Additional predictive value of disagreement beyond visible asset risk

Determine whether forecast disagreement among the seven-framework AI council predicts additional return error beyond historical asset volatility and a no-change baseline.

Background

The paper finds that numerical disagreement is strongly associated with subsequent absolute return error, but the same pattern appears for a simple forecast of no price change. Disagreement is also correlated with historical volatility, which provides an observable measure of asset risk before forecast generation.

After controlling for specified risk and other covariates, the estimated association between disagreement and excess error remains inconclusive. The unresolved issue is whether disagreement contains a distinct reliability signal or merely identifies assets whose outcomes are intrinsically more variable. This distinction matters before dispersion can be used as a confidence or forecast-selection measure.

References

The relationship between disagreement and additional forecast error is not resolved after accounting for visible asset risk.

— When AI Advisors Agree: Distinguishing Forecast Risk from Forecast Reliability  (2610.13431 - Ramdowar, 9 Oct 2026) in Abstract; Section 3.2, “Disagreement tracks risk visible before generation”

Generalization to other models, markets, horizons, or regimes remains untested.

— When AI Advisors Agree: Distinguishing Forecast Risk from Forecast Reliability  (2610.13431 - Ramdowar, 9 Oct 2026) in Limitations, paragraph “One system and few periods”