Analysis on Intraday Volatility from an Econophysics Perspective: Characteristics and Determinants

Abstract

This dissertation examines intraday volatility as well as related issues using econophysical methodologies. The financial landscape has been dramatically changing during the past decades, i.e., deregulation of markets and growing complexity of products. In a technical point of view, the ever increasing speed of networks and computational power paired with their decreasing costs have led to the emergence of huge databases, which record order book variations up to the last millisecond in all transactions. Volatility is the key variable when pricing various derivative securities, whose trading volume have surged in recent years. To price an option, the volatility of the underlying asset must first be known. Volatility itself becomes the underlying asset. Hence, determining volatility is becoming increasingly important. Previous literature on volatility has used the classical economics approach. They have concentrated on finding a hypothetical and stable model and on fitting data into this model, widely categorized in the regression method. These regression methods are based on the assumption of stationary noise. However, no solid empirical evidence for the stationarity of any product and any market is available. Hence, this lack of stationary evidences reinforces the importance of intraday volatility and a relatively new methodology, i.e., econophysics. The main objective of the present dissertation is to understand the issues related to intraday volatility and econophysical methodologies. Two aspects of intraday volatility are studied through two modules. The first module focuses on understanding the empirical characteristics of the Korean financial market in a high-frequency level. Two individual studies are conducted for the first module, a microscopic observation of one asset and a market-wide observation regarding market volatility. The second module explores the determinants of intraday volatility and related issues. Only one study is conducted for the second module. Hence, this work consists of a total of three individual studies. Chapter 3 presents the microscopic observation by examining the characteristics of the order book and the market impact from the Korean stock index futures market (KOSPI 200 index futures). The distribution of the three types of order (market, limit, cancel) generally follows power-law distribution. The exponents are estimated as 1.9 for the market order, 2.5 for the limit order, and 2.1 for the cancel order. These results are different from stocks cases in previous studies where the exponent of the limit order is larger than that of the market order. In the distribution order size, regular peaks are observed in order volume intervals of 50, which reflect the behavioral characteristics of human preference in rounded numbers. The distribution of the bid-ask spread, the distribution of the best quotes volume, and the percentile of the best quotes size and the market order volume provide evidence for the liquidity of the KOSPI 200 index futures market. The market impact function, market response function, and market response time in various transaction sizes have been observed. The market impact function shows a concave pattern to the transaction size. The market response time is longer, and the amplitude of response becomes larger as the order transaction size increases. For the market-wide observation, the relationship between the market volatility and the network properties of MST in the intraday high frequency level dynamics is investigated using the intraday Korean stock market data and is discussed in Chapter 4. The network structure becomes denser as the market volatility moves higher. The NTL negatively responds with the market volatility so that the trend of the NTL series is a reverse image of the image of the market volatility series. The MOL tends to decrease as the market volatility increases, and k_max shows the tendency of increment for higher market volatility. All these network properties consistently support the notion that the characteristics observed from low frequency dynamics (daily or longer) are also found in intraday high frequency dynamics. In other words, as the market becomes volatile, the co-movement tendency of financial assets is also reinforced in an intraday high frequency level as in a daily or longer low frequency level. In Chapter 5, the relationship between the behavioral characteristics of market participants and the volatility patterns from the heterogeneous agent-based simulation is obtained for the investigation on the determinants of intraday volatility. The simulated return distribution generally follows the power-law distribution, and the volatility is positively correlated with the parameter participating rate α, information flow structure λ, and proportion of chartist σ_2. Hence, the large values of these parameters tend to induce large volatility. Autocorrelation is rarely observed and is in turn, mostly observed in volatility. Three analyses in the order book level have been conducted. The inter-arrival time of orders follows the power-law distribution with the slope of around 2.3 with regard to α and about 2.4 with respect to λ. However, with regard to σ_2, the power-law distribution is not observed. The distribution of the order size generally follows the power-law regarding λ (fixed α, σ_2) but two other parameters do not fit with the power-law. The distributions in average order ages listed in the order book display distinctive shapes. Some appear like Gaussian distributions and others have double peaked distributions. The double peak-shaped distributions are observed in a large value of α, λ, and σ_2. These large parameter values make the market more volatile, so the listed orders are swept away by large orders; or continuous one-way orders more frequently happen. This sweeping case makes a peak around age zero. The simulation results are generally consistent with previous studies. However, literature on order book characteristics are insufficient. Hence, no substantial comparison could be made. Nonetheless, most of the results are in line with available empirical observations. In conclusion, the present dissertation analyzes the microscopic phenomena and market-wide characteristics of intraday volatility through empirical data research. Moreover, the determinants that influence intraday volatility have been investigated by the agent-based simulation. Through these studies, the present dissertation extends economic understanding on intraday volatility, the importance of which has been gradually increasing in the present financial environment.