Journal of Asset Management (2020) 21:355–373

https://doi.org/10.1057/s41260-020-00164-3

ORIGINAL ARTICLE

Liquidity commonality beyond best prices: Indian evidence

Abhinava Tripathi1   · Vipul1 · Alok Dixit1

Revised: 7 March 2020 / Published online: 20 May 2020

© Springer Nature Limited 2020

Abstract
This paper investigates the phenomenon of liquidity commonality in the Indian market, for the 3-year period spanning
2015–2017. Principal component analysis and canonical correlation analysis are performed on a number of liquidity proxies
to provide the evidence of commonality. Following the market model of Chordia et al. (J Financ Econ 56(1):3–28, 2000), we
show that the level of liquidity commonality for this order-driven emerging market is higher, as compared to that reported
for quote-driven markets. There is evidence to suggest that both the inventory and information asymmetry hypotheses sig-
nificantly contribute to the commonality in this important emerging market. Commonality is observed even at the deeper
levels beyond the best prices, indicating a significant role of information asymmetry hypothesis. The study also provides
evidence of specific intraday patterns in liquidity and ‘free-entry and free-exit’ phenomenon.

Keywords  Liquidity commonality · Market microstructure · Principal component analysis · Limit-order book · Information
asymmetry

JEL Classifications  G12 · G14 · C33

Introduction 2004; Hameed et al. 2010; Huberman and Halka 2001).

The demand-side influences include the correlated trading
Liquidity commonality refers to the proposition that sys- activity of institutional investors, mutual funds, and index-
tematic factors affect the liquidity of individual securities. based funds (Chordia et al. 2000; Domowitz et al. 2005;
The literature on quote-driven markets has extensively docu- Hasbrouck and Seppi 2001; Kamara et al. 2008; Koch et al.
mented the phenomenon of liquidity commonality, particu- 2016). However, the evidence from these studies is contrast-
larly for the developed markets like the U.S. (Chordia et al. ing. For example, Chordia et al. (2000, henceforth "CRS")
2000; Comerton-Forde et al. 2010; Hasbrouck and Seppi and Huberman and Halka (2001) found reasonably strong
2001; Huberman and Halka 2001). Extant literature attrib- evidence, whereas Hasbrouk and Seppi (2001) found weak
utes this phenomenon to certain supply-side and demand- evidence in support of commonality. This contrast may be
side influences. The supply-side influences include financing on account of the differences in the computation of liquidity
constraints of market participants (Coughenour and Saad measures. Hasbrouck and Seppi work with levels, whereas
CRS works with first-differences (movements). To address
this issue, this study examines the commonality both at the
The authors thank Prof. Marielle de Jong (the Editor) and
anonymous reviewer for their in-depth review of the manuscript levels and first-differences for an important emerging order-
and insightful comments. driven market, i.e., India.
With decelerating growth in developed markets, emerging
* Abhinava Tripathi markets (e.g., China, India, Brazil) offer important portfolio
[email protected]
diversification opportunities and relatively high returns to
Vipul global investors.1 Emerging markets comprise sixty percent
[email protected]
of the worldwide GDP (PPP-adjusted). 2 These markets,
Alok Dixit
1
[email protected]   https ​ : //www.imf.org/en/Publi ​ c atio ​ n s/WEO/Issue ​ s /2019/07/18/
WEOup​dateJ​uly20​19.
1 2
Finance and Accounting, Indian Institute of Management   https​://www.imf.org/exter​nal/datam​apper​/NGDP_RPCH@WEO/
Lucknow, Lucknow, India OEMDC​/ADVEC​/WEOWO​RLD.

Vol.:(0123456789)

356 A. Tripathi et al.

however, fall behind developed markets on key parameters et al. 2014). For example, Fabre and Frino (2004) examined
related to market efficiency, institutional strength, financial 660 stocks listed on the Australian Stock Exchange (ASX)
liberalization and reforms. Therefore, the emerging markets for the calendar year 2000. They followed the market model
are likely to exhibit substantially disparate liquidity char- of CRS and employed the liquidity proxies: the quoted
acteristics, as compared to their developed counterparts. spread; percentage quoted spread; depth; and dollar depth.
Most importantly, these markets share a common charac- They found weak evidence for commonality with adj. R2 of
teristic, i.e., a majority of them employ limit-order book less than 1%. Bauer (2004) studied 19 stocks listed on the
(LOB) mode of trading. These LOBs do not have designated Swiss Stock Exchange (SWX) for a 3-month period between
market makers for the supply of liquidity and execution of May and July 2002. They performed PCA on the depth and
trades. Conversely, they rely on the de-facto market mak- spread measures, and found a strong evidence for common-
ing by limit order providers.3 However, the efficacy of this ality with the first principal component accounting for 43%
de-facto market making in emerging markets is less-well of the overall variation. Brockman and Chung (2002) stud-
understood (Tripathi et al. 2019). In this context, the Indian ied 725 stocks listed on the Stock Exchange of HongKong
market is particularly important. This market is similar to (SEHK) for the 4-year period spanning 1996–1999. Follow-
developed markets in terms of size, the stability of macro- ing the market model of CRS, they employed the liquid-
environment, and the robustness of trading infrastructure ity proxies: the relative and absolute spread measures, and
(trading, margining, and settlement systems). But, it exhibits volume and dollar depth measures. They found modest evi-
a risk profile that is comparable to emerging markets (e.g., dence in support of liquidity commonality with adj. R2 in
regulatory framework that governs the flow of capital, eco- the range of 1–3%. They also found positive and significant
nomic growth on the back of sizeable domestic demand, and levels of commonality for more than 57% of companies in
a pure LOB microstructure). their sample-set (as compared to 35% companies in CRS).
Using 1169 stocks listed on the NYSE, CRS showed that In short, there is evidence of commonality in order-driven
the market and industry aggregates of liquidity affect the markets, but there are very few studies on emerging mar-
liquidity of individual stocks. They contended that the inven- kets that examine the order book and liquidity characteristics
tory considerations drive the marketwide liquidity changes, beyond the best prices.
whereas the information asymmetry drives the industrywide In order-driven markets, the market participants are not
liquidity changes. They found the industry commonality to obliged to submit limit orders; i.e., no liquidity provider
be relatively weak, which supported the dominance of inven- of last resort exists. For such a market microstructure, the
tory considerations over the information asymmetry para- literature proposes the ‘free-entry and free-exit’ (hereaf-
digm. The relative importance of these two influences is an ter, “FE”) hypothesis (Brockman and Chung 2002). The
important research issue in the current study. Hasbrouk and hypothesis suggests that during liquidity shocks, various
Seppi (2004), employing canonical correlation and princi- (de-facto) market makers withdraw liquidity provisions, or
pal component analysis, found that common factors affect make them extremely costly. Further, as the liquidity pro-
the returns and order-flows. There is evidence that, in the visioning becomes costly, these limit order providers enter
quote-driven markets, the order-flow co-movement drives the market. Therefore, the large movements in liquidity, both
the return co-movement, and the order-type co-movement up and down (entry and exit of liquidity), are expected to
drives the liquidity co-movement (Domowitz et al. 2005; occur together. That is, high volatility in liquidity should
Hasbrouck and Seppi 2001). Following the canonical cor- indicate a high level of commonality. This study contributes
relation methodology, the current study attempts to address to the literature on liquidity-volatility (Petkova et al. 2011)
this question by employing a large number of liquidity by examining this hypothesis in the context of the Indian
proxies. market, which is an important pure order-driven emerging
In contrast to quote-driven markets, the order-driven mar- market.
kets (e.g., India) are devoid of designated market makers. The study makes the following important contributions
This has important implications for the inventory consid- to the literature. There are very few studies that examine
erations, which are central to the commonality literature the liquidity of emerging markets beyond the best prices
in quote-driven markets. The literature on emerging LOB (Anagnostidis et al. 2016). This study offers an important
markets is silent on this aspect of market microstructure and contribution in this direction. The study examines a plethora
offers contrasting evidence. (Alhomaidi et al. 2018; Anag- of order book and trade book-related measures, at levels and
nostidis et al. 2016; Kumar and Prasanna 2018; Syamala first-differences, by employing various methodologies (PCA,
canonical correlations, CRS regressions) to provide robust
and reliable evidence on liquidity commonality. Further-
3
 In LOBs, limit-orders supply liquidity, and therefore, limit order more, the study also contributes to the literature by providing
suppliers become the de-facto market makers. evidence on the role of liquidity-volatility on commonality
Liquidity commonality beyond best prices: Indian evidence 357

in the context of an order-driven emerging market. The study facilitate synchronous trading across these segments.7 This
makes two important methodological contributions. First, it is unlike the fragmented nature of developed markets where
employs an orthogonalized measure of industry commonal- trading takes place at cash equity and derivative segments
ity, which is less vitiated by the marketwide influences. Sec- in an independent manner; e.g., derivative trading at CBOE
ond, it employs volume-weighted average prices (VWAP) and cash equity trading at NYSE and NASDAQ in the U.S.
for the computation of returns and bid-ask spreads to contain Market.8 In this market microstructure, the securities that
the effect of noise. trade on both the derivatives and cash equity segments are
The study reports several interesting results. First, it particularly important to price discovery and market effi-
shows that the individual firm-specific liquidity co-moves ciency, and hence, are the focus of this study.
with the market and industry liquidity, and this commonal- For the year ending December 31, 2017,9 NSE was the
ity phenomenon extends beyond the best bid-ask prices. As third-largest stock market in terms of equity trading, with
compared to quote-driven markets, the industry commonal- a trading volume of 2.3 billion trades (preceded only by
ity plays a more important role in explaining the liquidity of Shanghai Stock Exchange and Shenzhen Stock Exchange).
individual stocks, and the levels of commonality are much The total turnover of NSE equity market segment for the
higher. Second, the results suggest that the order book levels year 2017 was USD 1013.34 billion. In terms of market capi-
beyond the best prices also carry the information available talization, it ranked ninth globally (USD 2351.46 billion).
with the de-facto market makers. Third, the results support NSE is a pure LOB market that operates without designated
the FE hypothesis of liquidity in LOB markets. The findings market makers. As the system receives orders, it time-stamps
of this study are important to investors, academics, policy- and processes these orders for the best match. The market
makers, and regulators—who are interested in the liquidity orders (buy/sell) are immediately executed at the best avail-
dynamics of emerging markets. able (sell/buy) prices. The limit orders are stored in the order
The remainder of this paper proceeds as follows. “Back- book as per the price-time priority. These limit orders are
ground and microstructure of the Indian market” section matched for execution with the incoming stream of orders,
provides background information about NSE. “Data and according to their position in the order book.
empirical measuresThe study employs the high-frequency
trade and Methodology” sections discuss the data and meth-
odology. “Results” section discusses the evidence from vari- Data and empirical measures
ous tests of commonality. “Summary” section summarizes
the results. The study employs the high-frequency trade book data and
the hourly LOB snapshot data for the equity market securi-
ties. The data is provided by NSE. The period of three years
Background and microstructure from January 1, 2015 to December 31, 2017 is covered. Each
of the Indian market trade book data file contains daily time-stamped transaction
records. The LOB data comprises hourly snapshots for each
The emerging market of India is ranked third (sixth)in day, at 10 A.M., 11 A.M., 12 Noon, 1 P.M., and 2 P.M. The
terms of PPP-adjusted (nominal) GDP, with a size of USD choice of the study period is guided by the prior literature in
10.51 trillion (USD 2.94 trillion, 2019).4 Foreign portfo- the area that employs high-frequency trade book and order
lio investors (FPI) hold more than 25% share of the Indian book/quote dataset (Anagnostidis and Fontaine 2019; Chor-
equity market5; over the 8-year period (2010–2018), these dia et al. 2000; Domowitz et al. 2005; Hasbrouck and Seppi
FPIs have made a net cumulative investment of about USD 2001; Huberman and Halka 2001; Kang and Zhang 2013;
164.20 billion. NSE, the leading stock exchange of India, Kempf and Mayston 2008; Sensoy 2016, 2019; Syamala
operates in the equity, derivatives, and debt market seg- et al. 2014). The most recent period has been selected for
ments. It accounts for more than Ninety percent of equity which the order book and trade book data were available at
trading in the Indian market.6 NSE executes trades in cash
and derivative segments with rules and trading protocols that
7
  For instance, in case the market-wide circuit breaker is initiated in
the cash segment, the F&O segment automatically shuts down.
8
4
 CBOE: Chicago Board of Exchange, NYSE: New York Stock
  https ​ : //econo ​ m icti ​ m es.india ​ t imes ​ . com/news/econo ​ m y/india​ Exchange, and NASDAQ: National Association of Securities Dealer
-becom​es-5th-large​st-econo​my-overt​akes-uk-franc​e-repor ​t/artic​lesho​ and Automated Quotations.
w/74181​024.cms?from=mdr. 9
5
  https​://www.world​-excha​nges.org/home/index​.php/stati​stics​/marke​
  https​://www1.nsein​dia.com/conte​nt/us/ismr_full2​019.pdf. t-highl​ights​. For the 6-month period ending June 2018, NSE was the
6
  https​://focus​.world​-excha​nges.org/issue​/janua​r y-2020/marke​t-stati​ second largest exchange in terms of the number of trades in equity
stics​. shares.

358 A. Tripathi et al.

the design stage of the study. The study period includes suf- both stacks [i.e., up to L5, L10, L50, L100, and complete13
ficient observations to permit statistically significant results (LFull)] represents the depth at the corresponding level. The
and reliable inferences. RS measure is computed as follows.
PA − PB
Security sample‑set RS = (1)
PM
At the first stage of filtration, we select those securities (226) where PM = (PA + PB )∕2 denotes the mid-price.
that traded on the Futures and Options (F&O) segment, dur- The application of VWAP measure for computation of
ing the study period. NSE follows a robust criterion10 to informationally efficient price measures is widely docu-
ensure that only the heavily traded stocks are included in mented (Cao et al. 2009; Stoll 2000). We compute a VWAP
their F&O segment. These securities broadly represent the measure of spread by employing the deeper LOB levels
entire equity market in terms of industry-sector classifica- beyond the best prices (e.g., L5, L10, L50, L100, and com-
tion, and constitute about eighty-five percent of the NSE plete). The VWAP bid/ask prices at the L-th level of LOB
market capitalization. Out of these, only those securities are are computed as in Eq. (2).
retained which traded for all the 743 trading days with a
minimum daily trade volume of 10,000 units. Further, the
( L )/( L )
∑ ∑
(2)
VWAP
following criteria are applied to the observations corre- PA,L = Qi Pi Qi
sponding to intraday snapshots: (1) at least 100 limit orders i=1 i=1

on both bid and ask sides, (2) observations with negative Here PVWAP represents the VWAP ask measure at the L-th
spreads are discarded,11 (3) at least one transaction during A,L
level of the order book. Pi and Qi are the price and quantity at
the 5-s window, after the snapshot, and (4) at least 100 share the i-th level of ask prices. The VWAP bid measures PVWAP
transaction volume during 1-min period after the snapshot. are computed in a similar manner. These VWAP measures
B,L

This reduces the final sample-set to 197 securities. These of bid/ask are used to compute the VWAP measure of RS
securities are heavily traded, and arguably are the most for the L-th level of LOB, using Eq. (1).
informationally efficient securities available on NSE. These
securities are particularly suitable for this study, as the effect Construction of return, volume, order‑imbalance,
of noise is expected to be considerably less on estimation.12 and volatility measures
All the variables are 0.5% truncated from the upper and
lower extremities. The analysis involves 1-, 5-, 10-, 15-, 30-, and 60-min
VWAP returns corresponding to each snapshot. A 30-s14
Liquidity measures window (henceforth “trade aggregation window” or “trade
window”) is employed to aggregate the trade prices and
The study employs order book snapshot data for the com- volumes to construct the VWAP based price measures.
putation of liquidity measures. The two important liquid- For example, at 11 A.M., a 30-s trade aggregation window
ity measures employed in the study are depth and relative from 11:00:01 A.M. to 11:00:30 A.M is used to construct
spread measures (RS: Amihud and Mendelson 1986). For VWAP for security j, Pj,11 . Similarly, at 1-, 5-, 10-, 15-, 30-
each snapshot (e.g., 10 A.M.), the buy and sell orders are and 60-min15 periods (i.e., 11:01:00, 11:05:00, 11:10:00,
stacked in descending and ascending order, respectively. The 11:15:00, 11:30:00, 12:00:00), VWAP prices ( Pj,t+k : t = 11,
orders in both these stacks at the top are the best ask or bid k = 1, 5, 10, … 60 min) are calculated by aggregating trades
prices (lowest ask price PA , and the highest bid price PB ). for 30-s windows each. These VWAPs are used to construct
The average of the share volumes associated with these bid
and ask prices represents the depth at level 1 (L1). Simi-
larly, the average of the share volumes up to deeper levels in
13
  The analysis employs deeper levels of the order book beyond the
best prices. For instance, the information pertaining to the top five
buy and sell limit orders is referred to as L5. Similarly, L10 captures
the information pertaining to the best ten buy and sell limit orders,
and so on for L50, L100, and complete order book.
10
  https​://www.nsein​dia.com/produ​cts/conte​nt/deriv​ative​s/equit​ies/ 14
 For robustness checks, in addition to the 30-s VWAP price, the
selec​tion_crite​ria.htm. study also employs 5-, 10-, 15-, 20-, and 25-s VWAP prices. The
11
  Negative spread measures indicate data recording errors. results remain very similar; therefore, for brevity, we report results
12
  We aim to select the stocks that are informationally efficient and based on 30-s window.
15
least influenced by the information asymmetry. This ensures that the  The 60-min VWAP returns are employed in the CRS (2000)
evidence in favor of information asymmetry paradigm is robust and method, and the other return measures (1-, 5-, 10-, 15-, 30-min) are
reliable. employed in the PCA and canonical correlations.
Liquidity commonality beyond best prices: Indian evidence 359

the VWAP returns for 1-, 5-, 10-, 15-, 30-, and 60-min inter- of all the common marketwide effects on securities. Next,
vals. The VWAP return calculation procedure with VWAPs, the canonical correlation analysis is a more robust method
Pj,t and Pj,t+k , for security j is as follows. as compared to the simple pairwise correlation analysis. It is
(t+30s )/(t+30s ) particularly useful while comparing a set of multiple covari-
∑ ∑ ates, where a pairwise comparison can become difficult to
Pj,t = Volj,𝜏 Pj,𝜏 Volj,𝜏 (3a)
𝜏=t 𝜏=t
carry out and comprehend.19

(t+k+30s )/(t+k+30s ) Market model of CRS

∑ ∑
Pj,t+k = Volj,𝜏 Pj,𝜏 Volj,𝜏 (3b)
𝜏=t+k 𝜏=t+k The CRS approach to commonality involves regressions of
individual stock liquidity on aggregate liquidity. These mar-
Here t represents the timing of snapshots, i.e., 10 A.M., 11 ket model regressions employ( L )first-differenced liquidity
A.M., 12 Noon, 1 P.M., and 2 P. M. k represents the periods measures. Here, DLj,t = ln L j,t denotes the change in the
of 1-, 5-, 10-, 15-, 30-, and 60-min after the snapshot, for j,t−1
liquidity measure Lj,t from time t − 1 to t for stock j. The
which the return is calculated. V ­ olj,τ and Pj,τ are volume and
study employs hourly changes in the RS measure.
price of security(j at time τ.)
Following CRS, the market liquidity ( DLM,t ) measure is
Now, rj,t = ln Pj,t+k∕ Pj,t is the VWAP return for security
computed as the cross-sectional average of the individual
j for the period k, corresponding to the snapshot time t .
stock liquidity measures ( DLj,t  ). Similarly, the industry
Return square is used as a proxy of the volatility measure.
liquidity measure ( DLI,t  ) is computed using stocks from
The volume and order imbalance (OIB16) measures are com-
Industry I, following NSE industry classification. The study
puted using trade volumes for 1-, 5-, 10-, 15-, and 30-min
improves upon the construction of industry liquidity variable
intervals, using snapshot time as the starting point (as for the
(of CRS) by orthogonalizing it against the market liquidity
returns). The differences in buy and sell volumes are divided
variable as follows (Elton and Gruber 2011).
by the aggregate volumes to compute the normalized OIB
measure (Chordia et al. 2002, 2008).17 “Summary statistics” DLI,t = 𝛼I + 𝛽0 DLM,t + eI,t (4)
section provides the descriptive statistics of the measures
employed in the study. The orthogonalized measure of industry liquidity, eI,t  ,
reflects the variation in industry liquidity measure that is
unrelated to the market liquidity. This measure of industry
Methodology liquidity is relatively less vitiated by the inventory concerns
of market participants, and predominantly reflects the effects
Principal component and canonical correlation of asymmetric information. The basic model of CRS is pro-
analyses vided in Eq. (5). Next, following CRS, additional control
variables are introduced in Eq. (6). These include: the indi-
The study employs principal component analysis (PCA) and vidual stock returns ( rj,t ), volatility ( 𝜎j,t
2
 ), volume ( DVolj,t ),
canonical correlations to test the commonality in liquidity and market returns ( rm,t).
measures at levels.18 These techniques have been widely
used for their robustness in investigating liquidity common-
DLj,t = 𝛼j + 𝛽1 eI,t + 𝛽2 DLM,t + 𝜀j,t (5)
ality (Bauer 2004; Hansch 2003; Hasbrouck and Seppi 2001;
Kempf and Mayston 2008; Korajczyk and Sadka 2008). The DLj,t = 𝛼j� + 𝛽1� eI,t + 𝛽2� DLM,t + 𝛽3� rm,t + 𝛽4� rj,t
literature generally considers the first principal component + 𝛽5� 𝜎j,t
2
+ 𝛽6� DVolj,t + 𝜀j,t
(which explains the maximum variation) as a reflection of (6)
the systematic latent factor. This factor captures the effect The control variables are computed for hourly intervals
corresponding to the hourly snapshot timings (from t − 1 to
t). Here, α’s and β’s are regression coefficients, and ε’s are
16
  Standard tick-test algorithm is employed to classify the trades into error terms. rj,t is the return of stock j computed from the
buy or sell transactions, following Chakrabarty et al. 2015; Ellis et al.
2000; Odders-White 2000. 19
17
  For example, if a set of 5 and 6 covariates have to be compared,
  The results are provided for all the three normalized order imbal- simple pairwise correlation tests will yield 30 correlations and the
ance measures: based on the number of trades, volume, and dollar corresponding 30 t-statistics. In this case, however, a canonical cor-
volume. relation analysis will provide only one comprehensive correlation
18
 To test commonality: (1) PCA and canonical correlations are measure and the corresponding Chi-square statistic. For detailed dis-
applied on liquidity measures at levels, and (2) CRS regressions are cussion on PCA and canonical correlations, please refer Hasbrouck
employed on the first-differenced liquidity measures. and Seppi (2001) and Korajczyk and Sadka (2008).

360 A. Tripathi et al.

hourly price changes; rm,t is market return measure computed accordingly. The estimation of each of these four regression
from the cross-sectional average of returns across the sample models has been done separately; the value of regression
stocks; 𝜎j,t
2
is the volatility of stock j, constructed from return coefficients for each equation may be different.
squares; DVolj,t is the
( first-differenced
) volume measure of
stock j, DVolt = ln Vol  , where ­Volt is the volume for the
Volt
t−1
Results
15-min interval, following each snapshot. For the volume
measure, the procedure is repeated with the intervals of 1-,
Summary statistics
5-, 10-, and 30-min to confirm the robustness of the results.
For brevity, the results are reported only for the 15-min
Table 1 provides the descriptive statistics of the key-varia-
interval.20 Overall, the results remain robust for different
bles, namely, the relative spread measure, depth measure,
variable specifications. The study employs fixed-effect panel
hourly and other short-term return measures,23 volume, and
regression models.21 White’s heteroscedasticity and autocor-
order imbalance measures. The study employs order book
relation consistent standard errors are estimated for the cal-
measures—relative spread and depth—for multiple levels,
culation of t-statistics.
i.e., L1, L5, L10, L50, L100, and complete order book. Fur-
CRS (2000) note that non-synchronicity can occur due to
thermore, the short-term trading activity related measures
thin trading, which may lead to spurious correlations with
(Return, Volume, and OIB) are utilized for 1-, 5-, 10-, 15-,
lead and lag terms. To address this issue, additional robust-
and 30-min periods.
ness tests are performed by incorporating the lead and lag
terms of independent variables in Eq. (7).
Intraday patterns in liquidity, volatility, volume,
∑ 14
∑14
∑14 and order‑flow
DLj,t = 𝛼j�� + ��
𝛽1,t,T eI,T + ��
𝛽2,T DLM,T + ��
𝛽3,T rm,T
T=11 T=11 T=11
The literature documents an intraday U-shaped pattern for
14 14 14
∑
��
∑
�� 2
∑
�� various liquidity measures (Chuang et al. 2009; Foster and
+ 𝛽4,T rj,T + 𝛽5,T 𝜎j,T + 𝛽6,T ∗ DVolj,T + 𝜀j,t
T=11 T=11 T=11
Viswanathan 1990; Kempf and Mayston 2008; Kyle 1985;
(7) Madhavan and Smidt 1993; Saad and Darrat 2007). An
Here t denotes the snapshot times, i.e., 11 A.M., 12 P.M., examination of intraday behavior of liquidity measures pro-
1 P.M., and 2 P. M. T are the different snapshot times [i.e., vides the primary intuition behind the liquidity commonal-
11 A.M. (T = 11), 12 P.M. (T = 12), 1 P.M. (T = 13), and 2 P. ity. Five key variables, namely, the spread, depth, volatil-
M. (T = 14)] of the same day as that of t. This equation cor- ity, volume, and order-flow capture the intraday liquidity
responds to four separate models, each indicating a unique dynamics. These measures reflect the trading activity, risk,
hourly interval. For instance, the equation corresponding information, and competition (Schwartz 1988).
to 11 A.M. has 3 lead terms of the independent variables Figures 1a–e depict the intraday time-specific averages
(at 12 P.M., 1 P.M. and 2 P.M.) and no lag terms; and the (at 10 A.M, 11 A.M., 12 P.M., 1 P.M. and 2 P.M.) of vari-
equation corresponding to 2 P.M. has 3 lag terms of the inde- ous measures of trading activity based on the order book
pendent variables (at 11 A.M., 12 P.M., and 1 P.M.) and no and trade book. The order book measures (relative spread
lead terms.22 The remaining two equations are constructed and depth) are shown for multiple levels of order book
(i.e., L1, L5, L10, L50, L100, and complete). The trade
book measures: volatility, volume, and normalized order
imbalances (dollar volume) are provided for the periods of
20
1-, 5-, 10-, 15-, and 30-min. These figures lucidly portray
 The window of 15-min offers a reasonable trade-off between a
‘less noisy’ and ‘contemporaneous with the state of liquidity’ meas- the intraday patterns of liquidity. Illiquidity is high in the
ure of volume. morning, falls, and then again rises in the afternoon. The
21
 The fixed-effect panel regression methods are widely applied to intraday patterns of relative spread measure, volume, and
contain the vitiating effect of unobserved heterogeneity in estimation volatility are similar, and conform to the widely docu-
(Baltagi 2008; Hsiao 2003). For the selection of an appropriate panel mented U-shaped pattern. The pattern is similar across
regression model, we conduct the standard Hausman and poolabil-
ity tests. The results from the other models (viz., the random-effects, all the levels of limit-order book, and is in sync with the
between, Hausman–Taylor, first-differenced, and pooled), however,
are similar. The intuition for the fixed-effect model is as follows.
The observation set comprises diverse firms from various industries.
Therefore, we expect the cross-sectional variations (firm-specific time
invariant effects) to contribute significantly to the unobserved hetero- 23
  Short-term return measures for 1-, 5-, 10-, 15-, and 30-min peri-
geneity across observations. ods are employed for the PCA and canonical correlation analyses. For
22
  Hence, this model does not include the first snapshot of10 A.M. CRS regressions, hourly returns are used.
Liquidity commonality beyond best prices: Indian evidence 361

Table 1  Summary statistics of the sample stocks

Number of obser- Mean Median Minimum Maximum SD
vations

Panel A: Order book measures

Relative spread measure (%)
SPR_L1 617,114 0.22 0.15 0.00 1.26 0.20
SPR_L5 617,114 0.27 0.21 0.02 1.26 0.21
SPR_L10 617,114 0.33 0.27 0.04 1.27 0.22
SPR_L50 617,114 0.66 0.57 0.15 2.61 0.36
SPR_L100 617,114 1.01 0.85 0.25 4.69 0.62
SPR_LFull 617,114 9.44 8.62 4.51 21.48 3.36
Depth measure (based on the volume of order book)
Depth_L1 617,114 620 99 1 2,473,083 5081
Depth_L5 617,114 2840 775 5 2,501,838 13,221
Depth_L10 617,114 4272 1234 8 2,537,563 17,863
Depth_L50 617,114 25,217 8863 75 3,383,940 77,205
Depth_L100 617,114 46,217 16,863 177 5,646,927 134,107
Depth_LFull 617,114 412,182 160,557 814 35,403,800 1,045,569
Panel B: Trade book measures
Hourly VWAP returns (%)
Rt 146,618 − 0.02 0.69 − 3.21 3.77 − 0.01
Short-term VWAP returns (%)
R1min 251,834 0.00 0.00 − 0.49 0.53 0.11
R5min 251,834 0.00 0.00 − 0.82 0.90 0.20
R10min 251,834 − 0.01 − 0.01 − 1.03 1.10 0.26
R15min 251,834 − 0.01 − 0.01 − 1.13 1.19 0.30
R30min 251,834 − 0.01 − 0.01 − 1.57 1.78 0.42
Volume measure (number of stocks)
Vol_1min 251,834 14,254 4925 112 5,023,814 46,901
Vol_5min 251,834 58,725 24,051 210 15,526,286 147,022
Vol_10min 251,834 110,481 47,944 297 25,882,680 254,215
Vol_15min 251,834 160,711 71,799 409 35,074,764 352,226
Vol_30min 251,834 307,894 143,607 526 60,781,876 633,977
Normalized order imbalance (%)
OIB_1min 251,834  − 0.16 0.01  − 100.00 100.00 43.25
OIB_5min 251,834  − 0.57  − 0.48  − 99.96 99.93 27.87
OIB_10min 251,834  − 0.81  − 0.71  − 99.94 99.80 23.79
OIB_15min 251,834  − 0.78  − 0.73  − 99.93 99.78 21.89
OIB_30min 251,834  − 0.60  − 0.50  − 97.49 98.93 19.38

The table provides descriptive statistics of the sample stocks for the period of study (from January 1, 2015 to December 31, 2017). It sum-
marizes the liquidity and trading activity of the stocks: the relative spread measure (SPR L1, L5, L10, L50, L100, and LFull); depth measure;
hourly VWAP returns; short-term VWAP returns; volume; and order imbalance (dollar volume) measures. Here, L1, L5, L10, L50, L100, and
complete (LFull)) represent the levels of the order book. For example, L50 captures the price and volume information of the best fifty bid and
ask orders. Short-term VWAP returns, volume, and order imbalance measures are constructed over periods of 1, 5, 10, 15, and 30 min, starting
from the snapshot timings
(1) The study employs hourly VWAP returns for the CRS (2000) market model regressions. (2) The short-term VWAP returns are utilized for the
Principal Component Analysis and Canonical correlations. The number of observations for the trade book measure reduces significantly owing
to the additional filtration criteria (e.g., at least 1 transaction during the 5-s window, after the snapshot) applied in the study for CRS regressions

362 A. Tripathi et al.

Fig. 1  a Intraday movement of hourly average relative spread for vari- ity (1, 5, 10, 15 and 30 min), based on 197 securities for the 3-year period
ous levels of order book (L1, L5, L10, L50, L100, and complete (LFull)), (2015–2017). Volatility values are multiplied by a factor of 10^4 for proper
based on 197 securities for the 3-year period (2015–2017). b Intraday presentation. d Intraday movement of hourly average volume (1, 5, 10, 15
movement of hourly average depth for various levels of order book (L1, and 30 min), based on 197 securities for the 3-year period (2015–2017).
L5, L10, L50, L100, and complete (LFull)), based on 197 securities for the e Intraday movement of hourly average order imbalance (1, 5, 10, 15 and
3-year period (2015–2017). c Intraday movement of hourly average volatil- 30 min), based on 197 securities for the 3-year period (2015–2017)
Liquidity commonality beyond best prices: Indian evidence 363

During the closing periods, the intraday traders square-

off their short-positions.24 This creates moderately positive
order imbalances (buy pressure), and causes an increase in
trade volumes. These trade-demands are inelastic, and there-
fore, the (de-facto) market makers gain a certain level of
market power to charge higher spreads. These profits reflect
the monopolistic nature of the trading environment (Stoll
and Whaley 1990). This is also indicated by the increasing
spread and depth measures.

PCA analysis

The results from PCA analysis are presented (Table 2) for

various measures (at levels), namely, the depth, dollar depth,
relative spread, impact cost,25 volume, and OIB (number
of orders, volume, and dollar volume) among others. The
first principal component (PCA1) that explains the highest
common variation is reported. There is evidence to suggest
that liquidity across the stocks co-moves. Moreover, this
commonality is witnessed even at the deeper levels of LOB
beyond the best prices. The measures employed in the study
capture different dimensions of liquidity, and hence, exhibit
varying levels of commonality. For example, with the rela-
tive spread measure, a commonality of 30–45%26 is detected
Fig. 1  (continued) across multiple levels of the order book. With the impact
cost measure also, a consistently high level of commonality
various measures of volume and volatility. The intraday (21–44%) is observed; depth measures, however, show a siz-
patterns in the order-flows and depth are different. During able level of commonality only at the deeper levels of order
the opening times, the order imbalances are negative (and book, i.e., beyond L10 (18–39%). This is an indication of the
large in magnitude), indicating the existence of selling presence of informed traders at the deeper levels in the order
pressure. As the day progresses, this measure increases book, and that of the best bid/ask levels being noisy. Further-
and becomes moderately positive by the closing times, more, the PCA is also performed at different points of time
indicating the buildup of buying pressure. Similarly, the (e.g., 10 A.M., 11 A.M., 12 P.M., 1 P.M., 2 P.M.) separately.
depth is low during the opening, and increases gradually The results (not reported for brevity) indicate that the levels
over the day. of commonality are materially higher in the morning at 10
The intraday pattern is a reflection of competitive market A.M. (when liquidity is significantly low), as compared to
making, inventory considerations, and information asym- the other points of time during the day.
metry (Chan et al. 1995). First, it indicates that during the
opening time, there is a high degree of uncertainty about
prices and trade execution. At this time, a high degree of 24
  Short-sale restrictions require the naked short-positions to be com-
information asymmetry exists along with the accumulation pulsorily squared-off during the same day, and cannot be carried for-
of overnight information (Stoll 1989, 2000). This causes ward.
order imbalances (sell pressures), which, in turn, leads to 25
 NSE defines the impact cost as “percentage price movement
widening of spreads, lower depth, and high levels of volatil- caused by an order size of INR 0.1 million from the average of the
best bid and offer price in the order book snapshot. The impact cost
ity (Madhavan and Smidt 1993). The increase in trading vol-
is calculated for both, the buy and the sell side in each order book
umes indicates the presence of informed traders, who make snapshot”. The study computes the impact cost for the order sizes of
above average expected returns at the expense of (de-facto) INR 0.1 million, INR 0.2 million, INR 0.3 million, INR 0.5 million,
market makers. To contain their cost of adverse selection, INR 1.0 million, and complete order book. However, the trade vol-
umes above INR 0.3 million are not as frequent, and therefore, not
the market makers increase the spread and decrease their
very pertinent from the commonality perspective.
offers (leading to reduced depth). 26
  The first principal component explains nearly 40% of the total var-
iation in relative spreads for the cross-section of securities over the
study period. That broadly indicates the level of commonality in the
liquidity of securities.

364 A. Tripathi et al.

Table 2  Commonality in the sample stocks using Principal Component Analysis

Panel A: Level of commonality across various measures based on the limit-order book (%)
Order size
Levels of limit-order book
Variables L1 L5 L10 L50 L100 LFull

Depth 1.23 2.09 2.75 18.20 26.89 38.83

Dollar Depth 1.42 2.35 3.83 18.99 27.88 44.23
Relative Spread 44.96 46.02 45.87 44.56 43.49 30.58
Order size
INR 0.1 million INR 0.2 million INR 0.3 million INR 0.5 million INR 1.0 million Complete order book

Impact cost 44.15 39.69 32.65 25.17 21.23 30.69

Panel B: Level of commonality across various measures based on the trade book (%)
Time intervals from the snapshot time
Variables 1-min 5-min 10-min 15-min 30-min

Return 15.81 18.42 20.35 19.82 20.26

Volatility 16.83 9.26 10.83 9.76 10.65
Volume 8.48 12.38 13.95 15.61 19.50
OIB_Num 8.40 8.53 8.63 8.44 8.04
OIB_Qty 4.40 3.89 3.92 3.78 3.53
OIB_Doll 4.40 3.91 3.95 3.81 3.56

The table reports the highest common variation explained by the first principal component, for a comprehensive set of measures, namely, the
depth, dollar depth, relative spread, impact cost, returns, volatility, volume, and OIB (number of trades, volume, dollar volume). Panel A reports
the commonality based on the depth and spread measures, for various levels of order book (L1, L5, L10, L50, L100, and LFull), and impact cost
measures. Impact cost measures the average percentage change in prices, for the buy and sell orders of a given size (reported for INR 0.1 million
to INR 1.0 million, and the complete order book). Panel B presents the commonality in return, volatility, volume, and OIB measures for various
time intervals (1, 5, 10, 15, 30 min)

The levels of commonality in the returns (16–20%), vol- Following Hasbrouck and Seppi (2001), the canonical
ume (8–20%), volatility (9–17%), and order-flow measures correlation analysis is employed to find the correlation
(about 8%) are relatively lower than those in the relative between the measures of order-type and order-flow with the
spread, depth (at deeper levels), and impact cost measures. measures of liquidity and returns. The details of covariates
This suggests that the economic forces driving these com- employed in the canonical analysis are as follows: (a) Fol-
monalities are different. lowing Domowitz et al. (2005), the order-type variable27
The analysis so far offers considerable evidence for the denotes the ratio of market orders (proxied by executed
phenomenon of commonality in liquidity at NSE. Its nature order volume) to limit orders (proxied by depth volume).
and causal factors are further investigated with canonical This analysis measures the market order volumes for five
correlation analysis and market model approach of CRS in time-intervals, viz., 1-, 5-, 10-, 15-, and 30-min; and utilizes
the following discussion. six levels of limit-order book (L1, L5, L10, L50, L100, and
complete (LFull)) to construct the depth volume measures.
Canonical correlation analysis of return The order-type measure has five sets of covariates (cor-
commonality and liquidity commonality responding to each market order time interval), and each
covariate comprises of six variables (corresponding to each
The order-type (market orders and limit orders) co-move- level of order book); (b) The three order-flow covariates are
ment is related to liquidity co-movement, as the limit orders the number of orders, volume, and dollar volume variants of
provide liquidity that market orders consume. On the other the normalized OIB measure. Each variant of the normalized
hand, the order-flow (buy-sell order imbalance) co-move-
27
ment is related to return co-movement, as the buy orders   The study assumes that each trade involves two legs- a limit order
increase prices, and sell orders tend to lower them (e.g., and a market order that crosses against the limit order. Therefore, the
traded volume provides a good proxy of the market orders (Chordia
Domowitz et al. 2005; Hasbrouck and Seppi 2001). et al. 2008).
Liquidity commonality beyond best prices: Indian evidence 365

Table 3  Commonality in the sample stocks using Canonical correla- with returns. In contrast, the order-flow measures have a
tions strong and significant correlation with returns (30–60%), and
Coefficient (CV Chi-Sq an economically weak correlation with liquidity. The order-
in %) flow variables represent the variation in the demand-side,
Panel A: Canonical correlations with return variables
particularly the institutional investors, mutual funds, index
Order-type variables (Market order/Limit orders)
funds among others (Chordia et al. 2000; Karolyi et al. 2012).
Volume_1min/Depth_Qty 1.82*** 157
Overall, the return commonality appears to be related to the
Volume_5min/Depth_Qty 4.15*** 746
co-movement in demand-side factors (driven by the correla-
Volume_10min/Depth_Qty 4.42*** 852
tion in order-flows); and, the liquidity commonality relates
Volume_15min/Depth_Qty 4.41*** 847
to the co-movement in the supply and demand of liquidity
Volume_30min/Depth_Qty 6.03*** 1560
(caused by the ratio of limit orders to market orders).
Order-flow variables (Normalized OIB measure)
OIB_Num 64.07*** 468,000
Market model (CRS, 2000) approach to commonality
OIB_Qty 32.84*** 115,000
OIB_Doll 32.93*** 116,000
Following the CRS model, we investigate the co-move-
Panel B: Canonical correlations with liquidity variables (relative
ment of the liquidity of individual securities with that of
spread- L1, L5, L10, L50, L100, LFull) market and industry. CRS suggests that the variables be
Order-type variables (Market order/Limit orders) defined in the first-difference form (rather than at their
Volume_1min/Depth_Qty 18.95*** 28,100 levels), i.e., hourly changes in the liquidity measures. This
Volume_5min/Depth_Qty 30.28*** 72,000 approach, not only standardizes the values of the variables,
Volume_10min/Depth_Qty 32.96*** 85,100 but also obviates the unit root problem. We regress the
Volume_15min/Depth_Qty 30.95*** 77,700 liquidity measures of individual securities on those of the
Volume_30min/Depth_Qty 30.48*** 79,500 market and industry liquidity [Eq. (5)]. Following CRS,
Order-flow variables (Normalized OIB measure) the market model is augmented by introducing the control
OIB_Num 2.02*** 160 variables—market return, individual stock return, volatil-
OIB_Qty 0.02*** 82 ity, and volume [Eq. (6)]. The analysis estimates the mar-
OIB_Doll 1.32*** 80 ket model by applying the fixed-effect panel regression
methodology; these models are analyzed individually for
Panels A and B provide the canonical correlations for the order-type
the different levels of order book.
and order-flow variables with the return and liquidity variables. The
order-type variable signifies the ratio of market orders (proxied by Table 4 (Panels A and B) provides the regression coef-
executed order volume) to limit orders (proxied by depth volume). ficients and t-statistics for the market and industry liquidity,
This analysis measures the market order volumes for five time- adj.R2 , and F-values for the market model regressions. The
intervals, viz., 1-, 5-, 10-, 15-, and 30-min; and utilizes six levels of
adj.R2 from the regressions is in the range of 10–18%, for all
limit-order book (L1, L5, L10, L50, L100, and LFull) to construct
the depth volume measures. The order-type measure has five sets of the levels of order book; much higher than that reported by
covariates (corresponding to each market order time interval), and CRS (< 4%). Panel C shows the results of robustness tests
each covariate comprises of six variables (corresponding to each level (following the method of CRS), i.e., the regressions on indi-
of the order book). The three order-flow covariates are the number
vidual stocks by employing Eq. (6). The proportion of posi-
of orders, volume, and dollar volume variants of the normalized OIB
measure. Each variant of the normalized OIB is constructed for five tive market liquidity coefficients is considerably higher, about
time-intervals, viz., 1-, 5-, 10-, 15-, and 30-min. The return covari- 99%, than that reported by CRS (70–80%). The proportion
ate employs a set of five return variables (1-, 5-, 10-, 15-, and 30-min of ‘positive and significant’ market liquidity coefficients is
intervals). The liquidity covariate employs a set of six relative spread
also much higher, at about 95%, than that reported by CRS
variables (L1, L5, L10, L50, L100, and LFull)
(14–35%). Furthermore, the proportion of ‘positive and signifi-
(1) *, **, and *** denote statistical significance at 10%, 5%, and 1%
levels cant’ industry liquidity coefficients remains reasonably high (at
about 30–40%). These differences may be caused by the sam-
OIB is constructed for five time-intervals, viz., 1-, 5-, 10-, ple selection. CRS employed 1192 stocks from NYSE. Their
15-, and 30-min; (c) The liquidity covariate employs a set of sample-set included a sizable proportion of small securities
six relative spread variables (L1, L5, L10, L50, L100, and that may be influenced by noise. This study selects 197 stocks
complete (LFull)); (d) The return covariate employs a set of from about 1600 stocks listed on the NSE. The sample-set
five return variables (1-, 5-, 10-, 15-, and 30-min intervals). comprises the biggest and the most heavily traded securities on
Table 3 presents the results from the canonical correla- NSE. Hence, these securities are expected to be the most infor-
tion analysis. These results indicate that the order-type meas- mationally efficient as well. Moreover, the orthogonalization of
ures have a strong and significant correlation with liquidity market and industry liquidity segregates the effect of inventory
(18–30%), and a significant but economically weak correlation concerns and information asymmetry more explicitly.

366 A. Tripathi et al.

Table 4  Commonality with the industry and market liquidity using market model panel regressions
DLj,t = 𝛼j + 𝛽1 eI,t + 𝛽2 DLM,t + 𝜀j,t (5)
DLj,t = 𝛼j� + 𝛽1� eI,t + 𝛽2� DLM,t + 𝛽3� rm,t + 𝛽4� rj,t + 𝛽5� 𝜎j,t
2
+ 𝛽6� DVolj,t + 𝜀j,t (6)

Number of observations Market coefficient Industry coefficient adj.R2 (%) F-value

Panel A: Results without control variables (Eq. 5)

 L1 499,924 0.99*** 0.19*** 10.02 10,031.76***
(141.26) (23.91)
 L5 499,923 0.94*** 0.19*** 10.92 13,735.29***
(165.24) (25.51)
 L10 499,898 0.91*** 0.19*** 11.10 14,867.24***
(172.44) (24.15)
 L50 499,914 0.90*** 0.19*** 11.75 14,232.23***
(167.28) (25.70)
 L100 499,915 0.90*** 0.19*** 11.34 12,622.52***
(157.70) (26.78)
 LFull 499,919 0.83*** 0.29*** 9.68 4366.31***
(58.26) (20.49)
Panel B: Results with control variables (Eq. 6)
 L1 114,963 1.00*** 0.18*** 15.48 1061.20***
(72.35) (14.13)
 L5 114,230 0.98*** 0.20*** 16.53 1197.56***
(78.42) (16.30)
 L10 114,019 0.96*** 0.21*** 17.05 1436.42***
(85.57) (16.48)
 L50 113,812 0.97*** 0.22*** 18.46 1282.78***
(78.00) (17.30)
 L100 113,875 0.93*** 0.22*** 18.85 1261.76***
(72.12) (17.77)
 LFull 114,116 0.87*** 0.31*** 18.64 1814.17***
(40.20) (14.34)
Market coefficient Industry coefficient
Positive and significant (%) Positive and insignificant (%) Positive and significant (%) Positive and
insignificant
(%)

Panel C: Robustness tests (Eq. 6)

 L1 95.70 4.30 33.87 41.94
 L5 96.24 3.76 36.02 44.62
 L10 97.86 2.14 35.29 43.85
 L50 94.65 5.35 36.90 44.92
 L100 97.31 2.69 41.40 43.01

Panels A and B summarize the results of market model panel regressions [Eqs. (5) and (6)]. Results include the regression coefficient and their
t-statistics for the market and industry liquidity measures, adj.R2 , F-statistics, and the number of observations. The results are shown for the mul-
tiple levels of order book (L1, L5, L10, L50, L100, and LFull). For brevity, the table does not report the statistics pertaining to control variables.
Panel C provides results of the robustness test for Eq. (6). Following CRS (2000), the regression model of Eq. (6) is estimated separately for the
individual firms (197). We report the proportion of the t-statistics for the market and industry liquidity variables that are positive and significant,
and positive but not significant. Robustness tests on Eq. (5) yield qualitatively similar results (not shown here)
(1) *, **, and *** denote statistical significance at 10%, 5%, and 1% levels. (2) t-statistics are reported in parentheses ()

A comparison of t-statistics for the market and indus- and information asymmetry hypothesis permeates into
try liquidity coefficients provides interesting insights into LOB beyond the best prices. The evidence of commonal-
commonality. Both the coefficients are strongly significant ity of individual stock liquidity with the industry liquidity
for all the levels, suggesting that the effect of inventory is stronger than that in the quote-driven markets (CRS,
Liquidity commonality beyond best prices: Indian evidence 367

Table 5  Commonality with the industry and market liquidity using market model regressions on size-based portfolios
Market coefficient t-statistic Industry coefficient t-statistic Adj.R2 (%)
Large Medium Small Large Medium Small Large (%) Medium (%) Small (%)

Panel A: Portfolio regression results without control variables

 L1 87.15*** 72.97*** 46.32*** 7.14*** 5.58*** 1.60 73.74 64.39 40.72
 L5 95.30*** 71.34*** 47.21*** 8.04*** 8.06*** 4.08*** 76.15 65.35 44.96
 L10 96.81*** 64.40*** 45.74*** 8.64*** 6.75*** 5.15*** 77.17 65.79 45.42
 L50 92.99*** 68.47*** 43.13*** 9.58*** 6.88*** 5.04*** 79.28 67.48 43.54
 L100 88.34*** 62.30*** 44.47*** 10.40*** 7.62*** 5.11*** 77.25 63.04 42.97
 LFull 70.85*** 53.40*** 30.27*** 31.54*** 17.29*** 10.32*** 74.34 63.95 32.07
Panel B: Portfolio regression results with control variables
 L1 80.21*** 68.30*** 44.08*** 6.64*** 5.32*** 1.34 74.29 65.33 42.69
 L5 88.14*** 66.08*** 45.20*** 7.53*** 7.31*** 3.70*** 76.48 66.41 46.28
 L10 89.77*** 58.57*** 42.91*** 8.15*** 6.08*** 4.83*** 77.44 66.81 46.81
 L50 85.39*** 61.82*** 39.61*** 9.21*** 6.44*** 4.63*** 79.43 68.23 45.89
 L100 77.66*** 54.01*** 39.53*** 10.16*** 7.03*** 4.10*** 77.42 64.26 45.57
 LFull 65.35*** 47.70*** 24.91*** 31.82*** 16.18*** 9.15*** 76.21 66.32 35.64

The table reports results of market model time-series regressions [based on Eqs. (5) and (6)] on the size-based portfolios. The study constructs
three size-based equally weighted portfolios of sample stocks: Large, Medium, and Small. The daily average market capitalization for the past
three years (2015–17) is employed to rank the firms. Results show the t-statistics of the market and industry liquidity variables, and adj.R2 . For
brevity, the details of control variables are not shown
(1) *, **, and *** denote statistical significance at 10%, 5% and 1% levels

2000). The NSE microstructure is devoid of designated stocks. Table 5 presents the results from the portfolio regres-
market makers, and therefore, the role of information sions. As expected, the adj.R2 improves materially. 30 It
asymmetry attains a larger significance. This informa- ranges from nearly 75% (for large stock portfolios) to 46%
tion asymmetry exists despite the fact that our sample- (for small stock portfolios). Also, the effect of systematic
set includes the most informationally efficient securities factors (inventory concerns and information asymmetry)
traded on the NSE (because of selection criteria). Moving is lower for the smaller stocks. This is evident from the
deeper into the order book beyond the best prices (L5, corresponding t-statistics of market and industry liquidity
L10, L50, L10028), the adj.R2 measure increases, and the coefficients. This is intuitive, as we expect the large stocks
t-statistics for the market and industry coefficients remain to be less affected by the idiosyncratic factors. Such stocks
strongly significant. This suggests that the deeper levels are widely covered by the analysts and investors, and have
beyond the best prices are also informative, and capture a thick order book. In contrast, the order books of small
the effect of systematic factors (i.e., inventory concerns stocks are relatively less thick, and in-turn, the prices may
and information asymmetry). be noisy. Interestingly, in the portfolio regressions, the sig-
To investigate the effect of size of a firm on the com- nificance of industry liquidity is much lower as compared
monality, three size-based (large, medium, and small29) to that in the regression using individual stocks. Literature
portfolios are constructed from the sample stocks. The daily has widely reported that the effect of information asymmetry
average market capitalization for the three years’ period weakens in portfolios (Chordia et al. 2000, 2002, 2008). In a
(2015–2017) is used to rank the stocks on size. Market well-diversified portfolio of stocks, the information asymme-
model regressions [Eqs. (5) and (6)] are implemented for try effect tends to cancel out, while the marketwide effects
these portfolios in a manner similar to that for the individual sustain.
28
  It is presumed that the limit orders beyond 100 may also include To examine the robustness of these results, we also imple-
non-serious orders and may not carry much information. This is evi- ment the market model regressions [Eqs. (5) and (6)] using
dent from the drop in adj.R2 measure and the t-statistics pertaining to the impact cost measure (order sizes INR 0.10 Mn, INR
the industry and market liquidity coefficients. 0.15 Mn, and INR 0.20 Mn). The results remain qualitatively
29
  The portfolio measures of liquidity, returns, volume, etc., are com-
similar to those obtained with RS measure, and are reported
puted as the equally-weighted averages of these measures for the indi-
vidual stocks in that portfolio. in “Appendix A” (Table 7). A final set of robustness tests are
30
 The effect of noise is significantly reduced in portfolios, which also conducted with lead and lag terms, employing Eq. (7).
causes the evident increase in the adj.R2 measure (Chordia et al. 2000,
2008).

368 A. Tripathi et al.

of hourly movement in the aggregate market liquidity for

multiple levels of order book. Here, the cross-sectional aver-
ages of individual stock liquidity measure (first-differenced
relative spreads) represent the aggregate market liquidity.
Visual inspection of Fig. 2 suggests clustering of liquid-
ity–volatility. The observations with high liquidity-volatility
occur together, and those with low liquidity-volatility occur
together. This shows clustering in aggregate market liquid-
ity. The transition from the low to high aggregate market
liquidity-volatility forms a regime-switching pattern, which
supports the FE hypothesis.
This hypothesis suggests that the free-entry and free-exit
cause significant liquidity provisions and withdrawals (large
up and down movements) to occur together. Therefore, the
study argues that high liquidity-volatility should also exhibit
a high level of commonality. To test this hypothesis, the
market model [Eqs. (5) and (6)] is implemented across the
deciles of liquidity-volatility. Table 6 presents the regression
results from the sample-set (197 stocks). The decile 0.1 (1.0)
represents the observation sets corresponding to the lowest
(highest) liquidity-volatility regimes. The t-statistics of the
market and industry liquidity coefficients are statistically
Fig. 2  Time-series of hourly movement in the aggregate liquidity significant for most of the liquidity-volatility regimes. As
(using CRS, 2000) measure: first-difference of the relative spread) for one moves from low to high liquidity-volatility deciles, the
various levels of order book (L1, L5, L10, L50, L100, and complete t-statistics of the market and industry liquidity coefficients,
(LFull)), based on 197 securities over the 3-year period (2015–2017) and the adj.R2 increase almost monotonically. This is true for
all the levels of the order book. Overall, there is evidence
that the large up and down movements in liquidity exhibit a
The results are qualitatively similar, but are not reported for very strong commonality. This also corroborates our earlier
brevity. finding that the systematic factors (inventory concerns and
information effects) affect the LOB much deeper beyond
FE hypothesis of LOB markets the best prices.
In the Indian market, the de-facto market makers include
Without designated market makers, the order-driven mar- large institutional investors, mutual funds, hedge funds,
kets lack the balancing forces (Brockman and Chung 2002). pension funds, and high net-worth individuals, among oth-
Thus, they are susceptible to marketwide changes in liquid- ers. They hold multiple stocks in their portfolios, and invest
ity movements (liquidity-volatility). Unlike the oligopolistic considerably in gathering the pertinent information. These
quote-driven markets, the barriers to entry and exit in LOB market participants significantly contribute to the thickness
markets are very low. The limit order suppliers can freely of the order book in the form of limit orders, and account
withdraw liquidity (exit the market) or provide liquidity for the informativeness of the order book at deeper lev-
(enter the market) to maximize their utility. els beyond the best prices. These de-facto market makers
The study first examines the patterns of aggregate mar- dynamically respond to the marketwide factors (e.g., inter-
ket liquidity-volatility. Figure 2 illustrates the time-series est rate changes) and adjust their inventory levels (Amihud
Liquidity commonality beyond best prices: Indian evidence 369

Table 6  Commonality with the industry and market liquidity using market model panel regressions for liquidity-volatility deciles
DLj,t = 𝛼j + 𝛽1 eI,t + 𝛽2 DLM,t + 𝜀j,t (5)
DLj,t = 𝛼j� + 𝛽1� eI,t + 𝛽2� DLM,t + 𝛽3� rm,t + 𝛽4� rj,t + 𝛽5� 𝜎j,t
2
+ 𝛽6� DVolj,t + 𝜀j,t (6)

Regression quantiles

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Panel A: Results without control variables (Eq. 5)

t-statistic corresponding to Market liquidity variable
L1 3.8*** 10.0*** 21.6*** 34.0*** 52.6*** 42.3*** 59.4*** 70.8*** 67.3*** 66.7***
L5 10.2*** 13.7*** 22.5*** 31.2*** 46.3*** 60.3*** 63.2*** 71.9*** 69.3*** 66.2***
L10 6.0*** 15.7*** 22.5*** 36.4*** 41.8*** 57.7*** 66.6*** 68.6*** 67.7*** 66.0***
L50 3.8*** 12.9*** 20.7*** 30.6*** 38.4*** 41.6*** 54.6*** 54.3*** 63.4*** 63.5***
L100 8.1*** 14.7*** 23.2*** 31.0*** 38.2*** 42.7*** 48.6*** 52.1*** 55.8*** 59.9***
LFull 5.9*** 15.3*** 20.5*** 23.5*** 25.9*** 29.2*** 34.8*** 42.5*** 41.6*** 51.6***
t-statistic corresponding to Industry liquidity variable
L1  − 0.1 3.0*** 3.9*** 7.7*** 9.7*** 12.6*** 12.4*** 14.6*** 14.7*** 16.2***
L5 0.8 3.5*** 5.6*** 7.2*** 10.0*** 9.8*** 11.4*** 12.6*** 18.4*** 20.4***
L10 0.1 2.4** 5.5*** 7.4*** 9.1*** 10*** 12.5*** 13.5*** 15.6*** 23.0***
L50 0.6 3.8*** 5.9*** 7.5*** 9.0*** 12.5*** 14.9*** 16.3*** 15.9*** 21.3***
L100 1.5 2.0** 5.7*** 7.7*** 9.7*** 10.5*** 13.0*** 14.4*** 20.2*** 19.0***
LFull 1.8* 5.9*** 7.2*** 10.5*** 11.2*** 12.1*** 15.3*** 14.6*** 17.4*** 18.5***
Regression adj.R2 (%)
L1 – – 0.85 2.82 5.84 6.03 10.06 14.71 17.54 20.20
L5 – – 0.87 2.35 4.41 6.63 9.40 12.98 18.04 23.62
L10 – – 0.82 2.23 3.90 5.98 9.43 12.11 17.91 25.20
L50 – – 0.48 1.73 3.33 5.58 8.12 12.32 17.57 27.37
L100 – 0.07 0.85 2.01 3.75 5.39 8.44 12.46 17.34 25.92
LFull – 0.25 1.14 2.60 3.99 6.10 9.09 12.34 16.61 20.70
Panel B: Results with control variables (Eq. 6)
t-statistic corresponding to Market liquidity variable
L1 1.4 5.3*** 18.2*** 18.9*** 32.2*** 35.3*** 36.3*** 40.7*** 53.5*** 42.3***
L5 3.7*** 6.5*** 12.7*** 22.2*** 24.4*** 30.8*** 42.9*** 50.9*** 45.2*** 47.0***
L10 2.7*** 8.1*** 15*** 20.2*** 26.0*** 31.4*** 39.6*** 45.0*** 44.5*** 42.3***
L50 2.1** 5.0*** 11.2*** 22.8*** 20.1*** 28.3*** 31.1*** 37.0*** 37.5*** 40.3***
L100 3.9*** 6.0*** 12.8*** 18.5*** 21.9*** 24.3*** 32.9*** 39.3*** 41.0*** 39.1***
LFull 5.0*** 9.3*** 12.2*** 16.5*** 18.4*** 23.5*** 25.6*** 30.6*** 34.7*** 32.1***
t-statistic corresponding to Industry liquidity variable
L1 0.5 2.7*** 2.0** 2.8*** 4.3*** 8.6*** 5.5*** 6.0*** 8.2*** 7.0***
L5 0.5 1.50 1.9* 3.0*** 5.8*** 3.9*** 5.4*** 8.7*** 8.6*** 8.1***
L10 1.3 1.50 2.0** 4.3*** 4.3*** 7.3*** 5.0*** 8.6*** 9.2*** 8.1***
L50 1.2 2.8*** 2.8*** 4.4*** 6.2*** 7.1*** 7.2*** 8.6*** 9.2*** 8.2***
L100 0.6 1.50 4.2*** 3.4*** 3.5*** 6.0*** 9.2*** 7.6*** 9.9*** 10.5***
LFull 1.2 3.6*** 6.3*** 5.9*** 6.0*** 9.0*** 8.3*** 8.6*** 10.3*** 11.0***
Regression adj.R2 (%)
L1 1.30 – 1.06 4.56 9.58 19.04 21.38 23.14 26.08 28.59
L5 – – 0.17 3.77 6.70 9.62 13.62 19.25 27.34 33.74
L10 – – 0.46 2.34 5.42 9.11 13.50 18.83 26.34 37.25
L50 – – – 2.49 4.11 8.08 11.87 19.02 26.39 45.00
L100 – – 0.52 2.17 5.68 8.45 12.67 18.83 29.47 46.03
LFull – – 1.18 4.10 7.47 11.28 16.37 22.57 29.62 35.89
The analysis buckets the observations into deciles according to the levels of liquidity-volatility. The decile 0.1 (1.0) represents the observation
sets corresponding to the lowest (highest) liquidity-volatility bucket. Panels A and B present the results of market models [Eqs. (5) and (6)] on
these liquidity-volatility deciles. Results show the t-statistics of the market and industry liquidity coefficients and adj.R2 from the regressions
(1) *, **, and *** denote statistical significance at 10%, 5% and 1% levels. (2) The negative values of adj.R2 are not reported. The corresponding
R2 values, however, remain positive (generally low), with significant F-statistics primarily due to a large sample size. For all such cases, the val-
ues of R2, adj.R2 , and F-statistics are available upon request

370 A. Tripathi et al.

and Mendelson 1980; Hasbrouck 1988; O’Hara and Oldfield provides evidence that the return commonality is associ-
1986). Furthermore, they also gauge the information content ated with the order-flows, and the liquidity commonality is
of order-flows to examine the arrival of private information, associated with the order types.
and accordingly adjust their quotes (Cao et al. 2009; Cope- Following CRS (2000), both the inventory and informa-
land and Galai 1983; Easley and O’hara 1987). Both of these tion asymmetry hypotheses are found to contribute signifi-
actions drive liquidity commonality through the channels, cantly to the liquidity co-movement across the most liquid
namely, inventory considerations and information asymme- and informationally efficient stocks on the NSE. In particu-
try. These de-facto market makers, with their resources and lar, the impact of information asymmetry on this liquidity
significant investments in information acquisition, appear commonality appears to be more significant than that for
to be relatively more informed than the designated market the quote-driven markets. Moreover, there is significant
makers in quote-driven markets. This is evident from the evidence of commonality even at the deeper levels of LOB
higher importance of information asymmetry paradigm in beyond the best prices. Indeed, these deeper levels of LOB
the NSE than that in quote-driven markets. are more informative and capture the effect of systematic
factors (inventory concerns and information asymmetry).
The adj.R2 values from the market model regressions are
10–18% for individual stocks, and increase to 46–75% for
Summary size-based portfolios. This suggests that the level of liquidity
commonality is higher for NSE than those reported for the
The study provides robust empirical evidence to support the quote-driven markets. The study also provides evidence to
existence of liquidity commonality in the context of India, support the FE hypothesis of liquidity commonality in the
an order-driven emerging market. The intraday patterns in order-driven markets. That is, the large moves in liquidity,
liquidity are found to be U-shaped, and the commonality both up and down, occur together. It also contributes to the
levels are the highest in the morning, when the liquidity growing literature on liquidity-volatility, and supports the
levels are the lowest. The study employs the PCA, canoni- proposition that, in the order-driven markets, liquidity may
cal correlations, and CRS market model methodologies to not be present when it is most desired. The results presented
investigate liquidity commonality at the liquidity level and in this study are important to the global portfolio investors
its first-difference. The prior studies in the order-driven interested in the emerging markets, as the rising transac-
emerging markets are predominantly limited to best bid-ask tion costs due to commonality may eradicate the benefits of
prices. The study contributes to the thin body of literature diversification.
that examines the issue of liquidity commonality at the
deeper levels of LOB beyond the best prices.
The PCA, on a number of liquidity proxies, shows that
the systematic factors explain about 30% variation in the Appendix A
individual stock liquidity, in a cross-section of 197 stocks.
The commonality based on other microstructure variables See Table 7.
(returns, volatility, and order imbalances), however, is
relatively lower. Also, the canonical correlation analysis
Liquidity commonality beyond best prices: Indian evidence 371

Table 7  Commonality with the industry and market liquidity using market model panel regressions for impact cost

Order size Number of observa- Market coefficient Industry coefficient Adj.R2 (%) F-value
tions

Panel A: Results without control variables (Eq. 5)

 INR 0.10 Mn 459,920 0.990*** 0.099*** 10.63 10,551.951***
(143.80) (13.64)
 INR 0.15 Mn 403,005 0.981*** 0.098*** 10.62 11,055.131***
(141.72) (13.68)
 INR 0.20 Mn 338,870 0.975*** 0.097*** 10.82 11,417.073***
(141.87) (14.65)
Panel B: Results with control variables (Eq. 6)
 INR 0.10 Mn 110,431 1.031*** 0.107*** 15.66 1,576.426***
(95.20) (8.62)
 INR 0.15 Mn 100,514 1.024*** 0.116*** 15.18 1,097.255***
(80.03) (8.56)
 INR 0.20 Mn 88,154 1.021*** 0.119*** 14.65 1,054.149***
(78.28) (9.76)
Order size Market coefficient Industry coefficient
Positive and significant (%) Positive and insignificant (%) Positive and significant (%) Positive and
insignificant
(%)

Panel C: Robustness tests (Eq. 6)

 INR 0.10 Mn 96.61 3.39 24.29 44.07
 INR 0.15 Mn 97.13 2.87 23.56 48.28
 INR 0.20 Mn 92.86 6.55 22.62 48.81

Panels A and B summarize the results of market model panel regressions [Eqs. (5) and (6)], employing the impact cost measure. Results include
the regression coefficient and their t-statistics for the market and industry liquidity measures, adj.R2,F-statistics, and number of observations. The
results are shown for the impact cost corresponding to INR 0.10 Mn, INR 0.15 Mn, and INR 0.20 Mn. For brevity, the table does not report the
statistics pertaining to control variables. Panel C provides results of the robustness test for Eq. (6). Following CRS (2000), the regression model
of Eq. (6) is estimated separately for the individual firms. We report the proportion of the t-statistics for the market and industry liquidity vari-
ables that are positive and significant, and positive but not significant. Robustness tests on Eq. (5) yield qualitatively similar results (not shown
here)
(1) *, **, and *** denote statistical significance at 10%, 5% and 1% levels. (2) t-statistics are reported in parentheses (). (3) In the context of the
Indian market, NSE provides the definition of its impact cost measure as follows: “The impact cost is the percentage price movement caused by
an order size of INR 0.1 million from the average of the best bid and offer price in the order book snapshot. The impact cost is calculated for
both, the buy and the sell side in each order book snapshot”. We compute this measure for INR 0.10 million, INR 0.15 million, and INR 0.20
million

References Trading: Evidence from the Athens Stock Exchange. Applied

Economics 48(22): 2007–2021. https​://doi.org/10.1080/00036​
846.2015.11119​91.
Alhomaidi, Asem, M. Kabir Hassan, Duygu Zirek, and Abdulrah-
Baltagi, Badi. 2008. Econometric Analysis of Panel Data. Hoboken:
man Alhassan. 2018. Does an Islamic Label Cause Stock Price
Wiley.
Comovements and Commonality in Liquidity? Applied Eco-
Bauer, Wolfgang. 2004. “Commonality in Liquidity in Pure Order-
nomics 50(59): 6444–6457. https ​ : //doi.org/10.1080/00036​
Driven Markets.” In National Centre of Competence in Research
846.2018.14860​23.
Working Paper.
Amihud, Yakov, and Haim Mendelson. 1980. Dealership Market:
Brockman, Paul, and Dennis Y. Chung. 2002. Commonality in Liquid-
Market-Making with Inventory. Journal of Financial Economics
ity: Evidence from an Order-Driven Market Structure. Journal of
8(1): 31–53.
Financial Research 25(4): 521–539.
Amihud, Yakov, and Haim Mendelson. 1986. Asset Pricing and the
Cao, Charles, Oliver Hansch, and Xiaoxin Wang. 2009. The Informa-
Bid-Ask Spread. Journal of Financial Economics 17(2): 223–249.
tion Content of an Open Limit-Order Book. Journal of Futures
Anagnostidis, Panagiotis, and Patrice Fontaine. 2019. Liquidity Com-
Markets: Futures, Options, and Other Derivative Products 29(1):
monality and High Frequency Trading: Evidence from the French
16–41.
Stock Market. International Review of Financial Analysis 69:
Chakrabarty, Bidisha, Roberto Pascual, and Andriy Shkilko. 2015.
101428. https​://doi.org/10.1016/j.irfa.2019.10142​8.
Evaluating Trade Classification Algorithms: Bulk Volume Clas-
Anagnostidis, Panagiotis, George Papachristou, and Nikos S.
sification versus the Tick Rule and the Lee-Ready Algorithm.
Thomaidis. 2016. Liquidity Commonality in Order-Driven

372 A. Tripathi et al.

Journal of Financial Markets 25(September): 52–79. https​://doi. Karolyi, G. Andrew, Kuan-Hui Lee, and Mathijs A. Van Dijk. 2012.
org/10.1016/j.finma​r.2015.06.001. Understanding Commonality in Liquidity around the World. Jour-
Chan, Kalok, Y. Peter Chung, and Herb Johnson. 1995. The Intra- nal of Financial Economics 105(1): 82–112.
day Behavior of Bid-Ask Spreads for NYSE Stocks and CBOE Kempf, Alexander, and Daniel Mayston. 2008. Liquidity Commonality
Options. Journal of Financial and Quantitative Analysis 30(3): beyond Best Prices. Journal of Financial Research 31(1): 25–40.
329–346. Koch, Andrew, Stefan Ruenzi, and Laura Starks. 2016. Commonality in
Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam. 2000. Liquidity: A Demand-Side Explanation. The Review of Financial
Commonality in Liquidity. Journal of Financial Economics 56(1): Studies 29(8): 1943–1974.
3–28. Korajczyk, Robert A., and Ronnie Sadka. 2008. Pricing the Commonal-
Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam. 2002. ity across Alternative Measures of Liquidity. Journal of Financial
Order Imbalance, Liquidity, and Market Returns. Journal of Economics 87(1): 45–72.
Financial Economics 65(1): 111–130. Kumar, Suraj, and Krishna Prasanna. 2018. Liquidity in Asian Mar-
Chordia, Tarun, Richard Roll, and Avanidhar Subrahmanyam. 2008. kets: Intensity of Regional and Global Linkages. Applied Eco-
Liquidity and Market Efficiency. Journal of Financial Econom- nomics 50(55): 6010–6023. https ​ : //doi.org/10.1080/00036​
ics 87(2): 249–268. 846.2018.14891​12.
Chuang, Chia-Chang, Chung-Ming Kuan, and Hsin-Yi Lin. 2009. Kyle, Albert S. 1985. Continuous Auctions and Insider Trading. Econo-
Causality in Quantiles and Dynamic Stock Return-Volume Rela- metrica: Journal of the Econometric Society 53: 1315–1335.
tions. Journal of Banking & Finance 33(7): 1351–1360. https​:// Madhavan, Ananth, and Seymour Smidt. 1993. An Analysis of Changes
doi.org/10.1016/j.jbank​fin.2009.02.013. in Specialist Inventories and Quotations. The Journal of Finance
Comerton-Forde, Carole, Terrence Hendershott, Charles M. Jones, 48(5): 1595–1628.
Pamela C. Moulton, and Mark S. Seasholes. 2010. Time Vari- Odders-White, Elizabeth R. 2000. On the Occurrence and Conse-
ation in Liquidity: The Role of Market-Maker Inventories and quences of Inaccurate Trade Classification. Journal of Financial
Revenues. The Journal of Finance 65(1): 295–331. Markets 3(3): 259–286.
Copeland, Thomas E., and Dan Galai. 1983. Information Effects on O’Hara, Maureen, and George S. Oldfield. 1986. The Microeconomics
the Bid-Ask Spread. The Journal of Finance 38(5): 1457–1469. of Market Making. Journal of Financial and Quantitative Analysis
Coughenour, Jay F., and Mohsen M. Saad. 2004. Common Market 21(4): 361–376.
Makers and Commonality in Liquidity. Journal of Financial Petkova, Ralitsa, Ferhat Akbas, and Will Armstrong. 2011. The Volatil-
Economics 73(1): 37–69. ity of Liquidity and Expected Stock Returns.
Domowitz, Ian, Oliver Hansch, and Xiaoxin Wang. 2005. Liquidity Saad, Mohsen M., and Ali F. Darrat. 2007. Intraday Pattern in Liquidity
commonality and return co-movement. Journal of Financial Covariation: Evidence from NYSE Listed Firms. Applied Finan-
Markets 8(4): 351–376. cial Economics 17(11): 913–919. https​://doi.org/10.1080/09603​
Easley, David, and Maureen O’hara. 1987. Price, Trade Size, and 10060​06755​32.
Information in Securities Markets. Journal of Financial Eco- Schwartz, Robert Alan. 1988. Equity Markets: Structure, Trading, and
nomics 19(1): 69–90. Performance. New York: Harpercollins College Div.
Ellis, Katrina, Roni Michaely, and Maureen O’Hara. 2000. The Sensoy, Ahmet. 2016. Commonality in Liquidity: Effects of Monetary
Accuracy of Trade Classification Rules: Evidence from Nasdaq. Policy and Macroeconomic Announcements. Finance Research
Journal of Financial and Quantitative Analysis 35(4): 529–551. Letters 16(February): 125–131. https ​ : //doi.org/10.1016/j.
Elton, Edwin J., and Martin J. Gruber. 2011. “Marginal Stockholder frl.2015.10.021.
Tax Rates and the Clientele Effect.” In Investments And Portfo- Sensoy, Ahmet. 2019. Commonality in Ask-Side vs. Bid-Side Liquid-
lio Performance, 3–9. World Scientific. ity. Finance Research Letters 28(March): 198–207. https​://doi.
Fabre, Joel, and Alex Frino. 2004. Commonality in Liquidity: Evi- org/10.1016/j.frl.2018.04.020.
dence from the Australian Stock Exchange. Accounting & Stoll, Hans R. 2000. Presidential Address: Friction. The Journal of
Finance 44(3): 357–368. Finance 55(4): 1479–1514.
Foster, F.Douglas, and Sean Viswanathan. 1990. A Theory of the Stoll, Hans R., and Robert E. Whaley. 1990. Stock Market Structure
Interday Variations in Volume, Variance, and Trading Costs and Volatility. The Review of Financial Studies 3(1): 37–71.
in Securities Markets. The Review of Financial Studies 3(4): Syamala, Sudhakar Reddy, V. Nagi Reddy, and Abhinav Goyal. 2014.
593–624. Commonality in Liquidity: An Empirical Examination of Emerg-
Hameed, Allaudeen, Wenjin Kang, and Shivesh Viswanathan. 2010. ing Order-Driven Equity and Derivatives Market. Journal of Inter-
Stock Market Declines and Liquidity. The Journal of Finance national Financial Markets, Institutions and Money 33: 317–334.
65(1): 257–293. Tripathi, Abhinava, Alok Dixit, and Vipul. 2019. Liquidity of Financial
Hansch, Oliver. 2003. “Island Tides: Exploring ECN Liquidity.” Markets: A Review. Studies in Economics and Finance. https​://
Hasbrouck, Joel. 1988. Trades, Quotes, Inventories, and Information. doi.org/10.1108/SEF-10-2018-0319.
Journal of Financial Economics 22(2): 229–252.
Hasbrouck, Joel, and Duane J. Seppi. 2001. Common Factors in Prices, Publisher’s Note Springer Nature remains neutral with regard to
Order Flows, and Liquidity. Journal of Financial Economics jurisdictional claims in published maps and institutional affiliations.
59(3): 383–411.
Hsiao, Cheng. 2003. Analysis of Panel Data (Vol. 34). Econometric
Society Monographs.
Huberman, Gur, and Dominika Halka. 2001. Systematic Liquidity. Abhinava Tripathi  is a Ph.D. student at Indian Institute of Manage-
Journal of Financial Research 24(2): 161–178. ment, Lucknow. He has done his B-Tech. from Indian Institute of
Kamara, Avraham, Xiaoxia Lou, and Ronnie Sadka. 2008. The Diver- Technology, Roorkee and MBA from Indian Institute of Management,
gence of Liquidity Commonality in the Cross-Section of Stocks. Kozhikode. He has more than 5 years of industry experience into cor-
Journal of Financial Economics 89(3): 444–466. porate banking, credit rating, and project finance advisory firms. His
Kang, Wenjin, and Huiping Zhang. 2013. Limit Order Book and Com- current research focusses on the subject of market-microstructure and
monality in Liquidity. Financial Review 48(1): 97–122. liquidity in financial markets.
Liquidity commonality beyond best prices: Indian evidence 373

Vipul  is a Faculty of Finance and Accounting at Indian Institute of portfolio management, volatility forecasting and the return generation
Management, Lucknow. He has taught Investment Management, Finan- process in the Indian Stock Markets.
cial Derivatives, Strategic Financial Management, Financial Manage-
ment, Financial Accounting and Management Accounting to Post Alok Dixit  is a Faculty of Finance and Accounting at the Indian
Graduate Management students over the last 36 years. Prof. Vipul has Institute of Management Lucknow, India. He has completed his
industry experience of 6 years in various middle management posi- Ph.D. degree from the Indian Institute of Technology Delhi, India.
tions and has been a member of Accounting Standards Board of the He teaches postgraduate-level courses like Financial Derivatives and
Institute of Chartered Accountants of India. Prof. Vipul has published Risk Management, Quantitative Applications in Finance, and Manage-
research papers in international refereed journals including Journal ment Accounting. Prof. Alok Dixit has published research papers in
of Futures Markets, Applied Economics, Applied Financial Econom- international refereed journals including Journal of Futures Markets,
ics, Quarterly Review of Economics and Finance, Research in Inter- Studies in Economics and Finance, Finance Research Letters, and Jour-
national Business and Finance, Journal of Economic Studies, Studies nal of Quantitative Economics. His work primarily relates to financial
in Economics and Finance, and articles on issues of current interest in derivatives, portfolio management, volatility forecasting, and liquidity
professional journals including the Chartered Accountant and Manage- in financial markets.
ment Accountant. His work primarily relates to financial derivatives,