Economic Policy Uncertainty and Industrial Linkage in Japan: A Granger Causality in Quantile Test

ABSTRACT

This article examines the relationship between economic policy uncertainty and industrial linkage in Japan. Using a VAR-based connectedness approach to measure the monthly industrial linkage across 15 groups of Japanese industries, we examine the Granger causality in quantile between the industrial linkage and Japan’s monthly economic policy uncertainty (EPU) and its four components including fiscal policy uncertainty, monetary policy uncertainty, trade policy uncertainty, and exchange rate uncertainty during the period from January 1987 to May 2021. We find that changes in economic policy uncertainty and all four components Granger cause changes in industrial linkage, and this result is robust and consistent across nearly all quantiles. We do not find evidence of the Granger causality running from industrial linkage to economic policy uncertainty. However, considering all quantiles, changes in industrial linkage Granger cause the change in each of the four policy uncertainty.

KEYWORDS:

• Industrial linkage
• economic policy uncertainty
• Granger causality in quantile

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. There are several methods to measure the uncertainty in the economy including survey-based indicators; measures based on an econometric regression and news-based indicators. In this study, we focus on the news-based economic policy uncertainty, which refers to situations characterized by increased dispersion in agents’ expectations about governments’ future economic policy stands (Baker, Bloom, and Davis 2016; Caldara et al. 2020; Ghirelli et al. 2021). Literature also suggests that news-based indicator is a good indicator of uncertainty in explaining the relationship between uncertainty and other aggregate economic indicators (Baker, Bloom, and Davis 2016; Ghirelli et al. 2021). For more detailed, please refer to section 3.

2. While one can use the input-output tables to measure the industrial linkage, input-output tables are constructed annually and thus may not fully capture the dynamics of the economy. Meanwhile, Diebold and Yilmaz (2012) approach has some advantage over the input-output linkage by allowing one to split the forecast error variances of output growth in each industry into components attributable to the other industries in the economy (Li and Martin 2019).

3. There are a number of mechanisms through which shocks to individual sector propagate to other sectors and to the whole economy such as investment and capital accumulation responses in real business-cycle models or Keynesian multipliers or financial frictions facing firms, households, or banks; real and nominal rigidities and their interplay; and potentially inappropriate or constrained monetary policy (Acemoglu, Akcigit, and Kerr 2016).

4. Please refer to Appendix A for detailed exposition of the Diebold and Yilmaz (2012, 2014) approach.

5. As mentioned in Section 4, industrial linkage is measured by total spillover index which in turns is the average relative contribution of other industries in the economy to the forecast variance of an industry’s growth.

6. See equation 3(3) $E\left\{1\left[Y_t\le Q_{\tau }^{Y.Z}\left(Y_t|I_t^Y,I_t^Z\right)\right]|I_t^Y,I_t^Z\right\}=E\left\{\left.1\left[Y_t\le Q_{\tau }^Y\left(Y_t|I_t^Y\right)\right]\right|I_t^Y\right\}\mathrm{\forall }\tau \in \mathcal{T}$(3) in Appendix A.

7. We, however, carry out cointegration tests to reconfirm that there is no cointegration among our variables. In general, we did not find strong evidence of either linear or quantile cointegration between economic policy uncertainty and its components and industrial linkage. Please refer to the appendix B for the test results.

8. Although economic policy uncertainty and its four component series are stationary, implying that we can use the level value of these series to test the Granger causality in quantiles, we use the log difference for our major test results. Results from Granger causality in quantile test using the level of economic policy uncertainty (and its four components) and log difference of the industrial linkage series are used as the robustness check.