Dark trading in Australia Carole Comerton-Forde. Platypus Symposium 12 March 2013

Transcription

1 Dark trading in Australia Carole Comerton-Forde Platypus Symposium 12 March 2013

2 Overview What is dark trading? Why are regulators concerned about it? Dark trading and price discovery research Research question Key findings Empirical approach & results Conclusions & policy implications Recent regulatory changes 2

3 What is dark trading? Any trade that occurs without pre-trade transparency Generally captures: Negotiated block trades Broker internalization and client matching Dark pools Hidden orders on exchange order books 3

4 Dark trading in Australia Exceptions to pre-trade transparency: Block and portfolio crossings Trades at or within the spread New facilities introduced by ASX in June 2010: CentrePoint (separate dark order book) CentrePoint priority crossings Chi-X hidden orders ($20k limit removed May 2012) 18 broker crossing systems launched since 2005 Significant changes in the way dark trades are executed 4

5 Total dark 21% Below block dark 14% (10% of trades) Block dark 7.0% (0.1% of trades) 3 crossing syst. exist (UBS, Citi, CreditSuisse) Liquidnet Aus enters 11 other crossing syst. commence during sample ASX Centre Point commences Chi-X enters and fragments lit liquidity 5

6 % below block dark by stock size Small (1) Large (4) 6

7 Regulatory concerns Dark trading offers benefits to investors: Historically offered block liquidity Reduced market impact costs and information leakages Potential time priority for some clients Reduced exchange fees BUT 7

8 Regulatory concerns Dark trading also creates risks: Reduced liquidity and reduced incentives to display liquidity Fragmentation Increased bid ask spreads Less efficient price discovery process Lack of transparency about execution process Reduced monitoring and surveillance Unlevel playing field between market operators and crossing system operators 8

9 Dark trading and price discovery Joint research with Talis Putnins at UTS Our research question: What is the impact of dark trading on price discovery? 9

10 Key findings Aggregate price discovery impeded and prices become less informationally efficient as order flow migrates to the dark Order flow that migrates to the dark is less informed than that which remains lit Increase in adverse selection risk, bid ask spread and price impact on lit market No evidence that block trades harm price discovery 10

11 Sample All Ords index stocks 1 February 2008 to 30 October 2011 Normal trading hours Use flags to classify trades as: Lit - central limit order book Below block dark - priority crossings, Centre Point trades, Centre Point priority crossings Block dark - block specials and portfolio specials 11

12 Empirical strategy Estimate price discovery measures (stock-date level) Aggregate price discovery metrics (HF informational efficiency proxies) Price discovery shares (trades vs midquotes; lit vs dark trades) Informativeness of different types of order flow Spreads/adverse selection Relate price discovery measures via stock-date panel regressions to: Non-block dark share of $volume Block dark share of $volume 12

13 Aggregate price discovery Efficient prices will follow a random walk: Autocorrelation of mid-quote returns Variance ratios Return predictability 13

14 Aggregate price discovery Non-block dark Block dark 1. Increasing share of non-block dark trading deterioration of informational efficiency 2. Not true for block dark trades: improvement in 2 of the 3 informational efficiency measures 14

15 Non-block dark and price discovery 15

16 Block dark and price discovery 16

17 Price discovery shares Hasbrouck (1995) information share (IS) Measures relative speed at which innovations in fundamental value are reflected AND relative avoidance of noise Information leadership share (ILS) adapted from Yan and Zivot (2010) Measures relative speed at which innovations in fundamental value are reflected ( who moves first in price discovery) 17

18 Price discovery shares Estimate price discovery shares for: Trade prices vs. mid-quotes; and lit trade prices vs. dark trade prices Allows us to assess importance of pre-trade vs. post-trade transparency 18

19 Price discovery shares Mid-quotes impound slightly more information than trade prices Mid-quotes less noisy than trade prices (avoid bid-ask bounce) Lit trades impound more information than dark trades Lit trades are noisier than dark trades (possibly because dark trades can occur within the spread) 19

20 Price discovery share regressions Increases in non-block dark trading leads to: Mid-quotes becoming faster in reflecting innovations in fundamentals Trade prices become less noisy compared to mid-quotes Increases in block dark trading has no impact on midquotes vs. trades Increases in non-block and block dark trading decrease the contribution of lit trade prices (relative to dark prices) to price discovery Increases the role of dark trades in impounding new information Loss of pre-trade information may result in less than full information being impounded in prices 20

21 Conclusions Order flow that migrates to dark is less informed than order flow left behind, but is not entirely uninformative As order flow migrates to the dark: Aggregate price discovery (informational efficiency) deteriorates Contribution of quotes to price discovery increases (relative to trades) Contribution of dark trades to price discovery increases (relative to lit) Price impact, adverse selection and spreads increase in lit market 21

22 Conclusions Some dark trading is beneficial but too much is harmful Not all dark trading is the same i.e. blocks less harmful Regulation needs to: carefully consider differences in types of dark trading carefully consider how to avoid harmful levels of dark trading 22

23 New dark trading rules Effective 26 May, 2013 New block trade thresholds Trade below block size will need to offer economically meaningful price improvement Dark trading and HFT task forces reviewing conduct and need for further regulations to ensure market integrity and liquidity Treasury review of market licensing 23

24 Other market structure changes New extreme trading rules in cases of large price movements (Nov 2012) Direct control over trading algorithms, including 'kill switches' to immediately stop an algorithm if required (June 2014) Additional data reporting requirements to assist ASIC in performing market surveillance (Nov 2013) 24