What drives the volume–volatility relationship on Euronext Paris?

https://doi.org/10.1016/j.irfa.2011.03.001Get rights and content

Abstract

The goal of this paper is to shed light on the relationship between volume and volatility. More specifically, it aims to determine which component of trading volume (trade size or number of transactions) drives this relation. Our intraday analysis reveals several results. Firstly, we confirm the strong positive relationship between volume and volatility. Secondly, including volume in the conditional variance of stock returns significantly reduces the persistence of volatility. Thirdly, we show that the well-known positive relationship between volatility and volume is generated by the number of trades. These results are robust, even after controlling for the impact of the intraday patterns. Finally, our findings are available for the CAC40 Index as well as for individual stocks.

Introduction

A Wall Street adage says “It takes volume to make prices move.” This saying is confirmed by several empirical studies documented in the survey paper of Karpoff (1987). The author provides an interesting literature review of papers studying the relationship between price changes (measured as absolute or squared price changes) and volume. Earlier studies cited by Karpoff (1987) and recent works (Chuang et al., 2009, Giot et al., 2010, Malinova and Park, 2010) conclude that there is a strong positive correlation between volatility and trading volume. This means that volatile returns are associated with high trading volume.

Various microstructure models have attempted to provide theoretical justification for the well-known positive relationship between price changes and trading volume. The competing explanations are the mixture of distribution hypothesis and the asymmetric information hypothesis.

The seminal work of Clark (1973) has introduced the mixture of distribution hypothesis (MDH), which supposes that asset price changes are driven by information. As documented by Chen and Daigler (2008), this hypothesis was extended in recent works to highlight a strong relationship between information flow and market activity. The MDH consider information flow as a latent common factor that affects both of trading volume and stock prices. Thus, price changes and trading volume may be correlated as they depend jointly on the intensity of information flow (Li & Wu, 2006). Empirically, this means that volume and stock price react contemporaneously in response to information releases. In fact, the arrival of new information to the market induces a price adjustment process through the sequence of trades. The mixture distribution hypothesis has also been used to explain the well-known autoregressive conditional heteroskedasticity (ARCH) process that volatility follows. Lamoureux and Lastrapes (1990) use data from the US market and show that persistence in volatility is diminished when volume is introduced in the conditional variance equation of the GARCH model. This result is confirmed in the Korean market (Pyun, Lee, & Nam, 2000), in the Polish market (Bohl & Henke, 2003), in the Saudi market (Alsubaie & Najand, 2009) and in futures market (Pati & Rajib, 2010). These findings show that volume is driven by the same factors that generate the ARCH effects. In general, the mixture distribution hypothesis supports a strong, contemporaneous and positive relationship between volume and volatility.

Despite the interesting explanation given by the mixture distribution hypothesis, the models described above do not allow us to determine the component of trading volume that generates this relation. In fact, trading volume is composed of two components: the number of trades and size of trades. Thus, it would be interesting to test whether the volume–volatility relationship is driven by either one or both components. The asymmetric information hypothesis has focused on this issue. The microstructure literature distinguishes two groups of models: competitive asymmetric information models and strategic asymmetric information models. Competitive models suppose that informed investors prefer to trade large amounts and conclude that there is a positive relationship between price changes and trade size (Back and Baruch, 2007, Easley et al., 1997, Easley and O'Hara, 1987, Holthausen and Verrecchia, 1990, Ozsoylev and Takayama, 2010). Empirically this assumption leads to our first hypothesis:

Hypothesis 1

The well-known volume–volatility relationship is driven by the size of trades.

However, strategic models predict that informed traders may camouflage their private information by splitting large trades into several small trades (Cho, 2007, Chordia and Subrahmanyam, 2004, Foster and Viswanathan, 1996, Holden and Subrahmanyam, 1992, Kyle, 1985). Empirically this intuition leads to the following second hypothesis:

Hypothesis 2

The number of trades generates the well-known positive relationship between price changes and volume.

Despite the abundant literature on the volume–volatility relationship, few papers have focused on its origin. Jones, Kaul, and Lipson (1994) use daily data on NASDAQ-NMS firms and ordinary least squares (OLS) technique to test whether number of transactions per se or their size generates volatility. The authors show that the positive daily relationship between volatility and volume is due to the positive daily relationship between volatility and the number of transactions. Jones et al. (1994) conclude that the average size of trades has no incremental information content beyond that contained in the number of trades. Chan and Fong, 2006, Giot et al., 2010 use realized volatility and confirm the above result. Using data from the New York Stock Exchange (NYSE), Xu and Wu (1999) investigate the relationship between price changes, average trade size and the number of transactions. The authors highlight the information role of the frequency of transactions. However, contrary to Jones et al. (1994), they show that the average size of trades contains nontrivial information for return volatility. Chan and Fong (2000) consider a sample of stocks listed on the NASDAQ and New York Stock Exchange. Their findings are in line with those of Xu and Wu (1999). They support the significance of average trade size in the volume–volatility relationship in both markets. Huang and Masulis (2003) report that trade frequency and average trade size impact price volatility for small trades in London Stock Exchange. For large trades, the authors show that only trade frequency affects price volatility.

Our study is related to the empirical works mentioned above. It aims to test whether the number of trades or the size of trades drives the volume–volatility relationship on Euronext Paris. Our contributions concern the data used and the methodology adopted.

Although there are many empirical studies on the volatility–volume relation, there is no general consensus about what actually drives the relation. Moreover, most of the previous studies pertain to the U.S. market and it is unclear whether we can generalize its results to other markets. Euronext Paris is one of the most important markets in Europe. It is a pure automated order driven market, which has a specific microstructure that can impact the roles of the number of trades and size of trades in the volatility–volume relation. In fact, all investors can see at any time the five best limits at each side of the market, with the associated displayed depth. This transparency can incite investors to camouflage their large orders by splitting them. Market participants can also use hidden orders, i.e., orders whose some part of the quantity is not disclosed to other investors. In this case, the total order size is registered in the order book but only the disclosed quantity is displayed on the market screens. D'Hondt, De Winne, and François-Heude (2003), show that hidden orders are more involved in splitting strategies than usual orders. Hence, we expect that splitting orders may increase the information content of the number of trades on Euronext Paris. On the other hand, it may attenuate the relationship between the size of trades and volatility. To test this intuition, we have obtained intraday data that covers the period from January to December 2007. The intraday analysis allows us to avoid aggregating variables into daily sums, as proposed by Jones et al. (1994). The aggregation can smooth variables and affect their significance.

Our methodology differs in many aspects from the mentioned studies on the topic. We consider a conditional volatility measure instead of realized volatility (Chan and Fong, 2006, Giot et al., 2010) and absolute returns (Chan and Fong, 2000, Huang and Masulis, 2003, Jones et al., 1994). The use of a GARCH (general auto regressive conditional heteroskedasticity) model is appropriate for our study for two reasons. First, the family of ARCH models has been shown to provide a good fit for financial return time series (Bollerslev, 1987, Han and Park, 2008, Lamoureux and Lastrapes, 1990). In fact, the autoregressive process accounts for the persistence and for the clustering pattern of volatility. It captures some statistical artifacts in stock returns as the nonstability of the distributions documented by Mandelbrot, 1963, Fama, 1965. Second, the GARCH framework allows to test if returns are generated by a mixture of distributions, in which the trading volume is a stochastic mixing variable. Indeed, to study the interaction between volume and volatility, we introduce the trading volume in the conditional volatility equation. If the MDH hypothesis is validated, we expect that trading volume significantly influences the conditional volatility and reduces substantially its persistence. We apply our econometric model to 38 stocks listed on Euronext Paris and to the main Index of the French market (CAC40 Index). To test the robustness of our results, we control for the potential impact of the well-known intraday patterns.

Our research question is interesting and has several implications. First, it provides insight into the structure of financial markets. Indeed, this relation depends on the rate of information flow, information dissemination and the extent to which market prices convey the information (Karpoff, 1987). Second, this work tests if information flow is a latent common factor that affects both of trading volume and stock prices. In other words, we test if stock price movements and trading volume are influenced by the new information flow. The response to this question is important for event studies that use both returns and trading volume to investigate the market reaction around corporate disclosure (Louhichi, 2008, Miller, 2010). If price movements and volume depend jointly on the intensity of information flow, incorporating the price–volume relation will increase the power of these tests. Third, the results given by the decomposition of volume allow us to examine if the number of trades is a sufficient statistic for trading activity. Fourth, our findings can help investors to proxy the information flow. Fifth, our study is useful for market authorities as we shed light on the role of trade size and trade frequency as a signal of informed trading.

Our intraday analysis reveals several results. We confirm the strong positive relationship between volume and volatility. Moreover, including volume in the conditional variance of stock returns significantly reduces the persistence of conditional volatility. Furthermore, we highlight the fact that the average size of trades has no incremental information content beyond that contained in the number of trades. These results are robust even after controlling for the impact of the intraday patterns. Finally, our findings are available for the CAC40 Index as well as for individual stocks.

The remainder of the paper is organized as follows. Section 2 focuses on the integration of European exchanges and the creation of Euronext. Section 3 details the GARCH model and specifies the research methodology. Section 4 gives a description of the data and exposes the empirical results. The last section concludes.

Section snippets

European market integration and Euronext microstructure

In 1999, the euro1 was adopted as a common currency in the European Union. This event has played a very important catalyzing role for the financial integration process as documented by several studies. Arouri et al. (2010) focus on financial integration between 10 European stock markets during the period 1970–2007. Using a nonlinear model, the

The GARCH model

The goal of this paper is to shed light on the relationship between volume and volatility. More specifically, it aims to determine which component of trading volume (trade size or number of transactions) drives this relation. Several proxies have been used to measure volatility. Jones et al. (1994) use the absolute returns to calculate the realized volatility and set their analysis in the framework of ordinary least squares (OLS) technique to model the relationship between volume and

Data

The aim of this paper is to shed light on the relationship between trading volume and price volatility using high-frequency data from Euronext Paris. The study requires intraday data about trades, execution date and time, size, price, best limits and number of transactions. This information is obtained from Euronext database and covers the period from January to December 2007. During this period, Euronext Paris was open from 9:00 a.m. to 5:30 p.m. Transaction data in each day are divided into

Conclusion

The aim of this paper is to shed light on the relationship between return volatility and trading volume. Unlike the existing literature, we consider conditional volatility measure instead of realized volatility. We decompose trading volume into two components: the number of trades and the size of trades. Using intraday data from Euronext Paris, we test if the volume–volatility relationship is driven by one or both components. Our analysis gives several results. Firstly, we support the mixture

References (40)

  • H. Han et al.

    Time series properties of ARCH processes with persistent covariates

    Journal of Econometrics

    (2008)
  • R.D. Huang et al.

    Trading activity and stock price volatility: Evidence from the London Stock Exchange

    Journal of Empirical Finance

    (2003)
  • H.N. Ozsoylev et al.

    Price, trade size, and information revelation in multi-period securities markets

    Journal of Financial Markets

    (2010)
  • C. Pyun et al.

    Volatility and information flows in emerging equity markets: A case of the Korean stock exchange

    International Review of Financial Analysis

    (2000)
  • X.E. Xu et al.

    The intraday relation between return volatility, transactions, and volume

    International Review of Economics and Finance

    (1999)
  • M.H. Arouri et al.

    Stock market integration in the euro area: Segmentation or linear modelling misspecification?

    International Journal of Business

    (2010)
  • K. Back et al.

    Working orders in limit-order markets and floor exchanges

    Journal of Finance

    (2007)
  • B.M. Blau et al.

    Intraday stealth trading: Which trades move prices during periods of high volume?

    Journal of Financial Research

    (2009)
  • T. Bollerslev

    A conditionally heteroskedastic time series model for speculative prices and rates of return

    The Review of Economics and Statistics

    (1987)
  • Z. Chen et al.

    An examination of the complementary volume–volatility information theory

    Journal of Futures Markets

    (2008)
  • Cited by (27)

    • The impact of global financial crisis on informational efficiency: Evidence from price-volume relation in crude palm oil futures market

      2020, Journal of Commodity Markets
      Citation Excerpt :

      However, Liu et al. (2015) confirm that such a relation in the artificial stock markets is produced by trade size instead of information-driven trade. On the contrary, Stickel and Verrecchia (1994), Giot et al. (2010) and Louhichi (2011) state that informed traders can camouflage their private information by increasing their number of transactions. This can be done by splitting a large number of trades into several small numbers.

    • Trading intensity and the volume-volatility relationship on the Tunis Stock Exchange

      2018, Research in International Business and Finance
      Citation Excerpt :

      Nevertheless, a number of empirical studies (e.g., Bose and Rahman, 2015; Rahman et al., 2002) provides evidence for remaining volatility persistence even after incorporating the effect of volume. Although the positive volume-volatility relationship has been widely documented by empirical microstructure literature (Bohl and Henke, 2003; Louhichi, 2011; Orman and Mckenzie, 2000; Pyun et al., 2000; Rahman et al., 2002; Wang et al., 2005), total volume may appear as a coarse proxy for the latent information arrival (Bose and Rahman 2015; Clark, 1973; Li and Wu, 2006; Rahman et al., 2002; Slim and Dahmene, 2015). Alternatively, Jones et al. (1994) emphasize that the number of daily transactions, and not trade size, is the most appropriate measure of the rate of information flow.

    • Dynamic spillover effects in futures markets: UK and US evidence

      2016, International Review of Financial Analysis
    View all citing articles on Scopus

    I would like to thank Alexis Cellier for his help. Also, I thank the editor and the anonymous referees for their interesting comments. All remaining errors are mine.

    View full text