Market Illiquidity and Conditional Equity Premium

• DOI: http://doi.org/10.1111/fima.12162
• Author: Robert Savickas,Hui Guo,Sandra Mortal,Robert Wood
• Published date: 01 September 2017
• Date: 01 September 2017

Market Illiquidity and Conditional Equity

Premium

Hui Guo, Sandra Mortal, Robert Savickas, and Robert Wood∗

We examine the time-series relation between aggregate bid-ask spreads and conditional equity

premium. We document that average marketwiderelative effective bid-ask spreads forecast aggre-

gate market returns onlywhen controlling for average idiosyncratic variance. This controlallows

us to document the otherwise elusive relation between illiquidity and returns. The reason is that

idiosyncratic variance correlates positively with spreadsbut has a negative effect on conditional

equity premium, causing an omitted variable bias. Our results are robust to standard return pre-

dictors, alternative illiquidity measures, and out-of-sample tests. These findings are important

because they provide strongsupport for the literature’sconjecture that marketwide liquidity is an

important asset pricing risk factor.

Chordia, Roll, and Subrahmanyam (2000), Hasbrouck and Seppi (2001), Huberman and Halka

(2001), Lo and Wang (2000), and others document strong commonality in stock-level liquidity.

Pastor and Stambaugh (2003) and Acharya and Pedersen (2005) conjecture that liquidity is a

systematic risk factor because they find that covariances with marketwide liquidity help explain

the cross-section of stock returns.1For this inference to be validated, according to Campbell’s

(1993) intertemporal capital asset pricing model (ICAPM), we need a time-series relation—that

decreased market liquidity predicts higher future market returns.2The relation, however, is rather

weak over the post-World War II sample. In this article, we explore the possibility that aggregate

idiosyncratic risk confounds the time-series relation between aggregate liquidity and conditional

equity premium.

We thank the anonymous referee, Yakov Amihud, Turan Bali, Shmuel Baruch, Eric Chang, Rene Garcia, Brian Hatch,

Pankaj Jain, Abraham Lioui, Marc Lipson, Buhui Qiu, Michael Schill, Steve Slezak, Masa Watanabe, Ivo Welch, and

the seminar participants at Hong Kong University, EDHEC Business School, the Central University of Finance and

Economics, the Chicago Quantitative Alliance 2009 Spring Meetings in Las Vegas, and the FMA 2010 meetings in New

YorkCity for comments. Weare grateful to Nagpurnanand Prabhala,Buhui Qiu, Yufeng Han, and Amit Goyal forproviding

data. The paper formerly circulated under the title “Uncovering the Relation between Aggregate Stock Illiquidity and

Expected Excess Market Returns.”

∗Hui Guo is a Briggs Swift Cunningham Professorof Finance at the Carl H. Lindner College of Business at the University

of Cincinnati in Cincinnati, OH. Sandra Mortal is an Associate Professorof Finance at the FogelmanCollege of Business

& Economics at the University of Memphis in Memphis, TN. Robert Savickas is an Associate Professorof Finance at the

School of Business at George Washington University in Washington,DC. Robert Wood is a ProfessorEmeritus of Finance

at the FogelmanCollege of Business & Economics at the University of Memphis in Memphis, TN.

1Næs, Skjeltorp, and Ødegaard (2011) and Jensen and Moorman (2010) show that aggregate illiquidity and illiquidity

premium change countercyclicallyacross time. Amihud and Mendelson (1986) investigate whether expected stock returns

are related to the level of illiquidity as opposed to illiquidity covariance risk. Bali et al. (2014) find that expected stock

returns are related to illiquidity shocks in addition to the level of illiquidity.

2Jones (2002) proposes two specific channels through which aggregate illiquidity correlates with conditional equity

premium. First, illiquidity is a measure of information asymmetry, which correlates positively with the expected stock

returns (e.g., Glosten and Milgrom, 1985). Second, illiquidity moves closely with market makers’ financial constraints,

which tend to change countercyclically across time. Moreover, Baker and Stein (2004) argue that variation in illiquidity

reflects waves of investors’excessive optimism and pessimism.

Financial Management •Fall 2017 •pages 743 – 766

744 Financial Management rFall 2017

Using direct transaction cost measures for a long but cross-sectionally restricted sample of

Dow Jones firms, Jones (2002) uncovers a positive relation between aggregate illiquidity and

future market returns in a 1900 to 2000 sample but not in a post-1950 sample. Using indirect

measures of illiquidity for a large set of firms, Amihud (2002) and Baker and Stein (2004)

document a positive illiquidity-return relation as well; Fujimoto (2003), however, shows that

these illiquidity measures have negligible predictive power over 1966 to 2002. We document a

positive and significant relation between aggregate effective bid-ask spreads and future excess

stock market returns only when including aggregate idiosyncratic variance as a control. This

suggests the lack of predictive power in earlier studies is partly due to an omitted-variable bias.

There are two necessary conditions for the omitted-variable bias. First, aggregate illiquidity

and aggregate idiosyncratic variance correlate closely with each other. Indeed, their correlation

coefficient is positive and over 40%. Second, aggregate idiosyncratic variance correlates with

conditional equity premium in a way that is opposite to that of aggregate illiquidity; that is, the

relation between average idiosyncratic variance and future market returns is negative. Guo and

Savickas (2008) document such a negative relation in G7 countries.

The underlying relation between liquidity and returns as well as the confounding relation

between idiosyncratic variance and both liquidity and returns are both suggested by existing

economic theories and empirical findings. Constantinides (1986) argues that transaction costs

per se have a negligibleeffect on equity premium because investors choose optimally to rebalance

their portfolios infrequently to avoid high trading costs. Jang et al. (2007) and Lynch and Tan

(2011), however, point out that the effect of liquidity risk on equity premium can be economically

significant when there is a strong demand for hedging against changes in investment opportunities.

That is, both liquidity risk and hedging risk are important determinants of conditional equity

premium.

Guo and Savickas (2008) document a negative relation between value-weighted aggregate

idiosyncratic variance and conditional equity premium, using quarterly data. They argue that

this is because, by construction, idiosyncratic variance correlates closely with the variance of

an omitted hedging risk factor.3Specifically, Guo and Savickas (2008) find that (1) aggregate

idiosyncratic variance correlates closely with value premium variance—the most commonlyused

proxy for the hedging risk factor in empirical asset pricing research—and (2) the two variances

have similar forecasting power for excess market returns.4Similarly, we document that the

relation between aggregate bid-ask spreads and future excess market returns becomes positive

and significant when we replace idiosyncratic variance with value premium variance. In other

words, both idiosyncratic varianceand value premium variance have similar effects on the relation

between aggregate bid-ask spreads and conditional equity premium. Our results suggest that the

effect of idiosyncratic variance on the illiquidity/conditional equity premium relation is due to its

relation to the hedging risk factor.

Aggregate idiosyncratic variance correlates positively with aggregate bid-ask spreads because

an increase in uncertainty about investment opportunities, for example, the value premium, likely

leads to more information asymmetry and higher inventory costs and hence larger bid-ask spreads.

For firm-level idiosyncratic variance, a relation with spreads is consistent with the information

3Recent studies (e.g., Chen and Petkova2012; Duarte et al., 2014; Herskovic et al., 2016), document strong commonality

in stock-level idiosyncratic variance and show that innovations in aggregate idiosyncratic variance are priced in the

cross-section of stock returns.

4Fama and French (1996) interpret the value premium as a hedging risk factor.Campbell and Vuolteenaho (2004) show

that the value premium is a proxy for changes in discount rates in Campbell’s (1993) ICAPM. Guo et al. (2009) show

that the negative relation betweenvalue premium variance and conditional equity premium is consistent with Campbell’s

(1993) ICAPM.