Volatility Spillover Effects in Foreign Exchange Markets among China, Japan, and South Korea

I. Introduction

We investigate the dynamic spillover effects of exchange rate volatility among the foreign exchange markets of China, Japan, and South Korea from January of 2010 to March of 2024 using exchange rate determination theory, the GJR-GARCH model, and the TVP-VAR model.

A foreign exchange rate is defined as the rate of one currency against another, expressing the relative value or price of a country’s currency. Exchange rates are apparently determined by the supply and demand of currencies in the foreign exchange markets. However, exchange rates are influenced by a variety of macro-economic factors at the global as well as domestic level, such as interest rates, the money supply, and prices. Meanwhile, rapid exchange rate fluctuations can have negative effects on a country’s investments, foreign trade and productivity (Byrne and Davis, 2005; Arize et al., 2008; Aghion et al., 2009).

This paper aims to analyze the spillover effects of exchange rate volatility among nations. In particular, we focus the analysis on three Asian countries, China, Japan, and South Korea, for the following reasons. First, these countries are geographically close in East Asia. Some previous studies suggest that the gravity model1 in international trade can also be applied to international finance. They argue that geographical distance can act as a proxy for information pertaining to the corporate culture, markets, and institutions (Portes et al., 2001; Portes and Rey, 2005). Empirical research has also reported that the transmission effects of exchange rate volatility are stronger among geographically closer countries (Melvin and Melvin, 2003; Black and McMillan, 2004).

Moreover, these countries not only hold significant positions in international trade but also maintain close relations with each other. It is well known that trade interconnectedness can bring about financial co-movement or spillover between countries. These three countries account for nearly one-fifth (18.7%) of global trade, and their trade interconnections are numerous. As of 2023, China (12.7%) ranks first in world trade, with Japan (5.3%) and South Korea (5.2%) ranking second and third in Chinese trade, respectively, following the United States (11.2%). Japan (3.2%) holds the fifth position, with China (20.0%) and South Korea (5.2%) correspondingly ranking first and fifth in Japanese trade. South Korea (2.7%) ranks seventh, with China (21.0%) and Japan (6.0%) in first and fourth place, respectively, in South Korean trade (Korea International Trade Association, KITA).2 Accordingly, our study focuses on the spillover effects among the foreign exchange markets of three East Asian countries that are geographically proximate, occupy important positions in the global economy, and maintain very close economic relationships with one another. Analyzing the spillover effects of foreign exchange rate volatility among these countries is both significant and appropriate.3

This study is distinct and makes several important contributions. First, while random walk or AR models are often used in the conditional mean equations of GARCH models, we analyzed the spillover effects of exchange rate volatility between countries in a purer aspect by controlling for each country’s macroeconomic fundamentals and global common risk factors, based on conventional monetary exchange rate determination theory. Through our empirical analysis, we also demonstrated that macroeconomic fundamentals and global common factors could still be relevant in determining exchange rates between countries, thereby partially supporting the validity of monetary exchange rate determination theory.

Second, this paper emphasizes the importance of recognizing the expectations and behaviors of foreign exchange market participants, as well as institutional changes in geographically and economically close neighboring countries, by showing that even after controlling for each country’s macroeconomic fundamentals and global common factors, the foreign exchange markets of the three major East Asian countries can still influence each other.

Third, by employing the TVP-VAR model, this study was able to capture not only the dynamic effects of exchange rate volatility, which are time-varying rather than static parameters, but also revealed that the spillover effects of volatility between the countries differ significantly in terms of the timing, frequency, and persistence.

Section II reviews the existing literature. Section III describes the data and explains the methodology used in our analysis. Section IV reports the empirical results, and Section V concludes the paper.

II. Literature Review

As a first step in our research, we controlled for the effects of macroeconomic variables, including the global common risk on exchange rates, following the conventional exchange rate determination theory. Then, we estimated the GARCH volatility using the residuals and analyzed the cross-country spillover effects of these outcomes. A number of studies exist regarding the relationship between macroeconomic variables and exchange rates. This paper employs the monetary theory exchange rate determination model based on the widely used asset market model to explain the relationship between exchange rates and macroeconomic variables, following Frankel (1979).

Frankel (1979) constructed a real interest rate differential monetary model of exchange rate determination that incorporates price rigidity and the difference in expected inflation between two countries. 4 In this model, the difference in the money supply between two countries, the relative income differences, short-term interest rate differentials, and differences in expected inflation all play a role in determining the exchange rate. If the home country’s money supply is larger compared to that of the foreign country, the domestic currency depreciates, causing the exchange rate to rise (+). An increase in domestic income relative to that in the foreign country raises the demand for domestic currency, causing the exchange rate to appreciate (-).5 A relative increase in the domestic short-term interest rate compared to the foreign short-term interest rate attracts capital inflows according to the interest rate parity condition, also leading to an appreciation of the exchange rate (-).6 A relative increase in expected inflation in the domestic country leads to a depreciation of the domestic currency, resulting in a depreciation of the exchange rate (+). The author empirically tested the determinants of the mark/dollar exchange rate using monthly data from June of 1974 to February of 1978, showing results consistent with the present hypothesis. More specifically, increases in income and short-term interest rate differentials between Germany and the U.S. led to a decrease in the exchange rate, while differences in the money supply and expected inflation7 caused an increase in the exchange rate.

In contrast, Meese and Rogoff (1983) examined the explanatory and predictive power of the pound, mark, and yen exchange rates against the U.S. dollar using monthly data from March of 1973 to November of 1976 after the collapse of the Bretton Woods system. They found that structural models, including macroeconomic variables, had weak explanatory power for exchange rate movements and that their predictive power was not superior to the random walk model in the short and medium term (for horizons from one to twelve months). In addition, Frankel and Rose (1995), through their survey of various studies of exchange rate determination, concluded that macroeconomic fundamentals have limited explanatory power with regard to exchange rate fluctuations.

However, macroeconomic variables have reportedly shown some explanatory power for exchange rate movements since the 1990s, given the use of more sophisticated econometric analyses along with improved monetary policies such as inflation targeting. MacDonald and Taylor (1994) added cointegration techniques to existing monetary models of exchange rate determination, analyzing the relationship between the monthly dollar/sterling exchange rate and macroeconomic variables from January of 1976 to December of 1990. Their results showed that the model outfitted the random walk model in forecasting short-term exchange rate dynamics, and it was also valid for long-term exchange rate analyses. Johnston and Sun (1997) used data from the second quarter of 1975 to the second quarter of 1993, employing an error correction model, and compared the forecasting ability of the mark/dollar and yen/dollar exchange rates, concluding that a monetary model based on economic fundamentals was superior to the random walk model. Shin and Lee (2004) analyzed the determinants of the won/dollar exchange rate during the period before the Asian financial crisis (January 1990–August 1997) and the entire period at the time of their study (January 1990–December 2002). They used a basic model with only money supply and income, Bilson’s (1978) flexible price model including short-term interest rates along with these two variables, and Frankel’s (1979) real interest rate differential model with long-term interest rates in addition to the above-mentioned three variables. Their analysis showed that after the crisis, the macroeconomic variables of the money supply and income variables had a greater impact on the short-term and long-term fluctuations of the won/dollar exchange rate. Similarly, Chin et al. (2007) used quarterly data from the first quarter of 1980 to the first quarter of 2003, reporting that the monetary model suitably explained the long-term trends of the Philippine peso against the U.S. dollar with macroeconomic factors.

Meanwhile, Lilley et al. (2022), analyzing the period from 2007 to 2018, found that after the global financial crisis, global risk factors had significantly greater explanatory power for exchange rate movements across countries. They attributed this to the increasing role of the U.S. dollar as a safe-haven asset after the global financial crisis. When global financial market uncertainty rises, investors quickly shift their portfolios from riskier assets to safer ones. They also found that when the global risk appetite increases, riskier emerging market currencies generally depreciate, whereas safer currencies such as the Japanese yen and Swiss franc either show little changes or appreciate. Mo et al. (2023), using daily data from January 1, 2018 to June 17, 2021, analyzed the determinants of the U.S. dollar exchange rates for the G10 countries. They showed that interest rate differentials between countries were statistically insignificant. Instead, quantitative easing differences between countries, measured by central bank balance sheets, and the VIX (Volatility Index)8 had significant effects on exchange rate movements. Engel and Wu (2024) analyzed factors related to exchange rate determination for G10 countries using a monetary exchange rate model with global risk variables and monthly data from March of 1973 to December of 1998 and from January of 1999 to August of 2023, finding after the 2000s, the model’s fit and predictive power were greatly improved. This was attributed to the credibility of monetary policies with the introduction of explicit or implicit inflation targeting. Specifically, for Japan and Switzerland among the G10 countries, the impact of global risk variables on exchange rate movements was statistically insignificant or showed signs opposite to the expected direction. This can be explained by the fact that, unlike other currencies, these currencies have characteristics similar to those of safe-haven assets. When global real and financial stress and uncertainty increase, the demand for these currencies actually rises, leading to their appreciation.

Considering the comprehensive review of the existing literature of exchange rate determination, it can be said that constructing a mean equation for exchange rate determination is more reliable when based on a monetary theory model rather than a simple random walk model. Furthermore, it is reasonable to include global financial market uncertainty and investor risk preferences in addition to macroeconomic fundamentals in the model.

Next, there have been several salient studies of the cross-country spillover effects of exchange rate volatility in financial markets. Dornbusch et al. (2000) assert that asset price contagion between countries can be categorized conceptually as either based on fundamentals or contagion per se. Fundamental-based contagion includes spillover effects triggered by common global shocks as well as actual and financial linkages. Pure contagion contains the irrational phenomena of investors’ behavior stemming from liquidity constraints, imperfect information and dissimilar expectations, and changes in the “rules of the game,” among other factors. Dornbusch et al. (2000) also criticize how several studies of volatility spillover among countries with regard to asset prices, such as studies of interest rates and foreign exchange rates, did not control for fundamental macroeconomic variables. Leung et al. (2017), using the GARCH (1,1) model and hourly data from January 1, 2001 to April 26, 2013, examined volatility spillover effects among New York, London, and Tokyo stock markets and between the key exchange rates of the euro, pound, yen and New York stock market during normal periods and the crisis periods (the global financial crisis and the Eurozone crisis), finding that volatility spillover increased during the both crisis periods. Qin et al. (2018) analyzed volatility spillover between China’s foreign exchange and stock markets and Japan’s stock market using daily data from January 5, 1998 to June 1, 2018. Their results indicated that volatility spillover effects were mutual, with Japan’s financial market having a stronger spillover effect on China’s financial market. The authors attributed this to limited access to the Chinese financial market and to capital controls imposed by the Chinese government. Melvin and Melvin (2003) examined volatility spillover within and between regions in Asia, Europe, and America using high-frequency exchange rate data with 15-minute intervals for the mark/dollar and yen/dollar exchange rates. They found that intra-regional spillover effects were greater than inter-regional effects. Guo and Wang (2023) investigated the exchange rate transmission effects between the Chinese yuan and ten RCEP9 member countries’ currencies using daily data from August 23, 2010 to August 19, 2022. Regarding China, Japan, and Korea, China was a net receiver from Japan but a net transmitter to Korea,10 while Japan was a net transmitter to China but a net receiver from Korea. Finally, Korea was a net receiver from China, but a net transmitter to Japan. A limitation of this study is that it only analyzed the relationship between exchange rates and did not control for other variables that could influence the exchange rates of each country. During the COVID-19 period, countries with more severe cases of infection experienced greater depreciation of their currencies as their macroeconomic conditions weakened. Additionally, the spillover effects of exchange rate volatility between countries were found to be greater during this period compared to other periods (Wei et al., 2020; Narayan, 2022; Mo et al., 2023).

There are several major strands of the literature that employ TVP-VAR models to analyze spillover effects among domestic and cross-country economic variables. Antonakakis et al. (2019) examined monetary policy spillover among the United States, the Euro area, the United Kingdom, and Japan using daily shadow short rate data from 1995 to 2018. Their findings indicate that the United States and the Euro area act as net transmitters, whereas the United Kingdom and Japan are primarily receivers of spillover. Subsequently, Antonakakis et al. (2020) used monthly data from February of 1975 to January of 2019 to analyze volatility spillover in bilateral exchange rates for four major currencies against the U.S. dollar—the euro, pound sterling, Swiss franc, and Japanese yen. Their results show that the euro plays a dominant role as a transmitter of volatility to the other currencies. Byun and Cho (2023) investigated the interaction between policy uncertainty and financial markets in the United States and Korea using monthly data from November of 2000 to March of 2023. Their analysis reveals that in both countries, economic policy uncertainty and monetary policy uncertainty exert significant effects on the volatility of stock, bond, and foreign exchange markets. More recently, Liu et al. (2025) analyzed volatility spillover effects in exchange rates against the U.S. dollar across countries using daily data from January 6, 2019 to August 29, 2025, reporting that advanced-economy currencies, such as the euro and pound sterling, significantly influence the volatility of emerging market currencies.

III. Data and Methodology

A. Data

The exchange rates used in the empirical analysis of the CNY/USD, JPY/USD, and KRW/USD exchange rates are based on daily closing prices provided by Bloomberg. The analysis period ranges from January of 2010 to March of 2024, and the average monthly value of each dataset is used.

Figure 1 shows the exchange rates (level) and corresponding changes (natural log first difference) for China, Japan, and South Korea. The exchange rates of the three countries exhibit common fluctuations during key global events, including the early 2010 (sovereign debt crises in the Euro area), early and mid-2016 (Brexit), early 2020 (outbreaks of COVID-19), and early and mid-2022 (Russia-Ukraine war and continuation of COVID-19) periods.11 The CNY/USD increased in mid-2019, which appears to be partly due to U.S.-China trade tensions. The JPY/USD rate has risen since early 2023 due to the delay in key interest rate hikes by the Bank of Japan amid the continued strengthening of the U.S. dollar. Next, while changes in the exchange rates of the three countries show common movements since early to mid-2022, they exhibit some differences in other periods. For example, there was considerable change in the CNY/USD exchange rate in mid-2018, the JPY/USD change was pronounced in late 2014 and late 2016, and the KRW/USD change was notable in late 2011.

A unit root test was conducted to test the stationarity of these exchange rates, and as a result, the exchange rates of China, Japan, and South Korea were all found to be non-stationary. Regarding the natural log first-difference variable (), it appeared as a stationary time series without a unit root. In the empirical analysis, the natural log first difference variable was used, as shown in Equation (1). The trends of these exchange rates are presented in Figure 1.

FIGURE 1.

EXCHANGE RATES AND CHANGES IN EXCHANGE RATES OF CHINA, JAPAN AND KOREA

Note: The changes in the CNY/USD, JPY/USD, and KRW/USD rates were calculated based on natural log differences between each period (t and t-1).

Source: Bloomberg.

In Equation (1), S_t represents the spot exchange rate for period t , and the subscript i refers to China, Japan, and South Korea.

Table 1 presents descriptive statistics of the values of the natural log first differences in the CNY/USD, JPY/USD, and KRW/USD exchange rates. The average is 0.0003 for the CNY/USD change, 0.003 for the JPY/USD change, and 0.0008 for the KRW/USD change. The standard deviation is largest for the JPY/USD change (0.0214), followed by the KRW/USD change (0.0185), and the CNY/USD change (0.0102).12 In terms of skewness and kurtosis, the skewness of the CNY/USD and JPY/USD change is positive (+), while the skewness of the KRW/USD change is negative (−). The kurtosis values are all less than 3, indicating a platykurtic distribution.

TABLE 1

DESCRIPTIVE STATISTICS OF CHANGES IN EXCHANGE RATES

Note: 1) The null hypothesis is “no autocorrelation exists.”; 2) The null hypothesis is “the time series does not exhibit ARCH effects at lag k.”; 3) The null hypothesis is “time series has a unit root.”; 4) ***, **, and * denote statistical significance at the level of 1%, 5% and 10%, respectively.

In addition to exchange rates, the data include the M2 money supply, industrial production, short-term and long-term government bond interest rates, and the VIX index. Table 2 provides descriptions of the data and sources used as explanatory and dependent variables in the mean equations for China, Japan, and South Korea.

TABLE 2

DATA DESCRIPTIONS AND SOURCES

B. Methodology

This study adopts a two-stage approach to analyze volatility and the corresponding spillover effects of the CNY/USD, JPY/USD, and KRW/USD exchange rates. In the first stage, the mean equation and the variance equations are specified to extract the exchange rate volatility for each country. In the second stage, the spillover effects of exchange rate volatility among the three countries are examined based on the extracted volatility.

First, the mean equation was specified based on the Frankel (1979) model, which is widely used in the exchange rate determination theory. The mean equation includes the VIX, which is the most widely used proxy for global risk and uncertainty (Forbes and Warnock, 2012; Rey, 2015; Hansen and Krogstrup, 2019). We seek to ensure robustness by using the Global Economic Policy Uncertainty (GEPU) index as a proxy for broader forms of macroeconomic and policy uncertainty, taking into account that the VIX index primarily reflects financial market risk. The GEPU index used in this study was developed by Scott R. Baker, Nicholas Bloom, and Steven Davis, and it has been available since January of 1997. The index is constructed as the GDP-weighted average of national economic policy uncertainty (EPU) indices for the following 18 major countries: Australia, Brazil, Canada, Chile, China, France, Germany, Greece, India, Ireland, Italy, Japan, Russia, Korea, Spain, Sweden, the United Kingdom, and the United States (policyuncertainty.com).

The analysis period includes the period of the COVID-19 pandemic, which is controlled using a dummy variable. The dummy variable is set to 1 for the period from January of 2020 to April of 2023, and 0 for the remaining periods.13 The specific mean equation is expressed as Equation (2).

In Equation (2), ΔSR_t represents the change in the exchange rate (natural log first difference); ΔM2_t refers to the change rate of the M2 money supply (year-on-year); ΔIP_t indicates the industrial production change rate, serving as a proxy for economic growth (year-on-year); TBSR_t denotes short-term interest rates; and TBLR_t resents long-term interest rates, serving as proxy variables for expected inflation. VIX_t indicates the volatility index as a proxy for global common risk factors. The subscripts i and US represent China, Japan and South Korea, and the United States, respectively. e_t is a pure error term after controlling for country-specific macroeconomic fundamentals and global common risk factors. Equation (2) can be expressed more concisely as follows:

Following Frankel (1979), each explanatory variable in the mean equation is expected to have the following effects on exchange rate changes: differences in money supply change rates and long-term interest rates between each country and the U.S. are expected to lead to an increase (+) in exchange rates, while differences in industrial production change rates and short-term interest rates are expected to contribute to a decrease (−) in exchange rates. When the VIX rises, indicating a contraction in global investor sentiment, it is expected to exert upward (+) pressure on CNY/USD and KRW/USD exchange rates. However, due to the Japanese yen’s status as a safe-haven currency, it can have downward (−) pressure on the JPY/USD exchange rates in such situations.

The variance equation is based on the GJR-GARCH model proposed by Glosten, Jagannathan, and Runkle (1993), which extends the ARCH model of Engle (1982) and the GARCH model of Bollerslev (1986). The GJR-GARCH model is well suited for capturing the ‘leverage effect,’ where negative shocks have a greater impact on volatility than positive shocks in financial markets. This model overcomes the limitations of the traditional ARCH and GARCH models, which can capture volatility clustering and persistence but cannot measure the leverage effect. By incorporating dummy variables to reflect asymmetric effects in the foreign exchange markets, the GJR-GARCH model explains the dynamics of volatility more accurately.

The GJR-GARCH model is presented in Equation (4). It builds on the ARCH model (Engle, 1982) and the GARCH model (Bollerslev, 1986), where the volatility at time t (h_t) depends on the squared residuals (, ARCH effects) from period t − 1 and the volatility (h_t−1, GARCH effects) from period t − 1 . Additionally, by including a dummy interaction term (γI_t-1), the model captures the asymmetric effects of positive and negative shocks on volatility, commonly referred to as the leverage effects.

In Equation (4), h_t represents pure exchange rate volatility, as it is derived from the residual term obtained after controlling for macroeconomic fundamentals and global common risk factors. e_t is the error term, and the subscript i represents China, Japan, and South Korea.

To investigate the volatility spillover effects of exchange rates among China, Japan, and South Korea (CNY, JPY, KRW), we employed a time-varying parameter vector autoregression (TVP-VAR) model. This framework, widely popularized by Primiceri (2005) for its ability to capture dynamic interdependencies in variables, allows for structural changes in the transmission mechanism over time. The model parameters were estimated using the multi-move Gibbs sampling algorithm developed by Carter and Kohn (1994), which ensures the efficient simulation smoothing of state variables. However, to prevent model overfitting and maintain stability of the estimation stability, we adopted a specification by which the variance-covariance matrix of the error terms remains time-invariant.

The observation equation of the TVP-VAR(p) model for vector y_t , comprising three endogenous variables, is expressed as follow:

In Equation (5), y_t=[h_t of CNY_t, h_t of JPY_t, h_t of KRW_t]', B_j,t is the (3×3) coefficient matrix at time t, and c_t is the (3×1) constant vector. Here, p denotes the lag length.14 The error term ε_t is assumed to follow a multivariate normal distribution with a mean of zero and a covariance matrix ∑.

To reformulate the equation above into a state-space model, we define all time-varying parameters as a single vectort β_t. Letting (where x_t is a vector containing the constant term and lagged variables), the observation equation can be concisely expressed as follow:

In Equation (6), β_t represents an (m×1) vector containing all regression coefficients of the model, where m=3(3p+1) (the total number of estimated parameters including the constants terms). The time-varying parameter β_t is assumed to follow a random walk process, as described in the following state equation:

In Equation (7), Q is the (m×m) state covariance matrix, acting as a key hyperparameter that determines the time-variation of the parameters. If the elements of Q are close to zero, the parameters β_t remain nearly constant over time, and the model converges to a constant coefficient VAR.

To estimate the unobserved latent variable β_t and the parameters ∑ and Q simultaneously, we applied the Bayesian Markov chain Monte Carlo (MCMC) methodology. In particular, we employed the Gibbs sampling algorithm to avoid high-dimensional integration problems and to ensure efficient sampling. For the prior distribution, we used the inverse Wishart distribution, which acts as a conjugate prior. The specific estimation procedure is as follow:

Step 1. Sampling of the time-varying coefficient β_t

Given ∑ and Q, the conditional posterior distribution of β_t takes the form of a linear Gaussian state-space model. In this study, β_t was drawn by applying the simulation smoothing algorithm proposed by Carter and Kohn (1994).

Step 2. Sampling the error covariance matrix ∑

We employ the inverse Wishart distribution IW(v_S0, S₀) as the prior for ∑ . Given β_t , we can calculate the observation error ε_t=y_t-X_tβ_t . Because the posterior distribution of ∑ follows an inverse Wishart distribution, it is sampled as follows:

Step 3. Sampling the state covariance matrix Q

We employ the inverse Wishart distribution IW(v_S0, S₀) as the prior for Q .Q is sampled using the state equation errors, which are defined as u_t=β_t-β_t-1 for t=2, ⋯, T.

IV. Empirical Results

Table 3 shows maximum likelihood estimates of the mean equation for exchange rate changes and the variance equation for exchange rate volatility, according to the GJR-GARCH model. We also present here the empirical results of a model that employs the GEPU index, in addition to the VIX, as part of a robustness check. In interpreting the results, we take as the benchmark the model including the VIX, which is identified as the preferred specification based on both the AIC and BIC criteria.

TABLE 3

MAXIMUM LIKELIHOOD ESTIMATES OF GJR-GARCH MODEL

Note: 1) Numbers in parentheses are the standard errors; 2) ***, **, and * denote statistical significance at the level of 1%, 5% and 10%, respectively.

With regard to the mean equation, the CNY/USD rate is significantly affected by the short-term interest rate differential. In other words, when China’s short-term interest rates rise or U.S. short-term interest rates fall, leading to an increase in the short-term interest rate differential between the U.S. and China, the Chinese yuan appreciates against the U.S. dollar. This result aligns with the expected sign. The COVID-19 dummy variable had a negative effect on the CNY/USD rate, although it is not statistically significant, contrary to the JPY/USD and KRW/USD cases, indicating an appreciation of the Chinese yuan against the U.S. dollar during the COVID-19 period. This appears to be partly due to differences in the monetary policies between the U.S. and China during that period. The Federal Reserve in the U.S. implemented aggressive monetary easing by lowering the federal funds rate from 1.5–1.75% to 0.0–0.25% in March of 2020, shortly after the COVID-19 outbreak, and maintained near-zero interest rates until March of 2022, when the rate was raised to 0.25–0.5%. In contrast, the People’s Bank of China made a modest cut from 4.05% to 3.85% in April of 2020 and held the rate at 3.70% until July of 2022. The M2 money supply differential, industrial production differential, long-term interest rate differential, and the VIX did not have a statistically significant impact on the CNY/USD exchange rate.

The JPY/USD exchange rate is significantly affected by the VIX and the COVID-19 dummy variable. An increase in global risk aversion, indicated by a rise in the VIX, leads to an appreciation of the Japanese yen against the U.S. dollar as demand for the Japanese yen increases due to investors’ preference for safe-haven assets. In addition, the COVID-19 dummy variable shows a depreciation effect on the JPY/USD. This suggests that the COVID-19 pandemic had a negative impact on Japan’s macroeconomy, leading to a depreciation of the Japanese yen against the US dollar. The M2 money supply differential, industrial production differential, long-term interest rate differential, and short-term interest rate differential did not show a statistically significant impact on the JPY/USD exchange rate.

With regard to the KRW/USD exchange rate, the M2 money supply differential and the VIX registered a statistically significant impact. When Korea’s M2 money supply increases or the U.S. M2 money supply decreases, the Korean won depreciates against the U.S. dollar. Additionally, when global financial risks increase, international investors tend to prefer safe-haven assets, leading to a decline in the value of the Korean won against the U.S. dollar, which is considered a currency of an emerging economy. The industrial production differential, short-term interest rate differential, long-term interest rate differential, and the COVID-19 dummy variable also did not have statistically significant effects on the KRW/USD exchange rate.

In the variance equation, for the CNY/USD/ and the JPY/USD exchange rates, both the GARCH term and leverage term (γ), are statistically significant. This implies that the volatility of these currencies exhibits characteristics such as persistence and leverage effects. Regarding the KRW/USD exchange rate, the GARCH term (α₂) is statistically significant. This suggests that the volatility of the KRW/USD exchange rate is primarily driven by GARCH factors.

Figure 2 shows the time-varying effects of JPY/USD and KRW/USD exchange rate volatility on the CNY/USD exchange rate volatility. As shown in Figure 2 (a), the volatility in the JPY/USD exchange rate has a statistically significant effect on the volatility of the yuan/dollar exchange rate during the periods of May 2018–January 2019 and September 2021–September 2023. Notably, the JPY/USD exchange rate volatility is found to significantly amplify the CNY/USD exchange rate volatility in the period surrounding the outbreak of the Russia–Ukraine war. Figure 2 (b) shows that the volatility in the KRW/USD exchange rate increases significantly the CNY/USD exchange rate volatility across considerable periods of time, specifically November 2011–June 2012, August 2012–March 2014, and June 2014–April 2019.

FIGURE 2.

TIME-VARYING PARAMETER ESTIMATES FOR THE CNY/USD EXCHANGE RATE VOLATILITY

Figure 3 illustrates the time-varying effects of CNY/USD and KRW/USD exchange rate volatility on the JPY/USD exchange rate volatility. In Figure 3(a), the effects of both the CNY/USD and KRW/USD exchange rate volatility on the JPY/USD volatility appear to be limited over the full sample period. Regarding the CNY/USD volatility, it increases the JPY/USD volatility significantly only in December of 2014; however, during most other periods, such as March–April 2012 and November 2012–June 2013, it tends to reduce the JPY/USD volatility. The KRW/USD volatility decreases the JPY/USD volatility significantly over the period of October 2022–April 2023. These findings suggest that, unlike the Chinese yuan or the Korean won, the Japanese yen continues to function to some extent as a global safe-haven currency.

FIGURE 3.

TIME-VARYING PARAMETER ESTIMATES FOR THE JPY/USD EXCHANGE RATE VOLATILITY

Figure 4 presents the time-varying effects of CNY/USD and JPY/USD exchange rate volatility on the KRW/USD exchange rate volatility. As shown in Figure 4 (a), the CNY/USD exchange rate volatility appears to have a statistically significant effect on the KRW/USD exchange rate volatility throughout a substantial number of time periods, including December 2012–February 2013, April–May 2014, and March 2022–September 2023. In Figure 4 (b), the JPY/USD exchange rate volatility has a significant impact on the KRW/USD exchange rate volatility from May of 2012 to December of 2014.

FIGURE 4.

TIME-VARYING PARAMETER ESTIMATES FOR THE KRW/USD EXCHANGE RATE VOLATILITY

In addition, Table 4 shows the statistically significant period, number of months, and longest duration of the exchange rate volatility spillover effects among the three currencies, as derived from Figure 2, Figure 3, and Figure 4. This allows for a comparison of the frequency and persistence of exchange rate volatility spillover effects in the foreign exchange markets among the three countries. In this paper, frequency is measured using the number of months with significant spillover effects, and persistence is assessed by the longest duration of significant months.

TABLE 4

STATISTICALLY SIGNIFICANT PERIODS, NUMBER OF MONTHS AND THE LONGEST DURATION OF THE EXCHANGE RATE VOLATILITY SPILLOVER EFFECTS

Regarding the spillover effects of exchange rate volatility between the CNY/USD rate and the JPY/USD rate, the JPY/USD rate showed a significant effect over 34 months, with the longest duration being 25 months, while the longest durations for the CNY/USD rate were only eleven months and eight months. For the exchange rate volatility spillovers between the KRW/USD rate and the CNY/USD rate, while there were more statistically significant months and the longest duration, both the frequency and persistence of the spillover effects were greater or longer from the KRW/USD rate to the CNY/USD rate compared to the opposite case. As it were, frequency and persistence for the KRW/USD rate were 87 months and 59 months, respectively, compared with 24 months and 19 months for the CNY/USD rate. With respect to the exchange rate volatility spillover effects between the JPY/USD rate and the KRW/USD rate, both the frequency and persistence were higher from the JPY/USD to the KRW/USD rate. Specifically, both the frequency and persistence of the KRW/USD rate were seven months, whereas those of the JPY/USD rate were 32 months. These results indicate that the volatility of the CNY/USD exchange rate is affected to a greater extent by fluctuations in the KRW/USD and JPY/USD exchange rates than it affects them in return. This asymmetric relationship is attributable to China’s foreign exchange market regulations and daily exchange rate band constraints, which tend to dampen the transmission of CNY/USD volatility to neighboring foreign exchange markets. Moreover, the role of the Korean won as a hedging currency against the Chinese yuan, together with the strengthened status of the Japanese yen as a global safe-haven currency, may contribute to this asymmetry further. In addition, the relatively strong spillover effects observed between the Korean and Chinese foreign exchange markets, compared with those between China and Japan or Korea and Japan, are likely to reflect the substantial real economic linkages between Korea and China, as well as their common features of emerging market currencies.

V. Conclusion

China, Japan, and South Korea, located in the East Asian region, hold significant shares in global GDP and trade, making them key players in the world economy. These three countries are also geographically and economically interlinked. This implies that volatility in one country’s financial and foreign exchange markets can immediately and persistently impact the markets of the other countries. This paper analyzes the time-varying spillover effects of exchange rate volatility among the three countries from January of 2010 to March of 2024. We specifically aim to examine the pure spillover effects of exchange rate volatility by controlling for country-specific macroeconomic fundamentals and global common risk factors, based on the well-known monetary exchange rate determination theory. Furthermore, using the TVP-VAR model, we investigate not only the dynamic effects of volatility spillover over time but also asymmetric effects between the countries.

The main empirical results can be summarized as follows. First, using the GJR-GARCH model, we identified the determinants of exchange rate changes for the CNY/USD, JPY/USD, and KRW/USD rates. These determinants varied across the three countries. The CNY/USD rate is influenced by the short-term interest rate differential with the U.S. The JPY/USD rate is affected by the VIX and the COVID-19 dummy. The KRW/USD rate is impacted by the difference in the money supply change rate with the U.S. and the VIX. The VIX increased exchange rates for KRW/USD and CNY/USD,15 while it decreased exchange rates for JPY/USD. This indicates that as global financial risks rise, international investors tend to prefer a safe-haven currency such as the Japanese yen, which leads to the depreciation of emerging market currencies such as the Chinese yuan and Korean won. Second, the exchange rate volatility of the three countries exhibited characteristics of persistence (GARCH effects for CNY/USD, JPY/USD, and KRW/USD) and asymmetry (leverage effects for CNY/USD and JPY/USD). Third, regarding the time-varying spillover effects of exchange rate volatility between the three countries’ foreign exchange markets, the spillover effects of exchange rate volatility between the three countries demonstrated some differences in terms of timing, frequency, and persistence.

These analysis results show that, even after controlling for each country’s macro fundamentals and global common risk factors, the foreign exchange markets of the three major East Asian countries, geographically and economically close, can still influence each other. These characteristics of exchange rate volatility transmission effects among the three countries suggest that in order to maintain stability in their foreign exchange markets, each country must monitor and respond swiftly to the expectations and behaviors of participants in neighboring foreign exchange markets, and to institutional changes, along with macro fundamentals and global common risk factors.