Lit and Dark Liquidity with Lost Time Data: Interlinked Trading Venues around the Global Financial Crisis

Interlinked Trading Venues around the Global Financial Crisis

T. Vuorenmaa

• Publisher: Palgrave Pivot
• Year: 2014
• Print ISBN: 9781137432605
• eBook ISBN: 9781137396853
• Topic: Economía
• Subtopic: Gestión de riesgos financieros

Since the 2008 financial crisis, researchers and policy makers have been looking to empirical data to distil both what happened and how a similar event can be avoided in the future. In Lit and Dark Liquidity with Lost Time Data, Vuorenmaa analyses liquidity to better understand the crux of the financial crisis. By relating liquidity to jump activity, market microstructure noise variance, and average pairwise correlation, Vuorenmaa uncovers the dynamics and ramifications behind anonymous trades made outside of public exchanges, and measures its impact on the crisis. This volume is ideal for academics, students, and practitioners alike, who are interested in investigating the role of lost time in and after the recession.

Introduction

Trading volumes and volatility in the U.S. equity markets increased strongly in the second half of 2007, and continued to do so in 2008. As the U.S. finance sector problems became public knowledge around those times – widely referred to as the Global Financial Crisis – it would seem that the financial problems caused the activity increase. In reality, however, trading volumes had been increasing for years. While being relatively stable from 2001 to 2005, the average annual trading volume has been reported to have doubled from 2006 to 2009 [see, e.g., Anderson and Dyl (2014)].

Regulatory changes in securities trading, primarily the Regulation National Market System (RegNMS) and the decrease of the minimum price increment to one cent (decimalization), have facilitated this evolution. The new market microstructure level rules encourage automated trading (AT). It has, for example, become common to use cost-minimizing execution algorithms that slice large orders to smaller ones on exchanges while also sending a considerable amount of orders to dark pools outside of the lit markets [see SEC (2010)].¹ The evolution has changed how the trading industry handles data; there are now not only more data, but more data sources. Markets are notably more complex than before, and a large portion of the complexity is driven by regulations.

The traditional stock exchanges compete hard with numerous newcomers on both the “lit” and the “dark” side. In the U.S., the once-mighty NYSE (NYSE Euronext) face hardened competition not only from other lit exchanges, but also from Electronic Communication Networks (ECNs) and dark pools.² There are known to be more than fifteen trading venues operating in the U.S. currently, with more than 30 percent of the total volume traded off primary exchanges. Similarly, in Europe, Alternative Trading Systems (ATS) known as Multilateral Trading Facilities (MTFs) claim a large chunk of the total trading volume.³ In this fragmented scene,⁴ a subset of AT called high-frequency trading (HFT), or automated low-latency trading, has become influential. HFT firms can, for example, arbitrage away small price discrepancies between different trading venues. In practical terms, HFT glues the fragmented financial markets together as an interlinked network of trading venues, a feature that can be considered to be either good or bad [for a review, see Vuorenmaa (2013)]. In short, the empirical results we present below are illustrative of how well the U.S. trading venues are interlinked as measured by market quality related variables around the Global Financial Crisis – an extraordinary time period in the history of finance.

The state of liquidity in the U.S. equity markets has been surveyed in the past by Abrokwah and Sofianos (2006). But as the liquidity fragmentation process has progressed over the recent few years, traders have become more selective where they route their orders to. Heterogeneity of trader motivations is a key characteristic. Daniëls, Dönges, and Heinemann (2013) argue that impatient traders use lit markets and patient traders prefer dark markets. This has made the state of liquidity very dynamic and ambiguous. As a result of more order flow being routed away from the lit markets, the price formation process may have been impaired, potentially worsening market quality. There exists also evidence to the contrary, though. Jiang, McInish, and Upson (2012) find that in fragmented markets price discovery on exchanges is improved, because uninformed traders can route their order flow to off-exchange trading venues.⁵ Thus, the effect of fragmentation on market quality is not obvious.

Traded volume on the display and trade reporting facilitilities (ADF/TRF) give a rough idea of how fragmented the U.S. equity markets are. O’Hara and Ye (2011) use such an approach to conclude that fragmentation has not had a detrimental impact on market quality. Degryse, de Jong, and van Kervel (2013) come to the same conclusion for European markets. They also find that dark venues are associated with larger price impact and wider spreads. We find similar qualitative differences, but our focus is otherwise quite different from theirs. For us, the Global Financial Crisis is a special time period for studying liquidity fragmentation with respect to both lit and dark liquidity. This stands in contrast to Gresse (2013), for example, who specifically avoids the crisis period in her analysis. We approach the case from a purely statistical perspective with the belief that a statistical study of the relationships between liquidity, market microstructure noise, and correlations would reveal some unnoticed aspects of liquidity fragmentation. The overarching goal is that our results would be useful in improving regulations and market mechanisms by providing empirical evidence to develop the theory of modern markets. The results could also be applied to enhance the performance of trading and order routing algorithms.

From an academic point of view, liquidity fragmentation and the soaring level of automated trading present serious challenges. Standard market microstructure theory stresses the role of designated market-makers and asymmetric information in rather clean-cut cases. Nowadays, the reality is considerably messier as the rhythm of the markets has changed dramatically. In particular, the role of market-making has changed considerably over the past ten years or so. Transparency can be argued to have been diminished due to liquidity fragmentation, smaller tick sizes, faster pace, and more strategic order placing. Execution has become not only faster but more intelligent.

These changes in market microstructure affect the amount of liquidity. Here, we hypothesize the effect to be different on different trading venues, especially between the lit and dark ones. The statistical approach we apply is flexible with respect to institutional and regulatory developments, but exact causal relationships remain largely hidden under the surface. Uncovering them would require more data.

To be more specifc, we study if lit (NYSE, Arca, and Nasdaq) and dark (NASD ADF/TRF) liquidity show significantly different data characteristics from each other, especially around the Global Financial Crisis. The data sample we analyze comprises actively traded stocks that collectively form much of the Dow Jones Industrial Average (DJIA) index. Since we do not possess data directly from any dark pools [cf. Buti, Rinder, and Werner (2011)], and because only little reliable public information on their trading activity exists, we must rely on a dark pool proxy. The proxy we use is based on the fact that dar...

Table of contents

• Cover
• Half title
• Introduction