1: \documentclass{elsart}
2: %\documentclass[twocolumn,showpacs,preprintnumbers,superscriptaddress]{revtex4}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: \usepackage{amssymb,amsfonts,graphicx}
5: \usepackage{rotating}
6: \usepackage{subfigure}
7: \usepackage{dcolumn}
8: \usepackage{bm}
9: \usepackage{epsfig}
10:
11:
12: \begin{document}
13:
14:
15: \begin{frontmatter}
16:
17: \title{Information Flow between Composite Stock Index and Individual Stocks}
18:
19: \author[korea1,korea2]{Okyu Kwon},
20: \author[korea1]{Jae-Suk Yang\corauthref{cor1}}
21: \ead{mathphy@korea.ac.kr}
22: \corauth[cor1]{Corresponding author.}
23: \address[korea1]{Department of Physics, Korea University, Seoul 131-701, Korea}
24: \address[korea2]{National Creative Research Initiative Center for
25: Neuro-dynamics, Korea University, Seoul 136-701, Korea}
26:
27:
28:
29:
30: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
31: \begin{abstract}
32:
33: We investigate the strength and the direction of information
34: transfer in the U.S. stock market between the composite stock
35: price index of stock market and prices of individual stocks using
36: the transfer entropy. Through the directionality of the
37: information transfer, we find that individual stocks are
38: influenced by the index of the market.
39: %, and also they are clustered to the industrial sectors.
40:
41: \end{abstract}
42: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43:
44: \begin{keyword}
45: transfer entropy \sep information flow \sep econophysics \sep
46: stock market
47: % keywords here, in the form: keyword \sep keyword
48: \\
49: \PACS 05.45.Tp \sep 89.65.Gh \sep 89.70.+c
50: % PACS codes here, in the form: \PACS code \sep code
51: \end{keyword}
52:
53: \end{frontmatter}
54:
55: % main text
56:
57: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
58: \section{Introduction}
59: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
60:
61: Recently, economy has become an active research area for
62: physicists. Physicists have attempted to apply the concepts and
63: methods of statistical physics, such as the correlation function,
64: multifractal, spin models, complex networks, and information
65: theory to study economic problems
66: \cite{arthur,mategna,bouchaud,mandel,kullmann,giada,eguiluz,krawiecki,takaishi,kaizoji02,jbpark,jwlee,wsjung1,wsjung2,matal,yang,schae,jsyang2}.
67:
68: From the economic system, many empirical data reflecting the
69: economic conditions can be obtained. Among them the time series of
70: composite stock price index is one of the best data reflecting
71: economics conditions well. The index data is used to analyze and
72: predict the perspective of markets. The scientific interest in
73: studying financial markets stems from the fact that there is a
74: large amount of reasonably well defined data.
75:
76: Information is an important keyword in analyzing the market or in
77: estimating the stock price of a given company. It is quantified in
78: rigorous mathematical terms \cite{shannon}, and the mutual
79: information, for example, appears as meaningful choice replacing a
80: simple linear correlation even though it still does not specify
81: the direction. The directionality, however, is required to
82: discriminate the more influential one between correlated
83: participants, and can be detected by the transfer entropy
84: \cite{schreiber}.
85:
86: In many case, traders in the stock market refer to the index to
87: invest in stocks. Therefore, we can guess that prices of stocks is
88: affected by the composite stock index of the market. However, No
89: attempt to measure the influence of index quantitatively has been
90: accomplished, while it is found evident that the interaction
91: therein is highly nonlinear, unstable, and long-ranged from many
92: previous research on econophysics using financial time series.
93: Schreiber \cite{schreiber} introduced the transfer entropy which
94: measures dependency in time between two variables. We focus
95: quantitatively on the direction of information flow between the
96: index data and the price of individual companies using the method
97: of the transfer entropy. This concept of the transfer entropy has
98: been already applied to the analysis of financial time series by
99: Marschinski and Kantz \cite{marschinski}. They calculated the
100: information flow between the Dow Jones and DAX stock indexes and
101: obtained conclusions consistent with empirical observations. While
102: they examined interactions between two huge markets, we construct
103: its internal structure between stock index and individual stocks.
104:
105: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
106: \section{Transfer entropy}
107: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
108:
109: The transfer entropy which measures directionality of variable
110: with respect to time has been recently introduced by Schreiber
111: \cite{schreiber} based on the probability density function (PDF).
112: Let us consider two discrete and stationary process, $I$ and $J$.
113: The transfer entropy relates $k$ previous samples of process $I$
114: and $l$ previous samples
115: of process $J$ is defined as follows:
116: \begin{equation}
117: T_{J \rightarrow I} = \sum p(i_{t+1}, i_t^{(k)}, j_t^{(l)}) \log
118: \frac{p( i_{t+1} \mid i_t^{(k)}, j_t^{(l)})}{p( i_{t+1} \mid
119: i_t^{(k)})},
120: \end{equation}
121: where $i_t$ and $j_t$ represent the discrete states at time $t$ of
122: $I$ and $J$, respectively. $i_t^{(k)}$ and $j_t^{(l)}$ denotes $k$
123: and $l$ dimensional delay vectors of two time consequences $I$ and
124: $J$, respectively. The joint PDF $p(i_{t+1}, i_t^{(k)},
125: j_T^{(l)})$ is the probability that the combination of $i_{t+1}$,
126: $i_t^{(k)}$ and $j_t^{(l)}$ have particular values. The
127: conditional PDF $p(i_{t+1} \mid i_t^{(k)}, j_t^{(l)})$ and $p(
128: i_{t+1} \mid i_t^{(k)})$ are the probability that $i_{t+1}$ has a
129: particular value when the value of previous samples $i_t^{(k)}$
130: and $j_t^{(l)}$ are known and $i_t^{(k)}$ are known, respectively.
131:
132: The transfer entropy with index $J \rightarrow I$ measures how
133: much the dynamics of process $J$ influences the transition
134: probabilities of another process $I$. The reverse dependency is
135: calculated by exchanging $i$ and $j$ of the joint and conditional
136: PDFs. The transfer entropy is explicitly asymmetric under the
137: exchange of $i_t$ and $j_t$. It can thus give the information
138: about the direction of interaction between two time series.
139:
140: The transfer entropy is quantified by information flow from $J$ to
141: $I$. The transfer entropy can be calculated by subtracting the
142: information obtained from the last observation of $I$ only from
143: the information about the latest observation $I$ obtained from the
144: last joint observation of $I$ and $J$. This is the main concept of
145: the transfer entropy. Therefore, the transfer entropy can be
146: rephrased as
147: \begin{equation}
148: T_{J \rightarrow I} = h_I(k) - h_{IJ}(k,l),\label{eq:TE}
149: \end{equation}
150: where
151: \begin{eqnarray}
152: h_I(k) & = & -\sum p(i_{t+1}, i_t^{(k)}) \log p( i_{t+1} \mid i_t^{(k)}) \\
153: h_{IJ}(k,l) & = & -\sum p(i_{t+1}, i_t^{(k)}, j_t^{(l)}) \log p(
154: i_{t+1} \mid i_t^{(k)}, j_t^{(l)}).
155: \end{eqnarray}
156:
157: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
158: \section{Empirical data analysis}
159: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
160:
161: We analyze daily records of the S\&P 500 index (GSPC), Dow Jones
162: index (DJI) and stock price of selected 125 individual companies.
163: The dataset consists of about 4,000 simultaneously recorded data
164: points during the period June 1, 1983 to May 31, 2007. We use
165: logarithmic price difference as follows:
166: \begin{equation}
167: x_n \equiv \ln(S_n) - \ln(S_{n-1}),
168: \end{equation}
169: where $S_n$ means index or stock price of {\it n}-th trading day.
170: The first step in analysis for the transfer entropy is to
171: discretize the time series by some coarse graining. Quite often,
172: statistical studies which use the entropy assume that the
173: variables of interest are discrete, or may be discretized in some
174: straightforward manner. We partitioned the real value $x_n$ into
175: discretized price change $A_n$. In the concrete, $A_n=0$ for $x_n
176: \leq -d/2 $ (decrease), $A_n=1$ for $-d/2 < x_n < d/2$
177: (intermediate), $A_n=2$ for $x_n \geq d/2$ (increase) are chosen.
178:
179: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
180: \begin{figure}[tbph]
181: \begin{center}
182: \mbox{
183: {\scalebox{0.30}
184: {\hspace{0mm}\epsfig{file=fig1_a.eps, angle=0}} }
185: %} \mbox{
186: {\scalebox{0.30}
187: {\hspace{0mm}\epsfig{file=fig1_b.eps, angle=0}} }
188: {\scalebox{0.30}
189: {\hspace{0mm}\epsfig{file=fig1_c.eps,angle=0}} }
190: }
191: \end{center}
192: %\par
193: \vspace{0mm} \caption{Probability of states for (a) the GSPC, (b)
194: the DJI, and (c) individual stocks as a function of $d$, where (c)
195: is the average probability for all individual stocks. }
196: \label{fig:pstate}
197: \end{figure}
198: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
199:
200: When data is discretized, it is important to determine the size of
201: $d$ because probability of each state is varied by $d$. In case of
202: very small $d$, most of return value is belonged not to the
203: intermediate state but to the increase or decrease states.
204: Therefore, the data can be regarded as two-states practically.
205: Also, when $d$ is very high, the greater part of return is fallen
206: under intermediate state. So data is able to be considered in
207: one-state system. As the value of $d$, the range of intermediate
208: state, is changed, the probability of each state is varied. Fig.
209: \ref{fig:pstate} represents the probability of each state. The
210: probabilities of increase and decrease states are almost same.
211: Therefore, the probability of intermediate state increases as $d$
212: is increasing, while those of increase and decrease are reduced.
213: Around $d=0.003$, the probabilities of three states are
214: approximately same for both of composite stock index. On the other
215: hand, individual stocks represent the same probability at
216: $d=0.006$. The reason, why the value of $d$ which makes the same
217: probability for composite stock index is not same to that for
218: individual stock prices, is that index usually does not change its
219: value abruptly in a day compare with individual stocks, because
220: composite stock index is average or weighted average of individual
221: stock prices.
222:
223: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
224: \begin{figure}[tbph]
225: \begin{center}
226: \mbox{
227: {\scalebox{0.35}
228: {\hspace{0mm}\epsfig{file=fig2_a.eps, angle=0}} }
229: %} \mbox{
230: {\scalebox{0.35}
231: {\hspace{0mm}\epsfig{file=fig2_b.eps, angle=0}} }
232: }
233: \end{center}
234: %\par
235: \vspace{0mm} \caption{Mean value of the transfer entropy for (a)
236: the GSPC and (b) the DJI as a function of $d$: $\blacksquare$ for
237: $T_{I \rightarrow S}$, $\blacktriangle$ for $T_{S \rightarrow I}$,
238: $\square$ for $T^{\rm shuffle}_{I \rightarrow S}$, and
239: $\vartriangle$ for $T^{\rm shuffle}_{S \rightarrow I}$.}
240: \label{fig:meanTE-d}
241: \end{figure}
242: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
243:
244: Fig. \ref{fig:meanTE-d} shows the mean value of the transfer
245: entropy between composite stock index ($I$) and price of
246: individual stocks ($S$) for the GSPC and the DJI as a function of
247: $d$ with $k=1$ and $l=1$. The transfer entropy from the stock
248: index to the stock prices, $T_{I \rightarrow S}$, is almost higher
249: than that from the stock prices to the stock index, $T_{S
250: \rightarrow I}$. At $d=0$, discretized data is fallen into
251: two-states because the intermediate state is disappeared.
252: Therefore, it has smaller value of the transfer entropy compared
253: with that for three states. As $d$ is increasing, the number of
254: state turns to three, and the transfer entropy is maximized around
255: $d=0.015$.
256: %, and the probabilities of each state is almost same around $d=0.015$.
257: %Therefore, transfer entropy is maximized there.
258: Above $d$ which makes maximized the transfer entropy, the larger
259: $d$, the larger probability of the intermediate state. Moreover,
260: above about 0.02, $P(1)$ for the index goes close to 1. Therefore,
261: the transfer entropy is deceasing and finally goes to 0 because
262: all data is fallen into the intermediate state at very large $d$.
263:
264: Open squares ($\square$) and triangles ($\vartriangle$) of Fig.
265: \ref{fig:meanTE-d} represent the transfer entropy from shuffled
266: data. As expected, the transfer entropy from shuffled data is
267: smaller than that from the original data, and also the difference
268: between $T_{I \rightarrow S}$ and $T_{S \rightarrow I}$ is
269: disappeared below $d \approx 0.02$ and above $d \approx 0.04$ in
270: both indices. In the range from around 0.02 to around 0.04, number
271: of states for the indices is 1, while it is still 3 for individual
272: stocks. Therefore, this difference between them triggers
273: discrepancy of the transfer entropy between the indices and
274: stocks.
275:
276: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
277: %\begin{figure}[tbph]
278: %\begin{center}
279: %\mbox{
280: % {\scalebox{0.35}
281: % {\hspace{0mm}\epsfig{file=GSPC_diffmTE.eps,angle=0}} }
282: %} \mbox{
283: % {\scalebox{0.35}
284: % {\hspace{0mm}\epsfig{file=DJI_diffmTE.eps, angle=0}} }
285: %}
286: %\end{center}
287: %\par
288: %\vspace{0mm} \caption{The average difference of transfer entropy
289: %between composite stock index and stock prices for (a) the GSPC
290: %and (b) the DJI as a function of $d$.} \label{fig:diffmTE}
291: %\end{figure}
292: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
293:
294: %The average difference of transfer entropy between composite stock
295: %index and stock prices, $T_{I \rightarrow S} - T_{S \rightarrow
296: %I}$, is shown in Fig. \ref{fig:diffmTE}, where $I$ and $S$
297: %indicate index and stocks, respectively. The plots in Fig.
298: %\ref{fig:diffmTE} is obtained by averaging of transfer entropy
299: %over each stocks for given $d$.
300:
301: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
302: \begin{figure}[tbph]
303: \begin{center}
304: \mbox{
305: {\scalebox{0.35}
306: {\hspace{0mm}\epsfig{file=fig3_a.eps, angle=0}} }
307: %} \mbox{
308: {\scalebox{0.35}
309: {\hspace{0mm}\epsfig{file=fig3_b.eps, angle=0}} }
310:
311: } \mbox{
312: {\scalebox{0.35}
313: {\hspace{0mm}\epsfig{file=fig3_c.eps, angle=0}} }
314: %} \mbox{
315: {\scalebox{0.35}
316: {\hspace{0mm}\epsfig{file=fig3_d.eps, angle=0}} }
317:
318: }
319: \end{center}
320: %\par
321: \vspace{0mm} \caption{At $d=0.015$, the frequency of $T_{I
322: \rightarrow S}$ and $T_{S \rightarrow I}$ for (a) the GSPC and (b)
323: the DJI, and the frequency of difference between $T_{I \rightarrow
324: S}$ and $T_{S \rightarrow I}$ for (c) the GSPC and (d) the DJI.
325: %Numbers on $x$-axis indicate individual stocks.
326: } \label{fig:dTEstocks}
327: \end{figure}
328: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
329:
330: Figs. \ref{fig:dTEstocks}(a) and \ref{fig:dTEstocks}(b) show the
331: frequency of the transfer entropy between composite stock index
332: and stock prices at $d=0.015$. Frequency distribution of the
333: transfer entropy from index to stocks is more skewed to right than
334: that from stocks to index. Figs. \ref{fig:dTEstocks}(c) and
335: \ref{fig:dTEstocks}(d) show the difference between $T_{I
336: \rightarrow S}$ and $T_{S \rightarrow I}$. For the majority of
337: companies, the transfer entropy from index to stocks are larger
338: than the transfer entropy for the reverse. However, about 35\%
339: companies gives information to index of the next day.
340:
341: %For some stocks, transfer entropy from stock to index is lager
342: %than that from index to stock, though the average difference
343: %between index and stocks, $T_{I \rightarrow S} - T_{S \rightarrow
344: %I}$, has positive sign more frequently.
345:
346: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
347: \begin{figure}[tbph]
348: \begin{center}
349: \mbox{
350: {\scalebox{0.40}
351: {\hspace{0mm}\epsfig{file=fig4_a.eps, angle=0}} }
352: %} \mbox{
353: {\scalebox{0.40}
354: {\hspace{0mm}\epsfig{file=fig4_b.eps, angle=0}} }
355: }
356: \end{center}
357: %\par
358: \vspace{0mm} \caption{The relation between $T_{I \rightarrow S}$
359: and $T_{S \rightarrow I}$ for (a) the GSPC and (b) the DJI.}
360: \label{fig:relation}
361: \end{figure}
362: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
363: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
364: \begin{table}[tbp]
365: \centering%
366: \begin{tabular}{lllll}
367: \hline
368: &GSPC $\rightarrow$ stock & stock $\rightarrow$ GSPC \\
369: \hline\hline
370: 1&Pepsico Inc.&Centerpoint Energy Inc.\\
371: 2&FPL Group Inc.&Duke Energy Corp.\\
372: 3&Xerox Corp.&Xerox Corp.\\
373: 4&Entergy Corp.&Bristol-Myers Squibb Co.\\
374: 5&Consolidated Edison Inc.&International Business Machines Corp.\\
375: 6&Walt Disney co.&American Electric Power Co. Inc.\\
376: 7&Union Pacific Corp.&PG \& E Corp.\\
377: 8&United Technologies Corp.&TXU Corp.\\
378: 9&Clorox Co.&Wyeth\\
379: 10&Centerpoint Energy Inc.&Consolidated Edison Inc.\\
380: \hline
381: &\\
382: \hline
383: &DJI $\rightarrow$ stock & stock $\rightarrow$ DJI\\
384: \hline\hline
385: 1&Walt Disney Co.&Xerox Corp.\\
386: 2&Consolidated Edison Inc.&Centerpoint Energy Inc.\\
387: 3&Xerox Corp.&Willams Companies Inc.\\
388: 4&Whirlpool Corp.&Duke Energy Corp.\\
389: 5&Pepsico Inc.&Southern Co.\\
390: 6&FPL Group Inc.&PG \& E Corp.\\
391: 7&Coca-Cola Co.&American Electric Power Co. Inc.\\
392: 8&United Technologies Corp.&Honeywell International Inc.\\
393: 9&Corning Inc.&Entergy Corp.\\
394: 10&PG \& E Corp.&Bristol-Myers Squibb Co.\\
395: \hline
396: \end{tabular}%
397: \vspace{5mm} \caption{The top 10 companies of the transfer
398: entropy.} \label{table:company}
399: \end{table}
400: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
401: Fig. \ref{fig:relation} shows the positive relation between $T_{I
402: \rightarrow S}$ and $T_{S \rightarrow I}$. The value of
403: correlation between them is 0.51(9) for the GSPC and 0.40(9) for
404: the DJI. In Table \ref{table:company}, the top 10 companies of the
405: transfer entropy is listed. Among the top 10 companies, Xerox
406: Corp., Entergy Corp., Consolidated Edison Inc., Centerpoint Energy
407: Inc., and PG \& E belong to the top 10 companies for both $T_{I
408: \rightarrow S}$ and $T_{S \rightarrow I}$. Both Fig.
409: \ref{fig:relation} and Table \ref{table:company} show that the
410: higher $T_{I \rightarrow S}$, the higher $T_{S \rightarrow I}$,
411: though the average value of $T_{I \rightarrow S}$ is higher than
412: one of $T_{S \rightarrow I}$. Consequently, individual stocks are
413: able to be divided into highly connected stocks and lowly
414: connected stocks to the market.
415:
416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
417: %\begin{figure}[tbph]
418: %\begin{center}
419: %\mbox{
420: % {\scalebox{0.35}
421: % {\hspace{0mm}\epsfig{file=GSPC_TEmean_aslen.eps, angle=0}} }
422: %} \mbox{
423: % {\scalebox{0.35}
424: % {\hspace{0mm}\epsfig{file=DJI_TEmean_aslen.eps, angle=0}} }
425: %}
426: %\end{center}
427: %\par
428: %\vspace{0mm} \caption{Transfer entropy for various data size of
429: %(a) the GSPC and (b) the DJI as a function of $d$.}
430: %\label{fig:TEasLeng}
431: %\end{figure}
432: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
433:
434: %Transfer entropy for various data size of (a) the GSPC and (b) the
435: %DJI as a function of $d$.
436:
437: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
438: %\section{Artificial data analysis}
439: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
440:
441: %artificial data analysis
442:
443: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
444: \section{Conclusion}
445: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
446:
447: The concept of the transfer entropy has been proposed for finding
448: direction of casuality. Using the measure, we are able to
449: investigate the information flow between stock index and
450: individual stocks. Our results indicate that there is a stronger
451: flow of information from the stock index to the individual stocks
452: than vice versa, and the transfer entropy for both direction has
453: positive correlation. Moreover, we expect similar result to the
454: U.S. market for other stock markets. As a matter of fact, the
455: result of the information flow in Japan stock market also produces
456: the same directional casuality although it is not shown in this
457: paper.
458:
459: We have desire to find the correlations between the direction of
460: information flow and company profile. However, we could not find
461: it yet. The division of individual stocks due to the direction of
462: casuality between composite stock index and companies may be
463: useful for the stock investment strategies.
464:
465: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
466: \begin{ack}
467:
468: This work is supported in part by Creative Research Initiatives of
469: the Korea Ministry of Science and Technology (O.K.), the Second
470: Brain Korea 21 project (J.-S.Y.), and Grant No.
471: R01-2004-000-10148-1 from the Basic Research Program of KOSEF
472: (J.-S.Y.).
473:
474: \end{ack}
475: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
476:
477: %\newpage
478: \begin{thebibliography}{99}
479: \bibitem{arthur} W. B. Arthur, S. N. Durlauf, D. A. Lane, {\it The Economy as an Evolving Complex System
480: II} (Perseus Books, 1997).
481:
482: \bibitem{mategna} R. N. Mantegna, H. E. Stanley, {\it An Introduction to
483: Econophysics} (Cambridge University Press, 2000).
484:
485: \bibitem{bouchaud} J.-P. Bouchaud, M. Potters, {\it Theory of Financial
486: Risks} (Cambridge University Press, 2000).
487:
488: \bibitem{mandel} B. B. Mandelbrot, Quant. Finance \textbf{1} (2001) 124.
489:
490: \bibitem{kullmann} L. Kullmann, J. Kert\'{e}sz, R. N. Mantegna,
491: Physica A \textbf{287} (2000) 412.
492:
493: \bibitem{giada} L. Giada, M. Marsili, Physica A \textbf{315}
494: (2002) 650.
495:
496: \bibitem{eguiluz} V. M. Eguiluz, M. Zimmermann, Phys. Rev. Lett. \textbf{85} (2000)
497: 5659.
498:
499: \bibitem{krawiecki} A. Krawiecki, J. A. Ho\l yst, D. Helbing, Phys. Rev.
500: Lett. \textbf{89} (2002) 158701.
501:
502: \bibitem{takaishi} T. Takaishi, Int. J. Mod. Phys. C \textbf{16} (2005) 1311.
503:
504: \bibitem{kaizoji02} T. Kaizoji, Physica A \textbf{287} (2000) 493.
505:
506: \bibitem{jbpark} J. B. Park, J. W. Lee, J.-S. Yang, H.-H. Jo, H.-T. Moon, Physica A \textbf{379} (2007)
507: 179.
508:
509: \bibitem{jwlee} J. W. Lee, J. B. Park, H.-H. Jo, J.-S. Yang, H.-T.
510: Moon, physics/0607282, 2006.
511:
512: \bibitem{wsjung1} W.-S. Jung, S. Chae, J.-S. Yang, H.-T. Moon,
513: Physica A \textbf{361} (2006) 263.
514:
515: \bibitem{wsjung2} W.-S. Jung, O. Kwon, J.-S. Yang, H.-T. Moon,
516: J. Korean Phys. Soc. \textbf{48} (2006) S135.
517:
518: \bibitem{matal} K. Matal, M. Pal, H. Salunkay, H. E. Stanley, Europhys.
519: Lett. \textbf{66} (2004) 909.
520:
521: \bibitem{yang} J.-S. Yang, S. Chae, W.-S. Jung, H.-T. Moon, Physica A
522: \textbf{363} (2006) 377.
523:
524: \bibitem{schae} S. Chae, W.-S. Jung, J.-S. Yang, H.-T. Moon,
525: J. Korean Phys. Soc. \textbf{48} (2006) 313.
526:
527: \bibitem{jsyang2} J.-S. Yang, W. Kwak, T. Kaizoji, I.-m. Kim,
528: physics/0701179, 2007.
529:
530: \bibitem{shannon} C. E. Shannon, W. Weaver, {\it The Mathematical Theory of
531: Information} (University of Illinois Press, 1994).
532:
533: \bibitem{schreiber} T. Schreiber, Phys. Rev. Lett. \textbf{85}
534: (2000) 461.
535:
536: \bibitem{marschinski} R. Marschinski, H. Kantz, Eur. Phys. J. B.
537: \textbf{30} (2002) 275.
538:
539: %\bibitem{skbaek} S. K. Baek, W.-S. Jung, O. Kwon, H.-T. Moon,
540: %physics/0509014, 2005.
541: \end{thebibliography}
542:
543: \end{document}
544: