Volume shocks and stock returns: An alternative test

https://doi.org/10.1016/j.pacfin.2018.01.001Get rights and content

Highlights

  • There is a strong relationship between volume shocks and Australian equities returns.

  • The relationship is particularly stronger for firms that are less visible to investors.

  • Trading volume is higher after high volume shocks.

  • Both institutional and individual investors are responsible for the volume shock effect.

  • The volume shock effect is not priced.

Abstract

Using an alternative measure for abnormal trading volume, this article examines the role of volume shock in the generation of stock returns. We find a strong high volume effect at both portfolio and individual stock levels. A strategy that buys stocks experiencing high volume shocks and sells stocks experiencing low volume shocks generates positive returns up to 12 months after formation. The effect is robust after controlling for other stock characteristics that are known to affect stock returns. Our results show that trading volume becomes relatively higher after high volume shocks. Moreover, the relation between volume shocks and stock returns is stronger for stocks that previously failed to catch investors' attention. This finding is consistent with the view that abnormal trading volume proxies for unobserved attention-grabbing events. However, we find no evidence that volume shocks are priced.

Introduction

The relation between trading volume and price movements has long been the subject of academic study, resulting in a voluminous literature. Early studies focus on the contemporaneous relation between trading volume and price changes and find that trading volume is positively related with the absolute value of price changes (e.g., Karpoff, 1986; Gallant et al., 1992; Blume et al., 1994). More recent studies focus on whether there exists a relation between trading volume and expected returns. The seminal work of Gervais et al. (2001) documents that stocks with an unusually high trading volume over a day or a week (compared with their own trading volumes in the prior 50 days or 10 weeks) have higher subsequent returns than stocks with an unusually low trading volume. This finding results in a high volume return premium implying that abnormal trading volume contains information about future price movements. Gervais et al. (2001) interpret the high volume return premium following the investor recognition hypothesis originated by Miller (1977).

Abnormal trading volume can be driven by factors such as hedging, overconfidence, differences of opinion, liquidity, and the portfolio rebalancing needs of investors. However, Chuang and Lee (2006) and Hong and Stein (2007) argue that hedging, liquidity, and portfolio rebalancing needs are too small to account for the large trading activity observed in the markets. Thus, abnormal trading volume is more likely attributable to the result of differences of opinion or overconfidence (Glaser and Weber, 2007, Glaser and Weber, 2009). However, each explanation results in a different relation between volume shocks and subsequent stock returns. According to Gervais et al. (2001), a high volume shock induced by a high level of differences of opinion causes a stock to have higher subsequent returns after increased visibility.1 The high volume shock therefore acts as an attention-grabbing event that draws investors' attention to the stock (Miller, 1977). This view is also in line with recent emerging literature on investor attention and asset pricing dynamics (e.g., Hirshleifer and Teoh, 2003; Sims, 2003; Peng and Xiong, 2006; Hou et al., 2008) that uses trading volume as a proxy for investor attention. Conversely, as documented by Odean (1998), a high volume shock induced by overconfidence will lead to inferior subsequent returns as a result of excessive trading on risky securities.

Empirically, the positive relation between volume shocks and stock returns has also been found in a number of studies (e.g., Watkins, 2006; Barber and Odean, 2008; Huang and Heian, 2010; Kaniel et al., 2012). In markets outside the US, however, the findings are mixed. Kaniel et al. (2012), using the same method as Gervais et al. (2001) to define volume shocks, find that the relation is most pronounced in developed markets compared to emerging markets. Emerging markets exhibit weaker and inconsistent high volume premiums. Similar findings have also been documented by Huang et al. (2011), who find no evidence of high volume premiums (except for large firms in some countries) in six Asian markets, including Japan. The authors attribute this finding to the structural difference between US and Asian markets. Specifically, compared to Western financial markets, Asian markets are mostly dominated by individual investors who are believed to be less rational and to exhibit greater overconfidence and cognitive bias (Kim and Nofsinger, 2003; Kaniel et al., 2008). The inconsistency highlights that the role of volume shocks in stock returns is an empirical question and calls for further investigation for reasons put forward by Lo and MacKinlay (1990). The relation depends on what drives volume shocks and the extent of volume shocks presented in the market. The latter is strongly affected by the structure of the market.

This paper attempts to provide further evidence on the relation between volume shocks and stock returns using data from the Australian equities market. The paper makes the following contributions to the literature. First, we propose a measure of volume shocks in the spirit of Chen et al. (2001) and Bali et al. (2014). Specifically, trading volume in month t is classified as abnormally high if it is higher than the average trading volume of the prior n months. The difference is then scaled by the volatility of the trading volume over the prior n-month period.2 The advantage of using this measure is twofold. First, monthly trading volume is relatively easy to obtain compared to daily trading volume, especially in markets outside the US. It is also less noisy than daily or weekly data. This enables us to capture month-by-month variations in the volume of stocks and allows for the examination of the volume shock effect across a large number of stocks over a long period. Second, we aim to examine the longevity of the high volume effect, and monthly trading volume is also consistent with the frequency of the number of shares outstanding data available in this market. This approach allows us to conduct formal asset pricing tests on the role of volume shocks in the generation of stock returns.

Second, Kaniel et al. (2012) point out that the volume shock effect is related to a market's composition, investor demographics and confidence, and the extent of information dissemination. The Australian market is one of the developed capital markets and the second largest in Asia-Pacific based on free-float market capitalization. Similar to the US, the market is dominated by institutional investors. However, its composition is heavily skewed toward a large proportion of small and illiquid stocks, which are outside the investment universe of institutional investors. As of December 2013, over 70% of stocks in our sample were priced below AU$1, which is below the minimum price listing requirement for Nasdaq. Dyl et al. (2002), using Australian data, find that small and less well-known firms appear to attempt to increase their investor bases by having low stock prices, whereas large and more visible firms tend to have high stock prices. The market therefore has a small group of stocks that are well known to investors and a large group of stocks that are neglected by market participants, providing an interesting setting for studying the role of volume shocks in stock returns.3

Third, recent studies show that institutional and individual investors respond differently to return changes (e.g., Barber and Odean, 2008; Li et al., 2017) and the return-volume relation is found to be stronger when there is an increase in institutional ownership (Huang et al., 2011). Limited by data availability, prior studies either focus on a specific group of institutional investors, such as mutual funds, or rely on annual or quarterly institutional holdings, which may not fully reflect the dynamic trading of institutional investors. Using the monthly data of institutional holdings available in the Australian market, we further address how institutional trading affects the relation between equity returns and trading volume.

We find a strong relation between volume shocks and stock returns at both portfolio and individual stock levels. A strategy that buys stocks with high volume shocks and sells stocks with low volume shocks generates a risk-adjusted return exceeding 6% in the year after portfolio formation. However, the magnitude of the returns declines as the holding periods are extended. This finding is consistent with the view that attracted attention wears off over time (Barber and Odean, 2008), causing a decline in returns. At the individual stock level of analysis, the volume shock effect persists, even after controlling for other stock characteristics such as size, book-to-market, momentum, share turnover, idiosyncratic volatility, short-term reversals, beta, and demand for extreme positive returns. This implies that none of these firm characteristics, which are known to be important in explaining the cross-section of stock returns, account for the volume shock effect. Our results are also robust when we control for seasonality and price-sensitive announcements and are comparable across subperiods.

The evidence confirms the predictability of trading volume on future returns. Barber and Odean (2008) argue that investors purchase only stocks that have caught their attention and an increase in investor attention leads to temporary positive price pressure. In other words, potential investors have to be aware of a firm before they can become familiar with it and eventually decide to invest in it, implying that attention is a necessary condition for a firm to be publicly recognized. Accordingly, our evidence is in line with the notion that high volume shocks attract investor attention and subsequently increase a stock's visibility, which lends credence to the investor recognition hypothesis.4 We provide a number of tests to support this view.

An important methodological concern is to what extent trading volume proxies for stocks' visibility.5 If a high volume shock attracts investor attention and subsequently increases a stock's visibility, we should observe an increase in trading volume after the shock. Empirically, we find increased trading volumes after high volume shocks relative to prior trading volumes. This is particularly the case for stocks that were previously less visible in the market. These findings are consistent with the view that high volume shocks increase the probability of an investor investigating a stock (Miller, 1977) and are a suitable proxy for the unobserved attention-grabbing event.

In the absence of investor attention, stock prices underreact to public information about firm fundamentals.6 If high-volume shocks do attract investor attention, the effect should be stronger in stocks that were previously less known to investors. Merton (1987) points out that firms with less or no coverage by institutional investors and professional analysts are less visible to investors. In line with this view, Barber and Odean (2008) argue that attention is not as scarce a resource for institutional investors as it is for individuals. Dyl and Elliott (2006) find that small and less well-known firms appear to attempt to increase their investor bases by having low stock prices. Less visible firms should also have relatively lower trading activities. In this paper, we use the proportion of institutional ownership, stock prices, firm size, and past share turnover as indicators of a stock's visibility.7 We find that the volume shock effect is stronger for stocks with low share prices and small market capitalization, and with a low level of institutional ownership and past share turnover. The effect also vanishes quickly in stocks that were previously more visible to investors. This result is expected, since the effect of a high volume shock due to attention should be marginal for stocks that already have high visibility.

We further categorize stocks experiencing a high volume shock based on changes in their institutional ownership. Huang et al. (2011) argue that high volume shocks accompanied by increased institutional ownership are more likely attributable to increased recognition, resulting in stronger high volume return premiums than otherwise. However, we argue that changes in the percentage of individual investors' holdings of a stock are also indicative of investor recognition. First, institutional investors have an information advantage and are able to access large databases consisting of nearly the whole universe of publicly traded companies. Second, Merton's (1987) model assumes that individual investors consider only stocks they know before investing and does not explicitly recognize institutional investors. In a similar vein, Barber and Odean (2008) argue that individual investors are net buyers of stocks that draw their attention. Accordingly, if a high volume shock does increase a stock's attention and the stock subsequently becomes more recognized, then higher returns should logically present, irrespective of which group of investors is buying. This should particularly be the case for stocks where attention matters the most. Our empirical results confirm this conjecture. High volume shock premiums are present for both the largest and smallest increases in institutional ownership, particularly among stocks that were previously less recognized.

Given the strong evidence supporting the existence of the volume shock effect, a natural question is whether the volume-shock effect is priced. To explore whether volume shock proxies for sensitivity to a priced risk factor, we construct a volume shock mimicking portfolio. Using both portfolio sorts and cross-sectional regressions, we find no evidence that stock-level sensitivities to the volume shock factor, obtained by running time-series regressions, are related to future stock returns. This result suggests that the volume shock effect is not related to the underlying economic source of risk.

The remainder of this study is organized as follows. Section 2 describes the data and the research design. Section 3 reports the main findings. Section 4 explores the underlying causes of the volume shock return premium. Section 5 examines whether volume shock is a priced factor. Finally, Section 6 concludes the paper.

Section snippets

Data

The analysis in this study is conducted at the monthly level, from 1992 to 2013. The data are obtained from four sources. Monthly stock information is obtained from the Share Price and Price Relative (SPPR) file from the Securities Industry Research Centre of Asia-Pacific (SIRCA). We obtain monthly stock returns, closing prices, number of shares outstanding, market capitalization, share types, return on the value-weighted market index (as a proxy for the market portfolios), and returns on the

Portfolio sorts

We start our empirical analysis by investigating whether portfolios that contain stocks experiencing high volume shocks outperform portfolios that contain stocks experiencing low volume shocks. Table 1 displays the summary statistics of the decile portfolios formed based on VOSHOCK. Both firm size and share price tend to exhibit a U-shaped relation with VOSHOCK. Other firm characteristics, such as book-to-market, past returns, and the Amihud illiquidity ratio, do not exhibit an apparent

Volume shocks and investor attention

Merton (1987, p. 500) shows analytically that, when a stock is more publicly recognized, there will be an increase in its shareholder base and concurrent share price, resulting in lower expected returns and improved risk sharing across investor portfolios. This investor recognition hypothesis is used by a number of studies to explain the high volume return premium (Gervais et al., 2001; Kaniel et al., 2012). Stocks that are less recognized by investors are generally small in size and have less

Are volume shock effects priced?

The empirical findings in the previous sections show a robust volume shock effect in stock returns. In Section 4, we demonstrated that VOSHOCK captures investor attention and predicts future price movements. A natural question that has arisen is whether VOSHOCK is priced. To answer this question, we follow the procedure outlined by Bali et al. (2014). The procedure includes forming a volume shock-mimicking factor, denoted VOF, constructed from portfolios double-sorted on size and VOSHOCK in a

Conclusion

Using an alternative measure of volume shocks, this article investigates whether a volume shock effect exists in the Australian equity market. The study provides out-of-sample evidence to the literature and contributes to the understanding of investor behaviors in the Australian market. Academically, the link between volume shocks and stock returns will advance our understanding on the risk–return relation. Practically, trading volume has long been used in technical analysis and the findings of

References (64)

  • M. Glaser et al.

    Which past returns affect trading volume

    J. Financ. Mark.

    (2009)
  • D. Hirshleifer et al.

    Limited attention, information disclosure, and financial reporting

    J. Account. Econ.

    (2003)
  • R. Kaniel et al.

    The high volume return premium: cross-country evidence

    J. Financ. Econ.

    (2012)
  • J. Kim et al.

    Institutional holdings and trading volume reactions to quarterly earnings announcements

    J. Account. Audit. Finance

    (1997)
  • W. Li et al.

    Differences in herding: individual vs. institutional investors

    Pac. Basin Financ. J.

    (2017)
  • L. Peng et al.

    Investor attention, overconfidence and category learning

    J. Financ. Econ.

    (2006)
  • C.A. Sims

    Implications of rational inattention

    J. Monet. Econ.

    (2003)
  • A. Zhong et al.

    The max effect: an exploration of risk and mispricing explanations

    J. Bank. Financ.

    (2016)
  • A. Ang et al.

    The cross-section of volatility and expected returns

    J. Financ.

    (2006)
  • A. Arbel et al.

    Giraffes, institutions and neglected firms

    Financ. Anal. J.

    (1983)
  • T. Bali et al.

    Liquidity shocks and stock market reactions

    Rev. Financ. Stud.

    (2014)
  • M. Bank et al.

    Google search volume and its influence on liquidity and returns on German stocks

    Fin. Mkts. Portfolio Mgmt.

    (2011)
  • B.M. Barber et al.

    All that glitters: the effect of attention and news on the buying behavior of individual and institutional investors

    Rev. Financ. Stud.

    (2008)
  • L. Blume et al.

    Market statistics and technical analysis: the role of volume

    J. Financ.

    (1994)
  • T. Brailsford et al.

    Size and book-to-market factors in Australia

    Aust. J. Manag.

    (2012)
  • M. Carhart

    On persistence in mutual fund performance

    J. Financ.

    (1997)
  • M.J. Cooper et al.

    Market states and momentum

    J. Financ.

    (2004)
  • K. Daniel et al.

    Evidence on the characteristics of cross sectional variation in stock returns

    J. Financ.

    (1997)
  • S. DellaVigna et al.

    Investor inattention and Friday earnings announcements

    J. Financ.

    (2009)
  • M.S. Drake et al.

    Investor information demand: Evidence from Google searches around earnings announcements

    J. Account. Res.

    (2012)
  • M.S. Drake et al.

    The determinants and consequences of information acquisition via EDGAR

    Contemp. Account. Res.

    (2015)
  • E. Dyl et al.

    The share price puzzle?

    J. Bus.

    (2006)
  • Cited by (7)

    • Investor sentiment and stock market anomalies in Australia

      2023, International Review of Economics and Finance
    • Fundamental volatility and informative trading volume in a rational expectations equilibrium

      2021, Economic Modelling
      Citation Excerpt :

      Karpoff (1986) provides a review of early studies. More recent, Chordia and Swaminathan (2000), Lee and Swaminathan (2000), Lee and Rui (2002), Fleming et al. (2006), Gagnon and Karolyi (2009), Chen (2012), Chuang et al. (2012), Barinov (2015), Akbas (2016), Zhong et al. (2018), and others add new insights into this topic. These authors show that the relation can be time- and market-dependent.

    • Trading volume and stock returns: A meta-analysis

      2021, International Review of Financial Analysis
      Citation Excerpt :

      These investors are supposed to be less rational and suffer from overconfidence and cognitive bias (Kaniel, Saar, & Titman, 2008; Kim, Kim, & Nofsinger, 2003). In which direction it affects the volume-return relationship is from the literature unclear (Kaniel, Ozoguz, & Starks, 2012; Zhong et al., 2018). Furthermore, Market size, measured in terms of GDP at the midyear of the data, helps distinguish larger markets from smaller ones on each continent.

    • Firm efficiency and stock returns: Australian evidence

      2021, International Review of Financial Analysis
      Citation Excerpt :

      Second, our results highlight the potential role of mispricing due to limits-to-arbitrage in perpetuating return predictability in the Australian equity market. In doing so, our paper also extends the findings of previous studies on the role of liquidity in asset-pricing in Australia (e.g., Chai, Faff, & Gharghori, 2013; Chan & Faff, 2003; Zhong, 2018; Zhong, Chai, Li, & Chiah, 2018). In general, our study is related to the broad literature on various other forms of firm-level efficiencies that are difficult to process, such as innovative efficiency (Hirshleifer, Hsu, & Li, 2013) and inventory productivity (Alan, Gao, & Gaur, 2014).

    • Stock return predictability: Evidence from moving averages of trading volume

      2021, Pacific Basin Finance Journal
      Citation Excerpt :

      In this subsection, we examine whether MAVD–return pattern is a repackage of volume shock effect in a regression framework that includes both MAVD and volume shocks. Following Bali et al. (2014) and Zhong et al. (2018), we construct two proxies of volume shocks, including (1) VOSHOCK1, defined as the negative difference between illiquidty and its past 12-month average, and standardize the difference by its volatility; (2) VOSHOCK2, defined as the difference between volume and its past 12-month average, and standardize the difference by its volatility. Table 17 provides the results of Fama and MacBeth (1973) regressions.

    View all citing articles on Scopus
    View full text