Efficiency gains from the integration of exchanges: Lessons from Euronext’s “natural experiment”

1. Efficiency gains from the integration of exchanges:Lessons from Euronext’s “natural experiment” Dr. A. Jorge Padilla LECG Europe www.lecgcp.com Leuven, 7 November 2006

2. The theory The integration of exchanges produces a number of significant efficiency gains: • Cost savings • Eliminates the duplication of costly infrastructure … • … which may lead to a reduction in trading fees • … and brokerage fees • Direct user benefits • Savings on operating and capital costs • Trading more diversified portfolios • Increased cross-border trading … • … leading to increases in liquidity, as reflected by lower bid-ask spreads, greater volume and lower volatility

3. CashTrading integration May 2001 October2001 November2003 Brussels Paris Amsterdam Brussels Paris Lisbon Amsterdam Brussels Paris Euronext’s natural experiment • Integration between the French, Belgian, Dutch and Portuguese stock exchanges to form Euronext (September 2000 – 2003) • “Before and after” analysis on costs and user benefits … • … controlling for confounding factors (i.e., time-variant effects that have nothing to do with integration) Chronology of integration of cash trading business

4. Euronext’s natural experiment • This experiment makes it possible to: • Evaluate the cost savings achieved through the integration process; • Investigate the pass-through of those savings; • Identify other sources of direct user benefits, and • Test the impact of integration on liquidity and, hence, on the implicit trading costs faced by the users of the exchange.

5. Cost savings • Significant reduction in operating costs: • Overall, the total annual costs of Euronext’s continental operations fell by 137 million euros (25%) between 2001 and 2004. • IT cost savings: Euronext’s total continental IT costs fell by 29% between 2001 and 2004. • Headcount reductions: Euronext reduced the staffing levels of its continental operations by 24% between 2001 and 2004. Evolution of continental IT costs following Euronext integration Euronext continental staff numbers 2001-2004

6. Trading fees • The evidence shows that the average trading fee charged in Paris fell by about 30% (in real terms) in the period from December 1999 to December 2004. • Average trading fees also fell in Brussels and Amsterdam. • From January 2002 to December 2004, the average trading fee in Brussels fell by 30%. • From January 2001 to December 2004, the average trading fee in Amsterdam fell approximately 45%. Our econometric results show that those fee reductions were to a large extent the result of the creation of Euronext

7. Direct user benefits • Improved access: • Integration has allowed Euronext members directly to access all the different Euronext markets • The process of integration has expanded the set of securities accessible to a Euronext member. • Investors now benefit from greater inter-broker competition. Share of cross-border trade undertaken by Euronext members (% of total trades of members at each location)

8. Direct user benefits • Members have benefited also from reduced internal operating costs. • Increased liquidity • Lower bid-ask spreads; • Greater volume; • Lower volatility.

9. Panel data estimation • We aim to estimate the impact of integration on liquidity. In order to do so, we have estimated a panel data model that relates liquidity measures with Euronext integration dummies. • Liquidity is measured by: • Volume: the higher the liquidity, the higher the volume. • Bid-ask spread: the higher the liquidity, the lower the spread. • Volatility: the higher the liquidity, the lower the volatility. • Therefore, we have tested whether Euronext integration had a positive impact on volumes and a negative impact on bid-ask spreads, and volatilities. • In this analysis, we assumed that Euronext integration took place in the following dates: • 21-May-2001: Brussels and Paris trading • 29-Oct-2001: Amsterdam, Brussels and Paris trading • 7-Nov-2003: Lisbon, Amsterdam, Brussels and Paris trading

10. Methodology: specification • Liquidity (volume, bid-ask spread and volatility) of security i in period t, or • A dummy variable that takes the value of 1 if the security i is traded in an integrated market in period t and 0 otherwise. • Alternatively, we define three different dummies in order to differentiate the impact of each integrated market: • “Integration Brussels” takes the value of 1 if if the security i is traded in the (at-least) integrated market Paris – Brussels in period t and 0 otherwise. • “Integration Amsterdam” takes the value of 1 if if the security i is traded in the (at-least) integrated market Paris – Brussels – Amsterdam in period t and 0 otherwise. • “Integration Lisbon” takes the value of 1 if if the security i is traded in the fully integrated market (Paris, Brussels, Amsterdam and Lisbon) in period t and 0 otherwise.

11. Specification (continued) • Monthly dummies. • Dummies related to relevant economic events (similar to the ones used in the first stage). • A deterministic time trend. • Other controls (depend on data availability): • In the volume regression: the volume of an index traded in non-integrated markets (FTSE 100 and DAX). • In the volatility regression: the volatility of the index of the own market to net out covariance risk. • Fixed effects to control for differences across securities. • This control is specially important when using data at the security level. Panel data models allow to include a fixed effect per security, therefore, netting out differences across securities.

12. Direct user benefits • Lower bid-ask spreads • The bid-ask spreads of the securities included in the main Paris index fell as a result of the creation of Euronext: approx 40%. • The analysis also shows that integration led to a reduction of the bid-ask spreads of the securities in the main indices of Brussels (25%-30%) and Amsterdam (approx. 10%)

13. Liquidity effectsBid-ask spreads (Bloomberg) • Our main findings, using Bloomberg data, are: • In general, Euronext integration had a negative, and statistically significant, impact on bid-ask spread. • Our results show that Brussels, Amsterdam and Lisbon integration had a similar impact on the bid-ask spreads, as measured by Bloomberg.

14. Liquidity effectsBid-ask spreads (Euronext) • Our main findings, using Euronext data, are: • In general, Euronext integration had a negative, and statistically significant, impact on bid-ask spread. • Our results show that Brussels, Amsterdam and Lisbon integration had a similar impact on the bid-ask spreads, as measured by Bloomberg.

15. Direct user benefits • Greater volume • Trading volume in Paris, Brussels, and Amsterdam increased as a result of the creation of Euronext. • According to our estimations, the creation of Euronext led to an increase in the traded volume of the main securities listed on the Paris, Brussels and Amsterdam exchanges of approximately 40%.

16. Liquidity effectsVolume

17. Liquidity effectsVolume • Our main findings are: • Euronext integration had a positive, and statistically significant, impact on volume (defined as number of shares traded). • These results are robust to different specifications of the panel data model, in particular when including the volume of an index traded in non-integrated markets (FTSE 100 and DAX) as control variables. • Results are also robust when defining volume in levels, except that the integration of Brussels is no longer statistically significant.

18. Direct user benefits • Lower volatility • The volatility of the large-cap securities traded in Paris, Brussels, Amsterdam and Lisbon fell as a result of the creation of Euronext. • The reduction in volatility following integration was between 9% and 18% of the initial levels

19. Liquidity effectsVolatility

20. Liquidity effectsVolatility • Our main findings are: • In general, Euronext integration had a negative, and statistically significant, impact on volatility (defined as 20-days volatility) when including the volatility of the index of the own market as a control variable. • Our results show that Amsterdam and Lisbon integration had the highest (negative) impact on volatility, while Brussels integration had no statistically significant impact on volatility.

21. Conclusions • The results of the natural experiment show: • Significant cost savings were achieved as a result of the integration process; • Those savings were passed on in part to users; • Users also enjoyed other benefits: access to more securities, increased brokerage competition, lower transaction costs and, perhaps, most importantly increased liquidity. • The integration of the Amsterdam, Brussels, Lisbon and Paris exchanges in a single platform resulted in a significant increase in liquidity.

22. Jorge Padilla LECG Jpadilla@lecg.com www.lecgcp.com Leuven, 7 November 2006