1: \documentclass{elsart}

2: 

3: \usepackage{amssymb,amsfonts,graphicx}

4: 

5: \begin{document}

6: 

7: \begin{frontmatter}

8: 

9: \title{Dynamics of the return distribution in the Korean financial market}

10: 

11: \author{Jae-Suk Yang\corauthref{cor1}},

12: \ead{yang@kaist.ac.kr}\corauth[cor1]{Corresponding author. Fax:

13: +82-42-869-2510.}\author{Seungbyung Chae, }

14: \author{Woo-Sung Jung, }

15: \author{Hie-Tae Moon}

16: 

17: \address{Department of Physics, Korea Advanced Institute of Science and Technology,

18: Daejeon 305-701, South Korea}

19: 

20: \begin{abstract}

21: 

22: In this paper, we studied the dynamics of the log-return

23: distribution of the Korean Composition Stock Price Index (KOSPI)

24: from 1992 to 2004. Based on the microscopic spin model, we

25: found that while the index during the late 1990s showed a power-law

26: distribution, the distribution in the early 2000s was exponential.

27: This change in distribution shape was caused by the

28: duration and velocity, among other parameters, of the information that

29: flowed into the market.

30: 

31: \end{abstract}

32: 

33: \begin{keyword}

34: Econophysics \sep Emerging market \sep Log return \sep Power law

35: distribution \sep Exponential distribution

36: % keywords here, in the form: keyword \sep keyword

37: \\

38: 

39: \PACS 89.65.Gh \sep 89.75.Fb \sep 89.75.Hc

40: % PACS codes here, in the form: \PACS code \sep code

41: \end{keyword}

42: 

43: \end{frontmatter}

44: 

45: % main text

46: 

47: \section{Introduction}

48: 

49: Interdisciplinary research is now routinely carried out, with

50: econophysics being one of the most active interdisciplinary fields

51: \cite{stanley,mantegna,mandelbrot2,jung2,jung3}. Many research

52: papers on mature markets have already been published. However,

53: since emerging markets show different characteristics to those of

54: mature markets, they represent an active field for

55: econophysicists. The Korean market, one of the foremost emerging

56: markets, has already been studied by physicists

57: \cite{jung2,jung3}. We concentrate on the particular properties of

58: the Korean market through the return distribution.

59: 

60: It is broadly assumed that the distribution of price changes takes the

61: form of a Gaussian distribution, and that all information is applied to

62: the market immediately by the efficient market hypothesis (EMH)

63: \cite{cootner}. Using the EMH, the trading profit with arbitrage cannot be

64: obtained from the superiority of information. The price changes in an

65: efficient market cannot be predicted and change randomly. This is suited to

66: classical economics theory. However, experimental proofs reveal that Gaussian

67: distributions of price changes do not exist in real markets \cite{mandelbrot1,fama}.

68: 

69: Mandelbrot determined empirically that the tail part of the

70: distribution is wider and the center of the distribution is

71: sharper and higher than a Gaussian distribution by examining price

72: changes of cotton; this distribution of price changes is termed

73: the L\'evy stable distribution \cite{mandelbrot1}. Fama also found

74: a L\'evy stable distribution for the New York Stock Exchange

75: (NYSE) \cite{fama}. After Mandelbrot's study, the distribution of

76: price changes was identified as non-Gaussian by many researchers

77: \cite{stanley2,mccauley,silva,vicente}. It was reported that

78: distributions in mature markets have a power-law tail, while those

79: in emerging markets have an exponential tail \cite{matal,miranda}.

80: 

81: Silva \emph{et al.} \cite{silva} and Vicente \emph{et al.}

82: \cite{vicente} reported that the distributions of price changes

83: vary with time lag. Price changes have a power-law distribution

84: for a short time lag, and an exponential distribution when the

85: time lag is long. Moreover, for a very long time lag, the

86: distribution becomes Gaussian. This transition problem was solved

87: analytically by Heston using a stochastic model \cite{heston}.

88: 

89: Financial markets are adaptive evolving systems, so the

90: distributions of price changes are also different for various

91: periods and countries. Especially, the Korean stock market has

92: different properties compared with other countries, and the

93: distribution of price changes is not stable and changes with time.

94: 

95: In this paper, we study the characteristics of the Korean stock

96: market using probability distribution functions (PDFs) of the

97: Korean Composition Stock Price Index (KOSPI) and investigate why

98: phenomena different to other countries occur. We also carry out a

99: simulation using the microscopic spin model to explain these

100: differences.

101: 

102: \section{Empirical data and analysis}

103: 

104: We use the KOSPI data for the period from 1992 to 2004, and

105: observe the PDFs of the KOSPI price changes for a time window

106: of 1 year. We use only intra-day returns to exclude discontinuous

107: jumps between the previous day's close and the next day's open

108: price due to overnight effects. The price change log return

109: is defined by:

110: 

111: \begin{equation}

112: S(t)\equiv \ln Y(t+\Delta t)-\ln Y(t),

113: \end{equation}

114: 

115: where $Y(t)$ is the price at time $t$ and $\Delta$$t$ is the time lag.

116: 

117: Fig. \ref{Fig1}a shows the log return distribution for the KOSPI.

118: The distribution for an 1-min time lag represents a power-law

119: distribution, that for 10 min is exponential, and that for 30 min

120: is also exponential but close to Gaussian. These results are in

121: accordance with previous reports \cite{silva,vicente}.

122: 

123: In the Korean stock market, the log return distribution for an

124: 1-min time lag shows some peculiar phenomena. Fig. \ref{Fig1}b

125: shows the PDFs of the KOSPI log return from 1998 to 2002, while

126: Fig. \ref{Fig1}c is a graph of the tail index, power law exponent

127: of tail part of the log return distribution, as a function of time

128: from 1993 to 2004. The shape of the distribution in 1998

129: ($\Diamond$) in Fig. \ref{Fig1}b is close to a L\'evy distribution

130: and the tail part shows a power-law distribution. However, the

131: tail index of the PDFs increases over the years and the shape of

132: tail part changes to exponential with increasing time. This

133: phenomenon can be confirmed in Fig. \ref{Fig1}c. As well the tail

134: index in the early 1990s is approximately 2.0, increases from the

135: mid-1990s to the early 2000s, finally changes to an exponential

136: tail. Although a discontinuity in the increasing trend occurred

137: during the 1997 Asian financial crisis, the tail index continued

138: to increase thereafter.

139: 

140: In Fig. \ref{Fig2}a, the decay time for the autocorrelation

141: function of log return is continuously decreasing, while the tail

142: index abruptly varied around the time of the 1997 Asian financial

143: crisis. This suggests that the decay time is related to the

144: increasing trend of the tail index, regardless of the Asian

145: financial crisis. Thus, we investigated the relation between the

146: decay time for the autocorrelation function of log return and the

147: tail index of the log return distribution to identify why this

148: phenomenon is happened. Fig. \ref{Fig2}b shows the relation

149: between the tail index and the decay time. The decay time of the

150: autocorrelation function is inversely proportional to the tail

151: index by a factor of 4.

152: 

153: The decay time of the autocorrelation function decreases as the

154: log return distribution of the KOSPI changes from a power-law to

155: an exponential distribution (Fig. \ref{Fig2}). This decrease in

156: decay time means that the duration of information in the market is

157: diminished compared with the past, that is, the market is more

158: rapidly affected by the information flow and the influence of past

159: information decreases more rapidly when the decay time is longer.

160: Information and communication technology such as high-speed

161: internet connections and electronic trading systems were not fully

162: utilized in the past, so it took a longer time to deliver

163: information to the market. Moreover, the information flow was

164: small because social structures in the past were relatively

165: simple. For this reason, much more information flows into the

166: market for a specific time interval now compared with the past.

167: Therefore, the time scale of the past differs from the current

168: scale. The amount of information and its velocity in reaching the

169: market for 1 min now may be the same as that for 2 or 3 min or

170: more in the past.

171: 

172: \section{Model and results}

173: Eguiluz and Zimmermann \cite{eguiluz}, Krawiecki \emph{et al.}

174: \cite{krawiecki}, Chowdhury and Stauffer \cite{chowdhury}, and

175: Cont and Bouchaud \cite{cont} used microscopic models to simulate

176: the financial market, and these models describe well the

177: characteristics of the financial market. Krawiecki \emph{et al.}

178: described the market as a spin model. Agents and traders

179: comprising the market are represented by spins, and interaction

180: between agents and external information is represented by fields.

181: 

182: We modify the Krawiecki microscopic model of many interacting

183: agents to simulate the variation of log return distribution for

184: the Korean stock market. The number of agents is $N$, and we

185: consider $i=1,2,\dots ,N$ agents with orientations $\sigma_{i} (t)

186: = \pm 1$, corresponding to the decision to sell ($-1$) or buy

187: ($+1$) stock at discrete time-steps $t$. The orientation of agent

188: $i$ at the next step, $\sigma_{i}(t+1)$, depends on the local

189: field:

190: 

191: \begin{equation}

192: I_{i} (t) = \frac{1}{N} \sum _{j} A_{ij} (t) \sigma_{j} (t) +

193: h_{i} (t),

194: \end{equation}

195: 

196: where $A_{ij} (t)$ represent the time-dependent interaction strength among

197: agents, and $h_{i} (t)$ is an external field reflecting the effect

198: of the environment. The time-dependent interaction strength among

199: agents is:

200: 

201: \begin{equation}

202: A_{ij} (t) = A\xi (t) + a \eta_{ij} (t),

203: \end{equation}

204: 

205: with $\xi(t)$ and $\eta_{ij} (t)$ determined randomly in every

206: step. $A$ is an average interaction strength and $a$ is a

207: deviation of the individual interaction strength. The external

208: field reflecting the effect of the environment is:

209: 

210: \begin{equation}

211: h_{i}=h \sum _{k=0}^{\infty} \zeta_{i} (t-k) e^{-k/\tau},

212: \end{equation}

213: 

214: where $h$ is an information diffusion factor, and $\zeta_{i}(t)$

215: is an event happening at time $t$ and influencing the $i$-th

216: agent. $\tau$ is the duration time of the information, which

217: represents how long the event at time $t$ retains influence on the

218: opinion of agents on market prices. At every step, agents are

219: assumed to receive newly generated information. The later the

220: event, the greater is the influence on agents and the market. The

221: most recent event has a strong influence on agents, while old

222: events have a weak influence. Influence on the market is

223: considered to be reduced exponentially. Moreover, for larger

224: $\tau$, an event at time $t$ retains a relatively strong influence

225: on agent opinions and market price changes for a long time. On the

226: other hand, for shorter $\tau$, the information from an event is

227: rapidly delivered to agents and applied to market prices

228: immediately, so the information quickly vanishes from the market

229: (Fig \ref{Fig3}). If we assume that the information flow is the

230: same ($=1$) whether $\tau$ is long or short, then $h$ is equal to

231: 1/$\tau$.

232: 

233: From the local field determined as above, agent opinions in the

234: next step are determined by:

235: 

236: \begin{equation}

237: \sigma_{i} (t+1) =\left\{ \begin{array}{ll} +1 & \mbox{with probability

238: $p$}

239: \\ -1 & \mbox{with probability $1-p$} \end{array} \right..

240: \end{equation}

241: 

242: where $p=1/(1+exp\{-2I_{i}(t)\})$. In this model, price changes are:

243: 

244: \begin{equation}

245: x(t) = \frac{1}{N} \sum \sigma_{i}(t).

246: \end{equation}

247: 

248: We simulate a stock market with 1000 agents, and the values

249: $\xi(t)$, $\eta(t)$, and $\zeta(t)$ are generated randomly within

250: the range $[-1, 1]$.

251: 

252: Fig. \ref{Fig4}a shows the relation between the diffusion factor,

253: $h$, and the tail index of the PDFs. The tail index is directly

254: proportional to $h$. This is similar to the trend line (solid

255: line) in Fig. \ref{Fig1}c. Fig. \ref{Fig4}b and c show the PDFs of

256: log return for various $h$ values. Similar to the experimental

257: result, a power-law tail is evident for small $h$ (large $\tau$),

258: and an exponential tail for large $h$ (small $\tau$). For very

259: large $h$, the distribution is close to Gaussian (Fig.

260: \ref{Fig4}d)

261: 

262: The time lag, $\Delta t$, can be determined for 1 min, 10 min, 1

263: h, 1 day, etc., when price changes are calculated. However, time

264: is not uniform, because the volume, the number of contracts, and

265: the information flow into the market are not homogeneous. In other

266: words, 1 min of today may not be same as 1 min of tomorrow. Thus,

267: it is necessary to study the time uniformity.

268: 

269: \section{Conclusions}

270: The distribution of the KOSPI log return has recently taken the form of an

271: exponential, while it showed a power-law tail in the 1990s. Moreover, the

272: decay time of the autocorrelation function is continually decreasing.

273: Thus, the duration of information received by agents is also decreasing as

274: the amount of information increases. According to the EMH, the distribution

275: of log return becomes Gaussian when the velocity of

276: information flow is very fast, and all information received immediately

277: affects the opinion of agents in the market.

278: We could identify and confirm a relationship between the distribution

279: of price changes, the velocity of information flow, and the duration of

280: the influence of information for a time series of the Korean

281: stock market. A similar phenomenon occurred in Japan around

282: 1990, as identified from daily data \cite{kaizoji}. However, in mature markets,

283: including the NYSE, the tail index is not increasing or changing in shape.

284: The reason for this is the robustness and maturity of the

285: market. Mature markets are solid enough to endure external shocks,

286: while emerging markets are susceptible to shocks and environmental

287: changes. Modeling of the robustness and maturity of the market is

288: planned as further work in the near future.

289: 

290: \begin{ack}

291: We wish to thank H. Jeong and O. Kwon for helpful discussions and

292: supports.

293: \end{ack}

294: 

295: \begin{thebibliography}{00}

296: 

297: \bibitem{stanley} H.E. Stanley, L.A.N. Amaral, D. Canning, P. Gopikrishnan,

298: Y. Lee, Y. Liu, Physica A 269 (1999) 156.

299: 

300: \bibitem{mantegna} R.N. Mantegna, H.E. Stanley, An Introduction to Econophysics:

301: Correlations and Complexity in Finance, Cambridge University

302: Press, 2000.

303: 

304: \bibitem{mandelbrot2} B.B. Mandelbrot, Fractals and Scaling in Finance:

305: Discontinuity, Concentration, Risk, Springer, New York, 1997.

306: 

307: \bibitem{jung2} W.-S. Jung, S. Chae, J.-S. Yang, H.-T. Moon, Physica A, in press (Physics/0504009, 2006).

308: 

309: \bibitem{jung3} W.-S. Jung, O. Kwon, J.-S. Yang, H.-T. Moon, J. Korean Phys. Soc.,

310: in press (Physics/0509090, 2006).

311: 

312: \bibitem{cootner} P.H. Cootner, Ed.,

313: The Random Character of Stock Market Prices, MIT Press, Cambridge

314: MA, 1964.

315: 

316: \bibitem{mandelbrot1} B.B. Mandelbrot, J. Bus. 36 (1963) 394.

317: 

318: \bibitem{fama} E.F. Fama, J. Bus. 38 (1965) 34.

319: 

320: \bibitem{stanley2} H.E. Stanley, L.A.N. Amaral, X. Gabaix, P.

321: Gopikrishnan, V. Plerou, Physica A 299 (2001) 1.

322: 

323: \bibitem{mccauley} J.L. McCauley, G.H. Gunaratne, Physica A 329 (2003) 178.

324: 

325: \bibitem{silva} A.C. Silva, R.E. Prange, V.M. Yakovenko, Physica A 344 (2004) 227.

326: 

327: \bibitem{vicente} R. Vicente, C.M. de Toledo, V.B.P. Leite, N. Caticha, Cond-mat/0402185, 2004.

328: 

329: \bibitem{matal} K. Matal, M. Pal, H. Salunkay, H.E. Stanley,

330: Europhys. Lett. 66 (2004) 909.

331: 

332: \bibitem{miranda} L.C. Miranda, R. Riera, Physica A 297 (2001) 509.

333: 

334: \bibitem{heston} S.L. Heston, Rev. Financ. Stud. 6 (1993) 327.

335: 

336: \bibitem{eguiluz} V.M. Eguiluz, M. Zimmermann, Phys. Rev. Lett. 85 (2000) 5659.

337: 

338: \bibitem{krawiecki} A. Krawiecki, J.A. Ho\l yst, D. Helbing, Phys. Rev. Lett. 89 (2002) 158701.

339: 

340: \bibitem{chowdhury} D. Chowdhruy, D. Stauffer, Eur. Phys. J. B 8 (1999) 477.

341: 

342: \bibitem{cont} R. Cont, J.-P. Bouchaud, Macroecon. Dyn. 4 (2000) 170.

343: 

344: \bibitem{kaizoji} T. Kaizoji, Physica A 343 (2004) 662.

345: 

346: \end{thebibliography}

347: 

348: \newpage

349: 

350: \begin{figure}

351: \includegraphics[angle=0,width=0.5\textwidth]{Fig1}

352: \caption{(a) Probability distribution functions (PDFs) of log

353: return for the KOSPI. Rectangle($\blacksquare$) represents 1-min

354: time lag, circle($\bullet$) 10-min, and triangle($\blacktriangle$)

355: 30-min, respectively. (b) PDFs of log return for the KOSPI from

356: 1998 to 2002: $\Diamond$ 1998, $\bigtriangledown$ 1999,

357: $\bigtriangleup$ 2000, $\bigcirc$ 2001, and $\Box$ 2002. (c)

358: Evolution of the KOSPI tail index. The tail first increases and

359: then changes to exponential. The dashed line represents the 1997

360: Asian financial crisis.} \label{Fig1}

361: \end{figure}

362: 

363: \begin{figure}

364: \includegraphics[angle=0,width=1.0\textwidth]{Fig2}

365: \caption{(a) Evolution of the decay time for the autocorrelation function (ACF)

366: and the tail index. (b) Relation between decay time and tail index for the ACF.

367: } \label{Fig2}

368: \end{figure}

369: 

370: \begin{figure}

371: \includegraphics[angle=0,width=1.0\textwidth]{Fig3}

372: \caption{Influence of information flow into the market. The

373: latest information has a strong influence, and influence on the

374: market diminishes as time passes. (a) If the information flow is small,

375: the influence of the information slowly decreases. (b) If the information flow

376: is large, the influence of the information rapidly decreases. The

377: amount of information used by traders to determine whether to sell or buy is

378: constant. Thus, the area of these functions has to be constant.}

379: \label{Fig3}

380: \end{figure}

381: 

382: \begin{figure}

383: \includegraphics[angle=0,width=1.0\textwidth]{Fig4}

384: \caption{(a) Relation between $h(=1/\tau)$ and the tail index of

385: PDFs. (b) Semi-log plot for the PDFs of price changes as a

386: function of $h$. $\Box,\ h=0$; $\bigcirc,\ h=1$; $\bigtriangleup,\

387: h=2$; and $\bigtriangledown,\ h=3$. (c) Log-log plot for the PDFs.

388: The straight line is guide line for power law distribution and the

389: curved line is for exponential distribution. Symbols are as for

390: (b). (d) PDF of price changes when $h$ is very large ($h=10$).}

391: \label{Fig4}

392: \end{figure}

393: 

394: \end{document}

395: