A September 2026 Finance and Economics Discussion Series paper by Edward Herbst and Karen Page examines how coefficients in Taylor-rule regressions should be pooled across forecasters, dates, and horizons when using multiple-horizon forecast panels. The authors treat the pooling structure itself as the object of inference, comparing pooling patterns through Bayesian marginal likelihoods in a participant-date-horizon Taylor-rule regression model.
The framework is applied to three forecast sources: the Blue Chip Financial Forecasts, the Survey of Professional Forecasters, and the Summary of Economic Projections. The preferred specifications place much of the systematic variation in policy-rate forecasts in intercepts that vary across forecast horizons and survey dates. The paper is published as FEDS 2026.064 with DOI 10.17016/FEDS.2026.064.
The central implication is that evidence of changing perceptions of monetary policy through economically meaningful time-varying response coefficients is weak overall. This suggests that apparent shifts in forecasters' perceived policy rules may largely reflect horizon- and date-specific level adjustments rather than changes in how forecasters respond to inflation or output gaps.
What remains unknown is how robust these findings are outside the three surveyed panels or under alternative model specifications. The authors note the research represents their own views and is preliminary, circulated to stimulate discussion and critical comment rather than indicating concurrence by other Board staff or the Board of Governors.