Research in International Business and Finance 52 (2020) 101165

Contents lists available at ScienceDirect

Research in International Business and Finance

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

Liquidity risk and stock performance during the financial crisis

T
Tung Lam Danga,*,1, Thi Minh Hue Nguyenb,1
a
University of Economics, The University of Danang, Viet Nam
b
School of Banking and Finance, National Economics University, Viet Nam

ARTICLE INFO ABSTRACT

JEL classification: We investigate whether and how ex-ante liquidity risk affects realized stock returns during the
G01 global financial crisis of 2008–2009 in international equity markets. We find that stocks with
G12 higher pre-crisis return exposure to global market liquidity shocks experience larger price re-
G15 ductions during the crisis period. Our findings provide further insight into the comprehensive
picture of the effect of liquidity risk on asset prices, especially in an international context and
Keywords:
under different market conditions.
Liquidity risk
Cumulative stock returns
Global financial crisis

1. Introduction

Liquidity and commonality in liquidity across stocks and markets are of major concern to investors. Prior research documents that
stock liquidity levels affect the required returns of stocks (e.g., Amihud and Mendeson, 1986; Brennan and Subrahmanyam, 1996;
Amihud, 2002). Recent studies show that liquidity risk is a priced factor (e.g., Pástor and Stambaugh, 2003; Acharya and Pedersen,
2005; Sadka, 2006; Bekaert et al., 2007; Lee, 2011). However, an important question that has more practical implications but remains
unexplored is how a stock’s liquidity risk captures its performance during periods of market turmoil.
In this paper, we investigate whether and how ex-ante liquidity risk affect realized stock returns during the financial crisis of
2008–2009 in a global context. We construct an extensive intraday trading data set from Thomson Reuters Tick History (TRTH) for
17,493 firms across 41 countries to measure stock liquidity. The recent availability of intraday transaction data from the TRTH for
most countries enables us to construct global stock liquidity measures that are more precise than those used in empirical studies
outside the U.S. Following Acharya and Pedersen (2005), we use individual stocks' return exposure to global (local) market liquidity
shocks, estimated over the period before the crisis, as a proxy for stocks’ liquidity risk.2
We find that stocks with high pre-crisis global liquidity risk experience more price declines during the global financial crisis of
2008–2009. In particular, stocks with higher pre-crisis return sensitivity to the global market liquidity shocks exhibit greater declines
in price during the crisis. The results are consistent across the emerging, developed, and global samples and are robust to various
spread-based measures that are used as proxies for stock liquidity. However, we find weaker evidence on the predictability of stocks’
local liquidity risk, which may be due to the global nature of the financial crisis of 2008–2009, in which the exposure of each country
or individual stock to the global factors seems to be more important than the local factors.

Corresponding author.
⁎

E-mail addresses: [email protected] (T.L. Dang), [email protected] (T.M.H. Nguyen).

1
Both authors contributed equally to this manuscript.
2
For the purpose of this paper, we refer to a stock's return exposure to global market liquidity shocks as the stock’s global liquidity risk, and a
stock's return exposure to local market liquidity shocks as the stock’s local liquidity risk.

https://doi.org/10.1016/j.ribaf.2019.101165
Received 12 March 2019; Received in revised form 20 November 2019; Accepted 18 December 2019
Available online 19 December 2019
0275-5319/ © 2019 Elsevier B.V. All rights reserved.
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

Our study advances the understanding of the relation between liquidity risk and stock performance during a crisis, especially with
a focus on an international setting. Most prior evidence on the pricing effect of stock liquidity risk focuses on a single country, mainly
in U.S. markets, and only a few papers investigate international firms. In addition, all prior studies focus on the pricing implications of
liquidity risk, that is, how liquidity risk relates to the expected returns of stocks. Whether and how stock liquidity risk captures stocks’
realized returns when the market experiences turbulence is still an empirical question.
Our paper is related to several recent studies that investigate the relation between liquidity risk and stock returns during the crisis,
including Lou and Sadka (2011) and Cao and Petrasek (2014). However, our study differentiates itself from those papers in several
distinct ways. Lou and Sadka (2011) examine whether the performance of U.S. stocks during the finance crisis of 2008–2009 can be
explained by their pre-crisis liquidity risk. In contrast, our focus is on international financial markets, which are arguably more
illiquid, and the effect of liquidity risk could be more severe. Our article differs from Cao and Petrasek’s (2014) study in both method
and focus. Cao and Petrasek (2014) employ the event study method to examine what factors affect the relative performance of stocks
during liquidity crises in U.S. markets, whereas we rely on the cross-sectional data of stocks in 41 countries to examine the relation
between ex-ante liquidity risk and stocks’ crisis performance.
To the best of our knowledge, we are the first to empirically investigate whether ex-ante liquidity risk has an effect on stock
returns during a market crash in the global context, using the financial crisis of 2008–2009 as a natural experiment and more refined
microstructure data to measure the liquidity risk of stocks. We show that pre-crisis liquidity risk can explain realized stock returns
during the crisis. Specifically, stocks whose returns are more exposed to market liquidity shocks exhibit poorer performance during
the crisis. Our results thus lend support to the predictability of liquidity risk for stocks’ crisis performance; they also provide further
insights into a comprehensive picture of the effect of liquidity risk on asset prices under different market conditions.
In Section 2, we briefly cover the related literature on liquidity risk and asset prices. Section 3 describes data sources, a variable
construction procedure, and summary statistics. Section 4 presents evidence on the effect of ex-ante liquidity risk on stocks’ crisis
performance, as well as robustness checks. We conclude the paper in Section 5.

2. Related literature

Since the seminal study of Amihud and Mendelson (1986), a large body of literature has explored the link between asset prices
and stock liquidity level (e.g., Brennan and Subrahmanyam, 1996; Datar et al., 1998; Eleswarapu, 1997; Amihud, 2002; Bali et al.,
2014; among others). More recent studies have emphasized the role of liquidity as a systematic risk factor. Karolyi et al. (2012) find
that liquidity risk, measured as commonality in liquidity, is greater during times of high market volatility, particularly when the
market experiences large declines.3 Pástor and Stambaugh (2003) investigate the relation between asset prices and the sensitivity of
stock returns to market-wide liquidity and show that the difference in expected returns between the most and least liquidity-sensitive
stock portfolios is economically significant. Sadka (2006) provides similar evidence that the liquidity factor is priced with a positive
risk premium using alternative measures of liquidity. Watanabe and Watanabe (2008) examine whether the effects of liquidity and
liquidity risk on stock returns vary over time. They find that the pricing of liquidity risk strengthens in the high liquidity beta state.
Acharya and Pedersen (2005) develop a broad pricing model that embeds the different aspects of liquidity risk by extending the
traditional capital asset pricing model (CAPM) to include transaction costs.4 They find that liquidity risk is priced for the U.S. market.
Specifically, the difference in annualized expected returns between the highest and lowest liquidity portfolio is attributable to both
expected illiquidity and the effect of liquidity risk. Chung and Chuwonganant (2018) investigate the role of liquidity provision in the
relation between market volatility and stock returns. They provide evidence that the effect of market volatility on stock returns
depends critically on how stock liquidity reacts to unexpected changes in market volatility. Chen et al. (2018) examine the predictive
content of aggregate liquidity for stock returns and real economic activity. They decompose liquidity proxies into a component
capturing aggregate volatility and a residual. These authors find strong evidence that the component of liquidity uncorrelated with
volatility forecasts stock market returns.
The pricing effect of liquidity risk is also studied in a single country other than the U.S. Foran et al. (2015) examine the pricing of
systematic liquidity risk using a sample of firms listing in the UK. They find systematic liquidity risk is positively priced in the cross-
section of stocks, specifically for the quoted spread liquidity measure. Ho and Chang (2015) investigate whether systematic liquidity
risk plays a role in explaining the cross-section of stock returns in China. They find that the expected stock returns are related to
stocks’ return sensitivities to fluctuations in aggregate market liquidity. Dinh (2017) investigates the link between stock returns, risk
and liquidity in the Norwegian market. The author shows that stock liquidity and idiosyncratic risk may jointly contribute toward
explaining stocks’ expected returns.
The pricing of liquidity risk in an international setting is investigated by Bekaert et al. (2007). Using data for 19 emerging
markets, they find that liquidity risk with respect to local markets is priced significantly, while the price of global liquidity risk is only

3
Recent empirical papers that study commonality in liquidity as a risk factor also include Kempf and Mayston (2008), Hameed et al. (2010),
Będowska-Sójka and Echaust (2019), among others.
4
Specifically, Acharya and Pedersen’s (2005) liquidity adjusted capital asset pricing model nests three components of liquidity risk, in addition to
the traditional market risk. β1 is similar to the traditional market beta of CAPM. β2 is the liquidity risk caused by the covariance between individual
stock liquidity with market liquidity (Chordia et al., 2000). β3 captures the liquidity risk that arises due to the covariance between an individual
stock’s returns with the market liquidity (Pástor and Stambaugh, 2003). β4 represents an individual stock’ liquidity sensitivity to the market returns
and is a new measure of liquidity risk.

2
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

marginally significant. Lee (2011) utilizes the liquidity adjusted capital asset pricing model of Acharya and Pedersen (2005) to
investigate the effect of liquidity both as a characteristic and a risk factor on asset prices for stocks from 50 countries. He provides
supporting evidence that liquidity risk is the priced factor in the global financial market. Moshirian et al. (2017) find that com-
monality in liquidity is priced in international stock markets and that the pricing effect is stronger in developed countries. Ma et al.
(2018) examine how the interaction between market volatility and liquidity shocks affects stock returns using a sample of stocks
across 41 countries. They find that stock liquidity has a greater effect on the relation between market volatility and stock returns
during crisis periods.
Although both U.S. and international evidence sheds some light on the importance of liquidity risk for asset prices, few re-
searchers investigate whether and how a stock’s liquidity risk predicts stock performance during a market crash, which may have
more practical implications. Considering the situation when market returns or market liquidity decline significantly, a wealth con-
strained investor would liquidate her position (s) to raise money for unexpected needs. If her asset liquidity or returns are highly
correlated with the market liquidity, distressed sales of assets are then more expensive. Unless equitably compensated, investors
would not be willing to hold these assets (Pástor and Stambaugh, 2003; Acharya and Pedersen, 2005). In addition, the flight-to-
quality theory suggests that investors tend to shift their portfolio towards less risky and more liquid assets in stressed market
scenarios (Beber et al., 2009). These effects imply that stocks with higher liquidity risk would experience a greater decline in price
during market downturns. In support of this, Lou and Sadka (2011) find that stocks with higher pre-crisis liquidity risk exhibit a
greater drop in price than those with the lower liquidity risk in the U.S market during the crisis of 2008–2009. Cao and Petrasek
(2014) employ the event study method and find that abnormal stock returns during liquidity crises are strongly negatively related to
liquidity risk. Given the global nature of the recent financial crisis and the fact that market liquidity differs significantly across
countries during the crisis, extending the study to an international setting could provide further understanding on the importance of
liquidity risk on asset prices worldwide.

3. Data and sample description

We collect data from several sources to construct variables in this study. Specifically, intraday transaction data to estimate
liquidity measures are from Thomson Reuters Tick History (TRTH), stock returns (in US dollars) from Datastream, and accounting
data for calculating control variables from Worldscope via Datastream.

3.1. Estimation of a stock's liquidity risk

3.1.1. Proxies for stock liquidity

Stock liquidity has multiple dimensions and is difficult to measure by nature (Lesmond, 2005). Previous literature has adopted a
broad range of measures to proxy for stock liquidity, suggesting that there is no consensus about the most appropriate measure.
Goyenko et al. (2009) analyze a wide range of both traditional and microstructure-based liquidity measures and find that the spread-
based liquidity measures win the majority of the horseraces. Fong et al. (2017) analyze several liquidity proxies based on low-
frequency data to figure out which of them are the best liquidity proxies for global research. Ahn et al. (2018) find that most of the
spread proxies performs relatively well in emerging markets. Abdi and Ranaldo (2017) develop a new method to estimate bid-ask
spreads using daily close, high and low prices, which uses wider prices and is independent of the direction of the trade.5 In this study,
we employ the percentage effective spread and the percentage quoted spread as proxies for stock liquidity, which are computed based
on real-time transaction data. We adopt these liquidity measures because spread-based measures more effectively capture stock
liquidity and are arguably considered as the most refined among liquidity measures (Goyenko et al., 2009). Specifically, for each
trade, we compute the percentage effective spread as twice the absolute value of the difference between trading price and the
midpoint of the bid and ask price, scaled by the midpoint of the bid and ask price. The percentage quoted spread is defined as the
absolute value of the difference between ask and bid price, scaled by the midpoint of the bid and ask price. Then, the daily spread is
calculated as the dollar-volume weighted average of intraday spreads. The monthly spread is equal to the average of daily spreads
over the given month. The higher value in these spread measures of a given stock indicates that the stock is more illiquid.
To estimate these liquidity measures, we collect intraday transaction data from TRTH, with an initial sample covering 26,322
stocks from 53 countries. We impose several filters on each stock to build the reliable sample for regression analyses. Specifically, for
a stock to be included in the sample, its trades and quotes must be submitted during regular trading hours. Irregular trades and trades
with negative trading prices are excluded. Quotes with bid-ask spreads that are larger than half their midpoint quote prices are also
excluded. Following Chordia et al. (2000), we eliminate the spread values that are greater than 0.40.

3.1.2. Estimation of liquidity risk

We use individual stocks' return exposure to market liquidity shocks estimated over the period before the crisis as a proxy for the
stocks’ liquidity risk, following Acharya and Pedersen (2005).6 Specifically, a stock’s liquidity risk is estimated as follows:

5
See also Díaz and Escribano (2020) for a comprehensive review on measuring liquidity in financial markets.
6
In this paper, we do not intend to test Acharya and Pedersen’s (2005) liquidity adjusted capital asset pricing model (LCAPM). Instead, we
examine the ability of a liquidity risk measure, which is measured by individual stocks' return exposure to market liquidity shocks and estimated
over the period before the global financial crisis of 2008-2009, to explain the crisis performance of stocks. Our measure of liquidity risk is analogous

3
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

LIQ Cov (rti, ctM )

=
(1)
i
var (rtM ctM )

where rit, rM
are the returns of stock i and the market return during period t, respectively;
t cit, cM
are the liquidity of stock i and the
t
market liquidity during period t, respectively.
This βLIQ measure captures the liquidity risk that arises due to the covariance between an individual stock’s returns with the
market liquidity (Pástor and Stambaugh, 2003; Acharya and Pedersen, 2005).7 βLIQ is expected to be negative, with the higher
absolute value indicating greater liquidity risk; thus, the stock being predicted to decline more in prices when the market becomes
illiquid.
Because the financial crisis of 2008–2009 is a global event, the exposure of each country or individual stock to the global market
factor may be more relevant than its exposure to the local market factor.8 Therefore, we use stocks’ return exposure to global market
liquidity shocks (global liquidity risk) in the primary analyses. However, we also report the results using stocks’ return exposure to
local market liquidity shocks (local liquidity risk) as an additional check.
As market liquidity is persistent (Pástor and Stambaugh, 2003; Acharya and Pedersen, 2005; Sadka, 2006; Lee, 2011), we use
innovations in liquidity to estimate liquidity risk. Following Liu’s (2006) and Lee’s (2011) approach, we obtain the innovations in
liquidity for the market portfolio and individual stocks through applying the AR(1) process to the first-difference of liquidity.

ΔcM M M
t = ρ Δct-1 + uM,t (2)
M
where c denotes the liquidity, and uM is the innovation in liquidity of the global (local) market (or stock i, with M replaced by i), and
Δ is the first-difference operator.
To estimate pre-crisis liquidity risk for each stock, we use monthly stock liquidity and return data before the crisis from January
2003 to December 2007. Several requirements are imposed for estimating stocks’ liquidity risk. First, to mitigate the effects of
potential data errors or other reasons from Datastream data, we set monthly stock returns that exceed 3.0 % to missing (Lee, 2011).
For a stock to be included in a monthly market portfolio, the stock must have at least 10 daily observations in that month. The
monthly returns of the 0.1 % extremes at the top and bottom of the return distribution in each country in a given month are discarded
(Amihud, 2002; Lee, 2011). Global (local) market return and liquidity in each month are computed as the equally weighted average
of individual stock returns and liquidity, respectively, across countries (in a country) in that month. Each country must have at least
10 stocks in a given month to calculate the market return and liquidity. Finally, following Lee (2011), we require that stocks have at
least 36 monthly observations over the five-year period to estimate liquidity risk.

3.2. Stock returns during the crisis period

There is no consensus on the exact span of the global financial crisis of 2008–2009. Previous studies of the crisis use various
definitions of the crisis period. For example, Bekaert et al. (2014) use the total equity market returns over the period from August
2007 to March 2009 as the returns of the crisis period. Tong and Wei (2011) examine whether stocks' pre-crisis characteristics predict
stock price changes over the crisis period, which is defined as from July 2007 to December 2008. In contrast, Raddatz and Schmukler
(2012) define the period of the global turmoil as being from September 2008 to June 2009. However, the crisis may have affected
global financial markets beginning only in early 2008 following events in the U.S. (the quant-event and the sale of Bear Sterns).
Therefore, as the crisis period in our primary analysis we define the period from January 2008 through March 2009, which is
considered the trough of the global equity market during the crisis (Bekaert et al., 2014). A stock’s crisis return is the cumulative
return of the stock from January 2008 to March 2009.

3.3. Control variables

Following prior studies (Lang and Maffett, 2011; Aragon and Strahan, 2012; Tong and Wei, 2011), we control in regression
analyses a battery of pre-crisis firm-specific characteristics that are likely to be correlated with liquidity and stocks’ crisis perfor-
mance. Control variables include the stock's return exposure to market return shocks (βM), the MSCI index dummy (MSCI), returns on
asset (ROA), the log of book-to-market ratio (BM), the log of market capitalization (MCap), the proportion of a firm’s shares that are
closely held (CHeld), a dummy for whether a stock lists in the U.S. (ADR), the number of stock analysts following the firm (Analyst),
annual stock returns (Return), stock return volatility (STD), the log of stock price at the end of 2007 (Price), and the log of the pre-
crisis liquidity level of individual stocks (Liq). βM is estimated using monthly data from January 2003 to December 2007, and the

(footnote continued)
to that used in Lou and Sadka (2011).
7
As shown in Acharya and Pedersen (2005), an unexpected decline in the market liquidity may have a wealth effect on stocks that are sensitive to
the market liquidity. βLIQ is negatively related to the expected return because investors are willing to accept a lower return on an asset with high
return in time of market illiquidity.
8
Tong and Wei (2011) show that the global financial crisis of 2008-2009 began in the U.S. in late 2007 and quickly spread to other countries
through several channels, including a reversal of global capital flows. They find that the severe effects of the crisis during the 2008–2009 global
financial crisis is systematically linked to the pre-crisis international capital inflows, suggesting that each country or individual firm’s exposure to
international financing matters.

4
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

other control variables are measured over or at the end of 2007. The definition of firm-specific characteristics variables is described in
Appendix.
We include only common stocks in the sample, and exclude stocks with special features, such as ADRs, GDRs, warrants, trusts,
funds, and nonequity securities. In addition, we use stocks from the single major exchange of each country, except for China
(Shanghai and Shenzhen), Japan (Tokyo and Osaka) and the U.S. (American Stock Exchange and New York Stock Exchange), where
we use the two exchanges given their equal importance in these countries.
In our analysis, we consider only countries with at least 10 firms that have the measures of liquidity risk. The final sample includes
17,493 stocks from 41 countries.9 Countries are grouped into 21 developed markets (Australia, Austria, Belgium, Canada, Denmark,
Ireland, Finland, France, Germany, Hong Kong, Italy, Japan, Netherlands, Norway, New Zealand, Singapore, Spain, Sweden, Swit-
zerland, United Kingdom, United States) and 20 emerging countries (Argentina, Brazil, China, Chile, Egypt, Greece, Indonesia, India,
Israel, South Korea, Mexico, Malaysia, Peru, Poland, Philippines, Russia, South Africa, Thailand, Turkey, Taiwan) based on the
International Finance Corporation classification.

3.4. Summary statistics

Table 1A and 1B presents the average of the pre-crisis global (local) liquidity risk measures and cumulative stock returns during
the global crisis period from January 2008 to March 2009 for each of the 41 sample countries. As shown, liquidity risk as measured by
βLIQ has the expected negative sign. China, Japan, and the U.S. are among countries with the lowest ex-ante return exposure to global
liquidity shocks, on average. Specifically, the average global liquidity risk is −0.004 in China, −0.006 in Japan, and −0.007 in the
U.S., well above the sample average of −0.014. In contrast, Indonesia, India, and Turkey are among countries with the highest
average global liquidity risk, with the βLIQ value of −0.028, −0.029, and −0.035, respectively. Local liquidity risk is generally larger
in such countries as Malaysia (βLIQ=−0.034), Singapore (βLIQ=−0.021), and Australia (βLIQ=−0.024).
We observe that global stock markets dropped significantly by March 2009, with average cumulative stock returns (Ret) of −47.4
%. Remarkably, although the U.S. is the crisis-origin country, the financial crisis seems to affect the equity markets of other countries
more severely, with 26 out of 41 sample countries exhibiting the average cumulative stock returns of less than −50 %. In addition,
stock returns indicate considerable variation across countries, with India and Russia experiencing the largest price declines (−73.8 %
and −71.6 %), while Japan and Chile experiencing the smallest (−27.1 % and −31.5 %, respectively).
Table 2 shows the Pearson correlation coefficients among variables used in our analyses.10 The correlation coefficient between
cumulative stock returns over the crisis period (Ret) and stock liquidity risk (βLIQ) is positive for both the global liquidity risk and
local liquidity risk. This sheds some light on the expected relation between pre-crisis liquidity risk and stocks’ crisis performance. In
general, the moderate correlation between variables eliminates concerns about multicollinearity issues in the regression analyses.

4. Ex-ante liquidity risk and stocks’ crisis performance

We start this section by performing a univariate analysis to investigate how the cumulative stock returns of portfolios formed on
the basis of pre-crisis liquidity risk vary over the months of the crisis period. We next conduct regression analyses to rigorously
examine the relation between the ex-ante liquidity risk and realized stock returns, which allows for firm characteristics and other
factors that are potentially related to liquidity risk and stock returns to be taken into account.

4.1. Univariate analysis

Table 3 reports the cumulative stock returns of liquidity risk quintile portfolios through each month of the crisis period for the
global sample. Quintile portfolios are formed on the basis of pre-crisis liquidity risk, and cumulative stock returns are measured from
January 2008 through each month of the crisis period. Panel A presents cumulative stock returns for the portfolios of stocks sorted on
the basis of global liquidity risk, while Panel B includes the portfolios formed on the basis of local liquidity risk.
As shown in Panel A of Table 3, all global liquidity risk portfolios underperform significantly during the crisis. However, the price
declines of the high liquidity risk portfolios are more severe. The declines of returns for the highest liquidity risk portfolio are almost
twice as large as those observed for the portfolio with the lowest liquidity risk. The difference in the mean of cumulative stock returns
between the two portfolios is significantly high with the magnitude being even up to −26.5 % in January 2009.
A similar pattern is shown for the cumulative stock returns of portfolios when formed on the basis of the pre-crisis local liquidity
risk (Panel B of Table 3), although the difference in the returns of the highest liquidity risk portfolio and the lowest liquidity risk
portfolio appears to be lower than that of the global liquidity risk portfolios.
Overall, the results in this section provide preliminary insights on the relation between stocks’ ex-ante liquidity risk and their
crisis performance. Stocks with high pre-crisis liquidity risk tend to experience greater declines in prices than low pre-crisis liquidity
risk stocks during the crisis months. Specifically, stocks whose returns are more sensitive to market liquidity shocks exhibit con-
sistently poorer performance during the crisis, which suggests better predictability of this liquidity risk measure for ex-post stock
returns.

9
The number of observations is lower in regression analyses due to missing data when variables are combined.
10
For ease of presentation, the global liquidity risk is denoted as GβLIQ, while the local liquidity risk is denoted as LβLIQ in this table.

5
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

Table 1A
Summary statistics. This table presents the average of pre-crisis liquidity risk and cumulative stock returns during the crisis period for each of 41
sample countries. N denotes the number of eligible stocks in 2007 (pre-crisis). Pre-crisis liquidity risk is estimated at the firm level using the monthly
returns and innovations in liquidity from January 2003 to December 2007 with respect to the innovation in global (local) market liquidity. The
cumulative returns are stock returns during the crisis period that is defined from January 2008 to March 2009.
Panel A: Emerging markets

Country N Global liquidity risk Local liquidity risk Cumulative returns

(1) (2) (3)

Argentina 64 −0.013 −0.015 −52.7%

Brazil 24 −0.021 −0.006 −59.6%
China 1329 −0.004 −0.003 −32.1%
Chile 111 −0.018 −0.003 −31.5%
Egypt 81 −0.016 −0.006 −66.4%
Greece 257 −0.022 −0.011 −63.2%
Indonesia 260 −0.028 −0.011 −44.6%
India 826 −0.029 −0.012 −73.8%
Israel 344 −0.006 0.009 −45.0%
South Korea 645 −0.018 −0.005 −55.3%
Mexico 72 −0.010 0.000 −45.2%
Malaysia 846 −0.016 −0.034 −43.1%
Peru 53 −0.012 −0.009 −32.8%
Poland 190 −0.023 −0.009 −67.4%
Philippines 148 −0.019 0.012 −50.6%
Russia 31 −0.025 −0.011 −71.6%
South Africa 237 −0.018 −0.008 −50.8%
Thailand 363 −0.017 −0.005 −35.3%
Turkey 235 −0.035 −0.006 −60.4%
Taiwan 653 −0.011 −0.004 −35.1%
Emerging Average −0.016 −0.009 −47.1%

Table 1B
Summary statistics.
Panel B: Developed markets

Country N Global liquidity risk Local liquidity risk Cumulative returns

(1) (2) (3)

Australia 1028 −0.024 −0.024 −63.2%

Austria 36 −0.011 −0.004 −57.6%
Belgium 121 −0.008 −0.014 −42.2%
Canada 604 −0.018 −0.020 −57.9%
Denmark 137 −0.011 −0.011 −61.7%
Ireland 34 −0.011 −0.014 −67.8%
Finland 101 −0.013 −0.014 −50.5%
France 576 −0.011 −0.015 −47.1%
Germany 625 −0.014 −0.008 −47.6%
Hong Kong 783 −0.018 −0.014 −57.0%
Italy 185 −0.011 −0.012 −58.0%
Japan 2418 −0.006 −0.006 −27.1%
Netherlands 111 −0.011 −0.010 −52.9%
Norway 114 −0.016 −0.016 −60.0%
New Zealand 67 −0.018 −0.009 −51.8%
Singapore 467 −0.021 −0.021 −59.9%
Spain 96 −0.008 −0.004 −62.6%
Sweden 223 −0.017 −0.014 −52.5%
Switzerland 189 −0.009 −0.014 −39.3%
United Kingdom 1064 −0.015 −0.011 −61.4%
United States 1745 −0.007 −0.005 −47.3%
Developed Average −0.013 −0.011 −47.7%
Global Average −0.014 −0.010 −47.4%
Global Std. Dev. 0.016 0.014 30.6 %

Although the univariate relation between the ex-ante liquidity risk and the realized stock returns during the crisis lends support
for the predictability of the liquidity risk, other factors that are related to liquidity risk and have potential effects on the stock’s crisis
performance may drive this relation. Therefore, we rely on regression analyses to provide more robust results in the next section.

6
 T.L. Dang and T.M.H. Nguyen

Table 2
Correlation coefficients matrix. This table presents Pearson correlation coefficients among the variables used in this paper. CRet is the cumulative stock returns during the crisis period, defined from
January 2008 to March 2009. The liquidity risk (βLIQs) are measured with respect to either the global market portfolio (GβLIQ) or the local market portfolio (LβLIQ). Control variables include a stock's
return exposure to global (local) market return shocks (βMs), the MSCI index dummy (MSCI), return on asset (ROA), the log of book-to-market ratio (BM), the log of market capitalization (MCap), the
proportion of a firm’s shares that are closely held (CHeld), a dummy for whether a stock lists in the U.S. (ADR), the number of stock analysts following the firm (Analyst), the pre-crisis liquidity level of
individual stocks (PESprd) (we take the log of PESprd, denoted by Liq, in regression analyses), annual stock returns over 2007 (Return), stock return volatility over 2007 (STD), and the log of stock price at
the end of 2007 (Price).
Variable CRet GβLIQ GβM LβLIQ LβM MSCI ROA BM MCap CHeld ADR Analyst PESprd Return STD Price

CRet 1.000
GβLIQ 0.274 1.000
GβM −0.283 −0.671 1.000
LβLIQ 0.189 0.465 −0.376 1.000

7
LβM −0.162 −0.330 0.578 −0.387 1.000
MSCI 0.068 0.115 −0.061 0.161 0.020 1.000
ROA 0.061 0.086 −0.097 0.159 −0.143 0.166 1.000
BM 0.038 −0.059 0.003 −0.064 −0.049 −0.232 −0.039 1.000
MCap 0.029 0.206 −0.165 0.254 −0.124 0.695 0.253 −0.372 1.000
CHeld 0.027 −0.070 −0.017 0.014 −0.119 −0.242 0.081 0.061 −0.219 1.000
ADR −0.015 −0.011 0.029 0.005 −0.028 0.113 0.002 −0.046 0.210 −0.091 1.000
Analyst −0.004 0.118 −0.105 0.122 −0.120 0.332 0.117 −0.139 0.679 −0.249 0.243 1.000
PESprd −0.088 −0.134 0.095 −0.219 0.035 −0.429 −0.259 0.134 −0.563 0.173 −0.062 −0.295 1.000
Return −0.043 −0.026 0.070 0.024 0.000 0.073 0.190 −0.351 0.101 0.160 −0.033 −0.089 −0.050 1.000
STD −0.158 −0.201 0.263 −0.203 0.277 −0.135 −0.185 −0.191 −0.248 0.073 −0.052 −0.249 0.272 0.432 1.000
Price 0.049 0.208 −0.223 0.262 −0.188 0.328 0.245 −0.175 0.570 −0.142 0.082 0.347 −0.502 −0.027 −0.339 1.000
Research in International Business and Finance 52 (2020) 101165
 T.L. Dang and T.M.H. Nguyen

Table 3
Liquidity risk and stocks' cumulative returns during the crisis – Univariate analysis. This table reports cumulative returns from the beginning of 2008 until the end of each month of the extended period
from January 2008 to March 2009 for liquidity risk quintile portfolios for the global sample. The cumulative returns of a quintile portfolio are calculated as the equally weighted average of cumulative
returns across stocks in the portfolio. Quintile portfolios are formed based on the pre-crisis global (local) liquidity risk. t-statistics test for the difference in mean of the cumulative returns of the highest
liquidity risk quintile portfolio and the lowest liquidity risk quintile portfolio.
Panel A: Sorted by global liquidity risk

Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec-08 Jan-09 Feb-09 Mar-09

Quintile 1 (High risk) −14.0% −10.5% −18.9% −14.9% −13.1% −21.6% −24.8% −30.4% −43.0% −61.0% −63.8% −60.7% −62.7% −64.7% −61.5%
Quintile 2 −10.0% −5.6% −10.4% −7.6% −6.7% −15.3% −18.1% −23.7% −36.0% −52.6% −55.1% −51.9% −53.9% −56.6% −52.6%
Quintile 3 −7.7% −3.6% −6.9% −4.8% −3.3% −11.4% −14.0% −19.1% −30.0% −45.5% −48.2% −45.0% −47.5% −50.7% −46.6%
Quintile 4 −5.5% −2.1% −4.4% −3.2% −1.6% −9.1% −11.7% −16.2% −25.3% −38.7% −40.5% −37.2% −39.6% −43.5% −39.6%
Quintile 5 (Low risk) −4.1% −0.5% −3.8% −4.1% −2.7% −10.5% −12.1% −18.0% −25.5% −37.2% −38.1% −34.4% −36.1% −40.1% −35.7%

8
(High risk) - (Low risk) −9.9% −10.0% −15.0% −10.8% −10.3% −11.1% −12.7% −12.5% −17.5% −23.8% −25.6% −26.3% −26.5% −24.6% −25.8%
(t-statistics) −26.07 −18.12 −23.3 −14.67 −11.44 −10.81 −13.84 −14.53 −21.7 −33.57 −35.13 −33.18 −34.13 −33.61 −33.36

Panel B: Sorted by local liquidity risk

Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec-08 Jan-09 Feb-09 Mar-09

Quintile 1 (High risk) −11.0% −7.7% −14.7% −11.9% −8.9% −15.8% −21.3% −27.6% −39.7% −56.6% −59.2% −56.9% −58.7% −60.3% −58.4%
Quintile 2 −10.8% −7.9% −12.2% −9.0% −7.3% −15.3% −18.3% −23.0% −35.1% −51.6% −54.8% −51.8% −54.6% −58.1% −55.6%
Quintile 3 −8.8% −5.1% −7.8% −4.9% −3.4% −11.0% −13.4% −17.3% −28.9% −45.0% −48.1% −43.9% −47.2% −52.0% −48.0%
Quintile 4 −6.1% −0.4% −5.5% −4.9% −4.5% −14.8% −15.2% −22.2% −31.1% −44.2% −44.5% −40.2% −41.2% −43.9% −37.2%
Quintile 5 (Low risk) −5.2% −2.1% −5.3% −4.8% −3.8% −11.5% −13.5% −18.2% −26.3% −39.4% −41.3% −38.6% −40.4% −43.4% −39.4%
(High risk) - (Low risk) −5.7% −5.6% −9.4% −7.1% −5.1% −4.3% −7.8% −9.4% −13.4% −17.2% −17.9% −18.3% −18.4% −17.0% −19.0%
(t-statistics) −15.57 −10.38 −14.89 −10.21 −5.67 −4.48 −8.75 −11.16 −16.82 −23.6 −24.31 −22.93 −23.8 −22.65 −24.13
Research in International Business and Finance 52 (2020) 101165
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

Table 4
Effect of pre-crisis liquidity risk on stocks’ crisis performance. This table presents the regression results of cumulative stock
returns on stocks' pre-crisis global liquidity risk and firm-level control variables for the emerging (EMG), developed (DEV), and
global samples (GLB). Country and industry fixed effects are included (unreported). The regression model is
CReti = 0 + 1 iLIQ + Controlsi + i , where CReti denotes the cumulative stock return of stock i from January 2008 to March
2009. βLIQ
i is the pre-crisis global liquidity risk of stock i. Controlsi are the pre-crisis firm-specific characteristics, including the
stock's return exposure to market return shocks (βM), the MSCI index dummy (MSCI), return on asset (ROA), the log of book-to-
market ratio (BM), the log of market capitalization (MCap), the proportion of a firm’s shares that are closely held (CHeld), a
dummy for whether a stock lists in the U.S. (ADR), the number of stock analysts following the firm (Analyst), annual stock
returns (Return), stock return volatility (STD), the log of stock price at the end of 2007 (Price), and the log of the pre-crisis
liquidity level of individual stocks (Liq). All models are estimated with robust standard errors to correct for heteroscedasticity in
the firm-level errors terms. t-statistics are in parentheses. ***, **, and * denote statistical significance at the 1 %, 5 %, and 10 %
levels, respectively.
Variable EMG DEV GLB
(1) (2) (3)

βLIQ 1.041*** 0.838** 0.962***

(3.42) (2.30) (3.85)
βM −0.041*** −0.069*** −0.059***
(−4.97) (−8.14) (−9.67)
MSCI 0.036*** 0.015 0.024***
(4.17) (1.50) (3.52)
ROA 0.155*** 0.101*** 0.113***
(3.91) (3.76) (5.04)
BM 0.002 −0.001 −0.001
(0.26) (−0.26) (−0.20)
MCap −0.025*** −0.018*** −0.020***
(−6.02) (−4.74) (−7.33)
CHeld 0.046*** 0.047*** 0.043***
(3.63) (3.72) (4.78)
ADR 0.054** 0.061*** 0.056***
(2.34) (3.56) (4.07)
Analyst −0.001 −0.002*** −0.001*
(−0.60) (−2.67) (−1.81)
Liq 0.014* −0.024*** −0.010**
(1.84) (−4.36) (−2.29)
Return −0.030*** 0.033*** 0.010
(−3.18) (3.80) (1.56)
STD −0.089*** −0.076*** −0.084***
(−4.89) (−6.44) (−7.90)
Price 0.016*** −0.006** 0.003
(5.03) (−1.99) (1.26)

Fixed effects CI CI CI
Nobs 5,555 8,315 13,870
Adj. R-squared 33.1 % 24.7 % 27.0 %

4.2. Regression analysis

In this section, we perform the cross-sectional regression of the cumulative stock returns during the crisis on ex-ante global
liquidity risk while controlling for pre-crisis firm characteristics. Formally, our regression model is given as follows:
LIQ
CReti = 0 + 1 i + Controlsi + i (3)

where CReti denotes the cumulative stock return of stock i from January 2008 to March 2009. βLIQ i is the pre-crisis global liquidity risk
of stock i, which is estimated using monthly data from January 2003 to December 2007. Controlsi are the pre-crisis firm-specific
characteristics, including the stock's return exposure to global market return shocks (βM), the MSCI index dummy (MSCI), return on
asset (ROA), the log of book-to-market ratio (BM), the log of market capitalization (MCap), the proportion of a firm’s shares that are
closely held (CHeld), a dummy for whether a stock lists in the U.S. (ADR), the number of stock analysts following the firm (Analyst),
annual stock returns (Return), stock return volatility (STD), the log of stock price at the end of 2007 (Price), and the log of the pre-
crisis liquidity level of individual stocks (Liq).
To the extent that the crisis causes a macroeconomic recession which results in a negative impact on overall stock markets, this
effect will be reflected in the intercept term. We also include country-fixed effects to account for potential determining factors that
vary across countries and industry-fixed effects to control for different industry responses to the crisis. Regression is estimated with
robust standard errors to correct for heteroscedasticity in the firm-level errors terms.
Table 4 presents the regression results of cumulative stock returns on the pre-crisis global liquidity risk for the emerging, de-
veloped, and global samples. Consistent with our expectation, we find that ex-ante global liquidity risk is positively associated with

9
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

stocks’ crisis performance, suggesting that stocks with high pre-crisis exposure to global liquidity shocks exhibit larger declines in
prices during the crisis. Specifically, the coefficient estimate of βLIQ is significantly positive across subsamples, which varies from
0.838 (t-stat = 2.30) for developed markets to 1.041 (t-stat = 3.42) for emerging markets. The magnitude of the results is eco-
nomically significant. Taking the global sample as an example (Column 3), a one-standard-deviation increase in βLIQ (0.016) from
lower liquidity risk stock to higher liquidity risk stock (i.e., more negative) results in an approximately 1.5 % (=0.016*0.962) decline
in cumulative stock returns.
The coefficient estimate of the market risk (βM) is negative and significant at the conventional 1 % level in all specifications,
suggesting a “flight to quality” effect during the crisis. The statistical significance of other firm-specific characteristics also shows the
strong relevance of the control variables. In addition, we note that the regression R2s are relatively high given that all independent
variables are predetermined with respect to the financial crisis, suggesting their high collective predictability on stocks’ crisis per-
formance.

4.3. Robustness checks

In this section, we perform several robustness checks to assess whether our findings in the previous section are reliable. First,
instead of estimating a stock’s liquidity exposure to global market liquidity shocks (global liquidity risk), we measure its exposure to
local market liquidity shocks (local liquidity risk). Second, we repeat the regression analysis using the percentage quoted spread as a
liquidity measure to check whether the results are sensitive to spread-based liquidity measures.11 Third, it is likely that the results are
subject to the definition of the crisis period. To alleviate this concern, we use the two alternative definitions of the crisis: one from
January 2008 to June 2009, and another from September 2008, the date of the collapse of Lehman Brothers, to March 2009.
Panel A of Table 5 reports the cross-sectional regression results of Eq. (3) with the pre-crisis local liquidity risk. We find that the
coefficient estimate of the local liquidity risk is significantly positive only for the emerging markets, while the estimate for the
developed markets and the global sample is statistically insignificant. The weaker predictability of the local liquidity risk relative to
the global liquidity risk may be due to the global nature of the financial crisis of 2008–2009 in which the exposure of each country or
individual stock to the global factors seems to be more significant than the local factors.
Panel B of Table 5 presents the cross-sectional regression results of Eq. (3) that use the percentage quoted spread as a liquidity
proxy. The results are essentially similar to those using the percentage effective spread. The stocks with higher pre-crisis global
liquidity risk experience greater declines in prices during the crisis.
Table 6 reports the regression results of Eq. (3) using the alternative definitions of the crisis period. The coefficient estimates are
significantly positive for the emerging and global samples but are insignificant for the developed markets. The lower statistical
significance of coefficient estimates when we use the alternative definitions of the crisis period is probably because the impact of the
crisis becomes weaker in the second quarter of 2009 (for the period from January 2008 to June 2009). Alternatively, some effects in
the early stage of the crisis may have been missed when the period from September 2008 to March 2009 is defined as the crisis period.
Overall, these additional results confirm that our findings in the previous section are less affected by whether the stock's return
exposure is measured with respect to either global or local liquidity shocks, alternative spread-based liquidity measures, or when
using alternative definitions of the crisis period. These results provide evidence that pre-crisis liquidity risk helps predict the per-
formance of stocks during the global financial crisis of 2008–2009.

5. Conclusion

The global financial crisis of 2008–2009 ha s further emphasized the importance of liquidity risk in asset pricing, which has
attracted increased attention among researchers and policy makers alike over the two last decades. While most previous studies focus
on the role of liquidity risk as a priced risk factor, the ability of liquidity risk to explain realized stock returns when the market
experiences turbulence remains an empirical question. In this paper, we examine the effects of pre-crisis liquidity risk on stock
performance during the global financial crisis of 2008–2009. Using real-time transaction data for stocks across 41 countries to
measure stocks' return exposure to market liquidity shocks as a proxy for the stocks’ liquidity risk, we find that ex-ante liquidity risk
helps explain the realized returns of stocks during the crisis. Specifically, stocks with higher return exposure to market liquidity
shocks before the crisis exhibit greater declines in prices during the crisis. The negative impact of liquidity risk on stocks’ crisis
performance is consistent across the emerging, developed, and global samples and is robust to various spread-based measures that are
used as proxies for stock liquidity.
Lou and Sadka (2011) provide evidence that U.S. stocks with high pre-crisis liquidity risk decline more in prices during the
financial crisis. Using an approach similar to that in Lou and Sadka (2011), we extend their findings to the international financial
market and show that the negative effect of liquidity risk on stocks’ crisis performance persists across countries. Lee (2011) in-
vestigates the effect of liquidity as a risk factor on asset prices for stocks from 50 countries and finds that stocks’ expected returns
depend on the stocks’ liquidity risk. Our findings thus provide further insights into the comprehensive picture of the effect of liquidity
risk on asset prices in an international context and under different market conditions.
Our results suggest important implications for investors. Specifically, investors should be concerned with stocks with high

11
Goldreich et al. (2005) find that the quoted bid-ask spread has more explanatory power than the effective bid-ask spread on the prices of the
U.S. Treasury securities.

10
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

Table 5
Using local liquidity risk or alternative spread-based liquidity measure. This table presents the regression results of cumulative stock returns on
stocks' pre-crisis global liquidity risk and firm-level control variables for the emerging (EMG), developed (DEV), and global samples (GLB). Country
and industry fixed effects are included (unreported). In Panel A, the regression uses the local liquidity risk; in Panel B, the regression uses the
percentage-quoted spread as a liquidity measure. The regression model is CReti = 0 + 1 iLIQ + Controlsi + i , where CReti denotes the cumulative
stock return of stock i from January 2008 to March 2009. βLIQ i is the pre-crisis local liquidity risk of stock i (Panel A; in Panel B, we use global
liquidity risk). Controlsi are the pre-crisis firm-specific characteristics, including the stock's return exposure to market return shocks (βM), the MSCI
index dummy (MSCI), return on asset (ROA), the log of book-to-market ratio (BM), the log of market capitalization (MCap), the proportion of a firm’s
shares that are closely held (CHeld), a dummy for whether a stock lists in the U.S. (ADR), the number of stock analysts following the firm (Analyst),
annual stock returns (Return), stock return volatility (STD), the log of stock price at the end of 2007 (Price), and the log of the pre-crisis liquidity level
of individual stocks (Liq). All models are estimated with robust standard errors to correct for heteroscedasticity in the firm-level errors terms. t-
statistics are in parentheses. ***, **, and * denote statistical significance at the 1 %, 5 %, and 10 % levels, respectively.
Panel A: Local liquidity risk Panel B: Alternative liquidity measure

Variable EMG DEV GLB EMG DEV GLB

(1) (2) (3) (4) (5) (6)

βLIQ 1.987*** −0.356 0.438 1.230*** 0.643* 0.909***

(4.75) (−0.98) (1.58) (4.04) (1.72) (3.56)
βM −0.099*** −0.105*** −0.102*** −0.038*** −0.072*** −0.060***
(−9.26) (−12.41) (−15.26) (−4.58) (−8.29) (−9.66)
MSCI 0.038*** 0.022** 0.028*** 0.036*** 0.014 0.024***
(4.40) (2.24) (4.24) (4.22) (1.44) (3.52)
ROA 0.135*** 0.105*** 0.110*** 0.156*** 0.100*** 0.113***
(3.45) (3.92) (4.90) (3.93) (3.74) (5.04)
BM 0.000 −0.004 −0.003 0.002 −0.001 −0.001
(0.01) (−0.83) (−0.92) (0.30) (−0.27) (−0.19)
MCap −0.029*** −0.021*** −0.025*** −0.025*** −0.018*** −0.020***
(−7.12) (−5.72) (−8.95) (−6.08) (−4.87) (−7.34)
CHeld 0.041*** 0.038*** 0.037*** 0.045*** 0.049*** 0.044***
(3.29) (3.07) (4.09) (3.61) (3.91) (4.81)
ADR 0.050** 0.060*** 0.056*** 0.054** 0.059*** 0.056***
(2.17) (3.49) (3.98) (2.32) (3.48) (4.03)
Analyst −0.002 −0.002*** −0.001** −0.001 −0.002*** −0.001*
(−1.38) (−2.89) (−2.02) (−0.62) (−2.83) (−1.85)
Liq 0.010 −0.025*** −0.012*** 0.014** −0.024*** −0.009**
(1.33) (−4.63) (−2.93) (2.07) (−4.64) (−2.28)
Return −0.026*** 0.023*** 0.004 −0.031*** 0.033*** 0.010
(−2.82) (2.67) (0.67) (−3.27) (3.83) (1.50)
STD −0.066*** −0.061*** −0.065*** −0.088*** −0.077*** −0.084***
(−3.81) (−5.43) (−6.63) (−4.83) (−6.47) (−7.91)
Price 0.011*** −0.005* 0.001 0.016*** −0.006** 0.003
(3.46) (−1.76) (0.51) (4.98) (−1.97) (1.29)

Fixed effects CI CI CI CI CI CI
Nobs 5,556 8,315 13,871 5,555 8,315 13,870
Adj. R-squared 34.1 % 25.1 % 27.6 % 33.2 % 24.7 % 27.0 %

liquidity risk because those stocks are more likely to experience greater declines in prices during a crisis. In addition, because
covariances (variances) tend to be more persistent, liquidity risk, albeit an estimate itself, could provide a better predictability for
stocks’ out-of-sample performance.

Acknowledgements

We would like to thank the members of the UE-UD Teaching and Research Team in Corporate Finance and Asset pricing (TRT-
CFAP) for their helpful comments and suggestions. A part of this paper was completed while Tung Lam Dang was at Institute of
Global Finance, UNSW Business School, The University of New South Wales. Thi Minh Hue Nguyen acknowledges the research grant
(Grant number 502.02-2015.15) from the Vietnam National Foundation for Science and Technology Development (NAFOSTED) for
this project. All errors remain our own.

11
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

Table 6
Alternative definitions of the crisis period. This table presents the regression results of cumulative stock returns on stocks' pre-crisis global liquidity
risk and firm-level control variables for the emerging (EMG), developed (DEV), and global samples (GLB). Country and industry fixed effects are
included (unreported). The regression model is CReti = 0 + 1 iLIQ + Controlsi + i , where CReti denotes the cumulative stock return of stock i over
two alternative crisis periods from January 2008 to June 2009 (Panel A) and from September 2008 to March 2009 (Panel B). βLIQ i is the pre-crisis
global liquidity risk of stock i. Controlsi are the pre-crisis firm-specific characteristics, including the stock's return exposure to market return shocks
(βM), the MSCI index dummy (MSCI), return on asset (ROA), the log of book-to-market ratio (BM), the log of market capitalization (MCap), the
proportion of a firm’s shares that are closely held (CHeld), a dummy for whether a stock lists in the U.S. (ADR), the number of stock analysts
following the firm (Analyst), annual stock returns (Return), stock return volatility (STD), the log of stock price at the end of 2007 (Price), and the log
of the pre-crisis liquidity level of individual stocks (Liq). All models are estimated with robust standard errors to correct for heteroscedasticity in the
firm-level errors terms. t-statistics are in parentheses. ***, **, and * denote statistical significance at the 1 %, 5 %, and 10 % levels, respectively.
Panel A: January 2008 to June 2009 Panel B: September 2008 to March 2009

Variable EMG DEV GLB EMG DEV GLB

(1) (2) (3) (4) (5) (6)

βLIQ 1.575*** 0.633 1.043*** 1.045*** 0.459 0.721***

(3.95) (1.31) (3.19) (3.14) (1.30) (2.91)
βM −0.039*** −0.055*** −0.045*** −0.011 −0.046*** −0.032***
(−3.65) (−5.16) (−5.93) (−1.32) (−5.08) (−4.95)
MSCI 0.074*** 0.074*** 0.076*** 0.018** −0.004 0.006
(6.41) (5.35) (8.12) (2.01) (−0.32) (0.79)
ROA 0.199*** 0.152*** 0.153*** 0.082* 0.065** 0.068***
(3.99) (3.88) (4.76) (1.92) (2.18) (2.73)
BM 0.006 −0.003 −0.002 −0.001 −0.011** −0.008**
(0.84) (−0.50) (−0.42) (−0.17) (−2.57) (−2.31)
MCap −0.039*** −0.029*** −0.032*** −0.018*** −0.012*** −0.014***
(−7.35) (−5.81) (−8.61) (−4.59) (−3.33) (−5.27)
CHeld 0.042*** 0.057*** 0.051*** 0.024** 0.029** 0.025***
(2.62) (3.42) (4.31) (2.00) (2.54) (2.97)
ADR 0.016 0.045* 0.030 0.049** 0.095** 0.080**
(0.66) (1.91) (1.61) (2.07) (2.07) (2.45)
Analyst 0.002 −0.001 0.001 −0.002 −0.001* −0.001*
(1.16) (−0.62) (0.65) (−1.43) (−1.96) (−1.76)
Liq −0.014 −0.015** −0.008 −0.002 −0.015*** −0.009**
(−1.54) (−2.15) (−1.39) (−0.34) (−3.24) (−2.39)
Return −0.061*** −0.003 −0.023** −0.010 0.029*** 0.015**
(−5.07) (−0.21) (−2.54) (−1.12) (3.31) (2.38)
STD −0.080*** −0.040** −0.058*** −0.047*** −0.042*** −0.046***
(−3.85) (−2.37) (−4.34) (−3.42) (−3.97) (−5.45)
Price 0.016*** −0.008** −0.000 0.011*** −0.005 0.001
(4.02) (−2.17) (−0.13) (3.75) (−1.59) (0.40)

Fixed effects CI CI CI CI CI CI
Nobs 5,531 8,240 13,771 5,562 8,325 13,887
Adj. R-squared 24.1 % 16.7 % 18.4 % 65.4 % 18.4 % 45.0 %

Appendix A. Variable definitions

Variables Acronym Definition Data Sources

(i) Liquidity measures

Percentage-effective spread PESprd Twice the absolute value of the difference between the trading price and the midpoint of the bid and TRTH
ask price, divided by the midpoint of the bid and ask price.
Percentage-quoted spread PRSprd The absolute value of the difference between the ask and bid price, divided by the midpoint of the TRTH
bid and ask price.
(ii) Liquidity risk measure
Ex-ante liquidity risk βLIQ The covariance between a stock’s returns with the global (local) market liquidity, which is estimated Datastream
using monthly stock liquidity and return data before the crisis from January 2003 to December 2007. and TRTH
(iii) Pre-crisis firm level controls
Stock's return exposure to m- βM The covariance between a stock’s returns with the global (local) market returns, which is estimated Datastream
arket return shocks using monthly return data before the crisis from January 2003 to December 2007.
Inclusion in country index MSCI An MSCI index member dummy that equals one if the firm is included in an MSCI country index at Worldscope
the end of 2007.
Return on asset ROA Return on asset ratio in 2007. Worldscope
Book-to-market ratio BM Log of book-to-market equity ratio as of June 2007. Worldscope
Firm size MCap Log of market capitalization denominated in U.S. dollars at the end of 2007. Worldscope

12
T.L. Dang and T.M.H. Nguyen Research in International Business and Finance 52 (2020) 101165

Closely held ownership CHeld Fraction of shares closely held by insiders and controlling shareholders in 2007. Worldscope
US cross-listing ADR An ADR dummy that equals one if the firm was cross-listed on a U.S. exchange at the end of 2007. Worldscope
Analyst coverage Analyst Number of financial analysts covering a firm in 2007. (I/B/E/S)
Pre-crisis liquidity level Liq The log of the average of the daily percentage-effective (quoted) spread over 2007. TRTH
Annual stock returns Return Annual stock returns in 2007. Datastream
Stock return volatility STD Annualized standard deviation of monthly stock returns in 2007. Datastream
Stock price Price Log of stock price at the end of 2007. Datastream

References

Abdi, F., Ranaldo, A., 2017. A simple estimation of bid-ask spreads from daily close, high, and low prices. Rev. Financ. Stud. 30, 4437–4480.
Acharya, V.V., Pedersen, L.H., 2005. Asset pricing with liquidity risk. J. financ. econ. 77, 375–410.
Ahn, H.-J., Cai, J., Yang, C.-W., 2018. Which liquidity proxy measures liquidity best in emerging markets? Economies 6, 67.
Amihud, Y., Mendelson, H., 1986. Asset pricing and the bid-ask spread. J. Financ. Econ. 17, 223–249.
Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series effects. J. Financ. Mark. 5, 31–56.
Aragon, G.O., Strahan, P.E., 2012. Hedge funds as liquidity providers: evidence from the Lehman bankruptcy. J. Financ. Econ. 103, 570–587.
Bali, T.G., Peng, L., Shen, Y., Tang, Y., 2014. Liquidity shocks and stock market reactions. Rev. Financ. Stud. 27, 1434–1485.
Będowska-Sójka, B., Echaust, K., 2019. Commonality in liquidity indices: the emerging European stock markets. Systems 7, 1–18.
Bekaert, G., Ehrmann, M., Fratzscher, M., Mehl, A., 2014. The global crises and equity market contagion. J. Finance 69, 2597–2649.
Bekaert, G., Harvey, C.R., Lundblad, C., 2007. Liquidity and expected returns: lessons from emerging markets. Rev. Financ. Stud. 20, 1783–1831.
Beber, A., Brandt, M.W., Kavajecz, K.A., 2009. Flight-to-quality or flight-to-liquidity? Evidence from the Euro-area bond market. Rev. Financ. Stud. 22, 925–957.
Brennan, M.J., Subrahmanyam, A., 1996. Market microstructure and asset pricing: on the compensation for illiquidity in stock returns. J. Financ. Econ. 41, 441–464.
Cao, C., Petrasek, L., 2014. Liquidity risk in stock returns: an event-study perspective. J. Bank. Financ. 45, 72–83.
Chen, Y., Eaton, G.W., Paye, B.S., 2018. Micro(structure) before macro? The predictive power of aggregate illiquidity for stock returns and economic activity. J.
Financ. Econ. 130, 48–73.
Chordia, T., Roll, R., Subrahmanyam, A., 2000. Commonality in liquidity. J. Financ. Econ. 56, 3–28.
Chung, K.H., Chuwonganant, C., 2018. Market volatility and stock returns: the role of liquidity providers. J. Financ. Mark. 37, 17–34.
Datar, V.T., Naik, N.Y., Radcliffe, R., 1998. Liquidity and stock returns: an alternative test. J. Financ. Mark. 1, 203–219.
Díaz, A., Escribano, A., 2020. Measuring the multi-faceted dimension of liquidity in financial markets: a literature review. Res. Int. Bus. Financ. 51, 101079.
Dinh, M.T.H., 2017. The returns, risk and liquidity relationship in high frequency trading: evidence from the Oslo stock market. Res. Int. Bus. Financ. 39, 30–40.
Eleswarapu, V.R., 1997. Cost of transacting and expected returns in the Nasdaq market. J. Finance 52, 2113–2127.
Fong, K.Y.L., Holden, C.W., Trzcinka, C.A., 2017. What are the best liquidity proxies for global research? Rev. Financ. 21, 1355–1401.
Foran, J., Hutchinson, M.C., O’Sullivan, N., 2015. Liquidity commonality and pricing in UK equities. Res. Int. Bus. Financ. 34, 281–293.
Goldreich, D., Hanke, B., Nath, P., 2005. The price of futures liquidity: time-varying liquidity in the U.S. treasury market. Rev. Financ. 9, 1–32.
Goyenko, R.Y., Holden, C.W., Trzcinka, C.A., 2009. Do liquidity measures measure liquidity? J. Financ. Econ. 92, 153–181.
Hameed, A., Kang, W., Viswanathan, S., 2010. Stock market declines and liquidity. J. Finance 65, 257–292.
Ho, T.-W., Chang, S.-H., 2015. The pricing of liquidity risk on the Shanghai stock market. Int. Rev. Econ. Financ. 38, 112–130.
Karolyi, A., Lee, K.-H., van Dijk, M., 2012. Understanding commonality in liquidity around the world. J. Financ. Econ. 105, 82–112.
Kempf, A., Mayston, D., 2008. Liquidity commonalilty beyond best prices. J. Financ. Res. 31, 25–40.
Lang, M., Maffett, M., 2011. Transparency and liquidity uncertainty in crisis periods. J. Account. Econ. 52, 101–125.
Lee, K.-H., 2011. The world price of liquidity risk. J. Financ. Econ. 99, 136–161.
Lesmond, D., 2005. Liquidity of emerging markets. J. Financ. Econ. 77, 411–452.
Liu, W., 2006. A liquidity-augmented capital asset pricing model. J. Financ. Econ. 82, 631–671.
Lou, X., Sadka, R., 2011. Liquidity level or liquidity risk? Evidence from the financial crisis. Financ. Anal. J. 67, 51–62.
Ma, R., Anderson, H.D., Marshall, B.R., 2018. Market volatility, liquidity shocks, and stock returns: worldwide evidence. Pacific-Basin Financ. J. 49, 164–199.
Moshirian, F., Qian, X., Wee, C.K.G., Zhang, B., 2017. The determinants and pricing of liquidity commonality around the world. J. Financ. Mark. 33, 22–41.
Pástor, L., Stambaugh, R.F., 2003. Liquidity risk and expected stock returns. J. Polit. Econ. 111, 642–685.
Raddatz, C., Schmukler, S.L., 2012. On the international transmission of shocks: micro-evidence from mutual fund portfolios. J. Int. Econ. 88, 357–374.
Sadka, R., 2006. Momentum and post-earnings-announcement drift anomalies: the role of liquidity risk. J. Financ. Econ. 80, 309–349.
Tong, H., Wei, S.-J., 2011. The composition matters: capital inflows and liquidity crunch during a global economic crisis. Rev. Financ. Stud. 24, 2023–2052.
Watanabe, A., Watanabe, M., 2008. Time-varying liquidity risk and the crosssection of stock returns. Rev. Financ. Stud. 21, 2449–2486.

13