Trading volume and prediction of stock return reversals: Conditioning on investor types' trading

DOIhttp://doi.org/10.1002/for.2582
AuthorOlena Onishchenko,Numan Ülkü
Date01 September 2019
Published date01 September 2019
RESEARCH ARTICLE
Trading volume and prediction of stock return reversals:
Conditioning on investor types' trading
Numan Ülkü
1
| Olena Onishchenko
2
1
Institute of Economic Studies, Charles
University, Prague, Czechia
2
Department of Accountancy and
Finance, University of Otago, Dunedin,
New Zealand
Correspondence
Numan Ülkü, Institute of Economic
Studies, Charles University, Opletalova 26,
CZ110 00 Prague, Czechia.
Email: numan.ulku@gmail.com
Funding information
Accounting and Finance Association of
Australia and New Zealand; University of
Otago; Commerce Research Grant 2015
Abstract
We show that contrasting results on trading volume's predictive role for short
horizon reversals in stock returns can be reconciled by conditioning on differ-
ent investor types' trading. Using unique trading data by investor type from
Korea, we provide explicit evidence of three distinct mechanisms leading to
contrasting outcomes: (i) informed buyingprice increases accompanied by
high institutional buying volume are less likely to reverse; (ii) liquidity sell-
ingprice declines accompanied by high institutional selling volume in insti-
tutional investor habitat are more likely to reverse; (iii) attentiondriven
speculative buyingprice increases accompanied by high individual buying
volume in individual investor habitat are more likely to reverse. Our approach
to predict which mechanism will prevail improves reversal forecasts following
return shocks: An augmented contrarian strategy utilizing our ex ante formu-
lation increases shorthorizon reversal strategy profitability by 4070% in the
US and Korean stock markets.
KEYWORDS
forecasting shorthorizon reversals in stock returns, trading of investor types, trading volume
1|INTRODUCTION
Shorthorizon reversals, first documented by Lehmann,
(1990), Atkins and Dyl, (1990), and Bremer and Sweeney,
(1991), are one of the main predictable patterns in stock
markets. The likelihood of reversals following stock return
shocks is related to the trading volume accompanying the
return shock. The performance of shorthorizon reversal
strategies significantly varies when conditioned on vol-
ume. However, there are contrasting empirical results
and theoretical predictions on trading volume's role: rever-
sals are more likely following high volume in some cases
and low volume in others. Explanations for the contrasting
outcomes have been only partly established, and backed by
only indirect evidence of the mechanisms proposed.
High trading volume represents mass action in the
stock market. Its potential association with the likelihood
of subsequent reversals would imply systematic mecha-
nisms, driving the behavior of crowds and leading to pre-
dictable outcomes. Therefore, a clearer understanding of
trading volume's role in forecasting subsequent reversals
in stock returns is warranted.
The series of contrasting findings on trading volume's
predictive role in seminal empirical studies starts with
Conrad, Hameed, and Niden, (1994), who find on a sam-
ple of NASDAQ stocks that reversals are more likely fol-
lowing higher volume. In contrast, on a sample of large
cap NYSEAMEX stocks, Cooper, (1999) reports that
reversals are more likely following lower volume.
Llorente, Michaely, Saar, and Wang, (2002) find that
reversals are more likely following higher volume among
largecap stocks, against Cooper's, (1999) result. All these
findings conflict with Stickel and Verrechia's (1994) ear-
lier result that reversals are more likely following lower
Received: 8 October 2018 Accepted: 13 February 2019
DOI: 10.1002/for.2582
582 © 2019 John Wiley & Sons, Ltd. Journal of Forecasting. 2019;38:582599.wileyonlinelibrary.com/journal/for
volume among both NYSE/AMEX and NASDAQ stocks.
Avramov, Chordia, and Goyal, (2006) find more reversals
with highturnover stocks after controlling for liquidity.
Beyond these seminal articles, the picture gets more com-
plicated as more studies provide further breakdowns
and/or outofsample evidence. Studies of return autocor-
relation conditional on trading volume (e.g., McKenzie &
Faff, 2005; Säfvenblad, 2000) or the dynamic (causal) rela-
tion between volume and returns (e.g., Balcılar, Bouri,
Gupta, & Roubaud, 2017; S. Chen, 2012; Chuang, Kuan,
& Lin, 2009; Gebka & Wohar, 2013; Hiemstra & Jones,
1994) also obtain inconsistent or weak results.
Theoretical work suggests that return shocks accompa-
nied by high volume should be more likely to reverse if
high volume is driven by noninformational (liquidity or
hedging motivated) trading and less likely to reverse if it
is driven by informed trading. This conclusion evolved
from Campbell, Grossman, and Wang's, (1993) model of
noninformational trading absorbed by incomplete liquid-
ity provision, as J. Wang's, (1994) and Llorente et al.'s,
(2002) models added trading on private information. Sub-
sequent to Llorente et al., there seems to exist no further
progress to account for different outcomes in the trading
volumesubsequent reversal link.
1
These rational
models leave a gap concerning a third mechanism
whereby return shocks accompanied by high volume
driven by an individual investor bias, attentiondriven
retail buying (Barber & Odean, 2008), are more likely to
reverse. This third mechanism is particularly relevant
for market segments where individual investors dominate
trading.
In this article, we offer a new approach to characterize
when high or low volume is more likely to predict rever-
sal, by conditioning on investor types' trading. This
approach provides a more complete characterization of
potential outcomes and yields ex ante predictions
regarding which outcome is likely to follow. Our key
contribution is providing explicit evidence of three dis-
tinct mechanisms (information trading, noninformational
liquidity consumption, and attentiondriven speculative
retail buying) prevailing under the hypothesized
circumstances and leading to the predicted outcomes.
We obtain this evidence utilizing the world's most
comprehensive trading dataset with investor type identifi-
cation from the Korea Stock Exchange (KRX). On the
practical side, our partitioning, with ex ante predictions
of which mechanism will prevail, improves forecasts of
subsequent reversals, and provides economically signifi-
cant increases in the profitability of a shortterm reversal
strategy.
The three mechanisms are associated with the trading
of different types of investors, and lead to specific out-
comes in different segments of the market (investortype
habitats). While all three mechanisms may potentially be
present everywhere, we conjecture that each can be rela-
tively dominant under specific circumstances. In Section
3, we first outline these mechanisms in generic terms
based on conjectured motivations for highvolume trad-
ing, and then develop our partitioning to predict which
mechanism will prevail in each market segment by utiliz-
ing information on investor type habitat in our empirical
setting.
Our empirical setting is based on a veryshorthorizon
reversal strategy: each day all stocks are ranked by their
1day returns; the strategy portfolio goes long (short) the
mostloser (winner) decile. This setup essentially cap-
tures idiosyncratic return shocks. We apply this strategy
to both the US stock market and the KRX. The loser
minuswinner (LMW) portfolio earns statistically signifi-
cant positive returns during a 4day subsequent holding
period in both the USA and KRX. Positive returns of
the LMW portfolio imply profits to the contrarian strat-
egy; hence reversals. We then condition these reversals
on high versus low volume on the ranking day. Based
on our partitioning by investor types' trading and habitat,
we form ex ante predictions regarding whether reversal is
more likely in highor lowvolume winners and losers.
Results indicate that the relative likelihood of reversal
in highversus lowvolume portfolios is successfully pre-
dicted by our partitioning in all cases. An augmented
reversal strategy, which utilizes this predictive framework
to choose between highand lowvolume winner and
loser portfolios on an ex ante basis, achieves an approxi-
mately 4070% increase in profitability over the plain ver-
sion in the US and KRX market segments.
In the key final step we examine, using our KRX
dataset, whether the volume increases accompanying
the return shocks are indeed associated with the trading
of the hypothesized investor type. Results from this step
yield direct evidence on the three mechanisms that pre-
vail under the hypothesized circumstances as predicted
by our partitioning: (i) Winner stocks' high volume in
the institutional investor habitat is associated with
1
Another class of theoretical models (e.g., L. Blume, Easley, & O'Hara,
1994; Schneider, 2009) establishes the importance of trading volume
for return prediction, allowing rational agents to utilize information
conveyed by volume, but does not generate specific predictions. Several
models, which focus on explaining the contemporaneous relationship
between volume and returns, yield implications regarding the prediction
of future returns; for example, Sentana and Wadhwani, (1992) point out
that positive feedback trading can contribute to the positive contempora-
neous volumereturn relationship. High volume induced by positive
feedback trading should predict reversals. Based on a similar argument,
Dasgupta et al.s, (2011) model predicts return momentum following
high institutional volume (institutional herding) over short horizons
and subsequent reversal over longer horizons.
ÜLKÜ AND ONISHCHENKO 583

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT