Uncertain and Asymmetric Forecasts

TL;DR

This paper develops methods to accurately measure inflation forecast uncertainty and risk directionality by correcting for contamination from expected inflation levels, revealing their true economic implications.

Contribution

It introduces two correction techniques, Normalized Uncertainty and Asymmetry Coherence, to improve the measurement of forecast uncertainty and risk directionality.

Findings

Contamination from expected inflation affects raw measures of uncertainty and asymmetry.

Corrected measures alter economic inferences, such as the significance of inflation effects and policy responses.

Higher moments are only informative when measurement issues are addressed.

Abstract

Measures of inflation uncertainty and directional risk derived from higher moments of forecast distributions are contaminated by the first moment, but in distinct ways. Using individual density forecasts from the ECB Survey of Professional Forecasters, this paper shows that 42% of the variation in raw forecast variance reflects the distance of expected inflation from target, a mechanical level effect, while raw asymmetry is too noisy to identify directional risk without reference to the central forecast. We propose two complementary corrections. Normalized Uncertainty (NU) purges dispersion of its predictable component linked to the policy anchor, recovering genuine belief imprecision. Asymmetry Coherence (AC) extracts directional risk only when asymmetry aligns with the central forecast, formalizing the balance of risks. These corrections alter inference. In a replication of Barro (1995), the volatility effect on growth disappears once level contamination is removed, while the inflation-level coefficient regains significance. In a VAR, the sign of the policy response reverses: raw asymmetry suggests easing, whereas coherent upside risk predicts tightening. In the credit channel, higher uncertainty slows and weakens pass-through from policy easing to loan pricing, especially at longer maturities. A division of roles emerges: NU governs transmission, AC informs policy response. Higher moments are informative only when measurement separates macroeconomic signals from first-moment contamination.