• GARCH
• Heteroskedasticity
• VAR
• Local Projection

While LPs and VARs are known to asymptotically estimate the same impulse response function, their relative performance in a small sample setting lacks in-depth exploration and understanding. This dissertation aims at filling the gap in literature underscored by this need presenting a comparative study of the two methodologies in i.i.d. and heteroskedastic setting. Heteroskedasticity in time series data has been exploited within the VAR framework to identify structural shocks, however the implications of its influence on the estimation of structural impulse responses have not been thoroughly examined. The undertaken analysis addresses this overlooked research area by proposing a simulation study that extends the examination of VAR and LP properties to a small sample setting in the presence of conditional heteroskedasticity. I conduct a comparative study of VAR and LP properties, testing the methodologies' reliability and robustness to misspecification in the specified settings. Misspecification and overspecification are modeled experimenting with different lag selection procedures. Specifically, the former is assessed experimenting with Akaike and Schwarz information criteria while the latter refers to the case in which the lag length is fixed and chosen to be double that of the DGP.