Disentangling Fiscal Policy:a Data-Driven StructuralVector-Autoregressive Procedure
The thesis provides an empirical analysis relative to the estimation of the effects of fiscal policy on GDP. The research is conducted through the the study of the causal relationships that emerge from the implementation of some Structural Vector-Autoregressive models, which are estimated with U.S. time series data. Although Fiscal multipliers are the important tools that are able to capture the effect on GDP of one dollar spent by the State in increasing government expenditure or in reducing Taxation, economists until now have not reached a consensus about their size (and even their sign). For this reason one of the main topic on which the thesis is concentrated is the computation of fiscal multipliers. Three are the main innovative aspects of the work, respect to the literature of the empirical fiscal studies. The first one is the data-driven approach that characterises all the aspects of the work. In particular, differently from the majority of researches in the literature, the "identification" of fiscal shocks relies completely on a data-driven procedure (The Independent Component Analysis). This important innovation permits to detect the causal effects of fiscal policy relying exclusively on the properties of the data, avoiding in this way the imposition of some restrictions that are driven by some ex-ante economic knowledge. The second point of differentiation from a standard fiscal SVAR analysis is the completely new interpretation of what is called a "fiscal shock". The Fiscal shock is considered in the work as a unique mixture of exogenous and interrelated measures on Taxes and Government expenditure at the same time. In fact, given the evidence of some recent studies about how the fiscal policy is practically designed and implemented, the common assumption about the existence of two separate and independent fiscal shocks, one on taxes and one on government expenditure, is rejected in this work. Finally the results of government expenditure and Tax multipliers computed in the data-driven environment of the work result to be positive (Government expenditure also above the unity) and even persistent 2 years after the shock, something that is not very common in the literature.