Uncertainty shocks and asymmetric dynamics in Korea: a non-linear approach

ABSTRACT

This study investigates the impact of uncertainty shocks on macroeconomic activity in Korea. For this purpose, a Smooth Transition VAR model is employed to document the state-dependent dynamics of two distinct types of uncertainty shocks, namely, financial market based and news-based. When non-linearity is allowed to play a role in our model, quantitatively very different asymmetric dynamics are observed. Following inflation targeting, the responses tend to be smoother and less pronounced. Our empirical results support the view that the link between uncertainty and macroeconomic activity is clear over both recessions and expansions. Furthermore, the impact of uncertainty shocks is more pronounced when economic activity is depressed especially after shocks originate from the financial market, and not from news-based policy uncertainty in Korea.

KEYWORDS:

• Uncertainty shocks
• smooth transition vector autoregression
• asymmetric dynamics
• recessions

JEL CLASSIFICATION:

• C32
• E32
• E52

Acknowledgments

This paper has been greatly improved by the constructive comments and suggestions of David Peel (editor) and an anonymous referee. We are also grateful to Wook Sohn, Byung Kwun Ahn, and Kook Ka for helpful comments and discussions. The views expressed herein are those of the authors and do not necessarily reflect the official views of the Bank of Korea. When reporting or citing this paper, the authors' names should be explicitly stated.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See Bloom (2009), Bloom et al. (2012) for real option e¤ects; Van Nieuwerburgh and Veldkamp (2006), Fajgelbaum, Schaal, and Taschereau-Dumouchel (2014) for feedback (uncertainty trap) channel; Fernández-Villaverde, Pablo Guerrón-Quintana, and Uribe (2011) for precautionary savings channel; Carri_ere-Swallow and Céspedes (2013) for credit channel; Gertler, Gilchrist, and Nataluccim (2007) for financial friction for small open economy; Basu and Bundick (2012) and Leduc and Liu (2012) for search frictions channel.

² The same was true for the U.K. during the event of the Brexit referendum. The U.K. policy uncertainty was the highest, but the U.K. stock market volatility was very low during the period.

³ Choi and Shim (2018)  compared the impacts of financial uncertainty and policy uncertainty on business cycles in Korea with a linear VAR. Shin et al. (2018) construct Korean macroeconomic uncertainty measure and find that their measure explains aggregate fluctuations well in Korea.

⁴ The beginning of the sample of 1991 is due to data on several key variables being unavailable prior and the lack of suitable proxy for those variables. Monthly variables are employed for a couple of reasons. A lower frequency data usually smoothens out much of the variation and fluctuations of the variables concerned. Further, a higher frequency observation alleviates any issues related to zero contemporaneous restrictions employed for structural modeling.

⁵ The VKOSPIX from the datastream is available only after April, 2009. We empolyed both and no difference is found between them.

⁶ The EPU index is the only available standardized policy uncertainty index.

⁷ For more details, see Granger and Teräsvirta (1993).

⁸ We set the threshold for the probability as indicated by the transition function at .8 in this paper following Auerbach and Gorodnichenko (2012) and Caggiano, Castelnuovo, and Groshenny (2014).

⁹ We calibrate it again for the sub-sample because for the full sample period, Korea experienced longer periods of recession. As robustness we estimated a range of $\gamma$ calibrations from $\gamma =1.25$ and $\gamma =2$. The results were robust to these calibrations.

¹⁰ Following Teräsvirta (1994) and Teräsvirta and Yang (2014), several tests are conducted to test for linearity as well as to confirm correct specification of transition function. See Appendix for details.

¹¹ No recursive ordering is right because we are using monthly data. See footnote 4 for details.

¹² For technical details, see Auerbach and Gorodnichenko (2012) and Chernozhukov and Hong (2003).

¹³ We also estimated the pre-crisis period (1991–1997). Unfortunately, however, the sample period is too short to reach a convergence for STVAR.

¹⁴ In a world with perfect capital mobility, when uncertainty rises, it is not surprising that the central banks raise the rate to avoid any capital outflows. See Rey (2016) for details.

¹⁵ We considered three cases from 1998:1, 1998:4, and 1999:1, and then reported the case with 1998:4 simply because the BOK initiated inflation targeting at the second quarter of 1998. There is no difference in terms of empirical results between the three cases.

¹⁶ Inflation targeting is important as one of the factors to stabilize the economy after uncertainty shocks, not the most important factor.

¹⁷ For details, see Caggiano, Castelnuovo, and Groshenny (2014).

¹⁸ See Hoffmaister (1999) and Bodenstein, Erceg, and Guerrieri (2008) for details.

¹⁹ We use five for the full sample and four for subsample based on AIC and BIC. The models were also run with higher and lower lag lengths.