Semi-structural model enhances professional inflation forecasts by 50 percent
A Bank of Canada staff working paper introduces a semi-structural model that significantly improves professional inflation forecasts. The model, which explicitly accounts for information frictions, reduces mean squared errors by 50 percent in pseudo-real-time tests.
Unlocking hidden signals in forecaster data
The paper develops a semi-structural state-space model to extract improved inflation forecasts from the US Survey of Professional Forecasters (SPF) data.
It explicitly models information frictions, finding that forecast rigidity systematically increases with the forecast horizon, from near zero for backcasts to 0.81 beyond two quarters.
In pseudo-real-time tests covering 2006-2023, the model's "Resetting Nowcasts" reduce mean squared errors by approximately 50 percent compared to SPF averages, demonstrating significant accuracy gains at short horizons.
The research also introduces a novel theoretical criterion that identifies when improved forecasts will dominate survey aggregates.
This criterion relies on easily computable sufficient statistics from any survey microdata, such as the unconditional variances of individual forecast errors, average forecast errors, and inflation, making it broadly applicable without estimating friction parameters.
Beyond the headline inflation numbers
Monetary policy operates with significant lags, making it crucial for policymakers to distinguish persistent from transitory inflation movements.
While professional forecasts offer valuable signals, previous research consistently documents information frictions that prevent strict rationality.
This paper addresses this by showing how modeling these frictions can extract more useful information.
The study's inflation decomposition also provides policy-relevant insights, revealing that only a small portion of the 2021-2023 US inflation surge reflected permanent shifts in trend inflation.
Most of the surge was attributed to medium-run transitory components, a finding with important implications for appropriate monetary policy responses.
A practical tool for central banks
This research offers a genuinely practical tool for central banks seeking to refine their understanding of inflation expectations.
The ability to identify optimal forecast horizons without complex estimation is a significant methodological advance, promising improved accuracy.
While the empirical focus is on US data, the generalizable framework holds substantial promise for diverse economies.