High-Frequency Trading and Networked Markets
Total Page:16
File Type:pdf, Size:1020Kb
High-frequency trading and networked markets Federico Musciottoa, Jyrki Piilob , and Rosario N. Mantegnaa,c,d,1 SPECIAL FEATURE aDipartimento di Fisica e Chimica, Universita` degli Studi di Palermo, I-90128 Palermo, Italy; bTurku Centre for Quantum Physics, Department of Physics and Astronomy, University of Turku, FI-20014 Turun Yliopisto, Finland; cComplexity Science Hub Vienna, A1080 Vienna, Austria; and dDepartment of Computer Science, University College London, London WC1E 6BT, United Kingdom Edited by Simon Asher Levin, Princeton University, Princeton, NJ, and approved March 5, 2021 (received for review August 14, 2020) Financial markets have undergone a deep reorganization dur- provide liquidity only under normal market conditions whereas ing the last 20 y. A mixture of technological innovation and their trading is not guaranteed in exceptional market states, mak- regulatory constraints has promoted the diffusion of market frag- ing the markets more fragile and prone to flash or microflash mentation and high-frequency trading. The new stock market crashes (14). There is also empirical evidence that HFTs com- has changed the traditional ecology of market participants and petition is deteriorating liquidity provision (15) and that the market professionals, and financial markets have evolved into interaction of HFTs with orders of large institutions is performed complex sociotechnical institutions characterized by a great het- in a strategic way (16). erogeneity in the time scales of market members’ interactions The technical and regulatory changes observed in markets in that cover more than eight orders of magnitude. We analyze the last 20 y are not minor but rather they deeply affect the three different datasets for two highly studied market venues strategic behavior of market participants (5). An example of this recorded in 2004 to 2006, 2010 to 2011, and 2018. Using methods huge impact concerns market making and liquidity provision. of complex network theory, we show that transactions between Market making is the trading activity providing liquidity to the specific couples of market members are systematically and per- market (i.e., the trading activity allowing one to find quickly a sistently overexpressed or underexpressed. Contemporary stock counterpart for a transaction in a market). In the past market markets are therefore networked markets where liquidity provi- making was done by institutionalized figures (called specialists sion of market members has statistically detectable preferences at the New York Stock Exchange and jobbers at the London or avoidances with respect to some market members over time Stock Exchange [LSE]) that paid fees to the market and had with a degree of persistence that can cover several months. privileges and obligations for their role. Today, in most settings, We show a sizable increase in both the number and persis- market making is not institutionalized and it is freely strategically ECONOMIC SCIENCES tence of networked relationships between market members in performed by specialized market participants. Technical innova- most recent years and how technological and regulatory inno- tions (e.g., the use of computer algorithms and the fast access and vations affect the networked nature of the markets. Our study process of market quotes) and regulatory changes (e.g., market also shows that the portfolio of strategic trading decisions of fragmentation and changes in the information production and high-frequency traders has evolved over the years, adding to the dissemination) provide a changing environment deeply affect- liquidity provision other market activities that consume market ing the profile and ecology of different classes of investors and liquidity. the way they perform their strategic choices. We believe this extraordinary transition into a changing financial market set- STATISTICS complex networks j financial markets j high-frequency trading j ting is a clear case study of evolution and adaptation of the statistically validated networks ecology of market participants to new states of the financial world (17–19). he last 20 y have seen deep changes in the way finan- Tcial markets operate (1). The adoption of the regulation Significance about the national market system (“NMS”) for equity securi- ties (2) from the Securities and Exchange Commission of the During the last two decades, technological innovation and United States and a similar adoption subsequently taken by regulatory requirements have deeply changed the way finan- the European Securities and Market Authority have affected cial markets work. Today, financial markets are characterized the structure and practice of trading of equity securities in US, by the presence of high-frequency traders (able to perform European, and several other markets worldwide. The most evi- financial transactions at a submillisecond time scale) and mar- dent change has been the proliferation of market venues in a ket fragmentation. Using methods of complex networks, we given country or in a group of countries, usually addressed as show that some market participants (specifically so-called fragmentation of markets. A related change has been the spe- market members) preferentially interact with or avoid other cialization of a number of market participants in high-frequency market members persistently over a time scale extending up traders (HFTs). HFTs are professional traders able to use high to several months. By investigating two financial venues at speed in the generation, routing, and execution of orders (3). three different periods of time from two different decades, The amount of transactions performed by HFTs is today esti- we show that the persistent networked nature of today’s mar- mated to be around 50% in most markets (4). Typical response kets is most pronounced since the diffusion of high-frequency time of these traders to a market state or information can be trading and market fragmentation. as fast as a few microseconds. This exceptional time perfor- mance is often achieved by colocating technical infrastructures of Author contributions: F.M., J.P., and R.N.M. designed research; F.M. analyzed data; F.M. HFTs near the computer infrastructure of large market venues. performed numerical simulations; F.M., J.P., and R.N.M. analyzed results; and F.M. and The influence of regulatory changes and technological innova- R.N.M. wrote the paper with contributions from J.P.y tions have changed financial markets into complex sociotechnical The authors declare no competing interest.y institutions (5–9). This article is a PNAS Direct Submission.y There is no shared view about how changes occurring in mar- Published under the PNAS license.y kets have modified the basis of financial asset trading. One view 1 To whom correspondence may be addressed. Email: [email protected] is that the presence of HFTs makes markets more efficient by This article contains supporting information online at https://www.pnas.org/lookup/suppl/ decreasing the transaction cost per unit of transaction and by doi:10.1073/pnas.2015573118/-/DCSupplemental.y facilitating price discovery (10–13). Another view is that HFTs Published June 25, 2021. PNAS 2021 Vol. 118 No. 26 e2015573118 https://doi.org/10.1073/pnas.2015573118 j 1 of 8 Downloaded by guest on September 25, 2021 Recently the empirical characterization of the trading deci- transaction, or cancelation time for each order. In addition to its sions of investors (20) has made it possible to detect an ecology standard information, our special release of the database con- of investors (21, 22). In the present study, we analyze two dif- tains the anonymized identity of the market member who placed ferent venues of stock markets in three different periods of the order. Here we investigate all orders submitted through the time at the level of market members, i.e., at the level of those electronic order book. With this unique dataset we track the trad- companies that are acting for their proprietary trading or are ing activity of anonymized market members of the LSE in the acting as brokers or dealers for the trading of customers. Our investigated years. For the Nasdaq OMX Stockholm venue, it aim is to investigate whether the liquidity provision of the most was possible to obtain historical information about the identity of active market members (notably the HFTs) presents a networked market members performing each transaction. Having this type structure (i.e., it is provided under a framework of strategic of information is a rather unique characteristic of this venue and decisions). With the term “networked structure” we mean that for this reason the venue has been investigated in several papers market members trading in a venue of a fragmented market dealing with HFTs (15, 16, 27–31). The dataset for the Stockholm in the presence of HFTs might establish statistically validated venue of the Nasdaq OMX is the ITCH INET data stream main- preferential or avoided trading relationships with specific mar- tained and provided by the Nasdaq OMX (32). For this venue, ket members. By using tools of complex networks (23), we detect we investigate the 20 most liquid stocks, traded during the Febru- a networked structure among market members and we docu- ary 2010 to December 2011 and the January 2018 to December ment that this networked structure has evolved over the past 2018 time periods. The ITCH data stream captures the complete 20 y from a poor and dynamically changing to a richer and dynamics occurring on the electronic order book through a set of dynamically more persistent network structure highlighting an messages specifically formatted for each kind of action. ever-increasing strategic provision of liquidity. For both periods, we are able to reconstruct the full dynam- Financial transactions occurring in financial markets can be ics of the book in terms of submission (except for a small set described in terms of trading networks (23). Examples are trad- of transactions of the order of less than 5% of all transactions), ing networks occurring in the Interbank market (24) and in modification, and deletion of limit orders, together with the equity markets (25).