Effectiveness of monetary policy under the high and low economic uncertainty states: evidence from the major Asian economies

Aastveit et al. (2017), Balcilar et al. (2017), and Pellegrino (2018) use interacted VAR (IVAR) model to examine the effects of monetary policy shocks on macroeconomic variables for different countries. Even if IVAR model differentiates impulse responses of monetary policy shocks at the high and low levels of uncertainty by calculating above and below the mean of the historical distribution of the uncertainty, the parameters are based on a mean-based estimation and do not allow dynamics vary across the support of the distribution.

Although this study investigates the same Asian countries as Balcilar et al. (2020b) do, there are significant differences between these two studies. While Balcilar et al. (2020b) investigate the spillover effects from internal and external economic policy uncertainties to specifically on Asian financial markets (bond, stock, and exchange rate markets), this study attempts to measure monetary policy effectiveness in these markets by utilizing a case-based approach. To do that, we draw variables used in this study at different quantiles. For example, we build the high economic uncertainty condition by drawing economic policy uncertainty variables at 0.95 quantiles while drawing lending rate, inflation, and industrial production at 0.05 quantiles. Balcilar et al. (2020b) focus basically spillover to financial variables while this study investigates the effect of international uncertainty on real aggregate macroeconomic variables.

Unlike other equation-by-equation estimation methods (see, e.g., Cecchetti & Li 2008; Schüler 2014; White et al. 2015; Chavleishvili and Manganelli 2016; Linnemann and Winkler 2016; Zhu et al. 2016; Ando et al. 2018; Han et al. 2019), Montes-Rojas (2017) we develop a multivariate model approach that solves for a fixed point on the multivariate quantile space based on directional quantiles. This approach solves for multivariate conditional quantiles given the covariates.

In the presence of nonlinearities, one may also use Markov switching VAR (MS-VAR), threshold VAR (TVAR), smooth transition threshold VAR (STVAR), and time-varying parameter VAR (TVP-VAR) models. Each of these models has merits to answer different questions. MSVAR, TVAR, and STVAR models assume a small number of regimes while TVP-VAR models are suitable for cases where there are large number of regimes. However, all these nonlinear models fit to the conditional mean of the data, although conditional mean might be different in different regimes. Moreover, these models force all variables be in the same regime simultaneously. The QVAR model with different quintile paths is more suitable in our case, because we jointly model the US and EU economic uncertainty with macroeconomics aggregates of an Asian economy. We also have specific questions such as high EPU with recession and low EPU with expansion. The QVAR model flexibly allows us to condition on these assumptions and estimate the spillover under these assumptions. Regime switching models do not have the flexibility to incorporate extreme conditions we would like to analyze. Therefore, we should not expect to get similar results from these regime switching models because regimes may not correspond to our quantile path settings.

Generally, a canonical case fixed with \(\theta = \left\{ {0.5, \ldots ,0.5} \right\}\) for all \(i = 1,2, \ldots ,h\) delivers similar estimates to the mean-based VAR forecasts (Montes and Rojas, 2019). So actually, we can compare different economic scenarios with the standard VAR model result with this analysis.

To check robustness, we re-estimate time-varying spillovers from lending rates to macroeconomic variables in various cases after taking the first difference of lending rate series in logarithms. Results with first differenced interest rates are analogous to the results reported here. These additional empirical results are available upon request.