with Todd E. Clark (Johns Hopkins University, Federal Reserve Bank of Cleveland)

• paper: pdf

• online appendix: pdf

• slides: pdf

• supplementary material: https://github.com/elmarmertens/clark-mertens-entropic-tilting-spf/tree/main

Abstract: We develop a new approach to incorporating information from survey density forecasts into model-based predictive distributions. Histogram forecasts from the U.S. Survey of Professional Forecasters (SPF) provide rich, nonparametric information about expected outcomes, yet existing approaches typically rely on moment-based approximations that discard part of this information. We instead propose a direct entropic tilting framework that matches histogram probabilities exactly, without relying on parametric approximation. We first reformulate the single-histogram tilting problem and then derive a novel analytic characterization for the multiple-histogram case, with iterative solutions based on Iterative Proportional Fitting. The resulting framework provides a tractable and broadly applicable method for incorporating multiple survey densities into multivariate predictive systems. Applying the method to real-time forecasts from a Bayesian vector autoregression with time-varying volatility, we find that tilting to SPF histograms substantially improves forecast performance relative to the model's baseline forecasts, particularly during periods of heightened uncertainty such as the Great Recession and the COVID-19 pandemic. Importantly, the gains extend beyond the variables directly targeted by SPF histograms, improving forecasts for other variables in the system as well. This result highlights how survey density information can propagate through multivariate forecasting models and improve predictive accuracy more broadly.

De Polis, Melosi and Petrella: "The Taming of the Skew: Asymmetric Inflation Risk and Monetary Policy" (@ESCBcluster1, 2026)

• • My discussion slides: pdf

  • their  paper: pdf

  • discussed at the ESCB cluster 1 workshop, Banque de France, Mar 26-27, 2026