Managerial Overconfidence and Miscalibration

Managerial Overconfidence and Miscalibration

Managerial miscalibration is the empirical finding that corporate executives, particularly CFOs, hold confidence intervals on their own forecasts that are far too narrow relative to realized outcomes. The result, due to Ben-David, Graham, and Harvey ($2013$), distinguishes overconfidence (underestimation of variance, or excessive narrowness of confidence intervals) from the related but distinct over-optimism (overestimation of mean) that drives Malmendier-Tate’s over-optimism results.

Two cognitive biases — distinct mechanisms.

• Over-optimism (Malmendier-Tate): The manager over-estimates the mean of future outcomes — the firm is more valuable than the market believes, projects have higher NPV than the spreadsheet calculation shows. Implication: over-investment, refusal to issue equity perceived as undervalued.
• Overconfidence / miscalibration (Ben-David-Graham-Harvey): The manager under-estimates the variance of future outcomes — her confidence intervals are too narrow. Implication: excessive risk-taking, under-hedging, over-leveraged capital structures, under-pricing of low-probability bad outcomes.

The CFO survey design. Ben-David, Graham, and Harvey ($2013$) use Duke University’s quarterly CFO survey, which asks CFOs to predict next-year S&P $500$ returns and to provide an $80\%$ confidence interval around their prediction. The survey ran from $2001$ to $2011$, generating roughly $13,000$ forecast-interval pairs. The authors compare each CFO’s stated $80\%$ confidence interval to the realized S&P $500$ return.

Headline finding. Realized S&P $500$ returns fall within the CFO’s stated $80\%$ confidence interval only $36\%$ of the time. By definition, a well-calibrated $80\%$ confidence interval should contain the realized outcome $80\%$ of the time. The actual CFO interval thus has roughly half the coverage that calibration would require — confidence intervals are far too narrow.

Magnitude. The average width of the CFO’s stated $80\%$ confidence interval is $15\%$ (annual return). The historical $80\%$ confidence interval for one-year S&P $500$ returns ($90$th minus $10$th percentile of historical returns) is approximately $42\%$. CFOs’ intervals are roughly one-third as wide as historical experience would justify.

Corporate-decision correlations. Miscalibrated CFOs (defined as those whose intervals are particularly narrow relative to peers) make distinctive corporate decisions: they invest more, take on more leverage, and report internal-rate-of-return distributions for firm projects with extremely low anticipated volatility relative to historical comparison firms. The pattern is consistent with overconfidence increasing the manager’s willingness to take risks she misperceives as smaller than they are.

Intuition

The Ben-David-Graham-Harvey result is one of the cleanest direct measurements of cognitive bias in a senior-corporate population. The Duke CFO survey asks executives to do something that should be straightforward — give a range of plausible outcomes for next year’s stock-market return — and the executives’ answers reveal that their probability calibration is dramatically off. The miscalibration is not a small effect at the margin; the realized-outcome coverage is less than half what calibration requires.

The conceptual distinction between over-optimism and miscalibration is important. An over-optimistic CEO who over-estimates her firm’s value will systematically miss high: actual outcomes are below her expectations. A miscalibrated CEO who has overly-narrow confidence intervals will miss in both directions — actual outcomes are sometimes above her interval and sometimes below. The two biases produce different investment signatures: over-optimism leads to over-investment in the direction of optimism; miscalibration leads to risk-taking without a directional bias.

The two biases often coexist in the same manager. Malmendier-Tate’s Holder67 CEOs are likely to be both over-optimistic and miscalibrated. The empirical literature has begun to disentangle the two, with Ben-David-Graham-Harvey’s CFO data providing the cleanest direct measurement of miscalibration and Malmendier-Tate’s option-exercise proxies providing the cleanest measurement of optimism. The two corporate-decision implications are partially overlapping (both lead to elevated investment) and partially distinct (only miscalibration leads to elevated risk-taking and leverage independently of investment direction).

Properties

(80% Interval Coverage Is 36%) CFO confidence intervals contain the truth less than half as often as calibration requires.

The headline finding from $\sim 13,000$ Duke CFO survey forecast-interval pairs from $2001$ to $2011$. CFOs are asked for $80\%$ confidence intervals on next-year S&P $500$ returns. Realized returns fall within those intervals only $36\%$ of the time. The shortfall is statistically overwhelming and economically large; CFOs’ probability assessments are not just imperfect but deeply miscalibrated. The pattern holds across sub-samples (different industries, different firm sizes, different individual CFOs) and across the full $2001$–$2011$ period.

(Intervals Are One-Third the Required Width) CFO 80% intervals are $15\%$ versus historical $42\%$.

The average width of a CFO-stated $80\%$ confidence interval on annual S&P $500$ returns is approximately $15$ percentage points. The historical record of S&P $500$ returns has $80\%$ of annual realizations falling within an interval of about $42$ percentage points. CFOs’ intervals are thus about one-third the width of what historical variation alone would justify. Even if the future is somehow less volatile than the historical past, the gap is too large to attribute to genuine future-volatility expectations.

(Distinct From Over-Optimism) Miscalibration is variance-underestimation, not mean-overestimation.

The conceptual distinction matters. Over-optimism implies the manager misses high: outcomes are systematically below her central forecast. Miscalibration implies the manager underestimates the spread of plausible outcomes: outcomes are sometimes far below and sometimes far above her central forecast. Ben-David-Graham-Harvey verify this distinction by checking the directional bias in CFO point forecasts. CFO point forecasts of S&P $500$ returns are roughly unbiased on average; the failure is in the confidence-interval width, not in the central forecast.

(Project-Level IRR Distributions) CFOs report extremely low anticipated volatility for firm projects.

The survey also asks CFOs about their internal-rate-of-return projections for typical firm projects. They report distributions with very low anticipated volatility — far lower than the historical volatility of comparable firms’ realized returns on capital. The same pattern appears at the firm-internal level as at the market-aggregate level, suggesting that miscalibration is a general cognitive feature of corporate decision-making rather than specific to forecasting external markets.

(Correlations With Corporate Decisions) Miscalibrated CFOs invest more and take more leverage.

CFOs whose stated confidence intervals are particularly narrow (relative to peer CFOs in the same period) make distinctive corporate decisions:

• Higher investment-to-assets ratios.
• Higher leverage (debt-to-equity).
• Higher firm-project investment-cash-flow sensitivity.
• Greater willingness to undertake projects with reported high IRRs.

The correlations are consistent with miscalibration making the CFO perceive risk as lower than it is, raising her willingness to commit firm resources to risky undertakings.

(Mechanism — Underestimation of Distribution Tails) CFOs underestimate low-probability bad outcomes.

The narrowness of CFO confidence intervals implies systematic underestimation of tail outcomes — the worst $10\%$ and the best $10\%$ of possible realizations. The bias is asymmetric in welfare consequences: underestimating bad tails leads to under-hedging, over-leveraging, and excessive risk-taking; underestimating good tails leads to missing valuable optionality. The corporate-decision correlations focus on the bad-tail consequences because those are more easily measured (leverage, hedging behavior) and more consequential during financial crises.

(Robustness Across Survey Samples) Pattern replicates in CFOs across countries and time.

Ben-David-Graham-Harvey’s U.S. CFO sample finding has been replicated in international CFO surveys. The miscalibration is therefore not a U.S.-specific or $2001$–$2011$-specific phenomenon; it appears to be a general cognitive feature of senior corporate executives. The robustness suggests structural rather than idiosyncratic causes — perhaps selection of overconfident individuals into senior corporate roles, perhaps career incentives that reward apparent confidence, or perhaps a deeper psychological tendency among high-status decision-makers.

Examples

A CFO survey response in mid- $2008$

A CFO surveyed in June $2008$ predicts next-year S&P $500$ return of $+5\%$ with $80\%$ confidence interval $[-3\%,+13\%]$. The interval is $16$ percentage points wide — close to the survey average. Realized return from June $2008$ to June $2009$ was approximately $-30\%$ (during the financial crisis). The realized outcome falls $27$ percentage points outside the lower edge of the CFO’s interval. By the Ben-David-Graham-Harvey criterion, the CFO is severely miscalibrated. The corporate-decision implication is that a firm whose CFO held this miscalibrated view would have been under-hedged against tail risk going into the crisis — a pattern observed across many over-leveraged firms in $2008$.

The cumulative density of realized returns

If CFOs were well-calibrated $80\%$ interval-providers, then $20\%$ of realized returns should fall outside the interval — half above, half below. In the Ben-David-Graham-Harvey data, $64\%$ of realized returns fall outside the interval. The shortfall is staggering: confidence intervals would need to be approximately $3\times$ wider on average to achieve calibrated coverage. The CFO population is not just slightly overconfident; it is overconfident by a large margin in a way that has direct implications for the prudence of corporate financial decisions.

Distinguishing over-optimism from miscalibration in the same firm

A firm has both a CEO and CFO. The CEO is classified as over-optimistic by Holder67 (refused to exercise deeply in-the-money options); the CFO is classified as miscalibrated by Ben-David-Graham-Harvey criteria (very narrow confidence intervals). The over-optimistic CEO pushes for over-investment in projects she perceives as high-NPV. The miscalibrated CFO endorses the projects with high reported IRRs and low anticipated volatility, and approves a financing plan with high leverage because she underestimates default risk. The combined effect is much larger over-investment and risk-taking than either bias alone would produce. The empirical literature has begun to study CEO-CFO interaction effects; preliminary evidence suggests the combination is super-additive.

See Also

• 8.13. Behavioral Corporate Finance — parent hub.
• 8.13.5. Managerial Over-Optimism — sibling: the related but distinct over-optimism bias.
• 8.13.7. Single Firm-Wide Discount Rate — sibling: another bounded-rationality pattern in corporate decisions.
• 8.6. Belief-Based Biases — broader category of belief biases that miscalibration belongs to.
• 8.6.5. Representativeness Heuristic — related cognitive bias that contributes to overly-narrow confidence intervals.
• 5. Corporate Finance — broader corporate-finance hub.