earticle

영어

This paper examines the relationship between BSI volatility and KOSPI volatility. BSI indicates economic conditions and trends for the near future. Business survey information reveals expectations of economic agents. Stock price is the present value of expected future cash flow stream. Because expected future cash flows are influenced by expectations of economic agents, BSI and KOSPI may be related each other. Prior studies investigate the relationship between the first moment of BSI and the first moment of KOSPI. This study is distinguished from prior studies in that it investigates the relationship between the second moment of BSI and the second moment of KOSPI. And another contribution is investigating the relationship between BSI permanent volatility and KOSPI permanent volatility and the relationship between BSI transitory volatility and KOSPI transitory volatility. The summary statistics show the standard deviation of BSI is higher than that of KOSPI. Excess kurtosis is observed in each series, and the Jarque-Bera normality test reveals non-normalities of two series. Therefore, GARCH type model is appropriate for analyzing these data. The procedure used in this paper involves the following steps. First, in order to estimate BSI observed volatility and KOSPI observed volatility, EGARCH model is employed. Because the autocorrelation coefficient of data is not significantly different from zero, the mean equation in EGARCH model does not contain AR(1) term or MA(1) term in each model. In estimated EGARCH model, it turns out that the coefficient of the asymmetric volatility term is statistically significant. In estimated EGARCH model, (alpha) coefficient in the variance equation is smaller than one. Therefore, each estimated conditional variance series is stationary. Second, to decompose observed volatility into permanent volatility and transitory volatility, this study employs state space model. State space model is useful in analyzing a time series model that involves unobservable variables. A state space model consists of two equations, namely, a measurement equation and a transition equation. Measurement equation describes the relation between observed variables and unobserved variables. Transition equation describes the behavior of unobserved variables. In this study, observed volatility calculated by EGARCH model is observed variable, and permanent volatility and transitory volatility are unobservable variables. Permanent volatility is the long-run component and is modeled as a random walk with drift. Transitory volatility is the short-run component and is modeled as a stationary process because it passes away with time. Third, to test for the existence of unit root, ADF test is used. The appropriate lag order is determined by the SIC. ADF test results show that observed volatility and transitory volatility are stationary series, and permanent volatility is nonstationary series. Therefore, in the cases of observed volatility and transitory volatility, raw data is used in this study. Because BSI permanent volatility and KOSPI permanent volatility are found to be unit root processes, the next step is to check the cointegration between BSI permanent volatility and KOSPI permanent volatility. I use Johansen cointegration tests to examine the existence of long-run relationship. If the series are found to be cointegrated, Granger causality in a VECM should be applied. If a time-series of system includes integrated variables of order one and cointegrating relations, then this system can be more appropriately specified as a VECM rather than VAR. But, if there is no cointegrating relation between variables, they are to be transformed to stationary process by differencing the series. Because Johansen cointegration tests show BSI permanent volatility and KOSPI permanent volatility are not cointegrated, they are to be transformed to stationary process by differencing the series. The null hypotheses of a unit root are rejected at 1% significance level when they are in first differences. Fourth, to examine the lead-lag relationship between BSI volatility and KOSPI volatility over the data from January of 1990 to June of 2008, this study employs the Granger causality test. Granger causality test is a useful statistical technique to help examine the direction of causality. Each pair of volatilities has different characteristics. In the cases of observed volatility and transitory volatility data, they are found to be stationary series. Because there is not a unit root, Granger causality test based on a VAR is applied. On the other hand, in the case of permanent volatility, it is found to be nonstationary and they are not cointegrated, Granger causality test based on a VAR is applied using differenced data. The Granger causality tests show that, in all cases of volatilities, there are two-way Granger causalities between BSI and KOSPI, in line with general belief. Permanent volatility is the trend component of observed volatility. As expected, there are two way Granger causalities between BSI permanent volatility and KOSPI permanent volatility. Transitory volatility is the cyclical component of observed volatility, that is, it is the short-run portion and passes away with time. The empirical test result shows that there are two way Granger causalities between BSI transitory volatility and KOSPI transitory volatility. This means that the cyclical portion of KOSPI volatility is influenced by the cyclical portion of BSI volatility. Fifth, the impulse response function analysis is applied to trace out the time path of the various shocks on the variables. The results exhibit that BSI volatility shock increases the magnitude of KOSPI volatility in general. KOSPI volatility shock also increases the magnitude of BSI volatility from lag 1 to lag 6 in all cases of volatilities. In short, the empirical evidence of this study suggests that BSI volatilities help to estimate KOSPI volatilities. Therefore, analyzing the BSI volatilities is crucial for risk management of KOSPI related financial instruments.

한국어

본 연구에서는 기업 경영자들의 기대에 있어서의 불안정성을 나타내는 BSI(Business Survey Index) 변동성과 주식시장의 불안정성을 나타내는 KOSPI 변동성의 관계를 분석하였다. 주가는 궁극적으로는 경제주체들의 기대에 의해 형성되기 때문에, 대표적인 경제주체라고 할 수 있는 경영자들의 기대를 나타내는 변수와 주가의 관계를 연구하는 것은 투자전략 수립에 있어 의미 있는 연구 분야가 될 것으로 생각된다. 본 연구의 특징은 관측변동성을 상태공간모형을 이용하여 추세 부분에 해당되는 영속적 변동성 부분과 순환적 부분에 해당되는 일시적 변동성 부분으로 분해하여 BSI 변동성과 KOSPI 변동성 간의 관계에 대하여 분석하였다는 것이다. 본 연구의 분석은 첫 번째 단계로, EGARCH모형을 이용하여 BSI 관측변동성과 KOSPI 관측변동성을 계산하였다. 두 번째 단계로, 상태공간모형과 칼만필터링을 이용하여 BSI와 KOSPI 각각의 영속적 변동성과 일시적 변동성을 추정하였다. 세 번째 단계로, 단위근검정을 거친 다음 각 변동성 간의 관계를 분석하기 위하여 그랜저 인과관계 분석을 하였다. 네 번째 단계로, 충격에 대한 각 변수들의 시간에 따른 반응결과를 분석하기 위하여 충격반응분석을 하였다. 그랜저 인과관계 분석에서는 관측변동성, 영속적 변동성, 일시적 변동성 모두에 있어서 BSI와 KOSPI 간에 양방향 그랜저 인과관계를 갖는 것으로 나타났으며 충격반응분석에서는 일반적으로 BSI 변동성과 KOSPI 변동성 간에 서로 정(正)의 관계가 있는 것으로 나타났다.