A0643
Title: Fiscal impulse responses in a high-dimensional setting
Authors: Davide Bucci - University of Surrey (United Kingdom) [presenting]
Abstract: The aim is to investigate the relationship between fiscal policy and its macroeconomic effects, focusing on the long-standing debate over the size and sign of the fiscal multiplier. Conventional VAR models struggle in this context because they cannot handle many variables without loss of precision. To address this limitation, the standard specification is replaced with an adaptive LASSO-VAR, which applies data-driven, variable-specific penalties and yields a more flexible yet parsimonious model. A Monte Carlo simulation shows that the adaptive LASSO-VAR recovers impulse-response functions more accurately than the traditional VAR and other leading approaches in the literature, including factor-augmented VARs and Bayesian VARs. In addition, a recoverability test is applied, showing that the proposed model better identifies government-spending shocks. Finally, preliminary empirical results for the United States indicate that the adaptive LASSO-VAR estimates a larger fiscal multiplier than competing models, suggesting stronger real effects of government spending than previously documented. Overall, the evidence confirms that combining adaptive machine-learning techniques with VAR analysis improves identification, inference, and policy-relevant measurement in high-dimensional settings.
