Excess returns to buying low options‐volume stocks and selling high options‐volume stocks: Information or characteristics?

• DOI: http://doi.org/10.1002/fut.21957
• Published date: 01 December 2018
• Date: 01 December 2018
• Author: Jian Du,Li Cai

Received: 21 September 2016

|

Revised: 22 June 2018

|

Accepted: 23 June 2018

DOI: 10.1002/fut.21957

RESEARCH ARTICLE

Excess returns to buying low options‐volume stocks and

selling high options‐volume stocks: Information or

characteristics?

Li Cai

1

|

Jian Du

2

1

Stuart School of Business, Illinois

Institute of Technology, Chicago, Illinois

2

Bank of America Merrill Lynch, Chicago,

Illinois

Correspondence

Li Cai, Stuart School of Business, Illinois

Institute of Technology, 565 W Adams St.,

Chicago, IL 60661.

Email: lcai5@stuart.iit.edu

Abstract

This paper documents a negative relationship between options‐trading volume and

stock returns. The relationship is remarkably robust and cannot be explained by

existing asset‐pricing theorems. We find that strategies that require buying stocks

with low options‐trading volume in the past and selling stocks with high options‐

trading volume in the past generate significant positive abnormal returns. Further

analysis indicates that the pattern mostly represents a characteristic effect, in which

options trading predicts stock returns through its relationship with determinant

characteristics such as beta, illiquidity, and idiosyncratic volatility.

KEYWORDS

characteristic, market efficiency, options‐trading volume, prediction

1

|

INTRODUCTION

In this study, we find a negative relationship between options volume and stock returns. The relationship is

remarkably robust and cannot be explained by existing asset‐pricing theorems. Specifically, we find a strong

negative predictability from options‐trading volume for a broad cross section of US stocks. We find that investors

canearnexcessaveragereturnsbybuyingstockswithpastlowoptions‐trading volume and selling stocks with

past high options‐trading volume. The relationship is present in both monthly and weekly volumes, and it persists

for up to 2 months.

Information may affect this relationship. In an efficient market, the underlying market and the derivatives market

simultaneously reflect information. However, due to market frictions, derivatives markets and the underlying

markets absorb information at different speeds when informed agents choose to trade in one market in certain

situations. Pan and Poteshman (2006) study the information effect on options trading. They find that the ratio of

put–call options volume initiated by buyers to open new positions predicts stock returns for up to a few weeks.

However, their data are special and their predictor variable is not publicly available. Relatively little is known about

whether publicly observable options trading contains information about future stock market movemen ts, but this

paper fills the void.

Alternatively, options‐trading variables predicting the cross section of average returns may represent an asset‐pricing

or firm characteristic effect. The capital asset‐pricing model (CAPM) predicts that only market risk (beta) is priced in

equilibrium. Later literature suggests including more systematic risks, such as volatility risk (Ang, Hodrick, Xing &

Zhang, 2006) and liquidity risk (Amihud, 2002; Pastor & Stambaugh, 2003). Other firm characteristics relating to

average stock returns are considered anomaly variables, including size (Banz, 1981), value (Rosenberg, Reid, &

J Futures Markets. 2018;38:1487–1513. wileyonlinelibrary.com/journal/fut © 2018 Wiley Periodicals, Inc.

|

1487

Initial version of the paper was conducted when the author was at Isenberg School of Management, University of Massachusetts, Amherst, MA.

Lanstein, 1985), momentum (Jegadeesh & Titman, 1993), profitability (Novy‐Marx, 2013), investment (Patricia,

Fairfield, & Yohn, 2003; Titman, Wei, & Xie, 2004), accruals (Sloan, 1996), and net stock issues (Daniel & Titman, 2006;

Pontiff & Woodgate, 2008).

1

Additional factors may predict cross‐sectional returns due to market frictions. For instance,

Frazzini and Pedersen (2014) present a model with leverage and margin constraints and show that because constrained

investors bid up high‐beta stocks, high beta is associated with low alpha. As such, options trading variables may predict

stock returns through a relationship with stock determinants, either known or unknown.

Thisstudyexaminesaverysimpleoptionstradingvariable: Aggregate options trading volume (volume).

2

It excludes

stocks with illiquid options trading by keeping 2,000 firms that have the highest options‐trading volume during our sample

period of January 1996 to August 2014. Compared to the existing literature about options, our cross‐sectional sample of

underlying firms is large. For instance, in Pan and Poteshman (2006), the average sample size is less than 400. Using both

portfolio sorting and regression analysis, we document robust findings that volume negatively predicts future cross‐

sectional stock returns. The predictability remains after controlling for size, book‐to‐market ratio, momentum, alternative

options‐trading variables, and earnings prediction.

3

The study also tests a set of scaled options volume variables and find

very similar results. The relation exists for between 1 week and 2 months for weekly volume and for 1–2 months for

monthly volume after the portfolio formation. In unreported results, we track stock returns beyond 2 months, and the

cumulative returns do not reverse. A trading strategy that longs the low‐volume portfolio and shorts the high‐volume

portfolio yields a significant monthly mean returns of 1.27% and a Carhart four‐factor αof 1.28%.

4

The study further examines the information and characteristic hypotheses, and our empirical tests provide strong

support for the latter. If the aforementioned pattern in returns is attributable to the asset‐pricing effect, then sorting on

older and/or average options volume should perform similarly to use the most recent options volume. On the other

hand, if the relationship is attributable to information that is slowly incorporated into prices, then recent options

volume would have a stronger effect because new information would be more relevant. In particular, we find that the

relationship between options volume and stock returns remains to a large extent when we sort on older and average

volume variables, clearly supporting the characteristic hypothesis. Some evidence is also consistent with a bearish‐

information effect, but only in the monthly‐volume‐sorted portfolios. Specifically, we find that high‐volume stocks

sorted on the most recent monthly volume underperform benchmarks.

Next, the characteristic effects are examined in detail. Previous research finds that size, value, net stock issues,

accruals, asset growth, profitability, and momentum are associated with anomalous average returns. The popular four‐

factor model in Carhart (1997) can explain the size, value, and momentum effect where they are deemed risk‐related

characteristics. First, we examine whether excess returns associated with the 5‐1 strategy that longs a quintile 5 options‐

volume portfolio and shorts a quintile 1 options‐volume portfolio are consistent with compensation for bearing risks.

A time‐series analysis of the 5‐1 strategy returns reveals that market risk and growth risk are strong and persistent; size

risk and momentum risk are significant as well. However, all risk premiums put together explain only a small portion of

excess return, still leaving a large and signifincant alpha unexplained.

Next, other known characteristics are examined. A comprehensive Fama–Macbeth regression suggests that options

volume may predict stock returns partly because of its links with beta, liquidity, and idiosyncratic volatility. In

particular, because investors have leverage constraints and bid up high‐beta stocks, high beta is associated with low

alpha (Frazzini & Pedersen, 2014). At the same time, high‐beta stocks tend to have high volatility, which is usually

associated with high options volume due to hedging demand. As such, we expect to see higher options volume

predicting lower abnormal returns. Another characteristic that channels the options volume–stock return relation is

stock illiquidity. The literature suggests a cross‐sectional, positive return–illiquidity relationship, and because options

volume tends to be lower for illiquid stocks, one may expect a negative relationship between a stock’s option volume

and its return. Idiosyncratic volatility is possibly an explanatory characteristic. Options volume responds positively to

idiosyncratic volatility due to hedging demand, and a high idiosyncratic volatility predicts low stock returns in the cross

section (Ang et al., 2006).

1

Some literature suggests that firm characteristics related to cross‐sectional returns are proxies of risks and thus also represent asset‐pricing phenomena.

2

In untabulated results, we test an alternative options‐trading variable: Call–put ratio from options volume (

∕C

P

). Although less significant than volume,

∕C

P

positively predicts future stock returns.

The results of

∕C

P

are available upon request.

3

Literature finds that options‐trading activities predict future earnings announcements (Amin & Lee, 1997; Jennings & Starks, 1986; Mendenhall & Fehrs, 1999; Skinner, 1990). To ensure that our

results do not simply manifest the predictability of options trading on earnings announcements, we run prediction regressions controlling for earnings‐announcements effect. We find that the

predictability of options trading remains significant when we include a dummy variable of earnings announcement and an interaction term. We find that the options trading is generally informative

during “normal”times.

4

A similar strategy of

∕C

P

yields a significant monthly mean returns of 0.65% and a Carhart four‐factor αof 0.67%.

1488

|

CAI AND DU