See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259390316

Uncertainty, market structure, and liquidity ☆

Article  in  Journal of Financial Economics · December 2013

DOI: 10.1016/j.jfineco.2014.05.008

CITATIONS READS
61 1,033

2 authors, including:

Kee H. Chung
University at Buffalo, The State University of New York
163 PUBLICATIONS   6,849 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Regulation NMS and market quality View project

All content following this page was uploaded by Kee H. Chung on 19 October 2017.

The user has requested enhancement of the downloaded file.

 Uncertainty, market structure, and liquidity☆

Kee H. Chung a,b,*, Chairat Chuwonganant c

a
School of Management, State University of New York (SUNY) at Buffalo, United States
b
School of Business Administration, Chung-Ang University, Republic of Korea
c
College of Business Administration, Kansas State University, United States
______________________________________________________________________________

ABSTRACT

In this study we show that market uncertainty (measured by VIX) exerts a large market-wide
impact on liquidity which gives rise to co-movements in individual asset liquidity. The effect of
VIX on stock liquidity is greater than the combined effects of all other common determinants of
stock liquidity. We show that the uncertainty elasticity of liquidity (UEL: % change in liquidity
given a 1% change in VIX) has increased around regulatory changes in the U.S. markets that
increased the role of public traders in liquidity provision, reduced the minimum allowable price
variation, weakened the affirmative obligation of NASDAQ dealers, and abolished the specialist
system on the NYSE.

JEL classification: G01; G02; G10; G18

Keywords: Liquidity commonality; VIX; Volatility; Market makers; Uncertainty; Bid-ask spread;
Market structure
________________________________________________________________________

__________
☆
The paper benefitted greatly from the comments and suggestions of an anonymous referee.
The authors also thank Bill Schwert (the editor), Stephan Dieckmann, Wayne Ferson, Jaehoon
Hahn, Daehee Jeong, Hyoung-Goo Kang, In Joon Kim, Jung-Wook Kim, Woojin Kim, Kuan-Hui
Lee, Sang Bin Lee, Soojin Lee, Pamela Moulton, Cristian Tiu, Jiang Wang, Betty Wu, Jin Yoo,
colleagues at SUNY-Buffalo and Kansas State University, and seminar participants at Hanyang
University, Seoul National University, Yonsei University, the 2012 KDI Journal of Economic
Policy Conference, and the 2013 FMA conference for valuable discussion, comments, and
suggestions. Chuwonganant is thankful for a grant research support from the College of Business
Administration at Kansas State University. The usual disclaimer applies.

* Corresponding author at: Department of Finance and Managerial Economics, School of

Management, State University of New York (SUNY) at Buffalo, Buffalo, NY 14260, USA.
Tel.: +1 716 645 3262; fax: +1 716 645 2131.
E-mail address: keechung@buffalo.edu (K.H. Chung).
1. Introduction

In this study we provide evidence regarding the effect of uncertainty on stock market

liquidity by analyzing the time-series relation between an index of stock market volatility (VIX)

and various measures of liquidity. Our study also sheds light on whether the impact of market

volatility on liquidity varies with market structure by examining the effects of four major

regulatory changes in the U.S. markets on the relation between VIX and liquidity.

Prior research finds commonality in liquidity. Chordia, Roll, and Subrahmanyam (2000),

Hasbrouck and Seppi (2001), and Huberman and Halka (2001) show that the liquidity of

individual stocks co-varies with both the liquidity of the market as a whole and the liquidity of

stocks in the same industry. Other studies show that liquidity-related risks are priced. For

instance, Acharya and Pedersen (2005) show that the risk premium is related to commonality in

liquidity with market liquidity, return sensitivity to market liquidity, 1 and liquidity sensitivity to

market returns. Korajczyk and Sadka (2008) show that only the across-measure global

systematic liquidity factor commands a risk premium. Sadka (2010) shows that hedge fund

returns are significantly related to the co-variation of fund returns with unexpected changes in

aggregate liquidity. Lee (2011) analyzes liquidity risks using international data and shows that

the pricing of liquidity risk varies across countries according to geographic, economic, and

political environments.

Prior studies offer both demand- and supply-side theories of liquidity commonality.

Demand-side theory suggests that liquidity commonality arises from the behavior of investors

and traders. Kamara, Lou, and Sadka (2008) show that an increase in institutional ownership

leads to an increase in both liquidity commonality and its cross-sectional variation. Koch, Ruenzi,

1
Pastor and Stambaugh (2003) is the first to report this finding.

1
and Starks (2010) conjecture that co-movements in liquidity could arise among stocks if they are

held by a group of investors that tend to trade in the same direction and at the same time and

show that stocks with higher mutual fund ownership exhibit larger co-movements in liquidity.

Karolyi, Lee, and Van Dijk (2012) show that liquidity commonality is greater during times of

high market volatility and in countries with a greater presence of international investors and

more correlated trading activity, and interpret the results as evidence for the demand-side theory.

Supply-side theory suggests that liquidity commonality arises from liquidity providers’

information sharing and capital constraints. For example, Coughenour and Saad (2004) hold that

liquidity co-variation arises because specialists within each firm make common adjustments in

liquidity provisions based on their shared capital and information. Chordia, Sarkar, and

Subrahmanyam (2005) find evidence that monetary policy gives rise to liquidity commonality in

the stock market. 2 Hameed, Kang, and Viswanathan (2010) show that liquidity commonality on

the NYSE increases during market decline when funding liquidity is tight. Most factors are likely

to affect both the demand for and supply of liquidity, and liquidity commonality would arise

from interactions of liquidity demanders and suppliers.

In this study we show that an important source of liquidity commonality is overall market

uncertainty using the Chicago Board Options Exchange Market Volatility Index (VIX). 3 This

index, often referred to as the fear index or the fear gauge, is a measure of the implied volatility

of S&P 500 index options. 4 To the extent that systematic liquidity variation is a priced factor,

2
Chordia, Sarkar, and Subrahmanyam (2005) also conjecture that common factors drive liquidity and
volatility in stock and bond markets based on their finding that innovations to stock and bond market liquidity and
volatility are highly correlated.
3
Chordia, Roll, and Subrahmanyam (2000) note that “The risk of maintaining inventory depends also on
volatility, which could have a market component.” However, they do not expand on the conjecture.
4
In 1993, the Chicago Board Options Exchange (CBOE) introduced the CBOE Volatility Index, VIX,
which measures the market’s expectation of 30-day volatility implied by at-the-money S&P 100 Index option prices.
In 2003, CBOE introduced a new method of estimating the VIX based on the S&P 500 Index. The new method

2
understanding the causes of liquidity co-variation should help investors and traders to better deal

with such risk. Empirical evidence regarding the sources of liquidity commonality could also

help financial economists to better understand risk premiums and asset prices. In addition, our

study sheds light on whether the regulatory changes that relaxed the affirmative obligation of

NASDAQ dealers and abolished the specialist system on the NYSE are responsible, at least in

part, for the recent fluctuations in market liquidity.

The present study contributes to a growing literature that uses VIX as a measure of

expected volatility. Bao, Pan, and Wang (2008) show that monthly changes in aggregate bond

market liquidity are strongly related to changes in VIX. Pan and Singleton (2008) and Longstaff,

Pan, Pedersen, and Singleton (2010) find a strong correlation between sovereign credit spreads

and VIX. Graham and Harvey (2010) show that equity risk premium closely tracks VIX over

time and increases sharply during financial crises. Adrian and Shin (2010) argue that risk-

management constraints reduce the risk appetite of financial intermediaries in times of high VIX.

Bekaert, Hoerova, and Lo Duca (2011) find a high correlation between VIX and monetary policy.

Nagel (2012) finds that the expected return from liquidity provision is time-varying and

increases with VIX.

Market microstructure theory (see, e.g., Ho and Stoll, 1981 and Glosten and Milgrom,

1985) predicts that liquidity providers widen the bid-ask spread (i.e., an inverse measure of

liquidity) when inventory holding or adverse selection risks are high. Consistent with this

prediction, prior research shows that the bid-ask spread of a stock increases with its own risk,

typically measured by the standard deviation of quote midpoint returns and/or the standard error

estimates expected volatility by averaging the weighted prices of SPX puts and calls over a wide range of strike
prices (source: http://www.cboe.com/micro/vix/vixwhite.pdf).

3
calculated from the market model (i.e., unsystematic risk). 5 In the present study, we provide

empirical evidence that the liquidity of an individual asset is related not only to its own risk, but

also to overall market uncertainly reflected in VIX.

How uncertainty exerts an impact on liquidity is likely to depend on market structure. We

conjecture that uncertainty exerts a larger impact on liquidity when public traders play a greater

role in liquidity provision, when the minimum price variation (i.e., tick size) is smaller, and when

market makers play a smaller role in liquidity provision. We test these conjectures using the

following four regulatory changes, which serve as natural experiments, in market structure: (1)

the implementation of the new order handling rules (OHR) on NASDAQ in 1997; (2) the

reduction of tick size from $1/8 to $1/16 in 1997 and from $1/16 to $0.01 (decimalization) in

2001; (3) the amendment of NASDAQ Rule 4613 (c) in 2007 that dealer quotes must be

reasonably related to the prevailing market; and (4) the implementation of the designated market

maker (DMM) system on the NYSE in 2008.

We show that uncertainty plays an important role in liquidity commonality after

controlling for aggregate market liquidity, industry-wide liquidity, and well known determinants

of individual stock liquidity, such as return volatility, trading volume, and price. For instance, we

show that a 1% increase in VIX leads to a 0.49% (0.5%) increase in the effective spread and a

0.41% (0.42%) decrease in the quoted depth of NASDAQ (NYSE) stocks. We also show that the

effect of VIX on stock liquidity is greater than the combined effects of all other common

determinants of stock liquidity.

Our results are consistent with the prediction of Brunnermeier and Pedersen (2009) that

market liquidity decreases with VIX because higher volatility reduces market makers’ liquidity-

5
See, for example, Benston and Hagerman (1974) and Stoll (2000).

4
provision capacity. The results support the view of Nagel (2012) that liquidity evaporates during

financial crises because liquidity providers require greater returns during such times. Our results

are also consistent with anecdotal evidence that liquidity providers adjust their positions

uniformly across stocks according to market volatility as reflected in VIX before they make

adjustments based on individual asset risks. We provide other alternative explanations.

Consistent with our conjecture, we show that the effects of VIX on liquidity have

changed over time across different market structures. Specifically, we show that the uncertainty

elasticity of liquidity (UEL: % change in liquidity given a 1% change in VIX) has increased

dramatically around regulatory changes in the U.S. markets that increased the role of public

traders in liquidity provision, reduced the minimum allowable price variation, weakened the

affirmative obligation of NASDAQ dealers, and abolished the specialist system on the NYSE.

These results support the idea that a direct reflection of expected volatility in prices and quotes,

without the filtering by market intermediaries, may increase the volatility of market liquidity.

Although some of these regulatory changes have been shown to increase pricing efficiency and

reduce trading costs, 6 they might also have the unintended consequence of increasing liquidity

volatility, leading to an increase in liquidity risk premiums in asset returns.

The rest of the paper is organized as follows. Section 2 describes our data sources and

variable measurement methods. Section 3 conducts regression analyses to examine the relation

between VIX and various liquidity measures after controlling for the effects of other

determinants of liquidity, including market liquidity, industry-wide liquidity, and risk of

individual assets. Section 4 analyzes the effects of four major regulatory changes in the U.S.

6
For instance, Chordia, Roll, and Subrahmanyam (2008) show that the reduction in tick size in the U.S.
stock markets led to an increase in market efficiency and a decrease in the bid-ask spread.

5
markets on the relation between VIX and liquidity. Section 5 concludes the paper with a brief

summary and suggestions for future research.

2. Data sources and variable description

Our initial study sample includes NYSE and NASDAQ stocks from January 2007 to

December 2009. We obtain trade and quote data from the NYSE’s Trade and Quote (TAQ)

database. We retrieve trading volume and number of shares outstanding from the Center for

Research in Security Prices (CRSP) database and book value of equity from the Compustat

database. We obtain quarterly 13F institutional holdings data from Thomson Financial and

analyst coverage data from the Institutional Brokers' Estimate System (I/B/E/S). We assume zero

analyst following for firms not included in the I/B/E/S database. We exclude from our sample

stocks with missing volume greater than one percent of the number of trading days in the study

period, stocks with price lower than $1, and stocks with average daily volume less than 1,000

shares. 7 Our final sample consists of 1,945 NYSE stocks and 1,743 NASDAQ stocks.

We use the TAQ data to construct national bid and offer (NBBO) quotes. We exclude the

following quotes and trades to minimize data errors: quotes if either the ask or bid price is non-

positive; quotes if either the ask or bid size is non-positive; quotes if the bid-ask spread is greater

than $5 or non-positive; before-the-open and after-the-close trades and quotes; trades if the price

or volume is non-positive; bid quote, Bidt, if |(Bidt – Bidt-1)/Bidt-1| > 0.5; ask quote, Askt, if |(Askt

– Askt-1)/ Askt-1| > 0.5; and trade price, pt, if |(pt – pt-1) /pt-1| > 0.5.

We measure the quoted and effective spreads, depth, price impact, and market quality

index using following formulas:

7
We apply these sample selection criteria throughout the paper.

6
 Quoted spreadi,t = (Aski,t – Bidi,t) / Mi,t, (1)

Effective spreadi,t = 2Di,t(Pi,t – Mi,t) / Mi,t, (2)

Depthi,t = Mi,t(Bid sizei,t + Ask sizei,t) / 2, (3)

Price impacti,t = Di,t(Mi,t+5 – Mi,t) / Mi,t, (4)

Market quality indexi,t = ½ (Bid sizei,t + Ask sizei,t) / [(Aski,t – Bidi,t) / Mi,t,], (5)

where Aski,t is the national best ask price of stock i at time t, Bidi,t is the national best bid price of

stock i at time t, Mi,t is the quote midpoint ((Aski,t + Bidi,t)/2) of stock i at time t, Mi,t+5 is the

quote midpoint at time t + 5 minutes of stock i, Pi,t is the transaction price of stock i at time t, Bid

sizei,t is the quoted size at the bid of stock i at time t, Ask sizei,t is the quoted size at the ask of

stock i at time t, and Di,t is an indicator variable which equals +1 for customer buy orders and -1

for customer sell orders. The quoted depth in the TAQ database does not truly reflect the total

available liquidity for NASDAQ stocks. We include the quoted depth and market quality index

in our analysis for completeness. We estimate Di,t using the algorithm in Lee and Ready (1991)

and Bessembinder (2003) with no allowance for trade-reporting lag. We use the time-weighted

quoted spread and trade-weighted effective spread in our study. We measure return volatility by

the standard deviation of quote midpoint returns. We measure trading volume by the dollar

trading volume.

We obtain daily observations of the Chicago Board Options Exchange Market Volatility

Index (VIX) for our study period from Yahoo Finance (http://finance.yahoo.com/q/hp?s=^VIX+

Historical+Prices). VIX is quoted in percentage points and it measures the expected (annualized)

movement in the S&P 500 index over the following 30-day period. For example, if VIX is 15, it

means that the index option markets expect the S&P 500 index to move up or down by less than

7
or equal to 15%/√12 = 4.33% over the next 30-day period with a 68% likelihood (one standard

deviation). Although VIX is a measure of the implied volatility of S&P 500 index options, we,

following other studies, use it as an empirical proxy for market-wide uncertainty. We present

evidence that this is a reasonable assumption later in the paper.

Table 1 shows descriptive statistics (i.e., mean, standard deviation, and select percentile

values) for our study sample of NASDAQ and NYSE stocks. The results show that the mean

values (0.0024, 0.002, and 0.0013) of the quoted spread, effective spread, and price impact for

NYSE stocks are all much smaller than the corresponding figures (0.006, 0.0053, and 0.0033) for

NASDAQ stocks. Not surprisingly, we find that NYSE stocks have, on average, higher share

price, larger trading volume, and lower return volatility than NASDAQ stocks.

3. Regression results for the effect of VIX on liquidity measures

In this section, we analyze the effect of VIX on liquidity and show how the relation

between VIX and liquidity varies with stock attributes.

3.1. Effect of VIX on liquidity

We estimate the following regression models for each stock using daily time-series data

from January 3, 2007 through December 31, 2009 to examine the relation between VIX and

various measures of liquidity after controlling for other determinants of liquidity. Our models are

based on those used in Chordia, Roll, and Subrahmanyam (2000, 2001) and Coughenour and

Saad (2004). We obtain qualitatively similar results and draw identical inferences when we

include additional control variables, e.g., a dummy variable for each month and lead and lag

8
values of the dollar trading volume and share price. We chose to report the results from a

parsimonious model specification.

log(LMi,t) = αi0 + αi1log(VIXt) + αi2log(VIXt-1) + αi3log(VIXt+1)

+ αi4log(MKTLMt) + αi5log(MKTLMt-1) + αi6log(MKTLMt+1)

+ αi7log(INDLMt) + αi8log(INDLMt-1) + αi9log(INDLMt+1)

+ αi10log(PRICEi,t) + αi11log(VOLi,t) + αi12log(VOLAi,t)

+ αi13log(VOLAi,t-1) + αi14log(VOLAi,t+1) + αi15MKTRETt

+ αi16MKTRETt-1 + αi17MKTRETt+1 + αi18TUESDAY

+ αi19WEDNESDAY + αi20THURSDAY + αi21FRIDAY + ε1i,t, (6)

DLMi,t = βi0 + βi1DVIXt + βi2DVIXt-1 + βi3DVIXt+1 + βi4DMKTLMt

+ βi5DMKTLMt-1 + βi6DMKTLMt+1 + βi7DINDLMt + βi8DINDLMt-1

+ βi9DINDLMt+1 + βi10DPRICEi,t + βi11DVOLi,t + βi12DVOLAi,t

+ βi13DVOLAi,t-1 + βi14DVOLAi,t+1 + βi15DMKTRETt

+ βi16DMKTRETt-1 + βi17DMKTRETt+1 + βi18TUESDAY

+ βi19WEDNESDAY + βi20THURSDAY + βi21FRIDAY + ε2i,t, (7)

where, in regression model (6), LMi,t is one of five liquidity measures (i.e., quoted spread,

effective spread, price impact, depth, and market quality index) of stock i on day t; VIXt-1 , VIXt,

VIXt+1 are the CBOE VIX index on day t-1, t, and t+1; MKTLMt-1, MKTLMt, and MKTLMt+1

are the market liquidity measures (e.g., the average quoted spread across all stocks); INDLMt-1,

INDLMt, and INDLMt+1 are the industry-wide liquidity measures (e.g., the average quoted

percentage spread across all stocks in the same industry); PRICEi,t is the price of stock i on day t;

VOLi,t is the dollar trading volume of stock i on day t; VOLAi,t-1,VOLAi,t, and VOLAi,t+1 are the

9
standard deviation of five-minute quote midpoint returns of stock i; MKTRETt-1, MKTRETt, and

MKTRETt+1 are the market returns; TUESDAY, WEDNESDAY, THURSDAY, and FRIDAY

are dummy variables for Tuesday, Wednesday, Thursday, and Friday; and ε1i,t is the error term.

We exclude stock i when we calculate the market and industry liquidity measures. Note that each

estimated coefficient in regression model (6) represents % change in each liquidity measure

given a 1% change (i.e., elasticity) in each independent variable.

It is important to note that we would not need to include the market- and industry-wide

liquidity variables [i.e., log(MKTLMt), log(MKTLMt-1), log(MKTLMt+1), log(INDLMt),

log(INDLMt-1), and log(INDLMt+1)] in regression model (6) if VIX were the only source of

liquidity commonality because these variables would merely capture the aggregate (i.e., market-

and industry-wide) effects of VIX on liquidity. We include them in the model to control for the

effects of other variables that may also cause liquidity commonality. For instance, changes in the

federal funds rate or money supply may affect the liquidity of individual stocks, giving rise to

another source of liquidity commonality. In this case, regression coefficients on the market- and

industry-wide liquidity variables would capture the co-movements between the liquidity of

individual stocks and the market- and industry-wide liquidity that are attributable to changes in

the federal funds rate or money supply.

We include share price, trading volume, and return volatility in the model because these

variables are known to have significant effects on the liquidity of individual stocks. 8 We include

the lead and lag values of return volatility to account for the possibility that liquidity providers

make anticipatory or delayed adjustments in liquidity positions in response to daily changes in

return volatility. Following Chordia, Roll, and Subrahmanyam (2000) and Chordia, Sarkar, and

8
See Benston and Hagerman (1974) and Stoll (2000).

10
Subrahmanyam (2005), we include market returns (i.e., MKTRETt-1, MKTRETt, and

MKTRETt+1) and dummy variables for day of the week in the model.

In regression model (7), prefix D denotes the percentage change of the variable from the

previous day [e.g., DLMi,t = (LMi,t – LMi,t-1) / LMi,t-1]. We estimate this alternative regression

model to assess the sensitivity of our results with respect to different estimation methods. 9 We

estimate regression models (6) and (7) for each stock and then calculate the mean values of the

regression coefficients.

Panel A in Table 2 shows the results of regression model (6) for NASDAQ stocks and

NYSE stocks. We report the results for NASDAQ and NYSE stocks separately because the

effect of VIX on liquidity could be different between the two groups due to their different market

structures. Because our main focus is to examine the effect of VIX, market liquidity, industry

liquidity, and stock return volatility on individual stock liquidity, we report only the regression

coefficients on these variables in the table. The first four columns show the mean coefficients

(αi1, αi2, and αi3) for the contemporaneous, lag, and lead values of VIX, and the sum of these

coefficients (αi1 + αi2 + αi13), together with their respective t-statistics. The next four columns

show the mean coefficients (αi4, αi5, and αi6) for the contemporaneous, lag, and lead values of

market liquidity, and the sum of these coefficients (αi4 + αi5+ αi6). The next four columns show

the results for industry liquidity and the last four columns show the results for stock return
𝑥
volatility. Numbers in parentheses are t-statistics. We report the t-statistics as �𝑦� , where x is the
𝑧

t-value of the average regression coefficient, y is the mean t-value across all individual stock

regressions, and z is the median t-value across all individual stock regressions. The first t-statistic

9
Regressions using changes in the variables are less likely to suffer from econometric problems (e.g.,
nonstationarity) than those using the level variables.

11
(x) is calculated under the assumption that the estimation errors in α’s are independent across

stocks. We provide evidence regarding this assumption later in the paper.

The mean coefficients on VIXt are positive and significant in the quoted spread, effective

spread, and price impact regression models, and negative and significant in the depth and market

quality index regression models for both NASDAQ and NYSE stocks. It is unlikely that the

relation between VIX and our liquidity measures is driven by reverse causality because most of

our sample stocks do not belong to the S&P 500 index. 10 The results also show that the mean and

median t-values are all statistically significant. For NASDAQ stocks, we find that a 1% increase

in VIXt leads to a 0.46%, 0.49%, and 0.43% increase in the quoted spread, effective spread, and

price impact, respectively, and a 0.41% and 0.49% decrease in the quoted depth and market

quality index, respectively. We find similar results for NYSE stocks. Hence, VIX has

incremental explanatory power on temporal liquidity variation, over that of individual asset

return volatility. Apparently, liquidity providers react strongly to general market uncertainty (in

addition to individual asset risk) and this behavior generates a commonality in liquidity. We find

weaker results for VIXt-1 and VIXt+1 with much smaller (in absolute values) regression

coefficients for both NASDAQ and NYSE stocks. The mean and median t-values are not

statistically significant. Together, these results indicate that liquidity providers make quick

contemporaneous adjustments in liquidity positions in response to daily changes in market

uncertainty with minimal delayed or anticipatory adjustments.

Our results are consistent with the implication of the theoretical model proposed by

Brunnermeier and Pedersen (2009) that market liquidity decreases with VIX because higher

volatility tightens funding constraints on market makers and thereby reduces their liquidity-

10
It is unlikely that liquidity of a stock that does not belong to the S&P 500 index can exert an impact on
VIX, which is constructed using the implied volatilities of a wide range of S&P 500 index options.

12
provision capacity. The results are also consistent with the view of Nagel (2012) that liquidity

may disappear during periods of financial market turmoil because liquidity providers demand a

higher expected return from liquidity provision during such times.

There are a few other possible interpretations of our results. Liquidity providers may

adjust their positions uniformly across stocks according to readily available VIX before they

make security-specific adjustments based on security-specific risks. (Anecdotal evidence from

conversations with industry people suggests that this is indeed the practice followed by many

portfolio managers.) Also, VIX may capture additional (inventory and adverse selection)

risks/costs faced by liquidity providers that are not captured by individual asset risk and other

control variables. Finally, liquidity providers may become more risk averse when VIX is higher.

Consistent with the finding of prior studies, we find a significant and positive

contemporaneous correlation between individual asset liquidity and both overall market liquidity

and industry-wide liquidity. For instance, for NASDAQ stocks, we find that a 1% increase in

MKTLMt is associated with a 0.18%, 0.18%, 0.16%, 0.21%, and 0.24% increase in the quoted

spread, effective spread, price impact, quoted depth, and market quality index, respectively.

Similarly, a 1% increase in INDLMt is associated with a 0.14%, 0.15%, 0.12%, 0.18%, and 0.19%

increase in the quoted spread, effective spread, price impact, quoted depth, and market quality

index, respectively. We find similar results for NYSE stocks. On the whole, these results indicate

that VIX is not the only source of liquidity commonality. We find much weaker and inconsistent

relations between individual stock liquidity and the lead and lag values of overall market

liquidity and industry-wide liquidity for both NASDAQ and NYSE stocks.

The mean coefficients on individual stock volatility (VOLAt) are positive and significant

in the quoted spread, effective spread, and price impact regression models, and negative and

13
significant in the depth and market quality index regression models for both NASDAQ and

NYSE stocks. The results also show that the mean and median t-values are all statistically

significant. For NASDAQ stocks, we find that a 1% increase in VOLAt leads to a 0.15%, 0.16%,

and 0.15% increase in the quoted spread, effective spread, and price impact, respectively, and a

0.11% and 0.21% decrease in the quoted depth and market quality index, respectively. We find

similar results for NYSE stocks. We find weaker results for VOLAt-1 and VOLAt+1 with much

smaller (in absolute values) regression coefficients for both NASDAQ and NYSE stocks. The

mean and median t-values are not statistically significant. On the whole, these results indicate

that liquidity providers make quick contemporaneous adjustments in liquidity positions to daily

changes in stock return volatility with minimal delayed or anticipatory adjustments.

Panel B shows the results of regression model (7). On the whole, the results from

regression model (7) are qualitatively identical to those from regression model (6). Hence we

conclude that our results are robust and not sensitive to different estimation methods.

The reliability of the t-statistics in Table 2 depends on the independence of the residuals

across stocks. We test the independence of the residuals from regression models (6) and (7) using

the method in Chordia, Roll, and Subrahmanyam (2000) and Coughenour and Saad (2004). We

first sort 1,743 NASDAQ stocks alphabetically using ticker symbols and assign each stock a

serial number i (i = 1,…., 1,743). We then estimate the following regression models: ε1i+1,t = θ0

+ θ1ε1i,t + μ1i,t and ε2i+1,t = δ0 + δ1ε2i,t + μ2i,t (i = 1,…., 1,742); where ε1i,t (ε1i+1,t), ε2i,t (ε2i+1,t) are

residuals for stock i (i+1) from regression models (6) and (7), respectively, and μ1i,t and μ2i,t are

disturbance terms. We also calculate the correlation coefficient between ε1i+1,t and ε1i,t and

between ε2i+1,t and ε2i,t.

14
 Panel C in Table 2 shows the average correlation coefficients, the average t-statistics of

the regression coefficients θ1 and δ1, and the proportions of absolute t-statistics of θ1 and δ1

greater than the critical significant level of five percent (|t| > 1.96). The results show that the

mean values of regression coefficients θ1 and δ1 are not significantly different from zero,

indicating the independence of the residuals across stocks. We find similar results for the NYSE

sample. Based on these results, we conclude that our t-statistics are reasonable metrics for

assessing the statistical significance of the mean regression coefficients.

3.2. Standardized coefficients

To compare the relative effects of VIX, market liquidity (MKTLM), industry liquidity

(INDLM), and return volatility (VOLA) on individual stock liquidity, we also estimate

standardized coefficients on these variables after they are standardized so that their variances are

all equal to one. Standardized coefficients refer to changes in individual stock liquidity (in

number of standard deviations) per one standard deviation increase in VIX, market liquidity, and

industry liquidity. Appendix Panel A shows the results for regression model (6) and Appendix

Panel B shows the results for regression model (7). Note that the standardized coefficients in the

Appendix are qualitatively similar to the non-standardized coefficients reported in Panel A and

Panel B of Table 2. Hence, our inferences would be identical regardless of whether we used the

standardized or non-standardized regression coefficients.

The regression coefficients on VIXt are much larger (in absolute values) than the

regression coefficients on market liquidity, industry liquidity, and return volatility. Appendix

Panel A shows that for NYSE stocks the standardized coefficients on VIXt in the regression

model for the quoted spread, effective spread, price impact, depth, and market quality index are

15
0.3985, 0.4107, 0.3182, -0.3604, and -0.3505, which are all greater (in absolute value) than the

corresponding figures (0.1572, 0.1878, 0.1590, 0.1844, and 0.2257) for MKTLMt, the

corresponding figures (0.1186, 0.1632, 0.1441, 0.152, and 0.1741) for INDLMt, and the

corresponding figures (0.1227, 0.1245, 0.1113, -0.1181, and -0.1425) for VOLAt. The

differences are all significantly different from zero according to the t-test. We find qualitatively

similar results for NASDAQ stocks.

Furthermore, the coefficients (0.3985, 0.4107, 0.3182, -0.3604, and -0.3505) on VIX for

NYSE stocks are mostly significantly larger (in absolute values) than the summation (0.2758,

0.3510, 0.3031, 0.3364, and 0.3998) of the corresponding coefficients on market liquidity and

the corresponding coefficients on industry liquidity. We find similar results for NASDAQ stocks.

In addition, Appendix Panel B shows that the results from regression model (7) are qualitatively

similar to those from regression model (6). Overall, these results suggest that the effect of

uncertainty on stock liquidity is greater than the combined effects of all other common

determinants (e.g., federal funds rate or money supply) of stock liquidity, indicating the critical

importance of uncertainty as a determinant of liquidity commonality.

3.3. Stock attributes, market structure, and the uncertainty elasticity of liquidity (UEL)

In this section, we examine how the uncertainty elasticity of liquidity (UEL) is related to

stock attributes and market structure using the following regression model for the combined

sample of 1,945 NYSE stocks and 1,743 NASDAQ stocks:

UELi = π0 + π1NASDAQ + π2PRICEi + π3VOLi + π4MVEi + π5VOLAi + π6NAFi

+ π7INSTi + π8MVTOBVi + π9PMINi + π10SP500 + εi, (8)

16
where UELi is either the estimate of (α1+ α2 + α3) for stock i from regression model (6) or the

estimate of (β1 + β2 + β3) for stock i from regression model (7); 11 NASDAQ is an indicator

variable which equals one for NASDAQ stocks and zero for NYSE stocks; PRICEi is the average

share price of stock i; VOLi is the daily dollar trading volume of stock i; MVEi is the average

market capitalization of stock i; VOLAi is the stock i’s return volatility; NAFi is the number of

analysts following stock i; INSTi is the percentage of stock i’s shares held by institutional

investors; MVTOBVi is the market-to-book value of equity ratio for stock i; PMINi is the

proportion of spread quotes that are equal to the minimum price variation for stock i; SP500 is an

indicator variable which equals one for S&P 500 stocks and zero otherwise; and εi is the error

term. We use the mean values of PRICEi, VOLi, MVEi,VOLAi, NAFi, INSTi, MVTOBVi, and

PMINi for each stock during the entire study period (i.e., January 3, 2007 through December 31,

2009) in the regression.

We include PMINi in the regression model to account for the fact that the uncertainty

elasticity of liquidity (UEL) is likely to be affected by the probability that the minimum price

variation (i.e., tick size) is a binding constraint on spreads. For instance, when the equilibrium

dollar spread of a given stock is smaller than the tick size, the stock’s spread will have a small

UEL simply because it is bounded by the tick size (e.g., the spread cannot go down below the

tick size when VIX becomes very low). In this case, all things being equal, liquidity providers

may adjust depth quotes more frequently, resulting in a large UEL for the stock’s depth.

Panel A in Table 3 shows the results of regression model (8) when UEL is estimated from

regression model (6) and Panel B shows the results when UEL is estimated from regression

model (7). The results show that the regression coefficients on the NASDAQ dummy variable

11
We obtain qualitatively similar results when we measure UEL with only α1 and β1.

17
are not significantly different from zero across all liquidity measures, indicating that the effect of

market uncertainty on liquidity is similar between NASDAQ and NYSE stocks. The results show

that UEL is positively and significantly related to share price, trading volume, and return

volatility, and negatively and significantly related to firm size (market value of equity) when

liquidity is measured by the quoted spread, effective spread, or price impact. These results

suggest that uncertainty exerts a greater impact on the liquidity of stocks with higher price, larger

trading volume, greater return volatility, or smaller market capitalizations.

The larger UEL of higher priced stocks may be explained partly by the fact that the tick

size is less likely to be a binding constraint on price quote changes. The larger UEL of high-

volume or high-volatility stocks may be due to the fact that these stocks are more likely to

exhibit information-based trading, prompting greater reactions from liquidity providers. That

uncertainty exerts a smaller impact on the liquidity of larger companies may be explained by the

fact that these companies have lower information-based trading because generally, more

information is available about them, and thus liquidity providers in these stocks may be less

sensitive to changes in market volatility.

When liquidity is measured by the dollar depth or market quality index, the signs of the

estimated coefficients on share price, trading volume, firm size, and return volatility in

regression model (8) are the opposite of those when liquidity is measured by the spreads and

price impact: UEL is positively and significantly related to firm size, and negatively and

significantly related to share price, trading volume, and return volatility. This is an expected

result because the UEL estimates when liquidity is measured by the dollar depth or market

quality index are likely to have opposite signs of the UEL estimates when liquidity is measured

by the spread or price impact.

18
 Uncertainty exerts a smaller impact on the liquidity of stocks with higher analyst

following and/or institutional ownership. One possible interpretation of this result is that

liquidity providers in these stocks are less sensitive to changes in market volatility because more

information is available through analysts’ information collection and dissemination as well as

institutional monitoring. 12 We find that uncertainty exerts a larger impact on the liquidity of

firms with higher market-to-book ratios. To the extent that firms with higher market-to-book

ratios have larger intangible assets (e.g., higher proportions of these firms’ market values are

accounted for by future growth options), our results support the notion that changes in market

volatility lead to larger swings in liquidity when firm value is highly subject to future managerial

actions.

We find that UEL is negatively and significantly related to PMIN when liquidity is

measured by the quoted spread, effective spread, or price impact, and positively and significantly

related to PMIN when liquidity is measured by the quoted depth and market quality index. These

results are consistent with our expectation that liquidity providers make more frequent

adjustments in depth quotes when they cannot make adjustments in spreads because the tick size

is a binding constraint on spreads. We find that the estimated coefficients on the S&P 500

dummy variable are not significantly different from zero in all regressions. We interpret this

result as evidence that liquidity providers consider VIX as a measure of overall market

uncertainty rather than as a measure of uncertainty associated with only those stocks that belong

to the S&P 500 index.

12
Kamara, Lou, and Sadka (2008) show that the sensitivity of the stock’s liquidity to aggregate liquidity
shocks increases with institutional ownership. Our results are not directly comparable to their results because we
measure liquidity commonality that arises from market uncertainty.

19
4. Market structure and the uncertainty elasticity of liquidity

In this section, we analyze how market structure affects the uncertainty elasticity of

liquidity using data surrounding four major structural changes in the U.S. securities markets.

4.1. Order handling rules

The Securities and Exchange Commission (SEC) began to enact major changes in the

order handling rules (OHR) in the U.S. securities markets on January 20, 1997. Although all U.S.

securities markets were subject to the new rules, the NASDAQ Stock Market was the primary

target of the new rules. Prior to the new rules, the NASDAQ Stock Market had operated as a

pure dealer market, where public traders were unable to compete with market makers with limit

orders. Public limit orders were considered bids and offers to NASDAQ dealers (but not to other

public traders) before the implementation of the new OHR. The new limit order handling rule

enabled public traders to directly compete with market makers in the price discovery process by

requiring the latter to display public limit orders in the BBO (i.e., best bid and offer) when they

were better than quotes posted by market makers. The new quote rule also gave the public access

to quotes posted by market makers in the Electronic Communication Networks (ECN). Under the

new rule, if a dealer places a limit order into Instinet or another ECN, the price and quantity are

incorporated in the ECN quote displayed on NASDAQ if it represents the best bid or offer in

ECN.

Prior research shows that the new order handling rules help reduce trading costs by

promoting competition among public traders and market makers. 13 In the present study, we

analyze a potential unintended consequence of the new rules. Greater direct participation of

13
See Barclay, Christie, Harris, Kandel, and Schultz (1999).

20
public traders in liquidity provision may imply a more direct reflection of market volatility in

liquidity. Consequently, we conjecture an increase in UEL after the implementation of the new

order handling rules.

The first phase-in of the OHR stocks was on January 20, 1997 and the last phase-in was

on October 13, 1997. Another major event that occurred on NASDAQ during this time period

was the tick-size reduction from $1/8 to $1/16 on June 2, 1997. To examine the impact of the

new OHR without the confounding effect of the tick-size change, we use the first three phase-ins

of the OHR stocks (50 securities for each phase-in) on January 20, February 10, and February

24, 1997 for our OHR sample. These NASDAQ stocks were subject to the new order handling

rules but were not subject to the tick-size change. The pre-period is 65 days from October 16,

1996 to January 17, 1997 and the post-period is 65 days from February 25, 1997 to May 28,

1997. The post-period ends before the tick-size reduction on June 2, 1997. In addition, our OHR

sample consists only of stocks that were selling at prices of $10 or greater because the new order

handling rules applied only to those securities.

To measure the effect of the OHR on NASDAQ stocks accurately, we construct a control

sample of NYSE stocks that are similar to NASDAQ stocks. To do this, we first calculate the

composite match score (CMS) of each NASDAQ stock that was subject to the new OHR against

each NYSE stock with the same two-digit North American Industry Classification System

(NAICS) code: CMS = Σ[(YjNASDAQ - YjNYSE)/{(YjNASDAQ + YjNYSE)/2}]2, where Yj represents one

of the four stock attributes (share price, dollar volume, volatility, and market capitalization) in the

pre-OHR period, Σ denotes the summation over j = 1 to 4, and superscripts NASDAQ and NYSE

represent NASDAQ and NYSE stocks, respectively. Then, for each NASDAQ stock, we select

the NYSE stock with the lowest score. Once we match a NYSE stock with a NASDAQ stock, we

21
no longer consider that particular NYSE stock for subsequent matches. Our final sample

consists of 119 matching pairs of NASDAQ/NYSE stocks.

We estimate regression models (6) and (7) using data for NASDAQ securities in the

periods before and after the implementation of the new OHR. Likewise, we estimate the models

using data for the control sample of NYSE stocks. Panel A in Table 4 shows the results using

UELs estimated from regression model (6) and Panel B shows the results using UELs estimated

from regression model (7). The first two columns in each panel show the mean UEL for

NASDAQ stocks in the pre- and post-OHR periods, respectively, and the third column shows the

difference in the mean UEL between the two periods (together with the t-statistic).

The results show that for NASDAQ stocks, the mean UELs in the post-OHR period are

all significantly larger (in absolute value) than those in the pre-OHR period across all five

liquidity measures, indicating that the implementation of the new OHR magnified the effect of

market uncertainty on liquidity. The next three columns show the results for the control sample

of NYSE stocks. The results show that the mean UELs in the post-OHR period are not

significantly different from those in the pre-OHR period. The last column in Table 4 shows that

the changes in the mean UELs between the pre- and post-event periods for NASDAQ stocks are

significantly greater in absolute value than the corresponding figures for the NYSE control

sample.

To determine whether the results in Table 4 are driven by concurrent changes in stock

attributes, we estimate the following regression models using only the NASDAQ sample:

UELi = π0 + π1 POST + π2 PRICEi + π3 VOLi + π4 MVEi + π5VOLAi + π6PMINi + ε1i, (9)

UELi
post
– UELipre = λ0 + Σλk (Xkipost – Xkipre) + ε2i, (10)

22
where POST is an indicator variable which equals one for the post-OHR period and zero

otherwise; the superscripts ‘post’ and ‘pre’ denote, respectively, the post- and pre-periods; Xk (k

= 1 through 5) represents share price, dollar trading volume, market capitalization, stock return

volatility, and the proportion of spread quotes that are equal to minimum price variation; and all

other variables are the same as defined in regression model (8). We do not include NAFi, INSTi,

and SP500 in the model because changes in these variables during the estimation period are

negligible. We use the mean values of PRICEi, VOLi, MVEi,VOLAi, and PMINi for each stock

during the pre- and post-OHR periods, respectively, in the regression. The coefficient (π1) for

POST in regression model (9) and the intercept (λ0) in regression model (10) measure the impact

of the new OHR on UEL after controlling for concurrent changes in stock attributes.

The results (see the first and second columns of Panel A and Panel B in Table 5) show

that the estimated coefficients (π1) on POST and the estimated intercepts (λ0) are positive and

significant when liquidity (i.e., LM in regression models (6) and (7)) is measured by the quoted

spread, effective spread, or price impact, and negative and significant when liquidity is measured

by the depth or market quality index. These results indicate that uncertainty has a great impact on

liquidity after the implementation of the new OHR. The estimates of π1 and λ0 are all very close

to corresponding values of ∆ (Post – Pre), indicating that the changes in UEL are largely

attributable to the new OHR rather than concurrent changes in stock attributes.

As a further robustness check, we also estimate the following regression model using the

combined sample of NASDAQ and NYSE stocks:

UELi = ω0 + ω1POST*NASDAQi + ω2NASDAQ + ω3POST + ω4PRICEi

+ ω5VOLi + ω6MVEi + ω7VOLAi + ω8PMINi + ε3i, (11)

23
where NASDAQ is an indicator variable which equals one for NASDAQ stocks and zero for

NYSE stocks and POST is an indicator variable which equals one for the post-rule change period

and zero otherwise. All other variables are the same as previously defined. We report the

estimates of ω1, ω2, and ω3 in the last three columns of Table 5.

The estimates of ω1 are positive and significant when we measure liquidity by the quoted

spread, effective spread, or price impact, and negative and significant when we measure liquidity

by the depth or market quality index. These results are consistent with those in Table 4 that the

changes in the mean UEL between the pre- and post-OHR periods for NASDAQ stocks are

greater in absolute value than the corresponding figures for the NYSE control sample, indicating

that the new OHR has a greater impact on NASDAQ. The estimates of ω2 are not significantly

different from zero, indicating that the effect of uncertainty on liquidity is similar between

NASDAQ and NYSE stocks before the implementation of the new OHR. The estimates of ω3 are

not significantly different from zero, indicating that the new OHR does not have a material effect

on how uncertainty affects liquidity for NYSE stocks. Overall, empirical results support our

conjecture that increased direct participation of public traders in liquidity services implies a more

direct reflection of market volatility in liquidity, which then magnifies the effect of market

volatility on liquidity.

It is important to note that the finding that liquidity is more sensitive to market volatility

in the post-OHR period does not necessarily imply a decrease in market quality because prior

research (see, Barclay, Christie, Harris, Kandel, and Schultz, 1999) shows that the new OHR

improved the level of liquidity (e.g., reduced the quoted and effective spreads). 14 The present

study underscores that the net benefit of the new OHR is likely to be smaller than the level of

14
We also find a significant decrease in the quoted and effective spreads from our study sample of stocks.

24
benefit implied by the improved level of liquidity. The same caveat applies to the results of the

1997 and 2001 tick size change analysis presented below. 15

4.2. Minimum price variation (tick size)

The SEC has reduced the minimum price variation (tick size) in the U.S. markets in an

effort to reduce trading costs, increase the informational efficiency of prices, and make U.S.

markets more competitive in the global market. Chordia, Roll, and Subrahmanyam (2008) show

that price movements in the U.S. markets have become increasingly closer to random walks as

the market friction imposed by the minimum allowable price variation has become smaller. The

study also shows that smaller tick sizes have led to smaller spreads and lower trading costs.

Although these results underscore the positive ramifications of smaller tick sizes, we consider a

potential unintended negative consequence of smaller tick sizes. Specifically, we conjecture that

smaller tick sizes may increase the speed at which liquidity providers react to market volatility,

resulting in large swings in liquidity.

4.2.1. The tick size change from $1/8 to $1/16

The tick size in the NASDAQ Stock Market was reduced from $1/8 to $1/16 on June 2,

1997. To measure the pure effect of the tick size change on liquidity elasticity that was not

contaminated by the effect of the new OHR, we use only those NASDAQ stocks that were

subject to the new OHR after the tick size change post-period (i.e., NASDAQ stocks that were

subject to the new OHR after September 3, 1997). The pre-period is 65 days from February 27,

1997 to May 30, 1997 and the post-period is 65 days from June 3, 1997 to September 3, 1997.

15
Prior research (see Chordia, Roll, and Subrahmanyam, 2008) shows that these tick size reductions
decreased spreads significantly.

25
Our sample consists of 1,587 NASDAQ stocks. The tick size on the NYSE was reduced from

$1/8 to $1/16 on June 24, 1997. Hence, for the NYSE sample, the pre-period is 100 days from

January 30, 1997 to June 23, 1997 and the post-period is 100 days from June 25, 1997 to

November 13, 1997. We use the 100-day pre- and post-period for the NYSE sample because it

was not constrained by the confounding effect of the new order handling rules. The sample

contains 2,123 NYSE stocks. The tick size change on the NYSE occurred only 15 days after the

tick size change on NASDAQ. Because the estimation of UELs using regression models (6) and

(7) requires at least 21 observations, we were unable to construct a control sample of NYSE

stocks as we did in the previous section. For the same reason, we were unable to construct a

control sample of NASDAQ stocks for the event sample of NYSE stocks.

Panel A in Table 6 shows the results using UELs estimated from regression model (6)

and Panel B shows the results using UELs estimated from regression model (7). In each panel,

we show the results for NASDAQ and NYSE stocks separately. For both NASDAQ and NYSE

stocks, mean UELs after the tick size reduction are significantly greater in absolute value than

those before the tick-size reduction, regardless of estimation methods. Furthermore, when we

estimate regression models (9) and (10) using data before and after the tick size reduction, we

find that the estimated coefficients (π1) on POST in regression model (9) and the estimated

intercepts (λ0) in regression model (10) are positive and significant when we measure liquidity

by the quoted spread, effective spread, or price impact, and negative and significant when we

measure liquidity by the depth or market quality index. (Here, POST is an indicator variable

which equals one for the post-tick size reduction period and zero otherwise.) Overall, these

results suggest that market uncertainty exerts a larger impact on liquidity after the tick size

reduction.

26
 As pointed out earlier, we include PMIN in the regressions to control for the change in

the extent to which the tick size is a binding constraint on spreads. Our results suggest that larger

UELs after the tick size reduction are not entirely due to the reduced binding constraint effect

because there are significant differences in spreads between the pre- and post-event periods even

after controlling for the binding constraint effect (PMIN) in the regressions. Liquidity providers

became more responsive to changes in expected volatility after the tick size reduction even when

the tick size is not a binding constraint on spreads. One possible explanation of these results is

that liquidity providers revise bid and ask quotes more readily because they could use finer price

increments under a smaller tick size.

4.2.2. Decimalization

The NASDAQ stock market began its decimal trading with 14 securities on March 12,

2001, followed by 197 stocks on March 26, 2001, and completed the decimalization for all

remaining stocks on April 9, 2001. For NASDAQ sample stocks, we use 100 days before March

12, 2001 (October 16, 2000 to March 9, 2001) as the pre-decimal period and 100 days after April

9, 2001 (April 10, 2001 to August 30, 2001) as the post-decimal period. Our sample contains

1,671 NASDAQ stocks. The NYSE started decimal trading with 7 stocks on August 29, 2000,

followed by 57 additional stocks on September 25, 2000, and 94 more securities on December 4,

2000. The NYSE completed its decimalization implementation for all stocks on January 29,

2001. The pre-decimal period for our NYSE sample stocks is 100 days before August 28, 2000

(April 5, 2000 to August 25, 2000) and the post-decimal period for the NYSE stocks is 100 days

after January 29, 2001 (January 30, 2001 to June 21, 2001). Our study sample consists of 1,903

NYSE stocks.

27
 When we reproduce Table 6 using the data before and after decimalization, we find that

the results (see Table 7) are qualitatively identical to those in Table 6. On the whole, these results

support the idea that market uncertainty exerts a greater impact on stock market liquidity when

the minimum allowable price variation is smaller because liquidity providers change their quotes

and orders more easily according to market volatility.

Decimal pricing on NASDAQ began 28 days after the completion of decimal pricing on

the NYSE. Hence, we do not have enough observations to estimate the UELs using regression

models (6) and (7) for a control sample of NASDAQ stocks for the post-decimal event sample of

NYSE stocks. For the same reason, we do not have enough observations for a control sample of

NYSE stocks for the pre-decimal event sample of NASDAQ stocks. As a result, we conduct the

matching sample tests using the abbreviated versions of regression models (6) and (7). We find

qualitatively similar results. (The results are available from the authors upon request.)

4.3. Amendment of NASDAQ Rule 4613 (c)

In October 2007, NASDAQ filed a proposed rule change to the SEC to eliminate the

requirement (i.e., NASDAQ Rule 4613 (c)) that the NASDAQ market makers’ quotes must be

“reasonably related to the prevailing market.” NASDAQ argued that such a requirement is not

necessary in highly competitive and automated trading environments. The SEC granted the

request for the rule change and the new rule became effective on November 7, 2007. Some

market observers believe that the maintenance of the stricter requirement on NASDAQ market

makers might have helped to prevent the “Flash Crash” of May 6, 2010.

As computerized trading firms increasingly dominate market making on most exchanges,

regulators have asked whether these firms provide liquidity when it is really needed. For instance,

28
SEC Chairman Mary Schapiro urged the agency to find ways to keep them from abandoning the

market at the first sign of trouble. Similarly, market observers and lawmakers have questioned

whether the high-frequency trading firms contributed to the Flash Crash by stepping away when

they were needed most. For instance, Senator Charles E. Schumer wrote in a letter to Schapiro:

"High-frequency traders pulled out during the free fall, leaving a dearth of liquidity and

exacerbating market volatility."

Prior to the rule amendment, NASDAQ dealers had the obligation to ensure that their

quotes reflected the prevailing market when public traders did not provide sufficient liquidity

(i.e., limit order prices are not representative of the prevailing market). After the rule amendment,

dealers no longer had such an affirmative obligation. Although some believe that high-frequency

trading (HFT) has filled any liquidity void, many believe that it is unwise to depend on high-

frequency traders for liquidity provision because they have no duty to provide liquidity and much

of HFT represents opportunistic liquidity provision. 16 For example, during the Flash Crash,

several major HFT firms (e.g., Tradebot Systems and Tradeworx) shut down their systems

completely to protect themselves, which led to the evaporation of liquidity when it was needed

most.

These two attributes of HFT–its opportunistic nature and its ability to shut down at any

time–differ fundamentally from the nature of the dealer’s liquidity provision. Upon close

inspection of the data surrounding the Flash Crash, Easley, Prado, and O’Hara (2011) conclude

16
High-frequency traders often generate profits by trading alongside larger players rather than by taking the
opposite side of the trade. Hendershott, Jones, and Menkveld (2011) find that algorithmic trading generally
improves liquidity and enhances the informativeness of quotes. Hence, the role of high-frequency traders in liquidity
provision may differ depending on market circumstances. Perhaps they are good liquidity providers in normal
markets, but poor liquidity providers during financial turmoil.

29
that the Flash Crash may indeed be viewed as a liquidity event arising from “structural features

of the new high frequency world of trading.”

In this section, we analyze the effect of the Rule 4613 (c) amendment on market liquidity.

We estimate regression models (6) and (7) for NASDAQ stocks using data before and after the

rule change event. Because the rule change applied only to the NASDAQ Stock Market, we use

100 days from June 18, 2007 to November 6, 2007 as the pre-event period and 100 days from

November 8, 2007 to April 3, 2008 as the post-event period. We construct a control sample of

NYSE stocks using the same matching method described in Section 4.1. We use stock attributes

in the pre-event period for matching. The sample contains 770 matching pairs of NASDAQ and

NYSE stocks.

Panel A in Table 8 shows the results using UELs estimated from regression model (6)

and Panel B shows the results using UELs estimated from regression model (7). The first three

columns in each panel show that for NASDAQ stocks, the mean UELs in the post-event period

are significantly greater in absolute value than the corresponding figures in the pre-event period.

In contrast, the next three columns show that the mean UELs in the post-event period are not

significantly different from those in the pre-event period for the control sample of NYSE stocks.

The last column in Table 8 shows that the changes in the mean UELs between the pre- and post-

event periods for NASDAQ stocks are significantly greater in absolute value than the

corresponding figures for the NYSE control sample.

To determine whether the results in Table 8 are driven by concurrent changes in stock

attributes, we estimate regression models (9) and (10) using only the NASDAQ sample. The

results (see Table 9) show that the estimates of π1 and λ0 are positive and significant when we

30
measure liquidity by the quoted spread, effective spread, or price impact, and negative and

significant when we measure liquidity by the depth or market quality index.

As a further robustness check, we also estimate regression model (11) using the

combined sample of NASDAQ and NYSE stocks. We report the estimates of ω1, ω2, and ω3 in

the last three columns of Table 9. The estimates of ω1 are positive and significant when we

measure liquidity by the quoted spread, effective spread, or price impact, and negative and

significant when we measure liquidity by the depth or market quality index. These results are

consistent with the results in Table 8 that the changes in the mean UEL between the pre- and

post-event periods for NASDAQ stocks are greater than the corresponding figures for the NYSE

control sample, indicating that the SEC rule change had a greater impact on NASDAQ stocks.

The estimates of ω2 are not significantly different from zero, indicating that uncertainty has a

similar liquidity effect on the NYSE and NASDAQ before the rule change. The estimates of ω3

are not significantly different from zero, indicating that the rule change does not have a material

effect on how uncertainty affects liquidity on the NYSE. Finally, note that the estimates of π1, λ0,

and ω1 are close to the corresponding values of ∆ (Post – Pre) in Table 8, indicating that changes

in UEL between the pre- and post-event periods are largely attributable to the relaxation of

NASDAQ dealers’ affirmative obligation rather than to concurrent changes in stock attributes.

4.4. Implementation of the designated market maker (DMM) system on the NYSE

The NYSE implemented a new market model which emphasizes speed, technology, and

less human intermediation from October 27, 2008 through November 13, 2008. 17 The new model

replaced specialists with designated market makers (DMMs) who have lesser obligations than

17
http://www.nyse.com/attachment/NewMarketModelRollout.xls

31
their predecessors. 18 In September 2010, SEC Chairman Mary Schapiro appealed to the agency

to investigate whether the loss of old specialist obligations contributed to the Flash Crash. In this

section we test whether the implementation of the new market model has changed the effect of

uncertainty on the liquidity of NYSE stocks.

The pre-event period is 100 days from June 5, 2008 to October 24, 2008 and the post-

event period is 100 days from November 14, 2008 to April 9, 2009. Because the DMM system

applied only to the NYSE, we use NASDAQ stocks as a control sample. We construct a control

sample of NASDAQ stocks using the same matching method described in Section 4.1. The final

sample contains 779 matching pairs of NYSE and NASDAQ stocks.

Panel A in Table 10 shows the results using UELs estimated from regression model (6)

and Panel B shows the results using UELs estimated from regression model (7). The first three

columns in each panel show that for NYSE stocks, the mean UELs in the post-event period are

significantly greater in absolute value than the corresponding figures in the pre-event period

across all five liquidity measures, indicating that the implementation of the DMM system

increased the effect of market uncertainty on liquidity. In contrast, the mean UELs in the post-

DMM period are not significantly different from those in the pre-DMM period for the control

sample of NASDAQ stocks. The last column in Table 10 shows that the changes in the mean

UELs between the pre- and post-event periods for NYSE stocks are significantly greater in

absolute value than the corresponding figures for the NASDAQ control sample. Hence, after

18
The NYSE moved to the new market model to offer its customers the speed of electronic trading while
preserving the certain benefits of having human-moderated trading. The new model offers high-tech for fast,
automated and anonymous execution and high-touch for discovering and improving prices. The model adapts the
NYSE’s trading to reflect the pace and patterns of fully electronic markets. Among other things, the new model
includes: (1) enhanced reserve capabilities for all market participants allowing displayed and non-displayed
(reserve) trading interest; (2) the creation of a designated market maker (DMM) and the phasing out of the specialist
role; (3) changes to priority and parity rules that reward displayed liquidity; (4) d-Quotes interaction with non-
marketable IOC to provide additional opportunities for execution; (5) inclusion of e-Quotes in NYSE OpenBook;
and (6) the elimination Convert and Parity Orders (“CAP”) as valid order types on the NYSE.

32
controlling for market-wide changes, the introduction of the DMM system led to a significant

increase in the uncertainty elasticity of liquidity on the NYSE.

To determine whether the results in Table 10 are driven by concurrent changes in stock

attributes, we estimate regression models (9) and (10) using only the NYSE sample. The results

(see Table 11) show that the estimates of π1 and λ0 are positive and significant when LM in

regression models (6) and (7) is the quoted spread, effective spread, or price impact, and negative

and significant when LM is the depth or market quality index. As a further robustness check, we

also estimate the following regression model using the pooled sample of NYSE and NASDAQ

stocks:

UELi = ω0 + ω1POST*NYSEi + ω2NYSE + ω3POST + ω4PRICEi

+ ω5VOLi + ω6MVEi + ω7VOLAi + ω8PMINi + ε3i, (12)

where NYSE is an indicator variable which equals one for the NYSE stocks and zero for the

NASDAQ stocks. All other variables are the same as previously defined. We report the estimates

of ω1, ω2, and ω3 in the last three columns of Table 11.

The results show that the estimates of ω1 are positive and significant when LM in

regression models (6) and (7) is the quoted spread, effective spread, or price impact, and negative

and significant when LM is the depth or market quality index. These results are consistent with

the results in Table 10 that the changes in the mean UEL between the pre- and post-event periods

for NYSE stocks are greater than the corresponding figures for the NASDAQ control sample.

Hence, after controlling for market-wide changes, the introduction of the DMM system led to a

significant increase in the uncertainty elasticity of liquidity on the NYSE. The results show that

the estimates of ω2 are negative and significant when LM in regression models (6) and (7) is the

33
quoted spread, effective spread, or price impact, and positive and significant when LM is the

depth or market quality index. These results are consistent with those in Table 10 that NYSE

stocks have, on average, smaller values of UEL than NASDAQ stocks during the pre-event

period. The estimates of ω3 are not significantly different from zero, indicating that the

implementation of the DMM system on the NYSE does not have a material effect on the UELs

of NASDAQ stocks.

In contrast to the OHR and tick size changes, we find that the mean quoted and effective

spreads did not change significantly after the amendment of NASDAQ rule 4613 (c) in 2007 and

the implementation of the designated market maker system on the NYSE in 2008. 19 These results

indicate that there is no counterbalancing change in the level of liquidity associated with these

rule changes.

5. Summary and concluding remarks

Prior research shows that the liquidity of individual assets tends to move together and

looks for probable causes of such co-movements. The supply-side theory of liquidity co-

movements posits that liquidity co-variation arises because liquidity providers share common

capital and information and therefore tend to react similarly to outside shocks. The demand-side

theory suggests that liquidity commonality arises from the trading behavior of liquidity

demanders. In the present study, we shed additional light on the cause of liquidity commonality

by analyzing the relation between market uncertainty and liquidity. We show that uncertainty

exerts a market-wide impact on liquidity which, in turn, gives rise to co-movements in individual

19
The results are available from the authors upon request.

34
asset liquidity. We also find evidence that uncertainty exerts a greater impact on liquidity when

public traders play a greater role and market makers play a reduced role in liquidity provision.

We find it intriguing that the liquidity of an individual asset is related not only to its own

risk but also strongly to the level of uncertainty in the market as a whole. One possible

interpretation of this result is that our individual asset risk measure (return volatility) is an

incomplete measure of risk and VIX captures a kind of risk that is not captured by return

volatility. To the extent that adverse selection and inventory holding costs increase with risk, and

that VIX captures the portion of risk not captured by return volatility, the liquidity of assets is

likely to be related to VIX.

Alternatively, liquidity providers may be sensitive to market volatility and thus at least

part of the relation between VIX and liquidity may be driven by liquidity providers’ reactions to

market volatility. As pointed out earlier, liquidity providers may adjust their positions uniformly

across stocks according to readily available VIX before they make security-specific adjustments

based on security-specific risks. An interesting area for future research is to determine whether

liquidity providers’ reaction to market uncertainty reflects their rational, optimizing behaviors or

irrational overreactions to market volatility. For example, an interesting case to examine is how

the liquidity of individual assets changes when VIX and individual asset risks move in opposite

directions.

Having established strong evidence of the uncertainty-induced liquidity commonality,

another fruitful area of future research would be to find out whether stock returns are related to

UEL. Prior research shows that the return premium is related to commonality in liquidity with

the market liquidity, return sensitivity to market liquidity, and liquidity sensitivity to market

returns. To the extent that the market-volatility induced liquidity risk cannot be diversified away,

35
such systematic liquidity risk would also be built into the required return of investors. It would

be interesting to find out whether stocks with greater UELs command higher return premiums

after controlling for other liquidity-related risk factors documented in prior research. Indeed, our

results show that the uncertainty elasticity of liquidity is greater than the liquidity commonality

with the market liquidity, suggesting that the risk premium associated with the former may be

even greater than the risk premium associated with the latter.

36
References

Acharya, V.V., Pedersen, L.H., 2005. Asset pricing with liquidity risk. Journal of Financial
Economics 77, 375-410.

Adrian, T., Shin, H.S., 2010. Liquidity and leverage, Journal of Financial Intermediation 19,
418-437.

Bao, J., Pan, J., Wang, J., 2008. Liquidity of corporate bonds. Available at SSRN:
http://ssrn.com/abstract=1106852 or http://dx.doi.org/10.2139/ssrn.1106852

Barclay, M., Christie, W., Harris, J., Kandel, E., Schultz, P., 1999. The effects of market reform on
the trading costs and depths of NASDAQ stocks. Journal of Finance 54, 1-34.

Bekaert, G., Hoerova, M., Lo Duca, M., 2011. Risk, uncertainty and monetary policy. Available
at SSRN: http://ssrn.com/abstract=1561171 or http://dx.doi.org/10.2139/ssrn.1561171

Benston, G., Hagerman, R., 1974. Determinants of bid-asked spreads in the over-the-counter
market. Journal of Financial Economics 1, 353-364.

Bessembinder, H., 2003. Issues in assessing trade execution costs. Journal of Financial Markets 6,
233–257.

Brunnermeier, M.K., Pedersen, L.H., 2009. Market liquidity and funding liquidity. Review of
Financial Studies 22, 2201-2238.

Chordia, T., Roll, R., Subrahmanyam, A., 2000. Commonality in liquidity, Journal of Financial
Economics 56, 3-28.

Chordia, T., Roll, R., Subrahmanyam, A., 2001. Market liquidity and trading activity, Journal of
Finance 56, 501-530.

Chordia, T., Roll, R., Subrahmanyam, A., 2008. Liquidity and market efficiency. Journal of
Financial Economics 87, 249-268.

Chordia, T., Sarkar, A., Subrahmanyam, A., 2005. An empirical analysis of stock and bond
market liquidity. Review of Financial Studies 18, 85-129.

Coughenour J., Saad, M., 2004. Common market makers and commonality in liquidity, Journal
of Financial Economics 73, 37-69.

Easley, D., Prado. M. L., O'Hara, M., 2011. The microstructure of the ‘flash crash’: flow toxicity,
liquidity crashes and the probability of informed trading. Journal of Portfolio
Management 37, 118–128.

37
Graham, J.R., Harvey, C.R., 2010. The equity risk premium in 2010. Available at SSRN:
http://ssrn.com/abstract=1654026 or http://dx.doi.org/10.2139/ssrn.1654026

Glosten, L., Milgrom, P., 1985. Bid, ask and transaction prices in a specialist market with
heterogeneously informed traders. Journal of Financial Economics 14, 71-100.

Hameed, A., Kang, W., Viswanathan, S., 2010. Stock market declines and liquidity. Journal of
Finance 65, 257–293.

Hasbrouck, J., Seppi, D.J., 2001. Common factors in prices, order flows and liquidity. Journal of
Financial Economics 59, 383–411.

Hendershott, T., Jones, C., Menkveld, A., 2011. Does algorithmic trading improve liquidity?
Journal of Finance 66, 1–33.

Ho, T., Stoll, H.R., 1981. Optimal dealer pricing under transactions and return uncertainty.
Journal of Financial Economics 9, 47–73.

Huberman, G., Halka, D., 2001. Systematic liquidity. Journal of Financial Research 24, 161-178.

Kamara, A., Lou, X., Sadka, R., 2008. The divergence of liquidity commonality in the cross-
section of stocks. Journal of Financial Economics 89, 444–466.

Karolyi, G. A., Lee, K-H., Van Dijk, M. A., 2012. Understanding commonality in liquidity
around the world. Journal of Financial Economics, forthcoming.

Koch, A., Ruenzi, S., Starks, L., 2010. Commonality in liquidity: A demand-side explanation.
Working paper, University of Texas, Austin.

Korajczyk, R.A., Sadka, R., 2008. Pricing the commonality across alternative measures of
liquidity. Journal of Financial Economics 87, 45–72.

Lee, K-H., 2011. The world price of liquidity risk. Journal of Financial Economics 99, 136-161.

Lee, C.M.C., Ready, M.J. 1991. Inferring trade direction from intraday data. Journal of Finance
46, 733–746.

Longstaff, F., Pan, J., Pedersen, L., Singleton, K,. 2010. How sovereign is sovereign credit risk?
American Economic Journal: Macroeconomics 3, 75–103.

Nagel, S., 2012. Evaporating liquidity. Review of Financial Studies 25, 2005-2039.

Pan, J., Singleton, K., 2008. Default and recovery implicit in the term structure of sovereign CDS
spreads. Journal of Finance 63, 2345-2384.

38
Pastor, L., Stambaugh, R.F., 2003. Liquidity risk and expected stock returns. Journal of Political
Economy 111, 642-685.

Sadka, R., 2010. Liquidity risk and the cross-section of hedge-fund returns. Journal of Financial
Economics 98, 54-71.

Stoll, H., 2000. Friction. Journal of Finance 55, 1479-1514.

39
Table 1
Descriptive statistics
This table shows descriptive statistics of the variables for our study sample of 1,945 NYSE stocks and
1,743 NASDAQ stocks. We measure the quoted and effective spreads, depth, price impact, and market quality index
using following formulas: Quoted spreadi,t = (Aski,t – Bidi,t) / Mi,t; Effective spreadi,t = 2Di,t(Pi,t – Mi,t) / Mi,t; Depthi,t
= Mi,t (Bid sizei,t + Ask sizei,t) / 2; Price impacti,t = Di,t(Mi,t+5 – Mi,t) / Mi,t; Market quality indexi,t = ½ (Bid sizei,t +
Ask sizei,t) / [(Aski,t – Bidi,t) / Mi,t,]; where Aski,t is the national best ask price of stock i at time t, Bidi,t is the
national best bid price of stock i at time t, Mi,t is the quote midpoint ((Aski,t + Bidi,t)/2) of stock i at time t, Mi,t+5 is
the quote midpoint at time t + 5 minutes of stock i, Pi,t is the transaction price of stock i at time t, Bid sizei,t is the
quoted size at the bid of stock i at time t, Ask sizei,t is the quoted size at the ask of stock i at time t, and Di,t is an
indicator variable which equals +1 for customer buy orders and -1 for customer sell orders. We use the time-
weighted quoted spread and trade-weighted effective spread in our study. We measure return volatility by the
standard deviation of five-minute quote midpoint returns and trading volume by the dollar trading volume. We
obtain daily observations of the Chicago Board Options Exchange Market Volatility Index (VIX) for our study
period from Yahoo Finance (http://finance.yahoo.com/q/hp?s=^VIX+Historical+Prices). VIX is quoted in
percentage points and it measures the expected (annualized) movement in the S&P 500 index over the following 30-
day period.
_____________________________________________________________________________________________
Percentile
Standard __________________________________________________
Variable Mean deviation 5 25 50 75 95
_____________________________________________________________________________________________
Panel A: Descriptive statistics for NASDAQ stocks

Quoted spread 0.0060 0.0062 0.0006 0.0017 0.0035 0.0082 0.0189

Effective spread 0.0053 0.0054 0.0007 0.0015 0.0032 0.0074 0.0163
Price impact 0.0033 0.0027 0.0006 0.0013 0.0024 0.0045 0.0091
Depth ($ thousand) 17.39 30.37 3.81 6.38 10.21 16.81 47.76
Market quality index 0.5180 1.0718 0.0048 0.1098 0.2012 0.4174 2.1334
Price 17.18 14.87 2.23 6.54 12.96 23.54 46.44
Volume ($ million) 24.59 212.60 0.17 0.71 2.64 9.81 77.82
Volatility 0.0051 0.0027 0.0025 0.0035 0.0043 0.0058 0.0106
VIX 27.24 13.18 12.07 19.58 23.95 30.24 55.78
_____________________________________________________________________________________________
Panel B: Descriptive statistics for NYSE stocks

Quoted spread 0.0024 0.0022 0.0004 0.0009 0.0016 0.0032 0.0066

Effective spread 0.0020 0.0019 0.0005 0.0008 0.0013 0.0026 0.0059
Price impact 0.0013 0.0010 0.0004 0.0007 0.0010 0.0017 0.0032
Depth ($ thousand) 27.70 26.39 8.68 15.02 21.58 31.27 60.26
Market quality index 0.8173 1.1186 0.1311 0.2773 0.4682 0.8603 2.7123
Price 27.51 18.84 6.53 13.07 21.99 37.84 65.20
Volume ($ million) 60.09 188.06 0.24 1.85 12.22 45.85 267.88
Volatility 0.0037 0.0018 0.0019 0.0026 0.0033 0.0042 0.0064
VIX 27.24 13.18 12.07 19.58 23.95 30.24 55.78
_____________________________________________________________________________________________

40
Table 2
The effect of VIX on various liquidity measures using market and industry liquidity
We estimate the following regression models for each of 1,945 NYSE stocks and 1,743 NASDAQ stocks using daily data from January 3, 2007 through
December 31, 2009 to examine the relation between VIX and various measures of liquidity after controlling for other determinants of liquidity:

log(LMi,t) = αi0 + αi1log(VIXt) + αi2log(VIXt-1) + αi3log(VIXt+1) + αi4log(MKTLMt) + αi5log(MKTLMt-1) + αi6log(MKTLMt+1) + αi7log(INDLMt)

+ αi8log(INDLMt-1) + αi9log(INDLMt+1) + αi10log(PRICEi,t) + αi11log(VOLi,t) + αi12log(VOLAi,t) + αi13log(VOLAi,t-1) + αi14log(VOLAi,t+1)
+αi15MKTRETt + αi16MKTRETt-1 + αi17MKTRETt+1 + αi18TUESDAY + αi19WEDNESDAY + αi20THURSDAY + αi21FRIDAY + ε1i,t, (a)

DLMi,t = βi0 + βi1DVIXt + βi2DVIXt-1 + βi3DVIXt+1 + βi4DMKTLMt + βi5DMKTLMt-1 + βi6DMKTLMt+1 + βi7DINDLMt + βi8DINDLMt-1 + βi9DINDLMt+1
+ βi10DPRICEi,t + βi11DVOLi,t + βi12DVOLAi,t + βi13DVOLAi,t-1 + βi14DVOLAi,t+1 + βi15DMKTRETt + βi16DMKTRETt-1 + βi17DMKTRETt+1
+ βi18TUESDAY + βi19WEDNESDAY + βi20THURSDAY + βi21FRIDAY + ε2i,t, (b)

where LMi,t is one of five liquidity measures (i.e., quoted spread, effective spread, price impact, depth, and market quality index) of stock i on day t; VIXt-1, VIXt, VIXt+1
are the CBOE VIX index on day t-1, t, and t+1; MKTLMt-1, MKTLMt, and MKTLMt+1 are the market liquidity measures (e.g., the average quoted percentage spread
across all stocks) on days t-1, t, and t+1; INDLMt-1, INDLMt, and INDLMt+1 are the industry-wide liquidity measures (e.g., the average quoted percentage spread across
all stocks in the same industry) on days t-1, t, and t+1; PRICEi,t is the price of stock i on day t; VOLi,t is the dollar trading volume of stock i on day t; VOLAi,t is the
standard deviation of returns calculated from five-minute quote midpoints of stock i on day t; MKTRETt-1, MKTRETt, and MKTRETt+1 are the market returns on days t-1,
t, and t+1; TUESDAY, WEDNESDAY, THURSDAY, and FRIDAY are dummy variables for Tuesday, Wednesday, Thursday, and Friday; and ε1i,t and ε2i,t are the error
terms. D denotes the percentage change of the variable from the previous day [e.g., DLMi,t = (LMi,t – LMi,t-1) / LMi,t-1]. Panel A shows the results from regression model
𝑥
(a) and Panel B shows the results from regression model (b). Numbers in parentheses are t-statistics. We report the t-statistics as �𝑦� , where x is the t-value of the
𝑧
average regression coefficient, y is the mean t-value across all stock regressions, and z is the median t-value across all stock regressions. The reliability of the t-statistics
depends on the independence of the residuals across stocks. We test the independence of the residuals from regression models (a) and (b) using the method in Chordia,
Roll, and Subrahmanyam (2000) and Coughenour and Saad (2004). We first sort 1,743 NASDAQ stocks alphabetically using their ticker symbols and assign each stock a
serial number i (i = 1,…., 1,743). We then estimate the following regression models: ε1i+1,t = θ0 + θ1ε1i,t + μ1i,t and ε2i+1,t = δ0 + δ1ε2i,t + μ2i,t (i = 1,…., 1,742); where ε1i,t
(ε1i+1,t), ε2i,t (ε2i+1,t) are residuals for stock i (i+1) from regression models (a) and (b), respectively, and μ1i,t and μ2i,t are disturbance terms. We also calculate the correlation
coefficient between ε1i+1,t and ε1i,t and between ε2i+1,t and ε2i,t. Panel C shows our test results of cross-equation dependence in estimation error.

41
Table 2 (continued)
Panel A: Regression coefficients from model (a)
__________________________________________________________________________________________________________________________________________________________________

Variable VIXt VIXt-1 VIXt+1 ∑VIX MKTLMt MKTLMt-1 MKTLMt+1 ∑MKTLM INDLMt INDLMt-1 INDLMt+1 ∑INDLM VOLAt VOLAt-1 VOLAt+1 ∑VOLA
__________________________________________________________________________________________________________________________________________________________________
Results for NASDAQ stocks
Quoted spread 0.4644** 0.0302** 0.0215** 0.5161** 0.1758** -0.0203* -0.0224* 0.1331** 0.1387** 0.0213* 0.0172 0.1772** 0.1460** 0.0350** 0.0382** 0.2192**
(68.32) (5.10) (3.70) (65.44) (15.02) (-2.01) (-2.08) (5.46) (11.74) (2.25) (1.69) (11.10) (24.98) (5.18) (5.97) (22.39)
(5.95)** (0.38) (0.30) (0.65) (-0.08) (-0.09) (0.94) (0.14) (0.12) (2.41)* (0.56) (0.57)
(5.97) (0.38) (0.32) (0.67) (-0.07) (-0.08) (0.93) (0.12) (0.11) (2.47) (0.55) (0.58)
Effective spread 0.4868** 0.0279** 0.0259** 0.5406** 0.1839** -0.0146 0.0033 0.1726** 0.1456** 0.0275** -0.0060 0.1671** 0.1611** 0.0368** 0.0391** 0.2370**
(54.86) (3.54) (3.14) (68.30) (16.28) (-1.39) (0.31) (7.85) (14.54) (2.90) (-0.61) (7.64) (22.92) (5.61) (6.16) (21.39)
(6.40)** (0.35) (0.30) (0.59) (-0.06) (-0.07) (0.99) (0.18) (0.04) (2.74)** (0.61) (0.60)
(6.39) (0.36) (0.33) (0.55) (-0.08) (-0.06) (0.97) (0.17) (0.04) (2.73) (0.63) (0.61)
Price impact 0.4318** 0.0228* 0.0273* 0.4819** 0.1557** -0.0055 -0.0129 0.1373** 0.1242** 0.0164 0.0159 0.1565** 0.1525** 0.0289** 0.0387** 0.2201**
(38.05) (2.19) (2.23) (43.85) (10.77) (-0.38) (-0.87) (5.80) (9.69) (1.29) (1.28) (6.67) (15.39) (3.06) (4.33) (14.76)
(4.43)** (0.22) (0.22) (0.31) (-0.03) (-0.04) (0.80) (0.13) (0.15) (2.36)* (0.36) (0.53)
(4.43) (0.22) (0.23) (0.35) (-0.02) (-0.03) (0.82) (0.12) (0.13) (2.34) (0.34) (0.51)
Depth ($) -0.4059** -0.0141* -0.0283** -0.4483** 0.2131** -0.0222 -0.0292** 0.1617** 0.1754** 0.0227* 0.0207* 0.2188** -0.1130** -0.0406** -0.0401** -0.1937**
(-51.84) (-2.06) (-4.11) (-59.41) (13.96) (-1.53) (-2.09) (5.59) (17.08) (2.24) (2.08) (10.63) (-21.25) (-6.22) (-5.82) (-20.74)
(-4.39)** (-0.17) (-0.31) (0.67) (-0.07) (-0.10) (0.63) (-0.09) (-0.08) (-2.23)* (-0.66) (-0.64)
(-4.35) (-0.18) (-0.32) (0.69) (-0.09) (-0.12) (0.62) (-0.10) (-0.09) (-2.24) (-0.67) (-0.65)
Market quality -0.4902** -0.0264** -0.0140 -0.5306** 0.2389** -0.0022 -0.0067 0.2300** 0.1933** 0.0092 -0.0023 0.2002** -0.2143** -0.0252** -0.0215** -0.2610**
index (-41.93) (-2.72) (-1.25) (-53.75) (16.08) (-0.16) (-0.49) (8.20) (17.76) (0.87) (-0.22) (9.05) (-22.57) (-3.54) (-2.92) (-21.74)
(-4.26)** (-0.22) (-0.15) (0.75) (-0.04) (-0.05) (0.93) (0.03) (-0.02) (-2.46)* (-0.28) (-0.21)
(-4.27) (-0.23) (-0.14) (0.74) (-0.03) (-0.04) (0.89) (0.03) (-0.01) (-2.49) (-0.30) (-0.21)
__________________________________________________________________________________________________________________________________________________________________
Results for NYSE stocks
Quoted spread 0.4852** 0.0211** 0.0276** 0.5339** 0.1859** -0.0224* -0.0290** 0.1345** 0.1490** 0.0307** 0.0167 0.1964** 0.1589** 0.0301** 0.0337** 0.2227**
(74.32) (3.59) (4.79) (97.61) (14.02) (-2.09) (-2.61) (5.83) (11.15) (3.01) (1.71) (10.74) (26.34) (4.78) (5.07) (24.09)
(5.94)** (0.28) (0.32) (0.81) (-0.08) (-0.09) (1.02) (0.22) (0.10) (3.41)** (0.67) (0.76)
(5.95) (0.25) (0.32) (0.80) (-0.07) (-0.08) (1.05) (0.21) (0.10) (3.45) (0.65) (0.73)
Effective spread 0.5015** 0.0371** 0.0269** 0.5655** 0.2044** -0.0199* -0.0164* 0.1681** 0.1674** 0.0272** 0.0112 0.2058** 0.1776** 0.0283** 0.0348** 0.2407**
(73.49) (6.09) (4.26) (101.76) (14.87) (-2.38) (-2.05) (8.87) (12.96) (3.53) (1.50) (11.14) (28.07) (4.49) (5.52) (26.31)
(5.73)** (0.45) (0.32) (1.11) (-0.12) (-0.09) (0.93) (0.13) (0.08) (3.00)** (0.42) (0.53)
(5.71) (0.46) (0.32) (1.13) (-0.11) (-0.08) (0.96) (0.14) (0.07) (3.05) (0.41) (0.54)
Price impact 0.4601** 0.0221* 0.0160 0.4982** 0.1784** -0.0244* -0.0168 0.1372** 0.1395** 0.0084 0.0224* 0.1703** 0.1611** 0.0043 0.0411** 0.2065**
(39.66) (2.17) (1.48) (52.11) (12.43) (-2.04) (-1.44) (7.42) (11.84) (0.85) (2.25) (9.71) (19.90) (1.58) (5.08) (17.35)
(4.48)** (0.22) (0.13) (0.49) (-0.06) (-0.03) (0.73) (0.03) (0.13) (2.94)** (0.07) (0.68)
(4.51) (0.21) (0.15) (0.51) (-0.05) (-0.04) (0.78) (0.02) (0.12) (2.96) (0.06) (0.69)
Depth ($) -0.4220** -0.0192** -0.0229** -0.4641** 0.1996** -0.0205 -0.0303* 0.1488** 0.1740** 0.0319** 0.0289** 0.2348** -0.1188** -0.0225** -0.0342** -0.1755**
(-66.92) (-3.49) (-4.48) (-73.43) (12.38) (-1.61) (-2.29) (5.49) (14.58) (3.04) (2.61) (9.78) (-22.27) (-4.37) (-5.72) (-23.12)
(-4.72)** (-0.21) (-0.25) (0.69) (-0.07) (-0.10) (0.85) (0.14) (0.12) (-2.48)* (-0.43) (-0.78)
(-4.72) (-0.20) (-0.26) (0.74) (-0.07) (-0.09) (0.90) (0.13) (0.11) (-2.42) (-0.40) (-0.76)
Market quality -0.5054** -0.0211** -0.0112 -0.5377** 0.2548** -0.0196 -0.0174 0.2178** 0.2037** 0.0168 0.0173 0.2378** -0.1974** -0.0351** -0.0423** -0.2748**
index (-58.96) (-2.63) (-1.37) (-57.69) (17.10) (-1.52) (-1.45) (9.24) (19.05) (1.62) (1.82) (12.03) (-21.92) (-4.18) (-4.73) (-19.89)
(-4.45)** (-0.18) (-0.08) (0.87) (-0.06) (-0.05) (0.94) (0.07) (0.08) (-2.88)** (-0.50) (-0.58)
(-4.46) (-0.18) (-0.09) (0.90) (-0.06) (-0.04) (0.95) (0.06) (0.07) (-2.90) (-0.52) (-0.56)
________________________________________________________________________________________________________________________________________________
**Significant at the 1% level.
*Significant at the 5% level.

42
Table 2 (continued)
Panel B: Regression coefficients from model (b)
__________________________________________________________________________________________________________________________________________________________________

Variable DVIXt DVIXt-1 DVIXt+1 ∑DVIX DMKTLMt DMKTLMt-1 DMKTLMt+1 ∑DMKTLM DINDLMt DINDLMt-1 DINDLMt+1 ∑DINDLM DVOLAt DVOLAt-1 DVOLAt+1 ∑DVOLA
__________________________________________________________________________________________________________________________________________________________________
Results for NASDAQ stocks
Quoted spread 0.4307** 0.0297** 0.0263** 0.4867** 0.1584** -0.0269** -0.0094 0.1221** 0.1226** 0.0286** 0.0077 0.1589** 0.1641** 0.0092** 0.0084** 0.1817**
(95.92) (6.90) (6.98) (71.87) (14.37) (-3.40) (-1.21) (7.33) (15.26) (3.77) (1.06) (9.97) (28.87) (3.58) (3.37) (25.51)
(6.99)** (0.44) (0.42) (0.62) (-0.09) (-0.05) (0.63) (0.09) (0.05) (3.24)** (0.22) (0.15)
(6.96) (0.44) (0.23) (0.64) (-0.10) (-0.05) (0.60) (0.08) (0.04) (3.25) (0.20) (0.17)
Effective spread 0.4485** 0.0279** 0.0277** 0.5041** 0.1621** -0.0209* 0.0073 0.1485** 0.1371** 0.0332** -0.0049 0.1654** 0.1723** 0.0110** 0.0287** 0.2120**
(79.77) (4.75) (5.14) (54.04) (16.39) (-2.28) (0.80) (8.11) (16.88) (4.27) (-0.61) (10.02) (29.74) (3.77) (5.09) (27.33)
(6.10)** (0.36) (0.33) (0.57) (-0.05) (0.02) (0.70) (0.15) (-0.04) (2.90)** (0.25) (0.46)
(6.12) (0.35) (0.34) (0.57) (-0.06) (0.02) (0.69) (0.15) (-0.03) (2.91) (0.23) (0.45)
Price impact 0.4348** 0.0210** 0.0120 0.4678** 0.1349** -0.0120 0.0018 0.1247** 0.1238** 0.0157 -0.0003 0.1392** 0.1674** 0.0203** 0.0323** 0.2200**
(53.81) (2.69) (1.28) (25.00) (12.84) (-1.20) (0.18) (5.47) (13.65) (1.67) (-0.04) (7.01) (21.80) (3.72) (5.91) (16.19)
(4.02)** (0.14) (0.09) (0.37) (-0.03) (0.01) (0.58) (0.09) (-0.01) (2.48)* (0.31) (0.44)
(4.04) (0.13) (0.08) (0.39) (-0.03) (0.01) (0.57) (0.10) (-0.01) (2.51) (0.28) (0.43)
Depth ($) -0.3881** -0.0313** -0.0135** -0.4329** 0.1815** -0.0131 -0.0078 0.1606** 0.1722** 0.0112 0.0093 0.1927** -0.1256** -0.0181** -0.0204** -0.1641**
(-74.64) (-6.35) (-2.90) (-51.53) (13.41) (-1.12) (-0.68) (6.77) (18.60) (1.32) (1.09) (11.22) (-21.37) (3.20) (-3.79) (-20.55)
(-5.02)** (-0.33) (-0.16) (0.52) (-0.03) (-0.02) (0.62) (0.05) (0.03) (-2.46)* (-0.36) (-0.37)
(-5.02) (-0.33) (-0.17) (0.50) (-0.03) (-0.02) (0.59) (0.04) (0.04) (-2.43) (-0.36) (-0.38)
Market quality -0.4396** -0.0350** -0.0089 -0.4835** 0.2364** -0.0059 -0.0053 0.2252** 0.1941** 0.0101 -0.0004 0.2038** -0.1923** -0.0271** -0.0300** -0.2494**
index (-67.39) (-5.45) (-1.38) (-46.38) (18.84) (-0.55) (-0.51) (9.84) (22.40) (1.15) (-0.05) (11.45) (-25.73) (3.71) (-4.31) (-22.47)
(-4.75)** (-0.33) (-0.09) (0.73) (-0.02) (-0.03) (0.92) (0.05) (-0.01) (-2.85)** (-0.36) (-0.43)
(-4.73) (-0.33) (-0.09) (0.69) (-0.02) (-0.02) (0.89) (0.04) (-0.01) (-2.84) (-0.34) (-0.43)
__________________________________________________________________________________________________________________________________________________________________
Results for NYSE stocks
Quoted spread 0.4585** 0.0296** 0.0223** 0.5104** 0.1868** -0.0298** -0.0281** 0.1289** 0.1298** 0.0294** 0.0182* 0.1774** 0.1383** 0.0269** 0.0254** 0.1906**
(104.69) (6.43) (5.47) (70.55) (15.46) (-3.29) (-3.11) (6.61) (11.69) (3.74) (2.29) (9.79) (25.23) (4.99) (4.61) (23.25)
(6.68)** (0.38) (0.34) (0.73) (-0.11) (-0.10) (0.67) (0.13) (0.09) (3.47)** (0.63) (0.61)
(6.69) (0.37) (0.32) (0.74) (-0.10) (-0.09) (0.68) (0.12) (0.08) (3.45) (0.65) (0.59)
Effective spread 0.4753** 0.0217** 0.0389** 0.5359** 0.2133** -0.0357** -0.0066 0.1710** 0.1493** 0.0392** 0.0022 0.1907** 0.1825** 0.0243** 0.0109** 0.2177**
(101.11) (4.45) (8.31) (67.41) (18.76) (-5.24) (-0.95) (10.68) (14.26) (4.17) (0.34) (11.63) (31.11) (4.17) (3.03) (26.39)
(6.44)** (0.30) (0.39) (1.00) (-0.14) (-0.03) (0.74) (0.17) (0.03) (3.42)** (0.47) (0.19)
(6.42) (0.29) (0.40) (0.98) (-0.12) (-0.04) (0.77) (0.16) (0.04) (3.40) (0.47) (0.18)
Price impact 0.4426** 0.0253** 0.0105 0.4784** 0.1476** -0.0147 0.0088 0.1417** 0.1336** 0.0239** -0.0061 0.1514** 0.1542** 0.0062 0.0205** 0.1809**
(54.78) (3.17) (1.35) (25.97) (14.24) (-1.47) (1.15) (6.26) (18.27) (3.76) (-0.98) (10.32) (19.34) (1.02) (3.44) (15.32)
(4.64)** (0.25) (0.11) (0.39) (-0.11) (0.02) (0.63) (0.09) (-0.04) (2.26)* (0.09) (0.29)
(4.65) (0.26) (0.12) (0.39) (-0.10) (0.02) (0.62) (0.11) (-0.05) (2.27) (0.08) (0.29)
Depth ($) -0.3997** -0.0218** -0.0212** -0.4427** 0.2109** -0.0203 -0.0092 0.1814** 0.1878** 0.0279** 0.0112 0.2269** -0.1066** -0.0249** -0.0264** -0.1579**
(-90.98) (-5.26) (-5.67) (-61.30) (14.40) (-1.50) (-0.77) (7.09) (17.71) (2.94) (1.21) (11.26) (-19.83) (4.68) (-5.16) (-18.78)
(-5.49)** (-0.26) (-0.21) (0.52) (-0.05) (-0.01) (0.46) (0.05) (0.03) (-2.29)* (-0.50) (-0.52)
(-5.49) (-0.24) (-0.22) (0.53) (-0.05) (-0.02) (0.46) (0.03) (0.02) (-2.25) (-0.52) (-0.53)
Market quality -0.4742** -0.0236** -0.0031 -0.5009** 0.2557** -0.0038 -0.0145 0.2374** 0.1893** 0.0183* 0.0061 0.2137** -0.1633** -0.0267** -0.0336** -0.2236**
index (-85.62) (-4.23) (-0.62) (-54.63) (19.52) (-0.41) (-1.26) (10.76) (19.92) (2.33) (0.75) (12.05) (-21.99) (-4.05) (-4.87) (-20.66)
(-5.08)** (-0.22) (0.02) (0.76) (-0.02) (-0.04) (0.55) (0.06) (0.03) (-2.95)** (-0.49) (-0.60)
(-5.07) (-0.24) (0.03) (0.80) (-0.02) (-0.05) (0.52) (0.07) (0.04) (-2.96) (-0.48) (-0.59)
__________________________________________________________________________________________________________________________________________________________________
**Significant at the 1% level.
*Significant at the 5% level.

43
Table 2 (continued)
Panel C: Tests of cross-equation dependence in estimation error
______________________________________________________________________________________________________________________

Results from regression model (a) Results from regression model (b)
_________________________________________ ______________________________________________
Average Average |t| > 1.96 Average Average |t| > 1.96
correlation t-statistic (%) correlation t-statistic (%)
______________________________________________________________________________________________________________________
Results for NASDAQ stocks
Quoted spread 0.0099 0.26 10.98 0.0107 0.41 10.04
Effective spread 0.0141 0.37 10.72 0.0129 0.36 9.13
Price impact 0.0073 0.18 6.48 0.0042 0.12 4.74
Depth 0.0057 0.16 7.27 0.0048 0.14 6.07
Market quality index 0.0085 0.24 5.37 0.0077 0.21 7.02
______________________________________________________________________________________________________________________
Results for NYSE stocks
Quoted spread 0.0164 0.37 10.45 0.0178 0.44 10.33
Effective spread 0.0134 0.28 10.91 0.0128 0.24 10.07
Price impact 0.0035 0.11 6.12 0.0015 0.08 6.20
Depth 0.0126 0.49 5.16 0.0142 0.39 5.73
Market quality index 0.0113 0.43 6.01 0.0133 0.31 5.14
______________________________________________________________________________________________________________________

44
Table 3
Effects of stock attributes on the uncertainty elasticity of liquidity

This table reports the results of the following regression model using combined sample of 1,945
NYSE stocks and 1,743 NASDAQ stocks:

UELi = π0 + π1NASDAQ + π2PRICEi + π3VOLi + π4MVEi + π5VOLAi + π6NAFi + π7INSTi

+ π8MVTOBVi + π9PMINi + π10SP500 + εi,

where UELi is the uncertainty elasticity of liquidity for stock i computed as the sum of contemporaneous,
lag, and lead VIX regression coefficients from model (6) and (7); NASDAQ is an indicator variable which
equals one for NASDAQ stocks and zero for NYSE stocks; PRICEi is the share price of stock i; VOLi is the
dollar trading volume of stock i; MVEi is the market capitalization of stock i; VOLAi is the stock i’s return
volatility; NAFi is the number of analysts following stock i; INSTi is the percentage of institutional holding
of stock i; MVTOBVi is stock’s i ratio of market value of equity to book value of equity; PMINi is the
proportion of spread quotes that are equal to the minimum price variation for stock i; the SP500 is an
indicator variable which equals one for S&P 500 stocks and zero otherwise; and εi is the error term. We use
the mean values of PRICEi, VOLi, MVEi,VOLAi, NAFi, INSTi, MVTOBVi, and PMINi for each stock during
the entire study period (i.e., January 3, 2007 through December 31, 2009) in the regression. Panel A shows
the results when UEL is the estimate of (α1+ α2 + α3) in regression model (6) and Panel B shows the results
when UEL is the estimate of (β1 + β2 + β3) in regression model (7). Numbers in parentheses are t-statistics.
___________________________________________________________________________________________________________
Panel A: Panel B:
Results based on the estimate of (α1+ α2 + α3) in Results based on the estimate of (β1 + β2 + β3) in
regression model (6) regression model (7)
_____________________________________________________ _________________________________________________

Quoted Effective Price Depth

Market Quoted Effective Price Depth Market
spread spread impact quality spread spread impact quality
index index
___________________________________________________________________________________________________________
Intercept 0.5744** 1.0275** 0.4615* -0.4981* -0.4369* 1.7661** 1.6345** 1.5323* -0.3302** -0.3049**
(2.72) (4.29) (2.31) (-2.02) (-2.12) (6.97) (5.25) (2.24) (-2.71) (-2.85)
NASDAQ -0.0040 -0.0117 -0.0132 0.0125 0.0142 -0.0115 -0.0082 -0.0119 0.0098 0.0178
(-0.38) (-1.04) (-0.75) (1.08) (0.87) (-0.96) (-0.56) (-0.37) (0.74) (1.05)
PRICE 0.2158** 0.1300** 0.1062** -0.0576** -0.0685** 0.1398** 0.1091** 0.1402** -0.0494** -0.0895**
(15.77) (8.69) (4.56) (-3.75) (-3.16) (8.84) (5.61) (3.28) (-2.81) (-3.98)
VOL 0.0577** 0.0523** 0.0480** -0.0333** -0.0349** 0.0385** 0.0485** 0.0796** -0.0338** -0.0385**
(7.50) (6.23) (3.67) (-3.86) (-2.86) (4.34) (4.45) (3.32) (-3.42) (-3.05)
MVE -0.0450** -0.0372** -0.0347** 0.0307** 0.0369** -0.0460** -0.0422** -0.0749** 0.0297** 0.0452**
(-5.36) (-4.06) (-3.83) (3.26) (2.77) (-4.74) (-3.54) (-2.86) (2.75) (3.28)
VOLA 0.1783** 0.1414** 0.2195** -0.1244** 0.1776** 0.1606** 0.2302** 0.1487** -0.1516** -0.1824**
(4.52) (3.28) (3.27) (-2.81) (-2.84) (3.52) (4.11) (3.32) (-3.78) (-2.97)
NAF -0.0173** -0.0157** -0.0231** 0.0152** 0.0314** -0.0213** -0.0224** -0.0211** 0.0178** 0.0270**
(-3.66) (-3.04) (-2.87) (2.85) (4.18) (-3.79) (-3.34) (-3.54) (2.93) (3.47)
INST -0.1311** -0.3575** -0.4070** 0.1098** 0.1473** -0.1465** -0.1902** -0.2499** 0.1049** 0.1527**
(-4.56) (-8.20) (-8.32) (3.40) (3.23) (-4.41) (-4.66) (-3.90) (2.84) (3.87)
MVTOBV 0.0267** 0.0227** 0.0283** -0.0202** -0.0244** 0.0256** 0.0271** 0.0251** -0.0262** -0.0399**
(4.76) (3.71) (2.97) (-3.21) (-2.74) (3.95) (3.41) (3.15) (-3.64) (-4.33)
PMIN -0.0310** -0.0322** -0.0435** 0.0427** 0.0559** -0.0319** -0.0477** -0.0359** 0.0338** 0.0508**
(-3.54) (-3.36) (-2.91) (4.33) (4.02) (-3.45) (-3.83) (-3.28) (3.03) (3.53)
SP500 0.0180 0.0211 0.0169 -0.0145 -0.0194 0.0086 0.0176 0.0199 -0.0246 -0.0155
(1.09) (1.07) (0.60) (-0.78) (-0.74) (0.45) (0.75) (0.58) (-1.16) (-0.57)

Adj. R2 0.15 0.11 0.07 0.09 0.07 0.13 0.10 0.06 0.08 0.07
___________________________________________________________________________________________________________
**Significant at the 1% level.
*Significant at the 5% level.

45
Table 4
Effects of the new order handling rules (OHR) on the uncertainty elasticity of liquidity (UEL)

We calculate the composite match score (CMS) of each NASDAQ stock that was subject to the new OHR against each NYSE stock with the same two-
digit NAICS code: CMS = Σ[(YjNASDAQ - YjNYSE)/{(YjNASDAQ + YjNYSE)/2}]2, where Yj represents one of the four stock attributes (share price, dollar volume,
volatility, and market capitalization) in the pre-OHR period, Σ denotes the summation over j = 1 to 4, and superscripts NASDAQ and NYSE represent NASDAQ and
NYSE stocks, respectively. Then, for each NASDAQ stock, we select the NYSE stock with the lowest score. Once we match a NYSE stock with a NASDAQ
stock, we no longer consider that particular NYSE stock for subsequent matches. We use matching pairs of NASDAQ and NYSE stocks for UEL comparisons.
The first two columns of Panel A show the mean UEL [i.e., the estimate of (α1 + α2 + α3)] from regression model (6) in the pre- and post-event periods, and the
third column shows the difference in the mean UEL between the two periods for NASDAQ stocks. Likewise, the first two columns of Panel B show the mean
UEL [i.e., the estimate of (β1+ β2+ β3)] from regression model (7) in the pre- and post-event periods, and the third column shows the difference in the mean UEL
between the two periods for NASDAQ stocks. Columns 4 through 6 in both panels show the corresponding results for the control sample of NYSE stocks. Last
column in both panels show the difference in ∆ (Post – Pre) between NASDAQ and NYSE stocks. Our sample consists of 119 matching pairs of
NASDAQ/NYSE stocks. Numbers in parentheses are t-statistics.
_________________________________________________________________________________________________________________________________
NASDAQ NYSE (Control Stocks) NASDAQ - NYSE
_________________________________________________________________________________________________________________________________
Pre Post ∆ (Post – Pre) Pre Post ∆ (Post – Pre) ∆NASDAQ - ∆NYSE
_________________________________________________________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.0795 0.1965 0.1170** (3.27) 0.0649 0.0784 0.0135 (0.36) 0.1035** (3.64)
Effective spread 0.0967 0.2097 0.1130** (3.70) 0.0672 0.0826 0.0154 (0.43) 0.0976** (3.32)
Price impact 0.0826 0.1659 0.0833** (2.97) 0.0422 0.0567 0.0145 (0.47) 0.0688** (2.86)
Depth -0.1012 -0.1901 -0.0889** (-3.16) -0.0741 -0.0845 -0.0104 (-0.36) -0.0785** (-3.18)
Market quality index -0.1071 -0.1981 -0.0910** (-2.90) -0.0799 -0.0918 -0.0119 (-0.42) -0.0791** (-3.53)
_________________________________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β 1+ β2+ β3) in regression model (7)
Quoted spread 0.0749 0.1909 0.1160** (3.61) 0.0681 0.0758 0.0077 (0.28) 0.1083** (3.50)
Effective spread 0.0882 0.2001 0.1119** (4.19) 0.0695 0.0801 0.0106 (0.35) 0.1013** (3.75)
Price impact 0.0759 0.1760 0.1001** (3.38) 0.0401 0.0552 0.0151 (0.25) 0.0850** (3.16)
Depth -0.0967 -0.1748 -0.0781** (-2.99) -0.0759 -0.0866 -0.0107 (-0.38) -0.0674** (-3.25)
Market quality index -0.1139 -0.2102 -0.0963** (-3.60) -0.0714 -0.0851 -0.0137 (-0.46) -0.0826** (-3.37)
_________________________________________________________________________________________________________________________________
**Significant at the 1% level.

46
Table 5
Regression results for the effects of the implementation of Order Handling Rules (OHR) on the
uncertainty elasticity of liquidity (UEL)

Panel A shows the results based on the estimate of UEL (α1 + α2 + α3) in regression model (6) and
Panel B shows the results based on the estimate of UEL (β1 + β2 + β3) in regression model (7) for 119
matching pairs of NASDAQ/NYSE stocks. The first and second columns in both panels show the
estimates of π1 and λ0 in the following regression models for NASDAQ stocks:

UELi = π0 + π1POST + π2PRICEi + π3VOLi + π4MVEi + π5VOLAi + π6 PMINi + ε1i, (a)

UELipost – UELipre = λ0 + Σλk (Xkipost – Xkipre) + ε2i, (b)

where POST is an indicator variable which equals one for the post-event period and zero otherwise;
PRICEi is the share price of stock i; VOLi is the dollar trading volume of stock i; MVEi is the market
capitalization of stock i; VOLAi is the stock i’s return volatility; PMINi is the proportion of spread quotes
that are equal to the minimum price variation for stock i; the superscripts ‘post’ and ‘pre’ denote,
respectively, the post- and pre-periods; and Xk (k = 1 through 5) represents share price, dollar trading
volume, market capitalization, return volatility, or proportion of spread quotes that are equal to the
minimum price variation. We use the mean values of PRICEi, VOLi, MVEi,VOLAi, and PMINi for each
stock during the pre- and post-OHR periods, respectively, in the regression. The third and fourth columns
show the estimates of ω1 and ω2 in the following regression model for the combined sample of matching
NASDAQ and NYSE stocks:

UELi = ω0 + ω1POST*NASDAQ + ω2NASDAQ + ω3POST + ω4PRICEi + ω5VOLi + ω6MVEi

+ ω7VOLAi + ω8PMINi + ε3i, (c)

where NASDAQ is an indicator variable which equals one for NASDAQ stocks and zero for NYSE
stocks and POST is an indicator variable which equals one for the post-period and zero for the pre-period.
Numbers in parentheses are t-statistics.
_____________________________________________________________________________________

π1 λ0 ω1 ω2 ω3
____________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.1164** (3.87) 0.1158** (3.71) 0.1038** (4.32) 0.0148 (0.63) 0.0125 (0.43)
Effective spread 0.1120** (4.25) 0.1125** (3.61) 0.0960** (4.16) 0.0273 (0.82) 0.0146 (0.40)
Price impact 0.0825** (3.32) 0.0830** (2.84) 0.0696** (3.13) 0.0386 (1.27) 0.0149 (0.56)
Depth -0.0857** (-3.88) -0.0894** (-3.57) -0.0749** (-3.18) -0.0282 (-0.97) -0.0097 (-0.47)
Market quality index -0.0904** (-3.18) -0.0893** (-3.44) -0.0762** (-3.58) -0.0276 (-0.84) -0.0112 (-0.55)
________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β1 + β2 + β3) in regression model (7)
Quoted spread 0.1136** (3.96) 0.1152** (3.62) 0.1069** (4.43) 0.0065 (0.49) 0.0072 (0.39)
Effective spread 0.1110** (3.94) 0.1105** (3.25) 0.1025** (3.87) 0.0166 (0.56) 0.0102 (0.45)
Price impact 0.0956** (3.06) 0.0961** (3.36) 0.0841** (3.53) 0.0331 (1.06) 0.0142 (0.31)
Depth -0.0761** (-3.59) -0.0802** (-3.43) -0.0691** (-3.31) -0.0198 (-0.83) -0.0094 (-0.36)
Market quality index -0.0958** (-3.46) -0.0925** (-3.18) -0.0830** (-3.86) -0.0415 (-1.41) -0.0130 (-0.51)
_____________________________________________________________________________________
**Significant at the 1% level.

47
Table 6
Effects of the tick-size reduction in 1997 on the uncertainty elasticity of liquidity (UEL)
The first two columns of Panel A show the mean UEL [i.e., the estimate of (α1 + α2 + α3)] from regression model (6) in the pre- and post-event periods, and
the third column shows the difference in the mean UEL between the two periods for NASDAQ stocks. Likewise, the first two columns of Panel B show the mean
UEL [i.e., the estimate of (β1 + β2 + β3)] from regression model (7) in the pre- and post-event periods, and the third column shows the difference in the mean UEL
between the two periods for NASDAQ stocks. The fourth and fifth columns in both panels show the estimates of π1 and λ0 in the following regression models for
NASDAQ stocks:

UELi = π0 + π1POST + π2PRICEi + π3VOLi + π4 MVEi + π5VOLAi + π6PMINi + ε1i, (a)

UELipost – UELipre = λ0 + Σλk (Xkipost – Xkipre) + ε2i, (b)

where POST is an indicator variable which equals one for the post-event period and zero otherwise; PRICEi is the share price of stock i; VOLi is the dollar trading
volume of stock i; MVEi is the market capitalization of stock i; VOLAi is the stock i’s return volatility; PMINi is the proportion of spread quotes that are equal to the
minimum price variation for stock i; the superscripts ‘post’ and ‘pre’ denote, respectively, the post- and pre-periods; and Xk (k = 1 through 5) represents share price,
dollar trading volume, market capitalization, return volatility, or proportion of spread quotes that are equal to the minimum price variation. We use the mean values
of PRICEi, VOLi, MVEi,VOLAi, and PMINi for each stock during the pre- and post-event periods, respectively, in the regression. Columns 6 through 10 in both
panels show the corresponding results for NYSE stocks. Our sample consists of 1,587 NASDAQ stocks and 2,123 NYSE stocks. Numbers in parentheses are t-
statistics.
____________________________________________________________________________________________________________________________________
NASDAQ stocks NYSE stocks
_____________________________________________________________ _________________________________________________________________

Pre Post ∆ (Post – Pre) π1 λ0 Pre Post ∆ (Post – Pre) π1 λ0

____________________________________________________________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.0831 0.1742 0.0911** (4.15) 0.0887** (3.58) 0.0881** (3.82) 0.0883 0.1583 0.0700** (5.42) 0.0696** (4.51) 0.0710** (4.46)
Effective spread 0.0935 0.1892 0.0957** (4.47) 0.0944** (3.64) 0.0924** (4.73) 0.0854 0.1602 0.0748** (5.97) 0.0757** (5.43) 0.0729** (4.29)
Price impact 0.0759 0.1530 0.0771** (3.08) 0.0741** (3.23) 0.0752** (3.71) 0.0743 0.1456 0.0713** (3.90) 0.0705** (3.02) 0.0692** (2.87)
Depth -0.0975 -0.1591 -0.0616** (-3.85) -0.0614** (-4.16) -0.0631** (-3.54) -0.1083 -0.1865 -0.0782** (-4.82) -0.0787** (-3.72) -0.0765** (-3.82)
Market quality index -0.1030 -0.1710 -0.0680** (-3.92) -0.0656** (-3.42) -0.0663** (-3.22) -0.1066 -0.1834 -0.0768** (-4.61) -0.0737** (-3.60) -0.0749** (-3.48)
___________________________________________________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β1 + β2 + β3) in regression model (7)
Quoted spread 0.0876 0.1726 0.0850** (3.80) 0.0836** (3.34) 0.0844** (3.83) 0.0943 0.1572 0.0629** (4.71) 0.0615** (3.84) 0.0620** (4.81)
Effective spread 0.0954 0.1820 0.0866** (3.61) 0.0845** (4.49) 0.0856** (3.96) 0.0867 0.1553 0.0686** (5.27) 0.0651** (3.45) 0.0667** (3.91)
Price impact 0.0734 0.1484 0.0750** (3.45) 0.0758** (3.78) 0.0745** (3.90) 0.0750 0.1371 0.0621** (3.91) 0.0632** (3.10) 0.0613** (3.05)
Depth -0.0823 -0.1450 -0.0627** (-4.47) -0.0637** (-4.10) -0.0656** (-3.60) -0.0999 -0.1750 -0.0751** (-4.58) -0.0773** (-4.31) -0.0730** (-4.09)
Market quality index -0.1103 -0.1734 -0.0631** (-3.72) -0.0644** (4.60) -0.0628** (-3.56) -0.1045 -0.1819 -0.0774** (-4.49) -0.0780** (-3.51) -0.0761** (-3.82)
___________________________________________________________________________________________________________________________________________________
**Significant at the 1% level.

48
Table 7
Effects of decimalization in 2001 on the uncertainty elasticity of liquidity (UEL)
The first two columns of Panel A show the mean UEL [i.e., the estimate of (α1 + α2 + α3)] from regression model (6) in the pre- and post-decimalization
periods, and the third column shows the difference in the mean UEL between the two periods for NASDAQ stocks. Likewise, the first two columns of Panel B show
the mean UEL [i.e., the estimate of (β1+ β2+ β3)] from regression model (7) in the pre- and post-decimalization periods, and the third column shows the difference in
the mean UEL between the two periods for NASDAQ stocks. The fourth and fifth columns in both panels show the estimates of π1 and λ0 in the following regression
models for NASDAQ stocks:

UELi = π0 + π1POST + π2PRICEi + π3VOLi + π4MVEi + π5VOLAi + π6PMINi + ε1i, (a)

UELipost – UELipre = λ0 + Σλk (Xkipost – Xkipre) + ε2i, (b)

where POST is an indicator variable which equals one for the post-decimalization period and zero otherwise; PRICEi is the share price of stock i; VOLi is the dollar
trading volume of stock i; MVEi is the market capitalization of stock i; VOLAi is the stock i’s return volatility; PMINi is the proportion of spread quotes that are
equal to the minimum price variation for stock i; the superscripts ‘post’ and ‘pre’ denote, respectively, the post- and pre-periods; and Xk (k = 1 through 5) represents
share price, dollar trading volume, market capitalization, return volatility, or proportion of spread quotes that are equal to the minimum price variation. We use the
mean values of PRICEi, VOLi, MVEi,VOLAi, and PMINi for each stock during the pre- and post-event periods, respectively, in the regression. Columns 6 through 10
in both panels show the corresponding results for NYSE stocks. Our sample contains 1,671 NASDAQ stocks and 1,903 NYSE stocks. Numbers in parentheses are t-
statistics.
___________________________________________________________________________________________________________________________________________________

NASDAQ stocks NYSE stocks

_____________________________________________________________ _______________________________________________________________

Pre Post ∆ (Post – Pre) π1 λ0 Pre Post ∆ (Post – Pre) π1 λ0

___________________________________________________________________________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.2321 0.3063 0.0742** (4.71) 0.0734** (3.51) 0.0720** (3.23) 0.2254 0.2977 0.0723** (4.93) 0.0744** (4.75) 0.0714** (4.68)
Effective spread 0.2422 0.3232 0.0810** (5.23) 0.0790** (3.87) 0.0783** (4.70) 0.2387 0.3118 0.0731** (5.33) 0.0718** (4.19) 0.0737** (4.14)
Price impact 0.2016 0.2690 0.0674** (3.83) 0.0686** (3.23) 0.0652** (3.32) 0.2147 0.2729 0.0582** (3.86) 0.0573** (3.93) 0.0567** (3.23)
Depth -0.2264 -0.2979 -0.0715** (-3.21) -0.0706** (-4.15) -0.0682** (-4.93) -0.2481 -0.3051 -0.0570** (-3.74) -0.0557** (-3.51) -0.0539** (-3.37)
Market quality index -0.2458 -0.3042 -0.0584** (-3.31) -0.0561** (-3.30) -0.0551** (-4.24) -0.2412 -0.2957 -0.0545** (-3.20) -0.0551** (-4.12) -0.0528** (-3.45)
___________________________________________________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β1+ β2+ β3) in regression model (7)
Quoted spread 0.2268 0.2967 0.0699** (3.96) 0.0648** (3.98) 0.0657** (3.62) 0.2325 0.3010 0.0685** (4.19) 0.0672** (3.96) 0.0657** (4.10)
Effective spread 0.2370 0.3030 0.0660** (3.87) 0.0664** (3.68) 0.0632** (3.80) 0.2263 0.3062 0.0799** (4.76) 0.0776** (4.26) 0.0752** (3.50)
Price impact 0.2045 0.2654 0.0609** (3.58) 0.0597** (3.11) 0.0593** (3.20) 0.2071 0.2595 0.0524** (3.73) 0.0502** (3.06) 0.0519** (3.10)
Depth -0.2218 -0.2967 -0.0749** (-3.50) -0.0738** (-3.90) -0.0713** (-3.86) -0.2548 -0.3067 -0.0519** (-3.22) -0.0507** (-3.09) -0.0538** (-3.31)
Market quality index -0.2342 -0.3014 -0.0672** (-3.71) -0.0695** (-4.12) -0.0641** (-4.33) -0.2350 -0.2942 -0.0592** (-3.31) -0.0607** (-3.53) -0.0588** (-3.17)
__________________________________________________________________________________________________________________________________________________
**Significant at the 1% level.

49
Table 8
Effects of the amendment of NASDAQ Rule 4613 (c) on the uncertainty elasticity of liquidity (UEL)

We first calculate the composite match score (CMS) of each NASDAQ stock against each NYSE stock with the same two-digit NAICS code: CMS =
Σ[(YjNASDAQ - YjNYSE)/{(YjNASDAQ + YjNYSE)/2}]2, where Yj represents one of the four stock attributes (price, dollar volume, volatility, and market value of equity) in
the pre-period, Σ denotes the summation over j = 1 to 4, and superscripts NASDAQ and NYSE represent NASDAQ and NYSE stocks, respectively. Then, for each
NASDAQ stock, we select the NYSE stock with the lowest score. Once we match a NYSE stock with a NASDAQ stock, we no longer consider that particular
NYSE stock for subsequent matches. We use 770 matching pairs of NASDAQ and NYSE stocks for UEL comparisons. The first two columns of Panel A show
the mean UEL [i.e., the estimate of (α1 + α2 + α3)] from regression model (6) in the pre- and post-event periods, and the third column shows the difference in the
mean UEL between the two periods for NASDAQ stocks. Likewise, the first two columns of Panel B show the mean UEL [i.e., the estimate of (β1+ β2+ β3)]
from regression model (7) in the pre- and post-event periods, and the third column shows the difference in the mean UEL between the two periods for NASDAQ
stocks. Columns 4 through 6 in both panels show the corresponding results for the control sample of NYSE stocks. Last column in both panels show the
difference in ∆ (Post – Pre) between NASDAQ and NYSE stocks. Numbers in parentheses are t-statistics.
_________________________________________________________________________________________________________________________________
NASDAQ NYSE (Control sample) NASDAQ - NYSE
_________________________________________________________________________________________________________________________________
Pre Post ∆ (Post – Pre) Pre Post ∆ (Post – Pre) ∆NASDAQ - ∆NYSE
_________________________________________________________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.4095 0.5162 0.1067** (5.79) 0.4120 0.4425 0.0305 (1.23) 0.0762** (4.19)
Effective spread 0.4271 0.5292 0.1021** (5.36) 0.4067 0.4353 0.0286 (0.93) 0.0735** (3.78)
Price impact 0.3689 0.4518 0.0829** (4.79) 0.3623 0.3845 0.0222 (0.86) 0.0607** (3.51)
Depth -0.3580 -0.4595 -0.1015** (-4.82) -0.3516 -0.3807 -0.0291 (-1.04) -0.0724** (-3.97)
Market quality index -0.3733 -0.4862 -0.1129** (-5.48) -0.3406 -0.3668 -0.0262 (-0.71) -0.0867** (-3.22)
_________________________________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β 1+ β2+ β3) in regression model (7)
Quoted spread 0.3855 0.4826 0.0971** (5.90) 0.3873 0.4138 0.0265 (0.90) 0.0706** (3.83)
Effective spread 0.4087 0.5188 0.1101** (6.32) 0.3945 0.4179 0.0234 (0.76) 0.0867** (4.36)
Price impact 0.3506 0.4315 0.0809** (4.54) 0.3589 0.3783 0.0194 (0.63) 0.0615** (3.23)
Depth -0.3530 -0.4407 -0.0877** (-4.91) -0.3372 -0.3645 -0.0273 (-1.33) -0.0604** (-3.89)
Market quality index -0.3433 -0.4513 -0.1080** (-5.88) -0.3301 -0.3603 -0.0302 (-1.17) -0.0778** (-3.35)
_________________ ________________________________________________________________________________________________________________
**Significant at the 1% level.

50
Table 9
Regression results for the effects of the amendment of NASDAQ Rule 4613 (c) on the uncertainty
elasticity of liquidity (UEL)

Panel A shows the results based on the estimate of UEL (α1 + α2 + α3) in regression model (6) and
Panel B shows the results based on the estimate of UEL (β1 + β2 + β3) in regression model (7) for 770
matching pairs of NASDAQ and NYSE stocks. The first and second columns in both panels show the
estimates of π1 and λ0 in the following regression models for NASDAQ stocks:

UELi = π0 + π1POST + π2PRICEi + π3VOLi + π4MVEi + π5VOLAi + π6PMINi + ε1i, (a)

UELipost – UELipre = λ0 + Σλk (Xkipost – Xkipre) + ε2i, (b)

where POST is an indicator variable which equals one for the post-event period and zero otherwise; PRICEi
is the share price of stock i; VOLi is the dollar trading volume of stock i; MVEi is the market capitalization of
stock i; VOLAi is the stock i’s return volatility; PMINi is the proportion of spread quotes that are equal to the
minimum price variation for stock i; the superscripts ‘post’ and ‘pre’ denote, respectively, the post- and pre-
periods; and Xk (k = 1 through 5) represents share price, dollar trading volume, market capitalization, return
volatility, or proportion of spread quotes that are equal to the minimum price variation. We use the mean
values of PRICEi, VOLi, MVEi,VOLAi, and PMINi for each stock during the pre- and post-event periods,
respectively, in the regression. The third and fourth columns show the estimates of ω1 and ω2 in the following
regression model for the combined sample of matching NASDAQ and NYSE stocks:

UELi = ω0 + ω1POST*NASDAQ + ω2NASDAQ + ω3POST + ω4PRICEi + ω5VOLi + ω6MVEi

+ ω7 VOLAi + ω8PMINi + ε3i, (c)

where NASDAQ is an indicator variable which equals one for NASDAQ stocks and zero for NYSE stocks
and POST is an indicator variable which equals one for the post-period and zero for the pre-period. Numbers
in parentheses are t-statistics.
_____________________________________________________________________________________

π1 λ0 ω1 ω2 ω3
____________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.1064** (6.87) 0.1062** (6.97) 0.0754** (4.52) -0.0027 (-0.30) 0.0343 (1.27)
Effective spread 0.1019** (6.15) 0.1015** (6.66) 0.0730** (3.82) 0.0197 (0.81) 0.0283 (1.07)
Price impact 0.0852** (4.17) 0.0843** (4.52) 0.0598** (3.49) 0.0065 (0.42) 0.0215 (0.84)
Depth -0.1017** (-5.81) -0.1013** (-5.32) -0.0714** (-3.44) -0.0056 (-0.37) -0.0295 (-1.02)
Market quality index -0.1109** (-5.35) -0.1106** (-5.52) -0.0841** (-3.62) -0.0319 (-0.90) -0.0293 (-0.87)
________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β1 + β2 + β3) in regression model (7)
Quoted spread 0.0997** (6.26) 0.0981** (6.89) 0.0718** (3.91) -0.0020 (-0.39) 0.0239 (1.08)
Effective spread 0.1080** (6.52) 0.1087** (5.55) 0.0872** (4.27) 0.0137 (0.58) 0.0238 (0.49)
Price impact 0.0797** (4.97) 0.0817** (4.20) 0.0581** (3.64) -0.0076 (-0.44) 0.0186 (0.54)
Depth -0.0844** (-4.26) -0.0888** (-4.03) -0.0609** (-3.91) -0.0150 (-0.46) -0.0263 (-0.75)
Market quality index -0.1074** (-4.82) -0.1063** (-5.68) -0.0766** (-4.97) -0.0131 (-0.42) -0.0312 (-0.79)
_____________________________________________________________________________________
**Significant at the 1% level.

51
Table 10
Effects of the implementation of the designated market maker system (DMM) on the NYSE on the uncertainty elasticity of liquidity
(UEL)

We first calculate the composite match score (CMS) of each NYSE stock against each NASDAQ stock with the same two-digit NAICS code: CMS =
Σ[(YjNYSE - YjNASDAQ)/{(YjNYSE + YjNASDAQ)/2}]2, where Yj represents one of the four stock attributes (price, dollar volume, volatility, and market capitalization) in the
pre-period, Σ denotes the summation over k = 1 to 4, and superscripts NYSE and NASDAQ represent NYSE and NASDAQ stocks, respectively. Then, for each NYSE
stock, we select the NASDAQ stock with the lowest score. Once we match a NASDAQ stock with a NYSE stock, we no longer consider that particular NASDAQ
stock for subsequent matches. We use 779 matching pairs of NYSE and NASDAQ stocks for UEL comparisons. The first two columns of Panel A show the mean
UEL [i.e., the estimate of (α1 + α2 + α3)] from regression model (6) in the pre- and post-event periods, and the third column shows the difference in the mean UEL
between the two periods for NYSE stocks. Likewise, the first two columns of Panel B show the mean UEL [i.e., the estimate of (β1+ β2+ β3)] from regression model
(7) in the pre- and post-event periods, and the third column shows the difference in the mean UEL between the two periods for NYSE stocks. Columns 4 through 6
in both panels show the corresponding results for the control sample of NASDAQ stocks. Last column in both panels show the difference in ∆ (Post – Pre) between
NASDAQ and NYSE stocks. Numbers in parentheses are t-statistics.
_____________________________________________________________________________________________________________________
NYSE NASDAQ (Control sample) NYSE - NASDAQ
_____________________________________________________________________________________________________________________
Pre Post ∆ (Post – Pre) Pre Post ∆ (Post – Pre) ∆NYSE - ∆NASDAQ
_____________________________________________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.4721 0.5942 0.1221** (6.19) 0.5288 0.5509 0.0221 (0.92) 0.1000** (5.32)
Effective spread 0.4819 0.6067 0.1248 ** (6.32) 0.5338 0.5642 0.0304 (1.16) 0.0944** (5.83)
Price impact 0.4277 0.5287 0.1010** (5.08) 0.4713 0.4947 0.0234 (0.80) 0.0776** (3.71)
Depth -0.3972 -0.4942 -0.0970** (-4.25) -0.4624 -0.4837 -0.0213 (-0.84) -0.0757** (-3.62)
Market quality index -0.4069 -0.5328 -0.1259** (-5.13) -0.4984 -0.5296 -0.0312 (-1.33) -0.0947** (-4.76)
_____________________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β1+ β2+ β3) in regression model (7)
Quoted spread 0.4480 0.5504 0.1024** (5.86) 0.5110 0.5373 0.0263 (0.84) 0.0761** (4.00)
Effective spread 0.4590 0.5692 0.1102** (5.78) 0.5204 0.5523 0.0319 (1.38) 0.0783** (4.34)
Price impact 0.4056 0.5022 0.0966** (4.73) 0.4525 0.4737 0.0212 (0.76) 0.0754** (3.85)
Depth -0.3797 -0.4772 -0.0975** (-4.81) -0.4568 -0.4857 -0.0289 (-1.01) -0.0686** (-3.78)
Market quality index -0.3826 -0.4836 -0.1010** (-5.44) -0.4773 -0.5031 -0.0258 (-1.24) -0.0752** (-3.95)
_________________ ____________________________________________________________________________________________________
**Significant at the 1% level.

52
Table 11
Regression results for the effects of the implementation of the designated market maker (DMM)
system on the NYSE on the uncertainty elasticity of liquidity (UEL)

Panel A shows the results based on the estimate of UEL (α1 + α2 + α3) in regression model (6) and
Panel B shows the results based on the estimate of UEL (β1 + β2 + β3) in regression model (7) for 779
matching pairs of NYSE and NASDAQ stocks. The first and second columns in both panels show the
estimates of π1 and λ0 in the following regression models for matching NYSE stocks:

UELi = π0 + π1POST + π2PRICEi + π3VOLi + π4MVEi + π5VOLAi + π6PMINi + ε1i, (a)

UELipost – UELipre = λ0 + Σλk (Xkipost – Xkipre) + ε2i, (b)

where POST is an indicator variable which equals one for the post-event period and zero otherwise;
PRICEi is the share price of stock i; VOLi is the dollar trading volume of stock i; MVEi is the market
capitalization of stock i; VOLAi is the stock i’s return volatility; PMINi is the proportion of spread quotes
that are equal to the minimum price variation for stock i; the superscripts ‘post’ and ‘pre’ denote,
respectively, the post- and pre-periods; and Xk (k = 1 through 5) represents share price, dollar trading
volume, market capitalization, return volatility, or proportion of spread quotes that are equal to the
minimum price variation. We use the mean values of PRICEi, VOLi, MVEi,VOLAi, and PMINi for each
stock during the pre- and post-event periods, respectively, in the regression. The third and fourth columns
show the estimates of ω1 and ω2 in the following regression model for the combined matching sample of
NYSE and NASDAQ stocks:

UELi = ω0 + ω1POST*NYSE + ω2NYSE + ω3POST + ω4PRICEi + ω5VOLi + ω6MVEi

+ ω7 VOLAi + ω8PMINi + ε3i, (c)

where NYSE is an indicator variable which equals one for NYSE stocks and zero for NASDAQ stocks
and POST is an indicator variable which equals one for the post-period and zero for the pre-period.
Numbers in parentheses are t-statistics.
_____________________________________________________________________________________

π1 λ0 ω1 ω2 ω3
____________________________________________________________________________________
Panel A: Results based on the estimate of UEL (α1 + α2 + α3) in regression model (6)
Quoted spread 0.1199** (7.12) 0.1216** (6.91) 0.1004** (5.49) -0.0560** (-2.80) 0.0214 (1.02)
Effective spread 0.1228** (6.60) 0.1226** (6.81) 0.1014** (5.67) -0.0512** (-2.74) 0.0288 (1.11)
Price impact 0.1002** (4.82) 0.0982** (4.61) 0.0756** (4.53) -0.0417* (-2.28) 0.0225 (0.75)
Depth -0.0967** (-4.74) -0.0957** (-5.16) -0.0746** (-4.06) 0.0642** (3.06) -0.0196 (-0.71)
Market quality index -0.1231** (-5.57) -0.1218** (-4.96) -0.0956** (-4.85) 0.0909** (2.91) -0.0297 (-0.80)
________________________________________________________________________________________________________
Panel B: Results based on the estimate of UEL (β1 + β2 + β3) in regression model (7)
Quoted spread 0.1028** (5.20) 0.1029** (5.18) 0.0758** (4.22) -0.0627** (-3.12) 0.0260 (0.92)
Effective spread 0.1129** (5.88) 0.1134** (5.49) 0.0825** (4.81) -0.0598** (-2.85) 0.0308 (1.16)
Price impact 0.0969** (4.62) 0.0954** (4.49) 0.0743** (4.54) -0.0483* (2.40) 0.0209 (0.82)
Depth -0.0975** (-4.68) -0.0967** (-4.55) -0.0676** (-3.96) 0.0773** (3.13) -0.0279 (-0.83)
Market quality index -0.1022** (-4.90) -0.1005** (-4.70) -0.0642** (-3.76) 0.0949** (3.32) -0.0251 (-0.94)
_____________________________________________________________________________________
**Significant at the 1% level.
*Significant at the 5% level.

53
Appendix
The effect of VIX on various liquidity measures using market and industry liquidity: standardized coefficients
We estimate the following regression models for each of 1,945 NYSE stocks and 1,743 NASDAQ stocks using daily data from January 3, 2007 through
December 31, 2009 to examine the relation between VIX and various measures of liquidity after controlling for other determinants of liquidity:

log(LMi,t) = αi0 + αi1log(VIXt) + αi2log(VIXt-1) + αi3log(VIXt+1) + αi4log(MKTLMt) + αi5log(MKTLMt-1) + αi6log(MKTLMt+1) + αi7log(INDLMt)

+ αi8log(INDLMt-1) + αi9log(INDLMt+1) + αi10log(PRICEi,t) + αi11log(VOLi,t) + αi12log(VOLAi,t) + αi13log(VOLAi,t-1) + αi14log(VOLAi,t+1)
+αi15MKTRETt + αi16MKTRETt-1 + αi17MKTRETt+1 + αi18TUESDAY + αi19WEDNESDAY + αi20THURSDAY + αi21FRIDAY + ε1i,t, (a)

DLMi,t = βi0 + βi1DVIXt + βi2DVIXt-1 + βi3DVIXt+1 + βi4DMKTLMt + βi5DMKTLMt-1 + βi6DMKTLMt+1 + βi7DINDLMt + βi8DINDLMt-1 + βi9DINDLMt+1
+ βi10DPRICEi,t + βi11DVOLi,t + βi12DVOLAi,t + βi13DVOLAi,t-1 + βi14DVOLAi,t+1 + βi15DMKTRETt + βi16DMKTRETt-1 + βi17DMKTRETt+1
+ βi18TUESDAY + βi19WEDNESDAY + βi20THURSDAY + βi21FRIDAY + ε2i,t, (b)

where LMi,t is one of the five liquidity measures (i.e., quoted spread, effective spread, price impact, depth, and market quality index) of stock i on day t; VIXt-1, VIXt,
VIXt+1 are the CBOE VIX index on day t-1, t, and t+1; MKTLMt-1, MKTLMt, and MKTLMt+1 are the market liquidity measures (e.g., the average quoted percentage
spread across all stocks) on days t-1, t, and t+1; INDLMt-1, INDLMt, and INDLMt+1 are the industry-wide liquidity measures (e.g., the average quoted percentage
spread across all stocks in the same industry) on days t-1, t, and t+1; PRICEi,t is the price of stock i on day t; VOLi,t is the dollar trading volume of stock i on day t;
VOLAi,t is the standard deviation of returns calculated from five-minute quote midpoints of stock i on day t; MKTRETt-1, MKTRETt, and MKTRETt+1 are the market
return on days t-1, t, and t+1; TUESDAY, WEDNESDAY, THURSDAY, and FRIDAY are dummy variables for Tuesday, Wednesday, Thursday, and Friday; and
ε1i,t and ε2i,t are the error terms. D denotes the percentage change of the variable from the previous day [e.g., DLMi,t = (LMi,t – LMi,t-1) / LMi,t-1]. We estimate model (a)
and model (b) using the standardized regression method. To compare the relative effects of VIX, market liquidity (MKTLM), industry liquidity (INDLM), and
volatility (VOLA) on individual stock liquidity, we also estimate standardized coefficients on these variables after they are standardized so that their variances are all
equal to one. We show the results in the Appendix. Panel A shows the results from regression model (a) and Panel B shows the results from regression model (b).
Numbers in parentheses are t-statistics.

54
Appendix (continued)

Panel A: Standardized regression coefficients from model (a)

__________________________________________________________________________________________________________________________________________________________________

Variable VIXt VIXt-1 VIXt+1 ∑VIX MKTLMt MKTLMt-1 MKTLMt+1 ∑MKTLM INDLMt INDLMt-1 INDLMt+1 ∑INDLM VOLAt VOLAt-1 VOLAt+1 ∑VOLA
__________________________________________________________________________________________________________________________________________________________________
Results for NASDAQ stocks
Quoted spread 0.3729** 0.0265** 0.0263** 0.4257** 0.1753** -0.0054 -0.0127 0.1572** 0.1476** 0.0049 0.0130 0.1655** 0.1272** 0.0208** 0.0206** 0.1686**
(62.08) (5.52) (5.44) (77.26) (16.10) (-0.56) (-1.24) (7.27) (13.43) (0.58) (1.43) (11.51) (29.87) (4.69) (4.48) (34.85)
Effective spread 0.3957** 0.0262** 0.0359** 0.4578** 0.1969** -0.0109 -0.0129 0.1731** 0.1496** 0.0172** 0.0123 0.1791** 0.1380** 0.0352** 0.0351** 0.2083**
(64.99) (4.88) (6.35) (87.59) (22.37) (-1.40) (-1.60) (10.61) (19.82) (2.71) (1.69) (11.15) (28.68) (4.89) (4.86) (36.53)
Price impact 0.3331** 0.0166* 0.0364** 0.3861** 0.1605** -0.0246** -0.0224** 0.1135** 0.1194** 0.0170** 0.0015 0.1379** 0.1496** 0.0188** 0.0229** 0.1913**
(37.80) (2.20) (4.93) (95.73) (23.65) (-3.87) (-3.40) (10.65) (21.19) (3.01) (0.26) (13.54) (24.79) (2.97) (3.34) (32.85)
Depth ($) -0.3206** -0.0187** -0.0121* -0.3514** 0.2020** -0.0398** -0.0308** 0.1314** 0.1313** 0.0199** 0.0235** 0.1747** -0.0956** -0.0134** -0.0127** -0.1217**
(-49.03) (-3.32) (-2.14) (-54.47) (27.19) (-6.21) (-4.88) (10.01) (24.01) (4.11) (4.83) (16.72) (-22.26) (-3.25) (-2.97) (-26.86)
Market quality -0.3464** -0.0222** -0.0074 -0.3760** 0.2407** -0.0017 -0.0166** 0.2224** 0.1619** 0.0187** 0.0047 0.1853** -0.1399** -0.0116** -0.0117** -0.1632**
index (-48.16) (-3.57) (-1.10) (-62.72) (34.99) (-0.31) (-3.04) (18.72) (33.23) (4.07) (1.02) (19.03) (-30.95) (-2.94) (-3.07) (-35.04)
__________________________________________________________________________________________________________________________________________________________________
Results for NYSE stocks
Quoted spread 0.3985** 0.0147* 0.0188** 0.4320** 0.1572** -0.0053 -0.0052 0.1467** 0.1186** 0.0267** 0.0215** 0.1668** 0.1227** 0.0124** 0.0149** 0.1500**
(66.81) (2.38) (3.13) (81.62) (13.87) (-0.58) (-0.57) (7.30) (10.42) (3.20) (2.70) (11.41) (28.29) (3.74) (3.99) (33.73)
Effective spread 0.4107** 0.0182** 0.0229** 0.4518** 0.1878** -0.0122 -0.0192** 0.1564** 0.1632** 0.0270** 0.0201** 0.2103** 0.1245** 0.0190** 0.0173** 0.1608**
(69.25) (3.45) (4.20) (90.30) (16.02) (-1.78) (-2.89) (9.49) (15.02) (4.15) (3.21) (12.73) (25.09) (4.29) (3.98) (31.03)
Price impact 0.3182** 0.0146* 0.0158* 0.3486** 0.1590** -0.0230** -0.0369** 0.0991** 0.1441** 0.0179** 0.0079 0.1699** 0.1113** 0.0082** 0.0087** 0.1282**
(34.62) (1.99) (2.11) (94.86) (21.45) (-4.14) (-6.59) (10.39) (23.06) (3.64) (1.61) (18.13) (28.34) (2.81) (3.43) (29.94)
Depth ($) -0.3604** -0.0093 -0.0101* -0.3798** 0.1844** -0.0261** -0.0248** 0.1335** 0.1520** 0.0289** 0.0119 0.1928** -0.1181** -0.0089** -0.0064** -0.1334**
(-57.09) (-1.68) (-2.00) (-57.96) (18.01) (-3.36) (-3.11) (8.23) (17.91) (4.46) (1.72) (12.41) (-21.43) (-3.03) (-3.28) (-28.96)
Market quality -0.3505** -0.0187** -0.0079 -0.3771** 0.2257** -0.0045 -0.0118* 0.2094** 0.1741** 0.0185** 0.0217** 0.2143** -0.1425** -0.0011 -0.0018* -0.1454**
index (-53.75) (-3.12) (-1.30) (-63.38) (24.99) (-0.76) (-2.01) (15.12) (24.15) (3.46) (3.84) (16.95) (-24.26) (-1.26) (-2.11) (-31.84)
________________________________________________________________________________________________________________________________________________
**Significant at the 1% level.
*Significant at the 5% level.

55
 Appendix (continued)

Panel B: Standardized regression coefficients from model (b)

__________________________________________________________________________________________________________________________________________________________________

Variable DVIXt DVIXt-1 DVIXt+1 ∑DVIX DMKTLMt DMKTLMt-1 DMKTLMt+1 ∑DMKTLM DINDLMt DINDLMt-1 DINDLMt+1 ∑DINDLM DVOLAt DVOLAt-1 DVOLAt+1 ∑DVOLA
__________________________________________________________________________________________________________________________________________________________________
Results for NASDAQ stocks
Quoted spread 0.2967** 0.0021* 0.0030** 0.3018** 0.1589** -0.0255** -0.0065* 0.1269** 0.1054** 0.0113** 0.0048 0.1215** 0.1319** 0.0093** 0.0063** 0.1475**
(268.74) (2.02) (3.39) (185.36) (45.70) (-9.75) (-2.47) (22.12) (35.90) (4.19) (1.83) (21.11) (65.71) (7.43) (5.12) (47.43)
Effective spread 0.2831** 0.0035** 0.0033** 0.2899** 0.1394** -0.0048* -0.0024 0.1322** 0.1136** 0.0098** 0.0005 0.1239** 0.1607** 0.0055** 0.0110** 0.1772**
(274.32) (3.30) (3.44) (170.66) (59.27) (-2.45) (-1.17) (31.12) (53.14) (4.87) (0.24) (28.85) (86.16) (4.93) (9.63) (65.53)
Price impact 0.2511** 0.0064** 0.0013 0.2588** 0.1197** -0.0005 0.0009 0.1201** 0.0988** 0.0020 -0.0004 0.1004** 0.1226** 0.0023* 0.0066** 0.1315**
(241.75) (5.97) (1.28) (133.52) (61.35) (-0.26) (0.50) (28.96) (48.57) (1.11) (-0.23) (23.17) (81.61) (2.08) (6.02) (48.19)
Depth ($) -0.2594** -0.0026* -0.0030** -0.2650** 0.1686** 0.0012 0.0057** 0.1755** 0.1168** 0.0029 0.0028 0.1225** -0.1036** -0.0043** 0.0003 -0.1076**
(-231.11) (-2.42) (-2.98) (-144.00) (82.63) (0.78) (3.71) (51.90) (67.20) (1.70) (1.87) (38.62) (-69.50) (-3.68) (0.28) (-44.07)
Market quality -0.2636** -0.0137** -0.0032** -0.2805** 0.2143** -0.0040* -0.0009 0.2094** 0.1677** 0.0018 -0.0001 0.1694** -0.1330** -0.0060** -0.0033** -0.1423**
index (-245.01) (-13.52) (-3.19) (-162.22) (86.68) (-2.22) (-0.54) (50.68) (87.28) (0.99) (-0.06) (44.55) (-92.75) (-4.92) (-2.93) (-57.75)
__________________________________________________________________________________________________________________________________________________________________
Results for NYSE stocks
Quoted spread 0.2885** 0.0119** 0.0110** 0.3114** 0.1312** -0.0083** -0.0089** 0.1140** 0.1306** 0.0094** 0.0043* 0.1443** 0.1288** -0.0004 0.0021 0.1305**
(269.42) (11.01) (11.27) (172.80) (40.76) (-3.79) (-4.09) (23.03) (42.09) (4.46) (2.03) (28.87) (70.67) (-0.38) (1.87) (50.43)
Effective spread 0.3108** 0.0043** 0.0085** 0.3236** 0.1483** -0.0094** -0.0021 0.1368** 0.1176** 0.0127** 0.0018 0.1321** 0.1320** 0.0064** 0.0053** 0.1437**
(315.03) (4.22) (8.71) (193.53) (39.13) (-5.14) (-1.09) (27.48) (33.76) (6.74) (0.96) (26.41) (65.11) (6.28) (4.66) (50.22)
Price impact 0.2615** 0.0026* 0.0023* 0.2664** 0.1348** -0.0075** 0.0020 0.1293** 0.1073** 0.0061** -0.0021 0.1113** 0.1004** 0.0066** 0.0036** 0.1106**
(256.18) (2.50) (2.49) (140.86) (70.16) (-4.60) (1.26) (34.87) (56.68) (3.75) (-1.33) (29.54) (72.46) (6.51) (3.53) (44.50)
Depth ($) -0.2750** -0.0032** -0.0062** -0.2844** 0.1707** -0.0005 0.0042** 0.1744** 0.1341** 0.0070** 0.0013 0.1424** -0.0966** -0.0033** 0.0012 -0.0987**
(-249.95) (-3.18) (-6.59) (-160.21) (81.97) (-0.34) (2.79) (51.22) (73.82) (4.72) (0.89) (43.03) (-61.75) (-2.99) (1.19) (-42.94)
Market quality -0.2789** -0.0120** -0.0042** -0.2951** 0.1977** -0.0011 -0.0025 0.1941** 0.1424** 0.0036* 0.0011 0.1471** -0.1214** -0.0037** -0.0002 -0.1253**
index (-272.65) (-11.80) (-4.59) (-174.48) (73.45) (-0.68) (-1.48) (47.27) (68.01) (2.28) (0.68) (39.42) (-75.02) (-3.27) (-0.15) (-52.78)
__________________________________________________________________________________________________________________________________________________________________
**Significant at the 1% level.
*Significant at the 5% level.

56

View publication stats