Order Flow and Expected Option Returns

• Author: DMITRIY MURAVYEV
• Date: 01 April 2016
• DOI: http://doi.org/10.1111/jofi.12380
• Published date: 01 April 2016

THE JOURNAL OF FINANCE •VOL. LXXI, NO. 2 •APRIL 2016

Order Flow and Expected Option Returns

DMITRIY MURAVYEV∗

ABSTRACT

I show that the inventory risk faced by market-makers has a first-order effect on option

prices. I introduce a simple approach that decomposes the price impact of trades into

inventory risk and asymmetric information components. While both components are

large for option trades, the inventory risk component is larger. Using the full panel of

daily option returns, I find that option order imbalances attributable to inventory risk

have five times larger impact on option prices than previously thought. Finally,I find

that past order imbalances have greater predictive power than any other commonly

used predictor of option returns.

IN THIS PAPER,ISTUDY how the inventory risk that market-makers face affects

prices in the equity options market. These variables have a well-established

theoretical relation. Market-makers are required to provide liquidity and ac-

commodate customer order imbalances and thus their position often deviates

substantially from the desired level. In turn, a risk-averse investor requires a

higher expected return on his inventory as compensation for holding a portfolio

with suboptimal weights. The extent of inventory risk is proportional to the po-

sition’s size and volatility, the market-maker’s risk aversion, and the expected

holding period. Option market-makers face inventory risk because, in practice,

options cannot be fully replicated by trading in the underlying.

I show that option market-maker inventory risk has a first-order effect on

option prices. This conclusion is important for at least two reasons. First, the

equity options market is economically large: options on about 1.5 billion shares

are traded daily in the United States. Second, option prices are extensively

used to measure variables ranging from equity return volatility to the stochas-

tic discount factor, and these measures may be biased if the inventory risk

∗Muravyev is at the Carroll School of Management, Boston College. I appreciate helpful com-

ments from Kenneth Singleton (Editor), two anonymous referees, Heitor Almeida, Hui Chen,

Prachi Deuskar, Slava Fos, TimJohnson, Alan Marcus, George Pennacchi, Allen Poteshman, Ron-

nie Sadka, and Mao Ye, as well as seminar participants at University of Illinois, Georgia Institute

of Technology, Fordham University, Boston College, University of Toronto, University of Amster-

dam, Stockholm School of Economics, Singapore Management University,City University of Hong

Kong, and Wharton School of Business and participants at the 2013 American Finance Association

meeting. I am especially grateful to Neil Pearson for his support, wisdom, and extensive discus-

sions. I thank Nanex and Eric Hunsader for providing the option trade and quote data and the

International Securities Exchange and Jeff Soule for providing the open/close data. Financial sup-

port from the Irwin Fellowship is gratefully acknowledged. The author does not have any conflicts

of interest, as identified in the Disclosure Policy.

DOI: 10.1111/jofi.12380

673

674 The Journal of Finance R

component is not removed from option prices. Thus, a new generation of option

pricing models that accounts for inventory risk is needed.

The importance of inventory risk for the options market suggests that its role

in other markets may be more important than previously thought.1Prior liter-

ature shows that buying pressure is associated with higher option prices, but

its economic magnitude is small compared to other factors, making inventory

risk appear to be of secondary importance. The difference in the results is due

to two methodological issues that this paper solves. First, order flow and prices

are endogenous, so concurrent factors such as news affect both. The popular

approach of regressing option returns on same-day order imbalances can thus

produce biased coefficients. Second, besides having an inventory risk impact,

order imbalance also reflects informed trading (Shleifer (1986)) and correlates

with changes in economic fundamentals. This problem is commonly ignored by

attributing the entire order imbalance to only one of these three factors.2I find

that, after addressing these problems, the market-wide order imbalance has a

particularly large effect on option prices, as market-makers manage inventory

risk on a portfolio basis.

The paper’s main conclusion is supported by two sets of results. First, I es-

tablish the importance of inventory risk at the intraday level. Using a novel

methodology, I find that the inventory risk component of the price impact of

option trades is larger than the asymmetric information component for each

stock in my sample. Second, using a full panel of daily option returns, I estab-

lish that inventory risk is a dominant factor at the daily level. The instrumental

variables (IV) approach identifies the inventory-risk component of price pres-

sure by studying how past order option imbalances predict future imbalances.

According to the IV approach, a typical inventory shock moves option prices

by 3.7% the same day, which is five times larger than the change implied by

conventional OLS. Finally, I conduct a direct horse race between more than

50 commonly used predictors of future option returns including the inventory-

related order imbalance, I find that order imbalance has the highest predictive

power by a large margin. Economically, if it increases by one standard devia-

tion, the next-day return is 1% higher.

With respect to the intraday analysis, the interaction between trades and

quotes is key to understanding how and why prices change. The literature

identifies two reasons why quoted prices increase after a buyer-initiated trade.

First, market-makers adjust upward their beliefs about fair value as the trade

may contain private information (e.g., Glosten and Milgrom (1985)). Second,

market-makers require compensation for allowing their inventory position to

deviate from the desired level, and thus a risk-averse market-maker will ac-

commodate a subsequent buy order only at a higher price (e.g., Stoll (1978)).

1See Bollen and Whaley (2004) and Garleanu, Pedersen, and Poteshman (2009).

2Pan and Poteshman (2006) attribute all order imbalance to informed trading, Bollen and

Whaley (2004) and Garleanu, Pedersen, and Poteshman (2009) to inventory risk, and Chen, Joslin,

and Ni (2014) to fundamentals.

Order Flow and Expected Option Returns 675

Both arguments imply that quotes change in the direction of trade, but for

different reasons.

Building on Huang and Stoll (1997), I introduce a novel microstructure

method for evaluating the size and relative importance of information asymme-

try and dynamic inventory control for price dynamics. The method decomposes

the price impact of trades into an inventory risk component and an asymmetric

information component. The main idea is that investors instantly receive iden-

tical information about a trade, while only the market-maker at the exchange

where the trade is executed experiences a change in inventory. Thus, the price

response at the trading exchange includes both the asymmetric information

and the inventory risk components, while the price impacts at the nontrading

exchanges contain only the asymmetric information component. The difference

between the price responses of the trading and nontrading exchanges identifies

the inventory risk component. Both price impact components can be easily es-

timated empirically from observed price responses. I formally implement this

idea by extending the framework of Madhavan and Smidt (1991)tomultiple

competitive market-makers.

This new method is needed because, as Huang and Stoll (1997) discuss, exist-

ing methods struggle to separate the information and inventory components.3

These methods commonly assume that permanent price changes are attributed

to asymmetric information, while the impact of inventory risk is temporary and

is reversed in a matter of minutes. However, empirical evidence suggests that,

although the effect of inventory risk on prices is by definition temporary, some-

times it takes weeks to unwind. Thus, at the intraday level, the effects of

inventory risk and asymmetric information are largely permanent, making it

hard for conventional methods to separate them. The method that I introduce

avoids this criticism as it makes no assumption about how long it takes for the

inventory effect to disappear. Overall, this method contributes to the literature

on the role and measurement of market-maker inventory risk and to the litera-

ture on distinguishing the inventory and asymmetric information components

of bid-ask spreads.

I apply this new method to the equity options market using tick-level data

for options on 39 stocks over the period April 2003 to October 2006, when the

options market had already converted to its modern form with predominantly

electronic trading. The method yields several results. First, contrary to Vijh

(1990), I find that option trades have a significant price impact. Thus, option

trades contain a lot of new information that drives changes in option prices.

Second, although both price impacts have a significant effect, inventory risk

has larger price impact (0.4%) than asymmetric information (0.2%). Moreover,

the inventory risk component is larger for each stock in my sample. Thus, at

3Popular methods to measure asymmetric information include Glosten and Harris (1988), Has-

brouck (1991), George, Kaul, and Nimalendran (1991), Lin, Sanger, and Booth (1995), Madhavan,

Richardson, and Roomans (1997), and Huang and Stoll (1997). Madhavant and Smidt (1993),

Hasbrouck and Sofianos (1993), and Naik and Yadav (2003) use actual inventory data to study

market-maker behavior in the equity market.