Pre-trade transparency and trade size

Abstract

We analyse how Pre-Trade Transparency (PTT) affects the behaviour of different stock traders. To do so, we exploit a natural experiment, that is the PTT change in the equity segment of Italian Stock Exchange (ISE) which occurred in July 2007, with the aim of reducing information asymmetries between individuals and intermediaries/institutional investors. We specify a dynamic empirical model for trade size and estimate it on a large panel of tick-by-tick data. Results suggest that increased transparency affects the dynamic trade pattern emerging from interacting strategic decisions of different traders. In addition, the contribution of the order flow disclosure both in reducing the adverse selection component of the bid–ask spread, and in weakening the sensitiveness to risk of the trade size also emerges. Overall, PTT enhancement should reduce the informative disequilibrium among market participants and improve the quality of the market.

Keywords:

• pre-trade transparency
• trade size
• stealth trading
• panel data
• tick-by-tick data

JEL Classification::

• G14
• D82
• C23
• C81

Acknowledgements

Preliminary versions of this article were presented at the Commodity Futures Trading Commission, Washington, DC (August 2009); the third Rimini Finance Workshop (May, 2011); the II World Finance Conference at Rhodes (June, 2011); the 2011 International Finance and Banking Society Conference held in Rome (June 2011). We would like to thank participants, for helpful comments. Special thanks go to Pete Kyle and Philip Molyneux for their insightful discussions of the results. We would also like to thank Enrico Bellazzi, Lloyd Blenman, David Bowman, Celso Brunetti, Roberto Golinelli, Clara Vega and an anonymous referee for their useful comments and advice. Excellent research assistantship was provided by Camilla Mazzoli and AnnaGrazia Quaranta.

We are grateful to Investnet Italia S.p.A (IW Bank) and Townsend Analytics for providing the trading platform used to obtain the proprietary data set; to Massimiliano Ghedini, Ivan Girotto and CINECA (Italian Ministry of Research) for providing the necessary resources to deal with the unusual dimension of the dataset. Finally, we are grateful to BIt for providing us with information and comments on the implication of this research. Any remaining errors are our own.

Notes

¹ The DLTB also affected the bond segment (the 'Mercato delle Obbligazioni Telematico’ (MOT)), the segment of Securitized Derivatives (SeDeX) and the Italian Derivative Market (IDEM) of the ISE. We focus our analysis on the ISE traditional equity segment because it allows us to apply theoretical models referred to equity trading and to compare our results with the existing literature.

² Chelley-Steeley and Li (2005) analyse the effect of a change in trading mechanism on the volatility of security returns in the London Stock Exchange. Evans (2006) examines the impact of the introduction of electronic trading system on the market efficiency of three UK financial futures markets.

³ Fill or kill orders are either immediately and fully matched against existing orders on the order book, or the entire order is cancelled.

⁴ Also named Execute and Eliminate (ENE) orders, these quotes are matched immediately and as fully as possible against the existing orders on the order book. Any volume that cannot execute against existing orders are cancelled.

⁵ These are the rules for modifications of the iceberg order size: if the order size increases, the hidden size increases and the peak size has no impact; if the order size decreases to more than the current remaining peak size, the hidden size decreases and the peak size has no impact; if the order size decreases to less than the current remaining peak size, the hidden size is fully removed and the peak size is reduced.

⁶ Intermediaries and their interconnected professional clients can rely on a transparent full-order book; thanks to the combination of the complimentary information feeds available through data flows used by member firms for the transmission of orders to the market and a feed with desegregated orders up to the fifth level real time. The order book was also not anonymous, at least until 2004. The BIt Order no. 3238 of 20 January 2004, entered into force with effect from 21 January 2004, following the introduction of the central counterparty on the cash markets, which imposed book anonymity as far as the MTA is concerned. However, orders that entered markets without a central counterparty guarantee system still continue to indicate the names of the intermediaries involved.

⁷ Precisely, DDM+ allowed vendors (and so the public) three solutions for the equity market: a 5, 10 or 20 level order book. The fixed cost applied by BIt slightly differs for the vendor in relation to the levels of visible orders he/she chooses to buy, but it is not considerably high. The difference of the fixed cost between the 5 level and the 20 level package is less than 1000 euros per month. Moreover, the final users of the data-feed (i.e. on-line traders) generally received the 20-level order book for free, especially if they were heavy traders.

⁸ There is no ‘starting day’ for the introduction of DLTB. BIt informed us that the migration of the platforms of market data vendor to DDMplus took place between July and August 2007 (a smooth migration period).

⁹ Another possibility would be the Probability of Informed Trading (PIN) model of Focault et al. (2007), which assumes a change in the visibility of market participants; however, given that DLTB did not involve a change in anonymity, we prefer the trade size model.

¹⁰ In a continuous auction market (such as the BIt one), the use of ‘market maker’ refers to a whoever limit order placing speculator.

¹¹ The size of iceberg orders is reasonably not higher or lower than the sizes ‘ordinarily’ used to quote the share because informed traders have to manage the trade-off between hiding the size of their informed deal and controlling for trading fees (a higher number of small size transactions means higher transaction costs) (see also Chakravarty, 2001).

¹² On this point, see also Ciner and Sackley (2007) for the emerging market of Taiwan.

¹³ To make the results comparable, the risk indicator measured using the Garman–Klass Range is multiplied by 100.

¹⁴ We compute backwardation by comparing two interest rate futures of different maturities and do not use the traditional formula – i.e. computing the difference between the spot and the theoretical future price because we need an intraday variable (short-term interest rates in the cash market are based on daily time frames). Details are available upon request.

¹⁵ For example, Rossi and Santucci De Magistris (2009) use this approach to estimate a bivariate model for the volumes and realized volatility.

¹⁶ In addition, the efficient full information approach needs the strong assumption of correct specification of all model's equations, while with IV we can only focus on the relationship of interest.

¹⁷ Shocks to trade, , cannot feedback to instrument realizations as the latter refer to the past. In this way, the direction of causality from each instrumented variable, rbas_i,k,t , R_i,k,t and oi_i,k,t , to trade size, ts_i,k,t , is identified, and its parameter is consistently estimated.

¹⁸ Aktas (2011) shows that macroeconomic announcements are related to the overreaction by market participants.

¹⁹ During mid-August, the sub-prime crisis started to spread its effects globally (Bank for International Settlements, 2009; Brunnermeier, 2009).

²⁰ Even if less precise, similar patterns moving from BOOK1 to BOOK2/BOOK3, or from BOOK2 to BOOK3, are observed in Model_gkr, 150-minute frame and in Model_gkr, 30-minute frame.