Deus ex Machina? A Framework for Macro Forecasting with Machine Learning

IMF Working Papers, February 28, 2020

Source details

Canonical URL: Deus ex Machina? A Framework for Macro Forecasting with Machine Learning

Develops a framework to nowcast (and forecast) economic variables with machine learning techniques.
Explains how machine learning methods can address common shortcomings of traditional OLS-based models.
Uses several machine learning models to predict real output growth.
Combines multiple machine learning models into ensembles to further reduce forecast errors.
Identifies measures of variable importance to help improve the transparency of machine learning-based forecasts.
Keywords and methodological terms used: Cross-validation, Ensemble, Factor models, Random Forest, RF algorithm, SVM regression, ML model, ML method, Nowcasting.

Machine learning models predict real output growth with lower forecast errors than traditional OLS-based models.
Combining multiple machine learning models into ensembles lowers forecast errors even further.
Applying the framework to Turkey reduces forecast errors by at least 30 percent relative to traditional models.
The framework better predicts economic volatility compared with traditional models.
Implication: Machine learning techniques could be an important part of the macro forecasting toolkit of many countries.

Incorporate machine learning techniques into macro forecasting toolkits to improve forecast accuracy and volatility prediction.
Use ensembles of diverse machine learning models to achieve further reductions in forecast error.
Employ measures of variable importance to increase transparency and interpretability of machine learning-based forecasts.
Address common shortcomings of OLS-based models by leveraging machine learning methods and cross-validation practices.

IMF Working Papers — Deus ex Machina? A Framework for Macro Forecasting with Machine Learning (Marijn A. Bolhuis; Brett Rayner), February 28, 2020.