French and U.S. Trading of Cross-Listed Stocks around the Period of U.S. Decimalization: Volume, Spreads, and Depth Effects

Transcription

1 French and U.S. Trading of Cross-Listed Stocks around the Period of U.S. Decimalization: Volume, Spreads, and Depth Effects Bing-Xuan Lin Assistant Professor of Finance College of Business Administration University of Rhode Island 7 Lippitt Road Kingston, RI blin@uri.edu Tel: (401) David Michayluk School of Finance and Economics University of Technology at Sydney PO Box 123 Broadway, NSW 2007 Australia david.michayluk@uts.edu.au Tel: Henry R. Oppenheimer Associate Professor of Finance College of Business Administration University of Rhode Island 7 Lippitt Road Kingston, RI HenryO@uriacc.uri.edu Tel: (401) and Sanjiv Sabherwal Assistant Professor of Finance College of Business Administration University of Texas at Arlington Arlington, TX sabherwal@uta.edu Tel: (817) This version: September 1,

2 French and U.S. Trading of Cross-Listed Stocks around the Period of NYSE Decimalization: Volume, Spreads, and Depth Effects Abstract This paper analyzes the impact of U.S. decimalization on the trading of both French and American Securities traded on both Euronext and the NYSE. The sample of 31 firms represents some of the largest firms in each country. We find that trading of U.S. firms does not increase in the U.S. after decimalization and the size of these trades, on average, declines. Further, the number of trades of these stocks decreases on Euronext, but the average trade size increases modestly. After NYSE decimalization French securities trade in the U.S. about as often as before decimalization, but the average trade size declines markedly. Conversely, the frequency of trading of French securities on Euronext increases and the average size of these trades is unchanged after U.S. decimalization. Both spreads and depths on the NYSE decline; spreads for both French and American firms decline modestly on Euronext, but measured depths of these firms on Euronext increase somewhat. Furthermore, we find that trading is far more frequent and in bigger transactions sizes when both markets are open. The frequency of market-maker quotes is also much greater during periods when both markets are open. Our study provides more insights regarding the impact of decimalization and the transaction costs for stocks listed on multiple exchanges. Given the pending proposal merger of the NYSE and Euronext, stakeholders in both venues might want to contemplate these results as they go forward. 2

3 French and U.S. Trading of Cross-Listed Stocks around the Period of NYSE Decimalization: Volume, Spreads, and Depth Effects 1. Introduction Many firms securities trade on multiple world markets. Management chooses to have their securities listed on markets other than their natural home market for many reasons, including, but not limited to: the firm s products, sales outlets, and production facilities (as well as employees) may be in many countries, the desire to obtain a broader, more geographically diversified investor base, facilitation of raising capital external to the home market, the belief that cross-listing leads to a lower cost of capital, the goal of facilitating ease of trading for investors in terms of their time zones, and the perceived prestige of being listed on many (or all) of the world s major markets. 1 Irrespective of the reasons for management s decision to have its firm s securities listed in multiple markets, simultaneous listing provides researchers an opportunity to examine many important financial issues relating to trader and market maker behavior. These issues include arbitrage opportunities in fully or not fully integrated markets, transaction size and frequency issues, market maker s spreads and depths and frequency of quotes for the same securities in the different markets and differences in these characteristics during periods when one versus two or more (trading) markets are open. In this paper we examine the impact of New York Stock Exchange (NYSE) decimalization on the trading of U.S. firms listed on both the NYSE and on Euronext- 1 However, management needs to balance these positive arguments supporting cross-listing against reasons for not cross-listing, including incremental direct listing costs on other exchanges, incremental regulatory and financial disclosure costs, costs associated with restatement of financial statements according to alternative accounting conventions, and trading frictions. 3

4 Paris and the trading of French firms traded on both Euronext-Paris and (as ADRs) on the NYSE. Irrespective of the fact that the NYSE and NASDAQ have larger trading value, annually, than all other world exchanges, they were the last two major exchanges to undergo decimalization (with the NYSE implementing full decimalization on January 29, 2001 and NASDAQ implementing on April 9, 2001); consequently examination of the trading of securities listed on them and cross-listed elsewhere provides a good opportunity to analyze the various microstructure and decimalization issues listed above as well as other issues. 2 The two samples we analyze, while small (16 American and 15 French firms), represent material portions of the two home markets (as of December 31, 2000 the American firms accounted for over 15 percent of the capitalization of the NYSE and the French firms accounted for over 23 percent of the Euronext market). 3 The shares of these companies trade sufficiently often in their non-home markets to provide valuable insights as to the impact of decimalization in terms of investor and market maker behavior. In this paper we consider the following issues. First we focus on the changes in the daily number of and size of trades in the U.S. and in France during the six weeks after NYSE decimalization as compared to the six-week period prior to decimalization. Next we compare the changes in market makers quoted and Euronext s bid-ask spreads and depths between the six weeks prior to and the six weeks subsequent to NYSE decimalization. Finally we focus on these issues number of and size of transactions and 2 For the background to U.S. decimalization, see U.S. Securities and Exchange Commission (2000a, 2000b). Early results of U.S. decimalization are presented in NYSE Research (2001a, 2001b) and Bacidore et al. (2001). 3 In terms of total trading volume, the NYSE had the second greatest trading volume in the world in 2001 (just slightly below that of NASDAQ) and Euronext was fourth (after the London Stock Exchange). As will be noted shortly, these securities also trade actively over-the counter. (see Bodie, Kane, and Marcus, 2005, p. 84). 4

5 bid-ask spreads and depths during three specific daily periods: the European morning (when the Euronext market is open and the NYSE is not), the European afternoon (when both Euronext and the NYSE are open) and the European evening (when only New York is open). In this analysis we consider how competition for order flow when both markets are open influences trading frequency and size, as well as market maker behavior and how NYSE decimalization influenced (changed) these important parameters. This paper represents an addition to the market microstructure and decimalization literature. In one sense this paper is a logical extension of the Werner and Kleidon (1996) research on the integration of the London Stock Exchange (LSE) and the major U.S. exchanges. This paper also presents an extension of both the Ahn et al. (1998) and Oppenheimer and Sabhewal (2003) investigations on the impact of decimalization on the trading of securities cross-listed in a market that is undergoing decimalization and one that is not. It is distinguished from those later two papers in that we look at US decimalization of very large (institutionally dominated) American and French securities on markets that have a two-hour overlap of trading and long periods of no overlap. We find, contrary to expectations, that trading of these cross-listed U.S. firms does not increase in the U.S. after decimalization. As expected, we find that the size of these trades, on average, declines. Further, the number of trades of these stocks does decrease on Euronext, but average trade size increases modestly. After decimalization French securities trade in the U.S. about as often as before decimalization, but average trade size declines markedly. Conversely, the frequency of the trading of French securities on Euronext increases and the average size of these trades is unchanged after NYSE decimalization. Both spreads and depths on the NYSE decline (more modestly for 5

6 French as opposed to American firms); spreads for both French and American firms decline modestly on Euronext, but measured depths of these firms on Euronext increase somewhat. We find that trading is far more frequent and in bigger transactions sizes when both markets are open. The frequency of market-maker quotes is also much greater during periods when both markets are open. The remainder of this paper is organized as follows: In the next section we provide the necessary background and develop the major issues of this paper. Section 3 describes the data and methodology. The empirical results are presented in Section 4 and Section 5 concludes. 2. Background and Issues 2.1 Background Werner and Kleidon (1996) provide an early look at the trading of British securities that are listed in both London and in New York as ADRs. They find that the New York intraday patterns for these securities are relatively unaffected by the fact that they have traded in London for six hours prior to the New York open. They also find that there is heavy trading in both markets during the two-hour simultaneous trading period (14:30 to 16:30 in London and 9:30 to 11:30 in New York). They find that spreads in London for these securities are somewhat lower than for comparable controls (securities listed only in London) during the simultaneous trading period. Werner and Kleidon (1996) conclude that there is (new) private information reflected in U.S. trading of the ADRs and that arbitrage activities in both markets occur during the simultaneous trading period. They also conclude that the dealers in each market react differently to the informed trading issue because the costs for U.S. based traders transacting in London are 6

7 sufficiently high to provide a barrier for U.S. traders placing a major part of their orders in London. Harris (1994) provides the first consideration of the issue of decimalization. He argues that small traders benefit from tick size reductions because spreads decline. Ahn et al. (1996) look at the impact of the reduction of tick size from eighths to sixteenths for low priced ($1.00 to $5.00) American Stock Exchange (AMEX) securities. They find that both quoted and effective spreads of these stocks decrease, that the decrease is larger for more frequently traded than for less frequently traded securities and that market maker bid ask depths do not change. They conclude that investors benefit from and market makers are hurt by the reduced spreads because there is not a compensatory increase in trading volume. In a subsequent paper, Ahn et al. (1998) look at Canadian and U.S. stocks crosslisted in both countries in relationship to the Toronto Stock Exchange (TSE) reduction of tick size from C$0.125 to C$0.05. They find that for the Canadian stocks cross-listed in the U.S. there is not a decline in spreads in the U.S. for NYSE/AMEX issues, but there is a reduction in spread on NASDAQ. They do find there are declines in spread on the TSE, but there is no evidence of order flow to the TSE from the U.S. They also find that TSE trading volume of NYSE/AMEX issues declines after the TSE reduction in tick size. Bacidore et al. (2001) present early results of the impact of NYSE decimalization on a cross-section of domestic securities, including the fifty most actively traded and random samples with sampling from permutations of high or low volume and price. They do not address changes in trading volume, but for their samples they do find that the average bid-ask spread declines by about 30 percent and depth of quote, on average, 7

8 declines by about 70 percent. They find that their results are not transitory; they persist for at least 13 weeks. More recently Oppenheimer and Sabherwal (2003) examine the impact of NYSE and NASDAQ decimalization (two temporally separate events) on the trading in the U.S. and on the TSE of Canadian firms trading in both countries. 4 They find that the U.S. trading of the Canadian stocks listed on the TSE and either the NYSE or NASDAQ increases materially after decimalization. Further, they find that the TSE volume increases substantially for those securities that are traded on NASDAQ and increases marginally for those securities that are traded on the NYSE. The bid-ask spreads decline on all exchanges, but by more in the U.S. than in Canada. The quote depths on the NYSE decline substantially, from being above the TSE depths to well below the TSE depths. Their regression analyses indicate that the decline in the TSE spread is directly related to the size of the firm and to the decline in the U.S. spread, and is inversely related to the pre-decimalization ratio of spreads on the U.S. exchange and the TSE. 2.2 Issues Both theory and prior empirical research suggest that decimalization should lead to smaller spreads in the (newly decimalized) markets. This should lower transactions costs for traders, but cut profit per share traded for market makers. Unless trading volume increases sufficiently to compensate market makers for the lower spread, their profits from trading will decline. One might expect an increase in trading activity in affected markets because the decrease in tick size ought to improve liquidity as a result of the decline in total transactions costs. In addition, traders will face a smaller barrier to trading 8

9 on new information reaching the market. The consequent hypotheses, consistent with prior work, are that the number of trades and the total dollar value of the trades should increase subsequent to decimalization. Similarly, it is expected that the quoted spreads will decline. Such results should hold for both American and French securities trading on the NYSE. Whether or not market maker profit at least remains constant is an important empirical issue that we can only indirectly address here. A second impact should be smaller spreads in France, especially for American firms. Market makers (and on Euronext the computerized limit order book) serve to provide liquidity to those interested in transacting. To the extent that these participants are indifferent to the choice of marketplace for their transactions, they should utilize that marketplace with the lowest total transactions costs. Consequently, for competitive reasons, Euronext participants should lower their spreads in response to U.S. decimalization to avoid losing market orders. 5 The activity of informed (as opposed to liquidity or uninformed) traders is important. As the spread narrows, market makers and those entering limit orders face increased risk of losing money on trades with informed traders. Consequently, market makers and those entering limit orders will attempt to protect themselves against these losses by lowering the depth of their quotes (or limit orders). We expect to observe such a lowering in the U.S. after decimalization. Similarly, as we expect spreads to narrow on the Euronext, we expect a lowering of depths on the Euronext for the same reasons. 4 Ahn et al. look at both U.S. and Canadian stocks. Sabherwal and Oppenheimer (2003) only look at Canadian stocks; they find that there are not enough U.S. stocks that trade frequently in Canada to perform their analyses. 5 Several studies, such as Eun and Sabherwal (2002), Foerster and Karolyi (1998), and Werner and Kleidon (1996), discuss competition among market makers for internationally cross-listed securities. 9

10 The impact of limit orders on spreads and depths must be carefully considered. Harris (1997) and Bacidore et al. (2001) provide a background of the relationship between decimalization and limit orders. All things being equal, decimalization will make limit order traders more cautious because traders purchasing (selling) in front of a bid (ask) only have to beat the limit order by a penny, versus 6.25 cents prior to decimalization. Similarly, unwinding such positions into a limit order would only involve sacrificing (losing) a penny rather than the prior to decimalization 6.25 cents. Identical reasoning in Angel (1997) leads to the hypothesis that liquidity will decline because limit order traders and market makers will become more cautious, with market makers bid-ask quotes reflecting less depth. Goldstein and Kavajecz (2000) provide an analysis of the impact of the change in tick size from eighths to sixteenths on limit orders, the limit order book, and NYSE transactions. They argue that a decrease in tick size benefits liquidity demanders but damages liquidity providers by increasing their costs and decreasing their willingness to provide liquidity. Consistent with their argument, they find an increase in spread between highest buy and lowest sell of $0.03 (9.1 percent) after the reduction in tick size to sixteenths and they also find that the depth at best prices in the limit order book declines by 48 percent and the cumulative depth in the limit order book declines as well. They find that the cost is reduced for small market orders but that large market orders do not benefit from tick size reduction. 6 Finally, they conclude that that their results are sensitive to trade size, trading frequency, and price level. The benefit of tick size reduction to small orders is reduced for infrequently traded and low priced stocks, especially if liquidity is derived from the limit order book. 10

11 Bacidore et al. (2001) had access to data relating to limit order placement. 7 They find that limit order traders decrease the size of their limit orders by over 33 percent, that limit orders are priced closer to bid-ask midpoint after decimalization than before (though this is not found for low volume stocks), and that limit order cancellation rates increase after decimalization. These findings are not inconsistent with consequent market maker actions such as widening spreads and reducing depths, actions inconsistent with the intent of decimalization. These results could occur on the NYSE. Both the Ahn et al. (1998) and Oppenheimer and Sabherwal (2003) analyses of decimalization consider two markets which are always simultaneously open or closed. That is not the case in the current study. Here we have the European market (home for the French firms) open for 6 ½ hours, both markets open for 2 hours and then the NYSE (home for American firms) open for 4 ½ hours. Assuming institutional investors are, to a certain extent, indifferent to trading venue, 8 one could assert a competitive argument that during the period when both markets are open there is more competition, leading to more quotes both in the home and non-domestic market and, consequently, the spread in the non-domestic market should be lower than it is during the remainder of its trading day and depths should be higher. Therefore, one would expect more trading of securities during the period that both markets are open than during the other periods when only one is open. 6 The securities analyzed here are securities quite likely to be traded in large blocks. 7 We do not have access to such data for the NYSE, NASDAQ or Euronext. 8 But Werner and Kleidon (1996) and others do qualify that assumption. 11

12 3. Data and Methodology 3.1 Sample Selection Our sample was created as follows. As a first step in sample selection, we obtained a list of all French securities listed on the NYSE web site and we also obtained a list of all NYSE firms that traded on Euronext-Paris. 9 We then determined whether the U.S. decimalization of trading of the firm s securities had occurred prior to the NYSE final decimalization date, 10 and whether the firm was listed on both Euronext-Paris and the NYSE at least 30 trading days prior to decimalization and continued to be listed at least 30 trading days subsequent to decimalization. We also excluded stocks that split during the study period, preferred stocks, and suspended stocks. As of December 31, 2000, there were 15 French firms listed on the NYSE (all traded as ADRs) and there were 16 NYSE firms traded on Euronext-Paris that met these criteria. The listing of the final sample is provided in Table 1. The American firms in the sample are very large wellknown firms that were (and still are) very actively traded and subject to significant institutional investor interest. These 16 firms amounted to percent of NYSE total capitalization (excluding preferred issues but including closed end funds and ADRs) as of December 31, Similarly, the French firms are among the largest French firms that trade on Euronext and also subject to substantial institutional investor interest. They constituted percent of the total Euronext market as of December 31, Thus 9 Euronext was formed on September 22, 2000 by merger of the Amsterdam, Brussels, and Paris exchanges. The only securities analyzed in the current paper are those that were a part of the Paris exchange and thus became part of Euronext-Paris. 10 The NYSE had a phase-in of decimalization during the year 2000, with implementation of decimalization for a small minority of firms (chosen by the NYSE) prior to January 29, The vast majority of NYSElisted securities decimalization was implemented on January 29, It was not clear what the total capitalization of Euronext-Paris was on December 31, 2000, but obviously these fifteen securities constituted a larger proportion of it than the total Euronext market. 12

13 the 31 firms analyzed represent material portions of both the NYSE and the Euronext total capitalizations and their daily trading activity. 3.2 Analytical Approach For each security in the sample, we obtain all U.S. and Euronext-Paris transactions and market maker posted quotes for the 30 trading days prior to NYSE decimalization through the 30 trading days after decimalization that both the NYSE and Euronext are simultaneously open. 12 We obtain similar data for each security in the sample that is cross-listed on the NYSE and Euronext-Paris. The NYSE transaction data (number of trades, number of shares traded in each trade, and price of each trade) and quote data (time, bid and ask prices, and their depths) are obtained from the Trade and Quote (TAQ) database created by the NYSE. We acquire similar data for European trades and quotes from the Euronext. 13 We use standard procedures to eliminate cancelled trades and errors in the data. To convert the prices in Euros to U.S.$ we use the average daily exchange rates for 12 It should be noted that both the NYSE and the Euronext are, for the most part, open on the same days. However, there are certain American and European holidays that result in one, but not the other being open. The method chosen here assures that all parameters analyzed are for days where both are open. 13 Trading mechanisms on Euronext-Paris are different from those on the NYSE. Euronext is, in effect, a fully automated system for order routing and execution. In fact investors can enter orders without going through their brokers and an order may be immediately executed if it can be matched against an order in the public limit order book; otherwise it becomes an entry in the public limit order book. In analysis of NYSE/NASDAQ trading statistics researchers frequently apply either a 50 percent or a 65 percent correction factor on NASDAQ trading data to achieve equivalence because a significant number of NASDAQ trades are market-maker broker trades while far fewer trades on the NYSE are market makerbroker trades. One might argue that one should apply a small correction factor to NYSE trades here. However, we choose to not apply any correction factor because in the case of these NYSE firms there are likely few market maker to broker trades and it is not clear what factor should be applied to French firms trading on the NYSE. 13

14 the six weeks prior and for the six weeks subsequent to decimalization. These were obtained from the Bank of New York Web Site. 14 In the initial analyses we examine pre- and post-decimalization average daily activity measures per security for both French and American securities on each exchange: average number of daily trades, average trade size in dollars, average number of daily market maker quotes per security, average percentage bid-ask spread and average bid and ask depths. For subsequent analyses we divide each European and American day into trading periods. In Europe we create four trading periods: 9:00-11:30, 11:30-13:30, 13:30-15:30 and 15:30-17:30. In the United States we create three trading periods: 9:30-11:30, 11:30-13:30, and 13:30-16:00. It should be noted that, by construction, the fourth European and first American period consist of the two hour period, daily, where both markets are open concurrently, during the time periods that we analyze. Thus the first three European periods consist of 6 ½ hours where trading of the French and American securities is on Euronext (and possibly at that time in London), the next period has trading in both Europe and the United States, and the final two periods in North American consist of 4 ½ hours of (only) NYSE trading. 4 Results 4.1 Aggregate Pre and Post Decimalization Trading Statistics A summary of daily aggregate trading activity during the period from 30 trading days prior to NYSE decimalization to 30 days after NYSE decimalization is provided in 14 A slightly better analysis would involve using daily exchange rates as Oppenheimer and Sabherwal (2003) did. During the two six week periods in question the exchange rate was quite stable indeed the difference in average exchange rates between the two six-week periods is less than 0.4 percent; the 14

15 Table 2 and Figures 1 and 2. In Panel A we present a summary of pre- and postdecimalization aggregate trading activity of NYSE firms on the NYSE and on Euronext- Paris. In Panel B we present a summary of the comparable aggregate trading activity for the French firms. As might be expected, the summary results for pre-decimalization trading in Panels A and B present a picture of these firms trading quite often and in relatively large trade sizes in their home markets. On average the American firms traded 1,605 times per day (or over 200 times per hour) on the NYSE during the six weeks prior to decimalization. The mean trade size $142,351 per trade is quite large and likely representative of significant underlying institutional activity. Similarly, the French firms trade, on average 2,932 times per day in Paris, with average trade size of $36,415. The trading of French firms (as ADRs) in New York seems more important, relative to their home market, than the trading of American firms in Paris relative to their home market. Average trade size of French firms on the NYSE prior to its decimalization is relatively large $65,261 per trade. It should be noted that prior to decimalization average trade sizes in New York are significantly larger than in France across all of these securities likely the result of these securities being largely institutional traded in New York as opposed to France. We turn now to changes in trading of the American and French securities after NYSE decimalization. Post-decimalization the average daily number of trades for the American firms on the NYSE is unchanged (1605 versus 1600 per day). We observe a statistically significant decline in average trade size, as expected, from about $142,000 to practical improvement from using daily exchange rates would be nominal and would have no quantitative impact on reported results. 15

16 $130,000 per trade (about 8.5 percent). Consistent with this, average daily dollar and share volume in these securities both declined significantly. 15 We speculate that the total decline in trading activity is structural in the sense that the pre-decimalization period is the six-week period around the turn of the year where institutions (and others) execute year-end window-dressing, tax-related, bonus-related and other portfolio revision strategies. 16 The French trading activity of these securities declined from 41 to 29 trades per day after decimalization, while the size of trades on Euronext did increase modestly, but insignificantly. The post-decimalization results for French firms are somewhat different from those for American firms. In aggregate, average daily trading of cross-listed French securities on Euronext-Paris increased by about eight percent (from 2932 to 3166 trades per day) during the post-nyse-decimalization period. This increase is not statistically significant. Average trade size in dollars (in number of shares) was unchanged (increased significantly). One might argue that these increases in activity might be related to Euronext s competitive responses to NYSE decimalization. 17 Average daily number trades on the NYSE of these firms was essentially unchanged from 139 pre- to 135 post-decimalization. Average trade size, however, declined by about thirty percent after 15 To conserve space we do not report over the counter trading of these securities in Table 2 (but do present their trading in Figures 1 and 2). However, we note that over the counter trading of these securities is material in both the pre-decimalization period (at about 1932 trades per day) and in the post-decimalization period (at about 1362 trades per day). The decrease is statistically significant. There is no significant change in trade size (32,771 pre-decimalization versus 33,113 post-decimalization). Thus inclusion of the over-the counter market accentuates the decline in trading of these NYSE issues during the six weeks after NYSE decimalization. As noted earlier, NASDAQ has not yet implemented decimalization during this post-nyse decimalization period; NASDAQ implemented decimalization in April Ahn et al. (1998) find that TSE volume of Canadian firms cross-listed in the U.S. declines after Canadian reduction in tick size from C$0.125 to C$0.05. They can not determine if this is due to decimalization or some other cause. In contrast, Oppenheimer and Sabherwal (2003) find a total increase in trading around NYSE decimalization for TSE dual-listed securities. However, those securities are far less likely to be as favored by U.S. institutions as the ones analyzed here. 17 These are considered in the next section. 16

17 NYSE decimalization. 18 As a result both dollar volume and share volume for these securities on the NYSE decline by over thirty percent. It is important to note that in the pre-decimalization period average U.S. trade size for these French firms was almost twice that on Euronext. After decimalization it is only 30 percent higher. Overall, it appears that on average, in contrast with expectations, that postdecimalization there is lower trading volume of NYSE securities than prior to decimalization, though it is unclear why. NYSE trades sizes for both American and French firms decline, by about 8 percent for American firms and about 30 percent for French firms. This is consistent with expectations. We now turn to the issues related to market-maker activity: the number of quotes they provide, their quoted spreads and the depth of their quotes Market Makers Numbers of Quotes, Spread and Depth In Table 3 and Figures 3 and 4, we present the mean number of daily quotes, the average daily percentage spreads and average daily bid and asked depths. The layout of the Panels of this Table is identical to that of Table 2. Panel A summarizes the U.S. firms parameters on the NYSE and in Paris. We find that NYSE specialists significantly decrease the number of quotes postdecimalization: an average of 2,299 per day per firm pre-decimalization versus about 2,196 per day per firm post-decimalization. As expected the dollar quoted spread declines by over 40 percent, from $ to $0.0612; average daily spread as a percentage of the mid-point of the bid and the asked declines from 0.31 percent to It should be noted that American over-the-counter trading of these large French firms is immaterial in both the pre- and post-decimalization periods. 17

18 percent (a decline of 43 percent). Depths of quotes decline by about 67 percent, consistent with the Bacidore et al. (2001) findings. In contrast, the number of Euronext- Paris quotes per day per firm increases by about 22 percent, from about 1,417 predecimalization to about 1,727 post decimalization. The spread decline (for American firms) in France is small and insignificant 5 percent: from $ to $ (or from about percent of midpoint of bid-ask to about percent of midpoint of bidask. Further, depths for American securities actually increase in France, albeit modestly. Turning to the French firms, we find that on the NYSE the number of quotes per day per firm increase significantly, from 338 to 421. We find that the NYSE spreads for the French firms decline significantly percent post-decimalization, from $ to $ (or from percent to percent of midpoint of bid-ask). In the U.S. the depth of quotes for French firms declines by about 40 percent. In France, the home market for these firms, the results are quite striking. First, the number of quotes increases by about ten percent after NYSE decimalization, while the quoted spread (in both a currency and percent of midpoint of bid-ask) decreases after NYSE decimalization. These results are statistically significant. Depths also increase significantly by about ten percent. The results presented in Panel B of Table 3 are consistent with expectations: declines in spread and depth in the United States and, due to competitive pressures, declines in spread in France for these firms. 4.3 Trading Statistics During the Trading Day 18

19 In Table 4 we present trading activity statistics by various times of day for the sample. In Panel A we present activity for both the American firms and the French firms on the NYSE. In Panel B we present activity for both the American firms and the French firms on Euronext-Paris. Panel A is divided into three parts, representing three periods of trading within the day. Period one is the first two hours of NYSE trading, Period two is the next two hours and Period three is the last 2 ½ hours. Panel B is divided into four parts, representing four period of trading within the day. Period one is the first 2 ½ hours of Euronext-Paris trading, Period two is the next two hours, Period three is the next two hours, and Period four is the last two hours. Period 1 in New York and Period 4 in Paris are the exact same two hours; it is during period 1 in New York (its opening two hours) and period 4 in Europe (its closing two hours) that the cross-listed firms simultaneously trade in both markets. It should be noted that in this Table the number of trades is measured on a daily basis in each period; to compare period 3 in New York with periods 1 and 2 in New York or to compare Period 1 in Paris with the other three periods one should multiply by.80 to adjust for 2 ½ hour versus 2 hour periods. In Figure 5-A we present comparisons of trading size for U.S. and French firms on the NYSE while both exchanges are open (Period 1) versus when only the NYSE is open (the average of Periods 2 and 3). In Figure 5-B we present comparisons of trading size for U.S. and French firms on Euronext when only Euronext is open (the average of periods 1, 2, and 3) versus when both exchanges are open (Period 4). These figures serve to illustrate the differences in impact of decimalization on the exchanges coupled with differences when both exchanges are open as opposed to when only one is open. 19

20 Considering first the trading of American firms on the NYSE, we find that there tends to be heavier trading during the first two trading hours than later in the day with about 280 trades per firm per hour in the pre-decimalization period versus about 210 trades per hour during period 2 and about 250 trades per hour during period 3. Further, the average size trade tends to be higher during the first two hours of the day versus later periods: the average sized trade in period 1 for these firms, prior to decimalization, is over $169,000; in contrast during period 2 it is about $119,000 and during period 3 it is about $133,000. In France there is somewhat more trading of these American firms during its first two hours of trading than during its other three two-hour periods: about 8 ½ trades per firm per hour during the first 2 ½ hours of trading on Euronext versus about four to six trades per hour during other periods. During the pre-decimalization period average trade size of U.S. firms on Euronext tends to be higher during the last two periods of the French trading day than during the first two periods. Post-decimalization the number of trades decreases significantly on the NYSE during its first two hours of trading and increases slightly during the other two periods, with an overall change in number of trades being close to zero, as presented in Table 2. Average trade size decreases in all three periods from its pre-decimalization levels; however only in period 2 (the middle of the day) is this statistically significant. The average number of transactions for these American securities on Euronext-Paris decreases during each period (but this decrease is only significant in the first two periods). We find little change in transactions sizes of American firms on Euronext- Paris. In aggregate the results on Panel A do not seem to seem to provide much evidence beyond that of Table 2: that decimalization seemed to result in decreased transaction size 20

21 on the NYSE. There is no obvious link between Period 1 NYSE trading and Period 4 Paris trading of NYSE firms evident in Table 4. Turning to the trading of French firms, a somewhat different pattern emerges. On the NYSE, prior to its decimalization, we find far more trading of French firms during its first two trading hours than later in the day: an average of 29 trades per hour per firm prior to 11:30 New York time, versus about 15 trades per hour per firm for Period 2 and about 20 trades per hour per firm from 1:30 to 4:00. In New York the transaction size of French firms prior to decimalization is somewhat higher during Period 1 than during Period 2 or 3. There is a large decline in transaction size each period: 28 percent in Period 1, 32 percent during period 2, and 36 percent during period 3. Daily volume declines significantly during each of the three periods. In France, both prior to and after decimalization, there are far more transactions in these firms executed during its first two hours of trading than during the other three periods. However, average trade size during the first two hours of trading is lower than during the other three trading periods. Average trade size is about 33 percent higher during the fourth period (during which the NYSE is open) than in the other periods. While average trade size during the first three periods actually goes up modestly post NYSE-decimalization, during the fourth period it actually goes down by about 4.6 percent; this decrease is not statistically significant. The results presented in Table 4 seem to indicate that the heaviest trading (in terms of both number of trades and size of trades) of French firms on the NYSE occurs during its first two hours of trading (while Euronext is open). While there are far more trades of the French firms on Euronext during its Period 1 (when the NYSE is closed), the average size of its Period 4 (when the NYSE is also open) trades is over 50 percent 21

22 larger than that of its Period 1 trades. Prior to NYSE decimalization average trade sizes of French securities on the NYSE dwarfed average trade sizes on Euronext. Subsequent to NYSE decimalization average transactions sizes in its period 2 and 3 are much closer to those of France, Periods 1-3 and, on average, about the same size of those in France, Period 4. These results are consistent with both NYSE decimalization lowering U.S. transaction size and competitive effects in the two markets in relationship to French securities. 4.3 Number of Quotes and Spreads and Depths during the Trading Day The analysis of number of quotes and spreads and depths during the trading day is presented in Table 5. The format of this Table is identical to that of Table 4. Graphical summaries of this information during periods when both markets are open (NYSE period 1 and Euronext period 4) versus when only one market is open (NYSE periods 2 and 3 and Euronext periods 1-3) by trading market are presented in Figures 6-A (quotes for all firms on the NYSE), 6-B (quotes for all firms on Euronext), 7-A (spreads on the NYSE), 7-B (spreads on Euronext), 8-A (depths on the NYSE), and 8-B (depths on Euronext). The results in Table 5 and these Figures are striking. For American firms, in New York prior to decimalization, on average the spread decreases from period 1 to 3 and depths gradually increase during the day. The number of quotes per hour is relatively constant. 19 Post decimalization, the average spread declines by 38 percent in Period 1 (from its pre-decimalization average level) and declines by 49 and 46 percent, 19 It should be noted again that Period 3 is 2 ½ hours; consequently the number of quotes should be multiplied by 0.80 to provide comparability to Periods 1 and 2. 22

23 respectively, in Periods 2 and 3. Depths decline by about 67 percent in each period after decimalization. In France prior to NYSE decimalization, spreads for U.S. firms gradually grow from period 1 and 2 to period 3 and then become lower than period 1 spreads during period 4. Depths at bid or ask are quite similar in periods 1, 2 and 3 and triple that size during period 4. Indeed, the average number of quotes during period 4 is about ten times that of prior periods. Post NYSE decimalization spreads in France for American firms decline each period; the decline in dollar quoted spread is significant in periods 2 and 3 and the decline in percentage quoted spread is significant in period 4. Depths in France after NYSE decimalization increase significantly during Periods 1-3 and remain at their pre-decimalization level during period 4. During Periods 1-3 the number of quotes for U.S. firms in France declines somewhat; in Period 4 it increases from its predecimalization level by almost 40 percent. The results in this Table, in essence disaggregated results from Table 3, are indicative of competitive activity in France in relationship to these American firms. With the NYSE having over 100 times the volume (in value) for these U.S. firms, the magnitude of these results seems quite surprising. In Table 5, on the NYSE there are far more quotes provided for the French firms during the first two hours of trading (the last two hours of trading in Paris) than for the rest of the day, both pre- and post-decimalization. Spreads, however, seem relatively stable throughout the day. Depths tend to be lowest during Period 2 and highest during Period 3, though Period 3 depths are not, on average, appreciably different from Period 1 depths. These results are at variance with typical results, which suggest spreads lowest in mid-day. One can speculate that these results are influenced by concerns about adverse 23

24 information about French firms occurring after the close of the business day in France (after 5:30 local time in Paris). On Euronext the spread for the French firms gradually declines from Period 1 to Period 3, both pre- and post decimalization. The Period 4 spread is modestly higher than the Period 3 spread. The post-nyse-decimalization decline in spreads on Euronext-Paris occurs fairly uniformly across the four trading periods. Prior to NYSE decimalization depths gradually grow through the day, but are considerably higher in Period 4 than in other periods. After decimalization, depths on Euronext are higher than before in each period. Finally the number of quotes seems to increase post-decimalization and the number is far greater during Period 4 than any other period. One might argue that these results are indicative of changed competition in the trading of these major French securities. 5. Conclusions This paper presents an analysis of the impact of New York Stock Exchange (NYSE) decimalization on the trading of U.S. firms listed on both the NYSE and on Euronext-Paris and the trading of French firms traded on both Euronext-Paris and (as ADRs) on the NYSE. We find that the U.S. trading of the American securities decreases materially after decimalization; French trading of these securities declines more modestly. While the average transaction size decreases in the U.S., it does not decrease on Euronext. The average transaction size of French stocks declines materially in the U.S. after NYSE decimalization. That is not accompanied by an increase in the number of transactions. In contrast, the Euronext number of transaction in French securities increased after NYSE decimalization and average transactions size remained constant. 24

25 As has been reported elsewhere, the NYSE spreads and depths for these American firms decline markedly after NYSE decimalization. The Euronext spread declined by only 4 percent and the Euronext depths actually increased. We find that the number of trades and the average transaction size for both French and U.S. firms are much higher on the NYSE during its first two hours of trading when Euronext is also open. We also find that the average transaction size for French firms on Euronext is much higher during its last two hours of daily trading when the NYSE is also open. Finally we find that there are far more quotes and far greater depths on Euronext during its final two hours of trading than at other times when the NYSE is not open. The results should be considered in relationship to the reasons for decimalization and received market microstructure research. While it might be argued that the overall pricing of the securities is more efficient especially on the NYSE we note there is little evidence of increased trading volume. Indeed with significantly lowered depths in the U.S., institutions may have found it more difficult to trade in their accustomed block sizes. One issue for further research is further exploration of the size of trades. Oppenheimer and Sabherwal (2003) decompose trading into various trade size categories to determine whether retail trading did increase and how institutional trading changed. The results presented here suggest that such further analysis is needed. An obvious issue: were large traders of these stocks disadvantaged by decimalization and how did the magnitude of any disadvantage change over the course of the day as trading opened on the NYSE (while Euronext was open) and then Euronext closed. This is also a particularly interesting issue for the French securities which are primarily traded by 25

26 institutions in the U.S. and retail traders on Euronext. As briefly discussed earlier, it is possible that the lack of a change in number of trades of the U.S. securities is at least partially related to turn of the year activity of the institutions increasing the number of trades in the pre-decimalization period above normal volume for these securities. That issue also suggests the need for further exploration. Finally, given the pending proposal merger of the NYSE and Euronext, stakeholders in both of these venues might want to contemplate these results as they go forward. 26

27 References Ahn, Hee-Joon, Charles Q. Cao, and Hyuk Choe, 1996, Tick size, spread, and volume, Journal of Financial Intermediation 5, 2-22 Ahn, Hee-Joon, Charles Q. Cao, and Hyuk Choe, 1998, Decimalization and competition among exchanges: Evidence from the Toronto Stock Exchange cross-listed securities, Journal of Financial Markets 1, Angel, J.J., 1997, Tick size, share price, and stock splits, Journal of Finance 52, Bacidore, Jeff, Robert Battalio, Robert Jennings, and Susan Farkas, 2001, Changes in order characteristics, displayed liquidity, and execution quality on the New York Stock Exchange around the switch to decimal pricing, New York Stock Exchange Working paper # Bodie, Zvi, Alex Kane, and Alan J. Marcus, 2005, Investments (6 th Edition), McGraw- Hill Irwin: Boston. Eun, Cheol S. and Sanjiv Sabherwal, 2002, Cross-border listings and price discovery: Evidence from U.S. listed Canadian stocks, Journal of Finance, forthcoming. Foerster, Stephen R. and G. Andrew Karolyi, 1998, Multimarket trading and liquidity: A transaction data analysis of Canada-U.S. interlistings, Journal of International Financial Markets, Institutions and Money 8, Goldstein, Michael A. and Kenneth A. Kavajecz, 2000, Eighths, sixteenths, and market depth: changes in tick size and liquidity provision on the NYSE, Journal of Financial Economics 56, Harris, Lawrence E., 1994, Minimum price variations, discrete bid-ask spread, and quotation sizes, Review of Financial Studies 7, Harris, Lawrence E., 1997, Decimalization: A review of the arguments and evidence, Working paper (University of Southern California, Los Angeles, CA). NYSE Research, 2001a, Comparing bid-ask spreads on the New York Stock Exchange and Nasdaq immediately following Nasdaq decimalization, New York Stock Exchange, July 26. NYSE Research, 2001b, Decimalization of trading on the New York Stock Exchange: A report to the Securities and Exchange Commission, New York Stock Exchange, September 7. 27