Does nuclear uncertainty threaten financial markets?: The attention paid to North Korean nuclear threats and its impact on South Korea’s financial markets*

Ju Hyun Pyun† Korea University Business School

In Huh‡ The Catholic University of Korea

This version: October 1, 2016 Abstract We explore how the investor attention paid to dangerous nuclear tests in an adjacent country influences financial market outcomes. To measure the attention paid to North Korean nuclear threats, we introduce a weekly Google search volume index for keywords related to North Korean nuclear events. Using a time-varying structural vector auto-regression model with block exogeneity restrictions, we find that the investor attention paid to nuclear threats has heterogeneous effects on South Korea’s stock price both across industries and over time: only attention paid to the first nuclear test was negatively related to the stock price index, and this negative association vanished thereafter. The attention paid to the second test had a significant depreciation impact on the foreign exchange rate. Our result also reveals that the investor attention paid to the nuclear risk reduced stock price, especially in the banking industry, during the whole sample period. JEL Classification: F3; G1 Key words: Investor attention, North Korean Nuclear Risk, Google SVI, Structural Vector Auto-Regression Model, Block Exogeneity, Political Risk

* We are grateful to Boyoung Choi, Minsoo Han, Zonglai Kou, Abul Shamsuddin, Haizhi Wang and participants in the 9th Annual Conference between Fudan University and Chonnam National University, China, the KIEP seminar, Korea, 7th IFABS conference, China and 2016 WEAI meeting for their helpful comments. This work was supported by the Catholic University of Korea, Research Fund 2015. All remaining errors are our own. † Korea University Business School, 145 Anam-Ro, Seoungbuk-Gu, Seoul 136-701, Korea, Tel:+82-2-32902610, E-mail: [email protected] ‡ The Catholic University of Korea, Department of Economics, 43 Jibong-ro, Wonmi-gu, Bucheon-si, Gyeonggi-do 14662, Korea, Tel:+82-2-2164-4569, E-mail: [email protected]

1

1. Introduction North Korea has been provoking South Korea since the Korean Armistice Agreement was signed in 1953. As recently as the 2000s, North Korea conducted several serious military actions, such as the Yellow Sea battle, attacks on the Cheonan ship and Yeonpyeong Island, and various missile launches, which can only be construed as grave threats to South Korea. North Korea also started conducting nuclear tests in 2006, which have been heavily criticized the world over and have intensified the political and geographical uncertainty in the Korean peninsula. North Korea’s ambition to be a nuclear state has not only raised concerns about a potential war between the two Koreas but also put world peace in peril. As the New York Times noted in April 2013, there are various views on North Korea’s nuclear tests: “Many analysts believe that North Korea is again seeking aid and other concessions, while some suggest that it merely wants to be recognized as a nuclear state, like Pakistan. Still others suggest that the North genuinely fears an attack by the United States or South Korea and views the warnings as deterrence. Highlighting a perceived threat from abroad is also a favorite tool the North Korean government uses to ensure internal cohesion.” Since it is difficult to pin down the real intentions behind North Korea’s provocations and threats (i.e., whether the threats are empty), it is hard to identify the “real” effect of such risks on South Korea’s financial markets. In this study, we analyze how North Korea’s nuclear risk influences the South Korean financial markets from an investor’s viewpoint. We provide a new measurement for the investor attention paid to North Korean nuclear threats in the form of a Google search volume index (SVI) (e.g., Da et al., 2011), which measures the online search frequencies of related keywords pertaining to North Korea’s nuclear events. This measurement helps quantify investor’s demand for information on North Korea’s nuclear threats and provocations. Then, using weekly data for 2004–2012, we employ a structural vector auto-regression (VAR) 2

model with block exogeneity restrictions by treating the attention paid to the nuclear threats as a given exogenous variable in the system and considering endogeneity among financial market variables. We find that the market’s attention to very early North Korean nuclear threats was negatively related to the South Korean stock market price index (KOSPI). This negative effect on the stock price of the attention paid to the nuclear threats attenuated afterward. Interestingly, only the attention paid to the second nuclear test was significantly associated with depreciation of the Korean won. Further, we investigate the heterogeneous responses of industry-specific stock prices to the nuclear risk. Our results show that only the banking industry was considerably hit by the nuclear risk in terms of industry stock price; the stock prices of other industries did not respond significantly. We show that our results are robust by controlling for alternative measures, specifications, and identification. Many previous studies show significant financial market reactions (price changes and volatility) to exogenous political conflicts or social crises. Frey and Kucher (2000) and Waldenstrom and Frey (2002) examine the impact of events during World War II on the prices of several countries’ government bonds traded in Sweden and Zurich (Switzerland), respectively. Amihud and Wohl (2004) and Rigobon and Sack (2005) focus on the Iraq War and its consequences on financial markets. Amihud and Wohl (2004) find that the likelihood of Saddam Hussein’s fall from power, as reflected in a traded futures contract that paid out if Saddam were to be ousted, is related to U.S. stock market returns.1 Rigobon and Sack (2005) investigate the impact of the news shock about the Iraq War on several U.S. financial variables. They find that an increase in the war risk caused a rise in oil prices, declines in treasury yields and equity prices, a widening of corporate yield spreads, and depreciation of

1

Wolfers and Zitzewitz (2009) use the keywords “Saddam Security” by to estimate the expected cost of the Iraq War. They show that the Iraq War was expected to lower the value of U.S. equities by around 15%, equivalent to USD 1.1 trillion, the market value of all stocks in the S&P 500 Index.

3

the U.S. dollar. Berkman et al. (2011), using a large sample of major international political crises, show that disaster risk affects future consumption growth and expected stock returns. Fisman et al. (2014) show that interstate political conflicts between China and Japan2 have significant negative effects on firm-level stock returns and profit measures in both countries.3 Our study contributes to the existing literature on the effects of political crises on the financial market in several ways. First, we shed light on the unique and real-time political crisis posed by North Korea, which has often taken center stage in world affairs. We utilize the latest weekly data for the period 2004–2015, which include thrice as many nuclear threats and tests by North Korea, and implement comprehensive analysis by comparing the effects of these nuclear risks on South Korea’s financial markets across industries and over time. Despite the importance of such events, few studies examine the effects of North Korea’s military or nuclear provocations on the South Korean financial market or the markets of adjacent countries. Even some existing studies show mixed findings. For instance, Kollias et al. (2014) show a greater adverse effect of North Korea’s second nuclear test on the stock prices of South Korea and other Asian countries than that of the first nuclear test.4 They argue that the qualitative difference between the two events was that the first was announced whereas the second was unexpected, which led to different reactions in the financial market.5

2

They consider two events: Japan’s approval of the new textbook and the Senkaku (Diaoyu in China) disputes. Another strand of research focuses on the relationship between political crises and the second moment of financial variable (i.e., stock market volatility) or financial linkages. Bittlingmayer (1998) finds that political uncertainty in the early 1920s in Germany generated greater stock volatility. Brown et al. (2006) argue that the low volatility of Consols during Pax Britannica (1816–1913) may be attributed to the political stability in that period. Choudhry (2010) shows that wartime events during World War II resulted in structural breaks in stock price movement and volatility. Frijns et al. (2012) examine the relationship between political crises and stock market integration and show that political crises significantly reduced the degree of stock market integration in 19 emerging markets for 1991-2006. In a similar vein, previous studies on international trade also find significantly negative effects of interstate military conflicts on trade, and vice versa (e.g., Glick and Taylor, 2010; Lee and Pyun, 2016; Martin et al., 2008). 4 Lee (2006), writing in Korean, finds that the daily news about North Korean nuclear threats decreased South Korea’s stock market index and devalued the won during 2002 and 2003. 5 Koillas et al. (2014) suggest that the prior announcement of nuclear weapon development provided time for markets to absorb the nuclear shock, whereas the unannounced second nuclear test was more of an exogenous shock that markets did not expect. 3

4

However, Kim and Roland (2011) do not find any significant effects of North Korean military provocations on South Korea’s financial markets from 2000 to 2008. A novel feature of this study is that, motivated by previous studies on investor attention such as Da et al. (2011), Mondria and Wu (2012), and Vlastakis and Markellos (2012), we introduce Google’s SVI using keywords about North Korean nuclear tests to measure the demand for information on the political risk posed by North Korea. This measure captures valuable information on whether the nuclear risk is realized in the market, as it reflects the “attention” economic agents pay to real-time nuclear threats.6 In this regard, our findings of a significant effect of the nuclear risk on financial market outcome are not limited to specific political risk but provide more general implications for how investors respond to political risk without loss of generality. Further, many previous studies on political risks (including those orchestrated by North Korea) approach such issues with an event study analysis or identify each catastrophic event using a dummy variable (e.g., Kim and Roland, 2011, Kollias et al., 2014). While these works deal with each political crisis event with the same degree of importance, the Google SVI successfully measures the intensity and variation of the risks perceived by agents and overcomes the shortcomings of treating various North Korean risks as identical events, as in previous studies. Thus, our work contributes to the literature not only on political risk and financial markets but also on investor attention and/or information demand and financial market outcome. Methodologically, we employ a time-varying structural VAR model. While exogeneity of political risk is endorsed in many previous studies, endogeneity and feedback of financial market variables are still important concerns. Our VAR approach with block

6

The term “attention” has been studied in psychology and is widely used to explain investor behavior while studying the financial market. The terms “inattention” and/or “limited attention” are even more widely used (Huang and Liu, 2007; Kahneman, 1973).

5

exogeneity restrictions allows for both exogenous political risk and endogeneity among financial market variables, which contributes to identifying the pure effect of market attention. The rest of the paper is organized as follows. Section 2 introduces the data and our measure of North Korean nuclear risks and also describes the movement of the variables of the South Korean financial market in response to the nuclear tests. Section 3 shows the empirical model, which considers possible endogeneity among financial market variables and reports the empirical results. Section 4 concludes.

2. Data 2.1. Measuring the attention paid to North Korea’s nuclear threats In this section, we introduce the method used in this study to quantify the attention paid to risks posed by North Korea’s nuclear threats. Table 1 shows the series of North Korean military provocations and nuclear threats since 2000. There have been various types of dissimilar events. For example, it is clear that North Korea’s announcements of missile launches have very different effects on the South Korean financial market than do real military attacks like those on the Cheonan ship and Yeonpyeong Island. Even the same types of nuclear tests may have different effects on South Korea; investors pay more attention to some events than others, depending not only on the type and intensity of North Korea’s provocations but also the accessibility of information about each event. An incident itself does not affect investment in financial markets. Rather, investors may later actively search online for more specific information about the incident and change their investing decisions because they may be unaware of the event when it just breaks out. Hence, it is important to identify the “real” risks posed by North Korea, namely, the risks that investors or market participants perceive or pay attention to. [Insert Table 1]

6

A new indicator that reflects the attention economic agents pay to the nuclear risks posed by North Korea is proposed by using weekly Google SVI data collected from the Google trend database (https://www.google.com/trends/). Barber and Odean (2001) provide evidence of a growing tendency among investors to rely more on the Internet for information and brokerage services because they are unwilling to pay for advisory services from off-line brokers. This supports the important role of Internet-based search in investment decisions. In fact, the SVI has been previously used to measure the attention paid by investors to a specific stock in the stock market (Da et al., 2011; Mondria and Wu, 2012; Vlastakis and Markellos, 2012). For instance, if investors are interested in Microsoft Corporation, they can look for financial information and real-time news about Microsoft by typing the stock’s ticker symbol, “MSFT,” into the Google search bar. Then, Google provides information on Microsoft, its financial statements, and news about that stock. The Google SVI project summarizes this searching activity on the Internet and constructs the index after computing the search frequency. The index standardizes the number of searches between 0 and 100, wherein the higher the index, the greater the number of searches. Da et al. (2011) propose this Google SVI for stock ticker symbols as a direct proxy for investor attention and find that an increase in SVI predicts higher stock prices in the following two weeks and an eventual price reversal within the year. Mondria and Wu (2012) analyze the SVI for overseas stocks (i.e., stocks outside the United States) to examine U.S. investors’ scrutiny of overseas stocks. A subsequent study by Vlastakis and Markellos (2012) finds that the demand for information on the specific stock at the market level (measured by the SVI) is closely associated with stock market volatility, particularly when there is an economic boom. Previous studies also apply the SVI measure to various fields in finance. Kita and Wang (2012) and Smith (2012) construct SVIs to measure the intensity of the global financial crisis and investigate its effect on the foreign exchange market. 7

We construct the SVI for North Korean nuclear threats using the following keywords: “North Korea nuclear” or “Korea nuclear.” We also use Google to search for words reflecting the same meaning in Korean and devise another SVI to analyze the effects on domestic investors of South Korean nationality only. Note that the SVIs using Korean keywords are only available at a monthly frequency. We believe that the SVI can measure the market attention paid to the nuclear risks posed by North Korea. Nevertheless, there are several caveats to constructing the index. First, unlike the SVI used by previous studies to capture investor scrutiny of the stock ticker symbol7, the SVI for North Korean nuclear threats could be noisy because the index may capture not only the attention of investors to North Korean nuclear threats but also that of any other agents interested in this issue but having no intention to invest. Another concern is whether this measure entirely captures the North Korean nuclear risks. One may argue that our baseline measure could also capture surprising positive news about North Korea’s nuclear status (e.g., progress in the six-party talks). To avoid the above concerns, we first refine our measure by limiting the index construction to investors/regions that tend to search for information about Korea, such as the United States, which is a major investor in South Korea. However, this refined alternative measure shows a very high correlation with the original measure without regional restriction, and the results are not sensitive to the refinement. Furthermore, to rule out the possibility that the SVI index of the nuclear risk captures good news about North Korea, we introduce another proxy for positive information shocks about North Korea, the SVI constructed using the keyword “six-party talks,” and compare the effects of this positive measure with our baseline SVI on financial market outcomes. Fig. 1 describes two SVIs using English and Korean keywords for North Korean

7

It could be rare for non-investors to search the ticker name of stocks without any intention of investing.

8

nuclear threats, as well as the SVI using the English keyword “six-party talks.” Gray areas indicate when North Korea conducted nuclear tests. We observe high correlation between the SVIs for the nuclear threat and actual North Korean nuclear tests. However, the SVI utilizes more detailed information on the investor attention paid to each nuclear test. For example, the SVI takes the highest value after the first nuclear test, which was conducted in October 2006. The second-highest value is observed for February 2013, when North Korea conducted its third nuclear test. The SVI for six-party talks also shows a somewhat different fluctuation from the baseline SVIs. It has the highest value in September 2005 and another peak right after the first nuclear test, but it has been decreasing over time. [Insert Fig. 1] 2.2. Other financial market variables Data used are weekly average data for the South Korean foreign exchange rate (won/U.S. Dollar), three-year bond interest rate, and Korean stock market index (KOSPI) from 2004 to 2015. We source daily data for the three-year bond interest rate (the most traded bond in South Korea), the won/dollar exchange rate, and the KOSPI from the Economic Statistics System of the Bank of Korea. We use weekly averages instead of values at the end of the week to allow all financial variables to reflect any changes owing to the North Korean risk within a given week. Unfortunately, industrial production is not available at a weekly frequency, so we interpolate monthly industrial production to weekly data to measure real economic activities. However, our results do not change by excluding the industrial production variable from the system. The descriptive statistics of the sourced data appear in Appendix Table 1. Further, in order to analyze heterogeneous industry stock price responses to the nuclear risk, we collect the monthly industry-specific stock price index from the KIS Line,

9

which accumulates financial information on Korean firms.8 This source provides data for the stock price index, trading volume, and turnover for 20 industries. Appendix Table 2 summarizes the means and standard deviations for industry stock price data.

2.3. Eyeball test: Responses in South Korea’s financial market on the day of North Korean nuclear threats It is natural to expect negative impacts on the South Korean financial markets arising from North Korea’s escalating nuclear threats. To motivate our analysis, Table 2 reports changes in prices in the South Korean financial markets on the days North Korea conducted nuclear tests. In general, the risks posed by North Korea decreased the value of the stock index, the value of the Korean won, and the interest rate of long-term government bonds. In the foreign exchange market, the Korean won showed a very small rage of depreciation: about 0.1% to 0.4%. The value of the Korean won against the U.S. dollar depreciated after the first and third nuclear tests. KOSPI exhibited a more dramatic fall in response to the first nuclear test than to the other tests. Interestingly, the magnitude of the negative price shocks on KOSPI reduced as North Korea repeated the nuclear tests. It seems that the effects of the nuclear risks on the South Korean stock markets vary over time. [Insert Table 2] However, the descriptive changes in the prices of the South Korean financial markets in Table 2 could have been caused not only by the nuclear tests but also by other shocks in South Korea. Therefore, we should be very careful while interpreting these price changes as the real (and sole) consequences of the nuclear tests. We can evaluate the cost of negative spillovers from the nuclear risk and prepare policy instruments for future provocations by North Korea only if we are able to identify the pure effect of the nuclear threats on the South

8

KISLINE: Knowledge Inside. http://www.kisline.com. Accessed on March 08, 2015.

10

Korean financial market. Thus, we try to estimate the pure effects of the nuclear risks on the South Korean financial markets by adopting a new method and variable.

3. Empirical analysis 3.1. Empirical model We estimate the impact of the investor attention to North Korean nuclear threats on South Korean financial markets using the structural VAR model with block exogeneity restrictions (see Lastrapes, 2005; Maćkowiak, 2007). ∑ where

(1)

is a vector of South Korea’s real macro-economic variables and financial

variables: industrial production, foreign exchange rate, long-term interest rate, and stock price index.

is the measure of the attention paid to North Korean nuclear risk based on

Google’s SVI. A21(s) = 0 for s = 0,1,…,T, which indicates the assumption of block exogeneity: the residuals in the South Korean financial markets, nuclear risks posed by North Korea, and E[

, even with time lags. ε

are residuals of the system that satisfy E[

′|

,

0

, do not affect the ≡

|

; ,

, where 0

0 and

. Additionally, we impose recursive zero restrictions on

contemporaneous structural parameters by applying the Cholesky decomposition (i.e., we set the order of variables in the vector

from the most exogenous to the most endogenous

variable as follows: industrial production, long-term bond yield, foreign exchange rate, and stock index). This identification scheme allows changes in the interest rate to have a contemporaneous impact on the other financial market variables, such as the exchange rate and stock price index, but not the reverse. To check the robustness of the results, we change the order of the variables in vector

, but the results are not sensitive to this change. We 11

include the Chicago Board Options Exchange (CBOE) volatility index (VIX) as an exogenous variable in the system to control for movements in the U.S./international equity market and/or the uncertain environment of the world economy. This VIX also captures possible exogenous shocks during the sample period, such as the global financial crisis. Lastly, we allow time lags of up to two weeks for the VAR system using the final prediction error (FPE) test and the Akaike information criterion. In the VAR system, we de-trend all variables except for the SVI with a quadratic time trend after taking a logarithm of them. For the robustness check, we also use first differenced variables, but the results do not alter.9

3.2. Empirical results 3.2.1. Benchmark results: Full sample vs. sub-samples Fig. 2 shows the impulse response functions of South Korea’s long-term interest rate, foreign exchange rate and stock price index to the SVI for North Korean nuclear threats (nuclear risk perceived by agents) using weekly data for 2004–2015. We compute the standard errors of the impulse response functions and draw 90% confidence intervals. During the whole sample period, while North Korean nuclear risk had insignificant effects on foreign exchange rate, it had negative impacts on South Korea’s long-term bond yield and stock price: the stock price and long-term interest rate responded significantly and negatively to the impulse of the shock. [Insert Fig. 2] Further, motivated by the data description in Table 2, we split our full sample into three subsamples around the time of each nuclear test within a four-year window and examine whether the effects of the nuclear risk on South Korea’s financial markets are

9

We perform the (augmented) Dickey-Fuller unit root test for de-trended (with quadratic trend and first differenced) variables. The null hypothesis that each time-series has a unit root is rejected at the 5% level. The test results are not sensitive to the number of lagged difference terms.

12

constant over time. The second column of Fig. 2 shows the impulse response functions to the nuclear risk for the first subsample, from 2004 to 2007. The attention paid to the nuclear threats significantly decreases the long-term bond rate and stock price two and three weeks after the impulse, respectively. A decrease in stock price in response to the nuclear risk implies that investors’ perception of nuclear risk may discourage their investment in the Korean stock market. The negative response of the bond yield to the nuclear risk also implies that the demand for South Korea’s long-term bonds (as safe assets) may increase as the risk emerges.10 The third and fourth columns in Fig. 2 show the impulse response functions to the nuclear risk shock for other subsample periods: 2008–2011 and 2012–2015. Unlike the result of the first sub-sample, the SVI for nuclear threats does not have significant impacts on stock prices and long-term interest rate in these cases. The significant effect of the SVI observed in the very first subsample period disappears. Interestingly, the impulse responses to the nuclear risk in the first sub-sample are only negative and much greater than those in the full sample, so we conjecture that the full sample result could be driven by the subsample result around the very first nuclear test. Moreover, the response of the long-term interest rate during 20122015 is negative but marginally significant. However, the response of foreign exchange rate to the SVI during the second subsample period turns out to be significantly positive three weeks after the impulse, which implies that the nuclear risk that emerged around the second nuclear test depreciated the South Korean won. This finding is consistent with a previous study by Kollias et al. (2012) but requires careful interpretation. Kollias et al. (2012) find that the second nuclear test had a greater negative effect on the currency markets of 10 Asian countries, including South Korea, 10

According to the data on portfolio flows provided by International Financial Statistics at the International Monetary Fund, equity liability inflows to South Korea shrank while debt liability inflows to South Korea increased from 2004 to 2007, indicating that foreign investors had been buying bonds and selling stocks, respectively, in South Korea’s financial market in that period.

13

than the first nuclear test did. They argue that North Korea announced the first test not only to South Korea but also to the international community (e.g., North Korea’s secession from the the Non-Proliferation Treaty (NPT) occurred in January 2003, and the market might have expected this test based on North Korea’s preparations for nuclear development), whereas the second test was relatively unexpected. Why, then, did domestic stock price and bond yield respond more to the SVI around only the first test while foreign exchange rate was sensitive to the SVI around the second test? Our stock price findings suggest that the nuclear risk affected the stock price significantly around the time of the first nuclear test, but that its impact has receded. In this regard, repeated risks of the same form of nuclear tests may have dulled investors’ sentiments toward the South Korean financial market. From the perspective of investor learning, investors may realize that the risks posed by North Korea’s nuclear tests tend to be quite short-lived and are not realized; thus, the impacts of the nuclear risk on financial markets appear to have become subdued over time. However, foreign exchange rate which is mainly driven by foreign investors in Korea did not respond significantly to the SVI for the nuclear threats around the first test. In fact, what news investors pay attention to totally depends on investors’ access to information. It is possible that the “information content” of the second test was larger for foreign investors than for domestic or residential investors in South Korea. In line with Kollias et al. (2012), the second test may have been more surprising to foreigners than to domestic investors and subsequently attracted more attention from foreigners. In the robustness check section, we also show that the impacts of the English SVI and the Korean SVI on exchange rate around the second nuclear test differ.

3.2.2. Time-varying structural VAR with a four-year rolling window 14

Previous subsample analysis treated each nuclear event as a separate sample, thus making the investor attention paid to each North Korean nuclear test mutually exclusive across the subsample analysis. One may argue that this arbitrary separation of subsamples can bias the results. In this subsection, we use the continuous time-varying structural VAR approach with a four-year rolling window and check whether the implication from the previous subsample regression continues to hold. The results are reported in Figs. 3 and 4. Fig. 3 is our baseline result and Fig. 4 is the result with first differenced variables for robustness. We trace the time-varying responses of each financial market variable to the SVI at three weeks from the impulse of the shock. Overall, the results in Figs. 3 and 4 are consistent with those in Fig. 2. In particular, the stock market response to the SVI shows distinctive changes over time: the nuclear risk significantly reduced the stock price index in the early 2000s, when North Korea conducted its first nuclear test. However, the negative impacts vanished over time, although North Korea has continued its provocations with its second and third nuclear tests. Around the time of the first nuclear test, the SVI also reduced the long-term bond yield significantly, which was related to the increased demand for bonds as safe assets. However, during the period 2008 to 2015, the nuclear risk did not decrease the yield significantly, as already shown in our subsample results in Fig. 2. These results show that investors had already become acclimatized to the “benign” effects of the North Korean nuclear risks and, hence, they even stopped seeking investment in safer assets.11 When discussing the combined responses of the stock price and bond yield to the SVI for North Korean nuclear threats, we observe that the significant negative effects of the perceived nuclear risk on stock price and bond yield are only observed around the first nuclear test.

11

It might also be possible that the investors became more cautious after the global financial crisis (namely, the Lehman Brothers bankruptcy in September 2008), which heightened global financial uncertainty. When events adding to the risk occurred, bond investors decided to withdraw investments from the bond market as well.

15

Again, we show that the depreciation effects of the nuclear risk on foreign exchange rate were distinct around the second nuclear tests. To evaluate the economic significance of the result, we compute a maximum marginal effect of the nuclear risk on stock price using the time-varying impulse responses. One standard deviation of the nuclear risk decreases stock price by 0.8% from the mean of the stock price index for the whole sample period, which corresponds to 13 points. This change is not negligible compared to the mean of stock price growth over the sample period, which is 0.05% (or the mean of stock price change during 2004–2007: 0.3%). [Insert Fig. 3] [Insert Fig. 4]

3.2.3. Validity of the measurement for nuclear risk One of the main contributions in this study is to introduce the SVI as a new measure for the North Korean nuclear risk from an investor’s viewpoint. To check the validity of our measurement of the attention to the nuclear threats, we introduce an alternative index that measures a positive news shock on the nuclear deadlock. As introduced in section 2.1, we use the SVI constructed using the keyword “six-party talks” and reiterate our time-varying analysis. Fig. 5 reports the results. Unlike the baseline results in Figs. 3 and 4, this new measurement had a positive effect on the South Korean financial market. Between the first and second nuclear tests, attention to the six-party talks is associated with appreciation of the Korean won. This implies that after the first nuclear test, the effort to resolve this nuclear issue via the six-party talks might have attracted more foreign capital flows into South Korea, which increased the demand for the Korean won. The measurement of this positive information shock about North Korea is also positively related to stock price. Significant positive effects of this measure on 16

stock price (two peaks in Fig. 5) are observed during 2007–2010 and 2010–2014. Thus, the results in Fig. 5 support that our baseline measurement related to North Korean nuclear threats is not contaminated with investor attention to positive news about North Korea. [Insert Fig. 5]

3.2.4. Cross-industry variation: Industry-specific stock price index We have focused on aggregate analyses to investigate the effect of the attention to North Korean nuclear threats on South Korea’s macro-financial variables. In particular, the stock price was hit severely by the perceived nuclear risk in the early 2000s. However, it is possible that the risk had a different effect on disaggregate stock prices (e.g., industry- or firm-level stock price indexes) for different market or industry characteristics. This subsection examines the responses of industry stock prices to the SVI for the nuclear threats. By introducing industry-specific stock price indexes, we gauge which industry was more sensitive to the nuclear risk. Thus, we include three de-trended industry stock variables following the aggregate stock price index in the structural VAR system of Equation (1). We consider industry stock price as the most endogenous variable, followed by turnover and industry stock trading volume. Fig. 6 shows the responses of the industry stock price index to the nuclear risk.12 Interestingly, our results show that the stock prices were significantly lowered by the attention to the risk only in banking industries, whereas those of the other industries showed insignificant responses.13 The vulnerability of the domestic banking industry to the nuclear risk is not surprising. It implies that the risk increased investors’ loan withdrawals from

12

The impulse responses of trading volume and turnover to the SVI are available from the authors upon request. In terms of trading volume and turnover, banking industry consistently shows negative responses to the SVI as well. 13

17

domestic banks, thus hurting the profitability of the banking industry. [Insert Fig. 6]

3.3. Robustness checks 3.3.1. Monthly frequency data, different identifications, etc. We conduct various tests to confirm the robustness of our results. First, we introduce a different identification scheme by changing the order of the financial variables and setting up the foreign exchange rate as the most endogenous. Second, we use different lags for our VAR model. Our baseline results use de-trended financial market variables. In the robustness check, we also introduce variables without de-trending. The results of all the robustness checks do not change our main findings in Figs. 3 and 4. Lastly, we repeat our baseline timevarying structural model using monthly data. In this analysis, we use real effective exchange rate instead of nominal foreign exchange rate.14 Fig. 7 reports the impulse response functions of three financial variables to the nuclear risk measurement at the impulse of the shock. The results in Fig. 6 are consistent with our main results in Figs. 3 and 4. [Insert Fig. 7] 3.3.2. Behavior of domestic South Korean investors We investigate whether the nationality of investors influenced their access to the information on the North Korean nuclear threats and shaped the responses to the nuclear risk. Different nationalities may reflect varying degrees of access to information, leading to different responses to the same incident. To distinguish the attention of the South Korean investors from that of all investors, we employ two different attention measurements to represent the differences in the degree of attention paid to the nuclear risk by South Korean and all investors and compare the results. For South Korean investors, the measure of 14

Monthly South Korean real effective exchange rate is used instead of nominal exchange rate, and is obtained from the Bank of International Settlement (BIS).

18

attention is constructed using the SVI, which refers to keywords for North Korea’s nuclear events in Korean only.15 Fig. 8 shows the results of the time-varying impulse responses to the nuclear risk. Overall, the results are consistent with our main findings in Figs. 3 and 4. The results also reveal that the attention to the nuclear threats paid by South Korean investors only has a slightly larger negative effect on stock price than that indicated by the same measure for all investors. However, the response of real exchange rate to the attention to the risk paid by South Korean investors (using the Korean SVI) is weaker than that paid by all investors (using the English SVI). Thus, it is arguable that while the second nuclear test was not as surprising as the first test for South Korean investors, this distinction is more clear for foreign investors. This is also consistent with our baseline findings on foreign exchange rate in Figs. 3 and 4. [Insert Fig. 8]

4. Conclusions and implications What is the “real” risk from political shocks and how does it affect financial market? Does an incident itself matter for investors in the financial market? Otherwise, what kinds of events do investors pay attention to and respond to? To answer these questions, this study focuses on the nuclear tests done by North Korea since early 2000s and investigates the dynamic impact of the investor attention paid to North Korean nuclear threats on South Korea’s financial markets. Using weekly data for the period 2004–2015, we first quantify the degree of investor attention paid to the North Korean nuclear threats using Google’s SVI. Then, we can determine the impact of the nuclear risk perceived by market participants on the South

15

One might argue that it is also possible that some South Korean investors search for information about North Korea’s nuclear threats in English. For example, we can imagine that a South Korean investor might use Google to search for “North Korean threats” and find and read a New York Times article about the topic. However, when discussing investments in the South Korean financial market, information about North Korea is likely to be delivered in more detail and faster through the Korean media than through the media in other countries. Thus, we rule out South Korean investors using Google to search for news about North Korea in the English media.

19

Korean financial markets more precisely over time. While the attention paid to nuclear threats has had negative impacts on the long-term bond yield and stock price index in South Korea, the impacts of the nuclear risk were particularly large and significant in the earlier period (2004–2007) compared to the later period (2008–2015). More importantly, its negative effects on the stock price and the bond yield had become subdued after the first North Korean nuclear test was completed. Another finding is that only the attention to the second nuclear threat was significantly associated with the depreciation of the Korean won. We also find heterogeneous responses of industryspecific stock prices to the nuclear risk. Stock price in the banking industry was particularly sensitive to the investor attention paid to the nuclear threats. North Korea’s continuing nuclear tests and claim to be a nuclear state have escalated geopolitical risk in the Korean peninsula and attracted severe criticism from the international community. However, our results imply that stock market investors did not respond significantly to the repeated North Korean nuclear threats, whereas a new provocation from North Korea, like the first North Korean nuclear test that occurred in 2006, could have a significant impact on the South Korean financial market outcome. Thus, additional nuclear threats from North Korea in the future could have different consequences for the South Korean financial market, which would depend not only on the characteristics of the threat but also how seriously investors pay attention to it. Our findings also support the idea that the investor attention paid to information plays an important role in their investing decision. Further, we provide the South Korean government and international organizations with practical ideas how to stabilize the financial market, especially in preparation for new types of threats from North Korea: Impeding the access to information never be a remedy. Rather, policy makers provide a clear source of information and help reduce uncertainty in the financial market, which will then make investor’s investing decision more predictable. 20

References Amihud, Y., Wohl, A., 2004. Political news and stock prices: The case of the Saddam Hussein contracts. J. Bank. Finance 28, 1185–1200. Barber, B.M., Odean, T., 2001. The Internet and the investor. J. Econ. Perspect. 15(1), 41–54. Berkman, H., Jacobsen, B., Lee, J. B., 2011. Time-varying rare disaster risk and stock returns. J. Financial Econ. 101(2), 313–332. Bittlingmayer, G., 1998. Output, stock volatility, and political uncertainty in a natural experiment: Germany, 1880–1940. J. Finance 53, 2243–2257. Brown, W., Burdekin, R., Weidenmier, M., 2006. Volatility in an era of reduced uncertainty: Lessons from Pax Britannica. J. Financial Econ. 79, 693–707. Choudhry, Taufiq. 2010. World War II events and the Dow Jones industrial index, J. Bank. Finance 34 (5), 1022-1031. Da, Z., Engelberg, J., Gao, P., 2011. In search of attention. The J. Finance 66(5), 1461–1499. Fisman, R., Hamao, Y., Wang, Y., 2014. Nationalism and economic exchange: Evidence from shocks to Sino-Japanese relations. Rev. Financial Stud. 27(9), 2626–2660. Frey, B., Kucher, M., 2000. History as reflected in capital markets: The case of World War II. J. Econ. Hist. 60, 468–496. Frijns, Bart, Alireza Tourani-Rad, and Ivan Indriawan. 2012. Political crises and the stock market integration of emerging markets, J. Bank. Finance 36 (3), 644-653. Huang, L., Liu, H., 2007. Rational inattention and portfolio selection. The J. Finance 62(4), 1999–2040. Jun, B.-K. 2011. Critical Review of and Lessons from the Nuclear Negotiations with North Korea for Twenty Years (in Korean), Korea World Politics 72, 183–212. Kahneman, D., 1973. Attention and Effort. Prentice Hall, New Jersey. Kim, B.-Y., Roland, G., 2011. Are the markets afraid of Kim Jong-Il? KIER Discussion 21

Paper 789. Kim, J.-C. 2009, North Korea’s strategies of crisis making and South Korea’s countermeasures (in Korean), Asia Stud. 12(1), 135–165. Kita, A., Wang, Q., 2012. Investor attention and FX market volatility. Mimeo. Kollias, C., Papadamou, S., Psarianos, I., 2014. Rogue state behavior and markets: The financial fallout of North Korean nuclear tests. Peace Econ., Peace Sci. Public Policy 20(2), 267–292. Lastrapes, William D. "Estimating and identifying vector autoregressions under diagonality and block exogeneity restrictions." Economics letters 87, no. 1 (2005): 75-81. Lee, J.-W., Pyun, J.H., 2016. Does trade integration contribute to peace? Rev. Dev. Econ. 20(1), 327–344. Lee, K.-Y. 2006. The effects of news about North Korea’s Nuclear Weapon on domestic stock and foreign exchange markets (in Korean). J. Northeast Asian Econ. Stud. 18(1), 61–90. Maćkowiak, B., 2007. External shocks, US monetary policy and macroeconomic fluctuations in emerging markets. J. Monet. Econ. 54(8), 2512–2520. Martin, P., Mayer, T., Thoenig, M., 2008. Make trade not war? Rev. Econ. Stud. 75(3), 865– 900. Mondria, J., Wu, T., 2012. Familiarity and surprises in international financial markets: Bad news travels like wildfire, good news travels slow! 2012 Meeting Papers, Society for Economic Dynamics. New

York

Times,

2013,

In

Focus:

North

Korea’s

Nuclear

Threats,

http://www.nytimes.com/interactive/2013/04/12/world/asia/north-korea-questions.html. Accessed on May 07, 2015. Rigobon, R., Sack, B., 2005. The effects of war risk on US financial markets. J. Bank. 22

Finance 29(7), 1769–1789. Smith, G.P., 2012. Google Internet search activity and volatility prediction in the market for foreign currency. Finance Res. Lett. 9, 103–110. Vlastakis, N., Markellos, R.N., 2012. Information demand and stock market volatility. J. Bank. Finance 36, 1808–1821. Waldenstrom, D., Frey, B., 2002. How government bond prices reflect wartime events: The case of the Stockholm market. University of Zurich, unpublished working paper. Wolfers, J., Zitzewitz, E., 2009. Using markets to inform policy: The case of the Iraq war. Economica 76, 225–250.

23

Appendix Appendix Table 1. Descriptive statistics of weekly data during 2004-2015 Variable

Obs.

Mean

S.D.

Min

Max

SVI

626

2.37

5.27

0.00

89.50

Won/USD

626

1091.73

108.08

906.70

1554.80

INT

626

3.78

1.08

1.58

6.07

KOSPI

626

1631.02

399.84

730.55

2205.94

VIX

626

19.36

9.21

10.19

72.92

Appendix Table 2. Monthly Industry stock price index, trading volume, and turnover Industry Food & Beverage

Textile & Apparel

Forestry & Paper

Chemicals

Pharmaceuticals

Nonferrous Metals

Iron & Steel

Industrial Engineering

Electronic & Electrical Equipment Biotechnology

Industrial Transportation

Retail trade & distribution

Variables Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume

Mean 2626.06 7852.628 93961.13 179.6952 10725.94 28672.84 299.7197 12080.09 21860.87 3005.767 22389.58 548198.8 3494.927 14629.16 87175.55 855.5948 4842.979 16818.46 4875.975 16988.42 253549.1 1118.056 22197.3 138958.2 7379.558 61176.65 977043.9 1211.214 18920.98 43525.49 1719.864 26102.46 601080.3 434.2785 27849.99

24

S.D. 840.4699 8106.909 44753.46 54.53294 11310.25 22211.12 64.99655 12902.1 11980.57 1305.395 10174.35 380103.8 973.3381 13889.03 65892.19 194.6133 4308.385 11836.52 1700.293 12763.94 153437.7 431.1462 16749.03 79422.69 2181.545 32174.07 325812.3 576.0944 40770.26 51012.7 836.0887 14133.71 367170 103.4008 19699.09

Min 1191.67 1050.8 28353 70.23 1054 5157 156.53 1984.3 6473 1030.53 6972.5 118022 1146.92 1959.5 9053 531.95 995.7 3733 1586.8 2720.6 53568 356.03 3620.9 28731 4111.47 20761.2 392293 227.31 1179.8 1487 472.46 6427.4 82257 216.23 5210.9

Max 4667.82 54043 433018 299.22 79927.5 158037 489.56 107263.8 95767 6158.12 59210.7 2410348 4902.18 65526.2 477876 1399.88 25855.7 61014 7904.44 68014.6 986151 2526.01 79943.2 445052 11804.37 195626.2 1730793 2179.51 312277.9 259661 3452.65 86611.6 2161468 642.48 115529.7

Utilities (Electricity & Gas)

Construction

Warehousing & Storage

Telecommunication

Banks

Equity Investments

Insurance

Services

Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover Stock price index Trading volume Turnover

256742.1 996.1322 2190.262 69651.89 190.1113 12993.37 186220.3 2151.79 22219.88 149422.1 310.3886 3366.605 107896.8 275.0801 5760.145 105348.2 2222.577 27986.67 221986.5 14748.34 4499.335 106020.2 759.6942 16697.84 354548.8

25

120387 191.3609 1881.661 45180.01 78.53776 9953.31 109350.5 720.7117 33898.98 102686.9 36.68737 1943.273 47340.12 65.44944 4946.91 89444.06 837.9159 26540.12 216918.8 4576.432 2662.931 83775.93 216.0594 10753.16 239121.7

49826 585.04 730.8 17033 55.09 2865.2 22376 774.69 3176.5 25253 205.61 921.1 39906 136.96 350.7 3930 751.07 3628.2 26740 4787.14 946.1 13790 355.96 2731.3 30405

753017 1380.87 20148.5 390340 444.31 52402 676137 4379.89 232650 754818 400 12583.1 300866 383.75 35311.3 464196 4714.93 152463.8 1212461 25294.5 14253.5 770036 1110.03 66425.1 1270463

Table 1. North Korea’s military and nuclear threats. Date 2002. 6. 29

2002. 10. 26 2003. 1. 10 2003. 2. 24

2005. 2. 10

Event Military dispute in the Yellow Sea around the Northern Limit Line (NLL) Revelations of its highly enriched uranium (HEU) program Secession from the NPT Missile launch Withdrawal from the Six Party Talks and official announcement of its capability to manufacture a nuclear weapon

2005. 5. 11

Extracts more fuel for nuclear weapons

2006. 7. 5

Missile launched

2006. 10. 9

First nuclear test conducted

2009. 4. 5 2009. 4. 29

2009. 5. 25 2009. 11. 10 2010. 3. 26 2010. 11. 23 2012. 4. 13

Missile launch Warning about another nuclear test and missile launch Second nuclear test conducted Military dispute occurs near the NLL Cheonan ship attack Yeonpyeong island attacks Satellites launched (ended in failure)

Note

Type

The dispute began with the unprovoked shooting by North Korean patrols.

Provocation

-

Nuclear

-

Nuclear Missile

North Korea officially announced its country’s capability to manufacture Nuclear nuclear weapons. North Korea shut down the Yongbyon Reactor, a move which could allow it to extract more fuel for nuclear weapons. North Korea test-fires six missiles, including a long-range Taepodong-2 rocket believed to be capable of reaching the west coast of the U.S. The test was conducted to coincide with the U.S.’ celebration of its Independence Day. The test was conducted to coincide with the U.S.’ celebration of Columbus Day. North Korea launched Bright Star-2.

Missile

North Korea launched the longrange Taepodong missile.

Nuclear /Missile

The test was conducted to coincide with the U.S.’ celebration of Memorial Day.

Nuclear

-

Provocation

-

Provocation Provocation

North Korea launched Bright Star-3.

26

Nuclear

Missile

Nuclear

2012. 12. 12

Satellites launched (successful)

North Korea relaunched Bright Star3 a week before the Presidential election was held in South Korea (2012. 12. 19)

2013. 2. 12

Third nuclear test conducted

-

2013. 3. 11

1953 War Truce nullified

North Korea declared that it will no longer abide by the 1953 Armistice that halted the Korean War. The announcement was conducted to coincide with the joint military drills between the U.S. and South Korea.

2013. 5.18

Missile launch

North Korea launched a total of six short-range projectiles.

Exchange fire across the North Korea and South Korea fired disputed western sea hundreds of artillery shells across border their disputed western sea border Source: Kim (2009), Jun (2011), and the NY Times (http://www.nytimes.com/interactive/2014/11/20/world/asia/northkoreatimeline.html?_r=0#/#time238_7128). 2014. 3. 31

27

Nuclear

Missile

Provocation

Table 2. Eyeball test: North Korea’s nuclear tests and their spillovers.

1st nuclear test (2006/10/09) 2nd nuclear test (2009/5/25) 3rd nuclear test (2013/2/12)

Foreign exchange market (Won/US Dollar) 0.8 ₩/$↑ (0.1%) 5.2 ₩/$↓ (0.4%) 4.9 ₩/$↑ (0.4%)

Bond market (3-year Korean treasury bill rate) 0.02%p ↑ (0.43%) 0.04%p ↓ (1.03%) 0.01%p ↓ (0.37%)

Stock market (KOSPI) 33p ↓ (2.4%) 3p ↓ (0.19%) 5p ↓ (0.26%)

Source: Bank of Korea, ECOS. Notes: Table 2 reports daily financial market responses on the days of North Korean nuclear tests. Arrows indicate the movement of financial variables. Values in the parentheses report the percent changes. Note that the won/U.S. dollar exchange rates data from another source of Federal Reserve Economic Data (FRED) show missing values on the days of the first and the second nuclear tests (October 9th, 2006 and May 25th, 2009). However, the exchange rate on the day after the first test (October 10th, 2006) is 959.2, which increases by 10.3 ₩/$ from that on October 6th, 2006. The exchange rate on the day after the second test (May 26th, 2009) is 1264.25, which also increases by 22.25 ₩/$ from that on May 22th, 2009.

28

Fig. 1. Google’s search volume index for North Korean nuclear threats 100 90 80 70 60 50 40 30 20 10 Jan-04 May-04 Sep-04 Jan-05 May-05 Sep-05 Jan-06 May-06 Sep-06 Jan-07 May-07 Sep-07 Jan-08 May-08 Sep-08 Jan-09 May-09 Sep-09 Jan-10 May-10 Sep-10 Jan-11 May-11 Sep-11 Jan-12 May-12 Sep-12 Jan-13 May-13 Sep-13 Jan-14 May-14 Sep-14 Jan-15 May-15 Sep-15

0

SVI (Nuclear)

Korean SVI (Nuclear)

SVI (Six-party talks)

Notes: 1) The red line with square markers indicates Google’s search volume index (SVI) measuring the attention agents give to the North Korean nuclear risk. The SVI is constructed by collecting data on Internet search activities that include the keywords “Korea nuclear” and “North Korea nuclear.” 2) The purple line with cross markers indicates Google’s SVI measuring the attention only South Korean investors give to the North Korean nuclear risk. The keywords include “북한핵실험” and “북한핵” in Korean. 3) The green line with circle markers indicates Google’s SVI using the keyword “six-party talks.” 4) We use data from January 2004 onwards because Google’s SVI debuted at that time. The gray area depicts the months in which North Korea conducted nuclear tests.

29

Fig. 2. Impulse responses of the South Korean financial markets to the SVI for the nuclear risk Full sample Sub-sample 1: 2004–2007 Sub-sample 2: 2008–2011 0.01 0.005 0 -0.005 -0.01 -0.015 -0.02 -0.025 -0.03 -0.035

NK risk → Long-term bond yield

-0.03

2

3

4

5

6

7

8

9 10

NK risk → Foreign exchange rate

0.002 0.001 0 -0.001 -0.002 0

1

2

3

4

5

6

7

8

9 10

-0.06

-0.03

-0.07

-0.04

-0.08

-0.05 1

0.004

2

3

4

5

6

7

8

9

NK risk → Foreign exchange rate

3

4

5

6

7

8

9 10

-0.03 -0.035

0.01

1

2

3

4

5

6

7

8

9

0

10

NK risk → Foreign exchange rate

0.004

0.003

0.008

0.003

0.002

0.006

0.002

0.001

0.004

0.001

0

0.002

0

-0.001

0

-0.001

-0.002

-0.002

-0.002

-0.004

-0.003 0

1

2

3

4

5

6

7

8

0

9 10 0.006

NK risk → Stock price

1

2

3

4

5

6

7

8

9 10

0.006 0.004

-0.005

0

0.002

-0.01

-0.002

0

-0.004

-0.002

-0.006

-0.004

-0.008 1

2

3

4

5

6

7

8

9 10

3

4

5

6

7

8

9 10

2

3

4

5

6

7

8

9 10

NK risk → Stock price

-0.006 0

30

1

0.008

NK risk → Stock price

2

NK risk → Foreign exchange rate

0

0.002

0

1

-0.003

0.004

-0.02 2

-0.02 -0.025

0

10

-0.015

1

-0.01

-0.02

NK risk → Long-term bond yield

-0.005 -0.015

-0.01

0

0

0

0

-0.05

0

0.005

0.01

-0.04

0.005

NK risk → Stock price

NK risk → Long-term bond yield

0.02

-0.02

1

0.04

Sub-sample 1: 2011–2015

0.03

-0.01

0.003

0.006 0.004 0.002 0 -0.002 -0.004 -0.006 -0.008 -0.01 -0.012

NK risk → Long-term bond yield

0

0 0.004

0.01

0.05

1

2

3

4

5

6

7

8

9 10

0

1

2

3

4

5

6

7

8

9 10

Note: The first column reports the impulse responses of three financial market variables (bond, foreign exchange rate, stock price) to one standard deviation shock on the SVI for the nuclear risk. The remaining columns include sub-sample analyses for 2004–2007, 2008–2011, and 2012–2015. The red lines indicate the impulse response over weeks. The dotted lines represent 90% confidence intervals. Financial market variables are de-trended with a quadratic time trend.

31

2004w1~2007w52 2004w11~2008w10 2004w21~2008w20 2004w31~2008w30 2004w41~2008w40 2004w51~2008w50 2005w9~2009w8 2005w19~2009w18 2005w29~2009w28 2005w39~2009w38 2005w49~2009w48 2006w7~2010w6 2006w17~2010w16 2006w27~2010w26 2006w37~2010w36 2006w47~2010w46 2007w5~2011w4 2007w15~2011w14 2007w25~2011w24 2007w35~2011w34 2007w45~2011w44 2008w3~2012w2 2008w13~2012w12 2008w23~2012w22 2008w33~2012w32 2008w43~2012w42 2009w1~2012w52 2009w11~2013w10 2009w21~2013w20 2009w31~2013w30 2009w41~2013w40 2009w51~2013w50 2010w9~2014w8 2010w19~2014w18 2010w29~2014w28 2010w39~2014w38 2010w49~2014w48 2011w7~2015w6 2011w17~2015w16 2011w27~2015w26 2011w37~2015w36 2011w47~2015w46

-0.08

0.01 0.008 0.006 0.004 0.002 0 -0.002 -0.004 -0.006

0.015

-0.02 2004w1~2007w52 2004w11~2008w10 2004w21~2008w20 2004w31~2008w30 2004w41~2008w40 2004w51~2008w50 2005w9~2009w8 2005w19~2009w18 2005w29~2009w28 2005w39~2009w38 2005w49~2009w48 2006w7~2010w6 2006w17~2010w16 2006w27~2010w26 2006w37~2010w36 2006w47~2010w46 2007w5~2011w4 2007w15~2011w14 2007w25~2011w24 2007w35~2011w34 2007w45~2011w44 2008w3~2012w2 2008w13~2012w12 2008w23~2012w22 2008w33~2012w32 2008w43~2012w42 2009w1~2012w52 2009w11~2013w10 2009w21~2013w20 2009w31~2013w30 2009w41~2013w40 2009w51~2013w50 2010w9~2014w8 2010w19~2014w18 2010w29~2014w28 2010w39~2014w38 2010w49~2014w48 2011w7~2015w6 2011w17~2015w16 2011w27~2015w26 2011w37~2015w36 2011w47~2015w46

0.04

2004w1~2007w52 2004w11~2008w10 2004w21~2008w20 2004w31~2008w30 2004w41~2008w40 2004w51~2008w50 2005w9~2009w8 2005w19~2009w18 2005w29~2009w28 2005w39~2009w38 2005w49~2009w48 2006w7~2010w6 2006w17~2010w16 2006w27~2010w26 2006w37~2010w36 2006w47~2010w46 2007w5~2011w4 2007w15~2011w14 2007w25~2011w24 2007w35~2011w34 2007w45~2011w44 2008w3~2012w2 2008w13~2012w12 2008w23~2012w22 2008w33~2012w32 2008w43~2012w42 2009w1~2012w52 2009w11~2013w10 2009w21~2013w20 2009w31~2013w30 2009w41~2013w40 2009w51~2013w50 2010w9~2014w8 2010w19~2014w18 2010w29~2014w28 2010w39~2014w38 2010w49~2014w48 2011w7~2015w6 2011w17~2015w16 2011w27~2015w26 2011w37~2015w36 2011w47~2015w46

Fig. 3. Time-varying impulse-responses to the SVI for the nuclear risk NK risk→ Long-term bond yield

0.02

0

-0.02

-0.04

-0.06

NK risk→ Foreign exchange rate

0.01

NK risk→Stock price

0.005

0

-0.005

-0.01

-0.015

Note: The red lines indicate the impulse response of each variable three weeks after the impulse of one standard deviation shock on the SVI for the nuclear risk. The dotted lines represent 90% confidence intervals. Financial market variables are de-trended with a quadratic time trend. We use a four-year rolling window from the first week of 2004 and the last week (52nd) of 2007.

32

0.006 0.005 0.004 0.003 0.002 0.001 0 -0.001 -0.002 -0.003 -0.004

2004w1~2007w52 2004w11~2008w10 2004w21~2008w20 2004w31~2008w30 2004w41~2008w40 2004w51~2008w50 2005w9~2009w8 2005w19~2009w18 2005w29~2009w28 2005w39~2009w38 2005w49~2009w48 2006w7~2010w6 2006w17~2010w16 2006w27~2010w26 2006w37~2010w36 2006w47~2010w46 2007w5~2011w4 2007w15~2011w14 2007w25~2011w24 2007w35~2011w34 2007w45~2011w44 2008w3~2012w2 2008w13~2012w12 2008w23~2012w22 2008w33~2012w32 2008w43~2012w42 2009w1~2012w52 2009w11~2013w10 2009w21~2013w20 2009w31~2013w30 2009w41~2013w40 2009w51~2013w50 2010w9~2014w8 2010w19~2014w18 2010w29~2014w28 2010w39~2014w38 2010w49~2014w48 2011w7~2015w6 2011w17~2015w16 2011w27~2015w26 2011w37~2015w36 2011w47~2015w46

-0.02 2004w1~2007w52 2004w11~2008w10 2004w21~2008w20 2004w31~2008w30 2004w41~2008w40 2004w51~2008w50 2005w9~2009w8 2005w19~2009w18 2005w29~2009w28 2005w39~2009w38 2005w49~2009w48 2006w7~2010w6 2006w17~2010w16 2006w27~2010w26 2006w37~2010w36 2006w47~2010w46 2007w5~2011w4 2007w15~2011w14 2007w25~2011w24 2007w35~2011w34 2007w45~2011w44 2008w3~2012w2 2008w13~2012w12 2008w23~2012w22 2008w33~2012w32 2008w43~2012w42 2009w1~2012w52 2009w11~2013w10 2009w21~2013w20 2009w31~2013w30 2009w41~2013w40 2009w51~2013w50 2010w9~2014w8 2010w19~2014w18 2010w29~2014w28 2010w39~2014w38 2010w49~2014w48 2011w7~2015w6 2011w17~2015w16 2011w27~2015w26 2011w37~2015w36 2011w47~2015w46

0.02

0.015

0.002 0.0015 0.001 0.0005 0 -0.0005 -0.001 -0.0015 -0.002 2004w1~2007w52 2004w11~2008w10 2004w21~2008w20 2004w31~2008w30 2004w41~2008w40 2004w51~2008w50 2005w9~2009w8 2005w19~2009w18 2005w29~2009w28 2005w39~2009w38 2005w49~2009w48 2006w7~2010w6 2006w17~2010w16 2006w27~2010w26 2006w37~2010w36 2006w47~2010w46 2007w5~2011w4 2007w15~2011w14 2007w25~2011w24 2007w35~2011w34 2007w45~2011w44 2008w3~2012w2 2008w13~2012w12 2008w23~2012w22 2008w33~2012w32 2008w43~2012w42 2009w1~2012w52 2009w11~2013w10 2009w21~2013w20 2009w31~2013w30 2009w41~2013w40 2009w51~2013w50 2010w9~2014w8 2010w19~2014w18 2010w29~2014w28 2010w39~2014w38 2010w49~2014w48 2011w7~2015w6 2011w17~2015w16 2011w27~2015w26 2011w37~2015w36 2011w47~2015w46

Fig. 4. Time-varying impulse responses with first differenced variables NK risk→ Long-term bond yield

0.01

0.005

0

-0.005

-0.015 -0.01

NK risk→ Foreign exchange rate

NK risk→Stock price

Note: The red lines indicate the impulse response of each variable three weeks after the impulse of one standard deviation shock on the SVI for the nuclear risk. The dotted lines represent 90% confidence intervals. Financial market variables are first differenced to avoid non-stationarity. We use a four-year rolling window from the first week of 2004 and the last week (52nd) of 2007.

33

2004w1~2007w52 2004w13~2008w12 2004w25~2008w24 2004w37~2008w36 2004w49~2008w48 2005w9~2009w8 2005w21~2009w20 2005w33~2009w32 2005w45~2009w44 2006w5~2010w4 2006w17~2010w16 2006w29~2010w28 2006w41~2010w40 2007w1~2010w52 2007w13~2011w12 2007w25~2011w24 2007w37~2011w36 2007w49~2011w48 2008w9~2012w8 2008w21~2012w20 2008w33~2012w32 2008w45~2012w44 2009w5~2013w4 2009w17~2013w16 2009w29~2013w28 2009w41~2013w40 2010w1~2013w52 2010w13~2014w12 2010w25~2014w24 2010w37~2014w36 2010w49~2014w48 2011w9~2015w8 2011w21~2015w20 2011w33~2015w32 2011w45~2015w44 2004w1~2007w52 2004w12~2008w11 2004w23~2008w22 2004w34~2008w33 2004w45~2008w44 2005w4~2009w3 2005w15~2009w14 2005w26~2009w25 2005w37~2009w36 2005w48~2009w47 2006w7~2010w6 2006w18~2010w17 2006w29~2010w28 2006w40~2010w39 2006w51~2010w50 2007w10~2011w9 2007w21~2011w20 2007w32~2011w31 2007w43~2011w42 2008w2~2012w1 2008w13~2012w12 2008w24~2012w23 2008w35~2012w34 2008w46~2012w45 2009w5~2013w4 2009w16~2013w15 2009w27~2013w26 2009w38~2013w37 2009w49~2013w48 2010w8~2014w7 2010w19~2014w18 2010w30~2014w29 2010w41~2014w40 2010w52~2014w51 2011w11~2015w10 2011w22~2015w21 2011w33~2015w32 2011w44~2015w43

-0.08 2004w1~2007w52 2004w12~2008w11 2004w23~2008w22 2004w34~2008w33 2004w45~2008w44 2005w4~2009w3 2005w15~2009w14 2005w26~2009w25 2005w37~2009w36 2005w48~2009w47 2006w7~2010w6 2006w18~2010w17 2006w29~2010w28 2006w40~2010w39 2006w51~2010w50 2007w10~2011w9 2007w21~2011w20 2007w32~2011w31 2007w43~2011w42 2008w2~2012w1 2008w13~2012w12 2008w24~2012w23 2008w35~2012w34 2008w46~2012w45 2009w5~2013w4 2009w16~2013w15 2009w27~2013w26 2009w38~2013w37 2009w49~2013w48 2010w8~2014w7 2010w19~2014w18 2010w30~2014w29 2010w41~2014w40 2010w52~2014w51 2011w11~2015w10 2011w22~2015w21 2011w33~2015w32 2011w44~2015w43

Fig. 5. Time-varying impulse responses to the SVI for six-party talks

0.06

0.04

Six-party talks → Long-term bond yield

0.02

0

-0.02

-0.04

-0.06

0.008 0.006 0.004 0.002 0 -0.002 -0.004 -0.006 -0.008 -0.01

0.015

Six-party talks → Foreign exchange rate

Six-party talks→Stock price

0.01

0.005

0

-0.005

-0.01

Note: The red lines indicate the impulse response of each variable three weeks after the impulse of one standard deviation shock on the SVI for the six-party talks. The dotted lines represent 90% confidence intervals. Financial market variables are de-trended with a quadratic time trend. We use a four-year rolling window from the first week of 2004 and the last week (52nd) of 2007.

34

Fig. 6. Impulse responses of the industry stock price indexes to the SVI for the nuclear risk NK risk -> Stock price

NK risk -> Stock price

NK risk -> Stock price

Food & Beverage

Textile & Apparel

Forestry & Paper

.02

.04 .02

0

NK risk -> Stock price Chemicals

.04

.04

.02

.02

0

0

0 -.02

-.02

-.04

-.04 0

5

10

-.02 0

5

-.02

10

0

5

10

0

5

10

NK risk -> Stock price

NK risk -> Stock price

NK risk -> Stock price

NK risk -> Stock price

Phamaceuticals

Nonferrous Metals

Iron & Steel

Industrial Engineering

.02

.02

.01

.04

.02

0

.02

0

-.02

0

-.02

0 -.01 -.02

-.04 0

5

10

-.02 0

5

-.04

10

0

5

10

NK risk -> Stock price

NK risk -> Stock price

NK risk -> Stock price

Electronic & Eletronical Equip.

Biotechnology

Industrial Transportation

.02

.04

.05

0 -.01

-.05 5

10

NK risk -> Stock price Retail trade & Distribution

0 -.02

-.02

0

10

0

0

-.02

5

.02

.02

.01

0

-.04 0

5

-.04

10

0

5

10

0

5

10

NK risk -> Stock price NK risk -> Stock price NK risk -> Stock price NK risk -> Stock price Electricity & Gas

Construction

Warehousing & Stroage .02

.02

.01 0

0

-.01

-.02

-.02

0

5

10

NK risk -> Stock price

5

10

-.01 0

Equity Investments

10

0

Insurance

.02

.02

-.02

0

.01

-.04

-.02

0

-.06

-.04

-.08

-.06 10

5

5

10

NK risk -> Stock price NK risk -> Stock price NK risk -> Stock price

0

5

0

-.04 0

Banks

0

.01

-.02

-.04 0

Telecommunication .02

Services .02

0

-.01 -.02

-.02 0

5

10

0

5

10

0

5

10

Note: Monthly industry stock price variables are used. In the structural VAR system, both industry stock price and stock trade volume and turnover are included. Industry stock price is considered most endogenous in the VAR system. The red lines indicate the impulse responses of industry stock prices to the SVI for the nuclear risk. The dotted lines represent 90% confidence intervals. All financial market variables are de-trended with a quadratic time trend.

35

-0.04

36 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

2011m5~2015m1

2011m10~2015m6

2012m3~2015m11

2009m4~2012m12

2008m11~2012m7

2008m6~2012m2

2008m1~2011m9

2007m8~2011m4

2010m12~2014m8

-0.03 2010m7~2014m3

-0.02

2010m7~2014m3

-0.01 2010m2~2013m10

0

2010m2~2013m10

0.01 2009m9~2013m5

NK risk → Stock Price

2009m9~2013m5

2009m4~2012m12

2008m11~2012m7

2008m6~2012m2

2008m1~2011m9

2007m8~2011m4

2007m3~2010m11

2006m10~2010m6

0.01

2007m3~2010m11

0.02 2006m5~2010m1

2005m12~2009m8

2005m7~2009m3

2005m2~2008m10

2004m9~2008m5

0.015

2006m10~2010m6

0.03

2006m5~2010m1

2005m12~2009m8

2005m7~2009m3

2005m2~2008m10

2004m9~2008m5

-0.015 2004m4~2007m12

2012m3~2015m11

2011m10~2015m6

2011m5~2015m1

2010m12~2014m8

2010m7~2014m3

2010m2~2013m10

2009m9~2013m5

2009m4~2012m12

2008m11~2012m7

2008m6~2012m2

2008m1~2011m9

2007m8~2011m4

2007m3~2010m11

2006m10~2010m6

2006m5~2010m1

2005m12~2009m8

2005m7~2009m3

2005m2~2008m10

2004m9~2008m5

2004m4~2007m12

0.08 0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08 -0.1 -0.12

2004m4~2007m12

Fig. 7. Time-varying impulse responses to the SVI for the nuclear risk with monthly data

NK risk → Long-term bond yield

NK risk → REER

0.005

0

-0.005

-0.01

Note: The red lines indicate the impulse response of each variable three weeks after the impulse of one standard deviation shock on the SVI for the nuclear risk. The dotted lines represent 90% confidence intervals. Financial market variables are de-trended with a quadratic time trend. We use a four-year rolling window from the first month of 2004 and the last month of 2007.

2004m4~2007m12 2004m9~2008m5 2005m2~2008m10 2005m7~2009m3 2005m12~2009m8 2006m5~2010m1 2006m10~2010m6 2007m3~2010m11 2007m8~2011m4 2008m1~2011m9 2008m6~2012m2 2008m11~2012m7 2009m4~2012m12 2009m9~2013m5 2010m2~2013m10 2010m7~2014m3 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

0.03

0.02 0.1

0.05

-0.05

0.012 0.01 0.008 0.006 0.004 0.002 0 -0.002 -0.004 -0.006 -0.008 -0.01

NK risk → REER

NK risk → Stock Price

0.01

0

-0.01

-0.02

-0.03

-0.04

37

2004m4~2007m12 2004m9~2008m5 2005m2~2008m10 2005m7~2009m3 2005m12~2009m8 2006m5~2010m1 2006m10~2010m6 2007m3~2010m11 2007m8~2011m4 2008m1~2011m9 2008m6~2012m2 2008m11~2012m7 2009m4~2012m12 2009m9~2013m5 2010m2~2013m10 2010m7~2014m3 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

NK risk → Interest rate

2004m4~2007m12 2004m9~2008m5 2005m2~2008m10 2005m7~2009m3 2005m12~2009m8 2006m5~2010m1 2006m10~2010m6 2007m3~2010m11 2007m8~2011m4 2008m1~2011m9 2008m6~2012m2 2008m11~2012m7 2009m4~2012m12 2009m9~2013m5 2010m2~2013m10 2010m7~2014m3 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

2004m4~2007m12 2004m9~2008m5 2005m2~2008m10 2005m7~2009m3 2005m12~2009m8 2006m5~2010m1 2006m10~2010m6 2007m3~2010m11 2007m8~2011m4 2008m1~2011m9 2008m6~2012m2 2008m11~2012m7 2009m4~2012m12 2009m9~2013m5 2010m2~2013m10 2010m7~2014m3 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

0.15 0.1

-0.1 -0.1

-0.15 -0.15

0.01 0.008 0.006 0.004 0.002 0 -0.002 -0.004 -0.006 -0.008 -0.01 -0.012

0.02 0.015 0.01 0.005 0 -0.005 -0.01 -0.015 -0.02 -0.025 -0.03 -0.035

2004m4~2007m12 2004m9~2008m5 2005m2~2008m10 2005m7~2009m3 2005m12~2009m8 2006m5~2010m1 2006m10~2010m6 2007m3~2010m11 2007m8~2011m4 2008m1~2011m9 2008m6~2012m2 2008m11~2012m7 2009m4~2012m12 2009m9~2013m5 2010m2~2013m10 2010m7~2014m3 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

2004m4~2007m12 2004m9~2008m5 2005m2~2008m10 2005m7~2009m3 2005m12~2009m8 2006m5~2010m1 2006m10~2010m6 2007m3~2010m11 2007m8~2011m4 2008m1~2011m9 2008m6~2012m2 2008m11~2012m7 2009m4~2012m12 2009m9~2013m5 2010m2~2013m10 2010m7~2014m3 2010m12~2014m8 2011m5~2015m1 2011m10~2015m6 2012m3~2015m11

Fig. 8. Time-varying impulse responses to the English SVI and Korean SVI English SVI Korean SVI only NK risk → Interest rate

0.05

0 0

-0.05

NK risk → REER

NK risk → Stock Price

Note: The red lines indicate the impulse response of each variable three weeks after the impulse of one standard deviation shock on the SVI for the nuclear risk. The dotted lines represent 90% confidence intervals. Financial market variables are de-trended with a quadratic time trend. We use a four-year rolling window from the first month of 2004 and the last month of 2007.