Macroeconomic Forecast Accuracy in a Data-Rich Environment

Arbeitspapier

We compare the performance of six classes of models at forecasting di↵erent types of economic series in an extensive pseudo out-of-sample exercise. Our findings can be summarized in a few points: (i) Regularized Data-Rich Model Averaging techniques are hard to beat in general and are the best to forecast real variables. Simulations results show that this robust performance is attributable to the combination of sparsity/regularization with model averaging. (ii) The ARMA(1,1) model emerges as the best to forecast inflation growth, except during recessions. (iii) SP500 returns are predictable by data-rich models and model averaging techniques, especially during recessions. Also, factor models have significant predictive power for the signs of future returns. (iv) The cross-sectional dispersion of out-of-sample point forecasts is a good predictor of macroeconomic uncertainty. (v) The forecast accuracy and the optimal structure of forecasting equations are quite unstable over time.

Object Details

• Language: Englisch

References and Relationships

• Bibliographic citation: Series: Document de travail ; No. 2017-02

Classification and Topics

• Classification: Wirtschaft ; Large Data Sets: Modeling and Analysis ; Multiple or Simultaneous Equation Models: Time-Series Models; Dynamic Quantile Regressions; Dynamic Treatment Effect Models; Diffusion Processes; State Space Models ; General Aggregative Models: Forecasting and Simulation: Models and Applications
• Subject: Data-Rich Models ; Factor Models ; Forecasting ; Model Averaging ; Sparse Models ; Regularization

Contributors, Places and Time

• Event: Geistige Schöpfung
• (who): Kotchoni, Rachidi ; Leroux, Maxime ; Stevanovic, Dalibor
• Event: Veröffentlichung
• (who): Université du Québec à Montréal, École des sciences de la gestion (ESG UQAM), Département des sciences économiques
• (where): Montréal
• (when): 2017

Further information

• Handle: http://hdl.handle.net/10419/234747
• Last update: 10.03.2025, 11:43 AM CET