OVERCONFIDENT? NO, MORE OFTEN WE SELL OURSELVES SHORT, SAYS NEW ECONOMIC RESEARCH
01 Jan 2009
Does everybody really think they are above average? New research by Dr Jeremy Clark and Dr Lana Friesen, published in the January 2009 issue of the Economic Journal, finds that when it comes to judging our own abilities, we are not as overconfident as is commonly thought. In fact, we are more likely to be right on average or even sell ourselves short.
The researchers asked 238 individuals to make predictions about their performance in two initially unfamiliar tasks; searching for the highest possible number hidden on a computer spreadsheet; and decoding five- letter words. Comparing people''s predictions with their actual performance revealed that in four out of eight cases, predictions were&##160;accurate on average. The next most common outcome (three out of eight cases) was under-confidence about the outcome, and in only one in eight, overconfidence.
The researchers comment:
''Self-assessment of ability is an important factor in many economic decisions, such as choosing a career, pursuing higher education, starting a new business or selecting investments.
''If people''s self-assessments are overconfident, as psychological evidence suggests, costly economic mistakes could be made, such as new business failures.
''But the results of our study suggest that overconfidence may not
be ubiquitous in our economic decision-making.''
The way the experiment worked was as follows: to succeed at the tasks, half of the participants had to outperform others in their group, while the other half had to achieve specific performance targets. Participants were rewarded for their performance in the tasks, and some were also rewarded for accurately forecasting how well they would do. They were given the opportunity to update their forecasts after experience with the tasks and feedback about their performance.
In psychological studies suggesting overconfidence, people have been asked to rate their likelihood of having various good or bad outcomes in the indefinite future relative to their peers. Others have been asked to rate their general attributes relative to their peers. From these studies came the famous findings of ''optimism bias''; most people think they are better-than-average drivers, or have above average chances of having&##160;higher salaries, job security, successful marriages, good health and so on.
In contrast, this study asked people to make predictions about something specific and unambiguous: the number of times they would be successful at a task. The accuracy of these predictions could then be immediately evaluated. Participants were given immediate feedback about the accuracy of their predictions, and the opportunity to learn and to revise them. The researchers argue that these conditions are more relevant for assessing overconfidence in most economic decisions.
The researchers compared the predictions that individuals made with their actual performance and found surprising results. They found overconfidence on average in only one of eight possible cases (of two possible tasks in two possible orders, with relative or absolute&##160;performance criteria).
Accuracy of prediction was the most common outcome at four out of eight cases (meaning an equal prevalence of overly low and high predictions). Under-confidence, or under-selling yourself, was the next most common outcome, at three out of eight cases.
People''s self-assessments also depended on the task. Participants had limited overconfidence in their ability to find a large number on the spreadsheet, but substantial under-confidence in their ability to decode words. Participants also had an easier time forecasting their relative than their absolute performance.
''Overconfidence in Forecasts of Own Performance'' by Jeremy Clark and Lana Friesen is published in the January 2009 issue of the Economic Journal.
Department of Economics, University of Canterbury | +643 364 2308 | firstname.lastname@example.org
School of Economics, University of Queensland | +617 3365 6564 | email@example.com