Physica A 391 (2012) 3484–3495

Contents lists available at SciVerse ScienceDirect

Physica A
journal homepage: www.elsevier.com/locate/physa

Price–volume multifractal analysis and its application in Chinese

stock markets
Ying Yuan ∗ , Xin-tian Zhuang, Zhi-ying Liu
School of Business Administration, Northeastern University, Shenyang, 110819, China

article info abstract

Article history: An empirical research on Chinese stock markets is conducted using statistical tools.
Received 2 October 2011 First, the multifractality of stock price return series, ri (ri = ln(Pt +1 ) − ln(Pt )) and
Received in revised form 18 November trading volume variation series, vi (vi = ln(Vt +1 ) − ln(Vt )) is confirmed using multifractal
2011
detrended fluctuation analysis. Furthermore, a multifractal detrended cross-correlation
Available online 6 February 2012
analysis between stock price return and trading volume variation in Chinese stock markets
is also conducted. It is shown that the cross relationship between them is also found to
Keywords:
Trading volume
be multifractal. Second, the cross-correlation between stock price Pi and trading volume
Generalized Hurst exponents Vi is empirically studied using cross-correlation function and detrended cross-correlation
Price Limits analysis. It is found that both Shanghai stock market and Shenzhen stock market show
Reform of Non-tradable Shares pronounced long-range cross-correlations between stock price and trading volume. Third,
Financial crisis a composite index R based on price and trading volume is introduced. Compared with stock
price return series ri and trading volume variation series vi , R variation series not only
remain the characteristics of original series but also demonstrate the relative correlation
between stock price and trading volume. Finally, we analyze the multifractal characteristics
of R variation series before and after three financial events in China (namely, Price Limits,
Reform of Non-tradable Shares and financial crisis in 2008) in the whole period of sample
to study the changes of stock market fluctuation and financial risk. It is found that the
empirical results verified the validity of R.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction

The financial market can be thought of as a complex system and thus the operating law of the financial market is difficult
to understand and describe. Traditional financial market theory relies on the Efficient Market Hypothesis (EMH) [1]. Under
the frame of EMH, stock returns follow a normal distribution and price behaviors obey ‘random-walk’ hypothesis. However,
in some recent papers, a deviation from EMH was recently reported for both transition and some developed economies [2–4].
For example, the deviation from EMH was reported for the European transition economies by Podobnik et al. [4]. The
results demonstrated that many time series of major indices exhibit power-law correlations in their returns. Therefore,
it is necessary to develop an alternative method to extract the characteristics of the price fluctuation to enable an accurate
estimate for risk prevention and control purpose.
In 1963, Mandelbrot observed the process of returns showing time scaling property in his analysis of cotton prices [5].
After that, Mandelbrot & Stantely introduced the method of scaling invariance from the complexity science into the economic
systems for the first time [6]. The fractal theory was applied to redescribe the scaling invariance property theoretically. The
existence of scaling invariance characteristics was further confirmed by some prominent approaches, including rescaled
range analysis [7], the Levy stable distribution [8] and detrended fluctuation analysis [9–11]. However, these models could
not be used to address the scaling behavior of the probability distributions in financial time series due to the limitation of
retrieving recapitulatory information.

∗ Corresponding author. Tel.: +86 24 83686784; fax: +86 24 23891569.

E-mail address: yuanying1980@126.com (Y. Yuan).

0378-4371/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.physa.2012.01.034
 Y. Yuan et al. / Physica A 391 (2012) 3484–3495 3485

In order to make up the limitation of single fractal methods, Mandelbrot advanced a new method, namely, multifractal
method [12]. Multifractal analysis can be used to divide a complex system into various regions according to their complexity
and therefore is able to describe market volatilities. Therefore, many scholars analyzed multifractality of different markets
using multifractal tools. In fact, the presence of multifractality has been a ‘‘stylized fact’’ in financial markets [13]. It has been
verified that multifractality widely existed in financial markets such as stock markets, future markets, spot markets, foreign
exchange markets, derivative markets, interest rate markets and so on [14–22]. Prior researches concentrated on confirming
the multifractality of different financial markets, and nowadays some scholars try to explore the complicated statistical
characteristics. Ho et al. conducted a multifractal analysis on the Taiwan stock price index. He found that one scaling
exponent is insufficient to capture the time dynamics of Taiwan stock market and the nature of multifractal phenomena
in Taiwan stock market might be interpreted using the multiplicative cascade process of stock market information [23]. Sun
et al. investigated statistically the correlations between the parameters of the multifractal spectrum and the variation of
close returns [24]. Chen et al. predicted the price movements using two kinds of sign sequences as given conditions. One is
the parameter of the multifractal spectrum 1f based on the minute indices, and the other is the variation of the closing index.
Results showed that large fluctuations of the closing price and the conditions are strongly connected in these two methods
and some sign sequences of the parameter 1f can be used to predict the probability of near future price movements [25].
Wei et al. proposed a multifractal volatility (MFV) model based on the multifractal spectrum of one trading day. His empirical
results showed that the MFV model has the forecasting accuracy [26]. We also measured the multifractality of stock price
fluctuation and advanced two risk measures based on generalized Hurst exponents [27]. In addition, another robust and
powerful technique is Multifractal Detrended Fluctuation Analysis (MF-DFA), which is proposed by Kantelhardt in 2002 [28].
The advantages of MF-DFA over many techniques are that it permits the detection of long-range correlations embedded in
seemingly non-stationary time series, and also avoids the spurious detection of apparent long-range correlations that are
an artifact of nonstationarity.
In addition, it should be noted that although some statistical characteristics of stock market data such as stock returns,
price series and volatility series have become stylized fact, behavior of other related variables such as trading volume
is less studied. In fact, as a most important index, the trading volume contains much of useful information about the
dynamics of price formation, which can help us understand the behavior of financial markets. Mu & Zhou studied the long
memory and multifractality in trading volumes for Chinese stocks and got some meaningful results [29]. Cross-correlation
function is a well-known statistical method used to establish the degree of correlation between two time series. This is
done considering that stationarity characterizes both time series under investigation. However, most time series are hardly
stationary. Therefore, when the time series need to be seen and analyzed as a whole cross-correlation function is not always
a valid choice. Podobnik and Stanley advanced a new method that deals with nonstationary time series, named Detrended
Cross-Correlation Analysis (DCCA) [9]. Then long-range power-law volatility cross-correlations in finance were reported
between couple of time series by using DCCA and between large number of simultaneously recorded time series [30–34].
For example, Podobnik et al. analyzed the cross-correlations between volume change and price change and power-law
cross-correlations were reported between the absolute values in price changes and volume changes [30]. On the basis of
DCCA, Zhou proposed a Multifractal DCCA (MF-DCCA) method to combine multifractal detrended fluctuation analysis and
DCCA method [10]. Furthermore, Arianos and Carbone advanced a method for estimating the cross-correlation Cxy (τ ) of
long-range correlated series x(t ) and y(t ), at varying lags τ and scales n [35].
On the basis of previous literature, in this paper, we conduct an empirical research on Chinese stock market using
statistical tools. This study extends previous work in several respects. Firstly, the multifractality of stock price return series
and trading volume variation series are confirmed using MF-DFA. Furthermore, a multifractal detrended cross-correlation
analysis between stock price return series and trading volume variation series in Chinese stock markets is also conducted.
It is shown that the cross relationship between them are also found to be multifractal. Secondly, the cross-correlation
between stock price Pi and trading volume Vi is empirically studied using cross-correlation function and detrended cross-
correlation analysis. It is found that both Shanghai stock market and Shenzhen stock market show pronounced long-range
cross-correlations between stock price and trading volume. Thirdly, a composite index R based on price and trading volume
is advanced. Compared with stock price return series ri and trading volume variation series vi , R variation series not only
remain the characteristics of original series but also demonstrate the relative correlation between stock price and trading
volume. Finally, we analyze the multifractal characteristics of R variation series before and after three financial events in
China (namely, Price Limits, Reform of Non-tradable Shares and financial crisis in 2008) in the whole period of sample to
study the changes of stock market fluctuation and financial risk. It is found that the empirical results verified the validity of R.
The structure of the paper is as follows. Section 2 describes the research method. Section 3 introduces the empirical data.
Section 4 presents the empirical results. Section 5 is a brief conclusion of the paper.

2. Research methods

2.1. Multifractal Detrended Fluctuation Analysis (MF-DFA)

Kantelhardt introduced a multifractal method called Multifractal Detrended Fluctuation Analysis (MF-DFA) [28]. The
operation of MF-DFA on the series x(i), where i = 1, 2, . . . , N and N is the length of the series, is as follows. With x we
indicate the mean value of series x(i).
3486 Y. Yuan et al. / Physica A 391 (2012) 3484–3495

We assume that x(i) are increments of a random walk process around the average x, thus the ‘‘trajectory’’ or ‘‘profile’’ is
given by the integration of the signal
i

y(i) = [x(k) − x] , i = 1, 2, . . . , N . (1)
k=1

Next, the integrated series is divided into Ns = int(N /s) nonoverlapping segments of equal length s. Since the length
N of the series is often not a multiple of the considered time scale s, a short part at the end of the profile y(i) may remain.
In order not to disregard this part of the series, the same procedure is repeated starting from the opposite end. Thereby,
2Ns segments are obtained altogether. We calculate the local trend for each of the 2Ns segments by a least-square fit of the
series. Then we determine the variance
s
1
F 2 (s, v) = {y[(v − 1)s + i] − yv (i)}2 (2)
s i=1

for each segment v , v = 1, . . . , Ns and

s
1
F 2 (s, v) = {y [N − (v − Ns )s + i] − yv (i)}2 (3)
s i=1

for v = Ns + 1, . . . , 2Ns . Here, yv (i) is the fitting line in segment v . Then, we average over all segments to obtain the q-th
order fluctuation function
  1q
2Ns
1  q
Fq (s) = [F (s, v)]
2 2 (4)
2Ns v=1

where, in general, the index variable q can take any real value except zero. Repeating the procedure described above, for
several time scales s, Fq (s) will increase as s increases. By analyzing log–log plots Fq (s) versus s for each value of q, we
determine the scaling behavior of the fluctuation functions. If the series x(i) is long-range power-law correlated, Fq (s) will
increase for large values of s as a power-law

Fq (s) ≈ shq . (5)

The value h0 corresponds to the limit hq as q → 0, and cannot be determined directly by using the averaging procedure
of Eq. (4) because of the diverging exponent. Instead, a logarithmic averaging procedure has to be employed,
 
2Ns
1 
F0 (s) = exp ln[F (s, v)]
2
≈ sh0 . (6)
4Ns v=1

In general the exponent hq will depend on q. For stationary series, h2 is the well-defined Hurst exponent H. Thus, we
call hq the generalized Hurst exponent. hq is independent from q, which characterizes monofractal series. The different
scaling of small and large fluctuations will yield a significant dependence of hq on q. For positive q, the segments v with
large variance (i.e. large deviation from the corresponding fit) will dominate the average Fq (s). Therefore, if q is positive, hq
describes the scaling behavior of the segments with large fluctuations; and generally, large fluctuations are characterized by
a smaller scaling exponent hq for multifractal time series. For negative q, the segments v with small variance will dominate
the average Fq (s). Thus, for negative q values, the scaling exponent hq describes the scaling behavior of segments with small
fluctuations, usually characterized by large scaling exponents.
Furthermore, in order to measure the degree of multifractality, we define 1h:
1h = hmax (q) − hmin (q) (7)
where 1h is the range of generalized Hurst exponents h(q). The greater is 1h, the stronger is the degree of multifractality,
therefore the greater is the stock risk, and vice versa.

2.2. Detrended cross-correlation analysis and multifractal detrended cross-correlation analysis

Detrended cross-correlation analysis was proposed by Podobnik to quantify long-range cross-correlations between time
series [9]. This method represents a generalization of the DFA and it is designed to analyze power-law cross-correlations
between two time series.
Consider two time series xi , i = 1, 2, 3, . . . , N and yi , i = 1, 2, 3, . . . , N. They are integrated to produce X (k) = i=1 x(i)
k

and Y (k) = i=1 y(i), where k is an integer between 1 and N. Next, the integrated series are divided into Nn segments of
k
equal length n and in each segment a linear (or higher order polynomial) regression is performed to capture the local trend.
 Y. Yuan et al. / Physica A 391 (2012) 3484–3495 3487

The integrated series X (k) and Y (k) are then detrended by subtracting the local trends Xn,s (k) and Yn,s (k) (ordinates of the
straight line or polynomials within each segment) from the data in each box and the detrended covariance is calculated as
Nn −1 n(s+1)
1   
2
(n) = X (k) − Xn,s+1 (k) Y (k) − Yn,s+1 (k) .
 
FDCCA (8)
nNn s=0 k=ns+1

Repeating this calculation for all segment sizes provides the relationship between FDCCA (n) and the segment size n. When
only one series is analyzed (X (k) = Y (k)) the detrended covariance FDCCA2
. If the original series xi and yi are power-law
cross-correlated, then FDCCA (n) ∼ nλ , and the scaling exponent λ are determined from the linear regression of log[FDCCA (n)]
versus log(n). The interpretation of λ is similar to that of the DFA exponent α . Long-range cross-correlations between two
series imply that each series has long memory of its own previous values and additionally has a long memory of the previous
values of the other series [9].
Multifractal detrended cross-correlation analysis is proposed by Zhou et al. and it is a multifractal modification of
DCCA [10].

3. Data analyzed

The original data are taken from Chinese stock markets, which are composed Shanghai Stock Exchange (SHSE) and the
Shenzhen Stock Exchange (SZSE). The Chinese stock markets were newly set up markets and have only a history of more
than 20 years. The Shanghai Stock Exchange was established on November 26, 1990, and put into operation on December
19. Shortly after, the SZSE was established on December 1, 1990 and put into operation on April 3, 1991. On a trading day
both Stock Exchanges open from 9:30 to 11:30 am and then from 1:00 to 3:00 pm, so there are 4 trading hours on one day.
The data analyzed are daily closing price series, trading volume series and their logarithmic variation of Shanghai Stock
Index (SHSI) and Shenzhen Component Index (SZCI). The sample period for SHSI is from December 20, 1990 to December 30,
2010 and the sample period for SZCI is from April 3, 1991 to December 30, 2010. In this paper, when we conduct multifractal
analysis we focus on the stock price return series (that is ln(Pt +1 ) − ln(Pt )), logarithmic trading volume variation series (that
is ln(Vt +1 ) − ln(Vt )) and logarithmic R variation series (that is ln(Rt +1 ) − ln(Rt )) of Chinese stock markets.

4. Empirical results

4.1. Multifractal analysis of Chinese stock markets

4.1.1. Multifractal characteristics of stock price return series in Chinese stock markets
We performed the MF-DFA1 over the stock price return series of SHSI and SZCI, respectively. We calculated the
fluctuations Fq (s) for scales s ranging from 10 events to N /4, where N is the total length of the series. Fig. 1 shows the
fluctuation functions Fq (s) of stock price return series for SHSI for q = −10, q = −8 · · · and q = +10, where the upper
and the lower curves are the curves of q = 10 and q = −10. As can be seen from Fig. 1, there is a crossover time scale
s∗ (about s∗ ≈ 28) in the double logarithmic plots of Fq (s) versus s. This shows that there are different scaling laws and
scaling exponents for time scales s > s∗ and s < s∗ . Fig. 2(a) shows the q-dependence of the generalized Hurst exponent hq
determined by fits in the regime 10 < s < N /4. As shown in Fig. 2(a), when q varies from −10 to 10, h(q) decreases from
1.057 to 0.2817. h(q) is not a constant, indicating multifractality in time series. Similarly, Fig. 2(b) shows the generalized
Hurst exponent plot of stock price return series of SZCI. As can be seen from Fig. 2(b), when q varies from −10 to 10, h(q)
decreases from 1.1477 to 0.4538. h(q) is also not a constant, indicating multifractality in time series. The values of hq of stock
price return series in the regime 10 < s < N /4 for SHSI and SZCI are illustrated in Table 1, respectively.

4.1.2. Multifractal characteristics of trading volume variation series in Chinese stock markets
In order to better understand the dynamics of price formation, it is very useful to understand the structure of trading
volumes. Similarly, we performed the MF-DFA1 over the trading volume variation series of SHSI and SZCI, respectively.
Fig. 3(a) and (b) show the q-dependence of the generalized Hurst exponents hq determined by fits in the regime 10 < s <
N /4 for SHSI and SHCI, respectively. As shown in Fig. 3, when q varies from −10 to 10, h(q) decreases monotonously for
both of them, indicating multifractality in both time series. The values of h(q) in the regime 10 < s < N /4 for SHSI and SZCI
are illustrated in Table 2, respectively.

4.1.3. Multifractal detrended cross-correlation analysis between stock price return and trading volume variation in Chinese stock
markets
Furthermore, we also conducted an empirical analysis on price return series and trading volume variation series using
multifractal DCCA proposed by Zhou [10]. For detailed procedures, please see the Refs. [10,37]. Fig. 4(a) and (b) display the
relationship between cross-correlation exponent h(q) and q. As can be seen from Fig. 4(a) and (b), both of cross-correlation
relationships for Shanghai stock markets and Shenzhen stock markets are found to be multifractal because for different
3488 Y. Yuan et al. / Physica A 391 (2012) 3484–3495

Fig. 1. The multifractal fluctuation function Fq (s) of stock price return series for SHSI.

(a) SHSI. (b) SZCI.

Fig. 2. The generalized Hurst exponents h(q) of stock price return series.

Table 1
Values of the generalized qth-order Hurst exponents
h(q) of the stock price return series.
Order SHSI SZCI
q k=1 k=1

−10 1.057 1.1477

−8 1.0348 1.1246
−6 1.0004 1.0876
−4 0.942 1.0212
−2 0.8187 0.882
0 0.6096 0.6474
2 0.5439 0.626
4 0.413 0.5585
6 0.3422 0.508
8 0.3047 0.4753
10 0.2817 0.4538
1h 0.7753 0.6939

q, there are different exponents h(q). The values of h(q) for SHSI and SZCI are illustrated in Table 3, respectively. More
clearly, the generalized Hurst exponents of stock price return series, trading volume variation series and cross-correlation
relationships between them are shown in Fig. 5(a) and (b).
 Y. Yuan et al. / Physica A 391 (2012) 3484–3495 3489

(a) SHSI. (b) SZCI.

Fig. 3. The generalized Hurst exponents h(q) of trading volume variation series.

Table 2
Values of the generalized qth-order Hurst exponents
h(q) of the trading volume variation series.
q SHSI SZCI

−10 0.3982 0.4229

−8 0.3823 0.4048
−6 0.3612 0.3807
−4 0.3331 0.3488
−2 0.2911 0.3054
0 0.2145 0.235
2 0.1643 0.203
4 0.0992 0.1313
6 0.0591 0.0711
8 0.0352 0.034
10 0.0196 0.0109
1h 0.3786 0.412

(a) SHSI. (b) SZCI.

Fig. 4. The generalized Hurst exponents h(q) by MF-DCCA.

3490 Y. Yuan et al. / Physica A 391 (2012) 3484–3495

Table 3
Values of the generalized qth-order Hurst exponents
h(q) by MF-DCCA.
q SHSI SZCI

−10 0.7026 0.8281

−8 0.6713 0.7976
−6 0.6235 0.7481
−4 0.5604 0.6571
−2 0.5079 0.5256
0 0.477 0.4736
2 0.444 0.4312
4 0.3801 0.3679
6 0.3152 0.3027
8 0.2724 0.2604
10 0.2451 0.2348
1h 0.4575 0.5933

(a) SHSI. (b) SZCI.

Fig. 5. The generalized Hurst exponents h(q) for stock price return, trading volume variation and cross-correlation relationship between them.

4.2. Cross-correlation between stock price and trading volume

As two of the most important variables of stock market data, price and trading volume interact each other and form
together the dynamics of stock market. As we all known, trading volume can have a large impact on the stock prices. It is
an established fact that trading volume is a major propelling force for stock prices. Therefore, it is necessary to research the
correlation between price and trading volume.

4.2.1. Cross-correlation analysis between stock price and trading volume

The conditional dependence of trading volume on stock price explains the correlation between trading volume and stock
price to a certain degree. In order to further understand the relation between trading volume and price, we investigate the
cross-correlation between trading volume and price. The volume–price cross-correlation function is defined as ⟨V (t ′ )P (t ′ +
t )⟩, where t is the lagged time window [38].
Fig. 6 shows the volume–price cross-correlation function of the SHSI and SZCI. Each of Fig. 6(a) and (b) show a positive
cross-correlation between trading volume and price for both the positive and the negative time direction. For the positive
time direction, it indicates that high volume triggers high price, and small volume triggers low price. Large volume indicates
high liquidity of markets, which may result in sharp rise of markets, whereas small volume leads markets to become inactive
and display a low price. For the negative time direction, it suggests also that high price triggers a big trading volume, and
small price triggers a small volume [38]. The volume–price cross-correlation function shows that there exists a positive
long-range correlation between trading volume and price.

4.2.2. Detrended cross-correlation analysis between stock price and trading volume
It should be noted that cross-correlation function is done considering that stationarity characterizes both time series
under investigation. However, most time series are hardly stationary. Therefore, when the time series need to be seen and
 Y. Yuan et al. / Physica A 391 (2012) 3484–3495 3491

(a) SHSI. (b) SZCI.

Fig. 6. The volume–price cross-correlation functions.

analyzed as a whole cross-correlation is not always a valid choice. Podobnik and Stanley advanced a new method that deals
with nonstationary time series, named Detrended Cross-Correlation Analysis (DCCA) [9].
Podobnik et al. proposed a cross-correlation statistic in analogy to the Ljung–Box test [30]. For two series, {xt , t = 1,
2, 3, . . . , N } and {yt , t = 1, 2, 3, . . . , N }, the test statistic
m
 Xi 2
Qcc (m) = N 2 . (9)
i=1
N −i
Here, the cross-correlation function
N

xk yk−i
k=i+1
Xi =  . (10)
N
 N

xk 2 yk 2
k=1 k=1

The cross-correlation statistic QCC (m) is approximately χ 2 (m) distributed with m degrees of freedom. The statistic can
be used the null hypothesis that none of the first m cross-correlation coefficient is different from zero. If for a broad range
of m the test statistic of Eq. (9) exceeds the critical values of χ 2 (m)(QCC (m) > χ0.95 2 (m)), we claim that there are not only
cross-correlations, but there are long-range cross-correlations [30,39].
Fig. 7(a) and (b) demonstrated the cross-correlation statistics between stock price and trading volume in Shanghai stock
market and Shenzhen stock market. The degrees of freedom vary from 100 to 103 . As a comparison, we also plot the critical
values with m increasing. As can be seen from Fig. 7(a) and (b), for the whole range of m, all of the test statistics exceed the
critical values of χ 2 (m)(QCC (m) > χ0.95 2 (m)), indicating that there are not only cross-correlations, but there are long-range
cross-correlations between stock price and trading volume for both Shanghai stock market and Shenzhen stock market.
Furthermore, the DCCA should be used to test the presence of cross-correlations quantitatively [30]. We conducted an
empirical analysis on stock price and trading volume using DCCA. The empirical results are shown in Fig. 8(a) and (b).
As can be seen from Fig. 8, both Shanghai stock market and Shenzhen stock market show pronounced long-range cross-
correlations between stock price and trading volume because both of DCCA exponents are greater than 0.5(DCCA _SHSI ≈
1.3 and DCCA _SZCI ≈ 1.3).

4.3. Composite index based on price and trading volume and its multifractality

As seen from the empirical results in Section 4.2, there is a stable correlation between the trading volume and price.
Indeed, an important quantity that characterizes the dynamics of price movements is the number of shares traded (trading
volume) in a period of time. Based on the above-mentioned discussion, it is suggested that price and trading volume should
be used as a whole to analyze the Chinese stock market. Therefore, a composite index based on price and trading volume is
advanced according to the way of Ref. [36]:
V
R= (11)
P
where R is the ratio trading volume to stock price, V is the trading volume and P is the closing price of stock price index.
3492 Y. Yuan et al. / Physica A 391 (2012) 3484–3495

(a) SHSI. (b) SZCI.

Fig. 7. The volume–price cross-correlation statistics.

(a) SHSI. (b) SZCI.

Fig. 8. The volume–price detrended cross-correlation analysis.

Therefore, a new composite index series for each of SHSI and SZCI can be calculated through Eq. (11). Furthermore, it
is necessary to test the multifractality of the R series and compare the difference between R series and each of the original
data.
We performed the MF-DFA1 over the logarithmic variation of R series of SHSI and SZCI, respectively. Fig. 9(a) and (b)
show the q-dependence of the generalized Hurst exponents h(q) determined by fits in the regime 10 < s < N /4 for SHSI
and SHCI, respectively. As shown in Fig. 9 when q varies from −10 to 10, h(q) decreases monotonously for both of them,
indicating multifractality in both time series. The values of hq in the regime 10 < s < N /4 for SHSI and SZCI are illustrated
in Table 4, respectively. More clearly, the generalized Hurst exponents of price return series, trading volume variation series
and R variation series are shown in Fig. 10(a) and (b). As can be seen in Fig. 10(a) and (b), for both SHSI and SZCI, compared
with original series, R variation series basically remain the same trends and display similar characteristics. In another words,
R variation series do not change the characteristics of original series and meanwhile demonstrate the relative correlation
between stock price and trading volume. Therefore, more valuable information can be revealed about the dynamics of stock
market if the new index is valid.

4.4. Applications of the composite index in financial practices

Taking Chinese stock market as an example, we compared the multifractal characteristics before and after three financial
events in China (namely, Price Limits, Reform of Non-tradable Shares and financial crisis in 2008) in the whole period of
sample to study the changes of stock market fluctuation and financial risk. Price Limits started on December 27, 1996, Non-
tradable Shares Reform was on April 30, 2005 and Financial crisis in 2008 started on September, 15, 2008. (Because Lehman
 Y. Yuan et al. / Physica A 391 (2012) 3484–3495 3493

(a) SHSI. (b) SZCI.

Fig. 9. The generalized Hurst exponents h(q) of R variation series.

Table 4
Values of the generalized qth-order Hurst exponents
h(q) of R variation series.
q SHSI SZCI
k=1 k=1

−10 0.4364 0.503

−8 0.4185 0.4836
−6 0.3941 0.4574
−4 0.3603 0.422
−2 0.3082 0.3708
0 0.2126 0.30615
2 0.1543 0.2415
4 0.0836 0.1672
6 0.0421 0.1112
8 0.0183 0.0792
10 0.003 0.0597
1h 0.4334 0.4433

(a) SHSI. (b) SZCI.

Fig. 10. Comparison of the generalized Hurst exponents h(q) for different series.

Brothers filed for bankruptcy protection on September, 15, 2008, and this means the financial crisis has begun to break out
formally.) Therefore, the three financial events divide the whole series into four periods. The first one is before Price Limits;
the second one is after Price Limits and before Reform; and third one is after Reform and before financial crisis; the last one
is after financial crisis.
3494 Y. Yuan et al. / Physica A 391 (2012) 3484–3495

Table 5
Comparison of 1h before and after three events in Chinese stock markets.
Market states Before Price Limits After Price Limits before Reform After Reform before crisis After crisis

Time periods 12/19/1990–12/26/1996 12/27/1996–04/29/2005 05/09/2005–09/12/2008 09/16/2008–12/30/2010

1h (SHSI) 0.4614 0.4521 0.4911 0.5582
1h (SZCI) 0.597 0.4317 0.4647 0.6553

The 1h based on generalized Hurst exponents of different market states are shown in Table 5. As can be seen from Table 5,
the strength of multifractality after Price Limits is weaker than that of multifractality before Price Limits, and the 1h based
on generalized Hurst exponents after Price Limits is smaller than that of before Price Limits. These results indicate that after
the Price Limits, the stock price fluctuation decreases and the Price Limits have played a positive role in stabilizing sharp
fluctuations of market prices and stabilizing the stock market. In addition, the strength of multifractality after Reform of
Non-tradable Shares is stronger than that of multifractality before Reform of Non-tradable Shares (but after Price Limits),
and the 1h based on generalized Hurst exponents after Reform of Non-tradable Shares (but before financial crisis) is greater
than that of before Reform of Non-tradable Shares (but after Price Limits). These results indicate that after Reform of Non-
tradable Shares, the stock price fluctuation increases and the Reform of Non-tradable Shares increases the market risk.
Finally, the strength of multifractality after financial crisis is stronger than that of multifractality before financial crisis (but
after Reform), and the 1h based on generalized Hurst exponents after financial crisis is greater than that of before financial
crisis. All of these findings are consistent with the real situations. These empirical results verify the validity of R.

5. Conclusions

For multifractal model describes the local characteristics of the asset price process through a range of scalings, it
is acknowledged as the most appropriate model for price variations. The multifractal detrended fluctuation analysis,
performed in this study, leads to a better understanding of such complex stock market.
Our study extends previous work in several respects. Firstly, the multifractality of stock price return series and
trading volume variation series are confirmed using multifractal detrended fluctuation analysis. Furthermore, a multifractal
detrended cross-correlation analysis between stock price return and trading volume variation in Chinese stock markets
is also conducted. It is shown that the cross relationship between them are also found to be multifractal. Secondly, the
cross-correlation between stock price Pi and trading volume Vi is empirically studied using cross-correlation function
and detrended cross-correlation analysis. It is found that both Shanghai stock market and Shenzhen stock market show
pronounced long-range cross-correlations between stock price and trading volume. Thirdly, a composite index R based on
price and trading volume is advanced. Compared with stock price return series ri and trading volume variation series vi , R
variation series not only remain the characteristics of original series but also demonstrate the relative correlation between
stock price and trading volume. Finally, we analyze the multifractal characteristics of R variation series before and after
three financial events in China (namely, Price Limits, Reform of Non-tradable Shares and financial crisis in 2008) in the
whole period of sample to study the changes of stock market fluctuation and financial risk. It is found that the empirical
results verified the validity of R.
However, the potential of multifractal analysis is far from being fully exploited, since only very recently has attention been
drawn to the need for a thorough testing of the multifractal tools. How to deeply understand the essence of the multifractal
characteristics in capital markets and how to reveal more valuable information about market changes are two key issues in
the future. Study on these issues will help to make more accurate estimates for risk prevention and control.

Acknowledgments

We sincerely thank our anonymous referee. We also thank Zhou W.X. and Wei Y. for comments and discussion on this
article. We are appreciative for financial support by National Science Foundation of China (Nos 70901017, 70101022),
Science Foundation of Postdoctors in China (No. 20080441095), Special Science Foundation of Postdoctors in China
(200902546) and Fundamental Research Funds for the Central Universities (N100406003).

References

[1] E.F. Fama, Efficient capital markets: a review of theory and empirical work, Journal of Finance 25 (1970) 383–417.
[2] E. Panas, Estimating fractal dimension using stable distributions and exploring long memory through ARFIMA models in Athens stock exchange,
Applied Financial Economics 11 (4) (2001) 395–402.
[3] D.O. Cajueiro, B.M. Tabak, Evidence of long range dependence in Asian equity markets: the role of liquidity and market restriction, Physica A 342
(2004) 656–664.
[4] B. Podobnik, D. Fu, T. Jagric, I. Grosse, et al., Fractionally integrated process for transition economics, Physica A 362 (2006) 465–470.
[5] B.B. Mandelbrot, The variation of certain speculative prices, Journal of Business 36 (1963) 394–419.
[6] R.N. Mantegna, H.E. Stanley, Scaling behavior in the dynamics of an economic index, Nature 376 (1995) 46–49.
[7] J.A. Skjeltorp, Scaling in the Norwegian stock market, Physica A 283 (3–4) (2000) 486–528.
[8] B.H. Wang, P.M. Hui, The distribution and scaling of fluctuations for Hang Seng index in Hong Kong stock market, European Physical Journal B 20 (4)
(2001) 573–579.
 Y. Yuan et al. / Physica A 391 (2012) 3484–3495 3495

[9] B. Podobnik, H.E. Stanley, Detrended cross-correlation analysis: a new method for analyzing two non-stationary time series, Physical Review Letters
100 (2008) 084102.
[10] W.X. Zhou, Multifractal detrended cross-correlation analysis for two nonstationary signals, Physical Review E 77 (2008) 066211.
[11] X.T. Zhuang, X.Y. Huang, Y.L. Sha, Research on the fractal structure in the Chinese stock market, Physica A 333 (2004) 293–305.
[12] B.B. Mandelbrot, A. Fisher, L. Calvet, A multifractal model of asset returns, Yale University, Working paper, 1997.
[13] R. Cont, Empirical properties of asset returns: stylized facts and statistical issues, Quantitative Finance 1 (2001) 223–236.
[14] L. Liu, Y.D. Wang, J.Q. Wan, Analysis of efficient for Shenzhen stock market: evidence from the source of multifractality, International Review of Finance
Analysis 19 (2010) 237–241.
[15] Y. Yuan, X.T. Zhuang, Multifractal description of stock price index fluctuation using a quadratic function fitting, Physica A 387 (2008) 511–518.
[16] P. Norouzzadeh, G.R. Jafari, Application of multifractal measures to Tehran price index, Physica A 356 (2–4) (2005) 609–627.
[17] K.E. Lee, J.W. Lee, Probability distribution function and multiscaling properties in the Korean stock market, Physica A 383 (1) (2007) 65–70.
[18] D.O. Cajueiro, B.M. Tabak, Long-range dependence and multifractality in the term structure of LIBOR interest rates, Physica A 373 (2007) 603–614.
[19] P. Norouzzadeh, B. Rahmani, A multifractal detrended fluctuation description of Iranian rial–US dollar exchange rate, Physica A 367 (2006) 328–336.
[20] K.E. Lee, J.W. Lee, Probability distribution function and multiscaling properties in the Korean stock market, Physica A 383 (1) (2007) 65–70.
[21] G. Lim, S.Y. Kim, H. Lee, K. Kim, D.I. Lee, Multifractal detrended fluctuation analysis of derivative and spot markets, Physica A 386 (2007) 259–266.
[22] J. Karuppiah, C.A. Los, Wavelet multiresolution analysis of high-frequency Asian FX rates, summer 1997, International Review of Financial Analysis 14
(2005) 211–246.
[23] D.S. Ho, C.K. Lee, Scaling characteristics in the Taiwan stock market, Physica A 332 (2) (2003) 448–460.
[24] X. Sun, H.P. Chen, Z.Q. Wu, et al., Multifractal analysis of Hang Seng index in Hong Kong stock market, Physica A 291 (1–4) (2001) 553–562.
[25] H.P. Chen, X. Sun, Z.Q. Wu, et al., Enlightenment from various conditional probabilities about Hang Seng index in Hong Kong stock market, Physica A
335 (1–2) (2004) 183–196.
[26] Y. Wei, P. Wang, Forecasting volatility of SSEC in Chinese stock market using multifractal analysis, Physica A 387 (2008) 1585–1592.
[27] Y. Yuan, X.T. Zhuang, X. Jin, Measuring multifractality of stock price fluctuation using multifractal detrended fluctuation analysis, Physica A 388
2189–2197.
[28] J.W. Kantelhardt, S.A. Zschiegner, E. Koscielny-Bunde, et al., Multifractal detrended fluctuation analysis of nonstationary time series, Physica A 316
(1–4) (2002) 87–114.
[29] G.H. Mu, W.X. Zhou, W. Chen, K. Janos, Long-term correlations and multifractality in trading volumes for Chinese stocks, Physics Procedia (2010)
1631–1640.
[30] B. Podobnik, D. Horvatic, A.M. Petersen, et al., Cross-correlations between volume change and price change, The Proceedings of the National Academy
of Sciences of the USA 106 (2009) 22079–22084.
[31] T. Guhr, B. Kalber, A new method to estimate the noise in financial correlation matrices, Journal of Physics A 36 (2003) 3009–3032.
[32] L. Laloux, P. Cizeau, J.-Ph. Bouchaud, M. Potters, Noise dressing of financial correlation matrices, Physical Review Letters 83 (1999) 1467–1470.
[33] V. Plerou, P. Gopikrishnan, B. Rosenow, L.A.N. Amaral, H.E. Stanley, Universal and non-universal properties of cross-correlations in financial time
series, Physical Review Letters 83 (1999) 1471–1474.
[34] B. Podobnik, D. Wang, D. Horvatic, I. Grosse, H.E. Stanley, Time-lag cross-correlations in collective phenomena, Europhysics Letters (EPL) 90 (2010)
68001.
[35] S. Arianos, A. Carbone, Cross-correlation of long-range correlated series, Journal of Statistical Mechanism—Theory and Experiment P03037 (2009).
[36] L.Y. He, S.P. Chen, Multifractal detrended cross-correlation analysis of agricultural futures markets, Chaos, Solitons & Fractals 44 (2011) 355–361.
[37] T. Qiu, L.X. Zhong, G. Chen, X.R. Wu, Statistical properties of trading volume of Chinese stocks, Physica A 388 (2009) 2427–2434.
[38] Y.D. Wang, Y. Wei, C.F. Wu, Cross-correlations between Chinese A-share and B-share markets, Physica A 389 (2010) 5468–5478.
[39] S. Ma, Research on Risk Measures in Financial Markets Based on Multifractal Theory, University of Heilongjiang, Ha erbin, 2009 (in Chinese).