Journal of Financial Economics 111 (2014) 625645

Contents lists available at ScienceDirect

Journal of Financial Economics

journal homepage: www.elsevier.com/locate/jfec

Does option trading convey stock price information?$

Jianfeng Hu n
Lee Kong Chian School of Business, Singapore Management University, 50 Stamford Road, Singapore 178899, Singapore

a r t i c l e i n f o

abstract

Article history:
Received 4 June 2012
Received in revised form
11 July 2013
Accepted 23 July 2013
Available online 19 December 2013

After executing option orders, options market makers turn to the stock market to hedge
away the underlying stock exposure. As a result, the stock exposure imbalance in option
transactions translates into an imbalance in stock transactions. This paper decomposes the
total stock order imbalance into an imbalance induced by option transactions and an
imbalance independent of options. The analysis shows that the option-induced imbalance
significantly predicts future stock returns in the cross section controlling for the past stock
and options returns, but the imbalance independent of options has only a transitory price
impact. Further investigation suggests that options order flow contains important information
about the underlying stock value.
& 2013 Elsevier B.V. All rights reserved.

JEL classification:
G14
G12
G13
Keywords:
Options
Order flow
Information asymmetry
Delta hedging
Price discovery

1. Introduction

I thank G. William Schwert (the editor), Avanidhar Subrahmanyam

(the referee), Linda Allen, Turan Bali, Gurdip Bakshi, Jin-Chuan Duan,
Fangjian Fu, Harmeet Goindi, Armen Hovakimian, Jian Hua, Robert
Jarrow, Mehdi Karoui (discussant), Pete Kyle, Dmitriy Muravyev, Lin Peng,
Henry Schwartz, Robert Schwartz, Zhaogang Song (discussant), Jun Tu,
Mitch Warachka (discussant), Jason Wei, Liuren Wu, Yangru Wu, Shu Yan
(discussant), Rui Yao, Joe Zhang, Xiaofei Zhao (discussant), Hao Zhou, and
the seminar participants at Baruch College, Central University of Finance
and Economics, City University of Hong Kong, DePaul University, Drexel
University, Fordham University, Hong Kong University of Science and
Technology, Indiana University, National University of Singapore, Rutgers
University at Camden, Singapore Management University, Sungkyunkwan University, Trade Alert LLC, University of Georgia, University of
Massachusetts at Amherst, University of Missouri at Columbia, the 2013
China International Conference in Finance, the Sixth Annual Risk Management Conference in Singapore, the 2012 European Finance Association Doctoral Tutorial in Copenhagen, the 2012 Northern Finance
Association Conference, the 22nd Annual Derivatives Securities and Risk
Management Conference at the FDIC, and the Sixth Annual Conference on
Asia-Pacific Financial Markets. I owe special thanks to Henry Schwartz for
generously providing the option tick data at Trade Alert LLC.
n
Tel.: 65 6808 5477; fax: 65 6828 0427.
E-mail address: jianfenghu@smu.edu.sg

0304-405X/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jfineco.2013.12.004

Microstructure theories such as Ho and Stoll (1981),

Glosten and Milgrom (1985), Kyle (1985), and Easley and
O0 Hara (1987) suggest that a stock0 s order flow affects its
price and that, in particular, transactions initiated by buyers
(sellers) cause the stock prices to move up (down). The
empirical evidence that supports this prediction exists both
at the market level (Chordia, Roll, and Subrahmanyam,
2002) and in the cross section of stocks (Chordia and
Subrahmanyam, 2004).
The stock options market provides an alternative venue
for gaining stock exposure. For example, Easley, O0 Hara,
and Srinivas (1998) and Pan and Poteshman (2006) show
that options order flow also predicts the underlying stock
returns. This paper examines the interaction between
option transactions and the underlying stock transactions
and how this interaction affects the predictability of stock
returns.
When customers execute option trades, the options
market makers take the opposite side of the transaction

626

J. Hu / Journal of Financial Economics 111 (2014) 625645

and earn the bidask spread. Given the relative scarcity of

option transactions, the market makers have difficulty in
immediately unloading the positions through trades in the
opposite direction. They often have to hold the positions
for a long time, frequently until the options expire. To
reduce the risk exposure from these positions, a standard
practice of the market makers is to perform delta hedging
by trading on the underlying stocks. Therefore, if option
transactions generate an imbalance in the stock exposure,
the market makers transfer that imbalance to the stock
market through their delta hedging practice.
To understand the interactions between transactions in
the stock options market and the underlying stock market,
this paper decomposes the overall imbalance in stock
transactions into two components: an imbalance induced
by option transactions and the remaining imbalance
induced by stock market transactions unrelated to option
trading. To compute the option-induced stock order imbalance, the paper assumes that the options market makers
(who receive the bidask spread in the transaction) use
full delta hedging and that customers (who initiate the
option transactions by paying the spread) do not hedge.
Under these assumptions, the paper computes the optioninduced stock order imbalance by aggregating the signed
option transactions weighted by the respective delta
exposure of each option contract.
The direction of each option transaction is determined
according to the Lee and Ready (1991) algorithm. Over a
fixed time interval (say, a day), the volume of the signed
option transactions can be aggregated to generate an order
flow measure for each option contract. Weighting the
order flow of each option contract by its delta exposure,
the paper computes a measure for aggregated delta
exposure over all of the option transactions for each stock.
This measure provides an estimate of the stock order
imbalance induced by option trading. Subtracting this
option-induced order imbalance from the total stock order
imbalance results in the residual stock order imbalance
that is unrelated to the option trading activities. This paper
analyzes how these two components interact to predict
future stock returns.
The paper computes daily stock order imbalance for a
large cross section of stocks with options during the period
of April 2008 to August 2010. On average, the sample
contains 2,207 stocks per day. The paper examines the
cross-sectional relation between the two components of
the stock order imbalance and the stock returns over the
next day, and finds several results new to the literature.
First, out of the two components, only the option-induced
order imbalance positively predicts stock returns over a
day controlling for microstructural variables including the
past stock and options returns. Firms in the quintile of the
highest option-induced order imbalance outperform those
in the lowest quintile by 8.74 basis points (bp) on the next
day. This outperformance translates into an annualized
excess return of 22% with a t-statistic of 6.03. The remaining stock order imbalance unrelated to option trading has
no significant forecasting capability for stock returns,
although it shows a large contemporaneous price impact.
Furthermore, the return predictability from the optioninduced order imbalance does not reverse direction in

longer horizons. This finding suggests that this predictability is driven by a permanent information flow instead
of by temporary price pressure. Aggregating the order
flows over half-hour intervals generates similar patterns in
the predictability of intraday returns.
The finding is particularly interesting because option
trading generates only a small proportion of the total stock
order imbalance, but this small proportion contains most
of the predictive information about permanent stock price
changes.1 The finding suggests that the information from
the options market is not immediately incorporated into
the stock prices. To better understand the meaning and
source of the return predictability, the paper examines
how the return predictability from the option-generated
stock order imbalance varies with option contracts and
firm characteristics.
Subsequently, the paper divides the option-induced
order imbalance into three groups: for at-the-money
(ATM), in-the-money (ITM), and out-of-the-money (OTM)
options. Separate return prediction analysis shows that the
predictive power comes mainly from the delta exposures
of the ATM and the ITM options, but not the OTM options.
This is possible because most institutional investors use
the OTM options to gain volatility exposure while hedging
the associated delta exposure by using the underlying
stock. Their hedging activity offsets that of the market
makers so that the delta imbalance of the OTM options
does not translate into an actual stock order imbalance. On
the other hand, the investors who have private information about the stock price movement often take positions
in the ATM or the ITM options.
The paper also divides firms into different groups based
on their characteristics and analyzes the difference in
return predictability across these groups. First, the firms
are classified into three groups based on their degrees of
information asymmetry. The measurement components of
information asymmetry are the probability of informed
trading (PIN) proposed by Easley and O0 Hara (1987), the
number of stock analysts following the firm, the bidask
spread in the stock market, the adverse-selection component of the bidask spread from both Glosten and Harris
(1988) and Lin, Sanger, and Booth (1995), and the firm size.
Regardless of which measure is used, the option-induced
stock order imbalance always shows the largest predictive
coefficient and t-statistic in the group of informationally
opaque firms, that is, the firms with high PINs, low analyst
coverage, large spreads, large adverse-selection components of the spread, or small market capitalizations. These
findings suggest that informed trading in the options
market drives the return predictability from options
order flow.
The basis for the second type of classification is the
short-sale constraints. Short-sale constraints are measured
by institutional holdings because firms with more institutional holdings are less costly to borrow and sell short
(D0 Avolio, 2002). The analysis shows that the group of
firms with low institutional ownership has both the

1
On average, option trades account for only 3.41% of the total stock
order imbalance.

J. Hu / Journal of Financial Economics 111 (2014) 625645

largest predictive coefficient estimate and the highest

statistical significance for the option-induced order imbalance. This finding suggests that option trading, often
considered an effective way to mitigate short-sale constraints, conveys more information about the firms bound
by short-sale constraints. Informed traders with negative
news could place orders in the options market as an
alternative to short selling, especially when short selling
is more difficult (Figlewski and Webb, 1993; Johnson and
So, 2012).
The basis for the third type of classification is market
trading activities. The analysis shows that the optioninduced order imbalance predicts returns better when
option trading is active and the total options volume is
high. This finding is consistent with the prediction of
Easley, O0 Hara, and Srinivas (1998) that informed traders
prefer to trade in deep markets.
Further, the paper performs an event study around
earnings announcements and finds that the optioninduced order imbalance predicts the cumulative abnormal returns (CARs) five days before the announcement.
This predictability is robust to alternative event windows
and becomes greater when the earnings surprise is large
and the dispersion of analysts0 forecasts is high. Overall,
the analysis shows that option trading becomes more
informative when the information asymmetry is severe
and the profit from informed trading is high.
The findings in this paper have at least two implications. First, the analysis highlights the importance of
separating option-induced stock order imbalance from
order imbalance generated by pure stock market investors.
The prior literature often compares order flows and trading volumes from the options and stock markets directly.
Through a careful separation, this paper shows that
options order flow contains an important informative
component about future stock price movement that is
not found in trading activities of pure stock investors.
Therefore, this decomposition is critical to measuring the
information content in order flows.
Second, the findings shed light on the ongoing debate
regarding where informed traders place their orders.
A number of empirical studies examine the cross-market
information flow between the stock and the options
markets.2 In a closely related study, Pan and Poteshman
(2006) show that the volume ratio of put and call options
predicts future stock prices in the cross section. This paper
goes a step further and investigates the interaction of
the two markets via hedging activities. To the best of
my knowledge, this is the first paper to show that the
permanent price information in stock order flow is
induced mostly by option transactions, which suggests
that informed traders do use options.

2
See, for example, Manaster and Rendleman (1982), Bhattacharya
(1987), Anthony (1988), Stephan and Whaley (1990), Chan, Chung, and
Johnson (1993), Easley, O0 Hara, and Srinivas (1998), Chan, Chung, and
Fong (2002), Chakravarty, Gulen, and Mayhew (2004), Cao, Chen, and
Griffin (2005), Holowczak, Simaan, and Wu (2006), Pan and Poteshman
(2006), Ni, Pan, and Poteshman (2008), and Muravyev, Pearson, and
Broussard (2013).

627

The rest of the paper is organized as follows: Section 2

reviews the literature on the price impact of option
trading. Section 3 proposes a decomposition method for
the total stock order imbalance and develops the hypotheses for empirical testing. Section 4 describes the data
used in this study. Section 5 reports the empirical results
of the hypothesis tests. Section 6 provides additional
analyses on the predictive power resulting from option
transactions. Section 7 concludes.
2. Background and motivation
Derivative trading can convey important information in
a market with information asymmetry. Black (1975) first
notes the possibility that informed traders could use the
options market as an alternative trading venue because
option contracts provide higher leverage. Biais and Hillion
(1994) examine the impact from the introduction of
options and find that informed traders0 profits can either
increase or decrease depending on the type of liquidity
orders. Focusing on how private information gets incorporated into security prices, Easley, O0 Hara, and Srinivas
(1998) formalize a two-market microstructure model in
which informed traders choose to trade in the stock or
options markets. The authors show that a pooling equilibrium exists in which informed traders trade in both
markets when the leverage and liquidity of the options
are sufficiently high. They further argue that because the
availability of multiple option contracts presents difficult
learning problems for uninformed traders, option contracts can be more attractive to informed traders. If
informed traders trade in the options market, the options
order flow can contain information about fundamental
values of the underlying stocks.
The options order flow can gain return predictability
in another important yet less discussed way. Starting with
Black and Scholes (1973) and Merton (1973), the finance
literature has built a modern derivative pricing theory on
the no-arbitrage argument. Other than the price relation
between options and the underlying security, the noarbitrage rule also explains how to hedge option positions
by using the underlying security. This type of hedging
results in a reverse dependence of the stock order flow on
the options order flow. After an options trader executes a
transaction, the options market maker takes the opposite
position and gains risk exposures to the underlying price
movement as well as other factors such as return volatility.
Because the liquidity in the options market is usually not
high enough for market makers to unload inventory
immediately (liquidity is further diluted to hundreds of
different option contracts on the same underlying security), market makers need to hedge the underlying price
risk by transacting in the underlying market. The hedge
ratio is captured by the option price0 s sensitivity to the
underlying price movement, delta. Meanwhile, customers
can also perform delta hedging if they intend to gain risk
exposures other than the delta. In doing so, they not only
trade stocks to hedge the delta exposure themselves but
also induce the options market makers to perform delta
hedging. Therefore, the options order flow leads to subsequent stock order flow through delta hedging transactions.

628

J. Hu / Journal of Financial Economics 111 (2014) 625645

Even without private information, the hedging transactions

can cause transitory pressure on the stock price in the same
way as stock transactions unrelated to the options market.
Without perfect liquidity, such price pressure drives the
stock price away from its fundamental value temporarily,
followed by a reverse price movement to the fundamental
value later. Empirically, Ni, Pearson, and Poteshman (2005)
show that the practice of delta hedging causes the underlying stock prices to cluster around the option strike prices
on option expiration dates. In the structured equity product
market, Henderson and Pearson (2010) find that hedging
transactions raise the underlying stock price by almost one
hundred bp on the pricing date.
The theoretical discussion on the feedback effect from
options has motivated many empirical studies on the leadlag relation between the two markets. A group of researchers focus on the relation between the actual stock prices
and the implied stock prices from options. Early studies
such as Manaster and Rendleman (1982) and Bhattacharya
(1987) find that the options market leads the stock market
in price discovery. However, more evidence points in the
opposite direction. For example, Chakravarty, Gulen, and
Mayhew (2004) find that the information share of option
quotes is less than 20% on average. Holowczak, Simaan,
and Wu (2006) show that the information share of options
even decreases over time, and they argue that this decrease
is because of the prevailing use of computers for the automatic updating of quotes in the options market. In a recent
study, Muravyev, Pearson, and Broussard (2013) find that
option quotes, but not stock quotes, adjust themselves to
eliminate arbitrage opportunities across the two markets,
and they conclude that the options market does not play a
role in price discovery. Another strand of empirical research
directly investigates the options order flow. Many studies
find that this order flow predicts future stock returns
in time series regressions for a small group of stocks, for
example, Easley, O0 Hara, and Srinivas (1998), Poteshman
(2006), Dong and Sinha (2011), and Holowczak, Hu, and Wu
(2014).
Pan and Poteshman (2006) investigate the information
content of the options order flow in the cross section of
stocks. Using a unique data set, these authors construct
volume ratios of put and call options from buyers opening
new positions. They find that the open-buy put-call ratio
negatively predicts returns and that the return predictability lasts over three weeks. However, most studies on the
options order flow do not control for the stock order flow.
It is still unclear how options and stock order flows interact
to generate return predictability. Because informed traders
can trade in both markets, it is important to understand
whether the options market makes a marginal contribution
to embedding private information in stock prices.
Chan, Chung, and Fong (2002) investigate the informational role of order flows and quote revisions in the two
markets. They find that the options order flow does not
have a pricing effect but the stock order flow does. Cao,
Chen, and Griffin (2005) examine order flow information
around mergers and acquisitions. Their results suggest
that the options volume imbalance becomes significantly
informative about future stock prices right before merger
announcements but remains silent during normal periods.

Overall, no significant return predictability is found from

the options order flow once the stock order flow is
controlled for. However, the order imbalances analyzed
in these two studies come directly from trading volumes in
those two markets without separating out the hedging
transactions. Thus, the possibility exists that the contaminated total stock order imbalance absorbs both the information content and the return predictability from the
options order flow. Therefore, the information content in
different markets must be separated carefully before any
lead-lag analysis can be conducted.
This paper is also motivated by some recent developments in the options market. Most studies use data prior to
2000 when the options market in the United States was not
yet a consolidated national market (see Battalio, Hatch, and
Jennings, 2004). Multiple listings on different exchanges
and price competition were not common at that time,
and transaction costs in the options market were very high.
Since 2000, as a result of Securities and Exchange
Commission0 s efforts and the development of quote updating and order routing technologies, the options market has
experienced tremendous growth and consolidation. Market
competition drives down transaction costs and improves
liquidity. For informed traders, these changes are likely to
affect their decisions on order allocations. Therefore, it is
worthwhile to reexamine the information content in the
options order flow in a more recent sample period. Also, the
lack of liquidity in the past significantly limits the sample
size for an empirical analysis. For example, Chan, Chung,
and Fong (2002) have only 14 stocks in their sample in
which the stock with the most active option trading,
General Electric (GE), has an average daily trading volume
of 2,595 option lots. By comparison, GE0 s options volume
reached 134,874 lots on April 1, 2008, the first day in my
sample and the average options volume of all of the
common stocks is 3,255 lots on that day. The growth of
the options market significantly facilitates the expansion of
the sample coverage. This study covers all of the common
stocks with options and focuses on the cross-sectional
relation between the order flows and the future returns.
3. Methodology
This section explains the calculation of the order
imbalances in the options and the stock order flows. It
then describes the primary hypotheses and the empirical
testing methods.
3.1. Options order imbalance
A remarkable feature of the options market is that
hundreds of option contracts underlie the same security.
For example, on January 2, 2009, there were 478 option
contracts underlying Apple0 s common stock that spanned
61 strike prices from $12.5 to $400 and six expiration dates
from 15 days to 750 days.3 Because each contract has its
own unique combination of put or call type, strike price,
and maturity, the trading process for each option contract
3

The closing price of Apple0 s stock was $90.75 on January 2, 2009.

J. Hu / Journal of Financial Economics 111 (2014) 625645

discloses the underlying price information in different

ways. Measuring the price information in the options
order flow is challenging. In a recent study, Holowczak,
Hu, and Wu (2014) show that risk exposure to the underlying stock price (delta) is an important consideration
when extracting the stock price information from option
transactions. Following the authors0 reasoning, I propose a
standardized measure of the options order imbalance
(OOI):
OOIi;t

nj 1 100Diri;t;j  deltai;t;j  sizei;t;j

Num_shares_outstandingi

For stock i on day t, the Diri;t;j is a dummy variable equal to

one (negative one) if the jth option trade is initiated by
the buyer (seller) according to certain trade signing algorithms. The deltai;t;j is the option price0 s sensitivity to the
underlying stock price, and the sizei;t;j denotes the trade
size in option lots (one hundred shares of the underlying
stock). The numerator is thus the aggregate delta position
of the active options traders who initiate the trades.
The net delta exposure captures the directional trading
intention about the underlying stock price, which discloses
the price information in the options market if any. The
order flow is more likely to be imbalanced when trading
is inactive, and this imbalance can merely reflect noise
from uninformed trades. To address this issue and also for
cross-sectional standardization, the final OOI is the net
delta exposure scaled by the number of common shares
outstanding.4
3.2. Stock order imbalance
Because options market makers take the opposite side
of the OOI, the OOI also measures the net delta hedging
demand in the stock market. Then the stock order imbalance (SOI) unrelated to the options market is the difference between the total order imbalance (TOI) in the stock
market and the OOI. The SOI is defined as
SOIi;t TOIi;t OOIi;t

N
j 1 Diri;t;j  sizei;t;j
Num_shares_outstandingi

OOIi;t ;
2

where TOIi;t is in terms of the trading volume, and Diri;t;j

and sizei;t;j are the direction dummy and the size of the jth
stock transaction of firm i on day t, respectively. The TOI is
also scaled by the number of shares outstanding to be
consistent with the OOI. After subtracting the OOI from the
TOI, the remainder is the order imbalance caused by stock
market investors.

629

equation for the cross section of stocks:

5

k1

k1

Reti;t k1 SOIi;t  k k2 OOIi;t  k X i;t  1 i;t ;

3
where, for stock i on day t, the Reti;t is the stock return
calculated using the midpoint of the last National Best Bid
and Offer (NBBO) prices before the market close.5 Furthermore, X i;t  1 is a set of control variables composed of the
closing percentage bidask spread of the stock, the stock0 s
turnover ratio as the total trading volume scaled by the
number of shares outstanding, the log trading volumes in
the two markets, and the stock returns and the equally
weighted options returns from the past five days.6
Hypothesis 1. The SOI positively predicts the future stock
returns, at least for some k, k1 40. If the predictability
comes from the price pressure of liquidity trades, then the
predictive relation reverses its sign in longer horizons. If
the predictability comes from informed trading in the
stock market, then the predictive relation does not reverse
its sign.
Informed trading in the stock market allows the SOI
to predict the permanent price changes (see, e.g., Glosten
and Milgrom, 1985; Kyle, 1985; Easley and O0 Hara, 1987).
A positive (negative) SOI reflects the buying (selling) of
informed traders with good (bad) news and should predict
a higher (lower) stock price in the future. Alternatively,
even in the absence of information asymmetry, the stock
market0 s liquidity providers can lift the price quote after
receiving a buy order and reduce the price quote after
receiving a sell order to entice the offsetting orders and
return to their optimal portfolios. As a result, the SOI also
positively predicts future price changes. However, this
price impact is short-lived. Once the price pressure from
the order imbalance disappears, the stock price returns to
its fundamental value. Thus, a reverse (negative) predictive
relation exists in longer horizons.
Hypothesis 2. The OOI positively predicts the future stock
returns controlling for the SOI, at least for some k, k2 4 0.
If the predictability comes from the price pressure of the
options market makers0 delta hedging, then the predictive
relation reverses its sign in longer horizons. If the predictability comes from informed trading in the options market,
then the predictive relation does not reverse its sign.
If informed traders gain stock exposures in the options
market, then the OOI captures that information content.
Similar to the SOI, a positive (negative) OOI reflects the
private information about good (bad) news and predicts that
the underlying stock price will go up (down) permanently

3.3. The main test

To gauge the return predictability from the order flows
in the two markets, the paper estimates the following
4
Experiments with alternative scalers such as total delta volume and
total stock volume produce similar results. But these alternative scalers
are more likely to generate outliers during inactive trading periods.
Hence, I do not report those results in the paper.

5
In theory, the private information in the order flows is about
specific firms, and the imbalance should have a greater predictive ability
about the idiosyncratic returns. In unreported tests, I use the riskadjusted returns as the dependent variable and find stronger results.
I have also experimented with open-to-close NBBO midpoint returns and
the returns calculated by using transaction prices. The results are largely
the same.
6
I have also experimented with the value-weighted options returns
and found similar results.

630

J. Hu / Journal of Financial Economics 111 (2014) 625645

after that information becomes public. Meanwhile, the

options market makers perform delta hedging to reduce
their risk exposures and transfer the OOI to an imbalance
in the stock market. The OOI also causes price pressure on
the underlying stock, but the effect is temporary. Because
the private information and the price pressure can come
from both markets, understanding the OOI0 s marginal
contribution to the price discovery is important. Eq. (3)
investigates the return predictability from the OOI controlling for the SOI.
3.4. Conditional tests
If the OOI does predict future stock returns, it would be
interesting to investigate how the predictability depends
on option contract types and firm types. The following
tests focus on the sources of the potential return predictability from the OOI and serve as robustness checks on the
informational role of the OOI.
3.4.1. Option leverage
Hypothesis 3. If informed traders use options for leverage,
they should prefer options that give them the highest
leverage. Therefore, the order imbalance of these options
is more informative about future stock returns.
To test this hypothesis, the paper divides all of the option
transactions into three groups based on the moneyness of the
option, delta: out-of-the-money (OTM, jdeltaj o 0:375), atthe-money (ATM, 0:375 rjdeltaj r0:625), and in-the-money
(ITM, jdeltaj 4 0:625). Although the OTM options have the
smallest delta, they provide the highest leverage for every
dollar invested because of their low prices. The ITM options
have the lowest dollar leverage among the three groups. I
calculate the OOI separately for the three option groups and
estimate the following equation:
5

Reti;t k1 OTM_OOIi;t  k k2 ATM_OOIi;t  k

k1

k3

ITM_OOIi;t  k k4 SOIi;t  k X i;t  1 i;t ;

k1

the firm size. To calculate the adverse selection component

of the spread, the paper uses both the Glosten and Harris
(1988) and Lin, Sanger, and Booth (1995) models.
Based on each proxy of information asymmetry, the
paper first sorts all firms by ascending order every day and
divides the full sample into three groups (low, medium,
and high) with cutoff points at the 30th and 70th percentiles. I then reestimate Eq. (3) for each group and compare
the estimated coefficients and t-statistics of the OOI across
groups.

3.4.3. Institutional ownership

Other than the level of information asymmetry, the
short-sale costs also have an impact on the predictive
ability of the OOI. In practice, options frequently are used
as devices to circumvent the short-sale constraint in the
stock market. For stocks more difficult to sell short, option
trading can have greater informational benefits because
informed traders with negative news can be forced to
trade options only. Institutional ownership as a proxy for
the market supply of short interests has a negative relation
to the difficulty of selling short (D0 Avolio, 2002; Asquith,
Pathak, and Ritter, 2005). But, institutions are also more
likely to have an information advantage over retail investors. Thus, the paper proposes Hypothesis 5.
Hypothesis 5. A stock0 s institutional ownership can cause
the OOI0 s predictive ability to decrease because the institutional ownership reduces the short-sale costs. Alternatively, a stock0 s institutional ownership can cause the OOI0 s
predictive ability to increase because the institutional
investors could be better informed.
Similar to the tests discussed in Section 3.4.2, Eq. (3) is
reestimated for the three groups based on the institutional
ownership: low ( o 30%), medium (3070%), and high
( Z 70%). The analysis then compares both the statistical
and economic significance of the estimated OOI coefficients for the groups.

where Reti;t , SOIi;t  k , and X i;t  1 are the same as defined

before; and OTM_OOIi;t  k , ATM_OOIi;t  k , and ITM_OOIi;t  k
are delta order imbalances calculated using the OTM, the
ATM, and the ITM options, respectively.
3.4.2. Level of information asymmetry
Hypothesis 4. If the options order flow discloses private
information, then the predictive ability of the OOI should
be greater for the firms with more information asymmetry.
The likelihood of information-based trading is higher
when more private information exists to be exploited
ceteris paribus. For less transparent firms, the information
content in the OOI can be higher and, thus, the return
predictability can also be stronger. The paper uses five
proxies for the level of information asymmetry: the PIN
from Easley and O0 Hara (1992), the number of analysts
following the firm, the percentage bidask spread, the
adverse-selection component of the bidask spread, and

3.4.4. Market activity

The liquidity in the options market can also affect the
predictive ability of the OOI if the return predictability is
caused by informed trading.
Hypothesis 6. The price impact of the OOI increases when
the liquidity in the options market improves because the
informed traders can conceal their trades better in high
options volume stocks.
The same return prediction analysis is repeated for the
three groups based on the daily option trading volume,
which is a proxy for the options market liquidity.

4. Data
Analysis of the order flows in both markets requires the
merging of several databases. This section describes the
details of the data sources, the sample selection, and the
variable construction.

J. Hu / Journal of Financial Economics 111 (2014) 625645

631

Table 1
Summary statistics of the options market.
The table describes the options market activity between April 2008 and August 2010. Only single-name equity options with maturities of more than ten
days are included in the sample. The following trades are excluded: (i) off-hour trades, (ii) trades at the market open (the first 15 minutes) and the market
close (the last five minutes), (iii) trades that are reported to the Options Price Reporting Authority as late or cancel, (iv) trades flagged as part of a
structured trade, and (v) data errors such as zero strike prices or zero trade prices. All statistics are reported for the full sample as well as for the four
transaction-type groups based on the trade directions and the option types: buy call, sell call, buy put, and sell put. The trade direction is based on the Lee
and Ready (1991) algorithm. Panel A provides the statistics for the number of firms per day. Panel B reports the trade level statistics. The trading volume is
counted as option lots (one lot equals one hundred shares of the underlying stocks). Panel C shows the trading volume breakdown as percentages of the
total volume by the option moneyness, delta, and the maturity, T. The Black-Scholes-Merton model is used to calculate the delta with a zero interest rate
and a zero dividend rate. Panel D gives the volume breakdown by the options exchange. OTM out-of-the-money; ATMat-the-money; ITM in-themoney.
Statistics

All

Panel A: Number of firms per day

Mean
Maximum
Minimum
Standard deviation

Buy call

Sell call

Buy put

Sell put

1,670
1,909
1,395
98

1,390
1,709
1,091
120

1,400
1,738
1,086
115

1,179
1,634
839
120

1,164
1,503
874
102

17.45
273,102
4,765,903
1,031.36

17.74
74,114
1,314,526
254.07

16.16
75,321
1,217,082
246.89

18.80
49,145
924,072
223.15

17.67
47,147
832,880
206.45

51.94
33.91
14.14
36.64
27.91
35.45

13.21
10.11
4.26
10.25
7.78
9.55

11.77
9.52
4.25
9.22
7.05
9.27

11.62
5.75
2.03
7.12
5.46
6.81

10.21
5.21
2.06
6.46
4.82
6.20

7.15
0.05
4.70
29.66
27.43
2.50
11.59
16.92

1.93
0.01
1.24
8.76
7.06
0.64
2.69
4.79

1.82
0.01
1.17
8.05
6.88
0.65
2.62
4.40

1.32
0.01
1.01
5.79
5.58
0.45
2.08
3.22

1.28
0.01
0.97
5.06
4.97
0.44
1.90
2.91

Panel B: Transaction statistics

Mean
Mean
Mean
Mean

trade size
daily number of trades
daily trade volume
daily premium (in millions)

Panel C: Percentages of trading volume by moneyness and maturity

OTM: jdeltaj o 37:5%
ATM: 37:5% r jdeltajr 62:5%
ITM: jdeltaj4 62:5%
10 o T r 30 days
31 o T r 60 days
T 4 60 days
Panel D: Percentages of trading volume by exchange
American Stock Exchange
Better Alternative Trading System
Boston Options Exchange
Chicago Board Options Exchange
International Securities Exchange
Nasdaq
New York Stock Exchange ARCA
Philadelphia Stock Exchange

4.1. Options market activity

The options transaction data of 611 trading days in the
period of April 2008 to August 2010 come from Trade Alert
LLC, a specialized data vendor for the options market.
The data comprise all of the trade messages recorded by
the Options Price Reporting Authority (OPRA), a national
information processor that consolidates market information generated by option trading on all of the US options
exchanges. In real time, Trade Alert matches each option
transaction record with the underlying stock price and
computes the implied volatility from a binomial tree.
Following the quote rule, Trade Alert also classifies an
option trade as buyer-initiated (seller-initiated) if the
transaction price is above (below) the most recent mid
quote price. Such comprehensive trade data enable a
cross-sectional study over a relatively long period.
The quote rule fails to classify a trade when the trade
price falls at the mid quote or when no valid quote exists.
In such cases, I apply the tick rule: If the trade price is
above (below) the last different trade price, it is classified
as buyer-initiated (seller-initiated). This procedure of applying

the tick rule after the quote rule first appears in Lee and
Ready (1991).
The analysis excludes all indexes, units, American
Depositary Receipts (ADRs), Real Estate Investment Trusts
(REITs), closed-end funds, exchange-traded funds (ETFs),
and foreign firms and focuses on common stocks only.
For computing the OOI, I exclude the options that expire
within ten calendar days and the following trades:
(i) off-hour trades, (ii) trades at the market open (the first
15 minutes) and the market close (the last five minutes),
(iii) trades that are reported to the OPRA as late or
cancel, (iv) trades flagged as part of a structured trade,
and (v) data errors such as trades with zero prices or zero
strike prices. I filter out (i)(iii) because the trade direction
classification is less reliable for those trades; (iv) because
these trades are less likely to contain stock price information than volatility information; and (v) for obvious
reasons. Also excluded are the days with a stock-split or
with dividends because of their complex implications for
option pricing and trading.
Table 1 presents the primary statistics of the full
sample of options as well as those of four transaction-type

632

J. Hu / Journal of Financial Economics 111 (2014) 625645

groups: buy call, sell call, buy put, and sell put. Some
trades do not fit into any of the four transaction-type
groups. I do not report statistics for these trades because
they are not of interest to the study.7 On average, 1,670
stocks per day have valid option trades, with the maximum at 1,909 and the minimum at 1,395. The call options
are traded more often than the put options because the
average daily number of firms with call transactions
exceeds the average daily number of firms with put
transactions in both the buy and the sell categories in
Panel A. Trade level statistics are reported in Panel B. The
average trade size is 17.45 lots in the full sample. The call
option trades are, on average, smaller than the put option
trades. However, the call options have a much larger
average daily number of trades than the put options.
Therefore, the average daily volume of the call options
also exceeds that of the put options by approximately 0.77
million lots. The average premium of all single-name
options on a day exceeds $1 billion. Lakonishok, Lee,
Pearson, and Poteshman (2007) find that non-marketmaker participants take net long positions in option
trading volumes from 1990 to 2001. Panel B confirms their
finding by showing that for both call and put options, there
are more buy trades than sell trades in terms of both
volume and premium. Panel C presents the distribution
of the trading volumes (lots of options traded) across the
option moneyness and the time to expiration. Across the
moneyness, the OTM options are the most frequently
traded, accounting for over 50% of the entire market0 s
volume; the ATM options account for 33.91% of the total
volume; and the ITM options have the smallest share, with
only 14.14%. The same pattern exists for the four transactiontype groups. In addition, across the three moneyness regions,
the non-market-maker participants take net long positions
except for the ITM puts. The difference between the buy and
sell volumes is the largest for the OTM options and the
smallest for the ITM options. Across maturities, approximately 37% of the option transactions expire in 30 calendar
days. The options that expire in 3160 days account for
27.91% of the total volume. Given the fact that the analysis
excludes all options expiring in less than ten days, most
trades clearly are short-term options. Again, the same
pattern exists for the four transaction-type groups, and the
buy volume always exceeds the sell volume.
At the beginning of the sample period, there were
seven options exchanges in the OPRA plan: the American
Stock Exchange (AMEX), the Boston Options Exchange
(BOX), the Chicago Board Options Exchange (CBOE), the
International Securities Exchange (ISE), Nasdaq, the New
York Stock Exchange ARCA (NYSE), and the Philadelphia
Stock Exchange (PHLX). During the sample period, another
participant exchange, the Better Alternative Trading System (BATS), joined the OPRA on February 26, 2010. The last
panel of Table 1 presents the market share by exchanges.
The CBOE and the ISE lead in market share, as each of

7
The trade direction is assigned to zero for these trades. Therefore,
unclassified trades have no impact on the OOI in the empirical analysis.
However, omitting these trades can make a full-sample statistic different
from the sum or weighted average of the numbers in the four transactiontype groups.

them attracts more than one quarter of the total volume:

29.66% and 27.43%, respectively. The PHLX is also competitive, attracting 16.92% of the total volume, followed by
the NYSE at 11.59% and the AMEX at 7.15%. The BOX and
Nasdaq are relatively small markets; neither has more
than 5% of market share. The BATS does not have much of
a market share because it is newly established. The
Herfindahl index of the market share is 0.213, indicating
that the US options market is moderately concentrated
during the sample period.
4.2. Other data sources
The stock transaction data come from the NYSE Trade
and Quote (TAQ) database. I follow Lee and Ready (1991) to
assign stock transaction directions. However, unlike Lee
and Ready (1991), I collapse the trades that happen in the
same second into one record weighted by the dollar
volume and use the NBBO prices one second before the
trade time (t  1) for the trade signing. In addition to
excluding canceled trades and data errors, the analysis
excludes trades within the first 15 minutes and the last
five minutes of trading hours to increase the accuracy of
trade signing and to match the observation period of
the OOI.
The stock returns are calculated using the midpoints of
the closing bid and ask prices collected from the Center for
Research in Security Prices (CRSP). The number of analysts
following is extracted from the Institutional Brokers0
Estimate System (I/B/E/S), and the institutional ownership
data are from the Form 13-F filings in the Thomson Reuters
database.
To be included in the final sample, a stock-day observation must have valid stock price information from the CRSP
with share code 10 or 11 and the stock must have options
listed on an options exchange in the US. On average, the
final sample has 2,207 stocks a day.
4.3. Statistics of the main variables
The OOI and the SOI are constructed as discussed in
Sections 3.1 and 3.2. The option delta is calculated based
on the Black and Scholes (1973) model assuming a zero
interest rate and a zero dividend rate for simplicity. If no
option (stock) transaction exists on a particular day, the
OOI (SOI) is set to zero. Panel A of Table 2 presents the time
series averages of the cross-sectional statistics for the main
variables. For ease of reporting, I scale the stock returns
and all order imbalance variables by basis points. The
average TOI is close to zero (  0.782 bp) with the median
even smaller at 0.348 bp. The standard deviation of the
TOI reaches 21 bp, which is approximately 27 times its
mean. As a result, the TOI is not significantly different from
zero, which indicates that the stock market is balanced
overall. In extreme cases, the maximum imbalance goes
beyond 2.7% of all shares outstanding and the minimum is
below  3.98%. The SOI has similar statistics to those of the
TOI with a mean of  0.773 bp and a standard deviation of
21.71 bp. The OOI is even smaller, on average, with a mean
of  0.016 bp but the standard deviation is not that small
(5.517 bp). The skewness of the OOI (0.272) is also much

J. Hu / Journal of Financial Economics 111 (2014) 625645

633

Table 2
Descriptive statistics of the main variables.
Panel A reports the time series averages of the cross-sectional statistics. Ret is the daily stock return calculated by using the midpoint of the last National
Best Bid and Offer (NBBO) prices before the market close. TOI is the total stock order imbalance. SOI is the stock order imbalance independent of option
trading. OOI is the options order imbalance measured as the delta imbalance. Spread is the bidask spread as percentage of the mid quote in the stock
market. Turnover is the total stock trading volume scaled by the number of shares outstanding, reported as percentages. Volstock and Voloptions are the log
total stock volume and the log total options volume from 9:45 a.m. to 3:55 p.m., respectively. Ret, TOI, SOI, and OOI are reported as basis points. Panel B
presents the time series averages of the contemporaneous cross-sectional correlations. Panel C gives the cross-sectional averages of autocorrelations up to
five lags for each variable.
Panel A: Descriptive statistics
Variable

Mean

Std

Median

Minimum

Maximum

Skew

Kurt

Ret
TOI
SOI
OOI
Spread
Turnover
Voloptions
Volstock

1,348,555
1,348,555
1,340,407
1,340,407
1,348,555
1,348,552
1,348,555
1,346,313

4.602
 0.782
 0.773
 0.016
0.279
1.375
7.900
13.195

407.019
21.178
21.710
5.517
0.636
1.923
4.593
1.558

 6.005
 0.348
 0.343
 0.001
0.131
0.945
9.003
13.083

 3269.897
 397.915
 403.796
 106.223
 0.550
0.018
0.000
7.623

7193.655
269.771
282.378
109.196
14.153
46.133
17.667
19.464

2.698
 3.276
 3.058
0.272
9.929
10.558
 0.658
0.249

111.010
181.018
182.394
327.125
180.013
238.320
 0.663
0.149

Ret

TOI

SOI

OOI

Spread

Turnover

Voloptions

1
0.211
0.188
0.082
 0.007
0.068
0.027
0.025

1
0.962
0.067
 0.040
 0.072
0.019
0.008

1
 0.188
 0.040
 0.071
0.019
0.006

1
0.000
0.001
 0.000
0.005

1
 0.034
 0.258
 0.233

1
0.325
0.434

1
0.674

Ret

TOI

SOI

OOI

Spread

Turnover

Voloptions

 0.021
 0.031
0.016
 0.018
 0.028

0.105
0.053
0.035
0.028
0.026

0.101
0.051
0.034
0.028
0.026

0.028
0.009
0.006
0.004
0.003

0.170
0.157
0.152
0.153
0.152

0.436
0.309
0.253
0.226
0.207

0.322
0.265
0.233
0.215
0.202

Panel B: Correlations

Ret
TOI
SOI
OOI
Spread
Turnover
Voloptions
Volstock

Volstock

Panel C: Autocorrelations
Lag
1
2
3
4
5

smaller than that of the SOI (  3.058). Compared with the

stock market, the options market is a more balanced
market. Although the tails of the OOI are not as long as
those of the SOI, the OOI has a larger excess kurtosis (327)
than the SOI (182), indicating that both have fat tails. The
average size of the bidask spread is 0.28% of the midpoint
price in the stock market. It is interesting to compare
the imbalance variables with the turnover ratio because
the turnover ratio is also scaled by the number of shares
outstanding. The average turnover ratio is approximately
1.38%. At the ratio0 s maximum, over 46% of the shares
outstanding change hands on a given day. The stock and
options volumes are reported as log total volumes traded.
The average log stock volume is around 13.2, and the
average log options volume is 7.9. It is obvious that the
options market is still not the same size as the stock
market.
Panel B reports the time series averages of the daily
cross-sectional correlations. The stock returns have large
and positive contemporaneous correlations with all of the
order imbalance variables, and the correlation with the TOI
reaches 0.211. These positive correlations suggest that the

Volstock
0.358
0.295
0.256
0.239
0.219

imbalances have a strong impact on the contemporaneous

stock prices. The SOI has a larger contemporaneous correlation with the returns (0.188) than the OOI does (0.082).
The correlation between the TOI and the SOI is 0.962, but
the correlation between the TOI and the OOI is only 0.067,
suggesting that the majority of the TOI is determined
by the stock market investors. The correlation between
the SOI and the OOI is negative (  0.188); 50.63% of the
observations show the SOI and the OOI with different
signs. The negative correlation could be because of an
underestimation of structured trades in the sample, such
as the covered calls and the protected puts. For those
trades, the customers0 hedging transactions in the stock
market take the opposite direction of their stock exposure
in the options market. If the independent stock transactions are perfectly balanced, then the SOI has a different
sign from the OOI as a result of the structured trading. The
OPRA flags the structured trades when the customers
disclose such information to their brokers, and all of these
trades are excluded in the analysis. However, if customers
design the structured trades on their own, then these
trades cannot be identified. Thus, they weaken the return

634

J. Hu / Journal of Financial Economics 111 (2014) 625645

Table 3
Comparing the options order imbalance and the stock order imbalance.
This table compares the magnitude of the options order imbalance
(OOI) and the stock order imbalance (SOI). Two ratiosone original and
one using absolute valuesare calculated in the full sample, and detailed
statistics are presented. Both of the ratios are winsorized at 0.5%
and 99.5%.
Statistic
Minimum
1st percentile
10th percentile
25th percentile
Median
75th percentile
90th percentile
99th percentile
Maximum
Mean
Standard deviation

OOI/SOI

jOOI=SOIj

 19.689
 3.354
 0.309
 0.040
 0.000
0.028
0.202
3.088
25.094
 0.028
0.986

0
0
0
0.002
0.033
0.174
0.629
6.388
35.898
0.341
1.360

predictability by adding noise to the imbalance measures.

The imbalance variables have small correlations with the
spread, the turnover, and the trading volumes. The stock
market and the options market tend to be active at the
same time because the correlation between the stock and
the options volumes is 0.674.
Panel C presents the cross-sectional averages of the
autocorrelations for each variable up to five lags. The
autocorrelations of the TOI and the SOI are strong and
positive. For example, the autocorrelation of the TOI is
0.105 at the first lag, and it gradually decays to 0.026 at the
fifth lag. By comparison, the autocorrelation of the OOI is
much smaller. It is only 0.028 at the first lag and almost
dies out at the fifth lag. The bidask spread, the turnover
ratio, and the stock and options volumes all have highly
positive and persistent autocorrelations.
The relation between the OOI and the SOI is further
examined in Table 3. This table shows the statistics of
the two OOI-to-SOI ratios: one original and one using the
absolute values. To address the ratio explosion when the
SOI is close to zero, I winsorize both ratios at the 0.5% and
99.5% levels. The original OOI-to-SOI ratio has a mean
of  0.028. The standard deviation of the ratio is 0.986.
The average absolute ratio is 0.341, but the median is only
0.033. Even the 90th percentile absolute ratio is smaller
than one (0.629). Clearly, the OOI is small compared with
the SOI.
5. Empirical results
This section tests the hypotheses in Section 3.
5.1. Predicting stock returns using the SOI and the OOI
The empirical analysis investigates the predictive ability of the SOI and the OOI for stock returns as detailed in
Section 3.3. Table 4 contains the main results of the Fama
and MacBeth (1973) regressions. The table reports the
average slope coefficients estimated from the crosssectional regressions and the Newey and West (1987)

adjusted t-statistics. The demeaned coefficient estimates

show little autocorrelation. For many series, the autocorrelation estimates are insignificant even at the first lag. The
maximum length of a significant lag for all series is three.
To be conservative in the reported t-statistics, I choose
to use three lags for all of the Newey and West standard
error calculations. In preparation for the full model test of
Eq. (3), I test the predictive ability of the TOI in the first
column by estimating the following equation:
Reti;t TOIi;t  1 i;t :

The TOI has an average slope coefficient of  0.002, and

the t-statistic is  0.05. These values suggest that the TOI
has no significant predictive power. Column 2 presents the
regression results from using decomposed order imbalances. The OOI coefficient of 0.59 is statistically significant
at the 1% level (t-statistic 5.59). The SOI coefficient is
positive but insignificant (t-statistic 0.54). Only the OOI
positively and significantly predicts the next day0 s returns.
I then add the lagged order imbalance variables to the
model to investigate whether the predictive relations
reverse in longer horizons. Column 3 shows that the SOI
on day t 2 has a significant and negative coefficient of
 0.106 (t-statistic  2.66), but the lagged OOIs have only
insignificant coefficients. The full model in Eq. (3) is
examined in Column 4. With the microstructure control
variables, the OOI still significantly predicts the next day0 s
returns (t-statistic 4.17). The reversal effect of the SOI
becomes less significant as the t  2 SOI coefficient drops
to 0.065 (t-statistic 1.9). I also find that both of the
stock and options trading volumes significantly predict
returns, but in opposite directions. A large stock volume is
associated with positive future returns, but a large options
volume is associated with negative returns. This finding
is consistent with Johnson and So (2012), which shows
options-to-stock volume ratio negatively predicts stock
returns.
Some concerns might exist that the sample period
covers the major financial crisis in 2008 when unusual
market volatility and frequent regulatory intervention
could have distorted the generality of the results. For a
robustness check, the full model is reestimated using data
from 2009 and 2010 only. The results are reported in
Column 5. Comparing Columns 4 and 5, I find that the
results are similar, except that, in the after-crisis period,
the first lag of the SOI becomes marginally significant.
To make sure that the results are not driven by the choice
of the time window for constructing the order imbalances,
I also measure the SOI and the OOI at one hour before the
market close and repeat the full regression in Column 6.
The results are largely the same.
The results in this table have important implications for
the first two hypotheses. On the one hand, the SOI does
not positively predict future returns, although Table 2
shows that the contemporaneous correlation is strong.
The reversal effect on day t 2 indicates that the price
impact of the SOI is more likely to come from transitory
price pressure rather than from an information flow. On
the other hand, the OOI positively predicts the stock
returns on the next day, and the predictability does not
reverse in longer horizons. These findings thereby provide

J. Hu / Journal of Financial Economics 111 (2014) 625645

635

Table 4
Daily regressions of stock returns on lagged stock and options order imbalances.
The first column reports the Fama and MacBeth regression results of the following equation:
Reti;t TOIi;t  1 i;t :
Reti,t is stock i0 s return on day t calculated by using the midpoint of the last National Best Bid and Offer (NBBO) quote before the market close. TOIi,t  1 is
stock i0 s total order imbalance on day t  1. The rest of the columns present the Fama and MacBeth regression results based on the following equation:
5

k1

k1

Reti;t k1 SOIi;t  k k2 OOIi;t  k X i;t  1 i;t :

SOIi,t  k is stock i0 s order imbalance independent of option trading on day t  k. OOIi,t  k is the options order imbalance measured as the delta imbalance.
Xi,t  1 is a set of control variables on day t  1, including Reti,t  k, the stock returns for the previous five days; OpReti,t  k, the equally weighted options
returns across all of the option contracts on stock i for the previous five days; Spread, the percentage stock bidask spread; Turnover, the ratio of total stock
trading volume to the number of shares outstanding; Volstock, the log stock volume from 9:45 a.m. to 3:55 p.m.; and Voloptions, the log options volume from
9:45 a.m. to 3:55 p.m. Column 5 gives the regression results in the subsample period from 2009 to 2010. Column 6 reports the regression results by using
order imbalances measured at 3 p.m. every day. Standard errors are calculated with the Newey and West adjustment to three lags. The resulting t-statistics
are reported in parentheses. nnn, nn, and n denote statistical significance at the 1%, 5%, and 10% level, respectively.
Variable
Intercept
TOIt  1

(1)
4.957
(0.51)
 0.002
(  0.05)

SOIt  1

(2)
3.978
(0.41)

3.899
(0.40)

0.023
(0.54)

0.037
(0.85)
 0.106nnn
(  2.66)
0.046
(1.02)
0.057
(0.60)
 0.059
(  0.65)
0.531nnn
(5.12)
0.036
(0.27)
 0.070
(  0.71)
 0.058
(  0.49)
 0.018
(  0.16)

SOIt  2
SOIt  3
SOIt  4
SOIt  5
0.590nnn
(5.59)

OOIt  1

(3)

OOIt  2
OOIt  3
OOIt  4
OOIt  5
Rett  1
Rett  2
Rett  3
Rett  4
Rett  5
OpRett  1
OpRett  2
OpRett  3
OpRett  4
OpRett  5
Spread
Turnover
Volstock
Voloptions
No. of obs
Adj. R-squared

1,348,290
0.008

1,342,322
0.009

1,297,690
0.021

(4)

(5)
nnn

(6)
nn

 35.116
(  2.90)

 26.165
(  2.09)

 29.801nn
(  2.53)

0.019
(0.48)
 0.065n
(  1.90)
0.034
(1.12)
 0.008
(  0.27)
 0.028
(  0.38)
0.402nnn
(4.17)
0.077
(0.62)
 0.075
(  0.79)
 0.016
(  0.16)
 0.029
(  0.26)
0.003
(0.71)
 0.004
(  1.15)
 0.005
(  1.39)
 0.001
(  0.30)
 0.003
(  0.62)
 1.468
(  0.87)
1.059
(0.64)
0.686
(0.45)
1.092
(0.80)
 0.787
(  0.55)
4.258
(1.32)
0.008
(0.88)
2.773nnn
(3.52)
 0.763nnn
(  3.41)
1,294,304
0.067

0.073n
(1.65)
 0.054n
(  1.67)
0.027
(0.89)
 0.041
(  1.11)
 0.087
(  0.84)
0.336nnn
(2.96)
0.088
(0.72)
 0.090
(  0.83)
 0.094
(  0.89)
 0.088
(  0.72)
0.006
(1.21)
 0.006
(  1.37)
 0.005
(  1.19)
 0.001
(  0.37)
0.000
(  0.04)
 1.581
(  0.96)
0.517
(0.31)
1.816
(1.09)
1.098
(0.91)
 0.528
(  0.34)
8.342nn
(1.96)
0.000
(  0.01)
2.791nnn
(3.18)
 0.671nn
(  2.44)
892,120
0.063

0.120nnn
(2.73)
 0.064n
(  1.72)
0.045
(1.17)
0.002
(0.06)
0.042
(1.15)
0.641nnn
(3.57)
0.145
(0.59)
 0.043
(  0.30)
 0.157
(  0.94)
 0.169
(  1.06)
0.002
(0.54)
 0.004
(  1.15)
 0.005
(  1.49)
 0.001
(  0.34)
 0.002
(  0.54)
 1.689
(  0.99)
1.243
(0.75)
0.833
(0.54)
1.096
(0.81)
 0.829
(  0.58)
4.524
(1.40)
0.011
(1.23)
2.341nnn
(3.01)
 0.679nnn
(  3.06)
1,294,304
0.067

636

J. Hu / Journal of Financial Economics 111 (2014) 625645

Table 5
Excess returns from order imbalance strategies.
This table reports the average daily returns on the equally weighted
quintile portfolios based on the total order imbalance (TOI), the stock
order imbalance (SOI), and the options order imbalance (OOI) as well as
excess returns from the longshort portfolios. The portfolios are rebalanced every day at the market close based on the order imbalance
signals calculated at 3:55 p.m. The excess returns from the longshort
portfolios are reported in three forms: the original, the Fama and French
(FF) three-factor adjusted, and the four-factor adjusted returns. All
returns are reported as basis points. Sharpe is the annualized Sharpe
ratio. The resulting t-statistics are reported in parentheses. nnn, nn, and n
denote statistical significance at the 1%, 5%, and 10% level, respectively.
Quintile

TOI

SOI

OOI

Low1
2
3
4
High5

8.720
3.511
0.510
0.856
11.333

8.929
2.739
1.804
3.411
10.396

2.095
4.405
7.018
4.908
10.831

51

2.613
(1.08)
2.552
(1.07)
2.268
(0.95)
0.696

1.467
(0.81)
1.486
(0.85)
1.228
(0.71)
0.426

8.736nnn
(6.03)
8.787nnn
(6.28)
8.600nnn
(6.16)
3.754

FF3 alpha
FF4 alpha
Sharpe

unambiguous evidence that the options order flow contains a significant amount of private information about the
underlying stock0 s price movement.
Next, I perform an investment analysis to gauge the
economic significance of the return predictability. On each
day, the equally weighted quintile portfolios are formed
based on each order imbalance variable. The top quintile portfolio with the most negative order imbalance is
defined as the sell portfolio, and the bottom quintile portfolio with the most positive order imbalance is defined
as the buy portfolio. A zero-investment portfolio is constructed by buying all of the stocks in the buy portfolio and
selling all of the stocks in the sell portfolio at the market
close. All of the portfolios are rebalanced the next day.
Because all of the order imbalance variables are measured
daily at 3:55 p.m., each strategy has five minutes for
execution every day.
Table 5 gives the portfolio returns. The TOI strategy
generates a V-shaped pattern in quintile portfolio returns,
and the long-short portfolio does not generate a significant
return (t-statistic 1.08). The average returns on the quintile portfolios based on the SOI have the same V-shaped
pattern, and no significant return exists for the long-short
portfolio either. The returns on the OOI portfolios increase
almost monotonically across the quintile portfolios. The
sell portfolio has an average return of 2.095 bp per day,
and the buy portfolio has an average return of 10.831 bp.
The daily excess return reaches 8.736 bp (t-statistic 6.03),
which amounts to 22% annually. The excess return is
significant at the 1% level after controlling for the Fama
and French risk factors and the momentum factor. The
annual Sharpe ratio of the OOI strategy reaches 3.754. The
findings in this table show that the return predictability
from the OOI is economically significant.

5.2. An intraday analysis

To increase the number of observations given the
limited sample length, this subsection investigates the
high-frequency return predictability from the SOI and the
OOI at half-hour intervals. The empirical test is based on
the following equation:
2

Rett;t 1 k1 SOIt  k;t  k 1

k1
2

k2 OOIt  k;t  k 1 Y t t 1 ;
k1

where Rett;t 1 is the stock return calculated using the

midpoint of the last NBBO prices in the interval from time t
to t 1 in which the time unit is half an hour. Yt contains
the NBBO returns for the last two periods, the percentage
bidask spread at time t, and the log stock volume and the
log options volume in the last half hour. The firm subscription is omitted. The analysis excludes the first half
hour after the market open and the last hour before the
market close and does not use the order imbalances from
the previous day in the regressions. I choose two lags of
the imbalance variables in Eq. (6) to balance the goals of
detecting potential reversal effects and preserving the
sample size without missing values. Therefore, the analysis
starts in the fourth half hour of trading (11:00 a.m. to
11:30 a.m.) and ends in the 12th half hour (3:00 p.m. to
3:30 p.m.) because of the need for the lagged order
imbalance variables. Each stock then has nine half-hour
observations that can be used for estimating Eq. (6) on
each day.
Column 1 of Table 6 gives the Fama and MacBeth
regression results in the full sample. The SOI positively
predicts the next half-hour returns with a coefficient
estimate of 0.082 (t-statistic 2.12). However, the predictive relation becomes significantly negative (tstatistic 3.86) in the following half hour. The
SOIt  2;t  1 coefficient is  0.075almost offsetting the
impact of the SOIt  1;t completely. The OOIt  1;t coefficient
is 0.353 (t-statistic 3.48), and the second lag OOI has an
insignificant coefficient (t-statistic  0.18). The rest of the
columns show regression results for each half-hour interval separately. For example, Column 2 gives the time series
averages of the coefficients estimated in the crosssectional regressions that use only the first observation
of the day (11:00 a.m. to 11:30 a.m.). Out of the nine halfhour subsamples, both the SOIt  1;t and the SOIt  2;t  1 have
three coefficients at the 5% significance level, and the
OOIt  1;t has six coefficients at the 5% significance level.
The high frequency results confirm the previous findings
at the daily frequency that the OOI has a permanent price
impact and that the SOI generates only transitory price
pressure.
5.3. Conditional analysis
Having established the link between the options order
flow and the future stock prices, the analysis now turns to
examining the return predictability from the OOI conditioning on option contracts and firm characteristics.

J. Hu / Journal of Financial Economics 111 (2014) 625645

637

Table 6
Predicting stock returns using order imbalances at half-hour intervals.
This table investigates intraday relations between the stock returns and the order imbalances. Each trading day is divided into 13 half-hour intervals. The
Fama and MacBeth regression results of the following equation are presented for the full sample as well as for the nine half-hour intervals separately:
2

k1

k1

Rett;t 1 k1 SOIt  k;t  k 1 k2 OOIt  k;t  k 1 Y t t 1 :

The dependent variable Rett,t 1 is the stock return calculated by using the midpoint of National Best Bid and Offer (NBBO) prices from time t to t 1, and
the time unit is half an hour. SOIt  1,t is the stock order imbalance independent of option trading. OOIt  1,t is the options order imbalance measured as the
delta imbalance. Yt is a set of control variables at time t, including Rett  2,t  1 and Rett  1,t, the NBBO returns for the last two periods; Spread, the percentage
stock bidask spread; StVolt  1,t, the log stock volume in the last half hour; and OpVolt  1,t, the log options volume in the last half hour. The firm
subscription is omitted for all variables. The regressions do not use order imbalances from the previous day. The first three intervals and the last interval of
a day are excluded from the regression analysis because of missing variables. The resulting t-statistics are reported in parentheses. nnn, nn, and n denote
statistical significance at the 1%, 5%, and 10% level, respectively.
Variable

All
(1)

t 11:00
(2)

t 11:30
(3)

t 12:00
(4)

t 12:30
(5)

t 13:00
(6)

t 13:30
(7)

t 14:00
(8)

t 14:30
(9)

t 15:00
(10)

Intercept

 2.488nnn
(  3.78)
0.082nn
(2.12)
 0.075nnn
(  3.86)
0.353nnn
(3.48)
 0.017
(  0.18)
 0.020nnn
(  14.54)
 0.005nnn
(  3.94)
 1.773nnn
(  6.82)
0.242nnn
(4.12)
 0.058nnn
(  4.25)

 5.249nnn
(  2.77)
0.093nn
(2.04)
 0.106nnn
(  2.93)
0.098
(0.47)
 0.019
(  0.14)
 0.023nnn
(  6.54)
 0.011nnn
(  4.34)
 2.678nnn
(  3.70)
0.234
(1.34)
 0.093nn
(  2.27)

 2.448
(  1.34)
0.117nnn
(2.69)
 0.056
(  1.50)
0.540nnn
(3.16)
0.067
(0.43)
 0.011nnn
(  3.00)
 0.002
(  0.68)
 1.017
(  0.92)
0.190
(1.13)
 0.009
(  0.16)

 4.354nn
(  2.58)
0.090n
(1.82)
 0.070nn
(  2.02)
0.483nnn
(3.02)
0.071
(0.49)
 0.028nnn
(  7.46)
 0.006nn
(  2.32)
 0.620
(  1.04)
0.383nnn
(2.73)
 0.114nnn
(  3.61)

 2.789nn
(  2.00)
0.015
(0.33)
 0.068n
(  1.76)
0.314nn
(2.01)
0.091
(0.65)
 0.017nnn
(  4.54)
 0.001
(  0.19)
 0.806
(  1.39)
0.189
(1.47)
 0.033
(  1.11)

0.396
(0.13)
0.175nnn
(2.74)
 0.017
(  0.32)
0.446nnn
(2.79)
 0.168
(  0.89)
 0.034nnn
(  7.30)
 0.012n
(  1.92)
 1.546nn
(  2.00)
0.078
(0.29)
 0.086nn
(  2.41)

 1.004
(  0.69)
0.062
(1.45)
 0.104nnn
(  2.73)
0.166
(0.22)
 0.010
(  0.05)
 0.013nnn
(  3.35)
 0.002
(  0.56)
 1.801nnn
(  2.78)
0.032
(0.23)
 0.027
(  0.55)

 3.382nn
(  2.14)
0.067
(0.53)
 0.003
(  0.07)
0.509nnn
(3.05)
 0.402
(  0.54)
 0.034nnn
(  7.64)
 0.008n
(  1.93)
 2.815nnn
(  4.34)
0.211
(1.46)
 0.027
(  0.79)

 1.678
(  0.96)
 0.197
(  1.37)
 0.070
(  0.77)
0.271n
(1.72)
0.309n
(1.94)
 0.006
(  1.54)
 0.001
(  0.13)
 1.584nn
(  2.35)
0.466nnn
(3.26)
 0.039
(  1.17)

 1.885
(  0.88)
0.318
(1.02)
 0.176
(  1.55)
0.351nn
(2.25)
 0.092
(  0.59)
 0.014nnn
(  3.06)
 0.004
(  0.93)
 3.088nnn
(  4.04)
0.392nn
(2.21)
 0.097nn
(  2.65)

SOIt  1,t
SOIt  2,t  1
OOIt  1,t
OOIt  2,t  1
Rett  1,t
Rett  2,t  1
Spreadt
StVolt  1,t
OpVolt  1,t

5.3.1. Option leverage

Table 7 contains the regression results from Eq. (4).
I first show the regression results that use the OOI for each
option moneyness group separately. Columns 1 and 2
show that the OOI constructed by using only the OTM
options does not predict returns with or without the
control variables. This OOI measure is not informative,
although the OTM options provide the options investors
with the highest leverage. Columns 36 show that the OOI
constructed by using either the ATM options or the ITM
options significantly predicts the returns on the next day.
The results hold when the OOIs of the three moneyness
groups are examined together in Column 7, and when
the full control set is added to the model in Column 8.
Therefore, the analysis finds no empirical support for
Hypothesis 3. A possible explanation is that many institutional investors use the OTM options to gain volatility
exposures, and the order imbalance does not contain
much information about the underlying stock price.
Because the volatility traders usually perform delta hedging, their hedging transactions cancel out the options
market makers0 hedging transactions in the stock market.
There is little price pressure on the underlying stock either.
Another possibility is that the transaction costs associated
with the OTM options are too high. During the sample
period, the average percentage bidask spread for the OTM
options reaches 11.4%. For the ATM and the ITM options,
the percentage bidask spread is only 3.46% and 3%,

respectively. The spread difference in the option moneyness is consistent with the theoretical prediction of
Johnson and So (2012). If informed traders buy the OTM
options to gain higher leverage, then they are likely to
reverse the trade positions rather than exercise the
options to take profits. In doing so, they pay the full
amount of the bidask spread, which thereby sharply
increases their trading costs. The results in Table 7 suggest
that the consideration of the transaction costs can outweigh the desire for leverage when informed traders
allocate their orders across the options with different
moneyness.
5.3.2. Level of information asymmetry
This subsection tests Hypothesis 4that the OOI has
greater return predictability for firms with more information asymmetry. Based on each of the five proxies for
information asymmetry, I divide the full sample into three
groups: low, medium, and high. I then run full specification regressions for Eq. (3) in each group. Because the
main interest of the test is the predictive ability of the OOI
in subgroups, I report in Table 8 only the coefficient
estimates and the t-statistics of the OOIt  1 . Panel A shows
that the OOI coefficient is significant at the 1% level
(t-statistic 2.85) in the high PIN group, marginally significant in the medium PIN group (t-statistic 1.65), but
not significant in the low PIN group. Panel B shows that the
OOI coefficient is significant at the 1% level (t-statistic2.93)

638

J. Hu / Journal of Financial Economics 111 (2014) 625645

Table 7
Return predictability from options order imbalance by option moneyness.
This table presents the Fama and MacBeth regression results of the following equation:
5

k1

k1

k1

k1

Reti;t k1 OTM_OOIi;t  k k2 ATM_OOIi;t  k k3 ITM_OOIi;t  k k4 SOIi;t  k X i;t  1 i;t :

Reti,t is stock i0 s return calculated by using the midpoint of the last National Best Bid and Offer (NBBO) prices before the market close on day t. OTM_OOIi,
ATM_OOIi,t  k, and ITM_OOIi,t  k are the options order imbalances calculated by using the out-of-the-money (OTM), at-the-money (ATM), and in-themoney (ITM) option contracts, respectively. SOI is the stock order imbalance unrelated to options. Xi,t  1 is a set of control variables on day t  1, including
Reti,t  k, the stock returns for the previous five days; OpReti,t  k, the equally weighted options returns across all of the option contracts on stock i for the
previous five days; Spread, the percentage stock bidask spread; Turnover, the ratio of total stock trading volume to number of shares outstanding; Volstock,
the log stock volume from 9:45 a.m. to 3:55 p.m.; VolOTM, VolATM, and VolITM, the log trading volumes of the OTM, ATM, and ITM options from 9:45 a.m. to
3:55 p.m., respectively. Standard errors are calculated with the Newey and West adjustment to three lags. The resulting t-statistics are reported in
parentheses. nnn, nn, and n denote statistical significance at the 1%, 5%, and 10% level, respectively.
t  k,

Variable

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Intercept

5.080
(0.52)
0.578
(0.85)
 0.211
(  0.50)
0.833
(0.77)
0.727
(0.88)
 0.269
(  0.64)

 36.136nnn
(  3.01)
0.522
(0.89)
0.031
(0.08)
0.046
(0.15)
0.147
(0.46)
 0.222
(  0.61)

5.172
(0.53)

 31.924nnn
(  2.63)

5.172
(0.53)

 39.446nnn
(  3.18)

0.809nnn
(3.00)
0.524
(1.01)
0.417
(1.53)
 0.078
(  0.2)
0.057
(0.12)

0.646nnn
(2.68)
0.465
(0.95)
0.282
(1.07)
 0.359
(  1.39)
 0.120
(  0.48)

1.084nnn
(2.89)
 0.618
(  1.61)
 0.478
(  1.37)
 0.525
(  1.50)
 0.548
(  1.51)

0.747nn
(2.18)
 0.518
(  1.56)
 0.279
(  0.83)
 0.489
(  1.45)
 0.420
(  1.38)
0.002
(0.06)
 0.057n
(  1.75)
0.041
(1.39)
 0.006
(  0.22)
 0.029
(  0.40)
0.003
(0.78)
 0.004
(  1.02)
 0.005
(  1.39)
 0.001
(  0.36)
 0.003
(  0.62)
 1.772
(  1.04)
1.379
(0.83)
0.595

5.078
(0.52)
0.432
(0.72)
 0.136
(  0.34)
0.776
(0.73)
0.634
(0.85)
 0.178
(  0.44)
0.695nnn
(2.70)
0.518
(1.18)
0.348
(1.37)
 0.166
(  0.49)
0.010
(0.02)
0.922nn
(2.56)
 0.572
(  1.57)
 0.503
(  1.41)
 0.449
(  1.47)
 0.529
(  1.46)

 43.958nnn
(  3.46)
0.388
(0.80)
0.026
(0.07)
0.079
(0.25)
0.145
(0.45)
 0.174
(  0.48)
0.566nn
(2.33)
0.422
(1.02)
0.332
(1.26)
 0.341
(  1.33)
 0.132
(  0.52)
0.745nn
(2.18)
 0.459
(  1.27)
 0.375
(  1.08)
 0.374
(  1.16)
 0.384
(  1.30)
0.010
(0.25)
 0.056
(  1.74)
0.038
(1.28)
 0.005
(  0.17)
 0.032
(  0.42)
0.003
(0.65)
 0.005
(  1.24)
 0.005
(  1.46)
 0.001
(  0.26)
 0.002
(  0.59)
 1.677
(  0.98)
1.306
(0.79)
0.538

OTM_OOIt  1
OTM_OOIt  2
OTM_OOIt  3
OTM_OOIt  4
OTM_OOIt  5
ATM_OOIt  1
ATM_OOIt  2
ATM_OOIt  3
ATM_OOIt  4
ATM_OOIt  5
ITM_OOIt  1
ITM_OOIt  2
ITM_OOIt  3
ITM_OOIt  4
ITM_OOIt  5
SOIt  1
SOIt  2
SOIt  3
SOIt  4
SOIt  5
Rett  1
Rett  2
Rett  3
Rett  4
Rett  5
OpRett  1
OpRett  2
OpRett  3

0.001
(0.02)
 0.056n
(  1.77)
0.038
(1.28)
 0.006
(  0.21)
 0.029
(  0.38)
0.003
(0.80)
 0.004
(  1.14)
 0.005
(  1.45)
 0.001
(  0.37)
 0.003
(  0.64)
 1.615
(  0.95)
1.221
(0.74)
0.650

 0.001
(  0.02)
 0.051
(  1.59)
0.039
(1.33)
 0.011
(  0.39)
 0.027
(  0.36)
0.003
(0.77)
 0.005
(  1.27)
 0.005
(  1.45)
 0.001
(  0.26)
 0.002
(  0.61)
 1.711
(  1.00)
1.262
(0.75)
0.601

J. Hu / Journal of Financial Economics 111 (2014) 625645

639

Table 7 (continued )
Variable

(1)

(2)

(3)

(4)

(0.42)
1.078
(0.80)
 0.683
(  0.48)
4.453
(1.37)
0.008
(0.87)
2.699nnn
(3.52)
 0.686nnn
(  3.39)

OpRett  4
OpRett  5
Spread
Turnover
Volstock
VolOTM

(5)

(6)

(0.39)
1.156
(0.85)
 0.666
(  0.47)
4.476
(1.38)
0.008
(0.95)
2.282nnn
(2.99)

(0.39)
1.110
(0.82)
 0.678
(  0.48)
4.512
(1.39)
0.008
(0.86)
2.922nnn
(3.78)

 0.545nnn
(  2.69)

VolATM

(7)

(0.35)
1.236
(0.92)
 0.686
(  0.48)
4.365
(1.36)
0.009
(0.96)
3.394nnn
(3.97)
 0.329nn
(  2.11)
 0.158
(  1.05)
 0.654nnn
(  4.49)

 0.904nnn
(  4.70)

VolITM

(8)

Table 8
Predictive power of options order imbalance and firm characteristics.
This table investigates the return predictability from the options order imbalance for different types of firms. For each of the firm characteristics, the full
sample is divided into three groups: low ( o 30th percentile), medium (30th70th percentile), and high ( 4 70th percentile). Within each group, I estimate
the following equation using the Fama and MacBeth regressions:
2

k1

k1

Rett;t 1 k1 SOIt  k;t  k 1 k2 OOIt  k;t  k 1 X t t 1 :

All variables are the same as defined in Table 4. The average slope coefficients and t-statistics (in parentheses) are reported only for the previous day0 s
options order imbalance (OOIt  1). PIN is the probability of informed trading based on Easley and O0 Hara (1992). Analyst coverage is the number of analysts
following. Spread is the percentage stock bidask spread. GH adverse spread is the adverse selection component of the bidask spread based on the model
of Glosten and Harris (1988). LSB adverse spread is the adverse-selection component of the bidask spread based on the model of Lin, Sanger, and Booth
(1995). Size is the market capitalization. Ownership is the percentage of shares held by institutional investors. Options volume is the total options volume
traded. Standard errors are calculated with the Newey and West adjustment to three lags. The resulting t-statistics are reported in parentheses. nnn, nn, and n
denote statistical significance at the 1%, 5%, and 10% level, respectively.
Panel A: PIN
Variable
OOIt  1

Panel B: Analyst coverage

Low

Medium

High

0.219
(0.74)

0.253n
(1.65)

0.505nnn
(2.85)

Panel C: Spread
Variable
OOIt  1

OOIt  1

Low

Medium

High

0.399n
(1.66)

0.429nn
(2.12)

0.901nnn
(3.03)

OOIt  1

Medium

High

0.965nnn
(2.93)

0.455nn
(2.03)

0.416
(1.57)

Variable

Low

Medium

High

OOIt  1

0.220
(0.78)

0.358n
(1.87)

Small

Medium

1.216nnn
(2.82)

0.276
(1.24)

0.532nnn
(2.66)

Panel F: Size

Low

Medium

High

0.278
(1.44)

0.513nn
(2.09)

0.516nnn
(3.28)

Panel G: Ownership
Variable

OOIt  1

Low

Panel D: GH adverse spread

Panel E: LSB adverse spread

Variable

Variable

Variable
OOIt  1

Large
0.240n
(1.83)

Panel H: Options volume

Low

Medium

High

0.777nnn
(3.64)

0.688nnn
(2.69)

0.237n
(1.74)

in the low analyst coverage group, significant at the 5% level

in the medium coverage group (t-statistic2.03), but not
significant in the high coverage group. Panel C shows that
the OOI has significant coefficients in all of the groups

Variable
OOIt  1

Low

Medium

62.376
(1.46)

1.699nnn
(2.66)

High
0.474nnn
(4.18)

based on the bidask spread. However, the coefficient is

0.901 (t-statistic 3.03) in the high spread group, which is
much greater than the coefficient of 0.399 (t-statistic
1.66) in the low spread group. A similar pattern in the OOI

 0.758nnn
(  3.36)
2.661nnn
(3.43)
0.008
(0.87)
4.233
(1.32)
 0.488
(  0.35)
1.108
(0.81)
0.436
(0.28)
0.380
(0.23)
 0.879
(  0.50)
 0.004  0.004  0.002  0.002
(  1.24) (  1.20) (  0.56) (  0.55)
0.003
(0.62)
 0.020nnn
(  3.50)
 0.001
(  1.39)

Volstock
OpRett  1 OpRett  2 OpRett  3 OpRett  4 OpRett  5 Spread Turnover
Rett  5
Rett  4
Rett  3
Rett  2
Rett  1
S2  OOI2
S1  SOI2

0.122nn
(2.53)
33.537nnn
(  2.77)

1.111nnn
(5.75)

Voloptions

 0.630nnn
(  2.93)
2.612nnn
(3.40)
 0.007
(  0.76)
3.744
(1.17)
 0.529
(  0.38)
0.883
(0.65)
0.616
(0.40)
0.439
(0.27)
 0.785
(  0.45)
 0.004  0.002  0.002
(  1.16) (  0.55) (  0.57)
 0.004
(  1.09)
0.003
(0.71)
0.802nnn
(4.63)

Voloptions
Volstock
OpRett  1 OpRett  2 OpRett  3 OpRett  4 OpRett  5 Spread Turnover
Rett  5
Rett  4
Rett  3
Rett  2
Rett  1
OOI 
OOI

0.121
(0.71)

OOI
SOI
Intercept

where

OOIi;t
1
The results are reported in Panel A of Table 9. The 1
estimate is 0.286, statistically significant at the 1% level (tstatistic 4.24). The 2 estimate is 0.238, also significant
at the 1% level (t-statistic  3.63). These results suggest
that both the positive and negative SOIs significantly predict
stock returns, but the signs are inconsistent. Therefore, the
overall SOI does not exhibit an ability to predict stock

Panel B: Nonlinearity


maxSOIi;t  1 ; 0, SOIi;t
 1 minSOIi;t  1 ; 0,

maxOOIi;t  1 ; 0, and OOIi;t  1 minOOIi;t  1 ; 0.

 0.238nnn
(  3.63)

SOIi;t
1

 34.384nnn 0.286nnn
(  2.86)
(4.24)


4 OOIi;t
 1 X i;t  1 i;t ;

SOI 


Reti;t 1 SOIi;t
 1 2 SOIi;t  1 3 OOIi;t  1

SOI

I first investigate whether there is asymmetric price

response to the imbalances by estimating the following
equation:

Intercept

6.1. Asymmetric price response

Panel A: Asymmetric response

Section 5 shows that the options order flow contains

important price information about the underlying stock.
In this section, I perform additional analyses to better understand that information linkage.

Table 9
Nonlinear price impact from options and stock order imbalances.
Panel A presents the Fama and MacBeth regression results of the equation:

6. Further analysis


Reti;t 1 SOIi;t
 1 2 SOIi;t  1 3 OOIi;t  1 4 OOIi;t  1 X i;t  1 i;t ;

5.3.4. Market activity

Hypothesis 6 is tested in this subsection. Panel G in
Table 8 shows that the OOI does not significantly predict
the underlying returns when the options trading volume is
low. The OOI coefficients are statistically significant at the
1% level in both the medium (t-statistic 2.66) and the
high volume groups (t-statistic 4.18). This finding suggests that the predictive ability of the OOI comes from the
periods when the options market is active. The results are
consistent with the information explanation for the return
predictability of the OOI.

where S1i;t  1 and S2i;t  1 are dummy variables that equal one when SOIi;t  1 and OOIi;t  1 are positive, respectively, and negative one otherwise. Standard errors are calculated with the Newey and West adjustment
to three lags. The resulting t-statistics are reported in parentheses. nnn, nn, and n denote statistical significance at the 1%, 5%, and 10% level, respectively.

5.3.3. Institutional ownership

This subsection describes the testing of Hypothesis
5that the information content of the OOI varies with
a stock0 s institutional ownership. Panel G in Table 8 gives
the results. The OOI is statistically significant at the 10%
level in all of the ownership groups. However, the OOI
coefficient is 0.777 (t-statistic 3.64) in the low ownership
group and 0.237 (t-statistic 1.74) in the high ownership
group. The differences in the magnitude of the coefficients
and the t-statistics show that the OOI has a larger amount
of price information for the firms with low institutional
ownership. These stocks are associated with high shortsale costs. The results suggest that short-sale constraints in
the stock market increase the information content in the
options order flow.


OOIi;t
 1 maxOOIi;t  1 ; 0, and OOIi;t  1 minOOIi;t  1 ; 0. The control variable set X i;t  1 is the same as defined in Table 4. Panel B presents the regression results of the following equation:

coefficients is found in groups based on the adverseselection component of the spread in Panels D and E,
and in the groups based on the firm size in Panel F.
Collectively, these results suggest that, consistent with
Hypothesis 4, the OOI has a stronger predictive power
for returns when the firm has a higher level of information
asymmetry.

Reti;t 1 SOIi;t  1 2 OOIi;t  1 3 SOI2i;t  1  S1i;t  1 4 OOI2i;t  1  S2i;t  1 X i;t  1 i;t ;

J. Hu / Journal of Financial Economics 111 (2014) 625645


where Reti;t is stock i0 s return calculated by using the midpoint of the last National Best Bid and Offer (NBBO) prices before the market close on day t, SOIi;t
 1 maxSOIi;t  1 ; 0, SOIi;t  1 minSOIi;t  1 ; 0,

640

J. Hu / Journal of Financial Economics 111 (2014) 625645

returns. The 3 estimate is insignificant, and the 4 estimate

is significant at the 1% level (t-statistic4.63). These results
suggest that the predictive power of the OOI mainly comes
from the negative OOI, which support Hypothesis 5 that
informed traders use options to get around the short-sale
constraints in the stock market when they acquire negative
information.

This subsection investigates the nonlinearity in the

predictive relation. Both the SOI and the OOI have fat tails.
The possibility exists that the stock price0 s response is
different in the tails of the order imbalances. I estimate the
following specification:
Reti;t 1 SOIi;t  1 2 OOIi;t  1 3 SOI2i;t  1
8

where S1 and S2 are dummies that equal one when SOI

and OOI are positive, respectively, and negative one
otherwise. The interaction terms quadratically amplify
the two imbalance variables and parse out the tail effect.
Panel B of Table 9 gives the regression results. Both the SOI
and the OOI coefficients are positive and statistically
significant at the 5% level. No significant nonlinear effect
for the SOI is found because the 3 is insignificant.
However, the 4 is negative and significant at the 1% level
(t-statistic  3.5), indicating that the predictive ability of
the OOI reduces in the tails.
6.3. Moving average and shocks
Both the SOI and the OOI have positive autocorrelations. In this subsection, I decompose each imbalance
variable into a moving average (MA) component and a
shock component to further investigate the sources of the
price impact. For example,
OOI OOIMAk OOIshockk ;

where the OOIMAk is the average OOI from the previous k

trading days. For simplicity, I omit the firm and time
subscriptions in Eq. (9). The same decomposition is also
performed for the SOI. Then, I estimate the following
equation:
Reti;t 1 SOIMAi;t  1 2 SOIshocki;t  1 3 OOIMAi;t  1
4 OOIshocki;t  1 X i;t  1 i;t ;

both the MA and the shock have significant coefficient

estimates but the shock component has a larger t-statistic
(3.76) than the MA (1.79). As the MA horizon increases, the
predictive ability of the OOI MA becomes slightly weaker,
but the predictive ability of the OOI shock component
stays unchanged.
6.4. Earnings announcement

6.2. Nonlinear price impact

S1i;t  1 4 OOI2i;t  1  S2i;t  1 X i;t  1 i;t ;

641

10

Three MA periods of three days, five days, and ten days are
chosen, and the results are reported in Table 10. Panel A
shows that the correlations between the MAs and the
shocks are large and negative for both the SOI and the OOI.
For example, the three-day correlations are 0.468 and
0.514 for the SOI and the OOI, respectively. As the length
of the MA increases, the correlations become weaker.
Panel B reports the regression results from Eq. (10).
The first column reports the result from using a three-day
MA decomposition. Although the MAs and the shocks are
negatively correlated, both are positively correlated with
the future stock returns. Neither the MA nor the shock of
the SOI significantly predicts the stock returns. For the OOI,

This subsection investigates the price information in

the order imbalances around earnings announcements.
Several studies such as Skinner (1990) and Ho (1993) find
that stocks with options exhibit smaller price reactions to
earnings surprise, suggesting that option trading disseminates new information regarding earnings. Amin and Lee
(1997) show that an increase in the open interest in
options before an announcement is related to the direction
of the earnings surprise. More recently, Roll, Schwartz, and
Subrahmanyam (2010) and Johnson and So (2012) find
that the options-to-stock volume ratio predicts the earnings surprise and the cumulative abnormal returns around
the announcement. If option trading conveys private
information, then the return predictability from the OOI
should also be prominent around the earnings
announcements.
I collect quarterly earnings announcement data from
the I/B/E/S and merge the data with the options sample.
The final sample contains 21,103 observations on 2,647
firms. I follow Carhart (1997) and calculate the riskadjusted returns for each firm. Then the CARs are constructed for four alternative event windows: (0,2), (0,4),
(  1,1), and (  2,2). For example, the CAR0;2 is the threeday cumulative risk-adjusted return from the announcement day t to day t 2. Mirroring previous tests, the
analysis estimates the following equation:
5j

5j

i1

i1

CARj;k k1 SOIj  i k2 OOIj  i X j  1 j;k ;

11
where the independent variables are the same as defined
before with the firm and event subscriptions omitted and
the time subscription is relative to the announcement
day t. Table 11 presents the pooled ordinary least squares
(OLS) results with the standard errors clustered by firms.
Across different event windows, the OOI  5 always has
significant coefficient estimates and all of the other OOI
variables are insignificant. For example, in predicting the
CAR0;2 , the OOI  5 coefficient is 4.289 (t-statistic 2.64).
The predictive ability of the OOI  5 is largely the same in
the other event windows. The SOI coefficients are mostly
insignificant, except for the SOI  2 when predicting the
CAR  1;1 . The results indicate that the options market
conveys information about corporate earnings five days
before the announcement.
Informed trading should be more active when the
potential profit is high. The last empirical test investigates
whether the OOI also becomes more informative when
approaching such earnings announcements. Three proxies
for the potential profit from informed trading are used:
earnings surprise, defined as the absolute value of the

642

J. Hu / Journal of Financial Economics 111 (2014) 625645

Table 10
Moving averages and shocks of order imbalances and return predictability.
Panel A presents the time series averages of the cross-sectional correlations between the three-day, five-day, and ten-day moving averages (MA) and
the shock components for the stock order imbalance (SOI) and the options order imbalance (OOI). Panel B reports the Fama and MacBeth regression
results of the equation:
Reti;t 1 SOIMAi;t  1 2 SOIshocki;t  1 3 OOIMAi;t  1 4 OOIshocki;t  1 X i;t  1 i;t ;
where Reti;t is stock i0 s return on day t calculated by using the midpoint of the last National Best Bid and Offer (NBBO) prices before the market close and
X i;t  1 includes the stock returns and the equally weighted options returns on the previous five days, the percentage bidask spread, the ratio of total stock
volume to number of shares outstanding, the log total stock volume, and the log total options volume. Standard errors are calculated with the Newey and
West adjustment to three lags. The resulting t-statistics are reported in parentheses. nnn, nn, and n denote statistical significance at the 1%, 5%, and 10%
level, respectively.
Three-day MA

Five-day MA

Ten-day MA

 0.468
 0.514

 0.381
 0.435

 0.278
 0.343

 70.244nnn
(  4.43)
 0.033
(  0.51)
 0.014
(  0.27)
0.371n
(1.79)
0.359nnn
(3.76)
 0.002
(  0.27)
 0.006
(  1.40)
 0.005
(  1.32)
0.002
(0.44)
 0.003
(  0.82)
 1.061
(  0.56)
0.944
(0.56)
0.777
(0.47)
0.657
(0.44)
 1.114
(  0.77)
18.316nnn
(2.59)
0.007
(0.72)
 1.461nnn
(  4.65)
5.753nnn
(5.12)

 70.481nnn
(  4.44)
 0.069
(  1.00)
 0.013
(  0.26)
0.460n
(1.90)
0.360nnn
(3.79)
 0.002
(  0.28)
 0.006
(  1.44)
 0.005
(  1.23)
0.002
(0.44)
 0.003
(  0.81)
 1.129
(  0.60)
0.935
(0.55)
0.864
(0.52)
0.520
(0.35)
 1.006
(  0.70)
18.212nnn
(2.57)
0.008
(0.78)
 1.460nnn
(  4.65)
5.763nnn
(5.12)

 70.994nnn
(  4.43)
 0.142
(  1.47)
 0.015
(  0.29)
0.392
(1.44)
0.344nnn
(3.58)
 0.002
(  0.26)
 0.007
(  1.54)
 0.005
(  1.39)
0.001
(0.28)
 0.003
(  0.80)
 1.145
(  0.60)
0.709
(0.42)
0.809
(0.48)
0.450
(0.30)
 0.680
(  0.47)
18.642nnn
(2.61)
0.006
(0.67)
 1.475nnn
(  4.69)
5.868nnn
(5.16)

Panel A: Correlations
SOIMA and SOIshock
OOIMA and OOIshock
Panel B: Regression Results
Intercept
SOIMA
SOIshock
OOIMA
OOIshock
Rett  1
Rett  2
Rett  3
Rett  4
Rett  5
OpRett  1
OpRett  2
OpRett  3
OpRett  4
OpRett  5
Spread
Turnover
Volstock
Voloptions

difference between the actual earnings per share (EPS) and

the analysts0 consensus on the EPS forecasts immediately
prior to the announcement; the analysts0 dispersion,
defined as the standard deviation of the analysts0 EPS
forecasts; and the analyst coverage. The first two proxies
are event characteristics. The third proxy is a firm characteristic to capture the level of information asymmetry.
Similar to Table 8, Table 12 presents the coefficients and
the t-statistics of the OOI  5 from the subgroup regression
results. The dependent variable in these regressions is the

CAR0;2 because the alternative event windows generate

even stronger results. Panel A shows that the OOI  5
coefficient is statistically significant at the 1% level
(8.214, t-statistic 2.98) in the high earnings surprise
group, but it is not significant in the other two groups.
This finding suggests that the predictive ability of the
OOI  5 comes mainly from those events with large earnings surprises. Panel B shows that the OOI  5 significantly
predicts the CAR for the announcements with both positive and negative forecasting errors. However, the negative

J. Hu / Journal of Financial Economics 111 (2014) 625645

643

Table 11
Predicting cumulative abnormal returns (CARs) around earnings announcements using order imbalances.
This table presents the pooled ordinary least squares results of the following equation:
5j

5j

i1

i1

CARj;k k1 SOIj  i k2 OOIj  i X j  1 j;k ;

where CARj;k is the cumulative abnormal return from day j to day k relative to the earnings announcement day and X j  1 includes the stock returns and the
equally weighted options returns on the previous five days, the percentage bidask spread, the ratio of total stock volume to number of shares
outstanding, the log stock volume and the log options volume. The firm and time subscriptions are omitted for all variables. Standard errors are clustered
by firms. The resulting t-statistics are reported in parentheses. nnn, nn, and n denote statistical significance at the 1%, 5%, and 10% level, respectively.
Variable

CAR0;2

CAR0;4

CAR  1;1

CAR  2;2

Intercept

92.222
(0.98)
 1.653
(  1.13)
2.170
(1.43)
1.072
(0.75)
 0.650
(  0.53)
4.289nnn
(2.64)
 0.101
(  0.19)
0.846n
(1.76)
 0.527
(  1.23)
0.545
(1.01)
 0.590
(  0.93)
0.054
(0.02)
 0.018
(  0.53)
 0.054nn
(  2.02)
 0.008
(  0.29)
0.003
(0.10)
 3.798nnn
(  6.14)
6.589
(1.21)
 2.450nn
(  2.34)
 13.093n
(  1.72)
12.351n
(1.83)
 0.053
(  1.10)
 33.48
(  1.14)
 3.699
(  0.46)
 1.717
(  0.62)

79.531
(0.89)
 1.031
(  0.66)
2.407
(1.45)
0.429
(0.26)
 1.471
(  1.19)
4.860nnn
(2.55)
0.224
(0.43)
0.701
(1.25)
 0.508
(  1.29)
0.222
(0.40)
 0.320
(  0.49)
0.122
(0.04)
 0.021
(  0.70)
 0.076nn
(  2.52)
 0.029
(  0.94)
0.021
(0.79)
 3.760nnn
(  4.37)
2.269
(0.36)
 1.882
(  1.65)
 3.489
(  0.40)
6.378
(0.81)
 0.045
(  0.79)
 33.78
(  1.30)
 0.899
(  0.11)
 4.793
(  1.60)

145.773n
(1.70)

1.026
(0.79)
0.816
(0.59)
 0.389
(  0.31)
4.212nnn
(2.71)

1.265nn
(2.54)
 0.429
(  1.12)
0.619
(1.10)
0.077
(0.15)

 1.633
(  0.59)
 0.077nnn
(  2.72)
 0.010
(  0.39)
 0.015
(  0.62)

7.689
(1.55)
 1.550
(  1.34)
 14.239nn
(  2.06)
9.264
(1.62)
0.006
(0.12)
40.246
(1.27)
 9.834
(  1.33)
 0.147
(  0.06)

28.637
(0.27)

0.343
(0.16)
 0.596
(  0.44)
4.31nn
(2.08)

 0.516
(  0.89)
0.301
(0.42)
 0.533
(  0.58)

 5.129
(  1.54)
0.001
(0.04)
 0.032
(  1.10)

 1.534
(  1.40)
 8.508
(  1.10)
11.647n
(1.71)
0.019
(0.12)
 3.974
(  0.29)
3.775
(0.37)
 6.598n
(  1.81)

OOI  1
OOI  2
OOI  3
OOI  4
OOI  5
SOI  1
SOI  2
SOI  3
SOI  4
SOI  5
Ret  1
Ret  2
Ret  3
Ret  4
Ret  5
OpRet  1
OpRet  2
OpRet  3
OpRet  4
OpRet  5
Spread
Turnover
Volstock
Voloptions

error group has a larger coefficient estimate and a larger

t-statistic than the positive error group, possibly due to the
short-sale constraints in the stock market. Panel C shows
that the OOI  5 is not informative when the analysts0
dispersion is low. When analysts exhibit divergent

opinions, the OOI  5 significantly predicts the CAR. Panel

D shows that the return predictability is stronger for firms
with low analyst coverage, which supports the informational role of option trading. Using alternative measures of
information asymmetry generates largely the same results.

644

J. Hu / Journal of Financial Economics 111 (2014) 625645

Table 12
Order flow information and potential profit from informed trading around earnings announcements.
For each proxy for the potential profit from informed trading around earnings announcements, the full sample is divided into three groups: low ( o 30th
percentile), medium (30th70th percentile), and high ( 4 70th percentile). The slope coefficients and t-statistics (in parentheses) are reported only for the
options order imbalance five days before the announcement (OOI  5 ), but the regressions are based on the following equation with the same variables as
defined in Table 11:
5

i1

i1

CAR0;2 k1 SOI  i k2 OOI  i X  1 0;2 :

Surprise equals the difference between the actual quarterly earnings per share (EPS) and the analysts0 forecast consensus before the announcement.
Forecasting error is the signed earnings surprise defined as the actual EPS minus the analysts0 forecast consensus. Dispersion is the standard deviation of
analysts0 forecasts before the announcement. Analyst coverage is the number of analysts following. Standard errors are clustered by firm. nnn, nn, and n
denote statistical significance at the 1%, 5%, and 10% level, respectively.
Panel A: Surprise
Variable
OOI  5

Panel B: Forecast error

Low

Medium

High

0.657
(0.80)

2.983
(1.05)

8.214nnn
(2.98)

Panel C: Dispersion
Variable
OOI  5

Variable

Negative

Neutral

OOI  5

6.658nnn
(2.86)

1.321
(0.36)

Positive
4.221nn
(1.99)

Panel D: Analyst coverage

Low

Medium

High

0.875
(0.34)

8.829nnn
(3.23)

3.907n
(1.95)

7. Conclusion
In this paper, I investigate whether option trading
provides additional stock price information that is not in
the stock market. By decomposing the total stock order
imbalance into an option-induced imbalance component
and a residual component that is unrelated to option
trading, the analysis shows that the option-induced stock
order imbalance positively and significantly predicts future
stock returns and that the predictive relation does not
reverse direction in longer horizons. The remaining order
imbalance generates only temporary price pressure on the
stock price and exhibits a weak reversal effect later. The
results hold at both the daily and the half-hour frequencies. The evidence suggests that a substantial amount of
private information exists in the options order flow.
Specifically, the permanent price information comes
mainly from the option contracts with relatively narrow
bidask spreads (i.e., the ATM and the ITM options) but not
the contracts with higher leverage (i.e., the OTM options).
The paper also investigates the link between the
information content in options order flow and the level
of information asymmetry, short-sale costs, and market
liquidity. Consistent with the predictions, the paper finds
that the return predictability from the OOI is stronger for
the firms with high PINs, low analyst coverage, large bid
ask spreads, large adverse-selection components of the
spread, small market capitalizations, and low institutional
ownership. The concentration of informed traders and the
short-sale constraints enhance the informational role of
option trading in price discovery. The options order flow
also becomes more informative when option trading is
active, which supports the theoretical prediction by Easley,
O0 Hara, and Srinivas (1998) that informed trading is more
likely to occur in deep and liquid markets. An event study

Variable
OOI  5

Low

Medium

5.403nn
(2.25)

4.873n
(1.92)

High
1.379
(0.48)

shows that the OOI significantly predicts the CAR five days
before earnings announcements.
The findings indicate that the options market plays a
key role in the embedding of private information in stock
prices. However, this study does not further examine the
investors0 learning problem. It is important to understand
whether investors analyze the options order flow directly
to extract information or whether the information is
disclosed through the hedging transactions transmitted
from the OOI. The results suggest that market frictions can
exist that prevent investors from immediately learning
the stock price information in the options order flow.
For example, the cost of monitoring and collecting order
flow information can be too high for most investors. The
possibility also exists that the options order flow is too
complex for retail investors to understand.8 As a result, the
spot price does not fully adjust to the new information
from the options market immediately, and the return
predictability is preserved. For future research, it could
be interesting to explore the learning behavior of stock
market investors and investigate how such information in
the options market can be learned more efficiently.
The proposed OOI focuses on delta risk and shows
significant predictability regarding the underlying stock
price. Option contracts have other risk exposures that can
be analyzed in similar ways. It would be interesting to
investigate the information links between the derivative
markets and the underlying market in a multidimensional
framework because some risk can be traded only in the
derivative markets. Understanding that issue is important
to both academicians and practitioners.

8
In a related study, Henderson and Pearson (2011) show that retail
investors persistently overprice a popular structured equity product.

J. Hu / Journal of Financial Economics 111 (2014) 625645

References
Amin, K.I., Lee, C.M.C., 1997. Option trading, price discovery, and earnings
news dissemination. Contemp. Account. Res. 14, 153192.
Anthony, J.H., 1988. The interrelation of stock and options market
trading-volume data. J. Finance 43, 949964.
Asquith, P., Pathak, P.A., Ritter, J.R., 2005. Short interest, institutional
ownership, and stock returns. J. Financ. Econ. 78, 243276.
Battalio, R., Hatch, B., Jennings, R., 2004. Toward a national market system
for U.S. exchange-listed equity options. J. Finance 59, 933962.
Bhattacharya, M., 1987. Price changes of related securities: the case of call
options and stocks. J. Financ. Quant. Anal. 22, 115.
Biais, B., Hillion, P., 1994. Insider and liquidity trading in stock and options
markets. Rev. Financ. Stud. 7, 743780.
Black, F., 1975. Fact and fantasy in the use of options. Financ. Anal. J. 31,
3672.
Black, F., Scholes, M., 1973. The pricing of options and corporate liabilities.
J. Polit. Econ. 81, 637654.
Cao, C., Chen, Z., Griffin, J.M., 2005. Information content of option volume
prior to takeovers. J. Bus. 78, 10731109.
Carhart, M.M., 1997. On persistence in mutual fund performance.
J. Finance 52, 5782.
Chakravarty, S., Gulen, H., Mayhew, S., 2004. Informed trading in stock
and option markets. J. Finance 59, 12351258.
Chan, K., Chung, Y.P., Fong, W.M., 2002. The informational role of stock
and option volume. Rev. Financ. Stud. 14, 10491075.
Chan, K., Chung, Y.P., Johnson, H., 1993. Why option prices lag stock
prices: a trading-based explanation. J. Finance 48, 19571967.
Chordia, T., Roll, R., Subrahmanyam, A., 2002. Order imbalance, liquidity,
and market returns. J. Financ. Econ. 65, 111130.
Chordia, T., Subrahmanyam, A., 2004. Order imbalance and individual
stock returns: theory and evidence. J. Financ. Econ. 72, 485518.
D0 Avolio, G., 2002. The market for borrowing stock. J. Financ. Econ. 66,
271306.
Dong, W., Sinha, N., 2011. Where do informed traders trade? Trading
around news on Dow 30 options. Unpublished working paper.
University of Illinois, Chicago, IL.
Easley, D., O0 Hara, M., 1987. Price, trade size, and information in securities
markets. J. Financ. Econ. 19, 6990.
Easley, D., O0 Hara, M., 1992. Time and the process of security price
adjustment. J. Finance 47, 577605.
Easley, D., O0 Hara, M., Srinivas, P.S., 1998. Option volume and stock prices:
evidence on where informed traders trade. J. Finance 53, 431465.
Fama, E.F., MacBeth, J.D., 1973. Risk, return, and equilibrium: empirical
tests. J. Polit. Econ. 81, 607636.
Figlewski, S., Webb, G., 1993. Options, short sales, and market completeness. J. Finance 48, 761778.
Glosten, L., Harris, L., 1988. Estimating the components of the bidask
spread. J. Financ. Econ. 21, 123142.
Glosten, L.R., Milgrom, P.R., 1985. Bid, ask, and transaction prices in a
specialist market with heterogenously informed traders. J. Financ.
Econ. 14, 71100.

645

Henderson, B., Pearson, N., 2010. The price impact of large hedging trades.
Unpublished working paper. University of Illinois, Urbana-Champaign, IL.
Henderson, B., Pearson, N., 2011. The dark side of financial innovation: a
case study of the pricing of a retail financial product. J. Financ. Econ.
100, 227247.
Ho, L.-C.J., 1993. Option trading and the relation between price and
earnings: a cross-sectional analysis. Account. Rev. 68, 368384.
Ho, T., Stoll, H., 1981. The dynamics of dealer markets under competition.
J. Finance 38, 10531074.
Holowczak, R., Hu, J., Wu, L., 2014. Aggregating information in option
transactions. Journal of Derivatives 21 (3).
Holowczak, R., Simaan, Y., Wu, L., 2006. Price discovery in the U.S. stock
and stock options markets: a portfolio approach. Rev. Deriv. Res. 9,
3765.
Johnson, T.L., So, E.C., 2012. The option to stock volume ratio and future
returns. J. Financ. Econ. 106, 262286.
Kyle, A.S., 1985. Continuous auctions and insider trading. Econometrica
53, 13151336.
Lakonishok, J., Lee, I., Pearson, N.D., Poteshman, A.M., 2007. Option
market activity. Rev. Financ. Stud. 20, 813857.
Lee, C.M.C., Ready, M.J., 1991. Inferring trade direction from intraday data.
J. Finance 46, 733746.
Lin, J.-C., Sanger, G.C., Booth, G.G., 1995. Trade size and components of the
bidask spread. Rev. Financ. Stud. 8, 11531183.
Manaster, S., Rendleman, R.J., 1982. Option prices as predictors of
equilibrium stock prices. J. Finance 37, 10431058.
Merton, R.C., 1973. An intertemporal asset pricing model. Econometrica
41, 867887.
Muravyev, D., Pearson, N.D., Broussard, J.P., 2013. Is there price discovery
in equity options? J. Financ. Econ. 107, 259283.
Newey, W.K., West, K.D., 1987. A simple, positive semi-definite, heteroskedasticity, and autocorrelation consistent covariance matrix. Econometrica 55, 703708.
Ni, S.X., Pan, J., Poteshman, A.M., 2008. Volatility information trading in
option market. J. Finance 63, 10591091.
Ni, S.X., Pearson, N.D., Poteshman, A.M., 2005. Stock price clustering on
option expiration dates. J. Financ. Econ. 78, 4987.
Pan, J., Poteshman, A.M., 2006. The information in option volume for
future stock prices. Rev. Financ. Stud. 19, 871908.
Poteshman, A.M., 2006. Unusual option market activity and the terrorist
attacks of September 11, 2001. J. Bus. 79, 17031726.
Roll, R., Schwartz, E., Subrahmanyam, A., 2010. O/s: the relative trading
activity in options and stock. J. Financ. Econ. 96, 117.
Skinner, D., 1990. Options markets and the information content of
accounting earnings releases. J. Account. Econ. 13, 191211.
Stephan, J.A., Whaley, R.E., 1990. Intraday price change and trading
volume relations in the stock and stock option markets. J. Finance 45,
191220.