The impact of security concentration on adverseselection costs and liquidity: an examinationof exchange traded funds

Kenneth Small & James Wansley & Matthew Hood

Published online: 8 January 2010# Springer Science+Business Media, LLC 2009

Abstract We examine the determinants of liquidity and adverse selection costs in asample of basket securities. Using Exchange Traded Funds (ETFs), we find evidencethat adverse selection costs are decreasing in the number of equities held in theunderlying portfolio, but adverse selection costs do not increase as the concentrationamong the securities increases. We find no evidence that industry concentrationincreases basket security adverse selection costs or reduces liquidity. We alsodocument significantly lower levels of adverse selection costs in ETFs versus amatched sample of equities. In addition, ETFs have quoted dollar depth that is 35times larger than a matched sample of equities, but ETFs also have higher effectiveand quoted spreads. However, when considering spreads and depth in a singlemetric, ETFs have significantly higher levels of liquidity.

Keywords Exchange Traded Fund (ETF) . Liquidity . Adverse Selection

JEL Classification G00

J Econ Finan (2012) 36:261–281DOI 10.1007/s12197-009-9117-z

We would like to thank Gary Bauer, Carry Collins, Philip Daves, John McDermott, Michael McKee,Stanley Gyoshev, Laura Starks, and Robert VanNess for helpful comments. In addition, we thank seminarparticipants at the 2004 Southern Finance Association, 2004 Eastern Finance Association annualconferences, and the 2004 Loyola College Research Symposium. We also thank participants at the 2004Financial Management’s Dissertation Consortium in Zurich, Switzerland for their valuable comments.Any remaining errors are the authors.

K. Small (*)Coastal Carolina University, P.O. Box 261954, Conway, SC 29528-6054, USAe-mail: ksmall@coastal.edu

J. WansleyDepartment of Finance, College of Business Administration, Knoxville, TN 37996, USAe-mail: jwansley@utk.edu

M. HoodCollege of Business, University of Southern Mississippi, Hattiesburg, MS 39406, USAe-mail: matthew.hood@usm.edu

1 Introduction

In this work we explore the relationship between adverse selection costs, as apercentage of the bid-ask spread, and liquidity for a sample of basket securities,namely Exchange Traded Funds (ETFs), and we compare these relationships to amatched sample of equities. Prior theoretical works predict lower adverse selectioncosts in basket securities relative to individual equities (Subrahmanyam (1991),Gorton and Pennacchi (1993), and Kumar and Seppi (1994)).

While we present the first test comparing adverse selection costs for ETFs andequities, empirical tests comparing mutual funds and equities have generally borneout these predictions, although the differences in adverse selection component of thespread have typically been smaller than expected. Neal and Wheatley (1998)estimate the adverse selection component of the bid-ask spread for 17 mutual fundsand a control sample of 17 common stocks. They find the Glosten and Harris (1988)model estimates averaged 19% for the funds and 34% for the control stocks and thatthe George et al. (1991) model estimates averaged 52% and 65% for the mutualfunds and control stocks, respectively. The small difference between estimates of theadverse selection component of the spread for equities and closed-end mutual fundspresent a problem for Neal and Wheatley. They state: “Adverse selection arisesprimarily from factors other than a firm’s current liquidation value” (p.123), andthey also suggest that the adverse selection spread decomposition models may bemis-specified. We test whether their findings may also be related to common factorsin the portfolio of underlying securities, and we find evidence that adverse selectioncosts decrease as the number of equities held in the underlying ETF increases anddoes not increase as the concentration (using a Herfindahl index) among thesecurities increases.

We estimate measures of liquidity and the adverse selection component of the bid-ask spread and test for determinants of the spread component for exchange tradedfunds. Based on the prior theoretical work and empirical findings comparing adverseselection components of mutual funds to individual equities we expect to find loweradverse selection costs for ETFs than for matched equities. Following Gorton andPennacchi (1993), one possible reason for these differences is that informed agentsmay prefer to trade industry concentrated basket securities to avoid detection byregulatory agents or uninformed traders who might be monitoring their tradingactivities in the underlying securities.

Finally, we explore the determinants of liquidity and the adverse selectioncomponent of the spread, focusing on differences in portfolio construction andconcentration. Why do some basket securities rank as the most traded instruments inthe U.S. market (e.g., QQQQ) and some are in the lowest quartile of trading volume(e.g., MTK), even though they may be concentrated in the same industry or holdcommon securities?1 Some ETFs hold as few as 11 securities and some hold over2000. Does the addition of securities diversify away adverse selections costs?

1 QQQQ is the ticker for the Nasdaq-100 Index Tracking Stock represents ownership in the Nasdaq-100Trust, a unit investment trust established to accumulate and hold a portfolio of the equity securities thatcomprise the Nasdaq-100 Index; MTK is the ticker for the Morgan Stanley technology ETF. These twoETFs have considerable overlap in their holdings.

262 J Econ Finan (2012) 36:261–281

To explore these issues, we examine the adverse selection costs and liquidity of ETFsversus a matched sample of equity securities. We also explore factors that contribute tobasket security liquidity and adverse selection costs. Neal andWheatley (1998) and Chenet al. (2002) examined adverse selection costs in closed-end mutual funds, but we focuson ETFs because of their unique structure. ETFs trade intra-day and earn returns that arevery similar to those of their underlying portfolio of securities. Unlike many closed-endmutual funds, which often trade at a discount or premium to net asset value, exchangetraded funds are easily created and redeemed. This process reduces the differencebetween the price of the ETF and its net asset value. The elimination of the premium ordiscount also reduces investor uncertainty regarding the future value of the security. Byfocusing on ETFs we remove any noise that premiums and discounts introduced inprevious studies.

Our results indicate that exchange traded funds have significantly lower adverseselection costs than a matched set of equities, regardless of the model used toestimate adverse selection costs. We find Lin et al. (1995) ETF adverse selectioncosts, as a percentage of the bid-ask spread, to be 19.7% for ETFs and 34.3% for amatched sample of equities; these percentages are 29.6% and 72.6% using theGeorge et al. (1991) model, and they are 18.1% and 44.1% using the Glosten andHarris (1988) model. In a multivariate framework, ETFs also have significantlylower adverse selection costs than the sample of equities.

We also document significantly higher levels of quoted dollar depth for ETFscompared with matched equities. Actually, the ETFs have quoted dollar depth 35times larger than the sample of equity securities. However, the ETFs also havehigher effective and quoted spreads than the matched sample of equities. Whenconsidering spreads and depth in a unified framework, ETF liquidity is significantlygreater than that of a matched sample of securities.

An extended liquidity and adverse selection analysis indicates that sectorconcentrated ETFs do not exhibit lower levels of adverse selection costs ordecreased liquidity. In addition, we find no evidence that the concentration amongthe equities held in the underlying portfolio of a basket security has an impact onadverse selection costs or liquidity. We do find evidence that the number ofsecurities held in the basket has a significant impact on liquidity and adverseselection costs. As the number of securities held in the underlying portfolioincreases, adverse selection costs decrease and liquidity increases.

2 Exchange traded funds

The popularity of Exchange Traded Funds (ETFs) has steadily increased since theirintroduction in the early 1990s. The first U.S. exchange-traded fund2 was created asa result of action taken by Leland, O’Brien, Rubenstein Associates, who lobbiedthe SEC for the creation of a Standard & Poor’s 500 tracking instrument named theIndex Trust SuperUnit. The original SuperTrust was terminated in 1996, but theAmerican Stock Exchange (AMEX) took advantage of the SuperTrust order topetition for, and receive, an SEC Order in 1992 to create a stand-alone Standard &

2 The first exchange traded fund was listed on the Toronto Stock Exchange in 1989.

J Econ Finan (2012) 36:261–281 263

Poor’s 500 index-based ETF. This unit is commonly known as the Standard &Poor’s Depository Receipt, or SPDR, with the ticker symbol SPY (Novakoff 2001).Some of the most popular ETFs track the Dow Jones Industrial Average index(Diamonds, DIA), the NASDAQ 100 index (Qubes, QQQQ), and the previouslymentioned Standard & Poor’s 500 index (Spiders, SPY)

As of May 2008 there were over 650 ETFs listed in the U.S. valued at more than$600 billion. ETFs are popular investment vehicles because they offer investorscontinuous trading during exchange hours, low premiums or discounts, taxefficiency, diversification benefits, and transparency. Unlike open-ended mutualfunds, which are priced at the end of the day, ETFs trade continuously throughoutthe day. Annual expense ratios of exchange traded funds are often lower than thoseof mutual funds, because of decreased costs associated with marketing anddistribution. Index ETFs are passively managed and on average produce lowerlevels of capital gains than actively managed funds, so they offer an advantage to taxconscious investors.

The continual process of creation and redemption works to limit the deviation ofETF prices from their underlying net asset value. An ETF unit is created when aninvestor deposits the underlying securities, and a creation unit is issued. The averagecreation unit multiple is 50,000 and share creation units range from 25,000 to600,000. Closed-end mutual funds do not have this process and many of them tradeat a substantial discount or premium to net asset value. For instance, on June 1, 2004the average deviation of the Standard & Poor’s Depository Receipt’s (SPY) pricefrom its NAV since its inception was .0006%. The average deviation was .0004% forthe DIA and .0006% for the QQQQ. The average discount of all closed-end mutualfunds on June 30, 2001 was 4.8% and the average discount on equity closed-endfunds was 11.1% (Lipper 2001). The ability to create and redeem ETFs essentiallyeliminates the difference between the price of the ETF and its net asset value.

There has been a great deal of recent research on exchange traded funds research.Elton et al. (2002) find that the SPDRs slightly under-perform low cost index mutualfunds but offer immediacy in tracking the S&P 500 Index. Boehmer and Boehmer(2003) study the liquidity impact of the cross-listing of several exchange tradedfunds on the NYSE and Nguyen et al. (2007) study how liquidity trading in ETFsaffect the inter-market competition. Hasbrouck (2003) investigates price discovery inthe S&P 500 and NASDAQ 100 indices with their ETFs and futures contracts andTse et al. (2006) investigate price discovery in the DJIA with its ETF and futurescontracts. Hedge and McDermott (2004) examine changes in liquidity of thecomponent stocks of the NASDAQ 100 and the Dow Diamonds upon theintroduction of the tracking ETFs and Van Ness et al. (2005a) examine the changesin the market quality of the component stocks of the Dow Jones Industrial Indexwith the introduction of its ETF. Small and Wansley (2005) examine the impact ofthe sector SPDR funds introductions on the underlying securities. Barari et al. (2008)examine both short-term and long-term co-movements between the G7 exchangetraded funds. And, Marshall et al. (2008) use intraday transaction data for the SPDRsto find popular technical trading rules are not profitable.

When uninformed agents migrate to the basket securities, adverse selection costsincrease in the underlying securities. However, little research has examined thecharacteristic of basket securities that make them the preferred trading venue of

264 J Econ Finan (2012) 36:261–281

uninformed agents. Is it possible that some basket securities are the preferred tradingvenue of informed agents or are less desirable to uninformed trading agents? In the nextsection, we discuss the theoretical and empirical evidence surrounding this question.

3 Adverse selection costs and basket securities: theory and evidence

Theoretical models predict that basket security traders will incur lower adverseselection costs relative to trading in equities. Subrahmanyam (1991) provides amodel that demonstrates how markets in basket securities can provide a preferredtrading medium for uninformed liquidity traders. A positive benefit accrues touninformed liquidity traders because security-specific components of adverseselection are diversified away in basket securities. Consequently, market makersare exposed to lower levels of informed trading and as a result, adverse selection isdecreased in basket securities. Subrahmanyam’s theory holds that liquidity tradersare allowed to realize their trades more efficiently by trading in basket securitiesbecause their losses to informed trading are reduced.

Further theoretical research by Gorton and Pennacchi (1993) examines liquiditytrading and informed agents. They present a model where liquidity traders forminitial portfolios with knowledge of their future participation in markets whereinformed traders are present. The presence of informed traders places liquiditytraders at a disadvantage, so the introduction of baskets securities increases liquiditytraders’ utility. Liquidity traders can effectively reduce their expected losses oftrading with informed agents if they choose to trade in basket securities.

The empirical evidence regarding adverse selection costs in basket securities has beenmixed. Neal and Wheatley (1998) estimate the adverse selection component of the bid-ask spread for 17 mutual funds and a control sample of 17 common stocks. They findonly small difference between these estimates of the adverse selection for the sample ofequities and closed-end mutual funds. We argue later that these similarities may beexplained by common factors in the portfolio of underlying securities.

Chen et al. (2002) find evidence of decreased levels of adverse selection inclosed-end mutual finds. Using a sample of funds listed on the NYSE between 1994and 1999, they find adverse selection costs are significantly lower for the closed-endmutual funds than for the control sample of equities. Clarke and Shastri (2001)examine the effects of ownership structure, the expense ratio, portfolio turnover, anddiscount to net asset value on information asymmetry in closed-end mutual funds.They find block ownership significantly impacts adverse selection costs in closed-end mutual funds, while the other factors are not significant.

While it has generally been accepted that the adverse selection costs of basketsecurities are lower than those of individual equities, previous empirical research hasprovided conflicting evidence.We expand on prior work in the area of basket securities byestimating and comparing measures of liquidity, including spreads and depth, betweenETFs and a matched set of individual equity securities. Based on prior theoretical workand empirical work with mutual funds, we expect adverse selection costs to be lower andliquidity higher in the sample of ETFs. However, offsetting this decreased adverseselection cost due to diversification is the attractiveness of ETFs to informed traders whopossess industry-specific non-public information. In addition, informed traders, who

J Econ Finan (2012) 36:261–281 265

possess firm-specific non-public information, may choose to migrate from individualsecurities to industry specific baskets, because the return characteristics of the basketcould, in some cases, be very similar to those of the individual security and may allowtrading on material non-public information without detection.3 This would provideinformed traders a preferred venue for trading on non-public information. Also as aresult of possible legal scrutiny of trading the underlying equities, informed agents mayprefer to trade in assets that mask their intentions. Furthermore, informed agents maychoose to trade in securities that have the lowest probability of revealing their non-public information to uninformed traders who may monitor their trading in theindividual securities.

As the number of securities in a security basket increases, the costs associatedwith adverse selection should be diversified away. However, on the other hand, thecosts associated with reconstitution increases. These costs could result in adivergence of the price of the ETF and its underlying basket’s NAV (i.e trackingerror). In addition, the concentration of the securities held in the basket could have asignificant impact on adverse selection costs and liquidity. For example, a securitybasket may hold 100 securities, but two could comprise 90% of the NAV of theunderlying portfolio. The characteristics of a concentrated security would be muchdifferent than one that held the 100 securities in equal proportions.

To explore this, we develop a cross-sectional regression model in which theadverse selection component of the bid-ask spread is related to several controlfactors including price, volatility, and volume and several test variables that includea concentration measure, the number of securities in the ETF as well as binaryvariables for international, sector, and broad market coverage.

In the next section, we discuss the liquidity, adverse selection, and securityconcentration measures used to test the central hypotheses.

4 Methods and data

4.1 Liquidity

We employ four commonly used liquidity measures, and we develop a measure thatincorporates two dimensions of liquidity in a single metric. We evaluate thefollowing measures:

1) Quoted Spread Quotedð Þ ¼ Aski;t � Bidi;t

2) Effective Spread Effectiveð Þ ¼ 2 pi;t �MPi;t

�� ��

3 For example, suppose a “basket security” focused in the pharmaceutical industry exists. Informed traderswith firm level information from this industry may prefer to trade in the “basket security” to avoiddetection by a regulatory body or uninformed agents who monitor their trading for information signals.Informed agents may prefer these securities because the industry focused baskets are likely to exhibitreturn characteristics that are similar to those of the underlying securities, especially when the baskedholds a small number of securities in the same industry.

266 J Econ Finan (2012) 36:261–281

3) Depth DepthSharesð Þ ¼ NSAski;t þ NSBidi;t

4) Dollar Depth DollarDepthð Þ ¼ NSAski;t � Aski;t þ NSBidi;t � Bidi;t

5) Effective=Depth Effective=Depthð Þ ¼ EffectiveiDollar Depth

�100; 00

Where Aski,t, Bidi,t, pi,t , MPi,t, NSAski,t, and NSBidi,t are the best ask price, best bidprice, price, midpoint of the inside quotes, number of shares available at the best askprice, and number of shares available at the best bid price respectively, of firm i attime t. As in Chiyachanyana et al. (2004), the quoted spreads are time weighted andthe effective spreads are trade weighted. We time weight the quoted spreads by thenumber of seconds the quote is outstanding weighted by the trading time in eachtrading day. The effective spread is weighted by the size of the trade. This weightingis calculated by dividing the size of the trade by the total trade volume for the tradingday. These are summed over that trading day and then averaged over all trading daysin the year. Effective/Depth is similar to the measure DepSri used in Kumar et al.(1998) and Hedge and McDermott (2004).4 It is the average effective spreadaveraged across the trading year divided by average dollar depth averaged across theyear, where dollar depth is scaled by 100,000. This ratio captures increases in thebid-ask spread (width of the market) while also capturing changes in quoted depth(depth of the market) All liquidity measures are averaged for each day and thenaveraged across the year to produce one observation per ETF and matched equitysecurity for 2003.

4.2 Adverse selection

Kyle (1985) suggests market markers increase the bid-ask spread when trading withinformed agents. In cases where informed agents enter a market and market makersare unable to differentiate informed and uninformed agents, the costs of informedtrading (adverse selection costs) are often pooled across trading agents. Whentrading with informed agents, market makers could choose to increase the bid-askspread to offset losses associated with their information disadvantage, and in thiswork, we employ three commonly used bid-ask spread decomposition methodolo-gies to measure adverse selection costs.

First, we follow the method of Lin et al. (1995), LSB, which decomposes the bid-ask spread into order processing and adverse selection components. Second, weemploy a variant of the decomposition method of George et al. (1991), GKN, whichalso decomposes the spread into adverse selection and order processing components.Third, we use the model of Glosten and Harris (1988), GH, which decomposes thebid-ask spread into an order-processing/inventory-holding component and anadverse selection component. Van Ness et al. (2001), in an analysis of severaladverse selection models, find the adverse selection estimates from the LSB and GHmodels are highly correlated with accepted external measures of asymmetricinformation. We discuss each model in more detail below.

4 We thank John McDermott for suggesting its use.

J Econ Finan (2012) 36:261–281 267

The LSB adverse selection and persistence parameters are estimated from thefollowing equations:

Mtþ1 �Mt ¼ lZt þ "tþ1

Ztþ1 ¼ qZt þ htþ1

Zt ¼ Pt �Mt

; ð1Þ

where Mt is the quote midpoint at time t, Pt is the transaction price at time t, εt+1 andηt+1 are random error terms. The LSB model estimate is bounded between 0 and 1,and is the proportion of the effective spread that is attributed to adverse selection.

George et al. (1991) define transactions returns as:

Rt ¼ Et þ pSq2

� �Qt � Qt�1ð Þ þ 1� pð Þ Sq

2

� �Qt þ Ut; ð2Þ

where Et is the expected return from time t-1 to t, Qt takes the value of 1 when thetransaction is a purchase and −1 when the transaction is a sale,5 and Ut areunobservable public information innovations. Van Ness et al. (2005b) employ aparameterization of the GKN model that is similar to that of Neal and Wheatley(1998), which allows the quoted spread to vary with each observation. We followVan Ness et al. (2005b) when we specify the GKN model as:

2RDt ¼ p0 þ p1St Qt � Qt�1ð Þ þ ht; ð3Þwhere RDt ¼ RT

t � RQt . R

Tt are returns derived from transactions, RQ

t are returnsderived from mid-point quotes, St is the quoted percentage bid-ask spread, Qt takesthe value of 1 when the transaction is a purchase and −1 when the transaction is asale, and ht ¼ 2 Et � ETð Þ þ 2 Ut � UTð Þ. π1 represents the order processingcomponent of the bid-ask spread, and (1−π1) is the adverse selection componentof the bid-ask spread.

The last bid-ask spread decomposition model that we employ is the GH model.This model specifies the adverse selection and inventory-holding/order-processingcosts as a linear function of transaction volume. Their model can be expressed as:

$Pt ¼ c0$Qt þ c1$QtVt þ z0Qt þ z1QtVt þ et; ð4Þwhere Qt takes the value of 1 when the transaction is a purchase and −1 when thetransaction is a sale, Vt is volume traded at time t, and et captures public informationinnovations. As in Jiang and Kim (2005), we use the average transaction volume toestimate the adverse selection component of the bid-ask spread as:

2 z0 þ z1V� �

2 c0 þ c1V� �þ 2 z0 þ z1V

� � ð5Þ

5 As suggested in Bessembinder (2003), we use the Ellis et al. (2000) method to assign trades as buys orsells. Trades are classified buys (sells) when the trade occurs at the ask (bid), and trades not occurring atthe bid or ask prices are classified using a tick test. We use the information from one quote before thereported trade time to perform the tick test.

268 J Econ Finan (2012) 36:261–281

We estimate all adverse selection costs measures across all transactions in 2003.We report the raw percentages and the dollar cost estimates. The dollar costestimates are calculated by multiplying the percentage adverse selection costestimates by the quoted spreads for the GH and GKN models and multiplying theLSB model estimates by the effective spread.

4.3 Measures of basket security concentration

In this section we discuss the control variables and concentration measures used inthe analysis. First, there are three indicator control variables for some specific typesof ETFs: International, Sector, and Broad. To control for the impact thatinternational ETFs (i.e., ETFs that hold portfolios of international securities) haveon adverse selection costs and liquidity, we include the binary variable International.The information asymmetry between U.S. investors and foreign firms is welldocumented (Bacidore and Sofianos (2002), Jiang and Kim (2005), and Small et al.(2005)). International takes the value of one when the AMEX classifies the basketsecurity as an international ETF. To directly examine the effect of industryconcentration, the ETFs that are limited to a particular sector, Sector, and thosethat invest in broad-based basket of underlying securities, Broad, are also identified.The binary variable Sector takes the value of one when the AMEX classifies thesecurity as a sector fund (i.e., when the basket holds securities primarily from onesector), and zero otherwise. The binary variable Broad takes the value of one whenthe security basket is classified by the AMEX as broad based basket and zerootherwise. International, Sector, and Broad always take the value of zero for thecontrol group of individual equity securities in the analysis.

We also employ two measures of basket security concentration among thesecurities held in the basket. First, the natural log of the number of equities thatcomprise the basket security, ln(Number). As the number of securities held in thebasket increases, we expect the adverse selections costs of the basket security todecrease. Second, the concentration among the equities held in the security bycalculating the Herfindahl Index of the concentration of the top five holdings in thesecurity. We specify the Herfindahl measure as:

Herfindahl :X5i¼1

VSiNAV

� 100

� �2 !

ð6Þ

where VSi is the value of underlying security i, and NAV is the net asset value of theexchange traded fund. We also take the natural log of the Herfindahl Index, ln(Herfindahl). As the Herfindahl Index value for the ETFs increases, we expectadverse selection costs to also increase. Each individual equity security is consideredto be a basket of one security for the calculation of the Number and Herfindahlmetrics.

Empirical research suggests that quoted bid-ask spreads tend to increase withprice and volatility and decrease as trading volume increases (Demsetz (1968), Tinic(1972), Benston and Hagerman (1974), and Hamilton (1978)). To control for theimpact of securities prices, we include the natural log of the average end of day price

J Econ Finan (2012) 36:261–281 269

of the security, ln(Price). To control for volume, we include the natural log of theaverage daily volume of the security, ln(Volume). We also control for the volatility ofsecurity returns by including the natural log of the standard deviation of daily returnsestimated over the year, ln(STD).

4.4 Data

To conduct our tests, we identify all equity ETFs listed in the U.S. in 2003.6 For thisset of exchange traded funds and a matched sample of equities, we collecttransactions data for all trading days in 2003 from The New York Stock ExchangeTrade and Quote (TAQ) database. To estimate the spread and depth measures, wefirst calculate the NBBO (National Best Bid-Offer) of each security at each time t.We exclude the following data points from the NBBO calculation:

& Non-positive prices and quotes& All quotes with a time stamp before 9:30 am (market opening) or after 4:00 pm

(market closing)& Quotes with zero bid or offer sizes, and quotes that result in a negative spread& Quoted and effective spreads that are more than 7.5 standard deviations away

from the mean& Quotes that were reported in error.

Price, volume, and return data are collected from the Center for Research inSecurity Prices (CRSP) database. Classification for industry, broad market, andinternational ETFs are taken from the American Stock Exchange’s website, and ETFsecurity holding information is obtained from Morningstar. We discuss the matchingmethodology in the next section. Our final sample consists of 113 ETFs and theirmatched equity firms.

4.5 Matching methodology

To examine the levels of adverse selection between equities and ETFs, we first constructa matched sample of equity securities. Demsetz (1968) shows that bid-ask spreads arepositively correlated with price and trading volume. The matching method is similar tothe one used in Huang and Stoll (1996), Van Ness et al. (2005a, b), and Jiang and Kim(2005).7 Available matching equities, volume, return, and price data are obtained fromCRSP. We remove all firms with fewer than 227 trading days and all non-ordinarycommon shares (ADRs, Certificates, Shares of Beneficial Interest, and otherdepository receipts). The data are averaged daily over all trading days in 2003, and

6 Bond holders are exposed to a different set of informed trader incentives than equity holders and theunderlying portfolios for bond ETFs may exhibit microstructure characteristics, such as adverse selectioncosts, that are much different than those of equity ETFs.

7 Unlike studies that match equities to equities, we do not match on market capitalization, because themarket capitalization of ETFs and the market capitalization of equities do not capture the same factor.

270 J Econ Finan (2012) 36:261–281

the NYSE8 or AMEX equity security that minimizes the following objective isselected as the matching equity:

Score ¼X3

i¼1

XNon�ETFi � XETF

i

X Non�ETFi þ XETF

i½ �=2� 2

;

where Xi represent one of the three ETF matching attributes, which are the end of dayprice of the security averaged over the year, the standard deviation of daily returnsestimated over the year, and the daily volume averaged over the all trading days in2003. We select the stock with the lowest matching score for each ETF. The mean andmedian matching scores are 0.113 and 0.043, respectively.9

5 Results

5.1 Univariate analysis

We split our sample into ETFs and equities and examine the dimensions of liquidityand adverse selection costs in a univariate context. We present the results of thispartitioning in Table 1. As seen, the adverse selection percentage cost estimates areconsistently larger for equities than for ETFs, regardless of the model used. The LSBadverse selection costs bid-ask spread percentage estimates are 19.7% for ETFs and34.3% for the matched sample of equities. The GKN adverse selection bid-askspread component percentages are 17.2% for ETFs and 38.6% for equities. Finally,the GH adverse selection estimates are 14.2% for ETFs and 22.7% for equities. Inaddition, the dollar cost estimates for equities are also significantly greater for theGKN, but not for LSB or GH models. The univariate results clearly support theconjecture that basket securities have significantly lower levels of adverse selectioncosts than a matched sample of equities.

Table 1 also reports mean liquidity measures for the ETFs and matched equities.Lower adverse selection costs may or may not lead to an increase in liquidity. Higherlevels of liquidity would be associated with lower bid-ask spreads and greater depth.We find that quoted and effective spreads are statistically and economically larger forETFs than the sample of equities while dollar depth is also statistically andeconomically larger in ETFs versus the equities. The average quoted dollar depth forETFs is $2,531,200, while the quoted dollar depth for the sample of equities is only

9 This suggests that the ETFs and matched equities are similar along the pre-specified attributes, and thematched portfolio acts as a benchmark for drawing conclusions regarding spread and adverse selectioncosts in the cross-security analysis. The summary statistics of the other matching attributes are availablefrom the authors.

8 Because of a limitation with reported quoted depth in NASDAQ listed equities in the TAQ database, welimit our matching firms to NYSE and AMEX listed firms. TAQ only reports depth for one NASDAQdealer, even if more than one dealer is at the best bid or offer. Because of this underreporting, depth forNASDAQ firms may be understated in the TAQ database. Our results when allowing the inclusion of theNASDAQ firms are quantitatively similar to those sample including NYSE firms only. Restricting thesample to NYSE and AMEX firms resulted in twenty-six firms being replaced by NYSE firms.

J Econ Finan (2012) 36:261–281 271

$71,200. Thus, ETFs possess one characteristic of higher liquidity, namely greaterdollar depth, but their wider average and quoted spreads suggests lower liquidity.

To capture both the spread and depth dimensions of liquidity, we compute theEffective/Depth ratio, defined as the average effective spread over the trading yeardivided by the average dollar depth, where dollar depth is scaled by 100,000. Thisratio captures increases in the bid-ask spread (width of the market) while alsocapturing changes in quoted depth (depth of the market). Liquidity decreases as theratio increases (decreased dollar depth or increased effective spreads) and liquidity

Table 1 ETF and sample equity univariate characteristics

Variable Whole Sample ETFs Equities Difference t-stat

Price 45.80 48.30 43.38 4.91 1.21

STD 0.0136 0.0131 0.0140 0.0009 1.60

Volume 894,721 1,368,408 437,368 932,040 1.18

Herfindahl 5,347 571 10,000

Number 126 246 1

International 0.109 0.221 0

Sector 0.231 0.469 0

Broad 0.153 0.309 0

Quoted 0.1155 0.1420 0.0894 0.0527*** 4.39

Effective 0.1080 0.1340 0.0821 0.0523*** 5.24

DepthShares 27,558 53,545 2,035 51,510*** 12.80

DollarDepth 1,290,123 2,531,219 71,189 2,460,031*** 9.88

Effective/Depth 0.1020 0.0142 0.189 −0.175*** −8.07LSB 0.2710 0.1970 0.343 −0.1464*** −7.67LSBDollar 0.0302 0.0277 0.0326 −0.0048 −1.08GKN 0.2799 0.1718 0.386 −0.2144*** −17.79GKNDollar 0.0280 0.0217 0.034 −0.0124*** −2.94GH 0.1848 0.1422 0.2266 −0.0844*** −3.89GHDollar 0.0248 0.0227 0.02268 0.004 0.78

This table contains the mean values of the liquidity and adverse selection model estimates for the sampleof ETFs and equities. Price is the average closing price for the year (2003), STD is the standard deviationof daily returns estimated over the year, and Volume is the daily volume of security averaged over the year.Herfindahl is the Herfindahl Index concentration value of the security and Number is the number ofunderlying equities that comprise the security. International, Sector, and Broad are binary variables thattake the value of one when the ETF primarily holds: non-U.S. denominated securities (International),securities that are all in the same industry sector (Sector), and securities from many diverse industrygroups (Broad). Quoted is the quoted spread, Effective is the effective spread, DepthShares is the numberof shares quoted at the best bid and offer, DollarDepth is the dollar value of the shares quoted at the bidand ask price, and Effective/Depth is the average effective spread averaged across the trading year dividedby average dollar depth averaged across the year, where dollar depth is scaled by 100,000. LSB, GKN, andGH are adverse selection measures. The LSB model is from Lin et al. (1995), the GKN model is fromGeorge et al. (1991), and the GH model is from Glosten and Harris (1988) percentage bid-ask spreaddecomposition adverse selection estimates. The dollar estimates are the percentage adverse selection costestimate multiplied by the effective spread. Significance at the 10% level is indicated by *, significance atthe 5% level is indicated by **, and significance at the 1% level is indicated by ***

272 J Econ Finan (2012) 36:261–281

increases as the ratio decreases (increased dollar depth or decreased effectivespreads). The Effective/Depth ratio for equities is 0.189 but only 0.014 for ETFs.

5.2 Multivariate analysis of industry concentration

To control for the effects of price, standard deviation, and volume in security returnsacross the securities included in the sample, we estimate the following model:

Liquidity ¼ aþ b1 ln Priceð Þi þ b2 ln STDð Þi þ b3 ln Volumeð Þi þ b4ETFi

þ b5Internationali þ b6Sectori þ b7Broadi þ ei

In separate regressions Liquidity represents the Quoted Spread, Effective Spread,Dollar Depth, and Effective/Depth. Dollar Depth is the dollar value of the sharesquoted at the bid and ask prices and Effective/Depth is the effective spread averagedacross the trading year divided by average dollar depth. The average dollar depth isalso scaled by 100,000. The natural logs of the average end of day price of thesecurity, ln(Price), the standard deviation of daily returns estimated over the year, ln(STD), and the average daily trading volume over the year, ln(Volume), are used ascontrol variables. The variables of interest are the three that subdivide ETFs:International, Sector, and Broad, described in Section 4.3. The International binaryvariable is triggered for ETFs that primarily hold non-U.S. denominated securities.The Sector and Broad binary variables are triggered for ETFs that primarily holdsecurities in the same industry and ETFs that hold securities from many diverseindustry groups. In each of these cases the matched sample equity securities alwaystake the value of zero.

The results of the multivariate liquidity analysis10 can be found in Table 2. Thebinary variable ETF is positive and significant in the quoted spread and effectivespread regressions. The quoted spreads are 5.0 cents and effective spreads are 4.9cents higher for ETFs than those in a matched sample of equities, controlling for theprice, volume and standard deviation. ETFs have, on average, $2,460,000 morequoted dollar depth than the matched sample of equity securities. The positive andsignificant coefficient estimate on the ETF binary variable in the dollar depthspecification and the positive and significant coefficient estimate on the ETF binaryvariable in the quoted and effective spread analyses lead to an ambiguous result fortheir overall liquidity.

To reconcile this difference, we turn to the Effective/Depth ratio to aid indetermining the liquidity difference between ETFs and equities. Recall that liquiditydecreases as the ratio increases (decreased dollar depth or increased effectivespreads) and liquidity increases as the ratio decreases (increased dollar depth ordecreased effective spreads). The coefficient for ETFs on Effective/Depth (-0.185) isnegative and significant. Thus, when spreads and depth are considered in a unifiedframework, exchange traded funds have greater liquidity than the matched sample ofequities.

10 The multivariate results are similar when a control variable is added that measures the percentage oftrading that occurs off the primary exchange.

J Econ Finan (2012) 36:261–281 273

Focusing the attention on the characteristics of the ETFs we find that fundsconcentrated internationally or in a particular sector and those that have a broadmarket base all have significantly higher effective spreads than do the matched set ofequities. Quoted spreads are also significantly larger for broad market and sectorETFs, but are not significantly larger for internationally focused ETFs. The broad-based baskets have the highest spreads relative to equities, followed by the sectorconcentrated baskets, and then international baskets. Further, the parameter estimateson the variables in the depth specifications indicate that all forms of ETFs havesignificantly higher dollar depth than the matched set of equity securities.

Again, turning to the Effective/Depth ratio to examine the relationship theliquidity of the ETFs and the matched sample of equities using a single metric wefind the parameter estimates on International, Sector, and Broad are all negative andsignificant. This suggests that when spreads and depth are taken together, liquidity is

Table 2 Exchange traded funds and equity liquidity analysis (Industry concentration analysis). This tablecontains regression analysis coefficient estimates for the following model: Liquidityi ¼ a þ b1ln Priceð Þi þb2ln STDð Þi þ b3ln Volumeð Þi þ b4ETFi þ b5Internationali þ b6Sectori þ b7Broadi þ ei

Liquidity measures

Quoted Spread Effective Spread Dollar Depth Effective/Depth

Intercept .045 0.029 .047 .031 1,262,656 −371,275 .151 .149

(.486) (.311) (.589) (.399) (.667) (−.237) (.391) (.832)

ln(Price) −.014 −.014 −.011 −.010 −232,341 −136,019 −.028 −.028(−1.45) (−1.37) (−1.31) (−1.20) (1.19) (−1.11) (−1.24) (−1.22)

ln(STD) −.025 −.030 −.021 −.026 229,652 −301,326 −.058 −.058(−1.00) (−1.23) (−1.04) (−1.29) (.517) (−.804) (−1.15) (−1.13)

ln(Volume) −0.001 −.002 −.001 −.002 54,822 −32,578 −.009* −.0095*(−.351) (−.68) (−.622) (−.935) (.941) (−.743) (−1.77) (−1.69)

ETF 0.050*** .049*** 2,475,679*** −0.185***(4.02) (4.83) (9.94) (−7.78)

International .019 .033** 347,807*** −.166***(1.21) (2.24) (4.60) (−6.63)

Sector .0477*** .043*** 1,932,351*** −.192***(3.07) (3.30) (9.52) (−7.51)

Broad .077*** .071*** 4,883,989*** −.189***(4.54) (5.16) (8.66) (−8.45)

Adjusted R2 .07 .09 .10 .11 .30 .58 .23 .23

F Value 5.60 4.78 7.55 5.76 25.64 52.31 18.03 12.02

N 226 226 226 226 226 226 226 226

Liquidity is measured by Quoted Spread, Effective Spread, Dollar Depth, and Effective/Depth. DollarDepth is the dollar value of the shares quoted at the bid and ask prices and Effective/Depth is the averageeffective spread averaged across the trading year divided by average dollar depth averaged across the year,where dollar depth is scaled by 100,000. The natural log of the average price at the end of the day over theyear is ln(Price), the natural log of the standard deviation of daily returns over the year is ln(STD), and thenatural log of the daily volume of security averaged over the year is ln(Volume). ETF takes the value ofone when the security is an exchange traded fund. International, Sector, and Broad take the value of onewhen the ETF primarily holds: non-U.S. denominated securities (International), securities in the sameindustry sector (Sector), and securities from many diverse industry groups (Broad). By definition theInternational, Sector, and Broad variables are always zero for the equity securities. Significance at the10% level is indicated by *, significance at the 5% level is indicated by **, and significance at the 1%level is indicated by ***.

274 J Econ Finan (2012) 36:261–281

greater in international, sector, and broad-based exchange traded funds than in thematched sample of equity securities.

Recall that the concentration hypothesis suggests that as industry basketconcentration increases, the liquidity of the basket should decrease. Based on thishypothesis, we would expect to find the smallest coefficient on Sector, given theinverse relationship between the variable and liquidity. However, we find ETF sectorfunds to have the highest liquidity (most negative coefficient), when using theEffective/Depth as a proxy for liquidity, although the differences in coefficientsacross International, Sector and Broad are not significant. Thus, we fail to findsupport to the industry concentration liquidity hypothesis, and our results show that,although sector ETFs have greater quoted and effective spreads than matchedequities, they also have substantially greater depth, which dominates the relationshipbetween concentration and liquidity in our analysis. We now turn our attention toexamining the adverse selection costs of equity exchange traded funds versus thosethe matched sample of equities.

To explore the relationship between security industry concentration and adverseselection costs, we estimate the following OLS specification:

Adverse Selectioni ¼ aþ b1 ln Pr iceð Þi þ b2 ln STDð Þi þ b3 ln Volumeð Þi þ b4ETFi

þ b5Internationali þ b6Sector1 þ b7Broadi þ ei

In separate regressions, Adverse Selection takes the value of the adverse selectionestimates summarized in Table 1, with the explanatory variables defined earlier —ln(Price), ln(STD), ln(Volume), ETF, International, Sector, and Broad.

The results of the adverse selection estimation are shown in Table 3. Coefficientestimates on the ETF binary variable are negative and significant in all three adverseselection models. Regardless of the method used to determine adverse selection, wefind lower levels of adverse selection in ETFs than for the matched equities. In theLSB model, the percentage of the spread that is attributed to adverse selection costsis on average 14.3% smaller than for the matched sample of equities, and is 42.5%and 26.4% smaller for the GKN and GH models, respectively.

The coefficient estimates on Sector, Broad, and International are also negativeand significant in all percentage adverse selection specifications. We find noconsistent pattern and no statistically significant difference in the coefficients fromthe different specifications. This evidence suggests that ETFs concentrated on anindustry or concentrated internationally do not have significantly different adverseselection costs than a broad-based ETFs.

We do find that equities have significantly higher adverse selection costs than thesample of basket securities (ETFs), providing support for the adverse selectionbasket security hypothesis and strengthening prior research on this topic. Theindustry concentration hypothesis cannot be used to explain the result of Neal andWheatley (1998). Recall, we conjectured that commonalities in the underlyingmutual funds in their sample might have led to the similar parameter estimates onadverse selection models from their equity and mutual fund samples. Investoruncertainty associated with NAV price deviations would certainly create adverseselections costs. Since we use ETFs, which have essentially no premium or discount,

J Econ Finan (2012) 36:261–281 275

Tab

le3

Exchangetraded

fundsandequity

adverseselectioncostanalysis(industryconcentrationanalysis).Thistablecontains

regression

analysiscoefficientestim

ates

forthe

follo

wingmod

el:AdverseSelection i

¼aþb 1ln

Price

ðÞ iþ

b 2ln

STD

ðÞ iþ

b 3ln

Volum

eð

Þ iþb 4ETFiþb 5Internationa

l iþb 6Sector

iþb 7Broad

iþe i

Adverse

selectionmeasures

LSB(Percentage)

LSB(Percentage)

LSB(D

ollar)

GKN

(Percentage)

GKN

(Percentage)

GKN

(Dollar)

GH

(Percentage)

GH

(Percentage)

GH

(Dollar)

Intercept

.298*

.269

.0106

.456**

.503***

.009

.509***

.471***

.025

(1.71)

(1.62)

(.279)

(2.35)

(2.65)

(.157)

(2.82)

(2.66)

(.552)

ln(Price)

−.0038

−.002

−.003

−.003

−.006

−.011

−.011

−.009

−.006

(−.312)

(−.169)

(−1.04)

(−.238)

(−.451)

(1.41)

(−.960)

(−0.79)

(−1.35)

ln(STD)

.004

−.003

−.010

−.032

−.017

−.021

−.015

−.027

−.017

(.142)

(−.118)

(−1.27)

(−.801)

(−.420)

(1.08)

(−.426)

(−0.77)

(−1.43)

ln(Volum

e).006

.005

−.0008

.012

.014

.0002

−.007

−.009*

−.0028*

(.736)

(.558)

(−.477)

(1.49)

(1.82)

(.108)

(−1.40)

(−1.73)

(−1.85)

ETF

−.143***

−.425***

−.264***

(−6.88)

(−20.26)

(−14.1)

International

−0.121***

−.008

−.346***

−.028***

−.314***

−.029***

(−4.69)

(−1.48)

(−10.7)

(2.97)

(−14.6)

(−4.69)

Sector

−.170***

−.010*

−.415***

−.027***

−.277***

−.021***

(−5.94)

(−1.86)

(−15.3)

(−2.73)

(−13.6)

(−3.25)

Broad

−.117***

.003

−.500***

−.030***

−.208***

−.002

(−3.32)

(.464)

(−18.0)

(−3.36)

(−5.96)

(−.242)

AdjustedR2

.20

.20

.002

.68

.70

.042

.45

.47

.05

FValue

15.05***

10.67***

1.09

122.88***

90.27***

2.63**

47.7***

34.09***

3.15***

N226

226

226

226

226

226

226

226

226

Adverse

Selectionismeasuredby

threemod

els,each

with

apercentage

estim

ateandado

llarestim

ateT

henaturallogof

theaveragepriceattheendof

thedayov

ertheyear

isln

(Price),thenaturallog

ofthestandard

deviationof

daily

returnsover

theyear

isln(STD),andthenaturallog

ofthedaily

volumeof

security

averaged

over

theyear

isln(Volum

e).

The

naturallog

oftheconcentrationvariablesareln(H

erfin

dahl),theHerfindahlIndex

concentrationvalueof

thesecurity,and

ln(Num

ber),the

numberof

underlying

equitiesthat

comprisethesamplesecurity.ETFtakesthevalueof

onewhenthesecurity

isan

exchange

traded

fund.Internationa

l,Sector,andBroad

take

thevalueof

onewhentheETF

prim

arily

holds:non-U.S.denominated

securities(Internatio

nal),securitiesin

thesameindustry

sector

(Sector),andsecuritiesfrom

manydiverseindustry

groups

(Broad).By

definitio

ntheHerfin

dahl

andNum

bervariablesarealwaysonefortheequity

securitiesandtheInternational,Sector,andBroad

variablesarealwayszero

fortheequity

securities.Significanceat

the10

%levelisindicatedby

*,sign

ificance

atthe5%

levelisindicatedby

**,andsign

ificance

atthe1%

levelisindicatedby

***

276 J Econ Finan (2012) 36:261–281

we eliminate noise associated with deviations from NAV. So it is possible that thesimilarities found by Neal and Wheatley (1998) between mutual fund and equityadverse selection costs were driven by premiums and discounts and not bycommonalities in the underlying portfolio of securities held by the funds themselves.

We note that domestic equities even have higher adverse selection costs than theinternationally-concentrated basket securities, which is surprising given the amountof informational asymmetry between U.S. investors and foreign firms. However,given the diversification of adverse selection costs across the securities held in thebaskets, lower levels of total basket security adverse selection costs appear to beachievable. In the next section, we examine this conjecture more closely.

5.3 Multivariate analysis of security concentration

We next expand our analysis to determine whether security concentrationsignificantly affects liquidity and adverse selection costs. We construct the followingmodels:

Liquidityi ¼ aþ b1 ln Priceð Þi þ b2 ln STDð Þi þ b3 ln Volumeð Þi þ b4 ln Herfindahlð Þiþ b5 ln Numberð Þi þ b6Internationali þ b7Sectori þ b8Broadi þ ei

Adverse Selectioni ¼ a þ b1 ln Pr iceð Þi þ b2 ln STDð Þi þ b3 ln Volumeð Þiþ b4 ln Herfindahlð Þi þ b5 ln Numberð Þi þ b6Internationali

þ b7Sectori þ b8broadi þ ei

All variables are as defined previously in the paper. Each equity security isconsidered to be a basket of one security for the calculation of the Herfindahl andNumber metrics. These results are reported in Table 4 and Table 5.

In this analysis, we are interested in the coefficient estimates on the concentrationproxies, ln(Herfindahl) and ln(Number). The coefficient estimate for ln(Herfindahl)is insignificant in all specifications except the dollar depth model in which ourmeasure of the Herfindahl index is positively associated with dollar depth. Itscoefficient in the Effective/Depth model, however, suggests that it has no impact onsecurity liquidity. Since the concentration hypothesis posits a positive and significantparameter estimate on this variable, our evidence rejects the conjecture thatconcentration among the equities in the security leads to a decrease in liquidity.11

Next we turn our attention to the coefficient estimate on ln(Number). The numberof securities has no significant impact on quoted or effective spreads, but itsignificantly increases dollar depth and significantly decreases the Effective/Depthratio. Thus, as the number of securities in a basket security increases, the liquidity ofthe security increases.

11 It may also be that there is inadequate cross-sectional variation in Herfindahl for the relationship to besignificant. However, in Table 2, the minimum and maximum values and the standard deviation ofHerfindahl suggest considerable cross-sectional variation.

J Econ Finan (2012) 36:261–281 277

278 J Econ Finan (2012) 36:261–281

Table 5 presents the results of the adverse selection model after the addition of theconcentration proxies. The coefficient estimate for ln(Herfindahl) is insignificant inall specifications. Similar to the results of the liquidity analysis, the concentrationamong the equities held in the underlying portfolio of the basket security appears tohave no impact on the adverse selection costs. The coefficient estimate for ln(Number) is negative and significant in the partial adverse selection costsspecifications for the LSB and GKN models but insignificant in the GH model.

Table 4 Exchange traded funds and equity liquidity analysis (number and full model concentrationanalysis). This table contains regression analysis coefficient estimates for the following models:Liquidityi ¼ a þ b1 ln Priceð Þi þ b2 ln STDð Þi þ b3 ln Volumeð Þi þ b4 ln Herfindahlð Þi þ b5 ln Numberð Þiþb6Internationali þ b7Sectori þ b8Broadi þ ei

Liquidity measures

Quoted spread Effective spread Dollar depth Effective/Depth

Intercept .100 .134 .103 .128 −1,103,488 −3,692,916 .205 .174

(.987) (1.18) (1.20) (1.34) (−.424) (1.49) (1.15) (.993)

ln(Price) −.012 −.011 −.008 −.008 −51,114 −49,538 −.035 −.029

(1.19) (−1.15) (−1.03) (−1.01) (−.361) (−.479) (1.46) (−1.23)

ln(STD) −.031 −.029 −.028 −.026 247,577 −68,079 −.050 −.061

(1.23) (−1.14) (−1.32) (−1.21) (.731) (−.217) (.958) (−1.15)

ln(Volume) −.002 −.002 −.002 −.002 47,922 −6,755 −.007 −.009*

(.761) (−.751) (1.13) (−1.03) (1.00) (−.166) (−1.28) (1.71)

ln(Herfindahl) −.009 −.012 −.009 −.011 90,326 316,001** .0002 −.002

(1.22) (1.35) (−1.44) (−1.46) (.631) (2.86) (.071) (−.991)

ln(Number) .0045 .009 .003 .0061 886,970*** 813,869*** −.041*** −.007*

(.494) (1.02) (.47) (.731) (4.60) (2.86) (−5.88) (−1.81)

International −.053* −0.024 1,044,284 −.149***

(−1.86) (−.96) (1.36) (5.70)

Sector −.018 −.007 476,508 −.175***

(−.642) (−.311) (.66) (−6.59)

Broad −.051 −.031 2,865,816** −.164***

(−1.01) (−.711) (2.34) (−6.22)

Adjusted R2 .11 .11 .14 .13 .50 .63 .19 .22

F Value 6.66*** 4.60*** 8.36*** 5.35*** 45.66*** 49.70*** 11.37*** 8.97***

N 226 226 226 226 226 226 226 226

Liquidity is measured by Quoted Spread, Effective Spread, Dollar Depth, and Effective/Depth. DollarDepth is the dollar value of the shares quoted at the bid and ask prices and Effective/Depth is the averageeffective spread averaged across the trading year divided by average dollar depth averaged across the year,where dollar depth is scaled by 100,000. The natural log of the average price at the end of the day over theyear is ln(Price), the natural log of the standard deviation of daily returns over the year is ln(STD), andthe natural log of the daily volume of security averaged over the year is ln(Volume). The natural log of theconcentration variables are Herfindahl, the Herfindahl Index concentration value of the security, andNumber, the number of underlying equities that comprise the sample security. ETF takes the value of onewhen the security is an exchange traded fund. International, Sector, and Broad take the value of one whenthe ETF primarily holds: non-U.S. denominated securities (International), securities in the same industrysector (Sector), and securities from many diverse industry groups (Broad). By definition the Herfindahland Number variables are always one for the equity securities and the International, Sector, and Broadvariables are always zero for the equity securities. Significance at the 10% level is indicated by *,significance at the 5% level is indicated by **, and significance at the 1% level is indicated by ***

Tab

le5

Exchang

etraded

fundsandequity

adverseselectioncostanalysis(num

berandfullmod

elconcentrationAnalysis).Thistablecontains

regression

analysiscoefficient

estimates

forthe

follow

ing

mod

els:

AdverseSelection i

¼aþb 1

lnPrice

ðÞ iþ

b 2ln

STD

ðÞ iþ

b 3ln

Volum

eð

Þ iþb 4

lnHerfin

dahl

ðÞ iþ

b 5ln

Num

ber

ðÞ iþ

b 6International iþ

b 7Sector

iþb 8Broad

iþe i

Adverse

selectionmeasures

LSB

(Percentage)

LSB(Percentage)

LSB(D

ollars)

GKN

(Percentage)

GKN

(Percentage)

GKN

(Dollars)

GH

(Percentage)

GH

(Percentage)

GH

(Dollars)

Intercept

.311

.174

.012

.027

.469**

.032

.067

.476**

.040

(1.51)

(.942)

(.299)

(.467)

(2.26)

(.519)

(1.35)

(2.38)

(.782)

ln(Price)

−.008

−.0045

−.004

−.012

−.0102

−.011

−.007

−.013

−.007

(−.692)

(−.348)

(−1.08)

(−1.49)

(−.710)

(−1.38)

(−1.50)

(−1.05)

(−1.39)

ln(STD)

.012

−.0057

−.0112

−.021

−.023

−.022

−.014

−.034

−.018

(.383)

(−.174)

(−1.32)

(−1.09)

(−.567)

(−1.10)

(−1.20)

(−.973)

(−1.50)

ln(Volum

e).008

.005

−.0008

.0004

.014*

.0001

−.002

−.010*

−.003

(1.00)

(.553)

(−.502)

(.219)

(1.75)

(.060)

(−1.49)

(−1.79)

(−1.90)

ln(H

erfindahl)

.003

.011

−.00004

−.001

.005

−.002

−.003

.0015

−.001

(.498)

(1.52)

(−.023)

(−.627)

(.756)

(1.05)

(−1.22)

(.178)

(−.489)

ln(Num

ber)

−.028***

−.011

−.0026

−.007**

−.025**

−.001

−.006**

−.029*

−.004

(−3.10)

(−.638)

(−.743)

(.007)

(−2.12)

(−.59)

(−1.22)

(−1.69)

(−1.10)

International

−.046

−.0003

−.248***

−.031**

−.220***

−.020

(−.655)

(−.025)

(−4.25)

(2.46)

(−3.40)

(−1.32)

Sector

−.103*

−.003

−.323***

−.029**

−.188***

−.011

(1.71)

(−225)

(−6.31)

(−2.34)

(−3.1 1)

(−.806)

Broad

.012

.016

−.338***

−.037**

−.054

.012

(.090)

(.592)

(−3.83)

(−2.01)

(−.433)

(.395)

AdjustedR2

.15

.000

.21

.03

.03

.71

.02

.05

.47

FValue

9.30***

.892

8.37***

2.76**

2.00**

69.01***

2.01*

2.48**

26.46***

N226

226

226

226

226

226

226

226

226

Adverse

Selectionismeasuredby

threemodels,each

with

apercentage

estim

ateandadollarestim

ate.The

LSBmodelisfrom

(Lin

etal.1995),the

GKNmodelisfrom

Georgeetal.(1991),andthe

GHmodelisfrom

Glosten

andHarris(1988)

percentage

bid−

askspread

decompositionadverseselectionestim

ates.T

hedollarestim

ates

arethepercentage

adverseselectioncostestim

atemultiplied

bytheeffectivespread.T

henaturallog

oftheaveragepriceattheendof

thedayover

theyear

isln(Price),thenaturallog

ofthestandard

deviationof

daily

returnsover

theyear

isln(STD

),andthe

naturallog

ofthedaily

volumeof

securityaveraged

overtheyearisln(Volum

e).T

henaturallog

oftheconcentrationvariablesareHerfindahl,the

HerfindahlIndex

concentrationvalueof

thesecurity,

andNum

ber,thenumberof

underlying

equitiesthatcomprisethesamplesecurity.E

TFtakesthevalueof

onewhenthesecurity

isan

exchange

traded

fund.International,Sector,and

Broad

take

the

valueof

onewhentheETFprimarily

holds:non-U.S.denom

inated

securities(International),securities

inthesameindustry

sector

(Sector),and

securitiesfrom

manydiverseindustry

groups

(Broad).

BydefinitiontheHerfindahl

andNum

bervariablesarealwaysonefortheequity

securitiesandtheInternational,Sector,and

Broad

variables

arealwayszero

fortheequity

securities.Significanceat

the10%

levelisindicatedby

*,significance

atthe5%

levelisindicatedby

**,andsignificance

atthe1%

levelisindicatedby

***

J Econ Finan (2012) 36:261–281 279

These results provide evidence that as the number of securities in an ETF increases,adverse selection is diversified away, and general liquidity is improved. In addition,International, Sector, and Broad retain their negative and significant coefficientestimates in the GKN percentage and dollar estimate models, although the results aregenerally mixed in the other models. We interpret these results as providing modestsupport for the negative relationship between adverse selection costs of ETFs andsome specific characteristics, namely industry concentration, measured by Sectorand Broad and security concentration, measured by ln(Herfindahl) and ln(Number).

6 Conclusion

We examine liquidity and adverse selection costs in a sample of Exchange TradedFunds (ETFs) and a matched sample of equities. These relationships depend, to alarge extent, on the definition of liquidity. Defining liquidity as a spread measure,ETFs are less liquid; their quoted and effective spreads are larger than the matchsample. Defining liquidity as a depth measure, ETFs are more liquid; they havesubstantially larger dollar depth than the matched sample. We reconcile these twodimensions of liquidity by computing Effective/Depth, which is defined as theaverage effective spread divided by average dollar depth. With this measure we findthat ETFs are more liquid as their spread is more than compensated by their depth.

We focus primarily on the relationship between this measure of liquidity andadverse selection for ETFs and sample equities documenting significantly lowerlevels of adverse selection costs in the sample of equity ETFs versus the matchedsample of equities. In addition, we present evidence that adverse selection costs aredecreasing in the number of equities held in the underlying portfolio of the ETF. Weshow that adverse selection costs do not increase as the concentration among thesecurities increases and find no evidence that industry concentration increases ETFadverse selection costs or reduces liquidity. As a whole, ETFs, and security basketsin general, provide a beneficial trading medium for uninformed traders.

References

Bacidore JM, Sofianos G (2002) Liquidity Provisions and Specialist Trading in NYSE-Listed non-U.S.stocks. J. Financ. Econ. 63:133–58

Barari M, Lucey B, Voronkova S (2008) Reassessing Co-Movements Among G7 Equity Markets:Evidence From iShares. Appl. Financ. Econ 18:863–877

Benston G, Hagerman R (1974) Determinants of the Bid-Ask Spread in the Over-the-Counter Market. J.Financ. Econ. 1:353–364

Bessembinder H (2003) Quote-based competition and trade execution costs in NYSE-listed stocks. J.Financ. Econ. 70:385–422

Boehmer B, Boehmer E (2003) Trading Your Neighbor's ETFs: Competition or Fragmentation? J. Bank.Finance 27:1667–1704

Chen J, Jiang C, Kim J, McInish T (2002) Bid-ask Spreads Information Asymmetry, and InvestorSentiment: Evidence from Closed-end Funds, Rev. Quant. Financ. Account. 21:303–321

Chiyachantana C, Jiang C, Taechapiroontong N, Wood R (2004) The Impact of Regulation Fair Disclosureon Information Asymmetry and Trading: An Intraday Analysis. Finan. Rev 39:549–577

Clarke, J., and Shastri, K., 2001, Adverse Selection Costs and Closed-end Funds, University of PittsburghWorking Paper

280 J Econ Finan (2012) 36:261–281

Demsetz H (1968) The Cost of Transacting. Q. J. Econ. 82:33–53Ellis K, Michaely R, O’Hara M (2000) The accuracy of trade classification rules: Evidence from Nasdaq.

J. Finan. Quant, Anal. 35:529–552Elton E, Gruber M, Comer G, Li K (2002) Spiders: Where are the Bugs? J. Bus. 3:453–472George T, Kaul G, Nimalendran M (1991) Estimation of the Bid-Ask Spread and Its Components: A New

Approach. Rev. Financ. Stud. 4:623–656Glosten LR, Harris LE (1988) Estimating the Components of the Bid-Ask Spread. J. Financ. Econ.

21:123–42Gorton G, Pennacchi G (1993) Security Baskets and Index-linked Securities. J. Bus. 66:1–27Hamilton J (1978) Marketplace Organization and Marketability: NASDAQ, the Stock Exchange, and the

National Market System. J Financ 33:487–503Hasbrouck J (2003) Intraday Price Formation in U.S. Equity Index Markets. J Financ 58:2375–2399Hegde S, McDermott J (2004) Market Liquidity of Diamonds. Q’s and Their Underlying Stocks, J. Bank.

Financ 28:1043–1067Huang R, Stoll H (1996) Dealer Versus Auction Markets: A Paired Comparison of Execution Costs on

NASDAQ and the NYSE. J. Financ. Econ. 41:313–358Jiang CX, Kim JC (2005) Trading Costs in Non-U.S. Stocks on the New York Stock Exchange: The Effect

of Institutional Ownership, Analyst Following, and Market Regulation. J Financ Res 3:439–459Kumar P, Seppi DJ (1994) Information and Index Arbitrage. J. Bus. 67:481–509Kumar R, Sarin A, Shastri K (1998) The Impact of Options Trading on the Quality of the Market for the

Underlying Security: An Empirical Analysis. J Financ 53:717–732Kyle A (1985) Continuous Auctions and Insider Trading. Econometrica 53:1315–1335Lin J, Sanger G, Booth G (1995) Trade Size and Components of the Bid-Ask Spread. Rev. Financ. Stud.

8:1153–1183Lipper Investments, 2001, http://www.closed-endfunds.comMarshall B, Cahan R, Cahan J (2008) Does Intraday Technical Analysis in the U.S. Equity Market Have

Value? J. Empir. Finance 15:199–210Neal R, Wheatley S (1998) Adverse Selection and Bid-Ask Spreads: Evidence from Closed-End Mutual

Funds. J. Financ. Mark. 1:121–149Nguyen V, Van Ness B, Van Ness R (2007) Inter-market Competition for Exchange Traded Funds. J.

Econ. Finance 31:251–267Novakoff, J., 2001, Exchange Traded Funds, White Paper, 1–3Small, K., Flaherty, S., and Ionici, I., 2005, U.S. Cross-listing and Market Liquidity: ATest of the Bonding

Hypothesis, Loyola College Working Paper.Small, K., and Wansley, J., 2005, An Examination of Informed Trader Migration, Loyola College Working

PaperSubrahmanyam A (1991) A Theory of Trading in Stock Index Futures. Rev. Financ. Stud. 4:17–51Tinic S (1972) The Economics of Liquidity Services. Q. J. Econ. 86:79–93Tse Y, Bandyopadhyay P, Shen Y (2006) Intraday Price Discovery in the DJIA Index Markets. J. Bus.

Finance Account. 33:1572–1585Van Ness B, Van Ness R, Warr R (2001) How Well Do Adverse Selection Components Measure Adverse

Selection? Financ. Manage. 30:77–98Van Ness B, Van Ness R, Warr R (2005a) The Impact of the Introduction of Index Securities on the

Underlying Stocks: The Case of Diamonds and the Dow 30. Adv. Quant. Anal. Finance Account.2:105–129

Van Ness B, Van Ness R, Warr R (2005b) The Impact of Market Maker Concentration on Adverse-Selection Costs for NASDAQ Stocks 3. J Financ Res 3:461–485

J Econ Finan (2012) 36:261–281 281