Overconfidence Calibration: When Certainty Misleads

Key Takeaways

• Overconfidence calibration compares how confident you feel with how often you are correct.
• Overprecision, setting confidence intervals too narrow, is the most reliable form of overconfidence.
• Ninety percent ranges often contain the true value only about half the time, far below the claimed 90.
• Forcing wider ranges and tracking forecast hit rates are the practical fixes for poor calibration.

What It Is

Calibration measures the fit between confidence and accuracy. You are well calibrated if the things you are 90% sure of turn out true about 90% of the time. Overconfidence in this sense is being more certain than your accuracy warrants.

Researchers split overconfidence into three forms. Overestimation is thinking you performed better than you did. Overplacement is thinking you are better than others. Overprecision is being too sure your estimates are accurate, and it is the form most directly about calibration. Moore and Healy describe overprecision as the most reliable and persistent of the three.

The Intuition

Good decisions need not just a best guess but an honest sense of how wrong that guess could be. Calibration is about the width of your uncertainty, not the center of it.

The trouble is that uncertainty is uncomfortable, and a confident, narrow estimate feels more competent than a wide, hedged one. So people shrink their ranges. The center of the guess might be fine, but the range around it is too tight, leaving them blindsided when reality lands outside the band they swore it would stay in.

How It Works

The standard test asks for a 90% confidence interval around an unknown quantity, for example, the range you are 90% sure contains a company's earnings next year. If you are calibrated, the true value should fall inside your range 90% of the time. Across many studies, it falls inside only about 40 to 60% of the time. The intervals are far too narrow.

well calibrated:  90% ranges contain the truth ~90% of the time
overprecise:      90% ranges contain the truth ~40-60% of the time

This overprecision is unusually durable. While overestimation and overplacement come and go depending on task difficulty and framing, overprecision shows up consistently, which is why some researchers call it the most important of the overconfidence types. The CFA Institute notes that overconfident investors assign overly narrow confidence intervals to their forecasts and trade more than is optimal as a result. Research also shows that how you build the interval matters: methods that force you to consider each bound separately, or to assign probabilities to fixed ranges, produce better calibration than asking for a single tidy interval.

Worked Example

An analyst forecasts a company's earnings per share for next year. Their point estimate is 5.00 dollars, and they give a 90% range of 4.80 to 5.20.

That band is only 8% wide around the center. For a single firm a year out, with macro surprises, demand swings, and cost shocks all possible, real outcomes vary far more than that. When earnings come in at 4.40, the result sits well outside a range the analyst was 90% sure would hold.

Now picture this across a portfolio of forecasts. If every 90% range is this tight, the analyst will be surprised on roughly half their calls instead of one in ten. Position sizes and risk limits built on those ranges will be too aggressive, because they assume an accuracy the forecasts do not have. Widening each range to honestly reflect uncertainty, say 4.20 to 5.80, would size risk correctly even though it feels less impressive.

Common Mistakes

1. Setting ranges that feel smart instead of honest. A narrow band signals competence socially but fails calibration. Width should reflect real uncertainty, not confidence theater.

2. Anchoring the range on the point estimate. Building bounds as a small cushion around your guess guarantees overprecision. Set each bound by asking what would have to be true to hit it.

3. Never scoring your forecasts. Without tracking how often outcomes land inside your ranges, you cannot know if you are calibrated. Keep the record.

4. Sizing risk on overprecise inputs. If your inputs are too certain, your position sizes and stops will be too aggressive. Calibration errors compound into risk errors.

5. Assuming expertise fixes it. Overprecision appears even among specialists. Domain knowledge improves the center of the estimate more than the honesty of the range.

Frequently Asked Questions

What is overconfidence calibration in simple terms? Overconfidence calibration is checking whether your confidence matches how often you are actually right. If the things you are 90% sure of come true only half the time, your calibration is poor.

How does overconfidence calibration affect investment decisions? Poor calibration, especially overprecision, makes forecast ranges too narrow, so you are surprised far more often than you expect. As the earnings example shows, that leads to risk limits and position sizes that are too aggressive.

What is a real-world example of poor calibration? An analyst gives a 90% earnings range of 4.80 to 5.20 dollars, but actual earnings come in at 4.40, outside the band, because the range was far too narrow to reflect real uncertainty.

How can investors improve calibration? Deliberately widen your confidence ranges, set each bound separately rather than as a cushion around your guess, and keep a log scoring how often outcomes fall inside your ranges over time.

How is overconfidence calibration different from the better-than-average effect? Calibration is about overprecision, being too sure your own estimates are accurate. The better-than-average effect is overplacement, believing you rank above others. One is about your accuracy, the other about your ranking.

Sources

1. Moore, D. A., & Healy, P. J. "Overprecision in Judgment." https://learnmoore.org/mooredata/HOC.pdf
2. Cambridge / Judgment and Decision Making. "Effect of confidence interval construction on judgment accuracy." https://www.cambridge.org/core/journals/judgment-and-decision-making/article/effect-of-confidence-interval-construction-on-judgment-accuracy/BC53996661339539048C43C1D3905FB1
3. PMC. "A simple remedy for overprecision in judgment." https://pmc.ncbi.nlm.nih.gov/articles/PMC5386407/
4. CFA Institute. "The Behavioral Biases of Individuals." https://www.cfainstitute.org/insights/professional-learning/refresher-readings/2026/the-behavioral-biases-of-individuals

Disclaimer

This article is educational content only and is not financial advice. Nothing here is a recommendation to buy, sell, or hold any security. Consult a licensed advisor before making investment decisions.