DETERMINANTS OF COMMON FACTORS IN KOREAN BANKS’ CREDIT DEFAULT SWAP PREMIUMS

Using the panel analysis of non-stationarity in idi osycratic and common component method, we decompose Credit Default Swap (CDS) premium data of 11Korean banks into common factors and idiosyncratic shocks. We find that the CDS premium of all 11 banks is mostly explained by one common factor. We also find that the common factor of the banks’ CDS premium is mainly affected by the level and the volatility of stock market prices in develo ped markets and oil prices. It suggests that the Ko rean banking industry is susceptible to foreign shocks d ue to the heavy dependency of the Korean economy on export. We also find that a structural break in the common part of CDS premium occurred in mid-2007, implying that the exposure of credit risk in Korean banks jumped up after the 2007 financial crisis.


INTRODUCTION
The explosion and dramatic reversal of capital flows among international markets since the 1990s have ignited a heated debate. Some people argue that globalization has gone too far and that international capital markets have become extremely erratic. Conversely, others claim that globalization allows capital to move to where it is mostly needed in promoting economic growth. After the currency crisis in the late 1997, Korea has gradually opened its financial markets to promote foreign investment. Since the currency crisis, a series of institutional changes was implemented to facilitate the direct foreign investment. The changes included (1) opening the corporate bond market (December 1997), (2) allowing the purchase of shortterm financial products (February 1998), (3) abolishing the limit of domestic equity investment (May 1998), (4) allowing hostile M&A activities (April 1998), (5) opening more industries for foreign investment (May 1998) and (6) enacting the Foreign Investment Promotion Act (November 1998). Especially, in September 1998 the Foreign Exchange Management Act was abolished. Subsequently, in April 1999 the Foreign Exchange Trade Act was enacted and implemented to minimize regulations on foreign trade and to expand foreign exchange trading. With a series of institutional changes, Korean financial markets have become more volatile and more vulnerable to foreign shocks. When negative economic news comes from foreign countries, Korean financial markets could immediately be slashed by large capital outflows. Skeptical expectations on the Korean economy due to a decrease in exports and changes in portfolio may lead foreign investors to withdraw their fund from the Korean markets. As a result, sequential capital outflows induced shortage of liquidity in the domestic market, which had negative impacts on the Korean economy in the short run.
In particular, more liberalized Korean financial markets were thrown into turmoil when the subprime mortgage Science Publications AJEBA crisis in the United States broke out in September 2008. The subprime crisis is attributed to the problem of the United States economy due to the failures of asset management strategy of the U.S. financial institutions.
The crisis unfortunately affected Korean financial markets and the economy through withdrawal of foreign funds which led the Korean economy to be in severe liquidity shortage and a credit crunch. As Korean financial markets could not function well under a credit crunch, credit risks of lending increased and accordingly the credit default swap (hereafter CDS) premium soared. Korean banks' CDS premiums seemed to especially be susceptible to foreign shocks. Financial institutions would be exposed to a higher credit risk of lending due to the increase in bankruptcy risk of Korean firms. Therefore, their CDS premium rises.
Various methods have been employed to measure bank risk in the existing literature. Those methods include some alternative measures of firm risk such as subordinated debt spread (Krishnan et al., 2006) and expected default frequency calculated by an option pricing model (Altman and Hotchkiss, 2005). In addition, the CDS premium or spread has been increasingly popular as a simple indicator of bank credit risk. A CDS is a bilateral transaction under which the buyer is insured against credit risk and pays premium to the seller. The CDS premium is expressed as a function of the nominal value of the contract. Previous studies investigating the pricing of CDS premium claim that CDS premium is an efficient measure of credit risk. For example, Longstaff et al. (2005) claim that CDS spreads appropriately reflect credit risk. Kim et al. (2010) finds that the CDS spreads for Asian borrowers widened during the 2007-2009 crisis because of high expected default frequency.
Besides individual risk, researchers become more interested in systemic risk in the financial sector after a financial crisis. For instance, Bijlsma et al. (2010) review main literature investigating reasons for systemic risk and policy implications of systemic risk. Because of externalities, contagion and spillover inherent in financial markets, we must be concerned about systemic risk as well as individual risk. Systemic risk is measured by various indicators: principal components of the banks' CDS (Billio et al., 2010), spillover index (Diebold and Yilmaz, 2009), dynamic conditional correlation (Rahman, 2014) and co-risk measures (Adrian and Brunnermeir, 2011) and so on.
A few studies exploit the common factor as a measure of systemic risk. Kool (2006) investigates the role of common factors in European bank CDS spreads for financial stability and documents that the common factor is related to the European P/E ratio and the European 2-yaer nominal interest rate. Applying a dynamic factor model to the distance-todefault of EU banks, Brasili and Vulpes (2006) find that the commonality in bank risk appears to have increased since 1999. Eichengreen et al. (2012) recently report that common movement of banks' CDS spreads rose after the subprime crisis, using principal components analysis. Rahman (2014) also finds the extreme co-movements of financial institutions' default swap contracts in the aftermath of the subprime crisis.
We focus on common factors of Korean banks' CDS premium measure to estimate an indicator of systemic risk. Following the Bai and Ng (2004) method, we extract common factors of banks' CDS premium. After exploring the properties of the common factors, we attempt to select an optimal number of common factors and to find determinants of the common factors.
The main empirical findings are as follows: First, most variation of individual bank CDS premium is explained by a common factor. Second, the common factor of bank CDS premium is strongly affected by the level and volatility of stock prices in the developed market. In addition, the common factor is affected by spot oil price and sovereign bond rate. Finally, there was a structural break in the common part in August 2007as a result of contagion of the subprime crisis in the U.S.
We offer some policy implications from the findings. First, individual bank's CDS premium has a strong tendency to move in the same direction, indicating that the Korean banking in dustry is exposed to a substantial systemic risk. Second, because systemic risk is strongly susceptible to foreign capital out flows due to changes in the foreign macro-financial economic condition, regulatory efforts should be made to minimize the impact of foreign capital outflows on Korean financial markets and economy.

METHODOLOGY
The factors affecting the CDS premium can be categorized into macro-financial variables and firm specific variables mostly reflecting balance sheet information. The firm's specific variables include leverage, equity return, idiosyncratic volatility, the price to book ratio and credit ratings. On the other hand, macro-financial variables cover interest rates, term structure, equity market returns, equity market volatilities, macroeconomic conditions, sovereign bond yields and country credit ratings for sovereign bonds. In

Science Publications
AJEBA particular, the bank CDS premium in emerging markets would respond to the movements of the capital flows due to changes in macroeconomic conditions.
Because changes in the aggregate macroeconomic environment would affect CDS premium of all banks, the common factors extracted from CDS premium of banks would be explained by macro-financial variables. An approximate factor model is intuitively appealing in observing how the common factors of individual banks have reacted to the changes in the macroeconomic environment.
We decompose the CDS spreads across Korean banks into one or more common factors and idiosyncratic components attributable to individual firms by identifying common factors suggested by Bai and Ng (2004). Next, we attempt to find out what macrofinancial variables have determined the common factors among CDS premiums of Korean banks. The determinants might be closely related to the stability of the Korean banking industry and ultimately to the stability of the Korean economy.

Factor Model
Let X it be the observed CDS spread for the ith bank at time t, for i=1, …, N and t=1, …, T. Consider the following model: where, e it is the idiosyncratic component of X it with a zero mean and is orthogonal to F t , which is a vector of common factors. λ i is a vector of factor loadings related to F t . λ i F t is called the common component of X it .
Equation 1 is then the factor representation of the data which has two unobserved components-common factor and idiosyncratic components. Common factor F t can be estimated by taking the first difference of Equation 1 as follows: where, f t = ∆F t . By applying the principal component analysis to ∆x i,t , estimates of r factors of ˆt f are obtained. To determine the number of common factors r in Equation 2, the following criterion is adopted, which is the most robust under the presence of cross correlations among the idiosyncratic components: Where: The information criteria reflect the trade-off between the goodness-of-fit and over fitting. The first term on the right shows the goodness-of-fit given by the residual sum of squares, which depends on the estimates of the number of factors. If the number of the factors r increases, variance of the factors f t also increases while the sum of squared residuals decreases. The penalty of over fitting, which is the second term on the right, is an increasing function of the cross-section size N and time series length T. The optimal number of factors minimizes IC (r).
After the optimal number of the common factors is determined based on Equation 3, CDS spread data was decomposed into r common factors and idiosyncratic component of bank i's premium. With the calculatedcommon factor premium among Korean banks, this study investigates what affects common factor premium by employing regression analysis.

Data
Data consists of a balanced panel of daily CDS premium of 11 major Korean banks as a direct measure of credit spreads. The periodic payment expressed in basis points is called CDS premium. By definition, it provides a pure measure of the default risk of the reference entity. The sample covers data from January23, 2006 to April18, 2011 and includes 1,366 observations. The Korean banking industry consists of seven major commercial banks, five specialized banks and sixlocal banks. Because of some banks having insufficient trading records, the data includes only seven major banks and four specialized banks. Specialized banks were established with specific purposes of bolstering financing in specific areas facing funding difficulties due to profitability and expertise, based on the Special Act and run by the Korean government. The CDS data was extracted from Bloomberg. Other data-representing macroeconomic conditions-was derived from the Korean Center for International Finance. Table 1 reports summary statistics for the CDS premium of each bank. Most banks experienced a mean premium of 121.2 to 157.8 basis points over the sample period. The first four banks and Citibank Korea Inc. demonstrate relatively lower mean than the rest.   Since the first four banks are special banks considerably controlled by the Korean government, they would be perceived as relatively less risky. Standard deviations for most banks range from 124.4 to 169.8 basis points. In general, the larger the mean, the larger the standard deviation is. Private banks, in particular, experienced more volatile movements of premium over the sample period. The CDS premium for most banks soared above 800 basis points around late October 2008 right after the financial crisis triggered by the Lehman Brothers collapse in the United States. Since then, the CDS premium demonstrated alow and stable movement around 11.9 to 16.5 basis points until June 2007.

AJEBA
We implement unit root tests to check the stationarity of the CDS premium of Korean banks. The results in Table 2 show that every series are non-stationary in level while they are stationary in the first difference.

Common Part of CDS Premium
We use the method proposed by Bai and Ng (2004) to extract the common factors corresponding to the latent risk dimensions in the CDS premium. Before determining the number of common factors, we conduct the crosssection dependence test suggested by Breusch and Pagan (1980) in order to check whether the crosssection dependence exists among banks' CDS premium. The test result provides evidence that the CDS premium series are dependent upon each other. The results are presented in Table 3.
In order to find the optimal number of common factors, we employ Equation 3 and calculate the value of IC(r). The number of common factors is tested up to 8.
The result in Table 4 shows that the lowest value is-7.042 when the number of common factor is one (r = 1). Here, r represents a number of common components while IC stands for value of information criteria suggested by Bai and Ng (2002). Hence, the CDS premium data of 11 Korean banks is decomposed into one common factor and eleven idiosyncratic series.
The estimated common factor explains approximately 98.5% of the total variations of the CDS premiums. That is, the variations of the CDS premium are mostly explained by the estimated common factor.  The average during that period increased to 0.46. A major reason for continuous increase in the CDS premium is that the Korean capital market was so closely linked to the U.S. capital market and hence was affected by the subprime mortgage turmoil and global financial crisis. A tremendous outflow of foreign funds drove the Korean economy into a damaging situation due to a sharply decreasing liquidity supply. Naturally, the sequential credit crunch led to difficulties in financing for business firms. As the financial status of Korean firms worsened, the CDS premium of banks sharply increased. After adjusting and recovering from the financial crisis shock, the CDS premium gradually lowered down to 0.55 on January 12, 2010. It increased to around 0.69 since then but never returned to the same level that it reached in 2006 for a substantial period of time. This indicated that the global financial system was not yet fully recovered and stable. That is, the Korean economy could not completely be independent from financial crisis shocks and the Korean economy and banking industry were still in danger.

Determinants of Common Factor of the CDS Premium
Because of the frequency of the daily CDS premium, we select the macro-financial data publicly announced daily. The variables arelimited to the movements of foreign and domestic financial markets, currency markets and commodity markets. Suh and Lee (2011) take into account per capita GDP, GDP growth rate, foreign reserves, fiscal balance, current balance as macro variables that determine the CDS premium. Considering the limitation of daily data availability, we initially employa FTSE index for the developed markets (hereafter FTSED) and a FTSE index for the emerging markets (hereafter FTSEE), KOSPI500 (hereafter KOSPI), the CDS premium of Korean sovereign bond matured in 2025 (hereafter Korea bond CDS premium) and the Dubai oil spot price (hereafter Oil price) in the regression analysis. In addition, the volatility of each variable is added in the model for a better specification. Each variable is measured as moving averages of 20 trading days while the volatility of each variable is measured as moving standard deviations of 20 trading days. Looking at movements of volatilities in Fig. 2, we suspect a co-movement of volatilities and the common part of the CDS premium.
Acointegration test is conducted in order to check if variables are cointegrated. Level data would be used if cointegrated. Otherwise, the differenced data would be chosen in order to avoid a spurious regression problem. Johansen's Trace test is conducted with the lag length 2 and the result is presented in   Table 7. Regression results of determinants of the common component model with dummy Observing the common factor movement in Fig. 1, we suspect the existence of a structural break occurred inmid-2007. The structural break test suggested by Chow is conducted to detect structural breaks. As presented in Table 6, the null hypothesis of no structural break on August 1, 2007 is rejected at the Pvalue 0.00. That is, the CDS premium seems to jump due to the subprime crisis at that time. We add a dummy variable to Equation 4 to reflect the actual events which were related to the subprime mortgage crisis initiated in April 2007 and the following financial crisis in September 2008. In April 2007, New Century Financial filed for bankruptcy and triggered U.S. subprime mortgage crisis. Korea was affected when the American Home Mortgage Investment (AHMI) filed for bankruptcy protection in the court in August 2007.

AJEBA
Model I includes FTSED, FTSEE, KOSPI, Korea bond CDS premium, oil price and a dummy. All variables except the dummy are in logarithm because of its convenience for sensitivity analysis. To check if volatilities affect the common risk, Model II is estimated by adding the volatilities of each variable and also by dropping a statistically insignificant variable, FTSEE. Simple OLS method is used for estimation. As shown in Table 7, Model II offers a higher adjusted R2than Model I. Adding the volatilities and dropping FTSEE improve the explanatory power of the model. FTSED, KOSPI, Korea bond CDS premium, oil price and the volatility of FTSED appear to be statistically significant at 5%. The sign of the estimate of every statistically significant variable seems to be consistent with what we predicted. The estimate of the dummy variable is also statistically significant at 5% and positive.

DISCUSSION
This study examined what factors determine Korean banks' credit default swap premiums. As descrived in section 3, we first identify common factors in the CDS premium and further examine the determinants of the common factor employing an empiorical model shown in Equation 4. We discuss the major findings of the study as follows.
First, the estimates of FTSED and KOSPI were negative and statistically significant at 5%. It suggests that both foreign and domestic stock market movements had a negative impact on the CDS premium. When the developed foreign stock markets such as NYSE and EU sharply fell or collapsed, the foreign capital outflows for switching to a safer asset immediately exploded. The Korean banks were therefore faced with the shortage of liquidity and an increase in the default risk of loans. Accordingly, the CDS premiumsurged up. By the same token, when Korean stock markets fell sharply, exactly the same phenomena happened. Hence both movements of foreign and domestic stock markets negatively affect the CDS premium. The magnitude of responsiveness to KOSPI (-1.62) was slightly greater than that of the FTSED (-1.31).
Second, the volatility of FTSED and the Korean banks' common factor of the CDS premium turn out to be positively related. The common factor of the CDS premium jumps up as the foreign stock market gets more volatile. In addition, the common factor of the CDS premium appears to be the most responsive to the volatility of FTSED. The magnitude of the sensitivity to the volatility of FTSED is estimated at 4.78, which is approximately 3.5 times of FTSE developed markets and 2.8 times of KOSPI in terms of the absolute value. As the movement of FTSE developed markets became more unpredictable and riskier, it induced the foreign capital outflows to increase and caused the CDS premium to rise. However, the volatility of the Korean stock market had surprisingly no impact on the CDS premium.
Third, the CDS premium rises as the oil price increases. This implies that an increase in oil price tends to have negative impacts on the profits of Korean business firms by raising their production costs. The default risk of loans increases because of the weak profit structure and thus the CDS premium rises.
Fourth, as noticed, the structural break, which was caused by the subprime mortgage crisis, is incorporated into the model by adding a dummy variable. The estimate of the dummy variable turned out to be 0.41. That is, the subprime mortgage crisis period from August 1, 2007 to April 18, 2011 shifted the CDS premium up by 0.41. Consequently, the magnitude of the increase was considered to be an adjustment to the increased risk due to the subprime mortgage crisis and the following financial crisis.
Fifth, the estimate of the CDS premium of Korean sovereign bond variable appears to be positive and statistically significant at 5%. Since the Korean sovereign bond is issued by the Korean government in foreign currencies, the CDS premium is mainly affected by the country risk. As the country risk increases for various reasons, the CDS premium of bond rises and the CDS premium of banks increases accordingly.
Sixth, FTSEE is found to be statistically insignificant at 5%. That is, the movements of the CDS premium in the Korean market seem to be rather closely linked to the movements of the developed markets than on the emerging markets.
Lastly, the volatility of KOSPSI, the volatility of the CDS premium of Korea bond and the volatility of oil price appeared to be statistically insignificant at 5%. Only the volatility of FTSED is statistically significant at the conventional level. This implies that the Korean bank CDS premium is strongly affected by the volatility of the stock market in developed countries.

CONCLUSION
To find the determinants of the common factor of the CDS premium of Korean banks, we first decomposed the CDS premium of 11 Korean banks into common factors and idiosyncratic series by employing the method suggested by Bai and Ng (2004). We find that there is only one common factor deriving the CDS premium of Korean banks. Surprisingly, the most variation of each banks' CDS premium is explained by the common factor. It implies that the Korean banking industry confronts a substantial degree of systemic risk.

AJEBA
Next, we attempted to find the determinants of the common factor by regressing the common factor on macro-financial economic variables such as the daily stock composite index of foreign and domestic markets, Korean sovereign bond CDS premium, volatilities of each asset markets and the commodity prices. The regression results showed that the common factor was determined by the composite index of FTSE developed markets, KOSPI500, the Korea sovereign bond CDS premium, the Dubai spot oil price and the volatility of FTSE developed markets for the sample period. In particular, the common factor of the CDS premium appeared to be very sensitive to the FTSE level and its volatility. We also found a structure break for the CDS premium movement, which appeared to be affected by the subprime crisis in the United States since August 1, 2007.
These findings suggest that Korean banks are very susceptible to foreign capital movements, which are caused by changes in foreign economies. Not only is the Korean economy heavily dependent upon foreign economy through export, but also Korean financial markets are liberalized enough for foreign capital flows at foreign investors' convenience. In particular, an excessive amount of withdrawal of the foreign capital would induce a reduction in liquidity and a credit crunch. Accordingly, business firms' default rates of repayment rise and the CDS premium of the Korean banks increases.
The empirical findings suggest that the policy authority must pay heed to foreign stock markets to sustain the stability of banking industry. It is necessary to consider the stabilization of Korean financial asset markets, the maintenance of an appropriate level of foreign exchange reserves for emergencies and the expansion of foreign exchange swap agreements. In addition, financial supervision is needed to induce financial institutions to be less dependent on short-term financing to cushion against shocks resulting from exogenous capital outflows.