Who Wins When Exchanges Compete? Evidence from Competition after Euro Conversion

Who Wins When Exchanges Compete? Evidence from Competition after Euro Conversion Abstract Using euro conversion as the trigger, we examine what drives volume and spread changes when stock exchanges compete. Results show average trading costs on European exchanges decrease almost 9%, and turnover increases over 30%. Trading costs decline or remain unchanged on all exchanges, but volume deteriorates in some markets and improves in others. Frankfurt, Paris, London, and Milan are winners, while Madrid and Brussels lose volume. We examine the role of the spread-volume relation, firm characteristics, exchange trading rules, and country-level factors in determining these outcomes. Results suggest that euro conversion prompted competition by increasing transparency in market prices. 1. Introduction When European equity markets closed on December 30 1998, all stocks were priced in local currency, for example French francs in Paris, Deutsche marks in Frankfurt, and Italian lira in Milan. When markets reopened five days later on January 4 1999, stocks in eleven of the markets were priced in euros, allowing investors to compare the prices of stocks across all eleven markets in the same currency. Domowitz, Glen, and Madhavan (1998, p. 2011) define transparency as “the extent to which price information in the two markets is observable.” Euro conversion instantly increased price transparency across these markets. We propose that this increased transparency may have triggered a round of competition among European stock exchanges. We use this natural experiment to examine an important question in the literature on competition among stock exchanges: what drives spread and volume changes when stock exchanges compete? To address this question, we analyze changes in trading costs (measured here by bid-ask spreads) and trading activity for nine euro and three non-euro European exchanges around euro conversion, from the fourth quarter in 1998 to the fourth quarter in 1999. Academic models of exchange competition suggest that exchanges compete to attract firm listings and/or trading volume (see the discussion of related literature in sub-section 2.1). According to the models, exchanges have multiple competitive levers to pull (trading costs, listing requirements, fees, etc.). Previous research shows that competition is often prompted by a specific external trigger, and that it is associated with big decreases in trading costs and substantial shifts in trading volume. However, the exact mechanism linking the competitive trigger to the trading cost and volume shifts is still unclear. Exchanges may compete directly by reducing spreads to attract volume. Alternatively, changes in spreads and volume may reflect the characteristics of firms that list on the different exchanges, exchange trading rules, or the broader institutional environment. Three constraints hamper prior research on exchange competition. First, the limited sample size (in terms of number of firms and/or exchanges) means prior researchers do not observe much cross-sectional variation. It is hard, therefore, to analyze the role of firm- or exchange-level characteristics in competition, or whether competition spills over to nearby exchanges. Second, changes in volume and changes in bid-ask spreads are endogenously determined in equilibrium (e.g., Glosten and Milgrom, 1985; Admati and Pfleiderer, 1988), so it is challenging to identify the direction of causation in this relation. Finally, because competition often plays out within the context of one or more on-going general trends, it is difficult to establish whether the competitive trigger is responsible for the observed reactions. Our research design addresses each of these three limitations. First, our natural experiment, euro adoption on January 4 1999, allows us to analyze a much broader range of exchanges and firms than most previous studies. With this sample, we explore the role of firm- and country-level characteristics as well as changes in spreads as determinants of the outcome of a round of competition among euro and non-euro European exchanges. We estimate a system of simultaneous equations using two stage least squares (2SLS) to control for endogeneity in the relation between changes in costs and volume. Finally, we structure our analysis as a difference-in-differences test to control for other concurrent trends, enabling us to more confidently attribute changes in spreads and volume to the euro conversion event. We find that in the year following euro conversion, bid-ask spreads across European exchanges fall an average of almost 9% while turnover (defined as trading volume scaled by shares outstanding) rises over 30%. Trading costs either decrease significantly or remain unchanged on each exchange, but turnover increases on some exchanges (the “winners”), and falls on others (the “losers”).1 Both euro and non-euro European markets are affected. These changes are commensurate with other notable stock exchange events, such as London’s Big Bang and the New York Stock Exchange's (NYSE) decimalization. Changes in spreads and turnover are significantly larger than those that occur in the prior year, and are significantly greater than those for a control group of NYSE firms. Difference-in-differences analysis with European and NYSE firms, controlling for numerous firm characteristics, is consistent with our hypothesis that euro conversion prompted the sizeable spread and volume shifts we document. Lastly, our 2SLS results show that controlling for the endogeneity of volume and spreads crucially affects these inferences. Our results show that Milan, Frankfurt, Paris, and London are the winners with significant increases in turnover, while Brussels and Madrid are the biggest losers with significant declines. Variations in both firm-specific and country-level factors help explain which exchanges win and which lose volume. Volume increases more for exchanges that list smaller, lower volatility firms and firms in the technology, telecom, and consumer goods and services industries. Exchanges located in countries with higher Gross Domestic Product (GDP) growth, lower accounting standards, and larger increases in the number of analysts also gain more volume. Several of the significant firm- and country-level variables (such as firm size and accounting standards) often serve in the literature as proxies for asymmetric information. We find that exchanges that begin the period with the weakest information environments are the biggest volume winners. Furthermore, the effect of firm size on turnover is stronger for firms listed on less transparent exchanges. These results are consistent with our expectation that euro conversion prompted competition by increasing transparency in market prices. Finally, the welfare consequences of changes in spreads are large; investors save almost €568 (approximately $570) million in trading costs during the fourth quarter of 1999 from the decline in effective bid-ask spreads after euro adoption. Our results contribute to the literature on exchange competition by providing insights into the question of who wins when exchanges compete. Controlling for the endogeneity of volume and spreads, we find that both firm- and country-level factors affect volume changes, while firm, country, and exchange trading rules affect spreads. Our results consistently show that transparency is an important driver of changes in spreads and volume. Moreover, we find a strong role for the exchange’s industry composition. These results suggest that it is important to consider all of these factors (i.e., the spread/volume relation, firm characteristics, exchange-level characteristics, and country-level macroeconomic and institutional characteristics) simultaneously when examining the outcomes of exchange competition. Our results also contribute to the literature on the agglomeration versus fragmentation of trading. We show only marginal increases in the trading activity market share of Europe’s largest exchanges, suggesting that euro conversion did not contribute to a “winner-take-all” outcome. Our welfare calculations show that the effects of competition are economically large and differ substantially between the winning and losing exchanges. Finally, we are the first article to provide statistically significant evidence that competition spills across borders.2 In sum, our results provide a rich understanding of the factors that affect both spreads and volume when exchanges compete. The remainder of our article is organized as follows. In Section 2, we describe our research design and sample. Univariate results are presented in Section 3; multivariate difference-in-differences regressions are reported in Section 4. Section 5 discusses welfare consequences, and Section 6 provides additional evidence related to competition. Section 7 concludes. 2. Research Design and Sample 2.1 Theory and Related Literature Our research is related to models of competition among stock exchanges, which typically assume that exchanges compete to attract either trading volume (order flow) or firm listings. In models of the competition for order flow, exchanges adjust trading fees (Ramos and von Thadden, 2008; Pirrong, 2000), disclosure requirements (Huddart, Hughes, and Brunnermeier, 1999), margin requirements (Santos and Scheinkman, 2001), and not-for-profit status and membership size (Pirrong, 2000). In models by Chemmanur and Fulghieri (2006), Foucault and Parlour (2004), and Amira and Muzere (2011), exchanges modify listing standards or listing fees to attract more firm listings. Di Noia (2001) and Arnold et al. (1999) look at mergers and alliances as responses to exchange competition. Empirical investigations of competition for order flow generally use one of four different approaches. The narrowest approach is lab experiments that examine the effect of different trading rules on the volume and spreads for one asset (Lamoureux and Schnitzlein, 1997; Bloomfield and O’Hara, 2000). Perhaps the broadest approach is papers that look at the cross section of spreads or trading volume to measure the effect of country-level institutional factors and exchange characteristics on exchange competition (e.g., Eleswarapu and Venkataraman, 2006; Frost, Gordon, and Hayes, 2006; Lo, 2013). Microstructure papers that examine the effects of different trading mechanisms across multiple exchanges, while not directly motivated as studies of competition, would fall within this category (Domowitz, Glen, and Madhavan, 2001; Jain, 2006). The third approach is to examine the competition for order flow for the same firm listed on multiple exchanges to determine the effect of exchange and firm characteristics on volume and/or spreads. The cross-listing literature falls into this category. See, for example, Battalio, Greene, and Jennings (1997), Blume and Goldstein (1997), Kwan, Masulis, and McInish (2015), and Halling et al. (2008). The final approach, and the one closest to our experiment, examines how exchanges competitively respond to a specific event (e.g., Aggarwal and Angel, 1999; Arnold et al., 1999; Bessembinder, 2003; Brown, Mulherin, and Weidenmier , 2008; Dewenter, Kim, and Novaes, 2010). Most of these papers narrowly focus on only two exchanges.3 In order to draw conclusions about the effect of exchange competition on spreads and volume, we must first establish that euro conversion was a competitive event. The path to euro conversion was long and deliberate. (See Internet Appendix, Table A1 for a summary of major actions and a timeline of related events.) The Treaty of Maastricht on European Union, signed in 1992, created the framework and timetable for euro adoption. In the May 8 1998 Joint Communique, eleven European nations formally agreed to convert their currencies to the euro, announcing irrevocable exchange rates among themselves. On December 31 1998, participating countries announced the euro conversion rates (see Figure 1), keeping the cross-country exchange rates established in May unchanged. The next day, on January 1 1999, the euro was adopted for all electronic assets and transactions. When the European stock exchanges re-opened on January 4 1999, the prices for all firms listed on the eleven euro country exchanges were denominated in euros. Actual euro coins and notes were not introduced until 2002. For our purposes, it is important to note that the only major euro-related change during our sample year 1999 was euro conversion. Figure 1. View largeDownload slide Euro conversion rates. The December 31 1998 press release with the original conversion rates can be found at the European Central Bank website: http://www.ecb.int/press/pr/date/1998/html/pr981231_2.en.html. Figure 1. View largeDownload slide Euro conversion rates. The December 31 1998 press release with the original conversion rates can be found at the European Central Bank website: http://www.ecb.int/press/pr/date/1998/html/pr981231_2.en.html. Our argument that euro conversion served as a trigger for competition within the wider scope of changes made to prepare for the euro is similar to numerous other exchange-competition papers that identify a single, specific action within a broader trend, usually a regulatory change, as the trigger for competition (e.g., Arnold et al., 1999; Dewenter, Kim, and Novaes, 2010; Spankowski, Wagener, and Burghof, 2012). Ramos (2003) argues that a specific external trigger is necessary to spur competition or reform because exchanges are reluctant to change the status quo. Some might contend that euro conversion was trivial, a simple change of numeraire that was well known in advance. However, euro conversion had a real effect—it enabled investors to immediately and easily compare the prices of stocks across eleven different markets. Euro conversion therefore increased price transparency across those markets. Whether or not the change in transparency was trivial or significant enough to prompt a competitive response by the exchanges is an empirical question. A related challenge for our research design is adequately controlling for other broad trends occurring at the same time that could also affect trading costs and volume and thus confound interpretation of our results. The three most obvious forces are: changes to trading technology that prompted the creation of electronic communications networks, or ECNs (key events identified in Panel B of Table A1 in the Internet Appendix); portfolio rebalancing in response to the removal of investment barriers in preparation for the single market and the euro; and the internet bubble. We design our difference-in-differences analyses to control for more general trends and conduct several robustness tests (see sub-section 4.4) to address these potential alternative explanations. We extend existing research in several dimensions. First, while many of the empirical papers of exchange competition acknowledge endogeneity between spreads and volume, only one (Kwan, Masulis, and McInish, 2015) controls for it. We control for this endogeneity with a two-stage-least squares empirical specification, and our results show that this correction affects inferences. Second, Garvey (1944) and Arnold et al. (1999) address the potential for competition to spill over onto nearby exchanges. Our setting provides a natural set of exchanges that could be affected by spillover from competition triggered by euro conversion, the non-euro European exchanges, giving us an opportunity to test explicitly for spillovers.4 Third, the literature identifies four sets of factors that potentially drive the relation between competition and trading volume: spreads, firm characteristics, exchange trading environment, and country-level macroeconomic or institutional factors. Although previous research examines various subsets of these factors, no one to date has simultaneously examined all four. Because our experimental design includes multiple exchanges with many firms, we are able to examine all four factors, providing a more comprehensive examination of what drives volume (and spread) changes when exchanges compete. 2.2 Data and Sample Description The main source for our data is Thomson Reuters’ DataStream (DS). We collect the daily closing transaction price, ask price, bid price, and trading volume for firms listed on twelve exchanges in the euro-zone countries (Amsterdam-Netherlands, Brussels-Belgium, Dublin-Ireland, Frankfurt and Xetra-Germany, Helsinki-Finland, Lisbon-Portugal, Luxembourg-Luxembourg, Madrid-Spain, Milan-Italy, Paris-France, Vienna-Austria) and on three major European exchanges in non-euro zone countries (Copenhagen-Denmark, London-United Kingdom, and Swiss-Switzerland) for 1998 and 1999. We also collect data on NYSE trading from the Center for Research in Security Prices (CRSP) over the same period to use as a control sample. Alternatively, we considered using trading on the same European exchanges in the periods immediately prior to or after our sample period as the benchmark. Unfortunately, prior period data are incomplete, and post period trading is distorted due to the internet trading boom. Another possible source for control firms is Nasdaq. In Figure 2, we graph the value of shares traded for the sample of European exchanges we analyze relative to the NYSE and Nasdaq (all normalized to 100 in 1997) in Panel A, and the NYSE, Nasdaq and MSCI-Europe stock market indices in Panel B (all normalized to 100 at the beginning of 1998). Figure 2 shows that overall trading activity and market performance for NYSE firms track those in Europe very closely just prior to and during our sample period, whereas Nasdaq firms do not. The internet bubble strongly affected Nasdaq firms during our sample period.5 Therefore, we believe contemporaneous NYSE data are a more appropriate benchmark. Figure 2. View largeDownload slide Comparison of trading activity and market performance on the European exchanges, the NYSE, and Nasdaq. Panel A. Value of shares traded Panel B. Stock market indices Source. Panel A: World Stock Exchange Fact Book, various issues. European sample includes data for domestic firms listed on exchanges in our eleven European countries. Values normalized to equal 100 in 1997. Panel B: DataStream. Values normalized to 100 on January 1998. Figure 2. View largeDownload slide Comparison of trading activity and market performance on the European exchanges, the NYSE, and Nasdaq. Panel A. Value of shares traded Panel B. Stock market indices Source. Panel A: World Stock Exchange Fact Book, various issues. European sample includes data for domestic firms listed on exchanges in our eleven European countries. Values normalized to equal 100 in 1997. Panel B: DataStream. Values normalized to 100 on January 1998. Table I summarizes trading activity and other descriptive statistics for our exchanges. The first two data columns provide the total number of firms and domestic market capitalization for all domestic firms on the exchanges as of December 31 1998 from the World Stock Exchange Fact Book. In terms of market capitalization, both the NYSE and London are substantially larger than the other exchanges, at nine and two trillion euros, respectively. The smallest exchanges in market capitalization are Vienna, Luxembourg, and Lisbon. Table I. Summary descriptive statistics for the exchanges This table provides summary descriptive statistics for fifteen European exchanges and the NYSE for the fourth quarter of 1998. The first twelve exchanges listed in the table are located in countries which adopted the euro in 1999; the next three exchanges are located in European countries that did not adopt the euro. The last two rows report results for the Overall (Europe) and NYSE samples. Exchange data as of December 31 1998, in columns 1 and 2 are from the World Stock Exchange Fact Book. Irish exchange data are from The Handbook of World Stock, Commodity and Derivatives Exchanges, accessed via Factiva. Frankfurt and Xetra Exchange data are combined. Trading data in columns 3–6 for the European firms listed on the exchanges are from DataStream or KIT. NYSE data are from CRSP. Total Trading Volume is the average daily trading volume for firms in our sample (expressed in millions of shares traded) for the fourth quarter of 1998, summed over all the stocks with volume data for each exchange. Domestic Market Capitalization is the average daily market capitalization (the daily closing price times shares outstanding, expressed in billions of euros) for the fourth quarter of 1998 for firms located in the exchange’s domestic country summed over all the stocks with price and shares outstanding data in our sample for each exchange. The initial sample includes all firms with at least one ask quote between October 1 1998 and December 31 1999. The final sample includes all firms with at least fifteen non-missing trading volume and quote observations (for which the ask quote exceeds the bid quote) in both 1998Q4 and 1999Q4, average daily market capitalization of at least 50 million euros during December 1998, and a minimum average daily trading volume of 1000 shares during 1998Q4. Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 Table I. Summary descriptive statistics for the exchanges This table provides summary descriptive statistics for fifteen European exchanges and the NYSE for the fourth quarter of 1998. The first twelve exchanges listed in the table are located in countries which adopted the euro in 1999; the next three exchanges are located in European countries that did not adopt the euro. The last two rows report results for the Overall (Europe) and NYSE samples. Exchange data as of December 31 1998, in columns 1 and 2 are from the World Stock Exchange Fact Book. Irish exchange data are from The Handbook of World Stock, Commodity and Derivatives Exchanges, accessed via Factiva. Frankfurt and Xetra Exchange data are combined. Trading data in columns 3–6 for the European firms listed on the exchanges are from DataStream or KIT. NYSE data are from CRSP. Total Trading Volume is the average daily trading volume for firms in our sample (expressed in millions of shares traded) for the fourth quarter of 1998, summed over all the stocks with volume data for each exchange. Domestic Market Capitalization is the average daily market capitalization (the daily closing price times shares outstanding, expressed in billions of euros) for the fourth quarter of 1998 for firms located in the exchange’s domestic country summed over all the stocks with price and shares outstanding data in our sample for each exchange. The initial sample includes all firms with at least one ask quote between October 1 1998 and December 31 1999. The final sample includes all firms with at least fifteen non-missing trading volume and quote observations (for which the ask quote exceeds the bid quote) in both 1998Q4 and 1999Q4, average daily market capitalization of at least 50 million euros during December 1998, and a minimum average daily trading volume of 1000 shares during 1998Q4. Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 The final four columns of Table I provide information on our sample selection. The initial European exchange sample has 6,695 firm lists, satisfying the requirement that a firm has at least one non-missing ask quote observation for the 15 month period from October 1 1998 through December 31 1999.6 Trading volume data for German firms on DataStream are scarce for this period, so we replace the DataStream trading volume for Frankfurt and Xetra using volume data purchased from Karlsruhe Institute of Technologie (KIT).7 Around the euro conversion day of January 1 1999, our empirical analysis focuses on two periods: the last quarter of 1998 (1998Q4) and the last quarter of 1999 (1999Q4). Comparing 1998Q4 to 1999Q4 allows some time for competitive effects to take place and also controls for any potential calendar year effects.8,9 We further require each firm to have at least fifteen non-missing trading volume and quote observations (for which the ask quote exceeds the bid quote) in both 1998Q4 and 1999Q4. This screen leads us to drop the Dublin, Luxembourg, and Vienna exchanges from our sample due to lack of data. Finally, coverage on DataStream varies widely across exchanges. To control for potential selection biases associated with variations in coverage and to make sure that our comparisons across exchanges are between similar, large, and actively traded firms, we require that sample firms have a minimum average daily market capitalization of 50 million euros during December 1998, and a minimum average daily trading volume of 1,000 shares during 1998Q4.10 The resulting sample includes thirteen exchanges (nine euro exchanges, three non-euro European exchanges, and the NYSE), with a total of 2,183 European firm lists and 1,749 NYSE firm lists. The Frankfurt Stock Exchange and Xetra are both owned by Deutsche Borse. Xetra was created in 1997 as an electronic trading platform for the Frankfurt Stock Exchange.11 During our sample period, DataStream separately reported trading via these two systems. Since Xetra was a new electronic exchange with relatively rapidly expanding volume, we exclude Xetra from some of our analyses. Excluding Xetra reduces the sample of European firm lists to 2,063. We considered analyzing individual firms that are cross-listed on multiple exchanges. Unfortunately, when a firm is listed on more than one exchange during our sample period, most trading takes place on one exchange, so the second listing usually does not pass our data screens. Ramos and von Thadden (2008, Table I) also note that although there are many foreign firms cross-listed on the European stock exchanges, the total value of foreign trading on most exchanges is negligible. Furthermore, the coverage of cross-listing status on DataStream for firms that are no longer listed is very poor.12 To address our research questions, we compute several statistics to analyze changes in trading costs and trading volume around euro conversion. We focus on two alternative measures of trading costs: percentage spreads and effective spreads.13 Percentage spreads are defined as (Ai,t−Bi,t)/Mi,t ⁠, where Ai,t and Bi,t are the closing ask and bid prices, respectively, for stock i on day t, and Mi,t is the midquote, Mi,t=(Ai,t+Bi,t)/2. We define the (percentage) effective spread as 2*abs(Pi,t−Mi,t)/Mi,t ⁠, where Pi,t is the closing transaction price for stock i on day t. We use both daily share volume (⁠ Voli,t ⁠) and turnover to measure trading activity. Turnover is given by Voli,t/SHROUTi,t ⁠, where SHROUTi,t is the number of shares outstanding for stock i on day t. 3. Univariate Analysis 3.1 Univariate Changes Table II reports statistics for our trading cost and trading activity measures by exchange for 1998Q4 and 1999Q4. For each firm, we calculate the average statistic across all days in a given quarter. Averages reported in Table II reflect the value-weighted average across firms, where the weights are the average firm market capitalization during December 1998. We also report percentage changes from 1998Q4 to 1999Q4, a t-test for whether the percentage change differs significantly from zero, and a nonparametric sign test of changes in trading costs and volume.14 Statistics that are significant at the 5% level are noted in bold type. Table II. Summary statistics of changes in trading costs and trading activity This table provides mean values for measures of trading costs and trading activity for sample firms listed on nine euro exchanges, three non-euro exchanges and the NYSE. Mean daily values are provided for 1998Q4 and 1999Q4. For each firm, we compute the mean of each statistic each quarter; reported results are value-weighted averages across firms, where the value weights reflect the average December 1998 market capitalization. Panel A reports Percentage Spreads, defined as (Ai,t−Bi,t)/Mi,t where Ai,t ⁠, Bi,t and Mi,t are the closing ask price, bid price and midquote, respectively, for stock i on day t. Panel B reports the Effective Spread, defined as 2*abs(Pi,t−Mi,t)/Mi,twhere Pi,t is the closing transaction price. Panel C reports Volume (in thousands of shares). Panel D reports Turnover, defined as volume divided by shares outstanding. We also report the percentage change from 1998Q4 to 1999Q4, along with t-statistics testing whether the percentage change is statistically different from zero. The last three columns report the number of firms on each exchange for which the firm-level quarterly average statistic increased or decreased from 1998Q4 to 1999Q4, along with results of the z-statistic for a nonparametric sign test. This sign test is calculated as (Number of Firm Lists with Positive Changes − 0.5 × N)/ N/4 ⁠, where N is the total number of firm lists. Changes that are significantly different at the 5% level or better are in bold type. Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Table II. Summary statistics of changes in trading costs and trading activity This table provides mean values for measures of trading costs and trading activity for sample firms listed on nine euro exchanges, three non-euro exchanges and the NYSE. Mean daily values are provided for 1998Q4 and 1999Q4. For each firm, we compute the mean of each statistic each quarter; reported results are value-weighted averages across firms, where the value weights reflect the average December 1998 market capitalization. Panel A reports Percentage Spreads, defined as (Ai,t−Bi,t)/Mi,t where Ai,t ⁠, Bi,t and Mi,t are the closing ask price, bid price and midquote, respectively, for stock i on day t. Panel B reports the Effective Spread, defined as 2*abs(Pi,t−Mi,t)/Mi,twhere Pi,t is the closing transaction price. Panel C reports Volume (in thousands of shares). Panel D reports Turnover, defined as volume divided by shares outstanding. We also report the percentage change from 1998Q4 to 1999Q4, along with t-statistics testing whether the percentage change is statistically different from zero. The last three columns report the number of firms on each exchange for which the firm-level quarterly average statistic increased or decreased from 1998Q4 to 1999Q4, along with results of the z-statistic for a nonparametric sign test. This sign test is calculated as (Number of Firm Lists with Positive Changes − 0.5 × N)/ N/4 ⁠, where N is the total number of firm lists. Changes that are significantly different at the 5% level or better are in bold type. Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Results in Table II show that percentage spreads decrease almost 9% for the Overall (Europe) sample; the decrease in percentage spreads is significant based on both statistical tests. Results for effective spreads also show a significant decrease using the nonparametric test, although there is no significant change based on the t-statistic.15 Spreads either decrease significantly or remain unchanged for each of the European exchanges; there are no statistically significant increases in any trading cost measure. In particular, we see significant declines in trading costs on all three non-euro European exchanges in addition to many of the euro exchanges. The two exchanges that do not change percentage spreads based on either test statistic are Brussels and Madrid. Consistent with prior studies (e.g., Wahal, 1997; Foerster and Karolyi, 1998; Klock and McCormick, 1999; de Fontnouvelle, Fishe, and Harris, 2003), spreads decrease when there is more competition. For comparison, we also report results for the 1,749 NYSE firms that pass our data screens. Results for the NYSE show a significant increase in percentage spreads and mixed results for the change in effective spreads. Regarding trading activity, Table II shows that volume and turnover both increase over 30% for the full European sample. However, trading activity increases on some exchanges and decreases on others. Volume and/or turnover increase on Frankfurt, Milan, Paris, Xetra, and London, and also on the NYSE. Both measures decrease significantly on Brussels and Madrid.16 The effects we find are similar in economic magnitude to effects documented in prior analyses of other major events at stock exchanges (e.g., Brown, Mulherin, and Weidenmier, 2008; Pagano and Roell, 1990; Bessembinder, 2003). They are also significantly larger than changes that occurred in 1998 before euro conversion, suggesting that the effects we document are due to euro conversion itself rather than events associated with the longer path leading up to euro conversion.17 One potential concern is whether our results reflect euro adoption or perhaps something else that was going on at the same time. To address this concern, we conduct two additional tests. First, we repeat our test for a European Union (EU) country that adopted the euro after our event date.18 Specifically, we repeat our experiment for companies on the Athens stock exchange, using their euro adoption event date of January 1 2001. Results show that consistent with competitive effects, effective spreads decrease significantly more for firms on the Athens stock exchange around Greece’s euro adoption (from the fourth quarter of 2000 to the fourth quarter of 2001) than they do for a control sample of NYSE firms over the same period (Internet Appendix Table A9; see sub-section 3.2 for details on our NYSE matching procedure). Contrary to predictions, percentage spreads increase more in Athens than on the NYSE. t-statistics comparing changes in trading activity in Athens with those on the NYSE are not significant. However, nonparametric tests show that both volume and turnover increase significantly on Athens and do not change on the NYSE, and the chi-square test of medians shows that volume increases significantly more on Athens than the NYSE. It is difficult to draw strong inferences from these results given that we only have data for one exchange, and that euro adoption on this exchange occurred during a period of unusual stock market performance right after the internet bubble. Our overall conclusion is that the Athens stock exchange shows some of the same effects upon euro adoption in 2001 that we document for the original euro adoption in 1999. Second, we also rerun our experiment using our original exchanges for a placebo event day, January 1 2004. To implement this test, we collect data for the fourth quarter of 2003 and the fourth quarter of 2004, using the same exchanges and sample selection criteria we use for our original sample. Results (reported in the Internet Appendix, Table A10, Panel A) show that spreads also decrease significantly from 2003 to 2004 for the full sample of European firms. However, there is no significant change in trading volume, and turnover decreases significantly over the placebo window for the overall European sample. Moreover, Wilcoxon tests for changes in spreads and trading activity (Panel B) show that changes during 1998/99 are greater in magnitude than those in 2003/04 on most of the individual exchanges. Collectively, these results support our claim that the results we document during 1998/99 reflect a significant competitive event.19 3.2 Changes Relative to the NYSE To control for overall market trends, we next conduct univariate tests of whether changes for firms on the European exchanges differ significantly from our NYSE control firms. For each European firm, we select (with replacement) a matching NYSE firm with the closest market capitalization from the set of firms in the same industry.20 In Table III, we report the mean percentage change for the European firms relative to the corresponding change for their matching NYSE firms. So, for example, the first number in the panel (–0.604 for Amsterdam) shows that value-weighted percentage spreads for Amsterdam firms decline 60% more than spreads for their corresponding matched firms on the NYSE over the same period. Table III. Changes in trading costs and trading activity relative to NYSE This table reports percentage changes in trading costs and trading activity relative to the NYSE. Trading cost and trading activity measures are as defined in Table II. The table reports results for each individual European exchange relative to NYSE matching firms. For each European firm, the matching firm is the NYSE firm with the closest market capitalization from the set of firms in the same industry. We report the mean difference (Europe–NYSE) in percentage changes in each variable, and a t-test (in parentheses) of whether this difference is significantly different from zero. Significant differences at the 5% level or better are in bold type. Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Table III. Changes in trading costs and trading activity relative to NYSE This table reports percentage changes in trading costs and trading activity relative to the NYSE. Trading cost and trading activity measures are as defined in Table II. The table reports results for each individual European exchange relative to NYSE matching firms. For each European firm, the matching firm is the NYSE firm with the closest market capitalization from the set of firms in the same industry. We report the mean difference (Europe–NYSE) in percentage changes in each variable, and a t-test (in parentheses) of whether this difference is significantly different from zero. Significant differences at the 5% level or better are in bold type. Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Controlling for overall market movements, our main inferences hold. Spreads either decrease or remain unchanged relative to the matched NYSE firms. Trading activity declines significantly relative to matching firms on some exchanges (including Brussels, Helsinki, Madrid, and Copenhagen), and increases on others (Milan and Xetra). The relatively large jump in Xetra volume reflects the fact that Xetra was a very young and rapidly growing exchange at the time, which is why we exclude it from subsequent analyses. Finally, we note that, once we control for certain firm characteristics (market capitalization and industry) through our matching procedure, the list of volume winners and losers shifts somewhat. These results suggest to us that it is important to control for firm characteristics when evaluating the outcome of exchange competition. 3.3 A Comment on Tick Sizes Euro conversion on January 4 1999, changed the price unit (the currency), which in turn affected the relative tick size. (See Figure 1 for the euro conversion rates.) Conversion to euro prices lowered the price for all stocks except those in Ireland. Since the tick represents a lower bound on bid-ask spreads, changes in tick sizes may have affected bid-ask spreads which may confound our ability to identify the effect of competition. To explore this possibility, we calculate the relative tick size for each exchange.21 Controlling for changes in price levels, results indicate the minimum tick size decreased significantly on Amsterdam and Copenhagen. Therefore, decreasing tick size may be the cause of the decrease in spreads observed on Amsterdam and Copenhagen. In contrast, the effective tick size increased on most of the other euro exchanges. Despite the increase in tick size for these markets, however, previous results show that spreads decreased on many of them. We cannot reject that the correlation between exchange-level changes in spreads and changes in tick sizes equals zero. Overall, effects such as increased competition appear to overwhelm any tick size effects around euro conversion. 4. Multivariate Regressions So far we have shown that euro conversion is associated with significant changes in spreads and volume. Trading costs decrease or remain unchanged in all European markets. Trading activity increases significantly on some exchanges and decreases on others; there are some volume winners and some losers. These effects extend to non-euro European exchanges, and do not result mechanically from tick size changes. Univariate results in Section 3 are consistent with competition among exchanges. However, they do not control for confounding shocks or differences in the types of listed firms. To address these issues, we rely on a multivariate difference-in-differences methodology. This approach allows us to estimate the effect of the treatment (euro conversion) by controlling for confounding shocks with a set of firms that are similar except for the treatment. We conduct two sets of analyses. First, we consider the euro firms as the treatment group and the non-euro European firms as our control group. Because univariate results suggest that competitive effects extend to the non-euro European firms, we also conduct our difference-in-differences analysis combining euro and non-euro European exchange firms into the treatment group, and use NYSE firms as our control group. The NYSE firms should control for general global equity trends. Our sample screen requiring only large and actively traded firms provides some assurance that our treatment and control firms are similar. In addition, our multivariate regressions control for firm-specific characteristics as of the end of 1998, prior to conversion, that might differ across the treatment and control groups and therefore differentially affect either spreads or turnover. Use of these firm-level characteristics also allows us to explore the role of firm-level factors in determining winners and losers. We interpret the difference in the change in spreads or turnover between the treatment and control groups, controlling for firm-specific characteristics, as due to euro conversion. 4.1 Specification and Predictions Did euro conversion prompt exchanges to compete for order flow by reducing spreads? To control for the potentially endogenous relation between spreads and volume, we estimate a system of simultaneous equations using 2SLS to address this question. In the first stage, we estimate changes in percentage spreads and turnover from 1998Q4 to 1999Q4. Harris (1994), among others, suggests using lagged values of the regression variables as instruments. We use percentage change in average daily spread from the third quarter of 1998 to the fourth quarter of 1998 (⁠ ΔSpr98 ⁠) to instrument for changes in spreads over our event period, and percentage change in average daily turnover from the third quarter of 1998 to the fourth quarter of 1998 (⁠ ΔTurn98 ⁠) to instrument for change in turnover.22 ΔSpr^ and ΔTurn^ are the fitted values from the following equations: ΔSpri=α+β1ΔSpr98,i+β2ΔTurn98,i+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+ Industry Dummyi+εi  (1) ΔTurni=α+β1ΔSpr98,i+β2ΔTurn98,i+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+ Industry Dummyi+εi (2) where i denotes a firm-level variable, ΔSpri is the percentage change in firm i’s average daily closing (percentage) spread from 1998Q4 to 1999Q4 and ΔTurni is percentage change in turnover over the same period. In both equations, we include several firm-specific characteristics as of the end of 1998 that might differentially affect spreads or volume. Tick98Q4,i is the average relative tick size during the fourth quarter of 1998; EuroPrice98End,iis the average daily closing price for December 1998 in euros; 23MV98End,i is the average daily market value for December 1998 in billions of euros; Ret98Q4,i is the average daily return in the fourth quarter of 1998; Volat98Q4,i is the volatility of daily returns during the fourth quarter of 1998; and Foreigni is a dummy variable set equal to 1 if the firm’s home country is not the same as the country of the exchange on which it trades, and zero otherwise.24 According to Karpoff (1987), changes in price level and volatility should be positively related to volume. Industry dummies are based on the Industry Classification Benchmark Industry Codes from DataStream, which classify all the firms into 10 industries; see footnote 20 for more details. We also use industry clustered standard errors in both the first and second stage regressions. Using the fitted values ΔSpr^i and ΔTurn^i from Equations (1) and (2) should help control for an endogenous relation between changes in spreads and changes in turnover. To test whether euro conversion is associated with significant shifts in spreads and turnover, we estimate the following second stage equations, using the fitted values from the first stage above. ΔSpri=α+β0Di+β1ΔTurn^i+β2ΔTurn^i2+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi + εi (3) ΔTurni=α+β0Di+β1ΔSpri^+β2ΔSpr^i2+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi + εi  (4) We include the squared terms to allow for non-linearities in the spread-turnover relation. The dummy variable, Di, is set equal to one for the treatment firms and zero for the control firms; it serves as our difference-in-differences estimator for the effect of euro conversion on changes in spreads and volume. In our first test, Di equals one for euro firms and zero for non-euro European firms. In the second test, Di equals one for European firms and zero for NYSE firms. Controlling for other firm-specific factors and endogeneity in the relation between spreads and volume, we expect spreads to decrease and volume to increase more for treatment firms than for control firms. Therefore, the dummy variable should be negative in the spread regressions and positive in the turnover regressions. Competition should result in bigger reductions in spreads and increases in trading activity for the treatment firms relative to the control sample. 4.2 Regression Results The first four columns of Table IV report results of the first stage of our two-stage analysis [Equations (1) and (2) above]. Lagged change in spread (⁠ ΔSpr98,i ⁠) significantly explains change in spread from 1998Q4 to 1999Q4 in columns (1) and (3), and therefore satisfies the inclusion restriction for instruments. It is not significant in the regressions of change in turnover, and therefore does not violate the exclusion restriction. Analogous results hold for our instrument for change in turnover (⁠ ΔTurn98,i ⁠). Table IV. Two stage-least squares regressions of changes in spreads and turnover on firm-specific explanatory variables This table reports results of 2SLS regressions of percentage changes in spreads and turnover from 1998Q4 to 1999Q4 on firm-specific variables. Results from the first stage regressions (defined in Equations (1) and (2) in the text) are in columns (1) through (4). The dependent variables are ΔSpri (or ΔTurni ⁠), percentage change in the firm’s average daily closing percentage spread (or turnover) from 1998Q4 to 1999Q4. Columns (5)–(8) report results for the second stage equations: ΔSpri=α+β0Di+β1ΔTurn̂i+β2ΔTurn̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi (3) ΔTurni=α+β0Di+β1ΔSprî+β2ΔSpr̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi  (4) where ΔSprî and ΔTurnî are the fitted versions from the first stage equations. In columns (5) and (6), the dummy variable, Di, is set equal to one for euro firms and zero for non-euro European firms. In columns (7) and (8), Di equals one for European firms and zero for NYSE firms. Tick98Q4,i is the average relative tick size during the fourth quarter of 1998. EuroPrice98End,i and MV98End,i are the average daily closing price (in euros) and equity market capitalization (in billions of euros) for December 1998. Ret98Q4,i and Volat98Q4,i, are the daily average return and volatility of daily returns for 1998Q4. All continuous variables are winsorized at +/– three standard deviations around the mean. Foreigni equals 1 if a firm’s home country is not equal to its exchange location, otherwise 0. We include industry dummies in all regressions to control for industry effects. We use industry clustered standard errors in both the first and second stage regressions. The table reports coefficient estimates, with t-statistics below. Coefficient estimates that are different from zero at the 5% or 1% levels are in bold type. First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 Table IV. Two stage-least squares regressions of changes in spreads and turnover on firm-specific explanatory variables This table reports results of 2SLS regressions of percentage changes in spreads and turnover from 1998Q4 to 1999Q4 on firm-specific variables. Results from the first stage regressions (defined in Equations (1) and (2) in the text) are in columns (1) through (4). The dependent variables are ΔSpri (or ΔTurni ⁠), percentage change in the firm’s average daily closing percentage spread (or turnover) from 1998Q4 to 1999Q4. Columns (5)–(8) report results for the second stage equations: ΔSpri=α+β0Di+β1ΔTurn̂i+β2ΔTurn̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi (3) ΔTurni=α+β0Di+β1ΔSprî+β2ΔSpr̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi  (4) where ΔSprî and ΔTurnî are the fitted versions from the first stage equations. In columns (5) and (6), the dummy variable, Di, is set equal to one for euro firms and zero for non-euro European firms. In columns (7) and (8), Di equals one for European firms and zero for NYSE firms. Tick98Q4,i is the average relative tick size during the fourth quarter of 1998. EuroPrice98End,i and MV98End,i are the average daily closing price (in euros) and equity market capitalization (in billions of euros) for December 1998. Ret98Q4,i and Volat98Q4,i, are the daily average return and volatility of daily returns for 1998Q4. All continuous variables are winsorized at +/– three standard deviations around the mean. Foreigni equals 1 if a firm’s home country is not equal to its exchange location, otherwise 0. We include industry dummies in all regressions to control for industry effects. We use industry clustered standard errors in both the first and second stage regressions. The table reports coefficient estimates, with t-statistics below. Coefficient estimates that are different from zero at the 5% or 1% levels are in bold type. First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 We report the second stage of the 2SLS regressions in the last four columns of Table IV. Comparing euro to non-euro European firms [columns (5) and (6)], we see that the euro dummy is insignificant in the spread regression and significantly positive in the turnover regression. We interpret this coefficient as indicating that non-euro European exchanges also reduce spreads in an attempt to compete with euro exchanges. The significantly positive coefficient on the euro dummy in column (6) suggests that euro exchanges gain significantly more turnover than the other, non-euro European exchanges. Comparing firms on European exchanges to those on the NYSE [columns (7) and (8)], results show that, as predicted if euro conversion triggered competition, the coefficient on the European dummy is significantly negative in the spread regression and significantly positive in the turnover regression. Controlling for endogeneity in the relation between spreads and volume and for firm-specific factors, euro conversion results in lower spreads and higher turnover for European firms relative to a control sample of firms listed on the NYSE. Cross-sectionally, some models predict that the exchanges with lower transaction costs will take more volume away from the other exchanges (e.g., Admati and Pfleiderer, 1988; Chowdhry and Nanda, 1991; Ramos and von Thadden, 2008). We do not find that the coefficient on (fitted) change in spread significantly explains change in turnover. Aside from the endogenous relation between spreads and trading activity, therefore, we cannot conclude that changes in spreads drive changes in trading activity (or vice versa).25 If we estimate the second stage regressions using actual (as opposed to fitted) values for spreads and turnover, both of these explanatory variables are negative and significant, with their squared terms positive and significant (see Internet Appendix, Table A16). Furthermore, the European dummy in the turnover regression is still positive, but no longer significant. Thus, controlling for endogeneity affects inferences in important ways. From Table IV, we also see that changes in spread and turnover are associated with several firm-specific factors. Comparing European firms to the NYSE in column (7), spreads decrease more for firms with higher prices, smaller firms, firms with larger prior returns, and for firms whose home country differs from the exchange on which they list.26 Turnover increases more for smaller, less volatile firms [column (8)].27 Variations in these firm-specific factors help explain which exchanges win and which lose volume. We argue that euro conversion triggered competition by increasing the transparency of prices. Firm size is a common proxy for transparency (e.g., Halling et al., 2008). To explore the role of transparency more fully, we re-estimate the second stage regressions from columns (7) and (8), including an interaction between the European Dummy and firm size (MV98end).28 Results (reported in the Internet Appendix, Table A17) show that the interaction is significantly negative in the change in turnover regressions while the size coefficient by itself remains negative and significant. The negative relation between firm size and turnover is more pronounced for firms in Europe than for NYSE firms, consistent with increased transparency across Europe. Another dimension along which firms vary is by industry. One possibility is that some industries consistently win volume after euro conversion, while others lose. Winning exchanges may be those with the greatest concentration of firms in preferred industries. To examine the role of industry composition, we conduct an F-test of equality across the industry dummies included in columns (7) and (8). These tests reject equality at the one percent level. The sorted industry dummy coefficient estimates (not tabulated) suggest that spreads fall the most (or, rise the least) for telecom firms, with the next largest decreases observed by high tech, oil and gas, and financial firms. Utilities and consumer services firms have the smallest declines in spreads. For turnover, high tech firms have the biggest increase followed by telecom, and then by consumer goods, consumer services and industrials; utilities and oil and gas firms have the smallest turnover gains. 4.3 Country-level Institutional Factors Several of the intercepts in the second stage regressions in Table IV are statistically significant even after we control for the spread/volume relation, European listing, and firm-specific factors, suggesting that a substantial portion of the change in spreads and turnover is still unexplained. In this section, we explore broader country-level factors that may help explain variation in changes in spreads and turnover. These factors could include macroeconomic conditions such as GDP growth or inflation, institutional factors such as political stability or accounting standards, and exchange-level characteristics such as number of listed firms and trading hours or trading rules.29 We draw representative country factors from the competition models discussed in sub-section 2.1, plus the literatures on cross border mergers and acquisitions, law and finance, multi-market trading, and the transmission of crises and shocks (i.e., contagion).30 Unfortunately, we cannot simply include these variables in our regressions because measures for the country-level characteristics would be co-linear with the European dummy. The regression analysis in Equations (3) and (4) assumes that the relation between changes in spread or turnover and firm characteristics is the same across all exchanges. We modify these regressions and allow the intercepts and other coefficients to vary across eleven European exchanges (we exclude NYSE firms from this analysis). Specifically, in both stages, we break apart the constant term and instead include dummy variables for each European exchange. We also interact each explanatory variable with the exchange dummy coefficients to allow the coefficient on each firm-specific variable to vary across exchanges. We interpret the exchange dummy coefficient estimates in the second stage regression as a measure of the country-level unexplained change in spreads or turnover. Table V provides the correlations between the eleven European exchange-level intercept estimates from the second stage spread and turnover regressions with their respective country-level factors. (See the table legend for the definition and sources of the country-level variables. We report results for additional country-level variables in Internet Appendix, Table A18.) Table V. Correlations between exchange intercepts and country-level institutional factors This table reports correlations between the eleven European exchange dummy intercepts from the second stage spread and turnover regressions and country-level and exchange level factors. FDI Inflows, GDP, Exports, Wages, CPI, and LT Interest Rates are sourced from DataStream; changes are from 1998 to 1999. Political Stability, Regulations, Rule of Law, and Control of Corruption, all at 1998 values, are sourced from the World Bank Governance Indices. Insider Trading Enforcement equals the number of years since the first enforcement of insider trading laws, and is sourced from Bhattacharya and Daouk (2002). Shareholder Rights Index and the LLSV Accounting Standards are from La Porta et al. (1998). Anti-Self-Dealing Index is from Djankov et al. (2008). Institutional Ownership is from Ferreira, Massa, and Matos (2010). The accounting Transparency (CIFAR) and Disclosure indices are from Bushman, Piotroski, and Smith (2004). Change in the Number of Analysts, 1998 to 1999, equals the change in the Number of Analysts, which is the average for the thirty firms with the highest number of EPS analysts for each market in 1998 and 1999 from IBES. Number of Listed Firms, Value Traded and Market Cap/GDP are from the Handbook of International Stock Exchanges, with values for 1998. Average and Minimum Correlations of Returns for each exchange’s stock index with the indices of the other exchanges are from DataStream daily total market indices in 1999. Longer Trading Hours are sourced from a Factiva news search. Market Maker Obligatory, Centralization, Depth, Automatic Execution, and Mutual Ownership are provided by PK Jain and correspond to 2000. Correlations that are significant at the 10% or better level are in bold type and marked with *. Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Table V. Correlations between exchange intercepts and country-level institutional factors This table reports correlations between the eleven European exchange dummy intercepts from the second stage spread and turnover regressions and country-level and exchange level factors. FDI Inflows, GDP, Exports, Wages, CPI, and LT Interest Rates are sourced from DataStream; changes are from 1998 to 1999. Political Stability, Regulations, Rule of Law, and Control of Corruption, all at 1998 values, are sourced from the World Bank Governance Indices. Insider Trading Enforcement equals the number of years since the first enforcement of insider trading laws, and is sourced from Bhattacharya and Daouk (2002). Shareholder Rights Index and the LLSV Accounting Standards are from La Porta et al. (1998). Anti-Self-Dealing Index is from Djankov et al. (2008). Institutional Ownership is from Ferreira, Massa, and Matos (2010). The accounting Transparency (CIFAR) and Disclosure indices are from Bushman, Piotroski, and Smith (2004). Change in the Number of Analysts, 1998 to 1999, equals the change in the Number of Analysts, which is the average for the thirty firms with the highest number of EPS analysts for each market in 1998 and 1999 from IBES. Number of Listed Firms, Value Traded and Market Cap/GDP are from the Handbook of International Stock Exchanges, with values for 1998. Average and Minimum Correlations of Returns for each exchange’s stock index with the indices of the other exchanges are from DataStream daily total market indices in 1999. Longer Trading Hours are sourced from a Factiva news search. Market Maker Obligatory, Centralization, Depth, Automatic Execution, and Mutual Ownership are provided by PK Jain and correspond to 2000. Correlations that are significant at the 10% or better level are in bold type and marked with *. Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Correlations with the exchange-specific intercepts in the spread regressions indicate that, after controlling for all of the firm-specific characteristics, spreads fall the most in exchanges located in the countries with weak transparency as of the beginning of the period (measured with the Number of Analysts, 1998 or Disclosure), an increasing number of analysts over the period, and several exchange-level trading characteristics: fewer listed firms, lower stock index return correlations with the other European markets, no obligatory market makers, no centralization, and automatic execution.31 Exchange intercepts from the turnover regressions are related primarily to changes in GDP and measures of transparency, and not linked to any of the exchange-level variables.32 Turnover rises the most on the exchanges that are in countries with relatively rapidly growing economies (Change GDP), relatively weak transparency at the beginning of the period (measured with La Porta et al’s (1998),LLSV Accounting Standards index, CIFAR and Disclosure),33 and increases in the number of analysts. Frost, Gordon, and Hayes (2006) show, with levels, that an index of market development that includes two measures of trading volume is higher for exchanges located in countries with high transparency. These results show that the gains in turnover in response to competition come primarily to the exchanges in countries with low transparency. To further explore the effects of transparency, we also estimate regressions similar to those in Table IV using dummy variables to split exchanges into groups based on the accounting and transparency variables that are significant for change in turnover in Table V. These variables are LLSV Accounting Standards, Transparency (CIFAR), Change in the Number of Analysts, and Disclosure. We construct dummy variables for NYSE firms and for firms on exchanges in High, Medium, or Low groups based on their exchange-level country transparency factors. We re-estimate results reported in columns (7) and (8) of Table IV, splitting the intercept into these dummies, and also interacting each of these dummies with firm size. Competition due to increased transparency after euro conversion implies that the strongest effects should be present for firms listed on the least transparent exchanges. Results (see Internet Appendix, Table A19) show that for both LLSV and CIFAR, turnover increases more (higher intercepts) and the sensitivity of change in turnover to firm size is greater in magnitude (more negative) for firms in countries that start with low transparency as captured by these two measures.34 Firms listed on exchanges in countries that begin the period with the weakest information environments (as measured by LLSV and CIFAR) experience the biggest increase in turnover and see the greatest sensitivity of change in turnover to firm size. These results support our transparency interpretation. 4.4 Alternative Potential Explanations As noted earlier, there were several other broad trends occurring during our event window that could also affect trading costs and volume and thus confound interpretation of our results. The internet bubble was well under way during our sample period. The increases in trading activity and corresponding decreases in trading costs that we document may reflect changes due to internet-related trading rather than euro adoption. To address this concern, we repeat our 2SLS analysis, excluding firms in the technology industry, and also excluding those in both technology and telecom (two industries affected by the bubble). We also repeat our analysis excluding smaller firms (those in the bottom two NYSE size deciles), which may be more affected by the technology boom. Results from these three robustness tests are similar, and provide the same inferences, as those we report in Table IV (see the Internet Appendix, Table A20). The positive and significant coefficient estimate for the European dummy variable in the turnover regression [column (8) in Table IV] is also consistent with “cross market rebalancing,” where investors respond to a shock by rebalancing their portfolios (Kodres and Pritsker, 2002). It could be that euro conversion spurred investors to increase their holdings of euro stocks in general, resulting in a relatively larger increase in volume for euro exchanges than for the NYSE. The creation of euro-related indices could exacerbate this trend, particularly for large firms, through the formation of index-matching portfolios. However, the best known euro indices were created before euro conversion. For example, Amsterdam’s EuroTop 100 index was created in 1990, and the Dow Jones Euro STOXX indices in February of 1998. Also, as noted above, the euro conversion date was known well in advance, suggesting portfolio rebalancing would not necessarily concentrate right around the actual conversion date. Furthermore, if index-related rebalancing were driving our euro dummy results, we would expect that the dramatic increases in volume would be concentrated in the largest firms. However, the volume effects we observe are greater for smaller firms. This pattern suggests our volume results are not driven by general portfolio rebalancing efforts where investors moved into large, index-linked euro firms. Finally, the years around our sample period involved substantial changes to trading technology and a move toward electronic trading. Results in Table V show that exchanges with automatic execution have greater decreases in spreads. However, Internet Appendix Table A1, Panel B, shows no significant changes to trading technology during 1998 or 1999. Changes in trading technology began well before our sample period. Thus, we rely on our difference-in-differences test to control for this more general trend. 4.5 Discussion: Regression Results Overall, variations in both firm-specific and country-level factors help explain which exchanges win and which lose volume. Turnover increases more for exchanges with relatively smaller, lower volatility firms in countries with higher GDP growth, lower accounting standards, and larger increases in the number of analysts. Exchanges that win volume also have the biggest concentrations of firms in the industries that experience the biggest jumps in volume (such as high tech, telecom, consumer goods, consumer services, and industrials). These effects combine to determine the winning and losing exchanges. To illustrate, in Table VI we report exchange-specific averages for some of the explanatory characteristics that are significant determinants of change in turnover, along with the ranking (from high to low) for each factor. Our winning exchanges include Milan, Frankfurt, Paris, and London. Milan lists smaller firms (ranking 9th in terms of average market capitalization), and has many firms in some of the winning industries (especially telecom with 17.5% of all telecom firms in the sample, and consumer goods and services with 10.7%). Frankfurt benefits from relatively low accounting standards and having few firms in the oil and gas industry, only 2% of total; this exchange wins despite listing the most volatile firms. London wins because it has firms with lower volatility, the second biggest jump in analysts, and a higher concentration of telecom (27.5%) and technology (26.7%) firms. Paris primarily benefits from listing many technology (28.1%) and telecom (12.5%) firms, while its GDP growth and change in analysts are in the upper half of the group. Table VI. Exchange averages for significant firm-specific and institutional variables This table reports the average by exchange and ranking (highest to lowest) for some of the firm-specific and institutional variables that are significant in the 2SLS analysis (reported in Table IV) or in the country-level analysis (in Table V) for our sample. MV98end and Volat98Q4 are averages of firm specific values for our sample. Country-level GDP data are from DataStream, change is from 1998 to 1999. LLSV Accounting Index is from LaPorta et al. (1998). Change in Analysts is the change in the average number of analysts from 1998 to 1999 for the thirty firms with the largest number of EPS analysts, as reported by IBES. Rank #1 is assigned to the exchange with the largest number. Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 Table VI. Exchange averages for significant firm-specific and institutional variables This table reports the average by exchange and ranking (highest to lowest) for some of the firm-specific and institutional variables that are significant in the 2SLS analysis (reported in Table IV) or in the country-level analysis (in Table V) for our sample. MV98end and Volat98Q4 are averages of firm specific values for our sample. Country-level GDP data are from DataStream, change is from 1998 to 1999. LLSV Accounting Index is from LaPorta et al. (1998). Change in Analysts is the change in the average number of analysts from 1998 to 1999 for the thirty firms with the largest number of EPS analysts, as reported by IBES. Rank #1 is assigned to the exchange with the largest number. Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 In contrast, Brussels and Madrid are among the main losers in terms of changes in turnover. Brussels lists larger firms and has the largest decline in the number of analysts; it also suffers because it does not have a significant number of firms in any of the winning industries. Madrid has stronger accounting ratings, a decline in the number of analysts, and also few firms in the winning industries. Overall, these firm-specific and country factors combine to determine winners and losers. 5. Welfare Consequences To estimate whether changes based on competition in response to euro adoption are material, we perform several social welfare calculations. We begin with a calculation of the change in consumer welfare generated by the reduction in bid-ask spreads. This calculation is similar to that in Brown, Mulherin, and Weidenmier (2008). Specifically, we first estimate actual transactions costs paid by investors via effective bid-ask spreads during the fourth quarter of 1999. To do this, we multiply the value (in thousands of euros) of volume traded in our sample firms each day by one-half of the effective spread for that firm that day.35 We sum this cost across all firms and days on a particular exchange to derive an estimate of the total transactions costs paid by investors on that exchange during 1999Q4. We next estimate what transaction costs would have been during 1999Q4 if, hypothetically, effective spreads had remained constant at their 1998Q4 levels. So, for each firm we estimate the average effective spread during 1998Q4. We multiply the value (in thousands of euros) of volume traded each day during the fourth quarter of 1999 by one-half of the average effective spread for that firm during 1998Q4. This figure (summed across all trades for firms on a given exchange during 1999Q4) represents what the total cost to investors trading on that exchange would have been if effective spreads had remained constant at their 1998 levels (please see Panel A of Table VII). Table VII. Welfare consequences This table provides several estimates of the welfare consequences of changes around euro adoption. In Panel A, we estimate the change in welfare associated with change in effective spreads. This panel reports actual transaction costs paid by investors through effective bid-ask spreads during the1999Q4, and hypothetical 1999 transaction costs (multiplying 1999Q4 trading volume by the 1998Q4 effective spread). We also report the difference and percentage change between quarters. In Panel B, we report the difference in effective spreads, Market Cap/GDP and the exchange return from 1998 to 1999 by exchange. Results in Panel B are sorted by change in turnover from Table II. Losers are exchanges with a significant decrease in turnover based on at least one of the test statistics from Table II. Winners are those with a significant increase. We exclude Xetra from the analysis in Panel B. Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Table VII. Welfare consequences This table provides several estimates of the welfare consequences of changes around euro adoption. In Panel A, we estimate the change in welfare associated with change in effective spreads. This panel reports actual transaction costs paid by investors through effective bid-ask spreads during the1999Q4, and hypothetical 1999 transaction costs (multiplying 1999Q4 trading volume by the 1998Q4 effective spread). We also report the difference and percentage change between quarters. In Panel B, we report the difference in effective spreads, Market Cap/GDP and the exchange return from 1998 to 1999 by exchange. Results in Panel B are sorted by change in turnover from Table II. Losers are exchanges with a significant decrease in turnover based on at least one of the test statistics from Table II. Winners are those with a significant increase. We exclude Xetra from the analysis in Panel B. Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Results show that, for the full sample, traders paid a total of €2.748 billion in bid-ask-spread related transaction costs during 1999Q4. However, if effective bid-ask spreads had remained at their 1998 levels, this total would have been €3.316 billion. Investors therefore saved a total of €568 ($570) million (or about 17%) in trading costs because of the decline in effective bid-ask spreads around euro adoption. This value is calculated based on trading volume during one quarter, so annualized savings would be approximately four times higher.36 In general, bid-ask spreads have been decreasing in all markets over the last few decades, so some of this decline may represent a natural time trend.37 (Our difference-in-differences tests in Table III control for this time trend using a matching sample.) We note here that these numbers are economically large. We also provide two other estimates of welfare implications. To get a broader estimate of the impact of euro adoption, we calculate a time series of the ratio of equity market capitalization to GDP. Market Cap/GDP is a common measure of market depth, or of the development of the local capital markets.38 We also compare the returns to investors in these markets. We calculate the exchange return for each of our exchanges with exchange-level indices reported by the World Stock Exchange Handbook from year end 1998 to year end 1999. We report these results in Table VII, Panel B (along with the change in effective spreads from Panel A). In this panel, we sort exchanges into groups according to their change in turnover from Table II; those exchanges with a significant decrease in turnover are classified as losers and those with a significant increase are the winners. Results in Panel B show significant differences in all three measures between the losing and winning exchanges. The average losing exchange has savings due to effective spreads of only €2 million, a very small increase in Market Cap/GDP, and an average exchange return of only 2.6%. In contrast, the winning exchanges save an average of €76 million in transaction costs, have a 45% increase in Market Cap/GDP, and a 31% average exchange return. Other general trends affecting all firms or exchanges during our sample period would not explain these differential results we find across winning and losing exchanges. These amounts are economically large, and suggest to us that the effects we document have broad consequences. The winning exchanges win along several different dimensions. 6. More Evidence of Competition So far, we have presented evidence that is strongly consistent with exchange competition for trading volume. Competition among Europe’s exchanges during our event period is also apparent in the alternative competitive levers that exchanges pull and in outcomes other than trading volume. 6.1 Alternative Levers As noted in Section 2, exchanges have multiple levers to pull when they compete with each other. (We summarize major activities in these areas in the Internet Appendix, Table A1, Panel C.) One obvious lever is trading fees (also referred to as commissions or stamp taxes). Trading fees are an exchange decision, whereas bid-ask spreads are based on equilibrium outcomes of market participants. The World Stock Exchange Fact Book lists commissions and fees for each exchange, but shows no changes in these fees at any of our exchanges from 1998 to 2000. A Factiva news search indicates only two changes: Switzerland imposed a new tax on “remote members” so that they were treated the same as domestic brokers (Boland, 1998); and Denmark delayed dropping the share transaction tax from the beginning of 1999 to October 1 1999 in order to “generate state revenue” (Reuters News, November 25 1998).39 So, in the two cases with changes, the fees went up or a decrease was delayed. Competition does not appear to have played out in transaction fees. An alternative means to attract trading volume is expanded trading hours (e.g., Garvey, 1944). Five of the twelve exchanges did expand trading hours, albeit toward the end of the period (London, Milan, and Paris in September and Brussels and Madrid in October 1999). Two of these exchanges, Brussels and Madrid, were the univariate volume losers, suggesting that expanded trading hours may have been a response to lower trading volume. Di Noia (2001) and Arnold et al. (1999) argue that mergers and alliances are a response of exchanges to competition. Stockholm and Copenhagen announced an agreement to create a common Nordic securities market called Norex in early 1998. Frankfurt and London announced a “strategic alliance” to create a common exchange in July of 1998. During 1999, this bilateral alliance fell apart and talks shifted to a group of eight major exchanges (Amsterdam, Brussels, Frankfurt, London, Madrid, Milan, Paris, and Swiss). During 2000, these eight exchanges signed a memo of understanding to form a partnership, but talks broke down 4 months later. The first successful full merger of European exchanges, Euronext, took place in 2000 among Amsterdam, Paris, and Brussels. No doubt, all of these other levers take time to pull. Our evidence of big shifts in transactions costs and trading activity in the year following euro conversion may be viewed as the immediate competitive response, while the exchanges were simultaneously maneuvering on multiple other, longer-term fronts. 6.2 Other Competitive Outcomes In non-tabulated results, we find that the variation in spreads across firms on the twelve European exchanges falls significantly over our sample period.40 This convergence in spreads is similar to the convergence in other European financial prices found around this time.41 Since, as discussed in sub-section 2.1, no specific EU integration actions were taken during our sample period, it is hard to conclude that the convergence we find here is due to the general integration process associated with preparations for euro conversion. Instead, we interpret the convergence of equity trading costs around euro introduction as consistent with European market makers matching their competitors’ cost reductions. Lower average spreads, as well as a decrease in variation across spreads, are consistent with competition across all of the European exchanges. Previous literature discusses whether competition leads to more or less consolidation of trading across markets (e.g., Stigler, 1961; Pagano, 1989; Chowdhry and Nanda, 1991; Madhavan, 1995; Gehrig, 1998). To examine this question more fully, in Figure 3 we report some measures of market concentration among European exchanges. Panel A reports the Herfindahl–Hirshman Index (HHI), and Panel B reports the market share of the largest three exchanges based on value traded in US dollars, from the World Stock Exchange Fact Book. Results in Panel A show that the HHI trended downward over this period, with a slight interruption in 1999 immediately after euro conversion, especially for the sample including all European exchanges. Similar results hold in Panel B. Trading in Europe is fairly concentrated, with the top three exchanges holding 77% of the market. These results provide evidence that euro conversion did not contribute to a “winner-take-all” outcome (e.g., as in Chowdry and Nanda, 1991). Figure 3. View largeDownload slide Market share of Europe’s largest exchanges. Panel A. Industry concentration (Herfindahl–Hirshman Index, or HHI) based on the dollar value of shares traded for the eleven European exchanges in our sample. Panel B. Market share of Top 3 Euro and European Exchanges (based on value of shares traded) Source:World Stock Exchange Fact Book. Data covers twelve exchanges included in our sample; Frankfurt and Xetra are combined. Results are computed for all European exchanges, and separately for the euro exchanges. In Panel A, the Herfindahl–Hirshman Index is calculated as the sum of the squared values of the market share of each exchange. In Panel B, the three largest exchanges for volume of shares traded are London, Milan, and Frankfurt/Xetra. The three largest for value of shares traded are London, Frankfurt/Xetra, and Amsterdam. Figure 3. View largeDownload slide Market share of Europe’s largest exchanges. Panel A. Industry concentration (Herfindahl–Hirshman Index, or HHI) based on the dollar value of shares traded for the eleven European exchanges in our sample. Panel B. Market share of Top 3 Euro and European Exchanges (based on value of shares traded) Source:World Stock Exchange Fact Book. Data covers twelve exchanges included in our sample; Frankfurt and Xetra are combined. Results are computed for all European exchanges, and separately for the euro exchanges. In Panel A, the Herfindahl–Hirshman Index is calculated as the sum of the squared values of the market share of each exchange. In Panel B, the three largest exchanges for volume of shares traded are London, Milan, and Frankfurt/Xetra. The three largest for value of shares traded are London, Frankfurt/Xetra, and Amsterdam. Exchanges also compete for new firm listings (Chemmanur and Fulghieri, 2006). To examine changes in firm listings concurrent with euro conversion, we collect data on total firm listings as of the end of 1998 and 1999 for our exchanges from the World Stock Exchange Fact Book. Interestingly, rankings for changes in listings do not line up consistently with changes in trading volume (see Internet Appendix, Figure A1). The different set of winning and losing exchanges suggests that new listings are driven by factors distinct from those that affect trading activity. Nevertheless, these results also show large shifts in the number of firm listings after euro conversion. 7. Conclusions Euro conversion provides a natural experiment in which to examine the determinants of volume and spread changes when exchanges compete. We find first, that euro conversion lead to a significant round of cost cuts and volume shifts among the European exchanges. Second, shifts in volume were driven by an array of factors, including firm-specific characteristics (size, volatility, and industry), and country-level measures such as GDP growth, accounting standards, and change in the number of analysts. Several of these factors are linked to transparency, with turnover increasing more for less transparent firms. These results suggest that euro conversion triggered competition by changing price transparency. Changes in spreads are driven by these factors, plus exchange-level characteristics such as number of listed firms and the presence of automatic execution. Third, the competition we document spilled over to non-euro European exchanges. In sum, we show that even a seemingly trivial change—the redenomination of prices—can trigger a significant round of price cutting and volume shifts among exchanges. The large absolute and relative volume shifts among exchanges that we document confirm the predictions of many models that the removal of investment barriers can create new opportunities and threats for exchanges in their competition for order flow. Footnotes * We thank Hank Bessembinder, Ran Duchin, Jarrad Harford, Alan Hess, Pankaj Jain, Walter Novaes, Sofia Ramos, Ed Rice, Stephan Siegel, Ernst-Ludwig von Thadden, Thorsten Beck (the editor), the referees, and seminar participants at Boston College and at the University of Washington Finance brown bag seminar for helpful comments, the UW Center for International Business Education and Research for financial support, and Jon Kalodimos for valuable research assistance. Dewenter thanks the Joshua Green Family Professorship. Koski thanks the Kirby L. Cramer Endowed Chair in Finance. 1 Consistent with models that assume exchanges compete to attract trading volume (e.g., Ramos and von Thadden, 2008; Santos and Scheinkman, 2001), we define winners (losers) as the exchanges that win (lose) volume. 2 Arnold et al. (1999) provide evidence on spillovers in competition within a country, the USA. 3 Two exceptions are Arnold et al. (1999), who analyze competition across the NYSE, Amex, and multiple regional exchanges, and Garvey (1944), who looks at a series of exchanges created to compete with the NYSE during its early years. 4 Both the contagion and herding literatures provide guidance as to how competition spurred by euro conversion could spill over to nearby exchanges. The contagion literature primarily focuses on how crises move from one country to another (see Dornbusch, Park, and Claessens (2000) for a review of the contagion literature). Positive economic effects from euro conversion on firms in the euro countries could spill over to the other European countries through trade or financial links. The herding literature, see Devenow and Welch (1996) for a summary, suggests that just as a loss of confidence could lead to investment outflows across multiple countries, a jump in confidence from euro conversion might positively affect investment into all European exchanges. 5 NYSE firms also match the distribution of our European firms more closely based on firm size and industry composition (see Table A2 in the Internet Appendix). 6 A firm listed on a specific exchange is a “firm list.” One sample firm could be listed on two exchanges, or on both Frankfurt and Xetra. In both cases, we count it as two firm lists. We note that for several exchanges the number of firm lists in our initial dataset in column 3 of Table I is larger than the total number of listed stocks on the exchange from column 1. This difference is likely due to timing. The total number of listed stocks on the exchange is as of December 31 1998. The initial dataset includes any firm with at least one ask observation between October 1 1998 and December 31 1999. New listings during 1999 will not be included in the total number of listed stocks as of December 31 1998, but will survive the initial data screen. These firms will not, however, be included in our final sample, which requires at least fifteen non-missing observations for each of our sample quarters. We also note that for Lisbon and Paris, the subset market capitalization value in column 6 is larger than the total market capitalization value in column 2. We are not able to reconcile this discrepancy and note that the two sets of data are from different sources. 7 Only fifty-six firms on Frankfurt pass our data screens using DataStream. This compares to 183 Frankfurt firms in our final sample using KIT data. Results corresponding to Table II, using DataStream data for Frankfurt and Xetra are available in Table A3 of the Internet Appendix. 8 For example, Boomfield and O’Hara (2000) conduct lab experiments of their game theory model of competition between transparent and non-transparent market makers, where transparency refers to the speed with which trade information is revealed. In their lab experiments, quotes and dealer-selected levels of transparency evolve over multiple rounds as participants learn and respond to competitor moves. 9 See Corhay, Hawawini, and Michel (1987) and Das and Rao (2012) among others for studies of seasonalities in European equity markets. 10 About a third of the European firms in our sample fall in the smallest two size deciles based on NYSE cutoffs. Therefore, although we include a screen that eliminates the very smallest European firms (those with market capitalizations less than 50 million euros), our sample still includes many firms that are small relative to NYSE size deciles (see Panel A of Table A2 in the Internet Appendix). 11 In 2011, Frankfurt eliminated floor trading, so now all trading for the Frankfurt Stock Exchange takes place on Xetra. For additional information, see http://deutsche-boerse.com/dbg/dispatch/en/kir/dbg_nav/about_us/10_Deutsche_Boerse_Group/50_Company_History?horizontal=page5_DB_History_2010-heute_ . 12 For firms that are still listed, DataStream provides complete cross-listing information. However, DataStream only provides cross-listing information for about 10% of firms that are no longer listed. See Table A4 in the Internet Appendix for more details about the number and trading volume of cross-listed firms in our sample. 13 For robustness, we also estimate the Amihud (2002) illiquidity measure. Inferences based on the Amihud measure are essentially identical to those based on percentage and effective spreads, and are available in Internet AppendixTable A5. 14 We report results corresponding to Table II, equally weighting each firm, in Internet Appendix, Table A6. z-statistics for a nonparametric sign test are calculated as (Number of Firm Lists with Positive Changes – 0.5 × N)/ N/4 ⁠, where N stands for the total number of firm lists. For robustness, the Internet AppendixTable A7 also provides comparable analyses to those in Table II for short-run changes from 1998Q4 to 1999Q1. 15 Despite the fact that the value-weighted mean for the Overall (Europe) sample in Panel B declines from 0.017 to 0.014, the mean percentage change is (insignificantly) positive. The mean change is driven by a small number of extremely large positive changes. The median change in effective spreads is –17%. In our multivariate analysis, we winsorize these variables to mitigate the effect of outliers. 16 Our sample selection criteria require a firm to have at least 15 non-missing trading volume and quote observations during both 1998Q4 and 1999Q4. The sample of firms we analyze is constant across periods, so the changes in volume we document in our analyses are not attributable to changes in the number of firms due to mergers and acquisitions, delistings, or new listings. Untabulated t-statistics on the two (unbalanced) samples show that differences in all of our spread and trading activity measures differ significantly between the Overall(Europe) and the NYSE. 17 We do not have data for 1997, so we compare changes for the three quarters prior to our event (from 1998Q1 through 1998Q4) with changes over a comparable length of time during our event window (from 1998Q4 through 1999Q3). Please see the Internet AppendixTable A8. 18 The following EU countries adopted the euro after our event date: Greece (2001), Slovenia (2007), Cyprus (2008), Malta (2008), Slovakia (2009), Estonia (2011), Latvia (2014), and Lithuania (2015). Of these, only Greece has a nontrivial number of firms with adequate coverage on DataStream to pass our data screens. There were also several non-EU adopters, but there was no coverage on DataStream for companies in these other countries. 19 We also examine long-run changes, from 1998Q4 through 2004Q4, for the subset of firms with trading in both periods. Please see Internet AppendixTable A11. Inferences are similar. 20 For the European sample, we collect the Industry Classification Benchmark (ICB) Industry Codes from DataStream, which classify all the firms into ten industries. The ICB is a company classification system developed by Dow Jones and FTSE, and is used globally. For the NYSE sample, we collect the 4-digit Standard Industrial Classification (SIC) code from CRSP. We first classify companies into one of the forty-eight industries using the Fama-French forty-eight industry classification by SIC code. Then we further group the forty-eight industries into the ten ICB industries. 21 Relative tick size for a single firm is defined as the minimum tick over the quarter divided by the average transaction price during the quarter. To control for changes in price levels over time, we also divide the minimum tick during 1999Q4 by the average transaction price during 1998Q4 (adjusted for currency differences). See Internet Appendix, Table A12. We control for tick size effects in the multivariate analysis in the next section. 22 We considered using the percentage change from 1998Q1 to 1998Q4 as an instrument, but these instruments do not satisfy the inclusion and exclusion restrictions (see Internet Appendix, Table A13). Nevertheless, the European dummies in the second stage of this specification are similar to those we report in the paper. We choose to report results based on change in turnover to control for changes in the number of shares outstanding. Results are substantially similar when we use change in volume or change in effective spreads as our dependent variables (see Internet AppendixTable A14), although the European dummy is not significant in the second stage when we use change in volume as the dependent variable. 23 Dollar values for the NYSE firms are converted to euros using the exchange rate on January 4 1999. 24 We use DataStream code GGISN, which identifies the firm’s country of incorporation, to classify firms as Foreign. About 12% of the sample firms have Foreign = 1. This subset includes firms headquartered outside of the EU, for example US firms, and EU firms listed outside their home market. For robustness, we repeat the univariate analyses from Tables II and III for the subset of firms with Foreign = 1, and inferences are substantially similar. We also re-estimate our 2SLS regressions interacting the Foreign dummy with all other explanatory variables in the second stage regression. We find that all inferences hold. See Table A15 of the Internet Appendix for descriptive statistics and results relating to Foreign firms. 25 We note that the coefficient on the squared value of fitted change in turnover is significant in column (7), with change in spread as the dependent variable. This result suggests that change in turnover has a weak incremental effect where at higher increases in turnover, the drop in spreads is dampened. 26 These results are consistent with Eleswarapu and Venkataraman (2006) who find bigger spreads for larger firms and lower levels of spreads for firms with higher prices. 27 Halling et al. (2008), in a cross listing paper, show that the share of foreign/domestic trading is higher for smaller firms, firms with more volatility and high tech firms. On the other hand, they also show that trading on the domestic market is positively related to firm size and not related to volatility or being in the high tech industry. Kwan, Masulis, R., and McInish (2015) show that trading market share across exchanges is negatively related changes in volatility. 28 Although volatility is another potential proxy for transparency, it may also proxy for other factors such as risk. Therefore, we do not consider it a clean proxy for asymmetric information. 29 Eleswarapu and Venkataraman (2006) show that the level of spreads is related to macro-level institutions, while Lo (2013) shows that a ranking of the world’s forty-five largest exchanges in terms of “trading competition,” including turnover, is strongly affected by whether or not country-level factors are included in the analysis. 30 See Rossi and Volpin (2004), Erel, Liao, and Weisbach (2012), and Ferreira, Massa, and Matos (2010) for recent papers on cross border mergers and acquisitions; see La Porta et al. (1998) and Djankov et al. (2008) for law and finance papers; see Halling et al. (2008), Halling, Moulton, and Panayides (2013), and Baruch Karolyi, and Lemmon (2007) for multi-market trading papers; see Kaminsky and Reinhart (2000), Karolyi and Stulz (1996), and King and Wadhwani (1990) for representative contagion papers. 31 The transparency results are consistent with Eleswarapu and Venkataraman (2006). The market correlation results are inconsistent with Ramos and von Thadden’s (2008) model of exchange competition which predicts a negative relation. The automatic execution result is consistent with Jain (2006), but the market maker and centralization results are inconsistent. Our results may vary from Jain as he examines an even broader range of countries. The lack of significance for insider trading is consistent with Eleswarapu and Venkataraman (2006), but inconsistent with Halling et al. (2008) and Jain (2005, 2006). One explanation why we have no significance for insider trading enforcement is that all of our exchanges except for one had enforced insider trading by 1998. 32 The lack of any significant effects for exchange characteristics for turnover is inconsistent with Jain (2006, effects for centralization) and Jain (2005, for electronic trading). Our results may vary from Jain because of the different set of exchanges examined. His results show that the level of market development affects the level of turnover. Prior competition papers have not related volume changes to the local economy’s growth. 33 Both CIFAR and Disclosure are from Bushman, Piotroski, and Smith (2004). CIFAR equals the average number of ninety accounting and non-accounting items disclosed by a sample of large companies. Disclosure is based on the prevalence of disclosures for research and development expenses, capital expenditures, product and geographic segment data, subsidiary information and accounting methods. 34 As noted, we also estimate these regressions for Change in Analysts and Disclosure. Change in Analysts results are mixed (see Internet AppendixTable A19), suggesting it is the starting level of transparency rather than the change in transparency that determines winners and losers. Most of the countries in our sample have very high levels of Disclosure relative to the full cross-section presented in Bushman, Piotroski, and Smith (2004), so this variable does not seem to capture economically meaningful variation in transparency for our sample. 35 We multiply by one-half, because the effective spread is a measure of round-trip transaction costs, and each individual trade is only one way. 36 Corresponding savings for one quarter based on percentage spreads (rather than effective spreads) are €1.01 billion. These values represent the savings to investors who traded in the firms in our sample. To obtain a more general estimate applicable to all firms on these exchanges, we repeat this exercise, applying 1998 and 1999 effective spreads to the total 1999 volume on each exchange (in dollars, obtained from the World Stock Exchange Fact Book, 2004). Based on this calculation, investors saved over $5 billion in trading costs associated with the reduction in effective spreads. 37 In Internet Appendix, Table A21 we report the welfare savings in trading costs resulting from changes in effective spreads for European firms controlling for changes in effective spreads in matched NYSE firms over the same time period. These savings are even larger, at €743 (or 23.5%). This result is not surprising given our univariate results reported in Tables II and III. Spreads generally increased or decreased less on the NYSE than they did in Europe after euro adoption. 38 For example, this ratio is included in the World Bank’s Global Financial Development dataset. Market capitalization comes from the World Stock Exchange Fact Book over this period, and is estimated for domestic companies only. GDP data comes from the Organization for Economic Co-operation and Development (OECD). 39 There was also some discussion of removing the stamp duty on share trades in London. 40 For each quarter, we pool all firms and calculate the standard deviation of percentage and effective spreads. Across all firms, the standard deviation of percentage spreads falls from 0.028 in 1998Q4 to 0.023 in 1999Q4. The comparable numbers for effective spreads are 0.020 and 0.016. These changes are highly statistically significant. 41 See Adjaoute and Danthine (2004, equity premiums), Baele et al. (2004, government bond yield spreads), and Bekaert et al. (2013, average bilateral earnings yield differentials) who all argue that the economic and financial integration brought about by the European Monetary Union should be associated with a convergence in prices. None of the trends in their papers show a distinct change around the euro conversion event in January 1999. Indeed, Bekaert et al. test for this and find an insignificant effect of euro introduction on value differentials. References Adjaoute K. , Danthine J.-P. ( 2004 ) Equity returns and integration: is Europe changing?, Oxford Review of Economic Policy 20 , 555 – 570 . Google Scholar Crossref Search ADS Admati A. , Pfleiderer P. ( 1988 ) A theory of intraday patterns: volume and price variability , Review of Financial Studies 1 , 3 – 40 . Google Scholar Crossref Search ADS Aggarwal R. , Angel J. ( 1999 ) The rise and fall of the Amex Emerging Company Marketplace , Journal of Financial Economics 52 , 257 – 289 . Google Scholar Crossref Search ADS Amihud Y. ( 2002 ) Illiquidity and stock returns: cross-section and time-series effects , Journal of Financial Markets 5 , 31 – 56 . Google Scholar Crossref Search ADS Amira K. , Muzere M. ( 2011 ) Competition among stock exchanges for equity , Journal of Banking and Finance 35 , 2355 – 2373 . Google Scholar Crossref Search ADS Arnold T. , Hersch P. , Mulherin J. H. , Netter J. ( 1999 ) Merging markets , Journal of Finance 54 , 1083 – 1107 . Google Scholar Crossref Search ADS Baele L. , Ferrando A. , Hordahl P. , Krylova E. , Monnet C. ( 2004 ) Measuring European financial integration , Oxford Review of Economic Policy 20 , 509 – 530 . Google Scholar Crossref Search ADS Baruch S. , Karolyi G. A. , Lemmon M. ( 2007 ) Multimarket trading and liquidity: theory and evidence , Journal of Finance 62 , 2169 – 2200 . Google Scholar Crossref Search ADS Battalio R. , Greene J. , Jennings R. ( 1997 ) Do competing specialists and preferencing dealers affect market quality? , Review of Financial Studies 10 , 969 – 993 . Google Scholar Crossref Search ADS Bekaert G. , Harvey C. , Lundblad C. , Siegel S. ( 2013 ) The European Union, the euro, and equity market integration , Journal of Financial Economics 109 , 583 – 603 . Google Scholar Crossref Search ADS Bessembinder H. ( 2003 ) Trade execution costs and market quality after decimalization , Journal of Financial and Quantitative Analysis 38 , 747 – 777 . Google Scholar Crossref Search ADS Bhattacharya U. , Daouk H. ( 2002 ) The world price of insider trading , Journal of Finance 57 , 75 – 108 . Google Scholar Crossref Search ADS Bloomfield R. , O’Hara M. ( 2000 ) Can transparent markets survive? , Journal of Financial Economics 55 , 425 – 459 . Google Scholar Crossref Search ADS Blume M. , Goldstein M. ( 1997 ) Quotes, order flow and price discovery , Journal of Finance 52 , 221 – 244 . Google Scholar Crossref Search ADS Boland V. ( December 18 1998 ) Three more countries to join Euro NM, Financial Times. Brown W. , Mulherin J. H. , Weidenmier M. ( 2008 ) Competing with the New York Stock Exchange , Quarterly Journal of Economics 123 , 1679 – 1719 . Google Scholar Crossref Search ADS Bushman R. , Piotroski J. , Smith A. ( 2004 ) What determines corporate transparency? , Journal of Accounting Research 42 , 207 – 252 . Google Scholar Crossref Search ADS Chemmanur T. , Fulghieri P. ( 2006 ) Competition and cooperation among exchanges: a theory of cross-listing and endogenous listing standards , Journal of Financial Economics 82 , 455 – 489 . Google Scholar Crossref Search ADS Chowdhry B. , Nanda V. ( 1991 ) Multimarket trading and market liquidity , Review of Financial Studies 4 , 483 – 511 . Google Scholar Crossref Search ADS Corhay A. , Hawawini G. , Michel P. ( 1987 ) Seasonality in the risk-return relationship: some international evidence , Journal of Finance 42 , 49 – 68 . Google Scholar Crossref Search ADS Das P. , Rao S. P. ( 2012 ) Is the value effect seasonal? Evidence from global equity markets , International Journal of Business and Finance Research 6 , 21 – 33 . de Fontnouvelle P. , Fishe R. , Harris J. ( 2003 ) The behavior of bid-ask spreads and volume in options markets during the competition for listings in 1999 , Journal of Finance 58 , 2437 – 2453 . Google Scholar Crossref Search ADS Devenow A. , Welch I. ( 1996 ) Rational herding in financial economics, European Economic Review 40 , 603 – 615 . Google Scholar Crossref Search ADS Dewenter K. , Kim C.-S. , Novaes W. ( 2010 ) Anatomy of a regulatory race to the top: changes in delisting rules at Korea’s two stock exchanges, 1999–2002 , Journal of Corporate Finance 16 , 456 – 468 . Google Scholar Crossref Search ADS Di Noia C. ( 2001 ) Competition and integration among stock exchanges in Europe: network effects, implicit mergers and remote access , European Financial Management 7 , 39 – 72 . Google Scholar Crossref Search ADS Djankov S. , La Porta R. , Lopez-de-Silanes F. , Shleifer A. ( 2008 ) The law and economics of self-dealing , Journal of Financial Economics 88 , 430 – 762 . Google Scholar Crossref Search ADS Domowitz I. , Glen J. , Madhavan A. ( 1998 ) International cross-listing and order flow migration: evidence from an emerging market , Journal of Finance 53 , 2001 – 2027 . Google Scholar Crossref Search ADS Domowitz I. , Glen J. , Madhavan A. ( 2001 ) Liquidity, volatility and equity trading costs across countries and over time , International Finance 4 , 221 – 255 . Google Scholar Crossref Search ADS Dornbusch R. , Park Y. C. , Claessens S. ( 2000 ) Contagion: understanding how it spreads , The World Bank Research Observer 15 , 177 – 197 . Google Scholar Crossref Search ADS Eleswarapu V. , Venkataraman K. ( 2006 ) The impact of legal and political institutions on equity trading costs: a cross-country analysis , Review of Financial Studies 19 , 1081 – 1111 . Google Scholar Crossref Search ADS Erel I. , Liao R. , Weisbach M. ( 2012 ) Determinants of cross-border mergers and acquisitions , Journal of Finance 67 , 1045 – 1082 . Google Scholar Crossref Search ADS Ferreira M. , Massa M. , Matos P. ( 2010 ) Shareholders at the gate? Institutional investors and cross-border mergers and acquisitions , Review of Financial Studies 23 , 601 – 644 . Google Scholar Crossref Search ADS Foerster S. , Karolyi G. A. ( 1998 ) Multimarket trading and liquidity: a transaction data analysis of Canada–US interlistings , Journal of International Financial Markets 8 , 393 – 412 . Google Scholar Crossref Search ADS Foucault T. , Parlour C. ( 2004 ) Competition for listing , Rand Journal of Economics 35 , 329 – 355 . Google Scholar Crossref Search ADS Frost C. A. , Gordon E. , Hayes A. ( 2006 ) Stock exchange disclosure and market development: an analysis of 50 international exchanges , Journal of Accounting Research 44 , 437 – 483 . Google Scholar Crossref Search ADS Garvey G. ( 1944 ) Rivals and interlopers in the history of the New York Security Market , Journal of Political Economy 52 , 128 – 143 . Google Scholar Crossref Search ADS Gehrig T. ( 1998 ) Competing markets , European Economic Review 42 , 277 – 310 . Google Scholar Crossref Search ADS Glosten L. , Milgrom P. ( 1985 ) Bid, ask and transaction prices in a specialist market with heterogeneously informed traders , Journal of Financial Economics 14 , 71 – 100 . Google Scholar Crossref Search ADS Halling M. , Moulton P. , Panayides M. ( 2013 ) Volume dynamics and multimarket trading , Journal of Financial and Quantitative Analysis 48 , 489 – 518 . Google Scholar Crossref Search ADS Halling M. , Pagano M. , Randl O. , Zechner J. ( 2008 ) Where is the market? Evidence from cross-listings in the United States , Review of Financial Studies 21 , 725 – 761 . Google Scholar Crossref Search ADS Harris L. ( 1994 ) Minimum price variations, discrete bid-ask spreads, and quotation sizes , Review of Financial Studies 7 , 149 – 178 . Google Scholar Crossref Search ADS Huddart S. , Hughes J. , Brunnermeier M. ( 1999 ) Disclosure requirements and stock exchange listing choice in an international context , Journal of Accounting and Economics 26 , 237 – 269 . Google Scholar Crossref Search ADS Jain P. ( 2005 ) Financial market design and the equity premium: electronic versus floor trading , Journal of Finance 60 , 2955 – 2985 . Google Scholar Crossref Search ADS Jain P. ( 2006 ) Improving liquidity through efficient stock market structure and operational design , Journal of Financial Transformation 18 , 151 – 159 . Kaminsky G. , Reinhart C. ( 2000 ) On crises, contagion, and confusion , Journal of International Economics 51 , 145 – 168 . Google Scholar Crossref Search ADS Karolyi G. A. , Stulz R. ( 1996 ) Why do markets move together? An investigation of US–Japan stock return comovements , Journal of Finance 51 , 951 – 986 . Google Scholar Crossref Search ADS Karpoff J. ( 1987 ) The relation between price changes and trading volume: a survey , Journal of Financial and Quantitative Analysis 22 , 109 – 126 . Google Scholar Crossref Search ADS King M. , Wadhwani S. ( 1990 ) Transmission of volatility between stock markets , Review of Financial Studies 3 , 5 – 33 . Google Scholar Crossref Search ADS Klock M. , McCormick T. ( 1999 ) The impact of market maker competition on Nasdaq spreads , Financial Review 34 , 55 – 74 . Google Scholar Crossref Search ADS Kodres L. , Pritsker M. ( 2002 ) A rational expectations model of financial contagion , Journal of Finance 57 , 768 – 799 . Google Scholar Crossref Search ADS Kwan A. , Masulis R. , McInish T. ( 2015 ) Trading rules, competition for order flow and market fragmentation , Journal of Financial Economics 115 , 330 – 348 . Google Scholar Crossref Search ADS La Porta R. , Lopez-de-Silanes F. , Shleifer A. , Vishny R. ( 1998 ) Law and finance , Journal of Political Economy 106 , 1113 – 1155 . Google Scholar Crossref Search ADS Lamoureux C. , Schnitzlein C. ( 1997 ) When it’s not the only game in town: the effect of bilateral search on the quality of a dealer market , Journal of Finance 52 , 683 – 712 . Google Scholar Crossref Search ADS Lo S.-F. ( 2013 ) Which stock exchanges are more attractive? The competition analysis of listing and trading performance , Economic Modelling 30 , 501 – 509 . Google Scholar Crossref Search ADS Madhavan A. ( 1995 ) Consolidation, fragmentation, and the disclosure of trading information , Review of Financial Studies 8 , 579 – 603 . Google Scholar Crossref Search ADS Pagano M. ( 1989 ) Trading volume and asset liquidity , The Quarterly Journal of Economics 104 , 255 – 274 . Google Scholar Crossref Search ADS Pagano M. , Roell A. ( 1990 ) Trading systems in European stock exchanges: current performance and policy options , Economic Policy 5 , 63 – 115 . Google Scholar Crossref Search ADS Pirrong C. ( 2000 ) A theory of financial exchange organization , Journal of Law and Economics 43 , 437 – 472 . Google Scholar Crossref Search ADS Ramos S. ( 2003 ) Competition between stock exchanges: a survey, International Center for Financial Asset Management and Engineering Research Paper No. 77. Ramos S. , von Thadden E.-L. ( 2008 ) Stock exchange competition in a simple model of capital market equilibrium , Journal of Financial Markets 11 , 284 – 307 . Google Scholar Crossref Search ADS Reuters News ( November 25 1998 ) Danish share trade tax to remain until October 1 1999. Rossi S. , Volpin P. ( 2004 ) Cross-country determinants of mergers and acquisitions , Journal of Financial Economics 74 , 277 – 304 . Google Scholar Crossref Search ADS Santos T. , Scheinkman J. ( 2001 ) Competition among exchanges , Quarterly Journal of Economics 116 , 1027 – 1061 . Google Scholar Crossref Search ADS Spankowski U. , Wagener M. , Burghof H.-P. ( 2012 ) The role of traditional exchanges in fragmented markets, unpublished working paper, University of Hohenheim, Stuttgart Stock Exchange. Stigler G. ( 1961 ) The economics of information , Journal of Political Economy 69 , 213 – 225 . Google Scholar Crossref Search ADS Wahal S. ( 1997 ) Entry, exit, market makers, and the bid-ask spread , Review of Financial Studies 10 , 871 – 901 . Google Scholar Crossref Search ADS World Stock Exchange Fact Book, Meridian Securities Markets, Morris Plains NV, Electronic Commerce Inc. Various Issues. © The Authors 2017. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Review of Finance Oxford University Press

Who Wins When Exchanges Compete? Evidence from Competition after Euro Conversion

Loading next page...
 
/lp/ou_press/who-wins-when-exchanges-compete-evidence-from-competition-after-euro-u4sNcxBUo0
Publisher
Oxford University Press
Copyright
© The Authors 2017. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
1572-3097
eISSN
1573-692X
D.O.I.
10.1093/rof/rfx029
Publisher site
See Article on Publisher Site

Abstract

Abstract Using euro conversion as the trigger, we examine what drives volume and spread changes when stock exchanges compete. Results show average trading costs on European exchanges decrease almost 9%, and turnover increases over 30%. Trading costs decline or remain unchanged on all exchanges, but volume deteriorates in some markets and improves in others. Frankfurt, Paris, London, and Milan are winners, while Madrid and Brussels lose volume. We examine the role of the spread-volume relation, firm characteristics, exchange trading rules, and country-level factors in determining these outcomes. Results suggest that euro conversion prompted competition by increasing transparency in market prices. 1. Introduction When European equity markets closed on December 30 1998, all stocks were priced in local currency, for example French francs in Paris, Deutsche marks in Frankfurt, and Italian lira in Milan. When markets reopened five days later on January 4 1999, stocks in eleven of the markets were priced in euros, allowing investors to compare the prices of stocks across all eleven markets in the same currency. Domowitz, Glen, and Madhavan (1998, p. 2011) define transparency as “the extent to which price information in the two markets is observable.” Euro conversion instantly increased price transparency across these markets. We propose that this increased transparency may have triggered a round of competition among European stock exchanges. We use this natural experiment to examine an important question in the literature on competition among stock exchanges: what drives spread and volume changes when stock exchanges compete? To address this question, we analyze changes in trading costs (measured here by bid-ask spreads) and trading activity for nine euro and three non-euro European exchanges around euro conversion, from the fourth quarter in 1998 to the fourth quarter in 1999. Academic models of exchange competition suggest that exchanges compete to attract firm listings and/or trading volume (see the discussion of related literature in sub-section 2.1). According to the models, exchanges have multiple competitive levers to pull (trading costs, listing requirements, fees, etc.). Previous research shows that competition is often prompted by a specific external trigger, and that it is associated with big decreases in trading costs and substantial shifts in trading volume. However, the exact mechanism linking the competitive trigger to the trading cost and volume shifts is still unclear. Exchanges may compete directly by reducing spreads to attract volume. Alternatively, changes in spreads and volume may reflect the characteristics of firms that list on the different exchanges, exchange trading rules, or the broader institutional environment. Three constraints hamper prior research on exchange competition. First, the limited sample size (in terms of number of firms and/or exchanges) means prior researchers do not observe much cross-sectional variation. It is hard, therefore, to analyze the role of firm- or exchange-level characteristics in competition, or whether competition spills over to nearby exchanges. Second, changes in volume and changes in bid-ask spreads are endogenously determined in equilibrium (e.g., Glosten and Milgrom, 1985; Admati and Pfleiderer, 1988), so it is challenging to identify the direction of causation in this relation. Finally, because competition often plays out within the context of one or more on-going general trends, it is difficult to establish whether the competitive trigger is responsible for the observed reactions. Our research design addresses each of these three limitations. First, our natural experiment, euro adoption on January 4 1999, allows us to analyze a much broader range of exchanges and firms than most previous studies. With this sample, we explore the role of firm- and country-level characteristics as well as changes in spreads as determinants of the outcome of a round of competition among euro and non-euro European exchanges. We estimate a system of simultaneous equations using two stage least squares (2SLS) to control for endogeneity in the relation between changes in costs and volume. Finally, we structure our analysis as a difference-in-differences test to control for other concurrent trends, enabling us to more confidently attribute changes in spreads and volume to the euro conversion event. We find that in the year following euro conversion, bid-ask spreads across European exchanges fall an average of almost 9% while turnover (defined as trading volume scaled by shares outstanding) rises over 30%. Trading costs either decrease significantly or remain unchanged on each exchange, but turnover increases on some exchanges (the “winners”), and falls on others (the “losers”).1 Both euro and non-euro European markets are affected. These changes are commensurate with other notable stock exchange events, such as London’s Big Bang and the New York Stock Exchange's (NYSE) decimalization. Changes in spreads and turnover are significantly larger than those that occur in the prior year, and are significantly greater than those for a control group of NYSE firms. Difference-in-differences analysis with European and NYSE firms, controlling for numerous firm characteristics, is consistent with our hypothesis that euro conversion prompted the sizeable spread and volume shifts we document. Lastly, our 2SLS results show that controlling for the endogeneity of volume and spreads crucially affects these inferences. Our results show that Milan, Frankfurt, Paris, and London are the winners with significant increases in turnover, while Brussels and Madrid are the biggest losers with significant declines. Variations in both firm-specific and country-level factors help explain which exchanges win and which lose volume. Volume increases more for exchanges that list smaller, lower volatility firms and firms in the technology, telecom, and consumer goods and services industries. Exchanges located in countries with higher Gross Domestic Product (GDP) growth, lower accounting standards, and larger increases in the number of analysts also gain more volume. Several of the significant firm- and country-level variables (such as firm size and accounting standards) often serve in the literature as proxies for asymmetric information. We find that exchanges that begin the period with the weakest information environments are the biggest volume winners. Furthermore, the effect of firm size on turnover is stronger for firms listed on less transparent exchanges. These results are consistent with our expectation that euro conversion prompted competition by increasing transparency in market prices. Finally, the welfare consequences of changes in spreads are large; investors save almost €568 (approximately $570) million in trading costs during the fourth quarter of 1999 from the decline in effective bid-ask spreads after euro adoption. Our results contribute to the literature on exchange competition by providing insights into the question of who wins when exchanges compete. Controlling for the endogeneity of volume and spreads, we find that both firm- and country-level factors affect volume changes, while firm, country, and exchange trading rules affect spreads. Our results consistently show that transparency is an important driver of changes in spreads and volume. Moreover, we find a strong role for the exchange’s industry composition. These results suggest that it is important to consider all of these factors (i.e., the spread/volume relation, firm characteristics, exchange-level characteristics, and country-level macroeconomic and institutional characteristics) simultaneously when examining the outcomes of exchange competition. Our results also contribute to the literature on the agglomeration versus fragmentation of trading. We show only marginal increases in the trading activity market share of Europe’s largest exchanges, suggesting that euro conversion did not contribute to a “winner-take-all” outcome. Our welfare calculations show that the effects of competition are economically large and differ substantially between the winning and losing exchanges. Finally, we are the first article to provide statistically significant evidence that competition spills across borders.2 In sum, our results provide a rich understanding of the factors that affect both spreads and volume when exchanges compete. The remainder of our article is organized as follows. In Section 2, we describe our research design and sample. Univariate results are presented in Section 3; multivariate difference-in-differences regressions are reported in Section 4. Section 5 discusses welfare consequences, and Section 6 provides additional evidence related to competition. Section 7 concludes. 2. Research Design and Sample 2.1 Theory and Related Literature Our research is related to models of competition among stock exchanges, which typically assume that exchanges compete to attract either trading volume (order flow) or firm listings. In models of the competition for order flow, exchanges adjust trading fees (Ramos and von Thadden, 2008; Pirrong, 2000), disclosure requirements (Huddart, Hughes, and Brunnermeier, 1999), margin requirements (Santos and Scheinkman, 2001), and not-for-profit status and membership size (Pirrong, 2000). In models by Chemmanur and Fulghieri (2006), Foucault and Parlour (2004), and Amira and Muzere (2011), exchanges modify listing standards or listing fees to attract more firm listings. Di Noia (2001) and Arnold et al. (1999) look at mergers and alliances as responses to exchange competition. Empirical investigations of competition for order flow generally use one of four different approaches. The narrowest approach is lab experiments that examine the effect of different trading rules on the volume and spreads for one asset (Lamoureux and Schnitzlein, 1997; Bloomfield and O’Hara, 2000). Perhaps the broadest approach is papers that look at the cross section of spreads or trading volume to measure the effect of country-level institutional factors and exchange characteristics on exchange competition (e.g., Eleswarapu and Venkataraman, 2006; Frost, Gordon, and Hayes, 2006; Lo, 2013). Microstructure papers that examine the effects of different trading mechanisms across multiple exchanges, while not directly motivated as studies of competition, would fall within this category (Domowitz, Glen, and Madhavan, 2001; Jain, 2006). The third approach is to examine the competition for order flow for the same firm listed on multiple exchanges to determine the effect of exchange and firm characteristics on volume and/or spreads. The cross-listing literature falls into this category. See, for example, Battalio, Greene, and Jennings (1997), Blume and Goldstein (1997), Kwan, Masulis, and McInish (2015), and Halling et al. (2008). The final approach, and the one closest to our experiment, examines how exchanges competitively respond to a specific event (e.g., Aggarwal and Angel, 1999; Arnold et al., 1999; Bessembinder, 2003; Brown, Mulherin, and Weidenmier , 2008; Dewenter, Kim, and Novaes, 2010). Most of these papers narrowly focus on only two exchanges.3 In order to draw conclusions about the effect of exchange competition on spreads and volume, we must first establish that euro conversion was a competitive event. The path to euro conversion was long and deliberate. (See Internet Appendix, Table A1 for a summary of major actions and a timeline of related events.) The Treaty of Maastricht on European Union, signed in 1992, created the framework and timetable for euro adoption. In the May 8 1998 Joint Communique, eleven European nations formally agreed to convert their currencies to the euro, announcing irrevocable exchange rates among themselves. On December 31 1998, participating countries announced the euro conversion rates (see Figure 1), keeping the cross-country exchange rates established in May unchanged. The next day, on January 1 1999, the euro was adopted for all electronic assets and transactions. When the European stock exchanges re-opened on January 4 1999, the prices for all firms listed on the eleven euro country exchanges were denominated in euros. Actual euro coins and notes were not introduced until 2002. For our purposes, it is important to note that the only major euro-related change during our sample year 1999 was euro conversion. Figure 1. View largeDownload slide Euro conversion rates. The December 31 1998 press release with the original conversion rates can be found at the European Central Bank website: http://www.ecb.int/press/pr/date/1998/html/pr981231_2.en.html. Figure 1. View largeDownload slide Euro conversion rates. The December 31 1998 press release with the original conversion rates can be found at the European Central Bank website: http://www.ecb.int/press/pr/date/1998/html/pr981231_2.en.html. Our argument that euro conversion served as a trigger for competition within the wider scope of changes made to prepare for the euro is similar to numerous other exchange-competition papers that identify a single, specific action within a broader trend, usually a regulatory change, as the trigger for competition (e.g., Arnold et al., 1999; Dewenter, Kim, and Novaes, 2010; Spankowski, Wagener, and Burghof, 2012). Ramos (2003) argues that a specific external trigger is necessary to spur competition or reform because exchanges are reluctant to change the status quo. Some might contend that euro conversion was trivial, a simple change of numeraire that was well known in advance. However, euro conversion had a real effect—it enabled investors to immediately and easily compare the prices of stocks across eleven different markets. Euro conversion therefore increased price transparency across those markets. Whether or not the change in transparency was trivial or significant enough to prompt a competitive response by the exchanges is an empirical question. A related challenge for our research design is adequately controlling for other broad trends occurring at the same time that could also affect trading costs and volume and thus confound interpretation of our results. The three most obvious forces are: changes to trading technology that prompted the creation of electronic communications networks, or ECNs (key events identified in Panel B of Table A1 in the Internet Appendix); portfolio rebalancing in response to the removal of investment barriers in preparation for the single market and the euro; and the internet bubble. We design our difference-in-differences analyses to control for more general trends and conduct several robustness tests (see sub-section 4.4) to address these potential alternative explanations. We extend existing research in several dimensions. First, while many of the empirical papers of exchange competition acknowledge endogeneity between spreads and volume, only one (Kwan, Masulis, and McInish, 2015) controls for it. We control for this endogeneity with a two-stage-least squares empirical specification, and our results show that this correction affects inferences. Second, Garvey (1944) and Arnold et al. (1999) address the potential for competition to spill over onto nearby exchanges. Our setting provides a natural set of exchanges that could be affected by spillover from competition triggered by euro conversion, the non-euro European exchanges, giving us an opportunity to test explicitly for spillovers.4 Third, the literature identifies four sets of factors that potentially drive the relation between competition and trading volume: spreads, firm characteristics, exchange trading environment, and country-level macroeconomic or institutional factors. Although previous research examines various subsets of these factors, no one to date has simultaneously examined all four. Because our experimental design includes multiple exchanges with many firms, we are able to examine all four factors, providing a more comprehensive examination of what drives volume (and spread) changes when exchanges compete. 2.2 Data and Sample Description The main source for our data is Thomson Reuters’ DataStream (DS). We collect the daily closing transaction price, ask price, bid price, and trading volume for firms listed on twelve exchanges in the euro-zone countries (Amsterdam-Netherlands, Brussels-Belgium, Dublin-Ireland, Frankfurt and Xetra-Germany, Helsinki-Finland, Lisbon-Portugal, Luxembourg-Luxembourg, Madrid-Spain, Milan-Italy, Paris-France, Vienna-Austria) and on three major European exchanges in non-euro zone countries (Copenhagen-Denmark, London-United Kingdom, and Swiss-Switzerland) for 1998 and 1999. We also collect data on NYSE trading from the Center for Research in Security Prices (CRSP) over the same period to use as a control sample. Alternatively, we considered using trading on the same European exchanges in the periods immediately prior to or after our sample period as the benchmark. Unfortunately, prior period data are incomplete, and post period trading is distorted due to the internet trading boom. Another possible source for control firms is Nasdaq. In Figure 2, we graph the value of shares traded for the sample of European exchanges we analyze relative to the NYSE and Nasdaq (all normalized to 100 in 1997) in Panel A, and the NYSE, Nasdaq and MSCI-Europe stock market indices in Panel B (all normalized to 100 at the beginning of 1998). Figure 2 shows that overall trading activity and market performance for NYSE firms track those in Europe very closely just prior to and during our sample period, whereas Nasdaq firms do not. The internet bubble strongly affected Nasdaq firms during our sample period.5 Therefore, we believe contemporaneous NYSE data are a more appropriate benchmark. Figure 2. View largeDownload slide Comparison of trading activity and market performance on the European exchanges, the NYSE, and Nasdaq. Panel A. Value of shares traded Panel B. Stock market indices Source. Panel A: World Stock Exchange Fact Book, various issues. European sample includes data for domestic firms listed on exchanges in our eleven European countries. Values normalized to equal 100 in 1997. Panel B: DataStream. Values normalized to 100 on January 1998. Figure 2. View largeDownload slide Comparison of trading activity and market performance on the European exchanges, the NYSE, and Nasdaq. Panel A. Value of shares traded Panel B. Stock market indices Source. Panel A: World Stock Exchange Fact Book, various issues. European sample includes data for domestic firms listed on exchanges in our eleven European countries. Values normalized to equal 100 in 1997. Panel B: DataStream. Values normalized to 100 on January 1998. Table I summarizes trading activity and other descriptive statistics for our exchanges. The first two data columns provide the total number of firms and domestic market capitalization for all domestic firms on the exchanges as of December 31 1998 from the World Stock Exchange Fact Book. In terms of market capitalization, both the NYSE and London are substantially larger than the other exchanges, at nine and two trillion euros, respectively. The smallest exchanges in market capitalization are Vienna, Luxembourg, and Lisbon. Table I. Summary descriptive statistics for the exchanges This table provides summary descriptive statistics for fifteen European exchanges and the NYSE for the fourth quarter of 1998. The first twelve exchanges listed in the table are located in countries which adopted the euro in 1999; the next three exchanges are located in European countries that did not adopt the euro. The last two rows report results for the Overall (Europe) and NYSE samples. Exchange data as of December 31 1998, in columns 1 and 2 are from the World Stock Exchange Fact Book. Irish exchange data are from The Handbook of World Stock, Commodity and Derivatives Exchanges, accessed via Factiva. Frankfurt and Xetra Exchange data are combined. Trading data in columns 3–6 for the European firms listed on the exchanges are from DataStream or KIT. NYSE data are from CRSP. Total Trading Volume is the average daily trading volume for firms in our sample (expressed in millions of shares traded) for the fourth quarter of 1998, summed over all the stocks with volume data for each exchange. Domestic Market Capitalization is the average daily market capitalization (the daily closing price times shares outstanding, expressed in billions of euros) for the fourth quarter of 1998 for firms located in the exchange’s domestic country summed over all the stocks with price and shares outstanding data in our sample for each exchange. The initial sample includes all firms with at least one ask quote between October 1 1998 and December 31 1999. The final sample includes all firms with at least fifteen non-missing trading volume and quote observations (for which the ask quote exceeds the bid quote) in both 1998Q4 and 1999Q4, average daily market capitalization of at least 50 million euros during December 1998, and a minimum average daily trading volume of 1000 shares during 1998Q4. Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 Table I. Summary descriptive statistics for the exchanges This table provides summary descriptive statistics for fifteen European exchanges and the NYSE for the fourth quarter of 1998. The first twelve exchanges listed in the table are located in countries which adopted the euro in 1999; the next three exchanges are located in European countries that did not adopt the euro. The last two rows report results for the Overall (Europe) and NYSE samples. Exchange data as of December 31 1998, in columns 1 and 2 are from the World Stock Exchange Fact Book. Irish exchange data are from The Handbook of World Stock, Commodity and Derivatives Exchanges, accessed via Factiva. Frankfurt and Xetra Exchange data are combined. Trading data in columns 3–6 for the European firms listed on the exchanges are from DataStream or KIT. NYSE data are from CRSP. Total Trading Volume is the average daily trading volume for firms in our sample (expressed in millions of shares traded) for the fourth quarter of 1998, summed over all the stocks with volume data for each exchange. Domestic Market Capitalization is the average daily market capitalization (the daily closing price times shares outstanding, expressed in billions of euros) for the fourth quarter of 1998 for firms located in the exchange’s domestic country summed over all the stocks with price and shares outstanding data in our sample for each exchange. The initial sample includes all firms with at least one ask quote between October 1 1998 and December 31 1999. The final sample includes all firms with at least fifteen non-missing trading volume and quote observations (for which the ask quote exceeds the bid quote) in both 1998Q4 and 1999Q4, average daily market capitalization of at least 50 million euros during December 1998, and a minimum average daily trading volume of 1000 shares during 1998Q4. Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 Exchange country Exchange name Exchange symbol Local currency name Exchange data (12/31/1998) Initial dataset Final sample Total number of listed stocks Total domestic market capitalization (billions of euros) Total number of firm lists Total number of firm lists Total trading volume (mil. of shares) 1998Q4 domestic market cap. (billions of euros) (1) (2) (3) (4) (5) (6) Netherlands Amsterdam AMS Guilder 800 494 110 52 117 325 Belgium Brussels BRU Belgian Franc 176 208 287 108 6 129 Ireland Dublin DUB Pound 85 51 79 0 n.a. n.a. Germany Frankfurt FRA Mark 4,132 933 1,425 183 5 470 Finland Helsinki HEL Markka 123 129 151 40 25 76 Portugal Lisbon LIS Escudo 135 45 138 52 33 59 Luxembourg Luxembourg LUX Lux. Franc 308 28 65 0 n.a. n.a. Spain Madrid MAD Peseta 484 296 133 113 94 223 Italy Milan MIL Lira 304 483 315 224 293 396 France Paris PAR French Franc 911 574 986 355 93 686 Austria Vienna WBO Schilling 147 26 0 0 n.a. n.a. Germany Xetra XET Mark −-- −-- 332 120 12 459 Denmark Copenhagen CSE Krone 374 109 280 65 11 56 UK London LON Pound 2,921 1,935 1,909 616 895 1,620 Switzerland Swiss SWX Swiss Franc 486 590 485 255 61 546 Overall (Europe) n.a. n.a. n.a. 11,386 5,901 6,695 2,183 1,645 5,043 USA NYSE NYSE Dollar 3,382 9,310 2,027 1,749 952 8,313 The final four columns of Table I provide information on our sample selection. The initial European exchange sample has 6,695 firm lists, satisfying the requirement that a firm has at least one non-missing ask quote observation for the 15 month period from October 1 1998 through December 31 1999.6 Trading volume data for German firms on DataStream are scarce for this period, so we replace the DataStream trading volume for Frankfurt and Xetra using volume data purchased from Karlsruhe Institute of Technologie (KIT).7 Around the euro conversion day of January 1 1999, our empirical analysis focuses on two periods: the last quarter of 1998 (1998Q4) and the last quarter of 1999 (1999Q4). Comparing 1998Q4 to 1999Q4 allows some time for competitive effects to take place and also controls for any potential calendar year effects.8,9 We further require each firm to have at least fifteen non-missing trading volume and quote observations (for which the ask quote exceeds the bid quote) in both 1998Q4 and 1999Q4. This screen leads us to drop the Dublin, Luxembourg, and Vienna exchanges from our sample due to lack of data. Finally, coverage on DataStream varies widely across exchanges. To control for potential selection biases associated with variations in coverage and to make sure that our comparisons across exchanges are between similar, large, and actively traded firms, we require that sample firms have a minimum average daily market capitalization of 50 million euros during December 1998, and a minimum average daily trading volume of 1,000 shares during 1998Q4.10 The resulting sample includes thirteen exchanges (nine euro exchanges, three non-euro European exchanges, and the NYSE), with a total of 2,183 European firm lists and 1,749 NYSE firm lists. The Frankfurt Stock Exchange and Xetra are both owned by Deutsche Borse. Xetra was created in 1997 as an electronic trading platform for the Frankfurt Stock Exchange.11 During our sample period, DataStream separately reported trading via these two systems. Since Xetra was a new electronic exchange with relatively rapidly expanding volume, we exclude Xetra from some of our analyses. Excluding Xetra reduces the sample of European firm lists to 2,063. We considered analyzing individual firms that are cross-listed on multiple exchanges. Unfortunately, when a firm is listed on more than one exchange during our sample period, most trading takes place on one exchange, so the second listing usually does not pass our data screens. Ramos and von Thadden (2008, Table I) also note that although there are many foreign firms cross-listed on the European stock exchanges, the total value of foreign trading on most exchanges is negligible. Furthermore, the coverage of cross-listing status on DataStream for firms that are no longer listed is very poor.12 To address our research questions, we compute several statistics to analyze changes in trading costs and trading volume around euro conversion. We focus on two alternative measures of trading costs: percentage spreads and effective spreads.13 Percentage spreads are defined as (Ai,t−Bi,t)/Mi,t ⁠, where Ai,t and Bi,t are the closing ask and bid prices, respectively, for stock i on day t, and Mi,t is the midquote, Mi,t=(Ai,t+Bi,t)/2. We define the (percentage) effective spread as 2*abs(Pi,t−Mi,t)/Mi,t ⁠, where Pi,t is the closing transaction price for stock i on day t. We use both daily share volume (⁠ Voli,t ⁠) and turnover to measure trading activity. Turnover is given by Voli,t/SHROUTi,t ⁠, where SHROUTi,t is the number of shares outstanding for stock i on day t. 3. Univariate Analysis 3.1 Univariate Changes Table II reports statistics for our trading cost and trading activity measures by exchange for 1998Q4 and 1999Q4. For each firm, we calculate the average statistic across all days in a given quarter. Averages reported in Table II reflect the value-weighted average across firms, where the weights are the average firm market capitalization during December 1998. We also report percentage changes from 1998Q4 to 1999Q4, a t-test for whether the percentage change differs significantly from zero, and a nonparametric sign test of changes in trading costs and volume.14 Statistics that are significant at the 5% level are noted in bold type. Table II. Summary statistics of changes in trading costs and trading activity This table provides mean values for measures of trading costs and trading activity for sample firms listed on nine euro exchanges, three non-euro exchanges and the NYSE. Mean daily values are provided for 1998Q4 and 1999Q4. For each firm, we compute the mean of each statistic each quarter; reported results are value-weighted averages across firms, where the value weights reflect the average December 1998 market capitalization. Panel A reports Percentage Spreads, defined as (Ai,t−Bi,t)/Mi,t where Ai,t ⁠, Bi,t and Mi,t are the closing ask price, bid price and midquote, respectively, for stock i on day t. Panel B reports the Effective Spread, defined as 2*abs(Pi,t−Mi,t)/Mi,twhere Pi,t is the closing transaction price. Panel C reports Volume (in thousands of shares). Panel D reports Turnover, defined as volume divided by shares outstanding. We also report the percentage change from 1998Q4 to 1999Q4, along with t-statistics testing whether the percentage change is statistically different from zero. The last three columns report the number of firms on each exchange for which the firm-level quarterly average statistic increased or decreased from 1998Q4 to 1999Q4, along with results of the z-statistic for a nonparametric sign test. This sign test is calculated as (Number of Firm Lists with Positive Changes − 0.5 × N)/ N/4 ⁠, where N is the total number of firm lists. Changes that are significantly different at the 5% level or better are in bold type. Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Table II. Summary statistics of changes in trading costs and trading activity This table provides mean values for measures of trading costs and trading activity for sample firms listed on nine euro exchanges, three non-euro exchanges and the NYSE. Mean daily values are provided for 1998Q4 and 1999Q4. For each firm, we compute the mean of each statistic each quarter; reported results are value-weighted averages across firms, where the value weights reflect the average December 1998 market capitalization. Panel A reports Percentage Spreads, defined as (Ai,t−Bi,t)/Mi,t where Ai,t ⁠, Bi,t and Mi,t are the closing ask price, bid price and midquote, respectively, for stock i on day t. Panel B reports the Effective Spread, defined as 2*abs(Pi,t−Mi,t)/Mi,twhere Pi,t is the closing transaction price. Panel C reports Volume (in thousands of shares). Panel D reports Turnover, defined as volume divided by shares outstanding. We also report the percentage change from 1998Q4 to 1999Q4, along with t-statistics testing whether the percentage change is statistically different from zero. The last three columns report the number of firms on each exchange for which the firm-level quarterly average statistic increased or decreased from 1998Q4 to 1999Q4, along with results of the z-statistic for a nonparametric sign test. This sign test is calculated as (Number of Firm Lists with Positive Changes − 0.5 × N)/ N/4 ⁠, where N is the total number of firm lists. Changes that are significantly different at the 5% level or better are in bold type. Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Exchange N 1998Q4 1999Q4 Percentage Change Nonparametric Tests Mean Mean Mean t-stat Incr Decr z-stat Panel A. Percentage spreads Amsterdam 52 0.005 0.004 –0.091 –1.75 8 44 –4.99 Brussels 108 0.024 0.024 –0.024 –1.08 53 55 –0.19 Frankfurt 183 0.006 0.005 –0.132 –5.04 61 122 –4.51 Helsinki 40 0.006 0.004 –0.254 –9.45 5 35 –4.74 Lisbon 52 0.012 0.005 –0.322 –5.87 5 47 –5.82 Madrid 113 0.005 0.004 –0.057 –1.84 48 65 –1.60 Milan 224 0.006 0.005 –0.068 –2.93 75 149 –4.94 Paris 355 0.015 0.014 –0.018 –0.57 102 253 –8.01 Xetra 120 0.007 0.005 –0.170 –4.22 11 109 –8.95 Copenhagen 65 0.013 0.009 –0.238 –12.16 6 59 –6.57 London 616 0.009 0.008 –0.126 –9.03 222 394 –6.93 Swiss 255 0.046 0.035 –0.107 –1.82 43 212 –10.58 Overall(Europe) 2183 0.023 0.018 –0.086 –6.31 639 1544 –19.37 NYSE 1749 0.010 0.014 0.589 30.92 1053 696 8.54 Panel B. Effective spreads Amsterdam 52 0.005 0.005 0.132 1.93 14 38 –3.33 Brussels 108 0.020 0.019 –0.078 –3.56 43 65 –2.12 Frankfurt 183 0.004 0.003 –0.129 –3.76 58 125 -4.95 Helsinki 40 0.008 0.004 –0.465 –11.29 4 36 –5.06 Lisbon 52 0.013 0.005 –0.369 –6.88 3 49 –6.38 Madrid 113 0.005 0.005 0.005 0.12 49 64 –1.41 Milan 224 0.006 0.005 –0.027 –1.12 78 146 –4.54 Paris 355 0.013 0.013 –0.013 –0.40 103 252 –7.91 Xetra 120 0.009 0.007 –0.112 –4.83 24 96 –6.57 Copenhagen 65 0.011 0.009 –0.181 –7.53 18 47 –3.60 London 616 0.006 0.006 0.873 1.31 239 377 –5.56 Swiss 255 0.033 0.025 –0.110 –1.91 39 216 –11.08 Overall(Europe) 2183 0.017 0.014 0.083 0.58 672 1511 –17.96 NYSE 1749 0.003 0.004 0.446 20.26 819 930 –2.65 Panel C. Volume (/1000) Amsterdam 52 8129.0 8011.1 0.061 1.39 33 19 1.94 Brussels 108 62.4 63.3 –0.076 –1.64 43 65 –2.12 Frankfurt 183 205.0 115.8 0.488 2.53 97 86 0.81 Helsinki 40 8143.2 7372.5 0.203 1.36 22 18 0.63 Lisbon 52 2506.5 2807.0 0.558 1.79 27 25 0.28 Madrid 113 6560.0 6037.2 –0.009 –0.25 35 78 –4.05 Milan 224 7211.4 8895.0 0.624 5.91 123 101 1.47 Paris 355 556.7 534.3 0.418 6.26 203 152 2.71 Xetra 120 858.0 833.3 2.489 4.09 92 28 5.84 Copenhagen 65 1058.9 540.6 –0.092 –1.27 27 38 –1.36 London 616 8515.1 10618.1 0.301 12.38 331 285 1.85 Swiss 255 852.1 758.8 0.231 1.59 120 135 –0.94 Overall(Europe) 2183 2678.6 3025.4 0.371 7.96 1153 1030 2.63 NYSE 1749 2429.0 3536.6 0.524 18.73 1021 728 7.01 Panel D. Turnover (*1000) Amsterdam 52 5.459 5.547 0.051 1.08 32 20 1.66 Brussels 108 0.161 0.161 –0.086 –1.80 40 68 –2.69 Frankfurt 183 0.473 0.368 0.426 2.33 96 87 0.67 Helsinki 40 3.157 2.725 0.098 0.57 23 17 0.95 Lisbon 52 2.415 2.728 0.505 1.62 25 27 –0.28 Madrid 113 3.711 3.405 –0.071 –2.10 34 79 –4.23 Milan 224 3.040 3.415 0.583 5.69 119 105 0.94 Paris 355 0.857 0.792 0.359 5.48 191 164 1.43 Xetra 120 1.544 1.800 2.349 4.06 92 28 5.84 Copenhagen 65 2.405 1.802 –0.131 –1.84 25 40 –1.86 London 616 2.621 3.034 0.235 9.61 324 292 1.29 Swiss 255 0.518 0.418 0.214 1.48 114 141 –1.69 Overall(Europe) 2183 1.364 1.399 0.327 7.23 1115 1068 1.01 NYSE 1749 3.692 3.841 0.180 6.62 937 812 2.99 Results in Table II show that percentage spreads decrease almost 9% for the Overall (Europe) sample; the decrease in percentage spreads is significant based on both statistical tests. Results for effective spreads also show a significant decrease using the nonparametric test, although there is no significant change based on the t-statistic.15 Spreads either decrease significantly or remain unchanged for each of the European exchanges; there are no statistically significant increases in any trading cost measure. In particular, we see significant declines in trading costs on all three non-euro European exchanges in addition to many of the euro exchanges. The two exchanges that do not change percentage spreads based on either test statistic are Brussels and Madrid. Consistent with prior studies (e.g., Wahal, 1997; Foerster and Karolyi, 1998; Klock and McCormick, 1999; de Fontnouvelle, Fishe, and Harris, 2003), spreads decrease when there is more competition. For comparison, we also report results for the 1,749 NYSE firms that pass our data screens. Results for the NYSE show a significant increase in percentage spreads and mixed results for the change in effective spreads. Regarding trading activity, Table II shows that volume and turnover both increase over 30% for the full European sample. However, trading activity increases on some exchanges and decreases on others. Volume and/or turnover increase on Frankfurt, Milan, Paris, Xetra, and London, and also on the NYSE. Both measures decrease significantly on Brussels and Madrid.16 The effects we find are similar in economic magnitude to effects documented in prior analyses of other major events at stock exchanges (e.g., Brown, Mulherin, and Weidenmier, 2008; Pagano and Roell, 1990; Bessembinder, 2003). They are also significantly larger than changes that occurred in 1998 before euro conversion, suggesting that the effects we document are due to euro conversion itself rather than events associated with the longer path leading up to euro conversion.17 One potential concern is whether our results reflect euro adoption or perhaps something else that was going on at the same time. To address this concern, we conduct two additional tests. First, we repeat our test for a European Union (EU) country that adopted the euro after our event date.18 Specifically, we repeat our experiment for companies on the Athens stock exchange, using their euro adoption event date of January 1 2001. Results show that consistent with competitive effects, effective spreads decrease significantly more for firms on the Athens stock exchange around Greece’s euro adoption (from the fourth quarter of 2000 to the fourth quarter of 2001) than they do for a control sample of NYSE firms over the same period (Internet Appendix Table A9; see sub-section 3.2 for details on our NYSE matching procedure). Contrary to predictions, percentage spreads increase more in Athens than on the NYSE. t-statistics comparing changes in trading activity in Athens with those on the NYSE are not significant. However, nonparametric tests show that both volume and turnover increase significantly on Athens and do not change on the NYSE, and the chi-square test of medians shows that volume increases significantly more on Athens than the NYSE. It is difficult to draw strong inferences from these results given that we only have data for one exchange, and that euro adoption on this exchange occurred during a period of unusual stock market performance right after the internet bubble. Our overall conclusion is that the Athens stock exchange shows some of the same effects upon euro adoption in 2001 that we document for the original euro adoption in 1999. Second, we also rerun our experiment using our original exchanges for a placebo event day, January 1 2004. To implement this test, we collect data for the fourth quarter of 2003 and the fourth quarter of 2004, using the same exchanges and sample selection criteria we use for our original sample. Results (reported in the Internet Appendix, Table A10, Panel A) show that spreads also decrease significantly from 2003 to 2004 for the full sample of European firms. However, there is no significant change in trading volume, and turnover decreases significantly over the placebo window for the overall European sample. Moreover, Wilcoxon tests for changes in spreads and trading activity (Panel B) show that changes during 1998/99 are greater in magnitude than those in 2003/04 on most of the individual exchanges. Collectively, these results support our claim that the results we document during 1998/99 reflect a significant competitive event.19 3.2 Changes Relative to the NYSE To control for overall market trends, we next conduct univariate tests of whether changes for firms on the European exchanges differ significantly from our NYSE control firms. For each European firm, we select (with replacement) a matching NYSE firm with the closest market capitalization from the set of firms in the same industry.20 In Table III, we report the mean percentage change for the European firms relative to the corresponding change for their matching NYSE firms. So, for example, the first number in the panel (–0.604 for Amsterdam) shows that value-weighted percentage spreads for Amsterdam firms decline 60% more than spreads for their corresponding matched firms on the NYSE over the same period. Table III. Changes in trading costs and trading activity relative to NYSE This table reports percentage changes in trading costs and trading activity relative to the NYSE. Trading cost and trading activity measures are as defined in Table II. The table reports results for each individual European exchange relative to NYSE matching firms. For each European firm, the matching firm is the NYSE firm with the closest market capitalization from the set of firms in the same industry. We report the mean difference (Europe–NYSE) in percentage changes in each variable, and a t-test (in parentheses) of whether this difference is significantly different from zero. Significant differences at the 5% level or better are in bold type. Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Table III. Changes in trading costs and trading activity relative to NYSE This table reports percentage changes in trading costs and trading activity relative to the NYSE. Trading cost and trading activity measures are as defined in Table II. The table reports results for each individual European exchange relative to NYSE matching firms. For each European firm, the matching firm is the NYSE firm with the closest market capitalization from the set of firms in the same industry. We report the mean difference (Europe–NYSE) in percentage changes in each variable, and a t-test (in parentheses) of whether this difference is significantly different from zero. Significant differences at the 5% level or better are in bold type. Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Exchange name No. of firms Mean difference (European–NYSE) in percentage change Perc. spread Eff. spread Volume Turnover Amsterdam 52 –0.604 –0.024 –0.225 –0.051 (–4.64) (–0.21) (–2.13) (–0.74) Brussels 108 –0.761 –0.319 –0.301 –0.207 (–9.41) (–5.75) (–3.90) (–3.29) Frankfurt 183 –0.403 –0.214 0.118 0.142 (–6.71) (–3.34) (0.63) (0.78) Helsinki 40 –1.033 –2.204 –2.953 –0.877 (–7.42) (–7.01) (–5.44) (–3.31) Lisbon 52 –0.801 –0.433 0.249 0.315 (–8.82) (–3.58) (0.76) (0.98) Madrid 113 –0.721 –0.255 –0.576 –0.209 (–9.42) (–4.50) (–3.89) (–2.45) Milan 224 –0.599 –0.178 0.131 0.330 (–11.73) (–4.16) (0.86) (2.67) Paris 355 –0.624 –0.602 –0.561 0.029 (–12.01) (–7.32) (–4.86) (0.32) Xetra 120 –0.444 –0.196 2.182 2.083 (–5.37) (–2.88) (3.57) (3.59) Copenhagen 65 –0.595 –0.06 –0.241 –0.222 (–2.88) (–0.85) (–2.10) (–2.08) London 616 –0.782 0.427 –0.231 0.017 (–22.35) (0.64) (–6.24) (0.58) Swiss 255 –0.606 –0.612 –0.754 –0.138 (–7.77) (–5.88) (–3.79) (–0.92) Overall (Europe) 2183 –0.643 –0.336 –0.360 0.046 (–29.32) (–2.32) (–6.16) (0.98) Controlling for overall market movements, our main inferences hold. Spreads either decrease or remain unchanged relative to the matched NYSE firms. Trading activity declines significantly relative to matching firms on some exchanges (including Brussels, Helsinki, Madrid, and Copenhagen), and increases on others (Milan and Xetra). The relatively large jump in Xetra volume reflects the fact that Xetra was a very young and rapidly growing exchange at the time, which is why we exclude it from subsequent analyses. Finally, we note that, once we control for certain firm characteristics (market capitalization and industry) through our matching procedure, the list of volume winners and losers shifts somewhat. These results suggest to us that it is important to control for firm characteristics when evaluating the outcome of exchange competition. 3.3 A Comment on Tick Sizes Euro conversion on January 4 1999, changed the price unit (the currency), which in turn affected the relative tick size. (See Figure 1 for the euro conversion rates.) Conversion to euro prices lowered the price for all stocks except those in Ireland. Since the tick represents a lower bound on bid-ask spreads, changes in tick sizes may have affected bid-ask spreads which may confound our ability to identify the effect of competition. To explore this possibility, we calculate the relative tick size for each exchange.21 Controlling for changes in price levels, results indicate the minimum tick size decreased significantly on Amsterdam and Copenhagen. Therefore, decreasing tick size may be the cause of the decrease in spreads observed on Amsterdam and Copenhagen. In contrast, the effective tick size increased on most of the other euro exchanges. Despite the increase in tick size for these markets, however, previous results show that spreads decreased on many of them. We cannot reject that the correlation between exchange-level changes in spreads and changes in tick sizes equals zero. Overall, effects such as increased competition appear to overwhelm any tick size effects around euro conversion. 4. Multivariate Regressions So far we have shown that euro conversion is associated with significant changes in spreads and volume. Trading costs decrease or remain unchanged in all European markets. Trading activity increases significantly on some exchanges and decreases on others; there are some volume winners and some losers. These effects extend to non-euro European exchanges, and do not result mechanically from tick size changes. Univariate results in Section 3 are consistent with competition among exchanges. However, they do not control for confounding shocks or differences in the types of listed firms. To address these issues, we rely on a multivariate difference-in-differences methodology. This approach allows us to estimate the effect of the treatment (euro conversion) by controlling for confounding shocks with a set of firms that are similar except for the treatment. We conduct two sets of analyses. First, we consider the euro firms as the treatment group and the non-euro European firms as our control group. Because univariate results suggest that competitive effects extend to the non-euro European firms, we also conduct our difference-in-differences analysis combining euro and non-euro European exchange firms into the treatment group, and use NYSE firms as our control group. The NYSE firms should control for general global equity trends. Our sample screen requiring only large and actively traded firms provides some assurance that our treatment and control firms are similar. In addition, our multivariate regressions control for firm-specific characteristics as of the end of 1998, prior to conversion, that might differ across the treatment and control groups and therefore differentially affect either spreads or turnover. Use of these firm-level characteristics also allows us to explore the role of firm-level factors in determining winners and losers. We interpret the difference in the change in spreads or turnover between the treatment and control groups, controlling for firm-specific characteristics, as due to euro conversion. 4.1 Specification and Predictions Did euro conversion prompt exchanges to compete for order flow by reducing spreads? To control for the potentially endogenous relation between spreads and volume, we estimate a system of simultaneous equations using 2SLS to address this question. In the first stage, we estimate changes in percentage spreads and turnover from 1998Q4 to 1999Q4. Harris (1994), among others, suggests using lagged values of the regression variables as instruments. We use percentage change in average daily spread from the third quarter of 1998 to the fourth quarter of 1998 (⁠ ΔSpr98 ⁠) to instrument for changes in spreads over our event period, and percentage change in average daily turnover from the third quarter of 1998 to the fourth quarter of 1998 (⁠ ΔTurn98 ⁠) to instrument for change in turnover.22 ΔSpr^ and ΔTurn^ are the fitted values from the following equations: ΔSpri=α+β1ΔSpr98,i+β2ΔTurn98,i+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+ Industry Dummyi+εi  (1) ΔTurni=α+β1ΔSpr98,i+β2ΔTurn98,i+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+ Industry Dummyi+εi (2) where i denotes a firm-level variable, ΔSpri is the percentage change in firm i’s average daily closing (percentage) spread from 1998Q4 to 1999Q4 and ΔTurni is percentage change in turnover over the same period. In both equations, we include several firm-specific characteristics as of the end of 1998 that might differentially affect spreads or volume. Tick98Q4,i is the average relative tick size during the fourth quarter of 1998; EuroPrice98End,iis the average daily closing price for December 1998 in euros; 23MV98End,i is the average daily market value for December 1998 in billions of euros; Ret98Q4,i is the average daily return in the fourth quarter of 1998; Volat98Q4,i is the volatility of daily returns during the fourth quarter of 1998; and Foreigni is a dummy variable set equal to 1 if the firm’s home country is not the same as the country of the exchange on which it trades, and zero otherwise.24 According to Karpoff (1987), changes in price level and volatility should be positively related to volume. Industry dummies are based on the Industry Classification Benchmark Industry Codes from DataStream, which classify all the firms into 10 industries; see footnote 20 for more details. We also use industry clustered standard errors in both the first and second stage regressions. Using the fitted values ΔSpr^i and ΔTurn^i from Equations (1) and (2) should help control for an endogenous relation between changes in spreads and changes in turnover. To test whether euro conversion is associated with significant shifts in spreads and turnover, we estimate the following second stage equations, using the fitted values from the first stage above. ΔSpri=α+β0Di+β1ΔTurn^i+β2ΔTurn^i2+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi + εi (3) ΔTurni=α+β0Di+β1ΔSpri^+β2ΔSpr^i2+β3Tick98Q4,i+β4EuroPrice98End,i+  β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi + εi  (4) We include the squared terms to allow for non-linearities in the spread-turnover relation. The dummy variable, Di, is set equal to one for the treatment firms and zero for the control firms; it serves as our difference-in-differences estimator for the effect of euro conversion on changes in spreads and volume. In our first test, Di equals one for euro firms and zero for non-euro European firms. In the second test, Di equals one for European firms and zero for NYSE firms. Controlling for other firm-specific factors and endogeneity in the relation between spreads and volume, we expect spreads to decrease and volume to increase more for treatment firms than for control firms. Therefore, the dummy variable should be negative in the spread regressions and positive in the turnover regressions. Competition should result in bigger reductions in spreads and increases in trading activity for the treatment firms relative to the control sample. 4.2 Regression Results The first four columns of Table IV report results of the first stage of our two-stage analysis [Equations (1) and (2) above]. Lagged change in spread (⁠ ΔSpr98,i ⁠) significantly explains change in spread from 1998Q4 to 1999Q4 in columns (1) and (3), and therefore satisfies the inclusion restriction for instruments. It is not significant in the regressions of change in turnover, and therefore does not violate the exclusion restriction. Analogous results hold for our instrument for change in turnover (⁠ ΔTurn98,i ⁠). Table IV. Two stage-least squares regressions of changes in spreads and turnover on firm-specific explanatory variables This table reports results of 2SLS regressions of percentage changes in spreads and turnover from 1998Q4 to 1999Q4 on firm-specific variables. Results from the first stage regressions (defined in Equations (1) and (2) in the text) are in columns (1) through (4). The dependent variables are ΔSpri (or ΔTurni ⁠), percentage change in the firm’s average daily closing percentage spread (or turnover) from 1998Q4 to 1999Q4. Columns (5)–(8) report results for the second stage equations: ΔSpri=α+β0Di+β1ΔTurn̂i+β2ΔTurn̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi (3) ΔTurni=α+β0Di+β1ΔSprî+β2ΔSpr̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi  (4) where ΔSprî and ΔTurnî are the fitted versions from the first stage equations. In columns (5) and (6), the dummy variable, Di, is set equal to one for euro firms and zero for non-euro European firms. In columns (7) and (8), Di equals one for European firms and zero for NYSE firms. Tick98Q4,i is the average relative tick size during the fourth quarter of 1998. EuroPrice98End,i and MV98End,i are the average daily closing price (in euros) and equity market capitalization (in billions of euros) for December 1998. Ret98Q4,i and Volat98Q4,i, are the daily average return and volatility of daily returns for 1998Q4. All continuous variables are winsorized at +/– three standard deviations around the mean. Foreigni equals 1 if a firm’s home country is not equal to its exchange location, otherwise 0. We include industry dummies in all regressions to control for industry effects. We use industry clustered standard errors in both the first and second stage regressions. The table reports coefficient estimates, with t-statistics below. Coefficient estimates that are different from zero at the 5% or 1% levels are in bold type. First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 Table IV. Two stage-least squares regressions of changes in spreads and turnover on firm-specific explanatory variables This table reports results of 2SLS regressions of percentage changes in spreads and turnover from 1998Q4 to 1999Q4 on firm-specific variables. Results from the first stage regressions (defined in Equations (1) and (2) in the text) are in columns (1) through (4). The dependent variables are ΔSpri (or ΔTurni ⁠), percentage change in the firm’s average daily closing percentage spread (or turnover) from 1998Q4 to 1999Q4. Columns (5)–(8) report results for the second stage equations: ΔSpri=α+β0Di+β1ΔTurn̂i+β2ΔTurn̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi (3) ΔTurni=α+β0Di+β1ΔSprî+β2ΔSpr̂i2+β3Tick98Q4,i+β4EuroPrice98End,i+β5MV98End,i+β6Ret98Q4,i+β7Volat98Q4,i+Foreigni+Industry Dummyi+εi  (4) where ΔSprî and ΔTurnî are the fitted versions from the first stage equations. In columns (5) and (6), the dummy variable, Di, is set equal to one for euro firms and zero for non-euro European firms. In columns (7) and (8), Di equals one for European firms and zero for NYSE firms. Tick98Q4,i is the average relative tick size during the fourth quarter of 1998. EuroPrice98End,i and MV98End,i are the average daily closing price (in euros) and equity market capitalization (in billions of euros) for December 1998. Ret98Q4,i and Volat98Q4,i, are the daily average return and volatility of daily returns for 1998Q4. All continuous variables are winsorized at +/– three standard deviations around the mean. Foreigni equals 1 if a firm’s home country is not equal to its exchange location, otherwise 0. We include industry dummies in all regressions to control for industry effects. We use industry clustered standard errors in both the first and second stage regressions. The table reports coefficient estimates, with t-statistics below. Coefficient estimates that are different from zero at the 5% or 1% levels are in bold type. First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 First stage regressions Second stage regressions Euro versus non-euro European versus NYSE Euro versus non-euro European versus NYSE ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn ΔSpr ΔTurn (1) (2) (3) (4) (5) (6) (7) (8) Intercept –0.048 1.233 –0.105 1.282 –0.005 1.191 0.190 1.216 –1.31 8.63 –1.79 17.62 –0.04 7.53 2.07 11.50 ΔSpr98 –0.196 0.097 –0.178 0.011 –6.15 1.10 –8.97 0.21 ΔTurn98 0.008 –0.143 –0.005 –0.156 0.82 –6.24 –0.37 –6.46 Euro or European Dummy 0.058 0.241 –0.384 0.148 1.76 2.71 –12.24 2.47 Fitted ΔTurnover –0.016 –0.029 –0.20 –0.40 Fitted ΔTurnover2 –0.025 0.091 –0.68 2.17 Fitted ΔSpr –0.759 0.283 –1.63 0.91 Fitted ΔSpr2 –1.312 0.069 –0.76 0.22 Tick98Q4 –0.808 0.989 1.867 1.643 0.056 6.335 0.368 1.868 –0.70 0.13 1.28 0.25 0.05 0.85 0.26 0.29 Euro Price98end 0.000 –0.001 0.000 –0.001 0.000 –0.001 –0.000 –0.001 –1.94 –0.91 –0.23 –1.57 –1.58 –0.83 –2.05 –1.49 MV98end –0.001 –0.017 0.006 –0.009 –0.001 –0.016 0.007 –0.010 –0.66 –3.08 3.51 –6.08 –0.62 –2.54 4.22 –3.98 Ret98Q4 -11.840 –20.364 -13.621 –11.016 –8.568 –24.582 -14.688 –2.597 –4.81 –0.99 –4.28 –1.73 –3.63 –1.31 –6.54 –0.32 Volat98Q4 –2.745 –0.665 1.099 –7.449 –3.986 –5.104 –1.483 –8.235 –2.66 –0.12 0.70 –2.77 –3.94 –1.06 –1.23 –3.16 Foreign 0.070 0.259 –0.084 0.099 0.075 0.333 –0.070 0.127 1.24 1.11 –2.14 0.90 1.48 1.41 –2.26 1.09 Industry Dummies Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.067 0.040 0.046 0.038 0.041 0.036 0.136 0.031 Obs 2063 2063 3812 3812 2063 2063 3812 3812 We report the second stage of the 2SLS regressions in the last four columns of Table IV. Comparing euro to non-euro European firms [columns (5) and (6)], we see that the euro dummy is insignificant in the spread regression and significantly positive in the turnover regression. We interpret this coefficient as indicating that non-euro European exchanges also reduce spreads in an attempt to compete with euro exchanges. The significantly positive coefficient on the euro dummy in column (6) suggests that euro exchanges gain significantly more turnover than the other, non-euro European exchanges. Comparing firms on European exchanges to those on the NYSE [columns (7) and (8)], results show that, as predicted if euro conversion triggered competition, the coefficient on the European dummy is significantly negative in the spread regression and significantly positive in the turnover regression. Controlling for endogeneity in the relation between spreads and volume and for firm-specific factors, euro conversion results in lower spreads and higher turnover for European firms relative to a control sample of firms listed on the NYSE. Cross-sectionally, some models predict that the exchanges with lower transaction costs will take more volume away from the other exchanges (e.g., Admati and Pfleiderer, 1988; Chowdhry and Nanda, 1991; Ramos and von Thadden, 2008). We do not find that the coefficient on (fitted) change in spread significantly explains change in turnover. Aside from the endogenous relation between spreads and trading activity, therefore, we cannot conclude that changes in spreads drive changes in trading activity (or vice versa).25 If we estimate the second stage regressions using actual (as opposed to fitted) values for spreads and turnover, both of these explanatory variables are negative and significant, with their squared terms positive and significant (see Internet Appendix, Table A16). Furthermore, the European dummy in the turnover regression is still positive, but no longer significant. Thus, controlling for endogeneity affects inferences in important ways. From Table IV, we also see that changes in spread and turnover are associated with several firm-specific factors. Comparing European firms to the NYSE in column (7), spreads decrease more for firms with higher prices, smaller firms, firms with larger prior returns, and for firms whose home country differs from the exchange on which they list.26 Turnover increases more for smaller, less volatile firms [column (8)].27 Variations in these firm-specific factors help explain which exchanges win and which lose volume. We argue that euro conversion triggered competition by increasing the transparency of prices. Firm size is a common proxy for transparency (e.g., Halling et al., 2008). To explore the role of transparency more fully, we re-estimate the second stage regressions from columns (7) and (8), including an interaction between the European Dummy and firm size (MV98end).28 Results (reported in the Internet Appendix, Table A17) show that the interaction is significantly negative in the change in turnover regressions while the size coefficient by itself remains negative and significant. The negative relation between firm size and turnover is more pronounced for firms in Europe than for NYSE firms, consistent with increased transparency across Europe. Another dimension along which firms vary is by industry. One possibility is that some industries consistently win volume after euro conversion, while others lose. Winning exchanges may be those with the greatest concentration of firms in preferred industries. To examine the role of industry composition, we conduct an F-test of equality across the industry dummies included in columns (7) and (8). These tests reject equality at the one percent level. The sorted industry dummy coefficient estimates (not tabulated) suggest that spreads fall the most (or, rise the least) for telecom firms, with the next largest decreases observed by high tech, oil and gas, and financial firms. Utilities and consumer services firms have the smallest declines in spreads. For turnover, high tech firms have the biggest increase followed by telecom, and then by consumer goods, consumer services and industrials; utilities and oil and gas firms have the smallest turnover gains. 4.3 Country-level Institutional Factors Several of the intercepts in the second stage regressions in Table IV are statistically significant even after we control for the spread/volume relation, European listing, and firm-specific factors, suggesting that a substantial portion of the change in spreads and turnover is still unexplained. In this section, we explore broader country-level factors that may help explain variation in changes in spreads and turnover. These factors could include macroeconomic conditions such as GDP growth or inflation, institutional factors such as political stability or accounting standards, and exchange-level characteristics such as number of listed firms and trading hours or trading rules.29 We draw representative country factors from the competition models discussed in sub-section 2.1, plus the literatures on cross border mergers and acquisitions, law and finance, multi-market trading, and the transmission of crises and shocks (i.e., contagion).30 Unfortunately, we cannot simply include these variables in our regressions because measures for the country-level characteristics would be co-linear with the European dummy. The regression analysis in Equations (3) and (4) assumes that the relation between changes in spread or turnover and firm characteristics is the same across all exchanges. We modify these regressions and allow the intercepts and other coefficients to vary across eleven European exchanges (we exclude NYSE firms from this analysis). Specifically, in both stages, we break apart the constant term and instead include dummy variables for each European exchange. We also interact each explanatory variable with the exchange dummy coefficients to allow the coefficient on each firm-specific variable to vary across exchanges. We interpret the exchange dummy coefficient estimates in the second stage regression as a measure of the country-level unexplained change in spreads or turnover. Table V provides the correlations between the eleven European exchange-level intercept estimates from the second stage spread and turnover regressions with their respective country-level factors. (See the table legend for the definition and sources of the country-level variables. We report results for additional country-level variables in Internet Appendix, Table A18.) Table V. Correlations between exchange intercepts and country-level institutional factors This table reports correlations between the eleven European exchange dummy intercepts from the second stage spread and turnover regressions and country-level and exchange level factors. FDI Inflows, GDP, Exports, Wages, CPI, and LT Interest Rates are sourced from DataStream; changes are from 1998 to 1999. Political Stability, Regulations, Rule of Law, and Control of Corruption, all at 1998 values, are sourced from the World Bank Governance Indices. Insider Trading Enforcement equals the number of years since the first enforcement of insider trading laws, and is sourced from Bhattacharya and Daouk (2002). Shareholder Rights Index and the LLSV Accounting Standards are from La Porta et al. (1998). Anti-Self-Dealing Index is from Djankov et al. (2008). Institutional Ownership is from Ferreira, Massa, and Matos (2010). The accounting Transparency (CIFAR) and Disclosure indices are from Bushman, Piotroski, and Smith (2004). Change in the Number of Analysts, 1998 to 1999, equals the change in the Number of Analysts, which is the average for the thirty firms with the highest number of EPS analysts for each market in 1998 and 1999 from IBES. Number of Listed Firms, Value Traded and Market Cap/GDP are from the Handbook of International Stock Exchanges, with values for 1998. Average and Minimum Correlations of Returns for each exchange’s stock index with the indices of the other exchanges are from DataStream daily total market indices in 1999. Longer Trading Hours are sourced from a Factiva news search. Market Maker Obligatory, Centralization, Depth, Automatic Execution, and Mutual Ownership are provided by PK Jain and correspond to 2000. Correlations that are significant at the 10% or better level are in bold type and marked with *. Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Table V. Correlations between exchange intercepts and country-level institutional factors This table reports correlations between the eleven European exchange dummy intercepts from the second stage spread and turnover regressions and country-level and exchange level factors. FDI Inflows, GDP, Exports, Wages, CPI, and LT Interest Rates are sourced from DataStream; changes are from 1998 to 1999. Political Stability, Regulations, Rule of Law, and Control of Corruption, all at 1998 values, are sourced from the World Bank Governance Indices. Insider Trading Enforcement equals the number of years since the first enforcement of insider trading laws, and is sourced from Bhattacharya and Daouk (2002). Shareholder Rights Index and the LLSV Accounting Standards are from La Porta et al. (1998). Anti-Self-Dealing Index is from Djankov et al. (2008). Institutional Ownership is from Ferreira, Massa, and Matos (2010). The accounting Transparency (CIFAR) and Disclosure indices are from Bushman, Piotroski, and Smith (2004). Change in the Number of Analysts, 1998 to 1999, equals the change in the Number of Analysts, which is the average for the thirty firms with the highest number of EPS analysts for each market in 1998 and 1999 from IBES. Number of Listed Firms, Value Traded and Market Cap/GDP are from the Handbook of International Stock Exchanges, with values for 1998. Average and Minimum Correlations of Returns for each exchange’s stock index with the indices of the other exchanges are from DataStream daily total market indices in 1999. Longer Trading Hours are sourced from a Factiva news search. Market Maker Obligatory, Centralization, Depth, Automatic Execution, and Mutual Ownership are provided by PK Jain and correspond to 2000. Correlations that are significant at the 10% or better level are in bold type and marked with *. Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Correlation with European intercepts from regressions for Spreads Turnover Macro and Institutional Country-Level Factors Changes in the economy  Change FDI inflows 0.398 –0.226  Change GDP –0.175 0.522*  Change exports –0.212 –0.177  Change wages –0.174 –0.045  Change CPI –0.362 0.023  Change LT int. rates –0.082 0.050 Quality of political institutions  Political stability –0.411 0.336  Regulations –0.190 –0.175  Rule of law –0.215 –0.179  Control of corruption –0.118 –0.119 Governance  Shareholder rights –0.297 –0.037  Institutional ownership –0.065 –0.253  Anti-self dealing index 0.002 –0.082  Insider trading enforcement 0.215 –0.308 Accounting and transparency  LLSV accounting standards 0.400 –0.783*  Transparency (CIFAR) 0.095 –0.691*  Number of analysts, 1998 0.794* –0.286  Change in the number of analysts, 1998–1999 –0.717* 0.695*  Disclosure 0.637* –0.574*  Euro conversion rate 0.091 –0.056 Exchange-Level Factors Size and performance  Number of listed firms 0.606* –0.010  Value traded 0.454 –0.098  Market cap/GDP –0.026 –0.299  1998 Stock market return 0.122 –0.024  Avg corr returns w/ other EU markets 0.675* –0.333  Min corr returns w/ other EU markets 0.802* –0.167  Longer trading hours 0.465 –0.435 Trading rules  Market maker obligatory 0.557* 0.007  Centralization 0.684* –0.159  Depth 0.392 –0.251  Automatic execution –0.570* 0.126  Mutual ownership 0.049 0.168 Correlations with the exchange-specific intercepts in the spread regressions indicate that, after controlling for all of the firm-specific characteristics, spreads fall the most in exchanges located in the countries with weak transparency as of the beginning of the period (measured with the Number of Analysts, 1998 or Disclosure), an increasing number of analysts over the period, and several exchange-level trading characteristics: fewer listed firms, lower stock index return correlations with the other European markets, no obligatory market makers, no centralization, and automatic execution.31 Exchange intercepts from the turnover regressions are related primarily to changes in GDP and measures of transparency, and not linked to any of the exchange-level variables.32 Turnover rises the most on the exchanges that are in countries with relatively rapidly growing economies (Change GDP), relatively weak transparency at the beginning of the period (measured with La Porta et al’s (1998),LLSV Accounting Standards index, CIFAR and Disclosure),33 and increases in the number of analysts. Frost, Gordon, and Hayes (2006) show, with levels, that an index of market development that includes two measures of trading volume is higher for exchanges located in countries with high transparency. These results show that the gains in turnover in response to competition come primarily to the exchanges in countries with low transparency. To further explore the effects of transparency, we also estimate regressions similar to those in Table IV using dummy variables to split exchanges into groups based on the accounting and transparency variables that are significant for change in turnover in Table V. These variables are LLSV Accounting Standards, Transparency (CIFAR), Change in the Number of Analysts, and Disclosure. We construct dummy variables for NYSE firms and for firms on exchanges in High, Medium, or Low groups based on their exchange-level country transparency factors. We re-estimate results reported in columns (7) and (8) of Table IV, splitting the intercept into these dummies, and also interacting each of these dummies with firm size. Competition due to increased transparency after euro conversion implies that the strongest effects should be present for firms listed on the least transparent exchanges. Results (see Internet Appendix, Table A19) show that for both LLSV and CIFAR, turnover increases more (higher intercepts) and the sensitivity of change in turnover to firm size is greater in magnitude (more negative) for firms in countries that start with low transparency as captured by these two measures.34 Firms listed on exchanges in countries that begin the period with the weakest information environments (as measured by LLSV and CIFAR) experience the biggest increase in turnover and see the greatest sensitivity of change in turnover to firm size. These results support our transparency interpretation. 4.4 Alternative Potential Explanations As noted earlier, there were several other broad trends occurring during our event window that could also affect trading costs and volume and thus confound interpretation of our results. The internet bubble was well under way during our sample period. The increases in trading activity and corresponding decreases in trading costs that we document may reflect changes due to internet-related trading rather than euro adoption. To address this concern, we repeat our 2SLS analysis, excluding firms in the technology industry, and also excluding those in both technology and telecom (two industries affected by the bubble). We also repeat our analysis excluding smaller firms (those in the bottom two NYSE size deciles), which may be more affected by the technology boom. Results from these three robustness tests are similar, and provide the same inferences, as those we report in Table IV (see the Internet Appendix, Table A20). The positive and significant coefficient estimate for the European dummy variable in the turnover regression [column (8) in Table IV] is also consistent with “cross market rebalancing,” where investors respond to a shock by rebalancing their portfolios (Kodres and Pritsker, 2002). It could be that euro conversion spurred investors to increase their holdings of euro stocks in general, resulting in a relatively larger increase in volume for euro exchanges than for the NYSE. The creation of euro-related indices could exacerbate this trend, particularly for large firms, through the formation of index-matching portfolios. However, the best known euro indices were created before euro conversion. For example, Amsterdam’s EuroTop 100 index was created in 1990, and the Dow Jones Euro STOXX indices in February of 1998. Also, as noted above, the euro conversion date was known well in advance, suggesting portfolio rebalancing would not necessarily concentrate right around the actual conversion date. Furthermore, if index-related rebalancing were driving our euro dummy results, we would expect that the dramatic increases in volume would be concentrated in the largest firms. However, the volume effects we observe are greater for smaller firms. This pattern suggests our volume results are not driven by general portfolio rebalancing efforts where investors moved into large, index-linked euro firms. Finally, the years around our sample period involved substantial changes to trading technology and a move toward electronic trading. Results in Table V show that exchanges with automatic execution have greater decreases in spreads. However, Internet Appendix Table A1, Panel B, shows no significant changes to trading technology during 1998 or 1999. Changes in trading technology began well before our sample period. Thus, we rely on our difference-in-differences test to control for this more general trend. 4.5 Discussion: Regression Results Overall, variations in both firm-specific and country-level factors help explain which exchanges win and which lose volume. Turnover increases more for exchanges with relatively smaller, lower volatility firms in countries with higher GDP growth, lower accounting standards, and larger increases in the number of analysts. Exchanges that win volume also have the biggest concentrations of firms in the industries that experience the biggest jumps in volume (such as high tech, telecom, consumer goods, consumer services, and industrials). These effects combine to determine the winning and losing exchanges. To illustrate, in Table VI we report exchange-specific averages for some of the explanatory characteristics that are significant determinants of change in turnover, along with the ranking (from high to low) for each factor. Our winning exchanges include Milan, Frankfurt, Paris, and London. Milan lists smaller firms (ranking 9th in terms of average market capitalization), and has many firms in some of the winning industries (especially telecom with 17.5% of all telecom firms in the sample, and consumer goods and services with 10.7%). Frankfurt benefits from relatively low accounting standards and having few firms in the oil and gas industry, only 2% of total; this exchange wins despite listing the most volatile firms. London wins because it has firms with lower volatility, the second biggest jump in analysts, and a higher concentration of telecom (27.5%) and technology (26.7%) firms. Paris primarily benefits from listing many technology (28.1%) and telecom (12.5%) firms, while its GDP growth and change in analysts are in the upper half of the group. Table VI. Exchange averages for significant firm-specific and institutional variables This table reports the average by exchange and ranking (highest to lowest) for some of the firm-specific and institutional variables that are significant in the 2SLS analysis (reported in Table IV) or in the country-level analysis (in Table V) for our sample. MV98end and Volat98Q4 are averages of firm specific values for our sample. Country-level GDP data are from DataStream, change is from 1998 to 1999. LLSV Accounting Index is from LaPorta et al. (1998). Change in Analysts is the change in the average number of analysts from 1998 to 1999 for the thirty firms with the largest number of EPS analysts, as reported by IBES. Rank #1 is assigned to the exchange with the largest number. Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 Table VI. Exchange averages for significant firm-specific and institutional variables This table reports the average by exchange and ranking (highest to lowest) for some of the firm-specific and institutional variables that are significant in the 2SLS analysis (reported in Table IV) or in the country-level analysis (in Table V) for our sample. MV98end and Volat98Q4 are averages of firm specific values for our sample. Country-level GDP data are from DataStream, change is from 1998 to 1999. LLSV Accounting Index is from LaPorta et al. (1998). Change in Analysts is the change in the average number of analysts from 1998 to 1999 for the thirty firms with the largest number of EPS analysts, as reported by IBES. Rank #1 is assigned to the exchange with the largest number. Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 Exchange No. of firms MV98end Volat98Q4 Change GDP LLSV accounting index Change in analysts Average Rank Average Rank Value (%) Rank Value Rank Value (%) Rank (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Amsterdam 52 7.44 2 0.0328 3 6.5 2 64 6 –1.8 6 Brussels 108 7.38 3 0.0276 8 4.3 6 61 10 –10.8 11 Copenhagen 65 0.90 11 0.0230 11 3.4 9 62 9 –1.9 8 Frankfurt 183 2.34 6 0.0353 1 3.5 8 62 8 –10.5 10 Helsinki 40 2.12 7 0.0333 2 4.9 4 77 2 3.9 3 Lisbon 52 1.14 10 0.0273 9 7.0 1 36 11 23.0 1 London 616 2.53 5 0.0265 10 4.1 7 78 1 6.4 2 Madrid 113 2.11 8 0.0279 7 5.5 3 64 5 –7.6 9 Milan 224 2.05 9 0.0292 6 2.0 10 62 7 –1.8 7 Paris 355 5.39 4 0.0300 5 4.3 5 69 3 –1.3 5 Switzerland 255 9.61 1 0.0309 4 2.0 11 68 4 –1.3 4 In contrast, Brussels and Madrid are among the main losers in terms of changes in turnover. Brussels lists larger firms and has the largest decline in the number of analysts; it also suffers because it does not have a significant number of firms in any of the winning industries. Madrid has stronger accounting ratings, a decline in the number of analysts, and also few firms in the winning industries. Overall, these firm-specific and country factors combine to determine winners and losers. 5. Welfare Consequences To estimate whether changes based on competition in response to euro adoption are material, we perform several social welfare calculations. We begin with a calculation of the change in consumer welfare generated by the reduction in bid-ask spreads. This calculation is similar to that in Brown, Mulherin, and Weidenmier (2008). Specifically, we first estimate actual transactions costs paid by investors via effective bid-ask spreads during the fourth quarter of 1999. To do this, we multiply the value (in thousands of euros) of volume traded in our sample firms each day by one-half of the effective spread for that firm that day.35 We sum this cost across all firms and days on a particular exchange to derive an estimate of the total transactions costs paid by investors on that exchange during 1999Q4. We next estimate what transaction costs would have been during 1999Q4 if, hypothetically, effective spreads had remained constant at their 1998Q4 levels. So, for each firm we estimate the average effective spread during 1998Q4. We multiply the value (in thousands of euros) of volume traded each day during the fourth quarter of 1999 by one-half of the average effective spread for that firm during 1998Q4. This figure (summed across all trades for firms on a given exchange during 1999Q4) represents what the total cost to investors trading on that exchange would have been if effective spreads had remained constant at their 1998 levels (please see Panel A of Table VII). Table VII. Welfare consequences This table provides several estimates of the welfare consequences of changes around euro adoption. In Panel A, we estimate the change in welfare associated with change in effective spreads. This panel reports actual transaction costs paid by investors through effective bid-ask spreads during the1999Q4, and hypothetical 1999 transaction costs (multiplying 1999Q4 trading volume by the 1998Q4 effective spread). We also report the difference and percentage change between quarters. In Panel B, we report the difference in effective spreads, Market Cap/GDP and the exchange return from 1998 to 1999 by exchange. Results in Panel B are sorted by change in turnover from Table II. Losers are exchanges with a significant decrease in turnover based on at least one of the test statistics from Table II. Winners are those with a significant increase. We exclude Xetra from the analysis in Panel B. Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Table VII. Welfare consequences This table provides several estimates of the welfare consequences of changes around euro adoption. In Panel A, we estimate the change in welfare associated with change in effective spreads. This panel reports actual transaction costs paid by investors through effective bid-ask spreads during the1999Q4, and hypothetical 1999 transaction costs (multiplying 1999Q4 trading volume by the 1998Q4 effective spread). We also report the difference and percentage change between quarters. In Panel B, we report the difference in effective spreads, Market Cap/GDP and the exchange return from 1998 to 1999 by exchange. Results in Panel B are sorted by change in turnover from Table II. Losers are exchanges with a significant decrease in turnover based on at least one of the test statistics from Table II. Winners are those with a significant increase. We exclude Xetra from the analysis in Panel B. Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel A. Change in Welfare Associated with Change in Effective Spreads (1000s euros) Exchange 1999 Actual 1999 Hypothetical Difference (Act–Hyp) Percentage change (%) Amsterdam 424,577 398,327 26,250 6.6 Brussels 29,964 35,240 (5,276) –15.0 Frankfurt 24,575 31,417 (6,841) –21.8 Helsinki 42,881 89,242 (46,361) –52.0 Lisbon 22,771 62,364 (39,593) –63.5 Madrid 113,859 112,840 1,020 0.9 Milan 331,635 327,108 4,526 1.4 Paris 336,688 429,355 (92,667) –21.6 Xetra 118,071 136,125 (18,054) –13.3 Copenhagen 29,249 38,289 (9,040) –23.6 London 1,155,126 1,346,067 (190,941) –14.2 Swiss 118,932 309,932 (191,000) –61.6 Overall(Europe) 2,748,328 3,316,305 (567,977) –17.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Panel B. Welfare Measures for Winners and Losers Exchange Difference in % Change in Exchange return VW turnover (% change) effective spreads (1000 euros) market cap/GDP (% change) Brussels (5,276) –16.9 –8.3 –8.6 Loser Madrid 1,020 18.6 13.5 –7.1 Loser Copenhagen (9,040) 19.1 18.6 –13.1 Amsterdam 26,250 23.4 26.9 5.1 Helsinki (46,361) 146.3 159.0 9.8 Swiss (191,000) 10.7 4.8 21.4 Lisbon (39,593) 14.1 7.7 50.5 London (190,941) 20.2 16.0 23.5 Winner Paris (92,667) 68.1 50.3 35.9 Winner Frankfurt (24,896) 46.6 38.2 42.6 Winner Milan 4,526 43.2 20.3 58.3 Winner Average: losers (2,128) 0.9 2.6 –7.8 Average: winners (75,994) 44.5 31.2 40.1 Results show that, for the full sample, traders paid a total of €2.748 billion in bid-ask-spread related transaction costs during 1999Q4. However, if effective bid-ask spreads had remained at their 1998 levels, this total would have been €3.316 billion. Investors therefore saved a total of €568 ($570) million (or about 17%) in trading costs because of the decline in effective bid-ask spreads around euro adoption. This value is calculated based on trading volume during one quarter, so annualized savings would be approximately four times higher.36 In general, bid-ask spreads have been decreasing in all markets over the last few decades, so some of this decline may represent a natural time trend.37 (Our difference-in-differences tests in Table III control for this time trend using a matching sample.) We note here that these numbers are economically large. We also provide two other estimates of welfare implications. To get a broader estimate of the impact of euro adoption, we calculate a time series of the ratio of equity market capitalization to GDP. Market Cap/GDP is a common measure of market depth, or of the development of the local capital markets.38 We also compare the returns to investors in these markets. We calculate the exchange return for each of our exchanges with exchange-level indices reported by the World Stock Exchange Handbook from year end 1998 to year end 1999. We report these results in Table VII, Panel B (along with the change in effective spreads from Panel A). In this panel, we sort exchanges into groups according to their change in turnover from Table II; those exchanges with a significant decrease in turnover are classified as losers and those with a significant increase are the winners. Results in Panel B show significant differences in all three measures between the losing and winning exchanges. The average losing exchange has savings due to effective spreads of only €2 million, a very small increase in Market Cap/GDP, and an average exchange return of only 2.6%. In contrast, the winning exchanges save an average of €76 million in transaction costs, have a 45% increase in Market Cap/GDP, and a 31% average exchange return. Other general trends affecting all firms or exchanges during our sample period would not explain these differential results we find across winning and losing exchanges. These amounts are economically large, and suggest to us that the effects we document have broad consequences. The winning exchanges win along several different dimensions. 6. More Evidence of Competition So far, we have presented evidence that is strongly consistent with exchange competition for trading volume. Competition among Europe’s exchanges during our event period is also apparent in the alternative competitive levers that exchanges pull and in outcomes other than trading volume. 6.1 Alternative Levers As noted in Section 2, exchanges have multiple levers to pull when they compete with each other. (We summarize major activities in these areas in the Internet Appendix, Table A1, Panel C.) One obvious lever is trading fees (also referred to as commissions or stamp taxes). Trading fees are an exchange decision, whereas bid-ask spreads are based on equilibrium outcomes of market participants. The World Stock Exchange Fact Book lists commissions and fees for each exchange, but shows no changes in these fees at any of our exchanges from 1998 to 2000. A Factiva news search indicates only two changes: Switzerland imposed a new tax on “remote members” so that they were treated the same as domestic brokers (Boland, 1998); and Denmark delayed dropping the share transaction tax from the beginning of 1999 to October 1 1999 in order to “generate state revenue” (Reuters News, November 25 1998).39 So, in the two cases with changes, the fees went up or a decrease was delayed. Competition does not appear to have played out in transaction fees. An alternative means to attract trading volume is expanded trading hours (e.g., Garvey, 1944). Five of the twelve exchanges did expand trading hours, albeit toward the end of the period (London, Milan, and Paris in September and Brussels and Madrid in October 1999). Two of these exchanges, Brussels and Madrid, were the univariate volume losers, suggesting that expanded trading hours may have been a response to lower trading volume. Di Noia (2001) and Arnold et al. (1999) argue that mergers and alliances are a response of exchanges to competition. Stockholm and Copenhagen announced an agreement to create a common Nordic securities market called Norex in early 1998. Frankfurt and London announced a “strategic alliance” to create a common exchange in July of 1998. During 1999, this bilateral alliance fell apart and talks shifted to a group of eight major exchanges (Amsterdam, Brussels, Frankfurt, London, Madrid, Milan, Paris, and Swiss). During 2000, these eight exchanges signed a memo of understanding to form a partnership, but talks broke down 4 months later. The first successful full merger of European exchanges, Euronext, took place in 2000 among Amsterdam, Paris, and Brussels. No doubt, all of these other levers take time to pull. Our evidence of big shifts in transactions costs and trading activity in the year following euro conversion may be viewed as the immediate competitive response, while the exchanges were simultaneously maneuvering on multiple other, longer-term fronts. 6.2 Other Competitive Outcomes In non-tabulated results, we find that the variation in spreads across firms on the twelve European exchanges falls significantly over our sample period.40 This convergence in spreads is similar to the convergence in other European financial prices found around this time.41 Since, as discussed in sub-section 2.1, no specific EU integration actions were taken during our sample period, it is hard to conclude that the convergence we find here is due to the general integration process associated with preparations for euro conversion. Instead, we interpret the convergence of equity trading costs around euro introduction as consistent with European market makers matching their competitors’ cost reductions. Lower average spreads, as well as a decrease in variation across spreads, are consistent with competition across all of the European exchanges. Previous literature discusses whether competition leads to more or less consolidation of trading across markets (e.g., Stigler, 1961; Pagano, 1989; Chowdhry and Nanda, 1991; Madhavan, 1995; Gehrig, 1998). To examine this question more fully, in Figure 3 we report some measures of market concentration among European exchanges. Panel A reports the Herfindahl–Hirshman Index (HHI), and Panel B reports the market share of the largest three exchanges based on value traded in US dollars, from the World Stock Exchange Fact Book. Results in Panel A show that the HHI trended downward over this period, with a slight interruption in 1999 immediately after euro conversion, especially for the sample including all European exchanges. Similar results hold in Panel B. Trading in Europe is fairly concentrated, with the top three exchanges holding 77% of the market. These results provide evidence that euro conversion did not contribute to a “winner-take-all” outcome (e.g., as in Chowdry and Nanda, 1991). Figure 3. View largeDownload slide Market share of Europe’s largest exchanges. Panel A. Industry concentration (Herfindahl–Hirshman Index, or HHI) based on the dollar value of shares traded for the eleven European exchanges in our sample. Panel B. Market share of Top 3 Euro and European Exchanges (based on value of shares traded) Source:World Stock Exchange Fact Book. Data covers twelve exchanges included in our sample; Frankfurt and Xetra are combined. Results are computed for all European exchanges, and separately for the euro exchanges. In Panel A, the Herfindahl–Hirshman Index is calculated as the sum of the squared values of the market share of each exchange. In Panel B, the three largest exchanges for volume of shares traded are London, Milan, and Frankfurt/Xetra. The three largest for value of shares traded are London, Frankfurt/Xetra, and Amsterdam. Figure 3. View largeDownload slide Market share of Europe’s largest exchanges. Panel A. Industry concentration (Herfindahl–Hirshman Index, or HHI) based on the dollar value of shares traded for the eleven European exchanges in our sample. Panel B. Market share of Top 3 Euro and European Exchanges (based on value of shares traded) Source:World Stock Exchange Fact Book. Data covers twelve exchanges included in our sample; Frankfurt and Xetra are combined. Results are computed for all European exchanges, and separately for the euro exchanges. In Panel A, the Herfindahl–Hirshman Index is calculated as the sum of the squared values of the market share of each exchange. In Panel B, the three largest exchanges for volume of shares traded are London, Milan, and Frankfurt/Xetra. The three largest for value of shares traded are London, Frankfurt/Xetra, and Amsterdam. Exchanges also compete for new firm listings (Chemmanur and Fulghieri, 2006). To examine changes in firm listings concurrent with euro conversion, we collect data on total firm listings as of the end of 1998 and 1999 for our exchanges from the World Stock Exchange Fact Book. Interestingly, rankings for changes in listings do not line up consistently with changes in trading volume (see Internet Appendix, Figure A1). The different set of winning and losing exchanges suggests that new listings are driven by factors distinct from those that affect trading activity. Nevertheless, these results also show large shifts in the number of firm listings after euro conversion. 7. Conclusions Euro conversion provides a natural experiment in which to examine the determinants of volume and spread changes when exchanges compete. We find first, that euro conversion lead to a significant round of cost cuts and volume shifts among the European exchanges. Second, shifts in volume were driven by an array of factors, including firm-specific characteristics (size, volatility, and industry), and country-level measures such as GDP growth, accounting standards, and change in the number of analysts. Several of these factors are linked to transparency, with turnover increasing more for less transparent firms. These results suggest that euro conversion triggered competition by changing price transparency. Changes in spreads are driven by these factors, plus exchange-level characteristics such as number of listed firms and the presence of automatic execution. Third, the competition we document spilled over to non-euro European exchanges. In sum, we show that even a seemingly trivial change—the redenomination of prices—can trigger a significant round of price cutting and volume shifts among exchanges. The large absolute and relative volume shifts among exchanges that we document confirm the predictions of many models that the removal of investment barriers can create new opportunities and threats for exchanges in their competition for order flow. Footnotes * We thank Hank Bessembinder, Ran Duchin, Jarrad Harford, Alan Hess, Pankaj Jain, Walter Novaes, Sofia Ramos, Ed Rice, Stephan Siegel, Ernst-Ludwig von Thadden, Thorsten Beck (the editor), the referees, and seminar participants at Boston College and at the University of Washington Finance brown bag seminar for helpful comments, the UW Center for International Business Education and Research for financial support, and Jon Kalodimos for valuable research assistance. Dewenter thanks the Joshua Green Family Professorship. Koski thanks the Kirby L. Cramer Endowed Chair in Finance. 1 Consistent with models that assume exchanges compete to attract trading volume (e.g., Ramos and von Thadden, 2008; Santos and Scheinkman, 2001), we define winners (losers) as the exchanges that win (lose) volume. 2 Arnold et al. (1999) provide evidence on spillovers in competition within a country, the USA. 3 Two exceptions are Arnold et al. (1999), who analyze competition across the NYSE, Amex, and multiple regional exchanges, and Garvey (1944), who looks at a series of exchanges created to compete with the NYSE during its early years. 4 Both the contagion and herding literatures provide guidance as to how competition spurred by euro conversion could spill over to nearby exchanges. The contagion literature primarily focuses on how crises move from one country to another (see Dornbusch, Park, and Claessens (2000) for a review of the contagion literature). Positive economic effects from euro conversion on firms in the euro countries could spill over to the other European countries through trade or financial links. The herding literature, see Devenow and Welch (1996) for a summary, suggests that just as a loss of confidence could lead to investment outflows across multiple countries, a jump in confidence from euro conversion might positively affect investment into all European exchanges. 5 NYSE firms also match the distribution of our European firms more closely based on firm size and industry composition (see Table A2 in the Internet Appendix). 6 A firm listed on a specific exchange is a “firm list.” One sample firm could be listed on two exchanges, or on both Frankfurt and Xetra. In both cases, we count it as two firm lists. We note that for several exchanges the number of firm lists in our initial dataset in column 3 of Table I is larger than the total number of listed stocks on the exchange from column 1. This difference is likely due to timing. The total number of listed stocks on the exchange is as of December 31 1998. The initial dataset includes any firm with at least one ask observation between October 1 1998 and December 31 1999. New listings during 1999 will not be included in the total number of listed stocks as of December 31 1998, but will survive the initial data screen. These firms will not, however, be included in our final sample, which requires at least fifteen non-missing observations for each of our sample quarters. We also note that for Lisbon and Paris, the subset market capitalization value in column 6 is larger than the total market capitalization value in column 2. We are not able to reconcile this discrepancy and note that the two sets of data are from different sources. 7 Only fifty-six firms on Frankfurt pass our data screens using DataStream. This compares to 183 Frankfurt firms in our final sample using KIT data. Results corresponding to Table II, using DataStream data for Frankfurt and Xetra are available in Table A3 of the Internet Appendix. 8 For example, Boomfield and O’Hara (2000) conduct lab experiments of their game theory model of competition between transparent and non-transparent market makers, where transparency refers to the speed with which trade information is revealed. In their lab experiments, quotes and dealer-selected levels of transparency evolve over multiple rounds as participants learn and respond to competitor moves. 9 See Corhay, Hawawini, and Michel (1987) and Das and Rao (2012) among others for studies of seasonalities in European equity markets. 10 About a third of the European firms in our sample fall in the smallest two size deciles based on NYSE cutoffs. Therefore, although we include a screen that eliminates the very smallest European firms (those with market capitalizations less than 50 million euros), our sample still includes many firms that are small relative to NYSE size deciles (see Panel A of Table A2 in the Internet Appendix). 11 In 2011, Frankfurt eliminated floor trading, so now all trading for the Frankfurt Stock Exchange takes place on Xetra. For additional information, see http://deutsche-boerse.com/dbg/dispatch/en/kir/dbg_nav/about_us/10_Deutsche_Boerse_Group/50_Company_History?horizontal=page5_DB_History_2010-heute_ . 12 For firms that are still listed, DataStream provides complete cross-listing information. However, DataStream only provides cross-listing information for about 10% of firms that are no longer listed. See Table A4 in the Internet Appendix for more details about the number and trading volume of cross-listed firms in our sample. 13 For robustness, we also estimate the Amihud (2002) illiquidity measure. Inferences based on the Amihud measure are essentially identical to those based on percentage and effective spreads, and are available in Internet AppendixTable A5. 14 We report results corresponding to Table II, equally weighting each firm, in Internet Appendix, Table A6. z-statistics for a nonparametric sign test are calculated as (Number of Firm Lists with Positive Changes – 0.5 × N)/ N/4 ⁠, where N stands for the total number of firm lists. For robustness, the Internet AppendixTable A7 also provides comparable analyses to those in Table II for short-run changes from 1998Q4 to 1999Q1. 15 Despite the fact that the value-weighted mean for the Overall (Europe) sample in Panel B declines from 0.017 to 0.014, the mean percentage change is (insignificantly) positive. The mean change is driven by a small number of extremely large positive changes. The median change in effective spreads is –17%. In our multivariate analysis, we winsorize these variables to mitigate the effect of outliers. 16 Our sample selection criteria require a firm to have at least 15 non-missing trading volume and quote observations during both 1998Q4 and 1999Q4. The sample of firms we analyze is constant across periods, so the changes in volume we document in our analyses are not attributable to changes in the number of firms due to mergers and acquisitions, delistings, or new listings. Untabulated t-statistics on the two (unbalanced) samples show that differences in all of our spread and trading activity measures differ significantly between the Overall(Europe) and the NYSE. 17 We do not have data for 1997, so we compare changes for the three quarters prior to our event (from 1998Q1 through 1998Q4) with changes over a comparable length of time during our event window (from 1998Q4 through 1999Q3). Please see the Internet AppendixTable A8. 18 The following EU countries adopted the euro after our event date: Greece (2001), Slovenia (2007), Cyprus (2008), Malta (2008), Slovakia (2009), Estonia (2011), Latvia (2014), and Lithuania (2015). Of these, only Greece has a nontrivial number of firms with adequate coverage on DataStream to pass our data screens. There were also several non-EU adopters, but there was no coverage on DataStream for companies in these other countries. 19 We also examine long-run changes, from 1998Q4 through 2004Q4, for the subset of firms with trading in both periods. Please see Internet AppendixTable A11. Inferences are similar. 20 For the European sample, we collect the Industry Classification Benchmark (ICB) Industry Codes from DataStream, which classify all the firms into ten industries. The ICB is a company classification system developed by Dow Jones and FTSE, and is used globally. For the NYSE sample, we collect the 4-digit Standard Industrial Classification (SIC) code from CRSP. We first classify companies into one of the forty-eight industries using the Fama-French forty-eight industry classification by SIC code. Then we further group the forty-eight industries into the ten ICB industries. 21 Relative tick size for a single firm is defined as the minimum tick over the quarter divided by the average transaction price during the quarter. To control for changes in price levels over time, we also divide the minimum tick during 1999Q4 by the average transaction price during 1998Q4 (adjusted for currency differences). See Internet Appendix, Table A12. We control for tick size effects in the multivariate analysis in the next section. 22 We considered using the percentage change from 1998Q1 to 1998Q4 as an instrument, but these instruments do not satisfy the inclusion and exclusion restrictions (see Internet Appendix, Table A13). Nevertheless, the European dummies in the second stage of this specification are similar to those we report in the paper. We choose to report results based on change in turnover to control for changes in the number of shares outstanding. Results are substantially similar when we use change in volume or change in effective spreads as our dependent variables (see Internet AppendixTable A14), although the European dummy is not significant in the second stage when we use change in volume as the dependent variable. 23 Dollar values for the NYSE firms are converted to euros using the exchange rate on January 4 1999. 24 We use DataStream code GGISN, which identifies the firm’s country of incorporation, to classify firms as Foreign. About 12% of the sample firms have Foreign = 1. This subset includes firms headquartered outside of the EU, for example US firms, and EU firms listed outside their home market. For robustness, we repeat the univariate analyses from Tables II and III for the subset of firms with Foreign = 1, and inferences are substantially similar. We also re-estimate our 2SLS regressions interacting the Foreign dummy with all other explanatory variables in the second stage regression. We find that all inferences hold. See Table A15 of the Internet Appendix for descriptive statistics and results relating to Foreign firms. 25 We note that the coefficient on the squared value of fitted change in turnover is significant in column (7), with change in spread as the dependent variable. This result suggests that change in turnover has a weak incremental effect where at higher increases in turnover, the drop in spreads is dampened. 26 These results are consistent with Eleswarapu and Venkataraman (2006) who find bigger spreads for larger firms and lower levels of spreads for firms with higher prices. 27 Halling et al. (2008), in a cross listing paper, show that the share of foreign/domestic trading is higher for smaller firms, firms with more volatility and high tech firms. On the other hand, they also show that trading on the domestic market is positively related to firm size and not related to volatility or being in the high tech industry. Kwan, Masulis, R., and McInish (2015) show that trading market share across exchanges is negatively related changes in volatility. 28 Although volatility is another potential proxy for transparency, it may also proxy for other factors such as risk. Therefore, we do not consider it a clean proxy for asymmetric information. 29 Eleswarapu and Venkataraman (2006) show that the level of spreads is related to macro-level institutions, while Lo (2013) shows that a ranking of the world’s forty-five largest exchanges in terms of “trading competition,” including turnover, is strongly affected by whether or not country-level factors are included in the analysis. 30 See Rossi and Volpin (2004), Erel, Liao, and Weisbach (2012), and Ferreira, Massa, and Matos (2010) for recent papers on cross border mergers and acquisitions; see La Porta et al. (1998) and Djankov et al. (2008) for law and finance papers; see Halling et al. (2008), Halling, Moulton, and Panayides (2013), and Baruch Karolyi, and Lemmon (2007) for multi-market trading papers; see Kaminsky and Reinhart (2000), Karolyi and Stulz (1996), and King and Wadhwani (1990) for representative contagion papers. 31 The transparency results are consistent with Eleswarapu and Venkataraman (2006). The market correlation results are inconsistent with Ramos and von Thadden’s (2008) model of exchange competition which predicts a negative relation. The automatic execution result is consistent with Jain (2006), but the market maker and centralization results are inconsistent. Our results may vary from Jain as he examines an even broader range of countries. The lack of significance for insider trading is consistent with Eleswarapu and Venkataraman (2006), but inconsistent with Halling et al. (2008) and Jain (2005, 2006). One explanation why we have no significance for insider trading enforcement is that all of our exchanges except for one had enforced insider trading by 1998. 32 The lack of any significant effects for exchange characteristics for turnover is inconsistent with Jain (2006, effects for centralization) and Jain (2005, for electronic trading). Our results may vary from Jain because of the different set of exchanges examined. His results show that the level of market development affects the level of turnover. Prior competition papers have not related volume changes to the local economy’s growth. 33 Both CIFAR and Disclosure are from Bushman, Piotroski, and Smith (2004). CIFAR equals the average number of ninety accounting and non-accounting items disclosed by a sample of large companies. Disclosure is based on the prevalence of disclosures for research and development expenses, capital expenditures, product and geographic segment data, subsidiary information and accounting methods. 34 As noted, we also estimate these regressions for Change in Analysts and Disclosure. Change in Analysts results are mixed (see Internet AppendixTable A19), suggesting it is the starting level of transparency rather than the change in transparency that determines winners and losers. Most of the countries in our sample have very high levels of Disclosure relative to the full cross-section presented in Bushman, Piotroski, and Smith (2004), so this variable does not seem to capture economically meaningful variation in transparency for our sample. 35 We multiply by one-half, because the effective spread is a measure of round-trip transaction costs, and each individual trade is only one way. 36 Corresponding savings for one quarter based on percentage spreads (rather than effective spreads) are €1.01 billion. These values represent the savings to investors who traded in the firms in our sample. To obtain a more general estimate applicable to all firms on these exchanges, we repeat this exercise, applying 1998 and 1999 effective spreads to the total 1999 volume on each exchange (in dollars, obtained from the World Stock Exchange Fact Book, 2004). Based on this calculation, investors saved over $5 billion in trading costs associated with the reduction in effective spreads. 37 In Internet Appendix, Table A21 we report the welfare savings in trading costs resulting from changes in effective spreads for European firms controlling for changes in effective spreads in matched NYSE firms over the same time period. These savings are even larger, at €743 (or 23.5%). This result is not surprising given our univariate results reported in Tables II and III. Spreads generally increased or decreased less on the NYSE than they did in Europe after euro adoption. 38 For example, this ratio is included in the World Bank’s Global Financial Development dataset. Market capitalization comes from the World Stock Exchange Fact Book over this period, and is estimated for domestic companies only. GDP data comes from the Organization for Economic Co-operation and Development (OECD). 39 There was also some discussion of removing the stamp duty on share trades in London. 40 For each quarter, we pool all firms and calculate the standard deviation of percentage and effective spreads. Across all firms, the standard deviation of percentage spreads falls from 0.028 in 1998Q4 to 0.023 in 1999Q4. The comparable numbers for effective spreads are 0.020 and 0.016. These changes are highly statistically significant. 41 See Adjaoute and Danthine (2004, equity premiums), Baele et al. (2004, government bond yield spreads), and Bekaert et al. (2013, average bilateral earnings yield differentials) who all argue that the economic and financial integration brought about by the European Monetary Union should be associated with a convergence in prices. None of the trends in their papers show a distinct change around the euro conversion event in January 1999. Indeed, Bekaert et al. test for this and find an insignificant effect of euro introduction on value differentials. References Adjaoute K. , Danthine J.-P. ( 2004 ) Equity returns and integration: is Europe changing?, Oxford Review of Economic Policy 20 , 555 – 570 . Google Scholar Crossref Search ADS Admati A. , Pfleiderer P. ( 1988 ) A theory of intraday patterns: volume and price variability , Review of Financial Studies 1 , 3 – 40 . Google Scholar Crossref Search ADS Aggarwal R. , Angel J. ( 1999 ) The rise and fall of the Amex Emerging Company Marketplace , Journal of Financial Economics 52 , 257 – 289 . Google Scholar Crossref Search ADS Amihud Y. ( 2002 ) Illiquidity and stock returns: cross-section and time-series effects , Journal of Financial Markets 5 , 31 – 56 . Google Scholar Crossref Search ADS Amira K. , Muzere M. ( 2011 ) Competition among stock exchanges for equity , Journal of Banking and Finance 35 , 2355 – 2373 . Google Scholar Crossref Search ADS Arnold T. , Hersch P. , Mulherin J. H. , Netter J. ( 1999 ) Merging markets , Journal of Finance 54 , 1083 – 1107 . Google Scholar Crossref Search ADS Baele L. , Ferrando A. , Hordahl P. , Krylova E. , Monnet C. ( 2004 ) Measuring European financial integration , Oxford Review of Economic Policy 20 , 509 – 530 . Google Scholar Crossref Search ADS Baruch S. , Karolyi G. A. , Lemmon M. ( 2007 ) Multimarket trading and liquidity: theory and evidence , Journal of Finance 62 , 2169 – 2200 . Google Scholar Crossref Search ADS Battalio R. , Greene J. , Jennings R. ( 1997 ) Do competing specialists and preferencing dealers affect market quality? , Review of Financial Studies 10 , 969 – 993 . Google Scholar Crossref Search ADS Bekaert G. , Harvey C. , Lundblad C. , Siegel S. ( 2013 ) The European Union, the euro, and equity market integration , Journal of Financial Economics 109 , 583 – 603 . Google Scholar Crossref Search ADS Bessembinder H. ( 2003 ) Trade execution costs and market quality after decimalization , Journal of Financial and Quantitative Analysis 38 , 747 – 777 . Google Scholar Crossref Search ADS Bhattacharya U. , Daouk H. ( 2002 ) The world price of insider trading , Journal of Finance 57 , 75 – 108 . Google Scholar Crossref Search ADS Bloomfield R. , O’Hara M. ( 2000 ) Can transparent markets survive? , Journal of Financial Economics 55 , 425 – 459 . Google Scholar Crossref Search ADS Blume M. , Goldstein M. ( 1997 ) Quotes, order flow and price discovery , Journal of Finance 52 , 221 – 244 . Google Scholar Crossref Search ADS Boland V. ( December 18 1998 ) Three more countries to join Euro NM, Financial Times. Brown W. , Mulherin J. H. , Weidenmier M. ( 2008 ) Competing with the New York Stock Exchange , Quarterly Journal of Economics 123 , 1679 – 1719 . Google Scholar Crossref Search ADS Bushman R. , Piotroski J. , Smith A. ( 2004 ) What determines corporate transparency? , Journal of Accounting Research 42 , 207 – 252 . Google Scholar Crossref Search ADS Chemmanur T. , Fulghieri P. ( 2006 ) Competition and cooperation among exchanges: a theory of cross-listing and endogenous listing standards , Journal of Financial Economics 82 , 455 – 489 . Google Scholar Crossref Search ADS Chowdhry B. , Nanda V. ( 1991 ) Multimarket trading and market liquidity , Review of Financial Studies 4 , 483 – 511 . Google Scholar Crossref Search ADS Corhay A. , Hawawini G. , Michel P. ( 1987 ) Seasonality in the risk-return relationship: some international evidence , Journal of Finance 42 , 49 – 68 . Google Scholar Crossref Search ADS Das P. , Rao S. P. ( 2012 ) Is the value effect seasonal? Evidence from global equity markets , International Journal of Business and Finance Research 6 , 21 – 33 . de Fontnouvelle P. , Fishe R. , Harris J. ( 2003 ) The behavior of bid-ask spreads and volume in options markets during the competition for listings in 1999 , Journal of Finance 58 , 2437 – 2453 . Google Scholar Crossref Search ADS Devenow A. , Welch I. ( 1996 ) Rational herding in financial economics, European Economic Review 40 , 603 – 615 . Google Scholar Crossref Search ADS Dewenter K. , Kim C.-S. , Novaes W. ( 2010 ) Anatomy of a regulatory race to the top: changes in delisting rules at Korea’s two stock exchanges, 1999–2002 , Journal of Corporate Finance 16 , 456 – 468 . Google Scholar Crossref Search ADS Di Noia C. ( 2001 ) Competition and integration among stock exchanges in Europe: network effects, implicit mergers and remote access , European Financial Management 7 , 39 – 72 . Google Scholar Crossref Search ADS Djankov S. , La Porta R. , Lopez-de-Silanes F. , Shleifer A. ( 2008 ) The law and economics of self-dealing , Journal of Financial Economics 88 , 430 – 762 . Google Scholar Crossref Search ADS Domowitz I. , Glen J. , Madhavan A. ( 1998 ) International cross-listing and order flow migration: evidence from an emerging market , Journal of Finance 53 , 2001 – 2027 . Google Scholar Crossref Search ADS Domowitz I. , Glen J. , Madhavan A. ( 2001 ) Liquidity, volatility and equity trading costs across countries and over time , International Finance 4 , 221 – 255 . Google Scholar Crossref Search ADS Dornbusch R. , Park Y. C. , Claessens S. ( 2000 ) Contagion: understanding how it spreads , The World Bank Research Observer 15 , 177 – 197 . Google Scholar Crossref Search ADS Eleswarapu V. , Venkataraman K. ( 2006 ) The impact of legal and political institutions on equity trading costs: a cross-country analysis , Review of Financial Studies 19 , 1081 – 1111 . Google Scholar Crossref Search ADS Erel I. , Liao R. , Weisbach M. ( 2012 ) Determinants of cross-border mergers and acquisitions , Journal of Finance 67 , 1045 – 1082 . Google Scholar Crossref Search ADS Ferreira M. , Massa M. , Matos P. ( 2010 ) Shareholders at the gate? Institutional investors and cross-border mergers and acquisitions , Review of Financial Studies 23 , 601 – 644 . Google Scholar Crossref Search ADS Foerster S. , Karolyi G. A. ( 1998 ) Multimarket trading and liquidity: a transaction data analysis of Canada–US interlistings , Journal of International Financial Markets 8 , 393 – 412 . Google Scholar Crossref Search ADS Foucault T. , Parlour C. ( 2004 ) Competition for listing , Rand Journal of Economics 35 , 329 – 355 . Google Scholar Crossref Search ADS Frost C. A. , Gordon E. , Hayes A. ( 2006 ) Stock exchange disclosure and market development: an analysis of 50 international exchanges , Journal of Accounting Research 44 , 437 – 483 . Google Scholar Crossref Search ADS Garvey G. ( 1944 ) Rivals and interlopers in the history of the New York Security Market , Journal of Political Economy 52 , 128 – 143 . Google Scholar Crossref Search ADS Gehrig T. ( 1998 ) Competing markets , European Economic Review 42 , 277 – 310 . Google Scholar Crossref Search ADS Glosten L. , Milgrom P. ( 1985 ) Bid, ask and transaction prices in a specialist market with heterogeneously informed traders , Journal of Financial Economics 14 , 71 – 100 . Google Scholar Crossref Search ADS Halling M. , Moulton P. , Panayides M. ( 2013 ) Volume dynamics and multimarket trading , Journal of Financial and Quantitative Analysis 48 , 489 – 518 . Google Scholar Crossref Search ADS Halling M. , Pagano M. , Randl O. , Zechner J. ( 2008 ) Where is the market? Evidence from cross-listings in the United States , Review of Financial Studies 21 , 725 – 761 . Google Scholar Crossref Search ADS Harris L. ( 1994 ) Minimum price variations, discrete bid-ask spreads, and quotation sizes , Review of Financial Studies 7 , 149 – 178 . Google Scholar Crossref Search ADS Huddart S. , Hughes J. , Brunnermeier M. ( 1999 ) Disclosure requirements and stock exchange listing choice in an international context , Journal of Accounting and Economics 26 , 237 – 269 . Google Scholar Crossref Search ADS Jain P. ( 2005 ) Financial market design and the equity premium: electronic versus floor trading , Journal of Finance 60 , 2955 – 2985 . Google Scholar Crossref Search ADS Jain P. ( 2006 ) Improving liquidity through efficient stock market structure and operational design , Journal of Financial Transformation 18 , 151 – 159 . Kaminsky G. , Reinhart C. ( 2000 ) On crises, contagion, and confusion , Journal of International Economics 51 , 145 – 168 . Google Scholar Crossref Search ADS Karolyi G. A. , Stulz R. ( 1996 ) Why do markets move together? An investigation of US–Japan stock return comovements , Journal of Finance 51 , 951 – 986 . Google Scholar Crossref Search ADS Karpoff J. ( 1987 ) The relation between price changes and trading volume: a survey , Journal of Financial and Quantitative Analysis 22 , 109 – 126 . Google Scholar Crossref Search ADS King M. , Wadhwani S. ( 1990 ) Transmission of volatility between stock markets , Review of Financial Studies 3 , 5 – 33 . Google Scholar Crossref Search ADS Klock M. , McCormick T. ( 1999 ) The impact of market maker competition on Nasdaq spreads , Financial Review 34 , 55 – 74 . Google Scholar Crossref Search ADS Kodres L. , Pritsker M. ( 2002 ) A rational expectations model of financial contagion , Journal of Finance 57 , 768 – 799 . Google Scholar Crossref Search ADS Kwan A. , Masulis R. , McInish T. ( 2015 ) Trading rules, competition for order flow and market fragmentation , Journal of Financial Economics 115 , 330 – 348 . Google Scholar Crossref Search ADS La Porta R. , Lopez-de-Silanes F. , Shleifer A. , Vishny R. ( 1998 ) Law and finance , Journal of Political Economy 106 , 1113 – 1155 . Google Scholar Crossref Search ADS Lamoureux C. , Schnitzlein C. ( 1997 ) When it’s not the only game in town: the effect of bilateral search on the quality of a dealer market , Journal of Finance 52 , 683 – 712 . Google Scholar Crossref Search ADS Lo S.-F. ( 2013 ) Which stock exchanges are more attractive? The competition analysis of listing and trading performance , Economic Modelling 30 , 501 – 509 . Google Scholar Crossref Search ADS Madhavan A. ( 1995 ) Consolidation, fragmentation, and the disclosure of trading information , Review of Financial Studies 8 , 579 – 603 . Google Scholar Crossref Search ADS Pagano M. ( 1989 ) Trading volume and asset liquidity , The Quarterly Journal of Economics 104 , 255 – 274 . Google Scholar Crossref Search ADS Pagano M. , Roell A. ( 1990 ) Trading systems in European stock exchanges: current performance and policy options , Economic Policy 5 , 63 – 115 . Google Scholar Crossref Search ADS Pirrong C. ( 2000 ) A theory of financial exchange organization , Journal of Law and Economics 43 , 437 – 472 . Google Scholar Crossref Search ADS Ramos S. ( 2003 ) Competition between stock exchanges: a survey, International Center for Financial Asset Management and Engineering Research Paper No. 77. Ramos S. , von Thadden E.-L. ( 2008 ) Stock exchange competition in a simple model of capital market equilibrium , Journal of Financial Markets 11 , 284 – 307 . Google Scholar Crossref Search ADS Reuters News ( November 25 1998 ) Danish share trade tax to remain until October 1 1999. Rossi S. , Volpin P. ( 2004 ) Cross-country determinants of mergers and acquisitions , Journal of Financial Economics 74 , 277 – 304 . Google Scholar Crossref Search ADS Santos T. , Scheinkman J. ( 2001 ) Competition among exchanges , Quarterly Journal of Economics 116 , 1027 – 1061 . Google Scholar Crossref Search ADS Spankowski U. , Wagener M. , Burghof H.-P. ( 2012 ) The role of traditional exchanges in fragmented markets, unpublished working paper, University of Hohenheim, Stuttgart Stock Exchange. Stigler G. ( 1961 ) The economics of information , Journal of Political Economy 69 , 213 – 225 . Google Scholar Crossref Search ADS Wahal S. ( 1997 ) Entry, exit, market makers, and the bid-ask spread , Review of Financial Studies 10 , 871 – 901 . Google Scholar Crossref Search ADS World Stock Exchange Fact Book, Meridian Securities Markets, Morris Plains NV, Electronic Commerce Inc. Various Issues. © The Authors 2017. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Review of FinanceOxford University Press

Published: Oct 1, 2018

References