Economic Policy Uncertainty Index Meets Ensemble Learning.

Area of Science:

• Economics
• Data Science
• Computational Social Science

Background:

• The Economic Policy Uncertainty (EPU) index is a crucial metric for understanding economic dynamics.
• Traditional EPU calculations may not fully capture the complexity of uncertainty.
• Ensemble learning and gradient boosting techniques offer advanced methods for predictive modeling.

Purpose of the Study:

• To enhance the predictive accuracy of the Economic Policy Uncertainty (EPU) index.
• To compare the performance of various ensemble learning and gradient boosting techniques in modeling EPU.
• To investigate the relationship between media-based EPU and financial volatility versus consumer sentiment.

Main Methods:

• Utilized ensemble learning techniques: ensemble linear regression (LM) and random forest.
• Employed gradient boosting techniques: Gradient Boosting Decision Tree and Extreme Gradient Boosting (XGBoost).
• Applied methods to the Newsbank media database, expanding the scope of search terms for EPU calculation.

Main Results:

• LM and XGBoost models significantly outperformed other methods and the original EPU index.
• Enhanced EPU estimates demonstrated stronger countercyclicality and leading properties.
• EPU showed a higher correlation with financial volatility than with consumer uncertainty assessments, indicating media's focus on financial markets.
• Widening the keyword scope and using ensemble learning improved EPU's correlation with uncertainty proxies.
• Models performed better in monthly predictions (industrial production growth) than quarterly (GDP growth), suggesting uncertainty is a short-term phenomenon.

Conclusions:

• Ensemble learning, particularly LM and XGBoost, offers a superior approach to calculating and predicting the EPU index.
• The enhanced EPU is a more accurate and timely indicator of economic uncertainty, especially in short-term financial contexts.
• Findings highlight the media's emphasis on financial sector volatility in shaping public perception of economic uncertainty.