In this project, I delve into the intricate relationship between fiscal policy and its economic effects, addressing the ongoing debate on the fiscal multiplier’s magnitude and direction. The conventional VAR models used for such analysis face challenges due to their limitations in handling a large number of variables. To overcome this, I propose employing machine learning techniques, specifically LASSO regression, to enhance the model’s robustness and flexibility. Through a simulation study and preliminary empirical analysis, I demonstrate the effectiveness of the LASSO-VAR in capturing the true responses of macroeconomic variables to government spending shocks. The project addresses the challenge of deriving reliable impulse responses within high-dimensional systems. In addition, I explore innovative methods to conduct consistent inference in this setting.
JEL classification: C32, E62, C55