure at all on overconfidence.
As for health-care reform, if the Obama administration had expressed high confidence about the chances of successful reform, that would be overestimation. But studies of drivers and teachers who overplace and computer executives who are overprecise hardly let us conclude that health-care reform is tainted by overestimation. In fact, if health-care reform is difficult—and by all indications it may be very difficult to carry out successfully—the error may be in the opposite direction. We may actually underestimate our ability to bring about this sort of change. Far from undertaking too many difficult projects, we may in fact initiate too few complicated and ambitious projects.
WHAT’S THE RIGHT LEVEL OF CONFIDENCE?
So far we have seen that it’s not helpful merely to infer overconfidence after the fact, when things have gone wrong. Nor is it correct to use one word for three very different things. There’s little reason to suggest that overconfidence is as prevalent as is often claimed.
But we shouldn’t stop there. If overconfidence means “greater confidence than circumstances warrant,” we need to go another step. We saw previously that positive illusions often help improve performance. Believing we can do something may help us do it well. In that case, is a somewhat exaggerated level of confidence really excessive?
The way forward is to remember the topics of the previous chapters. First, can we exert control and influence outcomes, and second, is performance absolute or relative?
For things we cannot influence—the roll of dice, the weather, or the S&P 500—there’s nothing to be gained from overestimation. Any belief that we can control events is excessive. But when we can influence outcomes—whether sinking putts or pedaling a bicycle or carrying out some task—positive illusions can be helpful. A deliberate belief in high performance, maybe even a belief that is somewhat excessive given historical evidence, can boost results.
What’s the best level of confidence? An amount that inspires us to do our best, but not so much that we become complacent, or take success for granted, or otherwise neglect what it takes to achieve high performance. In Dr. Thompson’s cycling experiment (see Chapter Two), for example, athletes could match the avatar when it was made to go 2 percent faster, but couldn’t keep up at 5 percent. That limit surely had to do with oxygen reserves and may have been particular to the task at hand; for other tasks the numbers will vary. Of course knowing that precise balance in advance is difficult, and results will vary among people. Yet the general rule holds. When we can influence outcomes, it can be useful to hold opinions that are somewhat inflated—that is, to overestimate.
When performance is relative, however, the desired level of confidence can only be understood in the context of competition. What’s the best level of confidence? It’s what we need to do better than our rivals.
That’s not to say that a very high level of confidence will guarantee success. It won’t. The performance of our rivals matters, too. Still, when an ability to influence outcomes is combined with relative performance, only those who push themselves to go beyond what seems reasonable will be in a position to succeed. What might seem like an exaggerated level of confidence isn’t just useful, but in the context of competitive rivalry, it is essential.
The problem is, identifying that level of confidence in advance is difficult. There’s no formula we can use, which is why we so often resort to ex-post inferences. When things turn out well, we conclude that our level of confidence was appropriate. We were brimming with healthy confidence. When things turn out poorly, we conclude it was inappropriate—either too much or too little. We suffered from overconfidence—or maybe insufficient confidence. Of course that’s really just an easy way out. Ex ante, things are never so clear. Determining the right level of confidence demands more than a simple comparison to past achievements. We need to consider whether we can exert control over outcomes and also the nature of competition. When performance is relative and highly skewed, a very high level of confidence is not excessive but essential.
ARE WE REALLY OVERCONFIDENT?
At the start of this chapter I quoted Mark Twain’s remark that all we need in life is ignorance and confidence. That’s not a statement to be taken literally, of course. Success is never assured, at least not when it depends on the actions of others, and ignorance isn’t generally something we should recommend.
But as was typical of Mark Twain, he had his finger on a larger truth.35 When confidence can inspire and motivate, what seems excessive by one definition—a level of confidence that exceeds what’s objectively justified—may be useful. And when performance is relative and payoffs are skewed, such a level of confidence may be essential. Those who eventually succeed will have exhibited a level of confidence greater than what was, by some definitions, objectively justified: excessive by some definitions, but not by others.
Yes, there are many examples, not only in laboratory experiments but also in real life, of people exhibiting excessive confidence. They’re often overprecise, they frequently overestimate, and at times they also overplace. Once we tease them apart, however, we find that overprecision is widespread, but overestimation and overplacement aren’t inevitable. In fact, when it comes to difficult tasks, people are more likely to underplace than to overplace. The common image of people as overconfident is justified for routine tasks, but not for many of the greater challenges we face.
Far from people being “overconfidence machines,” I suspect a different interpretation is more accurate. As Henry David Thoreau observed, most people lead lives of quiet desperation. Glance at airport bookshelves, and you’d never guess that people suffer from overconfidence. Much more common are books that encourage us to raise our level of confidence. When I recently checked, I found these titles:
• Self-Confidence: The Remarkable Truth of Why a Small Change Can Make a Big Difference
• Brilliant Confidence: What Confident People Know, Say and Do
• Confidence: The Power to Take Control and Live the Life You Want
• Instant Confidence: The Power to Go for Anything You Want36
These books hardly suggest that the average person is overconfident. Most people seem to want more confidence, not less.* Even the most talented among us are at times prone to self-doubt. German violinist Christian Tetzlaff, one of the great musicians of our day and acclaimed for his original tone and stunning interpretations, remarked: “Most of the time, we tell ourselves ‘I’m confident’ or ‘I’m doing well.’ But then, in a moment alone at home, you feel how close you are to some kind of abyss.”37 We live in a society that is impressed with confidence, and we often try to project confidence because we think others expect it of us. But when we look closely at the evidence, it’s questionable whether people are best described as exhibiting overconfidence.
Why has it been reported so consistently that people are overconfident? One problem has to do with the design of experiments. If we ask for ranges that require 90 percent confidence, we shouldn’t be surprised to find that most errors are in the direction of too much confidence, not too little. If we ask about routine tasks like driving, we find people tend to overplace. When we run experiments with unbalanced designs, we shouldn’t be surprised that the errors are mostly in one direction. The more serious bias isn’t with the answers that are given, but with the questions that are asked.
In turn, this raises a deeper question: Why has research not been balanced? I suspect the answer is that for years, economic theory was based on the notion that people are rational actors, capable of making accurate judgments and sound decisions. We find evidence of biases and errors very interesting because it challenges the prevailing orthodoxy. In particular, we report finding overconfidence because that’s what is most surprising.
There is another reason, too. Some of the most important research about decision making has been conducted by cognitive psychologists, who have been interested in understanding basic mental processes, but to whom questions about competition among companies are not of central interest. We shouldn’t expect psychologists to ask when confidence might be excessive by one definition but useful for competitive purposes. Yet for those of us who are concerned with the world of management, questions of competition are central. We should be careful not to take findings that make sense in one domain and apply them to another—at least not without asking whether the circumstances are the same.
THINKING ABOUT CONFIDENCE . . . AND OVERCONFIDENCE
Early in this chapter, I also quoted psychologist Scott Plous, who wrote more than twenty years ago: “No problem in judgment and decision making is more prevalent and more potentially catastrophic than overconfidence.” By now, claims of overconfidence have been repeated so often that they are accepted as fact. Many recent authors have simply repeated the same phrase without taking a critical look.
Now, as I end this chapter, I suggest we turn the phrase on its head: No concept in judgment and decision making has led to as much erroneous thinking as overconfidence. Sure, almost all failure can be blamed ex post on overconfidence, but that’s not saying much. We know to be skeptical about retrospective attributions. They make good stories, but aren’t valid explanations. As for laboratory research, what we often call overconfidence turns out to be three very different errors: overprecision, overestimation, and overplacement. There’s very strong evidence of the first, but less of the next two. Evidence of overprecision cannot be used as evidence of overestimation or overplacement. The tendency to rate ourselves better than others—overplacement—is better understood as myopia, with most people sensing their abilities clearly and making reasonable inferences about others. Sweeping claims that people are overconfident do not stand up to close scrutiny.
Overconfidence has been taken to mean so many things, and has been used in so many ways, that the very term has been debased. My suggestion is that anyone who uses the term should have to specify the point of comparison. If overconfidence means excessively confident, then excessive compared to what? In much of our lives, where we can exert control and influence outcomes, what seems to be an exaggerated level of confidence may be useful; and when we add the need to outperform rivals, such a level of confidence may even be essential.
* You would be right to question whether a quick check of airport bookstores is the best indicator of how people behave. We can probably find a few books on just about any topic, and the fact that some people want to boost their confidence doesn’t mean that most share that concern. We could even argue that only underconfident people buy self-improvement books, because those who are truly overconfident don’t feel the need to read!