«by Robert Bloomfield, Maureen O’Hara, and Gideon Saar* June 2012 *Robert Bloomfield (rjb9 Maureen O’Hara (mo19 and ...»
Some new light on dark trading
Robert Bloomfield, Maureen O’Hara, and Gideon Saar*
*Robert Bloomfield (firstname.lastname@example.org), Maureen O’Hara (email@example.com), and
Gideon Saar (firstname.lastname@example.org) are from the Johnson Graduate School of Management,
Cornell University. We thank Alyssa Andersen for valuable research assistance. We
would like to thank Michael Goldstein, Andrew Karolyi, Yelena Larkin, and seminar
participants at Cornell University, FINRA, University of Notre Dame & NASDAQ OMX Conference on Current Topics in Financial Regulation, and the University of Illinois for helpful comments. We thank the Notre Dame finance faculty and NERA Economic Consulting for awarding us the Best Paper Prize at the Conference on Current Topics in Financial Regulations.
Abstract We use a laboratory market to investigate how the ability to hide orders affects traders’ strategies and market outcomes. We examine three market structures: Visible markets in which all orders must be displayed, Iceberg markets in which a minimum size must be displayed, and Hidden markets in which orders can be displayed, partially displayed, or completely non-displayed. We find that although order strategies are greatly affected by allowing hidden liquidity, most market outcomes are not. Our results on the robustness of informational efficiency and liquidity to opacity regimes have important regulatory implications for debates surrounding dark trading. We also find that opacity appears to increase the profits of informed traders but only when their private information is very valuable.
Hidden Liquidity: Some new light on dark trading
1. Introduction Hidden liquidity is now a standard feature of trading in equity markets. Virtually all exchanges allow traders to “hide” all or a portion of their orders on the book, resulting in market liquidity having both a displayed and a non-displayed component. Although non-displayed orders generally lose priority to displayed orders at a given price, the invisibility of these orders can be valuable for a variety of trading strategies. With orders hidden, however, market participants have only incomplete knowledge as to the overall depth in the market. Moreover, the ability to put hidden orders inside the displayed spread means that even the best prevailing prices are not observable. This evolution to dark trading in exchange markets has gained momentum in recent years, driven in part by the rise of crossing networks (which also allow traders to hide their trading intentions) and by competitive pressures from new exchanges and trading platforms. Despite the 1975 Congressional mandate that U.S. equity markets be transparent, the reality is that markets are becoming increasingly opaque.
Regulators both here and abroad are questioning the role that hidden liquidity plays in markets.1 Much of this regulatory scrutiny has focused on “dark pools” or crossing networks, but the hidden liquidity in exchange settings is actually of comparable or greater importance, with estimates of approximately 20% of marketable orders executing against non-displayed depth in U.S. markets.2 Advocates argue that hidden orders enhance market performance by helping traders shield their trading intentions from the predations of opportunistic traders.
Critics counter that these advantages to individual traders come at the expense of the market as whole by degrading the liquidity and informational efficiency of the market. Virtually everyone agrees, however, that the existence of hidden orders has increased uncertainty about the level of liquidity in the market and led to a variety of complications such as pinging (i.e., placing and
The SEC and the Ontario Securities Commission (the main Canadian regulator) have introduced new rules for dark trading, while European regulators are debating transparency rules in a revised MiFid framework.
Hasbrouck and Saar (2009) find that approximately 15% of marketable orders execute against hidden depth in a sample of stocks traded on Inet in 2004. These numbers increase to 17.3% and 19.0% in 2007 and 2008, respectively, for a NASDAQ dataset investigated in Hasbrouck and Saar (2011). By comparison, Rosenblatt securities estimates that U.S. dark pool activity was 11.26% of U.S. volumes in August 2011, see “Dark Pools Lose out to Exchanges,” Financial Times, Sept. 27, 2011. 3 cancelling orders simply to ascertain the existence of hidden orders on the book) and increased message traffic.3 Resolving debates over whether traders should be allowed to hide all, some, or none of their orders is complicated by a variety of factors. One is simply that all markets now feature hidden liquidity, making comparisons to the counterfactual difficult. Additionally, markets often adopt new opacity regimes in response to competitive pressures, complicating before-and-after analyses. Moreover, trader behavior should be affected by market design, suggesting that any analysis should examine how hidden liquidity affects order strategies (which are typically unobservable to both market participants and academic researchers). These difficulties have limited the ability of even the most insightful empirical and theoretical analyses to draw definitive conclusions as to the market consequences of different opacity regimes.
In this paper, we use an experimental methodology to investigate how non-displayed liquidity affects the market environment. Our analysis features informed traders who receive signals about the true value of securities and liquidity traders who must meet portfolio targets.
Markets operate continuously, allowing traders to implement a variety of trading strategies. Our trading platform features an electronic limit order book in which traders can enter orders of different sizes that can be cancelled at any time, choose to make liquidity (by placing limit orders in the book) or take liquidity (by hitting existing limit orders), and choose (depending on the rules of the market) to display or not display all or part of any order.4 Execution priority rules in our trading platform resemble those in actual markets, where displayed orders have priority over non-displayed orders. Overall, the functionality of our trading platform mimics that of current electronic markets, and it allows us to investigate the effects of transparency on trader and market behavior.
We investigate three market structures: Visible markets in which all orders must be displayed, Iceberg (or reserve) markets that allow both displayed and partially displayed orders
Nonetheless, we note that the issue of non-displayed liquidity existed before the advent of fully electronic markets.
Blume and Goldstein (1997), for example, discuss NYSE “not held” orders whereby clients instructed floor brokers to use their discretion in executing the orders. Floor brokers often chose not to display such orders in the book so that their existence is not revealed, but nonetheless participated in the trading process in a discretionary way. Nondisplayed orders in electronic markets can be viewed as an attempt to replicate at least some of the services performed by floor brokers.
See Bloomfield, O’Hara, and Saar (2005) for analysis of the make-or-take decision in an electronic market.
4 (i.e., a minimum size must be displayed and the remainder can be non-displayed), and Hidden markets in which orders can be displayed, partially displayed, or completely non-displayed.
Trading takes place in only one type of market at a time. We then compare equilibria across the three market structures, having employed experimental controls for learning, cohort, and other effects known to influence experimental studies. We test hypotheses suggested by the literature on how opacity (or the ability to utilize non-displayed liquidity in the book) affects trader behavior, with a particular focus on the disparate effects on informed and liquidity traders. We also investigate how opacity affects the overall performance of the market as captured by various liquidity and informational efficiency measures.
We seek to answer two basic questions. First, does the ability to hide liquidity affect trader behavior (and if so how)? Second, is market performance degraded when liquidity can be hidden? The short answer to the first question is yes, and by a lot, while the answer to the second question appears to be no, at least not by much.
Specifically, we find that both informed and liquidity traders use non-displayed orders if permitted, but informed traders strategies are more sensitive to changes in the opacity regime of the market. We observe that informed and liquidity traders respond differently to the ability to hide liquidity. For example, liquidity traders (who need to trade for reasons other than information about fundamentals) trade more aggressively by taking liquidity when the market is opaque. As transparency increases and liquidity traders are better able to assess depth in the book, they become less aggressive and are willing to wait for the execution of their limit orders.
In contrast, informed traders trade less aggressively (i.e., utilize limit orders to execute more of their trades) in an opaque environment that enables them to maintain their informational advantage for a longer time.
Our results concerning market performance highlight the remarkable manner in which trader strategies aggregate to create equilibrium outcomes. We find that giving traders the ability to control the exposure of their orders increases limit order book depth. Still, other liquidity measures such as “true” spreads (that reflect both displayed and non-displayed orders) or volume are not different across the three opacity regimes. As for informational efficiency, markets where all orders must be displayed exhibit more efficient prices at the open but this advantage fades 5 quickly as trading progresses. Throughout most of the trading period, whether or not traders can hide liquidity in the book has no impact on the informational efficiency of prices.
Thus, we find that while order strategies are greatly affected by allowing hidden liquidity, most market outcomes are not. Our results brings to mind Vickrey’s (1961) classic “irrelevance result” that auction design does not matter because trader strategies adjust to the rules of the auction so that revenue to the seller is unchanged.5 That market equilibrium features a similar robustness to opacity regimes should be an important, and reassuring, finding to market regulators. Still, it is important to recognize that the experience of market participants could differ markedly when markets are opaque even if market outcomes are similar. We find that displayed spreads are almost twice as large as true spreads in Hidden markets, and informed traders’ profits are higher in these markets as well when their private information is very valuable. Perhaps as a consequence, liquidity traders in our experiment prefer to trade in less opaque markets and consider them to be fairer.
Our research complements the literature on pre-trade transparency in markets, particularly studies of hidden liquidity in electronic limit order books. Madhavan, Porter, and Weaver (1999), Baruch (2005), and Boulatov and George (2012) compare markets that mandate the degree of pre-trade transparency (either transparent or opaque). Buti and Rindi (2008) and Moinas (2010), which are the most related theoretical analyses to our paper, look at how traders’ choices on whether to display all or part of their orders affect their strategies and market outcomes. These two papers provide interesting insights, but generally require a variety of restrictive assumptions for tractability. For example, the Buti and Rindi (2008) model includes only uninformed traders, while in Moinas (2010) informed traders can only supply liquidity (but not demand it). Our experiment has a more general structure for the information sets and strategies traders can adopt, allowing us to see if predictions from these models generalize to less restrictive settings.
Empirical studies, drawing on data from a variety of markets, also attempt to characterize the usage of non-displayed orders and investigate their information content (see Aitken,