STRUCTURAL VAR

This project extends the results of "Uncertainty Across Volatility Regimes" to the post-COVID-19 period using identification-through-heteroskedasticity SVAR analysis.

Using a nonrecursive identification scheme for uncertainty shocks that exploits breaks in the volatility of macroeconomic variables, this work addresses two key research questions:

• Is uncertainty a cause or an effect of declines in economic activity?
• Does the relationship between uncertainty and economic activity change across macroeconomic regimes?

---

Replication Files for the Paper

"Causes and Effects of Business Cycles: An Updated SVAR Identification Through Heteroskedasticity Using Uncertainty Data"

Authors:

• Davide Delfino, University of Bologna
• Giovanni Nannini, University of Bologna

---

Overview

This project replicates and extends the analysis of "Uncertainty Across Volatility Regimes"
by Giovanni Angelini, Emanuele Bacchiocchi, Giovanni Caggiano, and Luca Fanelli.

The contribution extends the observation period up to 2024, studying the effects of uncertainty even after the COVID-19 pandemic and introducing additional identification restrictions for this updated volatility regime.

Key Findings:

• Uncertainty is more likely a cause of declines in economic activity.
• Financial and Macroeconomic uncertainty need to be studied separatelly as uncertainty sources exhibits different dynamics.
• COVID-19 represent a specific kind of exogenous event. During this period both sources of uncertainty affect on impact the business cycle, but on the medium run are affected endogenously by the economic downturn.

Data Sources:

The data used in this project were retrieved from:

• FRED: Idustrial Production Total Index for the industrial production growth.
• Jurado, Kyle, Sydney C. Ludvigson, and Serena Ng. 2015. "Measuring Uncertainty." for the uncertainty measures;
• NBER database for the recession dates;

---

Scripts

Main Scripts:

• Structural_analysis: Estimates the on-impact effects for both endogenous and endogenous model and identifies the model that fits better the data. Using the latter model plots the IRF using confidence intervals using a iid bootrsap approach. The time required to run this estimated is estimated to be around 95 seconds.
• Preliminary_analysis: Implements a descriptive analysis. The time required to run this code is estimated to be around 51 seconds.

Supporting Scripts (in Functions/):

• Likelihood_SVAR_%...%: Performs all Maximum likelihood estimation needed to run the main scripts.