This project replicates some empirical results of Stock and Watson’s seminal paper, “Vector Autoregressions” (2001), using a three-variable VAR model comprising inflation, unemployment, and the federal funds rate. Drawing on publicly available U.S. macroeconomic data from the Federal Reserve Economic Data (FRED) database, I estimate the reduced-form and recursive VAR models. While diagnostic tests reveal violations of standard VAR assumptions—such as residual autocorrelation, heteroscedasticity, and non-normality—these issues are not explicitly addressed in the original paper. Despite this, the replication closely matches the original results in terms of forecast performance, impulse responses, and forecast error variance decompositions, with only minor deviations likely due to differences in data vintages.