This is an author-deposited version published in:
http://oatao.univ-toulouse.fr/
Eprints ID: 5808
To link to this article: DOI: 10.1111/j.1468-036X.2010.00553.x
URL:
http://dx.doi.org/10.1111/j.1468-036X.2010.00553.x
To cite this version:
Germain, Laurent and Kluger, Brian and Crina,
Pungulescu and Stolin, David Intra-Dealer Integration. (2010) European
Financial Management, Vol. 16, No. 4, 2010, 507–527, vol. 16 (n° 4). pp.
507-527. ISSN 1354-7798
O
pen
A
rchive
T
oulouse
A
rchive
O
uverte (
OATAO
)
OATAO is an open access repository that collects the work of Toulouse researchers and
makes it freely available over the web where possible.
Any correspondence concerning this service should be sent to the repository
administrator:
[email protected]
Intra-Dealer Integration
Laurent Germain
Universite de Toulouse, Toulouse Business School, 20 Boulevard Lascrosses, Toulouse 31068 Cedex, France and ISAE
Email: [email protected]
Brian Kluger
University of Cincinnati, University of Cincinnati, ML 195, Cincinnati, OH 45221-0195, USA Email: [email protected]
Crina Pungulescu
ESEC Toulouse Barcelona Business School, C/ Trafalgar 10, 08010 Barcelona, Spain Email: [email protected]
David Stolin
Toulouse Business School, 20 Boulevard Lascrosses, Toulouse 31068 Cedex, France Email: [email protected]
Daniel Weaver
Rutgers Business School, Rutgers University, 94 Rockafeller Road, Piscataway, NJ 08854-8054, USA Email: [email protected]
Abstract
This paper examines the quotation behaviour of dealers who made markets in the same stocks on both NASDAQ and either EASDAQ or the LSE. Whereas previous studies examine international integration at the market level, we examine integration at the dealer level. In other words, do dealers within the same market-making firm use information from their arm on the opposite side of the Atlantic in forming their own quotes? We find that while there is some evidence of integration at the market level, integration is hard to detect at the dealer level. The results are largely unaffected by differences in fungibility between our two samples.
1. Introduction
Over the past decades, there has been a clear movement towards globalisation of both corporations and securities markets. Investors in one country are increasingly able to
We are very grateful to two anonymous referees, Stephane De Maght of the Belgian Banking and Finance Commission, Vladimir Atanasov, Kuan-Hui Lee, Tim McCormick, Albert Menkveld, Dirk Tirez and his colleagues at EASDAQ, and Ingrid Werner for helpful comments and conversations. Kyriacos Lambrias provided outstanding research assistance. Weaver thanks the Whitcomb Center for Research in Financial Services for providing finan-cial support. All errors are ours. Correspondence: Daniel Weaver.Intra-Dealer Integration
easily purchase securities in other countries. In addition to securities markets becoming more global, market-making firms have also become global. There are now a number of market-making companies with operations in multiple markets around the world. The question we study is whether such firms operate in an integrated manner. This paper is a first look at international intra-dealer integration.
Previous studies of cross-listed shares trading on home and foreign markets generally
conclude that markets are less than fully integrated.1Several reasons have been advanced
as to why. In broad terms, these reasons fall into two categories; fungibility and home bias. Fungibility reasons focus on the barriers to buying cross-listed shares in one market and reselling on the other market. For example, in many instances, there are extra transactions costs or conversion fees incurred when trading across markets. In the USA, foreign shares are often cross-listed as American Depository Receipts (ADRs). An ADR is a financial instrument issued by a US bank representing foreign shares held in trust by the bank, and there is a conversion fee required to assemble or disassemble the ADR.
Another transaction cost example is the stamp duty reserve tax on UK shares.2This tax
would apply to a trader trying to buy and sell shares cross-listed in the UK. Still another fungibility issue is that the majority of cross-listed shares trade in different currencies in the home and foreign markets.
Conversely, the home-bias explanation of why markets are not fully integrated focuses on inherent differences between the two markets. There can be cultural and language differences between traders in home and foreign markets. These differences can potentially affect the release, dissemination and interpretation of information. However, as Chowdhry and Nanda (1991) and Menkveld (2008) point out, traders can differ in their ability to cross markets. For example, informed traders such as institutions may have the ability to easily access liquidity in foreign markets.
Similar to Chowdhry and Nanda (1991) and Menkveld (2008), we hypothesise that not all dealers have the same fungibility and information constraints. Dealers with multinational arms may be able to partially bypass fungibility barriers because they ostensibly maintain a supply of shares in both the markets where the cross-listed share is
traded.3Further, such firms might plausibly share information about stocks. If the home
dealer has superior information about a cross-listed stock, that information may be shared with the foreign arm of the same market-making firm. We call this type of information sharing intra-dealer integration, which in turn can promote market integration.
To examine the degree of intra-dealer integration, we analyse two samples containing cross-listed stocks on NASDAQ and EASDAQ, and NASDAQ and the LSE respectively. These two samples differ in their level of fungibility, allowing us to disentangle this impact. All companies in both samples have at least one market-making desk that operates on both sides of the Atlantic. We analyse these market-maker quotes to see whether there
1See for example Eun and Sabherwal (2003), Grammig et al. (2005), Hupperets and
Menkveld (2002), Karolyi (2006), Moulton and Wei (2009) and Phylaktis and Korczak (2004).
2http://www.direct.gov.uk/MoneyTaxAndBenefits/Taxes/TaxOnSavingsAndInvestments/
TaxOnSavingsAndInvestmentsArticles/fs/en?CONTENT_ID=10013514&chk=Tac6CP
3Porter et al. (2008) examine another interesting aspect of dealers with multinational arms.
They show that dealers with operations in multiple countries can avoid disruptions to trading by shifting trading to a non-affected location in the event of such interruptions as a power failure.
is any evidence of integration between multinational arms of these market-making firms. Surprisingly, we do not find much evidence that a dealer is influenced by the presence of colleagues in the same firm but on the other side of the Atlantic. We start by establishing market-level integration for our sample stocks and documenting the relative contributions of NASDAQ vs. EASDAQ/LSE to price discovery both over the full day and during the trading overlap. We then check whether dealers whose affiliates make markets in the stock on the other side of the Atlantic (henceforth ‘global’ or ‘affiliated’ dealers) are able to post more competitive quotes, i.e. whether their bid-ask spreads are lower and whether they spend more time alone on the inside of the market’s bid-ask spread than do unaffiliated dealers. The answer turns out to be no. Further, affiliated dealers do not appear to post more informative quotes when the respective markets open. Finally, we check more directly for evidence of global dealers’ coordination by studying whether, during the trading overlap, their European and American arms are more likely to be on the same side of the efficient price than a control pair of unrelated dealers. Again, the answer turns out to be no. Further, dealers fail to be influenced by the presence of foreign affiliates even when the affiliate is in the dominant market (in terms of price discovery) for the stock in question. In short, global securities firms do not make markets globally.
In the next section we discuss the literature related to this study. Section 3 describes the structure of the markets we examine, while Section 4 describes our sample selection criteria and data. Section 5 reports the results of our analysis and Section 6 concludes.
2. Related Literature
Although there is no literature on multinational market maker integration, there are studies that examine integration of domestic market making firms. For example, Naik and Yadav (2003) investigate whether market-making firms on the London Stock Exchange adopt a firm-wide portfolio approach in managing inventories. They conclude that these firms instead allow individual market makers to manage their inventories on a stock-by-stock basis. In contrast to Naik and Yadav, Coughenour and Saad (2004) find that on the NYSE, specialist firms do appear to employ a firm portfolio approach in managing individual stock inventories. Anand (2005) examines market quality measures for specialists employed by the same firm making markets on competing exchanges. He finds significant differences in the way in which the specialists compete for business, but that execution does not seem to vary significantly within a firm. The above studies then differ in their conclusions concerning market-making firm integration, with those examining US firms suggesting a higher level of intra-dealer integration than do the studies of UK firms. However, none of the studies examines the type of intra-firm integration studied in this paper – the co-movement of quotes within dealer firms but across markets for identical assets.
Although little investigation of intra-dealer quote integration has been produced to date, a relatively large body of research examines inter-market integration (sometimes referred to as fragmentation), where stocks are traded in more than one market. In the USA, Battalio et al. (1997) show that despite the substantial diversion of order flow to regional stock exchanges from the NYSE, market quality is relatively unaffected. Hasbrouck (1995) shows that for the thirty Dow stocks, nearly all price discovery takes place on the NYSE, with the regional exchanges contributing very little. Fong et al. (2000) study the determinants of off-market trading in Australian stocks, and find that
off-market volume is strongly related to various measures of liquidity on the primary market.
Internationally, Werner and Kleidon (1996) examine the intra-day volatility, spread and volume patterns for some British companies’ shares on the London Stock Exchange and their ADRs on the New York Stock Exchange and conclude that these markets are not integrated. Hupperets and Menkveld (2002) report similar findings for Amsterdam/ NYSE cross-listings. Further, they find that a given market’s share of price discovery during the overlap period (using the Hasbrouck (1995) methodology) varies widely across the stocks in their sample. Their overall conclusion is that the two markets are not perfectly integrated. Domowitz et al. (1998) study dual listed Mexican stocks at home and in the USA, and find that the fragmentation of order flow has an adverse effect on liquidity, but that the cross-border competition reduces spreads. Pulatkonak and Sofianos (1999) study the determinants of the market share of NYSE in the trading of multiple-listed non-US stocks. They find that the relative volume traded in the USA is strongly inversely related to the time zone difference between the home market and New York.
The theoretical literature contains a number of predictions regarding the effects of market fragmentation on security trading. Chowdhry and Nanda (1991) consider a model where the informed trader and some, but not all, liquidity traders have discretion to choose in which market to trade. At the equilibrium, there is a concentration of trading. Small liquidity traders with discretion over the location of their trade, concentrate their trades in the market that has the largest amount of trading by other liquidity traders who are unable to move to another market. In turn, this market will attract additional informed traders and liquidity traders. However, this model does not consider competition between the exchanges. Menkveld (2008) presents a model that specifically addresses instances where securities are traded on markets with overlapping trading hours. In the presence of sufficient (local) liquidity trading, informed traders who have access to both markets will split their orders across the two markets, and trading will concentrate in the overlap. Also related to this study, is the strand of literature devoted to price discovery for cross-listed stocks. These studies tend to examine a focused group of cross-listed firms. The results of these studies are mixed; with some showing that price discovery occurs primarily in the home country, while others find a larger share of price discovery in the foreign market. Among those finding home country price discovery dominance are Kato
et al. (1990) who examine 23 stocks cross-listed in New York (eight from Australia, eight
from Japan, and seven from the UK) The authors conclude that home country prices lead New York prices. Grammig et al. (2004) examine three German stocks cross-listed on the New York Stock Exchange and find results consistent with Kato et al. In contrast, Lau and Diltz (1994) find, for a sample of seven cross-listed Japanese stocks, that the foreign market (NYSE) dominates the home market in price discovery. Finally, Lieberman et al. (1999) obtain mixed results. They find that for five of the six cross-listed Israeli stocks in their sample, the Israeli market dominates the US market, while for one stock the reverse is true. Therefore, the directionality of price discovery for cross-listed stocks appears to vary depending on the stock traded, the exchanges, or both.
Based on the studies cited above, the general conclusion is that markets are less than fully integrated. There is also mixed evidence concerning domestic dealer firm integration with regard to inventory management. In this paper, we examine another dimension of intra-dealer integration – quotation behaviour. Do dealer firms allow market-making arms in different countries to establish quotes independently – similar to what Naik and Yadav (2003) find for individual dealer inventories? Or is there some
evidence of integration through information sharing – similar to what Coughenour and Saad (2004) find for specialists’ inventories? Intra-dealer quote integration would result in separate arms of the same dual-market dealer firm setting their quotes for identical assets in a like manner. Further, integrated dealers will quote more aggressively than unaffiliated dealers if integrated dealers can receive private information from their cross-Atlantic siblings. In the following sections we analyse our data for evidence of intra-dealer integration.
3. Markets and Integration
3.1 Institutional background
In this section, we briefly describe the EASDAQ market and compare the microstructures of EASDAQ and the London Stock Exchange to that of NASDAQ. For more institutional information, see Smith et al. (1998) and Barclay et al. (1999) for NASDAQ, Tirez (1999) for EASDAQ, and Werner and Kleidon (1996) for the London Stock Exchange.
In 1994 the European Association of Securities Dealers (EASD), created a European counterpart to NASDAQ, called EASDAQ. After securing the necessary approvals, the EASDAQ market opened for trading in November 1996. NASDAQ helped to create EASDAQ, and it also took a small stake in the EASDAQ market. In April 2000, NASDAQ acquired a majority stake in EASDAQ, renaming it NASDAQ Europe. In 2003 NASDAQ closed NASDAQ Europe.
3.2 Structural description
Not surprisingly, EASDAQ, the LSE, and NASDAQ are structurally similar. All are quote driven markets, although NASDAQ includes public limit orders and alternative trading system quotes (such as those from Island or Instinet) in the calculation of the best bid and offer. EASDAQ and the LSE are pure dealer markets during the time of our
study.4
The market structures of all three markets are closer than in many previous studies
of market integration.5 All three are decentralised markets where trades occur over the
phone or via e-mail (except for trades on NASDAQ ATSs that occur automatically). For trades done over the telephone, dealers on both markets know the identity of the broker placing the order – in the case of institutional trades this often means knowing the identity of the trader. In addition, all markets have similar post trade transparency. On NASDAQ transactions have to be reported to the market within 90 seconds and on both EASDAQ and the LSE within 180 seconds. On NASDAQ and EASDAQ the trade price and volume are disseminated as soon as the report is received. Block trades on the
4The LSE began SETS in October 1997 for the 100 most active LSE stocks. SETS is a
central limit order book that allows public limit order traders to compete with dealers on a time priority basis. Since then the LSE has included more stocks and established another system called SETSmm which is a combination of SETS and a dealer market. However, during 1999 all of our sample stocks were traded in a pure dealer market on the LSE.
5For example, Werner and Kleidon (1996) compare the London Stock Exchange (a pure
LSE are subject to delayed reporting to give market makers a chance to unwind their positions.
During the period of our study, on NASDAQ the minimum tick size is $1/32 or $1/16 while our sample of EASDAQ and the LSE stocks have a minimum tick of $0.01 and GBP0.01, respectively. EASDAQ-listed firms were allowed to decide on the currency in which its shares were quoted. In 1999, the only currencies used were US dollars, British pounds and Euros.
4. Sample Selection and Data
We identify firms that trade on NASDAQ and either EASDAQ or the London Stock Exchange during all of 1999. There are 15 such firms listed on EASDAQ/NASDAQ during 1999. As mentioned above, firms listing on EASDAQ can choose what currency their stocks are traded in. Previous studies of price discovery have found that exchange
rates explain a portion of innovation in price.6 Since a part of our study examines
the relative contribution of each market to price formation, we eliminate exchange risk by only choosing those stocks that trade on EASDAQ in US dollars. Three firms are then excluded that trade in Euros. As Werner and Kleidon (1996) point out, lack of full fungibility may impede full integration between markets by limiting
arbitrage opportunities.7The primary characteristic of fully fungible securities is that the
certificate traded in both markets is identical.8Therefore we further limit our EASDAQ
sample to those cases where the common stock of a firm is traded on both EASDAQ and NASDAQ. Five firms traded as depository receipts on EASDAQ and/or NASDAQ and one as trades as a share of beneficial interest. These firms are excluded from our sample. Our final selection criterion is that the security only be listed on EASDAQ and NASDAQ. This requirement allows us to limit the influence of exogenous factors. Eliminating the one firm that had three stock listings leaves us with five stocks in our EASDAQ/NASDAQ sample. All of the stocks in our EASDAQ/NASDAQ sample have
at least one dealer firm making markets in the stock in both markets.9
6See Kim et al. (2000), Grammig et al. (2005), and Phylaktis and Korzak (2004), among
others.
7See Pulatkonak and Sofianos (1999) and Gagnon and Karolyi (2006) for a discussion of
what constitutes full fungibility.
8According to Pulatkonak and Sofianos (1999), another characteristic is that no legal
restriction on cross-border trading and ownership exist. US citizens are not allowed to directly invest in foreign securities. However, as Werner and Kleidon (1996) point out, these restrictions are irrelevant for institutions who regularly trade in multiple markets simultaneously. Therefore, there are no real legal restrictions on cross-border ownership for institutions. One cost that does exist for institutions is the cost of clearing. EASDAQ securities were cleared and settled through the Cedelbank division of Euroclear while NASDAQ traded securities are cleared through a number of institutions and settled through the Depository Trust Corporation. Although the need to clear through multiple institutions increases the cost of arbitrage, previous studies of market integration have also considered these costs.
9
Even if we do not exclude ADRs and stocks listed on EASDAQ and NASDAQ in different currencies but simply form the EASDAQ/NASDAQ sample on the basis of data availability (excluding stocks listed for only a fraction of the year and those averaging less than one dealer quote per day during the overlap), the final sample will consist of the same five fully-fungible same-currency pairs of securities.
Stocks traded on the LSE are all traded in pound sterling. Therefore, we cannot impose the same currency requirement that we impose on EASDAQ stocks. However, we do impose the restriction that sample LSE stocks are not traded on any markets other than the LSE and NASDAQ. Seven stocks meet this requirement during the period of our study, but two did not have a dealer firm operating in both markets, so they are excluded. Lastly, the five sample firms are British and are traded on NASDAQ in the form of ADRs. Thus the main difference between our EASDAQ/NASDAQ and LSE/NASDAQ
samples is fungibility.10
Our selection criteria result in five stocks that trade on both NASDAQ and EASDAQ and five that trade on NASDAQ and the LSE. These sample sizes are similar to a number of previous studies that examine international market integration (See Kato et al. (1990), Grammig et al. (2004), Lau and Diltz (1994), Lieberman et al. (1999), and Karolyi and Stultz (1996), among others.) EU data is obtained from EASDAQ and from the LSE quote and trade disks for 1999. NASDAQ data are obtained from the 1999 NASTRAQ CDs. All three data sets are similar and contain time stamped trade data which indicate the price and quantity of each trade. Quote data includes all market maker quotes so that we are able to identify individual dealers in each stock. This is in contrast to some previous studies that could only examine the best bid and ask for a market (e.g., Werner and Kleidon, 1996).
Descriptive statistics for our two samples are contained in Table 1. Our set of dual-listed firms contains 3 US firms and 7 European firms. We improve the generality of our tests by examining our firms for an entire calendar year. Overall, our sample contains over 675,000 trades and over 685,000 quotes for the sample period.
Table 1 contains information regarding the number of dealers that make markets in our sample stocks during 1999. The number ranges from 6 dealers for NTL trading on EASDAQ to 42 dealers on NASDAQ for Danka Business Systems. The last 2 columns of Table 1 report the number of dealers that make markets in both the US and EU. There is at least one dealer making markets on both continents and the number reaches a high of 8 for Shire Pharmaceuticals.
As in previous studies such as Grammig et al. (2004) our analysis focuses on the period when both members of a market pair are actively trading. During the period of our study, the EASDAQ market is open from 09:30 to 16:30 Central European Time corresponding to 03:30 to 10:30 EST (except for the period from the last Sunday of March through the first Sunday of April, when this corresponds to the period from 02:30 to 09:30 EST). Quotes and trade prints were disseminated in real time through EASDAQ’s own system as well as through information vendors such as Bloomberg, Bridge, DataStream and Reuters; however, all trading was done by telephone.
The London Stock Exchange is open from 8:00 to 16:30 Western European Time corresponding to 03:00 to 11:30 EST (except for the period from the last Sunday of March through the first Sunday of April, when this corresponds to the period from 02:00 to 10:30 EST). NASDAQ is open from 9:30 to 16:00 EST. Therefore, for most of the year there is a one-hour overlap in trading hours on EASDAQ/NASDAQ and 2 hours on LSE/NASDAQ. Focusing on quotes in these overlap periods allows us to examine inter-market and intra-dealer integration most directly. We present a description of our methodology and the results of this analysis in the next section.
10Recall that the stamp-duty tax is a major impediment to fungibility for LSE-NASDAQ
Ta b le 1 Descripti v e statistics. This tab le presents descripti v e statistics for tw o samples of stocks that are duall y listed on mark et mak er based mark ets in Europe and the USA. F iv e st ocks are duall y listed on EASD A Q and N ASD A Q (P anel A) and fi v e are duall y listed on N ASD A Q and the London Stock Exchange (P anel B). C olumns contain the compan y’ s countr y of incor poration, the mark et v alue (in US$ millions) at the be ginning of 1999 and the tick er symbol in each mark et in w hich the stock is traded. T he number of dealers quoting the stock during 1999 as w ell as the dealer fir ms that operate in both N ASD A Q and EASD A Q or the LSE are also listed. F inall y, w e list the number of quotes and trades for each stock in both mark ets. USA Europe Common Mark et Mak ers Compan y Countr y V alue T ick er Dealers T rades Quotes T ick er Dealers T rades Quotes # Names A. Stoc ks Dual-listed on N ASD A Q and EASD A Q 4F ront T echnolo gies US 107.0 FFTI 21 48,290 42,198 FFTI 8 3,629 7,779 1 Herzo g ICOS V ision Systems Belgium 144.4 IVIS 10 5,144 7,929 IVIS 7 1,726 3,673 3 Herzo g, SG, Rober tson Stephens Ler nout & Hauspie Speech Products Belgium 1,839.3 L HSP 33 155,681 176,814 LHSP 16 32,138 38,591 3 Herzo g, SG, Rober tson Stephens NTL US 3,390.0 NTLI 25 140,348 156,337 NTLI 6 215 5,993 1 Herzo g Pixtech US 36.0 PIXT 13 29,324 12,798 PIXT 10 1,333 4,787 1 Herzo g B . Stoc ks Dual-listed on N ASD A Q and the LSE Danka Business Systems UK 227.0 D ANKY 42 124,795 119,713 DNK 7 6,329 9,712 4 Herzo g, Mer rill, Salomon, W arbur g Merant UK 250.2 M RNT 21 20,024 24,522 MRN 9 30,789 15,893 6 CSFB , DLJ , Goldman, Herzo g, DKB , Mer rill Shire Phar ma-ceuticals Group UK 903.2 SHPGY 19 19,337 17,555 SHP 13 10,249 11,774 8 Bear Ster ns, CSFB , DLJ , DMG, Goldman, Herzo g, Lehman, Mer rill Signet Group UK 861.2 SIGYY 12 1,402 5,115 SIG 13 16,098 13,818 4 DMG, Goldman, Herzo g, Mer rill, Wa rb u rg Sk y ephar ma UK 517.4 SKYE 9 3,986 5,497 SKP 9 28,717 8,290 3 Herzo g, Salomon, SG
5. Methodology and Results
5.1 Market integration
Before examining intra-dealer integration across multiple markets, we analyse the integration of the markets themselves, and seek to gauge the dominant market for each of our stocks. As a first step in this direction, we test for cointegration between European and US market prices for each of our sample stocks. To do so, we sample midquotes during the trading overlap at 30-second intervals and conduct Augmented Dickey-Fuller tests on the levels and first differences of the series of quotes. The null hypothesis of a unit root can be rejected only for the first differenced series and not for the series in level, which suggests that the series are integrated of order 1, hence we can test for the existence of a cointegration relationship between the European and American price quotes for all the stocks. Using the Johansen Trace and Maximum Eigenvalue statistics, the hypothesis that there is a cointegration relationship between European and US prices
cannot be rejected at the 0.05 level for any of the ten stocks we study.11
We next calculate Weighted Price Contributions (henceforth WPC, see Barclay and Hendershott (2003)) for each stock. These show what proportion of the total change in stock price is accounted for by different time periods. We split the 24-hour day into the following periods: US market close to European market open (i.e. the overnight period), European open to US open (i.e. the portion of the day when only the European market is open), US open to European close (i.e. the trading overlap), and European close to US close (i.e. the portion of the trading day when only the US market is open). As the basis for WPC calculations, we use inside quote midpoints of the open market (or their average when both markets are open). The results are presented in Table 2.
For the EASDAQ-NASDAQ sample (Panel A), we find that much more price discovery takes place during the US-only period than in the European-only period. Even for LHSP, where the share of the European-only period is the highest (0.248), US-only trading contributes nearly twice as much (0.451) to price discovery. The overlap period’s share is consistently around 20 percent (0.197 – 0.242). Thus, on a per-hour basis, the overlap period contributes the most to price discovery in each of the five stocks. The fact that price formation accelerates when a maximum number of market participants are involved may
be the result of the greater fungibility of shares for the EASDAQ-NASDAQ sample.12
Looking at the LSE-NASDAQ sample (Panel B), we find that once again, on a per-hour basis, the overlap period contributes the most to price discovery in each of the sample stocks. With the exception of MRNT and SIGYY, most of the remaining price discovery (as with our EASDAQ/NASDAQ sample) takes place when only NASDAQ is open. Although allocation of price discovery by time period is suggestive of the roles different markets play, a more direct way to examine which market is dominant for a given stock is to examine quote co-movement during the overlap period by using the Hasbrouck (1995) methodology. Table 3 shows the Hasbrouck (1995) information share range for each market in each of the stocks.
11
Results not reported here, but are available from the authors upon request.
12Greater fungibility should facilitate order splitting across markets. Menkveld (2008) shows
that the overlap period should attract increased activity from traders who split orders across markets, and that transaction prices become more revealing during the overlap owing to competition between informed traders of both markets.
Table 2
Weighted price contributions.
This table reports Weighted Price Contributions for each stock. The calculations follow Barclay and Hendershott (2003), and are based on inside midquotes of the open market, or the average of the midquotes when both markets are open.
US close - EU open - US open - EU close
-Market Ticker EU open US open EU close US close
A. Stocks dual-listed on NASDAQ and EASDAQ
Easdaq FFTI 0.046 0.122 0.229 0.602
Easdaq IVIS 0.178 0.181 0.213 0.421
Easdaq LHSP 0.056 0.248 0.226 0.451
Easdaq NTLI 0.077 0.105 0.197 0.614
Easdaq PIXT −0.032 0.144 0.242 0.628
B. Stocks dual-listed on NASDAQ and the LSE
LSE DANKY 0.047 0.143 0.366 0.448 LSE MRNT 0.145 0.341 0.234 0.276 LSE SHPGY 0.042 0.201 0.361 0.393 LSE SIGYY 0.362 0.451 0.143 0.046 LSE SKYE 0.252 0.251 0.220 0.282 Table 3
Information shares of European and US dealers.
This table reports, for each stock, Hasbrouck (1995) information shares for European and US markets, i.e. their proportional contributions to innovations in the common efficient price. The minimum and maximum information share bounds for each stock are obtained following Baillie et al. (2002). They are based on midpoints of best quotes for each market and are estimated for all trading days (after eliminating days with no more than three updates on each side of the market), with overnight periods excluded from the estimation. Quotes were sampled at 30-second intervals.
EU information share
US information share
Markets Nasdaq Ticker min max min max
A. Stocks dual-listed on NASDAQ and EASDAQ
Easdaq/Nasdaq FFTI 0.308 0.345 0.655 0.692
Easdaq/Nasdaq IVIS 0.496 0.547 0.453 0.504
Easdaq/Nasdaq LHSP 0.273 0.341 0.659 0.727
Easdaq/Nasdaq NTLI 0.150 0.174 0.826 0.850
Easdaq/Nasdaq PIXT 0.584 0.600 0.400 0.416
B. Stocks dual-listed on NASDAQ and the LSE
LSE/Nasdaq DANKY 0.143 0.166 0.834 0.857
LSE/Nasdaq MRNT 0.346 0.358 0.642 0.654
LSE/Nasdaq SHPGY 0.150 0.166 0.834 0.850
LSE/Nasdaq SIGYY 1.000 1.000 0.000 0.000
Starting with LSE-NASDAQ stocks, the stock that stands out is again SIGYY, where NASDAQ has a 0.000 information share during the overlap. This is consistent with both the particularly small contribution of the NASDAQ-only time period to price formation in this stock (Table 2) and with this stock having by far the lowest ratio of the number of US trades to the number of European trades (Table 1). Similarly, for SKYE, the only other LSE-NASDAQ stock where the LSE has an overwhelming share of trading, the European market has the greater information share (0.75). For the other three stocks, DANKY, MRNT, and SHPGY, NASDAQ is the dominant market in the overlap period. For the EASDAQ/NASDAQ sample, information share results similarly make sense given the distribution of trading across the two markets. Across the five stocks, NASDAQ’s information share is by far the highest, at around 84%, for NTLI, for which the ratio of the number of US trades to European ones is also by far the highest. NASDAQ’s information share is also above a half for FFTI and LHSP, while price discovery in IVIS and PIXT during the trading overlap is mainly EASDAQ-driven.
Overall, examining market integration for our two samples shows that neither sample is overwhelmingly dominated by a single exchange (as is typically the case when one studies stocks dual-listed on US regional exchanges). The fact that we have stocks where most information during the trading day overlap appears to originate from Europe (IVIS, PIXT, SIGYY, and SKYE) as well as those where most information comes from the USA (FFTI, LHSP, NTLI, DANKY, MRNT, SHPGY) suggests that having an affiliated dealer on either side of the Atlantic has the potential to be beneficial. Specifically, a dealer might benefit from receiving information from its affiliate on the other, dominant side of the market. Does this potential benefit of intra-dealer integration translate into actual quotation behavior? We now examine this issue.
5.2 Intra-dealer integration
Recall that Chowdhry and Nanda (1991) and Menkveld (2008) argue that not all traders have the same information and ability to trade in foreign markets. Ceteris paribus, dealers with foreign affiliates have the ability to trade in multiple markets simultaneously and the potential to share information cross-border. The degree of information sharing by affiliated dealers remains an empirical question, which this study seeks to answer. In order to examine the effect of dealer’s affiliation on their behavior, we need to construct a control sample of unaffiliated dealers. Accordingly, we match affiliated dealers to their controls based on quotation frequencies (the procedure is explained and the matches are listed in the Appendix). We start by examining the effect of dealer affiliation on quoted spreads. To do so, we calculate the best bid and the best ask for the group of affiliated dealers in each stock, and separately for the control set of unaffiliated dealers. Percentage spreads are based on these best bids and asks.
The results are in Table 4, which, for each market, shows spreads for the market opening (OPEN), the period before the trading overlap in the case of European markets (BEFR), the trading overlap (OVRL), the period after the trading overlap in the case of NASDAQ (AFTR), and the market closing (CLOS). For each of these four periods, and for each stock, four numbers are reported: the average percentage spread for affiliated dealers, the average percentage spread for unaffiliated control dealers, the difference between the two spreads, and the (two-tailed) p-value for the difference.
If affiliated dealers are integrated, there are two reasons to expect them to generate lower spreads than do otherwise similar unaffiliated dealers. First, the pooling of information sources across the Atlantic could be expected to reduce adverse selection
Ta b le 4 Dealer af filiation and quoted spreads. This tab le examines the ef fect of dealer af filiation on quoted spreads. F or each stock, w e constr uct the best bid and ask separatel y for af filiated dealers and for the control set of non-af filiated dealers (these are listed in the appendix). W e then calculate the spread for each g roup as the time-w eighted ratio of 2 ∗(best ask − best bid)/(best ask + best bid). W e subdi vide the trading da y into the follo wing time periods: the mark et’ s opening (OPEN), the period from the opening until the star t of the ov erlap (BEFR), the ov erlap (O VRL), the period from the end of the ov erlap until the mark et’ s closing (AFTR) and the mark et’ s closing (CLOS). F or each of these time periods, the tab le sho ws four numbers: the av erage spread for af filiated dealers, the av erage spread for control dealers, the dif ference b etw een these tw o spreads, and the p-v alue for the dif ference based on the sign rank test with the trading da y as the unit of obser v ation. Mark et T ick er OPEN BEFR O VRL AFTR CLOS A. Stocks dual-listed on N ASD A Q and EASD A Q Easdaq FFTI 5.34 4.35 0.99 0.00 5.31 4.25 1.07 0.00 5.28 4.30 0.98 0.00 5.28 4.28 0.99 0.00 Easdaq IVIS 5.87 5.35 0.52 0.00 5.58 5.11 0.47 0.00 5.39 4.91 0.48 0.00 5.37 4.85 0.52 0.00 Easdaq LHSP 2.28 1.95 0.34 0.00 2.15 2.03 0.12 0.00 1.97 1.83 0.14 0.00 1.95 1.77 0.18 0.00 Easdaq NTLI 4.71 3.72 1.00 0.00 4.78 3.78 1.00 0.00 4.80 3.70 1.10 0.00 4.80 3.67 1.13 0.00 Easdaq PIXT 9.39 8.30 1.09 0.00 9.56 8.25 1.32 0.00 9.51 8.24 1.27 0.00 9.51 8.27 1.24 0.00 Nasdaq FFTI 6.87 5.26 1.61 0.00 6.87 5.25 1.62 0.00 6.87 5.10 1.77 0.00 7.21 5.20 2.01 0.00 Nasdaq IVIS 3.66 4.34 − 0.68 0.00 3.57 3.79 − 0.22 0.28 3.56 3.67 − 0.11 0.88 3.63 3.61 0.02 0.59 Nasdaq LHSP 1.36 1.64 − 0.29 0.00 1.47 1.78 − 0.32 0.00 1.50 1.53 − 0.03 0.57 1.61 1.64 − 0.03 0.89 Nasdaq N TLI 4.18 2.81 1.37 0.00 4.22 3.00 1.22 0.00 4.57 3.16 1.41 0.00 4.27 2.88 1.39 0.00 Nasdaq PIXT 15.45 21.25 − 5.79 0.00 15.93 21.17 − 5.24 0.00 16.15 20.58 − 4.43 0.04 16.27 20.91 − 4.64 0.06 B . Stocks dual-listed on N ASD A Q and the LSE LSE DNK 3.89 3.95 − 0.06 0.43 4.09 4.09 0.00 0.64 3.96 3.94 0.01 0.91 3.91 3.93 − 0.01 0.99 LSE MRN 4.32 4.31 0.01 0.00 4.25 4.23 0.02 0.00 4.25 4.24 0.01 0.02 4.28 4.26 0.02 0.07 LSE SHP 1.54 1.40 0.14 0.00 1.58 1.50 0.08 0.00 1.58 1.47 0.11 0.00 1.56 1.44 0.13 0.00 LSE SIG 2.19 2.23 − 0.04 0.35 1.93 2.15 − 0.23 0.00 1.89 2.17 − 0.28 0.00 1.86 2.14 − 0.28 0.00 LSE SKP 4.97 4.25 0.73 0.00 4.84 4.32 0.52 0.00 4.80 4.14 0.65 0.00 4.79 4.12 0.67 0.00 Nasdaq D ANKY 3.67 5.63 − 1.96 0.00 3.68 5.27 − 1.59 0.00 3.45 4.79 − 1.34 0.00 3.42 4.90 − 1.48 0.00 Nasdaq MRNT 3.09 2.57 0.52 0.00 2.77 2.75 0.02 0.97 2.68 2.69 0.00 0.81 2.76 2.63 0.14 0.32 Nasdaq SHPGY 2.35 3.35 − 1.00 0.00 2.16 3.29 − 1.13 0.00 2.06 3.18 − 1.12 0.00 2.03 3.08 − 1.05 0.00 Nasdaq SIGYY 2.97 3.91 − 0.93 0.00 2.90 3.76 − 0.86 0.00 2.85 3.66 − 0.81 0.00 2.85 3.61 − 0.76 0.00 Nasdaq SKYE 4.18 6.72 − 2.54 0.00 4.22 6.50 − 2.28 0.00 4.13 6.49 − 2.36 0.00 4.14 6.43 − 2.29 0.00
costs. Second, being able to offset inventories could be expected to reduce inventory costs. There is little evidence, however, that dealer affiliation consistently contributes to narrower spreads. On the EASDAQ side, all EASDAQ-NASDAQ stocks have wider spreads generated by affiliated dealers than by unaffiliated ones, and this pattern is observed for each time period. On the NASDAQ side, IVIS, LHSP, and PIXT have wider spreads generated by affiliated than by unaffiliated dealers at the market open, although this pattern persists across all time periods only for PIXT.
For LSE-NASDAQ stocks, only SIG has significantly lower affiliated than unaffiliated dealers’ spreads on the LSE (with the exception of the opening period). On the NASDAQ side, however, the affiliated group produces lower spreads than the unaffiliated group in all stocks except MRNT. Although these four stocks include the two (SIGYY and SKYE) in which, according to Table 3, the dominant market is in Europe, supporting the idea that having an affiliate in a better-informed market can enable a dealer to quote lower spreads, this is undermined by the inclusion of DANKY, where EASDAQ contributes little to price formation. In all, it is difficult to regard Table 4 as offering strong support for the notion of intra-dealer integration in our dual-listed stocks.
Another way to detect if affiliation enhances dealers’ ability to compete for order flow would be to examine whether affiliated dealers spend more time being alone on the inside of the posted spread (either on the bid or on the ask side) than do unaffiliated dealers. This issue is addressed in Table 5, whose structure mirrors that of Table 4.
On the EASDAQ side, only LHSP has affiliated dealers quote significantly more aggressively than unaffiliated ones, and even that is only during the period before the US market opens. On the NASDAQ side (for the EASDAQ-NASDAQ sample), IVIS’s affiliated dealers spend more time alone on the inside than its unaffiliated dealer (as long as EASDAQ is also open), as do LHSP’s affiliated dealers (but only at the time of NASDAQ’s closing).
Quotations on the LSE likewise fail to link having an affiliated dealer with quoting more aggressively. Of the five stocks, only SIG has affiliated dealers spending longer on the inside (except at the time of the market’s opening). On the NASDAQ side, though, four of the five stocks have affiliated dealers being more aggressive than unaffiliated ones. However, the one exception is SIGYY (the NASDAQ counterpart of SIG), the stock where all of the price discovery during the trading overlap takes place in Europe – i.e. precisely the stock where one would expect that having a European affiliate would be especially helpful to a US-based dealer. This pattern undermines the argument that affiliation drives quoting aggressiveness.
Even if affiliated dealers do not quote more aggressively (or indeed, as is the case for a number of stocks, less aggressively) than unaffiliated ones, it may nonetheless be the case that ‘affiliated’ quotes are more informative. This issue is particularly interesting to examine at a crucial time in price formation, i.e. at the market open. We measure the informativeness of ‘affiliated’ (respectively, ‘unaffiliated’) opening quotes as the distance from the average of the best bid and ask constructed for the set of affiliated (respectively, control) dealers in the given stock to the first transaction price of the day. This distance is expressed as proportion of the midquote, and reported in Table 6. That table also reports quote informativeness for the market as a whole, and the difference between affiliated and control dealers’ informativeness.
Although it is conceivable that having a US affiliate can help a Europe-based dealer set better quotes at the time the European market opens, any effect is likely to be small, as the US market will have been closed for many hours. Indeed, across our ten dual-listed stocks, affiliated dealers appear to set more informative quotes than unaffiliated ones
Ta b le 5 Dealer af filiation and propor tion of time spent alone on the inside This tab le examines the ef fect of dealer af filiation on quoting agg ressi v eness. F or each stock, w e track w hen a member of the set of af filiated dealers , and of the control set of non-af filiated dealers, w as alone in quoting either the best bid or the best ask price. W e subdi vide the trading da y into the follo wing ti me periods: the mark et’ s opening (OPEN), the period from the opening until the star t of the ov erlap (BEFR), the ov erlap (O VRL), the period from the end of the ov erla p until the mark et’ s closing (AFTR) and the mark et’ s closing (CLOS). F or each of these time periods, the tab le sho ws four numbers: the propor tion of time w hen an af filiated dealer w as alone on the inside, the propor tion of time w hen a control dealer w as alone on the inside, the dif ference betw een these tw o propor tions, and the p-v alue for the dif ference based on the sign rank test with a trading da y as the unit of obser v ation. Mark et T ick er OPEN BEFR O VRL AFTR CLOS A. Stocks dual-listed on N ASD A Q and EASD A Q Easdaq FFTI 0.12 0.11 0.01 0.67 0.11 0.11 0.00 0.48 0.12 0.11 0.01 0.89 0.12 0.13 − 0.01 0.78 Easdaq IVIS 0.33 0.41 − 0.08 0.07 0.33 0.42 − 0.09 0.01 0.35 0.45 − 0.09 0.02 0.35 0.46 − 0.11 0.02 Easdaq LHSP 0.24 0.25 − 0.01 0.76 0.22 0.17 0.05 0.04 0.28 0.27 0.01 0.97 0.25 0.30 − 0.05 0.26 Easdaq NTLI 0.15 0.62 − 0.46 0.00 0.11 0.61 − 0.50 0.00 0.20 0.66 − 0.45 0.00 0.21 0.71 − 0.50 0.00 Easdaq PIXT 0.12 0.24 − 0.12 0.00 0.06 0.25 − 0.19 0.00 0.09 0.26 − 0.18 0.00 0.08 0.24 − 0.15 0.00 Nasdaq FFTI 0.15 0.17 − 0.03 0.45 0.11 0.14 − 0.03 0.32 0.08 0.10 − 0.03 0.03 0.08 0.11 − 0.04 0.17 Nasdaq IVIS 0.61 0.44 0.16 0.00 0.50 0.38 0.12 0.01 0.42 0.42 0.01 0.83 0.41 0.47 − 0.06 0.22 Nasdaq LHSP 0.37 0.42 − 0.05 0.30 0.25 0.22 0.02 0.11 0.21 0.24 − 0.03 0.11 0.25 0.16 0.09 0.03 Nasdaq N TLI 0.16 0.11 0.05 0.15 0.05 0.04 0.01 0.07 0.04 0.04 0.00 0.12 0.16 0.16 0.00 0.91 Nasdaq PIXT 0.18 0.16 0.02 0.63 0.14 0.16 − 0.01 0.57 0.11 0.16 − 0.05 0.05 0.13 0.14 − 0.01 0.80 B . Stocks dual-listed on N ASD A Q and the LSE LSE DNK 0.61 0.66 − 0.04 0.32 0.43 0.46 − 0.03 0.39 0.37 0.44 − 0.07 0.07 0.37 0.42 − 0.05 0.34 LSE MRN 0.10 0.18 − 0.08 0.01 0.06 0.14 − 0.08 0.00 0.09 0.08 0.00 0.84 0.07 0.09 − 0.02 0.32 LSE SHP 0.35 0.56 − 0.20 0.00 0.33 0.42 − 0.09 0.02 0.32 0.40 − 0.07 0.08 0.33 0.44 − 0.11 0.03 LSE SIG 0.56 0.58 − 0.02 0.69 0.51 0.36 0.15 0.00 0.52 0.31 0.21 0.00 0.54 0.29 0.24 0.00 LSE SKP 0.36 0.64 − 0.28 0.00 0.29 0.52 − 0.23 0.00 0.27 0.51 − 0.25 0.00 0.24 0.49 − 0.24 0.00 Nasdaq D ANKY 0.32 0.13 0.18 0.00 0.13 0.05 0.08 0.00 0.12 0.05 0.07 0.00 0.17 0.05 0.12 0.00 Nasdaq MRNT 0.49 0.53 − 0.04 0.43 0.43 0.31 0.12 0.00 0.37 0.23 0.14 0.00 0.34 0.23 0.11 0.01 Nasdaq SHPGY 0.60 0.32 0.28 0.00 0.46 0.19 0.27 0.00 0.40 0.19 0.22 0.00 0.39 0.23 0.16 0.00 Nasdaq SIGYY 0.14 0.19 − 0.05 0.16 0.17 0.21 − 0.04 0.20 0.14 0.20 − 0.06 0.09 0.13 0.23 − 0.10 0.01 Nasdaq SKYE 0.75 0.41 0.33 0.00 0.64 0.36 0.28 0.00 0.58 0.31 0.27 0.00 0.54 0.37 0.17 0.00
Table 6
Informativeness of opening quotes.
This table compares the informativeness of opening quotes for groups of dealers. For each stock, we construct the best bid and ask separately for affiliated dealers and for the control set of non-affiliated dealers (these are listed in the appendix). We then calculate the midquote (average of best bid and ask) for affiliated (“Global”) and unaffiliated control dealers (“Control”), as well as for the market as a whole (“All”). Informativeness for each group of dealers is then measured as the absolute value of the difference between the midquote for that group and the first transaction price of the day, scaled by the midquote. The column labeled “# days” reports the number of days used in the estimation. The next three columns show the average informativeness for Global, Control, and All groups. The last two columns show the difference between average informativeness of Global and Control groups, followed by the p-value for the difference based on the paired t-test.
Market Ticker # days Global Control All Global-Control A. Stocks dual-listed on NASDAQ and EASDAQ
Easdaq FFTI 68 0.016 0.012 0.012 0.004 0.00 Easdaq IVIS 142 0.017 0.015 0.015 0.001 0.12 Easdaq LHSP 210 0.007 0.007 0.006 0.001 0.10 Easdaq NTLI 63 0.017 0.019 0.018 −0.002 0.00 Easdaq PIXT 114 0.028 0.027 0.023 0.001 0.77 Nasdaq FFTI 237 0.017 0.013 0.006 0.004 0.00 Nasdaq IVIS 229 0.010 0.011 0.008 −0.001 0.19 Nasdaq LHSP 237 0.005 0.006 0.002 −0.001 0.03 Nasdaq NTLI 236 0.012 0.006 0.002 0.006 0.00 Nasdaq PLXT 219 0.041 0.059 0.014 −0.018 0.00
B. Stocks dual-listed on NASDAQ and the LSE
LSE DNK 241 0.017 0.017 0.017 0.000 0.44 LSE MRN 227 0.012 0.012 0.010 0.000 0.67 LSE SHP 241 0.005 0.005 0.005 0.000 0.67 LSE SIG 241 0.009 0.009 0.009 0.001 0.02 LSE SKP 241 0.017 0.014 0.014 0.003 0.00 Nasdaq DANKY 241 0.010 0.012 0.007 −0.001 0.04 Nasdaq MRNT 241 0.010 0.007 0.007 0.003 0.00 Nasdaq SHPGY 241 0.007 0.008 0.006 −0.001 0.11 Nasdaq SIGYY 205 0.011 0.012 0.011 0.000 0.46 Nasdaq SKYE 238 0.015 0.018 0.013 −0.003 0.01
only in the case of NTLI. On NASDAQ, however, one would expect the effect of having a European affiliate to be stronger, since at the time of the US open, European dealers are active and have been processing information about dual-listed stocks for almost the entire trading day. The table shows that there are four dual-listed stocks where having a European affiliate allows the US dealer to set more informative opening quotes: LHSP and PIXT (which are also quoted on EASDAQ) as well as DANKY and SKYE (which are also quoted on the LSE). However, this result is rendered more ambiguous by the fact that these four stocks are not all characterised by high contributions of the European market to price discovery, whether during the trading overlap (Table 3), or, as is more pertinent here, as measured by the weighted price contribution of the time period when
only the US market is open (Table 2). Once again, therefore, it is difficult to regard the evidence on opening quote efficiency as providing support for affiliated dealers being integrated.
Our final attempt to tease out evidence of intra-dealer integration from affiliated dealers’ price quotes investigates whether, during the trading overlap, a dealer is more likely to quote prices in concert with its overseas affiliate than with an unrelated control dealer. This investigation is conducted at the level of a dealer-stock combination and proceeds as follows. First, we estimate a time series of efficient prices for each stock
during the trading day overlap.13 These prices can be interpreted as the best estimate
of underlying value of the stock at each point in time, given the price quote dynamics on the two markets. The more two affiliated dealers are integrated, i.e. the more they act as single dealer, the more frequently we would expect their quotes to be on the same side of the efficient price. This is because they would have access to the same information about the stock, and because they would care about their common (netted out) inventory position. Thus, for each pair of affiliated market makers in each stock, we calculate the proportion of time that the affiliated market makers were on the same side of the efficient price, and compare it with the proportion of time than a control pair
of unaffiliated market makers is on the same side of the efficient price.14
The results are presented in Table 7. Remarkably, of the 19 instances of global market making firms being active in the same stock on NASDAQ and in Europe, there is not a single case where affiliated dealers behave significantly more alike than does a control pair of unaffiliated dealers (indeed, there are two cases where the reverse is true). In the case of LSE-NASDAQ stocks, one could attribute this in part to the fact that on NASDAQ they are traded in the form of ADRs, which may complicate the netting out of inventory positions. In the case of EASDAQ-NASDAQ stocks, however, there is perfect fungibility across the two markets. Failure to detect coordination among affiliated dealers adds to
our accumulated evidence against intra-dealer integration in cross-listed stocks.15
Presumably, belonging to separate profit centres, as European and US desks of
securities firms do, is enough to discourage within-firm transatlantic cooperation.16In
13
Efficient prices for each stock and at each point in time are estimated (as in Hasbrouck (2007, p. 98)) as the forecast price to which quote midpoints on both markets converge (given that these prices are cointegrated). We use 30-second intervals and 10 lags in the estimation of the Vector Error Correction Models (VECM). We also estimated efficient prices with alternative VECM specifications using the Akaike Information Criterion to determine the optimal number of lags, which leads to similar results.
14We only use cases where both members of a pair of affiliated dealers in a given stock have
matching unaffiliated dealers averaging at least one quote per trading day overlap.
15As a further test for intra-dealer integration, we study the patterns of affiliated dealers’
contributions to price discovery, following the Huang (2002) application of the WPC measure. Specifically, we check whether quote revisions due to affiliated dealers contribute more to price discovery than those due to matching unaffiliated dealers (results available upon request). In particular, we would expect a global dealer to be a particularly strong driver of price discovery if its affiliate is located in the dominant market for the stock in question. However, we do not find strong evidence in support of this hypothesis. On the ask side, global dealers with affiliates in the dominant market have higher WPCs than unaffiliated matching dealers for six of the ten sample stocks. On the bid side, this is the case for only five stocks.
16Conversations with several of the European dealers included in this study suggested that
Ta b le 7 Dealer af filiation and quote positioning. This tab le sho ws the propor tion of time that af filiated mark et mak ers w ere on the same side of the ef ficient price during the trading da y ov erlap, and co mpares it with the propor tion of time than a control pair of unaf filiated mark et mak ers w as on the same side of the ef ficient price. W e onl y use cases w here both mem bers of a pair of af filiated dealers in a gi v en stock ha v e matching unaf filiated dealers av eraging at least one quote per trading da y ov erlap. P-v alues for the dif ference betw een the propor tions of time are based on the sign test conducted with the trading da y as the obser v ation. % time quotes are on the same side of the ef ficient price Nasdaq Af filiated Control-Af filiated Control # da ys Mark ets tick er dealer Europe US dealer dealers Dif ference used p-v alue A. Stocks dual-listed on N ASD A Q and EASD A Q Easdaq/Nasdaq FFTI Herzo g Puilatco F irst Alban y 0.699 0.763 − 0.064 111 0.137 Easdaq/Nasdaq IVIS Herzo g Puilatco Knight 0.668 0.748 − 0.080 163 0.595 Easdaq/Nasdaq IVIS SG Quar tz Spear , L eeds & K ello gg 0.715 0.809 − 0.094 132 0.201 Easdaq/Nasdaq L HSP Herzo g Banque P opulaire Ma y er & Schw eitzer 0.799 0.790 0.009 221 0.654 Easdaq/Nasdaq L HSP SG Quar tz Sherw ood 0.718 0.793 − 0.075 221 0.013 Easdaq/Nasdaq L HSP Rober tson Ste v ens P aribas J.W . Genes is 0.652 0.707 − 0.056 107 1.000 Easdaq/Nasdaq NTLI Herzo g Quar tz Salomon 0.759 0.828 − 0.070 197 0.000 Easdaq/Nasdaq PIXT Herzo g BNP Spear , Leeds & K ello gg 0.687 0.700 − 0.013 160 0.777 B . Stocks dual-listed on N ASD A Q and the LSE LSE/Nasdaq D ANKY Herzo g HSBC Sherw ood 0.679 0.653 0.026 155 0.929 LSE/Nasdaq D ANKY Mer rill SG Allen C. E wing 0.688 0.675 0.014 237 0.261 LSE/Nasdaq D ANKY Salomon W interflood Rober t W . Baird 0.657 0.696 − 0.039 240 0.249 LSE/Nasdaq M RNT Herzo g W interflood Salomon 0.711 0.647 0.064 240 0.192 LSE/Nasdaq SHPGY Goldman W arb ur g Hambrecht & Quist 0.619 0.707 − 0.088 41 0.377 LSE/Nasdaq SHPGY Herzo g W interflood Spear , Leeds & K ello gg 0.684 0.732 − 0.048 98 0.057 LSE/Nasdaq SHPGY Mer rill DKB Knight 0.737 0.678 0.059 110 0.278 LSE/Nasdaq S I GYY Goldman W interflood Knight 0.603 0.574 0.029 239 0.363 LSE/Nasdaq SKYE Herzo g Mer rill Spear , Leeds & K ello gg 0.627 0.648 − 0.021 220 0.721 LSE/Nasdaq SKYE Salomon W arbur g V olpe Bro wn Whelan 0.530 0.512 0.018 229 0.691 LSE/Nasdaq SKYE SG W interflood Knight 0.532 0.570 − 0.038 185 0.677
light of previous evidence that informed traders split orders across markets (Menkveld, 2008) and earn increased profits (Chowdhry and Nanda, 1991) our findings suggest that affiliated dealers may increase profits by creating global rather than regional profit centres.
6. Conclusions
Our paper is the first to study what happens when the same firms are involved in market making for the same securities on trading venues in different countries. One part of our sample consists of stocks dually listed on NASDAQ and EASDAQ markets. These stocks have the advantage of near-perfect fungibility (as trades and quotes are in US dollars on both venues, and the same assets – rather than ADRs – are involved on both venues), but they have little liquidity and only several dual-market participants. Therefore, as the second part of our sample we use stocks dually listed on NASDAQ and the LSE, where liquidity is higher and dual-market dealers more numerous. The three markets are similar, during the period of our study, in that they are essentially quote-driven, enabling us to study quotation behaviour in the remarkable setting where securities, trading hours, and market makers overlap between two market centres.
Our main finding is easy to summarise. While the markets themselves are integrated, firm-level integration is hard to detect. There is little evidence that a dealer pays particular attention to the quotes of her colleagues in the same firm but on the other side of the Atlantic. This result offers a new insight into international stock market linkages at a time when a cross-Atlantic exchange appears inevitable. More broadly, it is also indicative of the difficulties multinationals face in truly integrating their activities. Global markets have become more integrated over time. It will be interesting to see whether global securities firms will become more integrated as well. This question is left to future research.
Appendix: Matching unaffiliated dealers
This Appendix shows how a set of matching unaffiliated dealers was selected for dealers who have affiliates across the Atlantic. To obtain a workable control sample of unaffiliated dealers, we proceeded as follows. If most dealers were unaffiliated (resp. affiliated), we matched affiliated (resp. unaffiliated) dealers to unaffiliated (resp. affiliated) ones based on the total number of valid quotes, starting from the affiliated (resp. unaffiliated) dealer with the smallest number of quotes, and without replacement. Then, in instances where the set of matching dealers for a given stock could be reordered so as to minimize the sum of differences between the quotation frequencies of affiliated dealers and their matches, we did so. Lastly, we dropped cases where the affiliated dealer has more than double the number of quotes of the matching dealer, or vice versa. The resulting set of matching dealers is listed below. When the control dealer is shown in parentheses, this indicates that the corresponding affiliated dealer averages less than one quote per trading day overlap in the given stock, and is therefore only used for those analyses where affiliated dealers are pooled rather than studied individually.
no formal system of sharing stock-specific information, possibly due to the fact that dual listings represented a relatively small proportion of a dealer’s stocks.
A. Stocks dual-listed on NASDAQ and EASDAQ
Stock Affiliated dealer Control – Europe Control – US 4Front Technologies Herzog Puilatco First Albany
ICOS Vision Systems Herzog Puilatco Knight
SG Quartz Spear, Leeds & Kellogg Robertson Stevens (KBC) Fleet
Lernout & Hauspie Speech Herzog Banque Populaire Mayer & Schweitzer
Products SG Quartz Sherwood
Robertson Stevens Paribas J.W. Genesis
NTL Herzog Quartz Salomon
PixTech Herzog BNP Spear, Leeds & Kellogg
B. Stocks dual-listed on NASDAQ and the LSE
Stock Affiliated dealer Control – Europe Control – US
Danka Business Systems Herzog HSBC Sherwood
Merrill SG Allen C. Ewing
Salomon Winterflood Robert W. Baird
Warburg – (Olde)
Merant CSFB – Jefferies
DLJ – National Financial
DKB – Montgomery
Goldman – Sherwood
Herzog Winterflood Salomon
Merrill – Spear, Leeds & Kellogg
Warburg – H.C. Wainwright
Shire Pharmaceuticals Bear Stearns – Natexis
CSFB – Sherwood
DLJ – –
DMG – –
Goldman Warburg Hambrecht & Quist Herzog Winterflood Spear, Leeds & Kellogg
Lehman WestLB –
Merrill DKB Knight
Signet DMG SG (Sherwood)
Goldman Winterflood Knight
Herzog Cazenove (Mayer & Schweitzer) Merrill Aitken Campbell (Spear, Leeds & Kellogg) Warburg ABN AMRO (Natexis Bleichroeder) Skyepharma Herzog Merrill Spear, Leeds & Kellogg
Salomon Warburg Volpe Brown Whelan
SG Winterflood Knight
References
Anand, A., ‘Specialist: the firm or the individual? Empirical evidence from the options markets’,
Baillie, R. T., Booth, G. G., Tse, Y. and Zabotina, T., ‘Price discovery and common factor models’,
Journal of Financial Markets, Vol. 5, 2002, pp. 309–21.
Barclay, M. J., Christie, W. G., Harris, J. H. Kandel, E. and Schultz, P. H., ‘Effects of market reform on the trading costs and depths of NASDAQ stocks’, Journal of Finance, Vol. 54, 1999, pp. 1–34. Barclay, M. and Hendershott, T., ‘Price discovery and trading after hours’, Review of Financial Studies,
Vol. 16, 2003, pp. 1041–73.
Battalio, R., Green, J. and Jennings, R., ‘Do competing specialists and preferencing dealers affect market quality?’ Review of Financial Studies, Vol. 10, 1997, pp. 969–93.
Chowdhry, B. and Nanda, V., ‘Multimarket trading and market liquidity’, Review of Financial Studies, Vol. 4, 1991, pp. 485–511.
Coughenour, J. F. and Saad, M. M., ‘Common market makers and commonality in liquidity’, Journal
of Financial Economics, Vol. 73, 2004, pp. 37–69.
Domowitz, I., Glen, J. D. and Madhavan, A., ‘International cross-listing and order flow migration: evidence from an emerging market’, Journal of Finance, Vol. 53, 1998, pp. 2001–27.
Eun, C. S. and Sabherwal, S., ‘Cross-border listings and price discovery: evidence from US-listed Canadian stocks’, Journal of Finance, Vol. 58, 2003, pp. 549–75.
Fong, K., Madhavan, A. and Swan, P. L., ‘Why do markets fragment? A panel-data analysis of off-exchange trading’, Working Paper (University of New South Wales, 2003).
Gagnon, L. and Karolyi, G.A., ‘Price and volatility transmission across borders’, Financial Markets,
Institutions and Instruments, Vol. 15, 2006, pp. 107–58.
Grammig, J., Melvin, M. and Schlag, C., ‘Internationally cross-listed stock prices during overlapping trading hours: price discovery and exchange rate effects’, Journal of Empirical Finance, Vol. 12, 2005, pp. 139–64.
Hasbrouck, J., ‘One security, many markets: determining the contributions to price discovery’, Journal
of Finance, Vol. 50, 1995, pp. 1175–99.
Hasbrouck, J., Empirical Market Microstructure (Oxford University Press, 2007).
Hau, H., ‘Location matters: an examination of trading profits’, Journal of Finance, Vol. 51, 2001, pp. 1959–83.
Huang, R. D., ‘The quality of ECN and Nasdaq market maker quotes’, Journal of Finance, Vol. 57, 2002, pp. 1285–1319.
Hupperets, E. and Menkveld, A., ‘Intra-day analysis of market integration: Dutch blue chips traded in Amsterdam and New York’, Journal of Financial Markets, Vol. 5, 2002, pp. 205–58.
Karolyi, G. A., ‘The world of cross-listings and cross-listings of the world: challenging conventional wisdom’, Review of Financial Studies, Vol. 10, 2006, pp. 99–152.
Kato, K., Linn, S. and Schallheim, J., ‘Are there arbitrage opportunities in the market for American Depository Receipts?’ Journal of International Financial Markets, Institutions, and Money, Vol. 1, 1990, pp. 73–89.
Kim M., Szakmary, A. C. and Mathur, I., ‘Price transmission dynamics between ADRs and their underlying foreign securities’, Journal of Banking and Finance, Vol. 24, 2000 pp. 1359–82. Lau, S. T. and Diltz, J. D., ‘Stock returns and the transfer of information between the New York and
Tokyo stock exchanges’, Journal of International Money and Finance, Vol. 13, 1994, pp. 211– 22.
Lieberman, O., Ben-Zion, U. and Hauser, S., ‘A characterization of the price behavior of international dual stocks: an error correction approach’, Journal of International Money and Finance, Vol. 18, 1999, pp. 289–304.
Menkveld, A. J., ‘Splitting orders in overlapping markets: a study of cross-listed stocks’, Journal of
Financial Intermediation, Vol. 17, 2008, pp. 145–74.
Mouton, P. and Wei, L., ‘A tale of two time zones: Cross-listed stock liquidity and the availability of substitutes’, Journal of Financial Markets, Vol. 12, 2009, pp. 570–91.
Naik, N. Y. and Yadav, P. K., ‘Do dealer firms manage inventory on a stock-by-stock or a portfolio basis?’ Journal of Financial Economics, Vol. 69, 2003, pp. 325–53.
Phylaktis, K. and Korczak, P., ‘Price discovery process in international cross-listings: evidence from US-listed British and French companies’, Working Paper (Cass Business School, 1994).
Porter, D. C., Tanggaard, C., Weaver, D. G., and Yu, W., ‘Dispersed trading and the prevention of market failure: the case of the Copenhagen Stock Exchange’, European Financial Management, Vol. 14(2), 2008, pp. 243–67.
Pulatkonak, M. and Sofianos, G., ‘The distribution of global trading in NYSE-listed non-US stocks’,
NYSE Working Paper 99-03 (1999).
Smith, J. W., Selway, J. P. and McCormick, D. T., ‘The NASDAQ stock market: historical background and current operation’, NASD Working Paper 98–01 (1998).
Tirez, D., ed., A Practitioner’s Guide to EASDAQ Rules, City and Financial Publishing, London, (1999). Werner, I. M. and Kleidon, A. W., ‘UK and US trading of British cross-listed stocks: an intra-day