Overconfidence Bias in Investing: Miscalibration, Better-Than-Average & Illusion of Knowledge

Published: March 31, 2026

Article by Ryan O'Connell, CFA, FRM

Overconfidence bias in investing is one of the most consistently documented emotional biases in behavioral finance. It causes investors to overestimate their knowledge, assign overly narrow confidence intervals to forecasts, and trade far more than is optimal — costing measurable percentage points in annual returns. This guide unpacks the three mechanisms through which overconfidence operates: miscalibration, the better-than-average effect, and the illusion of knowledge. You will learn the landmark evidence from Barber and Odean’s trading frequency studies, how self-attribution compounds overconfidence over time, and practical techniques for detecting and correcting the bias in your own portfolio decisions.

What Is Overconfidence Bias in Investing?

Michael Pompian’s behavioral finance framework classifies overconfidence as an emotional bias — even though it stems from cognitive weaknesses, it produces emotionally charged behaviors like excessive risk-taking and refusal to sell losing positions. Pompian identifies two core forms:

• Prediction overconfidence — assigning confidence intervals to forecasts that are too narrow
• Certainty overconfidence — holding judgments with excessive certainty, becoming blind to disconfirming information

In practice and in search literature, these map to three labels that capture distinct mechanisms:

Understanding which mechanism is operating helps target the right correction. For a broader view of how cognitive and emotional biases differ, see our cognitive bias investing overview.

Miscalibration: Confidence Intervals That Are Too Narrow

Miscalibration occurs when the confidence intervals you assign to predictions are systematically too narrow. When you say you are 90% confident a stock will return between 5% and 15%, but the actual outcome falls outside that range more than 10% of the time, you are miscalibrated.

Calibration research by Lichtenstein, Fischhoff, and Phillips (1982) documented this pattern across thousands of general-knowledge questions:

The pattern is consistent: actual accuracy falls systematically short of stated confidence at every level. This is not unique to novices — professionals exhibit the same gap.

Better-Than-Average Effect

Svenson’s 1981 survey of drivers remains the classic demonstration: over 80% of respondents rated their driving skill and safety above the median. Similar patterns appear across domains including academic ability, job performance, and investing.

Among professional fund managers, surveys consistently show that a majority believe their investment skill exceeds the median manager — yet by definition, only 50% can be above average. This collective overplacement fuels the active management industry and explains why so many investors pay premium fees for strategies that statistically cannot all outperform.

Gervais and Odean’s model “Learning to Be Overconfident” (2001) formalizes this dynamic: traders who experience early success take excessive credit, become overconfident, and trade more aggressively — ultimately lowering their expected profits while increasing market volatility.

Illusion of Knowledge: More Information Does Not Improve Decisions

The illusion of knowledge occurs when acquiring more information increases your confidence without proportionally improving your accuracy. You feel more certain, but you are not actually making better predictions.

Oskamp’s 1965 clinical psychology experiment provides the landmark evidence. Psychologists were given progressively more information about a patient case — first a brief summary, then increasingly detailed history, then full records. As information increased:

• Confidence rose from ~33% to ~53%
• Diagnostic accuracy rose only from ~26% to ~28%

The additional information made participants feel more certain without making them meaningfully more accurate. For investors, the parallel is direct: reading more analyst reports, consuming more financial news, and studying more earnings transcripts raises conviction about trade theses without genuinely improving forecast accuracy.

The Self-Attribution Loop: How Overconfidence Compounds Over Time

Self-attribution bias is the mechanism that sustains and amplifies overconfidence. It operates through two components:

• Self-enhancing bias — attributing successes to your own skill and foresight
• Self-protecting bias — attributing failures to external factors, bad luck, or market irrationality

Each success ratchets confidence upward. Each failure is discarded as non-representative. The net result: confidence increases over time regardless of actual performance. Gervais and Odean’s research shows that overconfidence peaks during early career success periods, when the feedback loop is most active.

Pompian identifies four harmful portfolio behaviors that result from sustained overconfidence:

A classic example from Pompian: former executives or family legacy stockholders who refuse to diversify concentrated positions in companies like AT&T, claiming “insider knowledge” of the business. The conviction persists even as the stock declines, because self-attribution filters out disconfirming evidence.

Trading Frequency and Portfolio Returns: The Barber-Odean Evidence

The most cited evidence on overconfidence’s portfolio cost comes from Brad Barber and Terrance Odean’s study “Boys Will Be Boys” (2001). They analyzed 35,000 household brokerage accounts from 1991-1997 at a large discount brokerage firm.

The least active investors outperformed the market benchmark. The most active investors underperformed it by nearly 7 percentage points annually. Both groups held the same types of stocks in the same market environment — the primary observable difference was trading frequency, which the authors interpret as consistent with overconfidence.

The study also documented a gender differential: men traded 45% more frequently than women, and single men traded 67% more than single women. This excess trading reduced men’s net returns by approximately 2.65 percentage points annually and women’s returns by 1.72 percentage points — a gap of roughly one percentage point attributable to the higher male trading frequency.

For strategies that eliminate the overconfidence expression channel entirely, consider index fund investing and dollar-cost averaging.

Overconfidence in Analyst Forecasts

Overconfidence is not limited to retail investors — professional equity analysts exhibit the same patterns. Analyst earnings forecast confidence intervals are systematically too narrow: actual earnings outcomes fall outside the stated range more often than the stated confidence level would predict.

Research has documented that analyst earnings forecasts tend to be excessively optimistic and insufficiently responsive to negative information. When positive news arrives, forecasts adjust upward quickly; when negative news arrives, adjustments are slower and smaller. This asymmetry is consistent with overconfidence and the self-attribution mechanism.

Forecast herding amplifies the problem: analysts cluster consensus estimates too tightly to avoid career risk from outlier predictions. When the consensus is wrong — as during the dot-com era or the 2008 financial crisis — the collective miss is large because individual confidence intervals overlapped almost completely.

The Barber-Odean evidence on investor trading behavior extends this pattern: the stocks individual investors chose to buy underperformed the stocks they chose to sell by 20 basis points per month for men and 17 basis points for women. Investors traded as if they had informational edges that did not actually exist.

How to Correct Overconfidence Bias in Investing

Pompian’s correction framework addresses each of the four harmful behaviors:

1. Poor stock selection (blind evaluation): Retrieve your actual trading records for the past 24 months and calculate realized returns. Pompian notes that “more often than not, the trading activity will demonstrate poor performance.” Odean’s 1999 study found the average investor underperformed the market by approximately 2% annually after costs.

2. Excessive trading: Track each trade with your thesis stated in advance. Record your predicted outcome, stated confidence level, and time horizon before you act. After 6-12 months, score your accuracy. This exercise converts the abstract critique (“you trade too much”) into quantified evidence (“your average trade lost X basis points net of costs”).

3. Underestimating downside risk: Use historical volatility data and maximum drawdown statistics to anchor your expectations. If you believe a stock has a 90% chance of being up next year, compare that stated probability with the historical base rate of annual positive returns for stocks with similar characteristics.

4. Underdiversification: Apply the “fresh-money question” for concentrated positions: “If I did not already own this stock, would I buy this much of it today?” When the answer is no, the overconfidence-driven anchor has been identified. Hedging strategies like costless collars or protective puts can reduce concentration without triggering immediate liquidation.

Overconfidence Bias vs. Illusion of Control

Overconfidence and illusion of control are related but distinct biases. Distinguishing them matters because they require different corrections.

The two biases frequently co-occur and reinforce each other. An investor exhibiting both will believe their analysis is more accurate than it actually is (overconfidence), and believe that their constant monitoring and portfolio adjustments directly improve outcomes (illusion of control). The combination produces the highest observed trading frequencies and the worst return outcomes. For the full illusion of control framework, see illusion of control in investing.

Limitations of Overconfidence Research

1. Context-dependence: Overconfidence is most pronounced in domains where feedback is delayed or ambiguous (investing, long-term planning) and least pronounced where feedback is immediate and unambiguous (chess, athletics). Investors who have experienced multiple market cycles may show partial calibration improvement.

2. Awareness does not eliminate the bias: The debiasing literature is mixed. Knowing about overconfidence does not reliably reduce it. The bias operates at a level that persists despite cognitive recognition — which is why Pompian’s advice focuses on process changes rather than education alone.

3. Distinguishing overconfidence from genuine expertise: A CFA charterholder or sector specialist may have a genuine informational edge in a narrow domain. The overconfidence framework should not be misapplied to delegitimize all active analysis. What matters is whether confidence intervals are accurately calibrated, not whether the individual is confident at all.

4. Measurement challenges: Most evidence comes from lab experiments or brokerage data from the 1990s. The modern information environment — algorithmic pricing, broader retail access to professional research, commission-free trading — may have shifted some dynamics. However, recent evidence from retail trading during 2020-2021 suggests overconfidence remains prevalent.

Common Mistakes

1. Using trading frequency as the only indicator of overconfidence. Many readers interpret the Barber-Odean findings as “low turnover = not overconfident.” This is incorrect. Overconfidence can manifest as underdiversification (concentrated conviction positions with infrequent trading), as systematically narrow confidence intervals in financial plans, or as false certainty about analyst research. Assess all four harmful behaviors, not just turnover.

2. Conflating overconfidence with illusion of control. These are distinct biases with different mechanisms. Overconfidence is about the accuracy of your judgments. Illusion of control is about believing your actions influence random outcomes. Conflating them leads to mismatched corrections.

3. Assuming that recognizing overconfidence eliminates it. Research consistently shows that awareness of a bias does not reliably reduce its effect on behavior. Investors who understand miscalibration intellectually still produce interval estimates that are too narrow when tested empirically. Implement structural debiasing tools — prospective tracking, decision audits, investment policy statements — not just education.

4. Mistaking good outcomes for well-calibrated decisions. A bet placed with 70% stated confidence that turns out correct does not prove you were well calibrated. Calibration is only assessable over a large sample of predictions. Investors who made money in the 2010-2021 bull market frequently attribute their returns to skill rather than the rising tide. This is the self-attribution feedback loop in action.

5. Equating more information with better forecasts. The illusion of knowledge research demonstrates that consuming more data increases confidence faster than it improves accuracy. Before acting on additional research, ask whether the new information genuinely shifts outcome probabilities or merely reinforces existing beliefs.