Uncertainty measures improve recession forecasting

Researchers at the Federal Reserve Bank of St. Louis find that economic uncertainty measures significantly enhance real-time recession forecasting. Their study shows that including uncertainty improves the predictive power of forecast models over baseline financial variables.

Uncertainty's varied predictive power

Researchers evaluated the ability of various economic uncertainty measures to forecast recessions in real time, comparing their predictive power against a baseline set of financial variables.

The study found that including uncertainty significantly enhances recession prediction, as measured by the area under the ROC or precision-recall curve.

The best-performing uncertainty index varies by forecast horizon: the categorical monetary policy uncertainty index (MPU) is most effective at short horizons, while the aggregate Economic Policy Uncertainty (EPU) index shows greater predictive ability at longer horizons.

A nonlinear maximum transformation of uncertainty, which captures when a measure exceeds its past year's maximum, also improved forecast performance for some measures.

The paper concludes that combining uncertainty information, especially through ex post Bayesian model averaging, yields the best predictive model.

Modeling recession probabilities

The study employs binary outcome models, specifically probit regression, to correlate financial and macroeconomic data with NBER business cycle turning points.

Baseline predictors include the S&P 500 percentage change, term spread, 1-year inflation-adjusted T-bill rate, and corporate bond risk.

Real GDP is also incorporated in some specifications, using real-time vintages from the Federal Reserve Bank of Philadelphia.

Uncertainty measures include the overall Economic Policy Uncertainty (EPU) index and its components, such as the Monetary Policy Uncertainty (MPU) categorical index, alongside the Jurado, Ludvigson, and Ng (JLN) uncertainty index, constructed with real-time vintages.

Models are estimated separately for each prediction horizon using a direct multistep forecasting approach.

Uncertainty's clear signal

This paper offers compelling evidence that economic uncertainty robustly predicts recessions, demonstrating real-time forecasting power beyond mere correlation.

The significant improvement in predictive accuracy via nonlinear transformations and Bayesian model averaging provides practical tools for policymakers and analysts.

This consistent enhancement across various uncertainty indices underscores uncertainty's critical role in economic foresight, even if the optimal measure varies by horizon.