Review of Quantitative Finance and Accounting, 14 (2000): 131±153
# 2000 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Return Volatility, Trading Imbalance and the
Information Content of Volume
School of Management, Syracuse University, Syracuse, NY 13244, USA
XIAOQING ELEANOR XU
School of Business Administration, Saint Louis University, 3674 Lindell Blvd., DS300 St. Louis, MO 63108
Abstract. In this paper, we examine the relationship between volume and return volatility using the transaction
data. We introduce transaction and volume imbalance measures to capture the information content of trades.
These two information measures are shown to have a strong explanatory power for return volatility and contain
incremental information about the asset values over and above that conveyed by the size and frequency of trades.
Also, return volatility is signi®cantly correlated with the percentage of trading volume taking place at NYSE. This
result suggests that NYSE trades are more informative and contribute more to price discovery. There is evidence
that price discovery concentrates in more heavily traded stocks, particularly the Dow Jones Stocks. Finally, return
volatility is found to be persistent at the intraday level. The persistence level is higher for less frequently traded
stocks. Return volatility also exhibits temporal variations. In particular, return volatility is signi®cantly higher in
the opening half-hour for less frequently traded stocks. Thus, stocks with different frequencies of trades may
follow different volatility processes.
Key words: trade imbalance, transaction size and frequency, GARCH
JEL Classi®cation: G0, G1.
Finance literature has documented a positive relation between return volatility and trading
volume (Karpoff, 1987; Schwert, 1989; Gallant, Rossi and Tauchen, 1992; Lamoureux and
Lastrapes, 1990; Anderson, 1996). Earlier studies on the volatility-volume relation
typically followed the framework of the Mixture of Distribution Hypothesis (MDH) that
posits a joint dependence of price change and volume on underlying latent events or
information ¯ows (Clark, 1973; Tauchen and Pitts, 1983). Although earlier tests were
supportive of the model, later studies found negative evidence against the MDH (Heimstra
and Jones, 1994, Lamoureux and Lastrapes, 1994; Richardson and Smith, 1994). Recent
empirical investigations of the volatility-volume relation have drawn more heavily on the
implications of the market microstructure theory (Jones, Kaul and Lipson, 1994; Foster
and Viswanathan, 1995; Anderson, 1996). The theory suggests that trading arises from
asymmetric information or differences in opinion, and volume re¯ects the extent of
disagreement among market participants about the value of the traded asset. The theory
thus predicts a positive relation between trade size and the quality of information and
hence a positive relation between volume and return volatility.
Recently, Jones, Kaul and Lipson (1994) documented striking evidence for the role of
trade frequency in determining return volatility. They decomposed the total volume into
the frequency of trades and average trade size. They found that the positive volatility-