4.2 VAR and short-run identification
\[B(L)y_t=d+\varepsilon_t\]
where \(y_t
\)is vector of endogenous variables, \(d\) is vector of constant terms , \(\varepsilon_t\) is reduced-form residuals(\(E(\varepsilon_t)=0, E(\varepsilon_t\varepsilon_t') = \Sigma\)) and \(B(L)=I+B_1L+B_2L^2+\dots+B_NL^N\).
Then estimating the model using the Cholesky decomposition. First, estimate the reduced-form VAR parameters and compute the residual variance-covariance matrix. The structural impact multiplier matrix is then estimated based on a lower-triangular Cholesky decomposition of the residual variance-covariance matrix.
4.3 TVP-VAR
5. Results
5.1 Uncertainty measurement
[Fig.1] shows that the uncertainty measures in 1997-2017. Economic Policy Uncertainty(EPU), an index of uncertainty that has been used in many previous papers, and the value (GU, KU) we used in this study, both variables are compared as time series data, respectively. EPU indices appear to be insufficient to represent economic uncertainty in some ways. First, EPU index is showing a increasing trend for both global and domestic. These need to be corrected because there is no clear reason for economic unexpected variable of uncertainty to be time dependent. Second, the EPU index does not fully reflect economic uncertainty. In the above graph in Fig1, the global EPU does not properly reflect the global financial crisis(2008), while the domestic EPU in the graph below does not reflect the 1997 financial crisis. The GU and KU indices estimated in this study complement the above limitations, reflecting past uncertainties as shown in the Fig1.
Uncertainty Index comparison: GU and KU indexes comparing with GEPU(Global Economic Policy Uncertainty) and KEPU(Korea Economic Policy Uncertainty), time spans from 1997Q1 to 2017Q4