...

A Review of High-frequency Financial Data Based on Characteristics in China

by user

on
Category: Documents
6

views

Report

Comments

Transcript

A Review of High-frequency Financial Data Based on Characteristics in China
EASTERN ACADEMIC FORUM
A Review of High-frequency Financial Data Based on Characteristics
in China
JIANG Xiangcheng, YANG Yedi
School of Business, Hohai University, Nanjing, Jiangsu, China
Abstract: With the growing risks of financial participation, high-frequency time series becomes more
and more important in determining the financial risks. This essay studied the current situation of the
characteristics of high-frequency financial data from four aspects, including the study of sampling
frequency, volatility, price discovery and liquidity. And then put forward some problems in present
research and advanced some outlooks to provide some help in future.
Keywords: High-frequency data, Characteristics, Current situation, Problems, Outlooks, China
The high-frequency data was used in physics for the study of natural phenomena originally, such as
analysis of sound waves, hydrologic time series and so on. In recent years, with the rapid development
of financial market, financial participation risks increase. High-frequency data was extended to research
financial time series gradually, to seek the law of financial market, so as to make the right prediction and
lower the investment risks in the future.
The research of High-frequency data is of great significance in both theory and practice. First of all, in
the operation of financial market, high-frequency data contains more abundant information resources
than low-frequency data. As we all know, the key factor deciding the success or failure of the investment
is the information obtained. In the financial market, information is continuous, discrete collection may
cause information loss in varying degrees. The lower sampling frequency is, the more the information
lost. So high-frequency data can reflect the real situation of financial markets better than low-frequency
data. Secondly, high-frequency financial data research can supply the gaps and overcome the defects of
low-frequency data research models, even can coordinate with the improved low-frequency time series
models mutually to study the operation of financial markets jointly. Finally, the studies of highfrequency financial data provide the possibility to describe the micro market movement from the angle
of quantification.
With the study of high-frequency financial data characteristics, we can know the working of financial
markets better. The performance of high-frequency financial data will be affected by the sampling
frequency. And volatility and price discovery can reflect its features. In addition, we could find out
different liquidity in different financial market by studying it. China’s studies in this field start late, but
grow fast. Moreover, they pay more attention on volatility and price discovery of high-frequency
financial time series.
1 Sampling Frequency
High-frequency financial time series is essentially a form of discrete data, and the discretization depends
on the sampling frequency. As is known to all, different sampling frequency will show different features.
What’s more, using high-frequency financial data needs longer operation time and higher collection
costs, so searching the suitable sampling frequency, studying different features under different sampling
frequency is of great importance for us to obtain accurate information resources.
Researchers in other countries study the sampling frequency of high-frequency financial data earlier,
such as Yacine et al. (2005) believed that the higher the sampling frequency is, the less the information

Corresponding author. Tel: 13851475475 (China)
Email address: [email protected]
Current address: Hohai University, Jiangning District, Nanjing, China, 211100.
3
EASTERN ACADEMIC FORUM
lose, so the higher, the better. Recently, some scholars go deep, for example, the research by Cleiton
(2011) using Nyquist-shannon sampling theory proposed by Shannon in1949, concluded that we could
find out the optimal sampling frequency by using Spectral Analysis.
However, China starts later, and fewer studies study sampling frequency of high-frequency financial
time series. But more and more people begin to realize the importance of sampling frequency studies
and pay close attention to them. As to the sampling frequency, some scholars consider the higher the
better, notable examples are the studies by Wang et al. (2010), who use Shanghai composite index and
Standard & Poor's 500 Index, calculate and use SPA methods to test the prediction accuracy of each
volatility models with bootstrap characteristics on the basis of applying different sampling frequency of
high-frequency financial data, conclude that high-frequency data volatility models are better than lowfrequency data volatility models to forecast volatility of financial market. Nevertheless, some scholars
don’t think so. Higher sampling frequency may not improve the prediction accuracy of financial risk
well, as in Miao (2011).
But, for the past years, almost all such studies focus on the analysis of operation results between highfrequency data and low-frequency data, and discuss the direct relations between sampling frequency and
prediction accuracy. There is no research studying the difference of high-frequency financial data
between different high sampling frequencies and cause, such as different characteristics between highfrequency financial time series with sampling frequencies in 1 minute, 5 minutes, 10 minutes, and 15
minutes, calculating and making comparisons between different calculation results.
2 Volatility
Volatility is one of the most important characteristics of high-frequency financial data. In recent years,
with the deepening of the reform in financial market, China's financial investors and institutions
strengthen their risk management consciousness constantly. Handling the volatility of high frequency
financial data, selecting the appropriate volatility models to discuss it, exploring its regularity, are of
increasingly importance to make accurate estimation of the volatility and prediction of financial assets.
2.1 The methods to deal with volatility
In the studies of high-frequency financial time series, foreign researchers do a lot in the processing of
volatility, for example, Lacopo et al. (2009) used improved Fourier Transform, Yuan et al. (2002) used
The Bayesian Spectrum Estimation.
In addition to remove the special value simply, some scholars take logarithm on the sample data, using
ADF and KPSS method to test the yield sequence, as in Zeng (2007). Some use Bayesian Spectrum
Estimation who believe this method is suitable for random characteristics of high-frequency financial
data and could give a more realistic conclusion, as in Li (2010). Also some scholars consider WRBV
method is a good way to solve the jump point problem and eliminate the calendar effect of stock market,
as in Fu (2013).
2.2 Models of volatility
With the early work by Engle’s ARCH model, Bollerslev’s GARCH model and Taylor’s Stochastic
Volatility Model, GARCH class models and stochastic volatility model stand their leading position in
the forecast of volatility gradually. And then the ACD models proposed by Engle and Russell were
applied to the study of volatility of high frequency financial time series gradually. Recently, the
shortened realized variance (TRV) is used to study the noise in high-frequency financial time series
(Mancini, 2013). Realized volatility models (Chaboud, 2010; Degiannakis, 2013) and ACD model
(Allen, 2013) are used to seek the volatility of high-frequency financial data.
In the past few years, most of the studies in China concentrate on the prediction and calculation of the
volatility using different models. As to the models of volatility, several improvements were made to
some low-frequency data models, including three categories mainly, one is the improved traditional
4
EASTERN ACADEMIC FORUM
low-frequency data models, and the second is the improved high-frequency data models, such as ACD
model, another is the improved models in the field of physics, computer and others.
2.2.1 Improved low-frequency financial data models
As far as the volatility models of High-frequency financial data is concerned, recently, most scholars in
China pay more attention to applying the improved traditional low-frequency data models to highfrequency financial data series to study the financial market volatility. Moreover, one of the improved
low-frequency data models, used more, is the expansion of the ARCH class models, including strong
GARCH class models, weak ARCH class models. Another is the improved SV model. Improved
GARCH models are used to study the volatility of high-frequency financial time series often (Xiong et
al., 2012) and improved SV models, such as realized volatility of heterogeneous autoregressive model is
used to measure the volatility of China's stock index futures returns (Wen et al., 2012). More
traditionally, after stationary test, Tang et al. (2012) examined the relationships between jump, noise,
and liquidity by using Granger causality test, and then analyzed the direct positive and negative
relationships between each factors through the regression analysis. The results of their research reflected
that noise was caused by some specific information, which would have a great impact on the volatility.
As for the performance of improved low-frequency models in measuring the volatility of high-frequency
financial time series, conclusions are various. Some studies show that low-frequency data models
perform well, while others don’t. Let’s take GARCH class models for example. Some researchers
believe GARCH class models could reveal the volatility clustering characteristic of high frequency
financial time series well (Xie et al., 2012), but some do think their forecasting ability is poor (Wei,
2012).
2.2.2 Improved high-frequency financial data models
ACD models and improved ACD models which are put some microstructure variables in are considered
to describe the dynamics correlation of the volatility jump phase, such as Liu et al. (2012) constructed
ACD-GARCH-Model to measure and analyze the volatility of high-frequency financial time series, and
discussed the direct relationships between volume, duration, the rate of return and volatility. They
expanded the use of ACD model, and thought the improved model reacted well in the measurement of
the volatility clustering of high-frequency financial data. And also some other models used to study the
jump duration of high-frequency financial data, such as ACH model (Yang et al., 2011) and its extension
model, extreme POT model (Gui et al., 2010).
2.2.3 Improved models of other fields
Over the years, more and more scholars apply multifractal theory to the field of high-frequency financial
time series. They put forward and use multifractal volatility to measure the volatility of financial market.
Some scholars compare the performance of multifractal theory in high-frequency financial time series to
low-frequency data improved models, and demonstrate its advantages, for example, papers by Wei
(2012) argue that multifractal volatility model is more suitable than improved low-frequency data
models in the measurement of the high-frequency financial time series’ volatility.
Neural network model which is generally used in Computer Science, is also improved and used in the
study of high frequency financial data. The improved neural network model is thought to perform well
in the description of volatility of high frequency financial time series, such as Xu et al. (2007),
combined the neural network model with the wavelet cross-correlation model, and proposed the Wavelet
Neural Network (WNN) Model for quantitative study of volatility in high-frequency data’s "calendar
effect" problem, and show that WNN model is able to depict the "calendar effect" well.
3 Price Discovery
Price discovery is an important aspect of high frequency financial time series features. The price
discovery of high-frequency financial data mainly refers to the forecast of future prices reflected by the
current price, as well as the price transport phenomena in the high-frequency financial time series.
Price discovery of high-frequency financial data is studied by many scholars in foreign countries. Some
5
EASTERN ACADEMIC FORUM
research shows there is no cointegration relation between futures and spot (Chevallier, 2010), but some
research make a different conclusion, manifesting that there are specific evidences showing that futures
market play an important role in price leadership, and we could make the prediction of the prices in the
spot market through the prices in futures market, there is a cointegration relationship between the spot
and futures.
Price discovery is one of the main research directions of high-frequency financial time series’
characteristics. Most of the research focuses on the leading relationship between futures and spot goods
prices and which has the leading role in the market. Different conclusion appears in different scholars’
research. As to the leading relationship between the prices of futures and spot goods as well as it’s
leading function, some learned men believe that spot goods plays a leading role in the price discovery,
such as Yan et al. (2009), also some scholars consider that the futures price run ahead of the spots index,
such as Ren (2010), in addition, a part of scholars argue that the prices of spots and futures lead each
other, such as Peng (2010).
As for the cause of price discovery, in other counties, the degree of information sharing is often thought
to be a significant cause that will influence the speed of price discovery. Additionally, the sum number
of information sharing is the signal of the price discovery speed. The more the information shares, the
faster the speed of price discovery is, as Frijins (2009). Similar results are drawn in China. Research
indicates that information will affect the function of price discovery in high frequency financial data. In
addition, information asymmetry and the existence of private information are important reasons that will
lead to abnormal relationships between prices and reversal effect, and will affect the prediction function
of price discovery, such as Zhu (2011).
As far as the models of Price discovery is concerned, in the field of high-frequency financial data,
expect the traditional analysis method of vector error correction model, the newly applied methods, such
as the wavelet time-frequency cross-correlation analysis method, not only pays attention to time
dimension, but also tests the dynamic correlation of time series from the view of the frequency domain
and time domain. It provides a new perspective for the dynamic causal analysis in high-frequency
financial time series.
4 Liquidity
After the financial crisis in the year of 2007, if the liquidity height of financial market can remind
investors to determine their ability of resisting unpredictable risks has caused great attention among
researchers. High-frequency financial data is thought to be more effective in the research of liquidity in
financial markets. Foreign people, Riordan (2012) thinks in a market, the increased waiting time of most
small and medium-sized stocks can lead to an increase in liquidity.
The liquidity of financial markets is studied a lot in China through researching high-frequency financial
time series. As far as how the block trade affects the liquidity of financial markets is concerned, Feng et
al. (2008) confirmed the asymmetry phenomenon behind block trade through studying the frequency
data with different frequency in the different regions, countries. He maintained that the liquidity was
poorer in the often city than in the bear market. But this scholar made a totally different conclusion in
2012. This time, his study showed that liquidity was higher in the often city significantly than in the
very city (bull market and bear market) environment. Investigating its reasons, maybe the highfrequency financial data intervals are different, as well as the sample size was different, and the methods
have been improved. Also the factors that influence liquidity are researched. Most researchers agree that
information affects the liquidity of securities, so we could improve the information owning rate and
reduce the investment risks by enhancing financial market liquidity.
6
EASTERN ACADEMIC FORUM
5 Problems and Outlooks
At present, there is gradually growing research about high-frequency financial data, more and more
scholars come to realize the importance of research in high-frequency financial data. It is necessary to
seek the operation rule of financial markets, reduce risks, and promote the health of the financial market
operation. But there are still many problems to be effective solved, still a long way to go.
5.1 Problems
Although the studies of characteristics in high-frequency financial time series develop fast, there are still
a lot of problems to be solved.
To begin with, sampling frequency research of high-frequency financial data is relative insufficient, it
also can be understood as the lake of the understanding to the process of high-frequency time series.
What are the differences under different sampling frequency, and what is the optimal sampling
frequency, which model is suitable for high-frequency financial time series with a certain kind of
sampling frequency. All of these need to be further explored.
In addition, the handling methods of volatility need to be further deepened. Ignoring special value is
Chinese traditional way to deal with volatility. Few scholars use Bayesian Method and other methods to
deal with high frequency financial data. Ignoring the special value or exploring other ways should be
further researched.
What’s more, volatility models need more studies. Different models have different features and scope of
application. Further research should not only include the suitable models, for example, using improved
low-frequency models or high-frequency models, but also should consider the options about variables,
such as volume, price, time and so on. Of course, we also need to compare the application ranges of
different models. Personally, to study the characteristics of high-frequency financial time series, future
study should enhance the research about specific high-frequency financial data models, rather than just
improve the low-frequency financial data models, such as GARCH model, SV model, and apply them to
the field of high-frequency time series.
Finally, we should go further through the nonlinear way, because high-frequency financial data is a kind
of nonlinear data basically. There are already some scholars in foreign countries nonlinear way to study
its characteristics, viewing high-frequency financial data from the point view of chaos, but there are
almost none essay studying it from this way in China.
5.2 A few directions about future study
Firstly, it is an important issue to research the sampling frequency about high-frequency financial time
series. With different sampling frequency, the nature of the financial data is different. So it is important
to study the sampling frequency, find the suitable models, and locate the optimal frequency.
Secondly, the handling measurement of particular value in high frequency financial time series needs
further research. Everybody knows, the treatment of data is as important as selection, and different
treating ways will produce different results. For this reason, we should seek the most suitable handling
way after studying sampling frequency.
Thirdly, further deepen the studies of volatility models, including the following aspects:
(1) Clear the applicability of the volatility models. We should clear if the models are suitable for using
no matter they are low-frequency date models or high-frequency data models. We should study how to
select the appropriate variables in the models, such as first use a model, selecting the same data sources,
the same sampling frequency, then add another variable, compare the differences between the two
models’ results and prove the cause about the differences, judge whether the variables can be added to
the model.
(2) Strengthen the study of continuous time models. The observed data in the reality is discrete, and
discrete models are widely used. However, future studies could go on from the view of continuous time
models. High-frequency financial time series, especially the ultrahigh frequency financial time series, is
7
EASTERN ACADEMIC FORUM
more similar to continuous time, so from the view of continuous time series models to study this
problem is a worthy of further method.
(3) Improve models in other fields. Find applicable models in other areas, improve them, and then
applied into high-frequency financial time series, not only including improve the current studies in
Neural Network Model, multifractal model, Time-varying wavelet model and so on, but also including
seek new models, for example, seek the chaotic characteristics of high frequency financial data, and
make models.
(4) As far as the studies of ultrahigh-frequency financial time series is concerned, since Engle put
forward ACD model, there are still a lot of problems to be further discussed, such as persistent problem
of the trading interval, stationary ergodic problem of ACD model, and variable structure of ACD model
and so on.
Finally, study the characteristics of high frequency financial data from the nonlinear angle and its
nonlinear models. Fundamentally speaking, high-frequency financial data have more of the nonlinear
characteristics, so it deserves to be studied from the nonlinear perspective, and seek appropriate
nonlinear models.
6 Conclusion
High-frequency financial data contains more information resources than low-frequency financial data,
so it is used to make prediction in financial market gradually. For the past few years, more and more
scholars study the characteristics of high-frequency financial data to make more accurate predictions.
This essay fist analyzes the current situation of studies on the characteristics of high-frequency financial
data, including sampling frequency, volatility (we analyze the treatment of volatility and volatility
models in this part), price discovery and liquidity. We could see that although China develops rapidly in
the studies of this field, there are still a large number of problems. Then we put forward some problems
and a few directions so as to make some help in the future.
References
[1]. Yacine Aï
t-Sachalia, Per A. Mykland, Lan Zhang. How Often to Sample a Continuous-Time
Process in the Presence of Market Microstructure Noise. I Review of Financial Studies, 2005, 18, 2,
351 – 416.
[2]. Cleiton Taufemback, Sergio Da Silva. Spectral Analysis Informs the Proper Frequency in the
Sampling of Financial Time Series Data. Physica A. 2011, 390, 11, 2067 – 2073.
[3]. C.E. Shannon. Communication in the Presence of Noise. Proceedings of the Institute of Radio
Engineers, 1949, 37, 1, 10.
[4]. Xiaoyu Miao. Measuring Financial Risks Using Different Frequency Data—Base on Ultra-high
frequency, High Frequency and Low Frequency Data Model. Journal of Shanxi Finance and
Economics University, 2011, 33, 2, 30 – 37.
[5]. Peng Wang, Yu Wei. Comparison of Forecasting Ability of Volatility Models with Different
Frequency. Chinese Journal of Management, 2010, 7, 8, 1258 – 1262.
[6]. Iacopo Giampaoli, Wing Lon Ng, Nick Constantinou. Analysis of Ultra-high-frequency Financial
Data Using Advanced Fourier Transforms. Finance Research Letters, 2009, 6, 1, 47 – 53.
[7]. Yuan Qi, Thomas P. Minka, Rosalind W. Picara. Bayesian Spectum Estimation of Unevenly
Sampled. Acoustics, Speech, and Signal Processing (ICASSP), 2002 IEEE International
Conference, 2002, 2, 1473 – 1476.
[8]. Yinqiu Zeng. Analysis on Long- term Memory in Petroleum Futures Price Returns and Volatilities.
Journal of Graduates Sun Yat-SEN University, 2007, 28, 2, 76 – 102.
[9]. Sufang Li, Huiming Zhu, Keming Yu, Liya Hao. Ultra High-frequency Financial Data
8
EASTERN ACADEMIC FORUM
Co-integration Relation Research Based on Gibbs Sampling of Bayesian. Statistics & Information
Forum, 2010, 25, 10, 40 – 44.
[10]. Qiang Fu, Xili Wu. Risk Research for Chinese Stock Market Based on ARFIMA-WRBV-VAR
Model. Journal of Southwest University, 2013, 35, 9, 9 – 14.
[11]. Robert F. Engle. Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of
United Kingdom Inflation. Econometrica 50, 4, 987 – 1008
[12]. Tim Bollerslev. Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics,
1986, 31, 3, 307 – 327.
[13]. Taylor SJ. Modelling Financial Time Series. World Scientific Publishing, 2008.
[14]. Robert F. Engle, Jeffrey R. Russell. Autoregressive Conditional Duration: A New Model for
Irregular Spaced Transaction Data, 1998, 66, 5, 1127 – 1162.
[15]. Cecilia Mancini. Measuring the Relevance of the Microstructure Noise in Financial Data, 2013,
123, 7, 2728 – 2751
[16]. Alain P. Chaboud, Benjamin Chiquoine, Erik Hjalmarsson, Mico Loretan. Frequency of
Observation and the Estimation of Integrated Colatility in Deep and Liquid Financial Markets.
Journal of Empirical Finance, 2010, 17, 2, 212 – 240.
[17]. Stavros Degiannakis, Christos Floros. Modeling CAC40 Volatility Using Ultra-high Frequency
Data. Research in International Business and Finance, 2013, 28, 68-81.
[18]. David Allen, K.H.Ng, Shelton Peiris. The Efficient Modeling of High Frequency Transaction Data:
A new Application of Estimating Functions in Financial Economics. Economics Letters, 2013, 120,
117-122.
[19]. Jian Xiong, Li Wang. Risk Analysis of China’s Stock Market Based on t-Copula-GARCH-t.
Journal of Guangzhou University, 2012, 11, 5, 1 – 5.
[20]. Rui Long, Chi Xie, Zhi Jian, Changqing Luo. Measurement of CSI 300 Stock Index Futures
Volatility Under High-frequency Environment — Methods Based on Realized Volatility and Its
Modification. Systems Engineering — Theory & Practice, 2011, 31, 5, 813 – 822.
[21]. Fenghua Wen, Xiaoqun Liu, Hairu Tang, Xiaoguang Yang. Research on China’s Stock Market
Fluctuations Based on LHAR-RV-V Model. Journal of Management Science in China, 2012, 15, 6,
59 – 67.
[22]. Yong Tang, Guiming Kou. Analysis on the Relationship among Microstructure Noise, Jump and
Liquidity in Stock Market. Chinese Journal of Management Science, 2012, 20, 2, 11 – 19.
[23]. Jun Xie, Chunpeng Yang, Wei Yan. The Impact of Investor Sentiment on Stock Index Futures
Market under High-frequency Environment. Systems Engineering, 2012, 30, 9, 27 – 36.
[24]. Yu Wei. Volatility Forcasting Models for CSI 300 Index Futures. Journal of Management Sciences
In China, 2012, 13, 2, 66 – 76.
[25]. Xiangli Liu, Siwei Cheng, Shouyang Wang, Yongmiao Hong. Volatility of Chinese Futures Market
Based on ACD Model. Systems Engineering – Theory & Practice 32, 2, 268 – 273.
[26]. Ke Yang, Langnan Chen. Empirical Research on Jump Behavior of High-frequency Volatility
Based on C_TMPV in Chinese Stock Market. Journal of Management Science, 2011, 24, 2, 103 –
112.
[27]. Wenlin Gui, Zhaozhou Han, Nianqing Pan. Tail Thickness of GPD in POT Model and Finance Risk
Measurement. The Journal of Quantitative & Technical Economics, 2010, 1, 1, 107 – 118.
[28]. Zhengguo Xu, Shiying Zhang. High-frequency Financial Data Calendar Effects’ Wavelet Neural
Network Analysis. Mathem Atics in Practice and Theory, 2007, 37, 15, 1 – 6.
[29]. Jing Gu, Delei Ye. Research of Stock Index Futures Price Discovery Function: A Literature Review.
Hainan Finance, 2013, 219, 2, 41 – 56.
[30]. Chevallier, Julien. A Note on Cointegrating and Vector Autoregressive Relationships between CO2
Allowances Spot and Futures Prices. Economics Bulletin, 2010, 30, 2, 1564 – 1584.
[31]. Daniel Rittler. Price Discovery and Volatility Spillovers in the European Union Emissions Trading
Scheme: A High-frequency Analysis. Journal of Banking & Finance, 2012, 36, 3, 774 – 785.
9
EASTERN ACADEMIC FORUM
[32]. Min Yan, Shusong Ba. Price Discovery and Volatility Spillovers of Stock Index Futures Markets in
China. Systems Engineering, 2009, 27, 10, 32 – 38.
[33]. Yuan Ren. The Lead Lag Relationship Between Stock Index Futures and the Spot Index. Securities
Futures of China, 2010, 7, 24 – 28.
[34]. Ziyun Peng. An Empirical Study of Relationship Between China's Stock Index Futures and
Shanghai & Shenzhen 300 Index. China Business, 2010, 6, 1 – 7.
[35]. Bart Frijins, Peter Schotman. Price Discovery in Tick Time. J. Empirical. Finance, 2009, 16, 5, 759
– 776.
[36]. Xi Zhu. To Discuss the Form Mechanism of Momentum and Contrarian Anomalies in Shanghai
and Shenzhen A-share Stock Markets: From Hasbrouck Model. Research of Shanghai Finance,
2011, 6, 1, 51 – 58.
[37]. Xiangli Liu, Yumeng Zhang. Price Discovery Function of the Stock Index Futures on the Basis of
VECM. Management Review, 2012, 24, 2, 71 – 77.
[38]. Zhi Su, Yanglong Chen. Research on Efficiency of Stock Index Futures’ Price Discovery Based on
Morlet Wavelet. The Journal of Quantitative & Technical Economics, 2012, 6, 140 – 151.
[39]. Jianfen Feng, Maobin Wang. Big Deals with Asymmetric Market Impact. Securities Market Herald,
2008, 11, 17 – 24.
[40]. Jianfen Feng, Maobin Wang. Block Trades and Asymmetrical Market Impact in Different Market
Situations. Securities Market Herald, 2012, 6, 44 – 51.
[41]. Xiangcun Xu, Ping Li, Yong Zeng. The Impact of Increase in the Transparency of Call Auction on
Market Liquidity: Evidence from the Opening in China Stock Markets. Chinese Journal of
Management, 2010, 7, 1, 123 – 130.
[42]. Fu Lu, Renhai Hua. Research on Collaborative Herding Behavior and Market Volatility: Based on
Computation Experiments. Journal of Management Sciences in China, 2010, 13, 9, 98 – 106.
[43]. Hui Chen, Qianyuan Wang. How does Institutional Investor Affect Stock Liquidity? Trading
Hypotheses or Information Hypotheses. Journal of Business Economics, 2012, 248, 6, 71 – 80.
10
Fly UP