금융시장의 비선형성 및 복잡성 연구

Abstract

We studied the long-term memory effects of the Korean agricultural market using the Detrended Fluctuation Analysis (DFA) method. In general, the return time series of various financial data, including stock indices, foreign exchange rates and commodity prices, are uncorrelated in time, while the volatility time series are strongly correlated. However, we found that the return time series of the Koreanagricultural commodity prices are anti-correlated in time, while the volatility time series are correlated. The n-point correlations of time series were also examined, and it was found that a multifractal structure exists in the Korean agricultural market prices. We studied the market efficiency of Korean agricultural market using the approximate entropy (ApEn). The products traded in Korean agricultural market may show diverse features according to their intrinsic properties. In order to observe the efficiency of the Korean agricultural products created according to a place-oforigin, we employ the approximate entropy and calculate the ApEn value for seven agricultural products. We found that ApEn values of the agricultural products in domestic are significantly higher than those for imported agricultural products. We investigate the characteristics of the Korean fund returns. There are several types of funds, we examined stock funds and balanced funds. Cummulative distribution of Korean fund returns are well fitted to power-law distribution function. All returns were uncorrelated, while all volatilities were strongly correlated similar to several previous works for financial time series. There exist Multifractality in every Korean fund returns and the width of multifractal spectrum of shuffled data and surrogatedata is narrower than original one. This result can be understood as nonlinearity, and temporal correlation and power-law distribution is suspected as the origin of multifractality. We investigate an agent based model of financial markets that is made up of numerous interacting agents with respective to local and global coupling and intrinsic randomness. The prices or return time series are constructed by the essential mechanism of the interactions between agents that is determined according to local and global coupling strength. The characteristics that can describe a financial time series are the fat-tail distribution in the return time series and the long-term memory and clustering behavior of volatility data. We find that for the proper parameter values, the probability distribution of return time series follow power-law distribution with various scaling exponents according to the control parameter set. In particular, the local coupling strength considered as the modularity in financial market play an important role in terms of the price formation.