How does overconfidence bias specifically distort CME confidence intervals, and what's the evidence that analysts get this wrong?

Overconfidence bias is arguably the most dangerous psychological bias in CME because it affects not just the point estimate but the entire distribution of possible outcomes. And the evidence is overwhelming: professionals are just as susceptible as amateurs.

The Calibration Evidence:

Research consistently shows:

• When professionals report 90% confidence intervals, the true outcome falls outside the range roughly 40-50% of the time
• Even after being warned about overconfidence, subjects only slightly widen their intervals
• Experts in a field are often MORE overconfident than non-experts (expertise increases confidence faster than it increases accuracy)

Example — Blackridge Asset Management:

Blackridge's CIO provides the following 2026 CMEs with '90% confidence intervals':

The CIO's confidence intervals are roughly one-third the width of what history suggests is appropriate. The reported ranges imply a world of mild variation; reality includes crashes, bubbles, and regime shifts.

Two Dimensions of Overconfidence:

1. Known unknowns — risks you're aware of but underweight. The CIO knows that recessions, geopolitical shocks, and policy errors are possible but assigns them implicitly low probability in the narrow interval.

1. Unknown unknowns — risks you haven't even considered. COVID-19 in early 2020, the SVB banking crisis in 2023, or sudden regulatory shifts in specific markets. By definition, these can't be captured by scenario analysis, which is why confidence intervals need to be wider than any explicit scenario set.

Why Professionals Are Vulnerable:

• Illusion of knowledge: More information creates the feeling of better forecasts, but beyond a point, additional information adds noise rather than signal
• Illusion of control: Active management creates the feeling that you can respond to emerging risks, but portfolio adjustments take time and markets move first
• Track record illusion: A few correct calls build confidence, even though the base rate of correct macro forecasts is low

Practical Defenses:

1. Historical calibration: Compare your past confidence intervals to actual outcomes. If outcomes fell outside your '90% range' more than 10% of the time, you're overconfident.
2. Pre-mortem analysis: Before finalizing CMEs, ask 'Assume this forecast turned out to be spectacularly wrong. Why?' This forces consideration of scenarios you would otherwise ignore.
3. Mechanical widening: After setting your initial confidence interval, systematically widen it by 50-100%. Research shows this crude adjustment actually improves calibration.
4. Diverse perspectives: Solicit forecasts from people with different information sets and frameworks. Aggregating diverse views typically produces better-calibrated intervals.

Practice overconfidence questions in our CFA Level III question bank.