On the effects of crude oil demand and supply shocks on Korea’s bilateral trade balance: are the effects asymmetry?

ABSTRACT

The contribution of the present article is to find any indication of whether three components of oil shocks – oil supply shocks, aggregate demand shocks, and oil-specific demand shocks – asymmetrically influence the bilateral trade balance of Korea with its top 10 partners. We discover that there is evidence that the three oil shocks asymmetrically influence Korea’s trade balance with some of its top 10 partners in the short- and long run. Additionally, the asymmetric impact appears to be a country-specific phenomenon and varies depending on different shock components.

KEYWORDS:

• Asymmetry
• oil demand
• oil supply
• NARDL
• SVAR

Acknowledgments

The author thanks the anonymous referee for the careful reading of our manuscript and the many insightful comments and suggestions that improved the quality of the initial manuscript. Any remaining errors are my sole responsibility. This research is financially supported by the award from the University of Alaska Foundation Harold T. Caven Professorship.

Disclosure statement

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

Ethical approval

This article does not contain any studies with human participants performed by any of the authors.

Notes

1. For details, see Kilian et al. (2009).

2. $\mathit{pr}{o}_t$and $\mathit{rp}{o}_t$ are log-transformed. The DF-GLS unit root tests reveal that $\mathit{pr}{o}_t,\mathit{re}{a}_t$, and $\mathit{rp}{o}_t$ are I(0) series and we use them in levels for the SVAR model. It should be noted that, since an SVAR is sensitive to lag lengths, choosing an appropriate number of lags is important for achieving reliable estimates and accurate results. The lag length for the model is determined by the Akaike information criterion (AIC) and Schwarz information criterion (SC). These methods commonly recommend taking a lag length of four quarters (n = 4) for the analysis.

3. For example, $\mathit{s}{s}_t^{+}=\sum _{\mathit{j}=1}^t\max \left[\mathrm{\Delta }ln\left(\mathit{s}{s}_t,0\right)\right]$, $\mathit{s}{s}_t^{-}=\sum _{\mathit{j}=1}^t\min \left[\mathrm{\Delta }ln\left(\mathit{s}{s}_t,0\right)\right]$.

4. Since both tests have non-standard distributions, Pesaran et al. (2001) tabulate new critical values.

5. Because the NARDL assumes that the errors should not be serially correlated, lag lengths (k) should be selected to mitigate the serial correlation in Equation (6)(6) $\mathrm{\Delta }ln\mathit{t}{b}_{\mathit{i},\mathit{t}}={\alpha }_o+\sum _{\mathit{k}=1}^{\mathit{n}1}{\alpha }_{\mathit{i}1,\mathit{t}-\mathit{k}}\mathrm{\Delta }ln\mathit{t}{b}_{\mathit{i},\mathit{t}-\mathit{k}}+\sum _{\mathit{k}=0}^{\mathit{n}2}{\alpha }_{\mathit{i}2,\mathit{t}-\mathit{k}}\mathrm{\Delta }ln{y}_{\mathit{i},\mathit{t}-\mathit{k}}+\sum _{\mathit{k}=0}^{\mathit{n}3}{\alpha }_{\mathit{i}3,\mathit{t}-\mathit{k}}\Delta \mathit{lne}{r}_{\mathit{i},\mathit{t}-\mathit{k}}+\sum _{\mathit{k}=0}^{\mathit{n}4}{\alpha }_{\mathit{i}4,\mathit{t}-\mathit{k}}\mathrm{\Delta s}{s}_{t-k}^{+}+\sum _{\mathit{k}=0}^{\mathit{n}5}{\alpha }_{\mathit{i}5,\mathit{t}-\mathit{k}}\mathrm{\Delta s}{s}_{t-k}^{-}+\sum _{\mathit{k}=0}^{\mathit{n}6}{\alpha }_{\mathit{i}6,\mathit{t}-\mathit{k}}\mathrm{\Delta d}{s}_{t-k}^{+}$(6) . When selecting k = 4 based on AIC, the Lagrange Multiplier (LM) statistics display that the null of homoskedasticity in the models cannot be rejected (Panel D of Tables 2). In addition, the NARDL cannot apply to I(2) series. The DF-GLS unit root tests discover that all variables are I(1) in Equation (6)(6) $\mathrm{\Delta }ln\mathit{t}{b}_{\mathit{i},\mathit{t}}={\alpha }_o+\sum _{\mathit{k}=1}^{\mathit{n}1}{\alpha }_{\mathit{i}1,\mathit{t}-\mathit{k}}\mathrm{\Delta }ln\mathit{t}{b}_{\mathit{i},\mathit{t}-\mathit{k}}+\sum _{\mathit{k}=0}^{\mathit{n}2}{\alpha }_{\mathit{i}2,\mathit{t}-\mathit{k}}\mathrm{\Delta }ln{y}_{\mathit{i},\mathit{t}-\mathit{k}}+\sum _{\mathit{k}=0}^{\mathit{n}3}{\alpha }_{\mathit{i}3,\mathit{t}-\mathit{k}}\Delta \mathit{lne}{r}_{\mathit{i},\mathit{t}-\mathit{k}}+\sum _{\mathit{k}=0}^{\mathit{n}4}{\alpha }_{\mathit{i}4,\mathit{t}-\mathit{k}}\mathrm{\Delta s}{s}_{t-k}^{+}+\sum _{\mathit{k}=0}^{\mathit{n}5}{\alpha }_{\mathit{i}5,\mathit{t}-\mathit{k}}\mathrm{\Delta s}{s}_{t-k}^{-}+\sum _{\mathit{k}=0}^{\mathit{n}6}{\alpha }_{\mathit{i}6,\mathit{t}-\mathit{k}}\mathrm{\Delta d}{s}_{t-k}^{+}$(6) .

6. Since $\mathit{pr}{o}_t$, $\mathit{re}{a}_t$, and $\mathit{rp}{o}_t$ are available monthly and have not been seasonally adjusted, the estimated results may not be robust. Thus, three structural shocks derived from monthly data are seasonally adjusted using the so-called Seasonal adjustment Census X-13 analysis before they are used for the NARDL.

7. These numbers represent the sum of the percentage of a partner’s trade to Korea’s total trade.

8. Since oil-specific demand shocks represent precautionary demand shocks, their impact on trade would be qualitatively the same as in the case of oil supply shocks (Killian, Rubucci, and Spatafora 2009).

9. Since the study period includes the 1997 Asian financial crisis and the 2008 Global financial crisis, two dummies are incorporated to account for these market shocks. It is unveiled that the 1997 Asian financial crisis (the 2008 Great Recession) is highly significant for 5 models (4 models). It should be noted that since the size of the Korean economy is small relative to the global economy, these market shocks are expected to behave exogenously in our model; thus, we treat them as exogenous variables in Eq. (6).

10. Since $\mathit{pr}{o}_t$, $\mathit{re}{a}_t$, and $\mathit{rp}{o}_t$ are available monthly, three-month averages are calculated to use for the quarterly series.

Additional information

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.