List of abbreviations
1 Introduction
1.1 Problem
1.2 Aim
1.3 Research Questions
1.4 Scientific Method
1.5 Structure of the paper
2 Fundamentals of HFT
2.1 History of HFT and its presence in current markets
2.1.1 Algorithmic Trading (AT)
2.1.2 Definition and characteristics of HFT
2.1.3 HFT’s fraction of market activity in the US and Europe
2.2. Technology used for HFT
2.2.1 Software for HFT
2.2.2 Hardware for HFT
2.3 Users of HFT
2.4 Strategies of HFTs
2.4.1 Liquidity Provision (Market Making)
2.4.2 Liquidity Detection
2.4.3 Arbitrage
2.4.3.1 Market Neutral Arbitrage (“Pairs Trading”)
2.4.3.2 Cross Asset-, Cross Market- & ETF-Arbitrage
3 The effect of HFT on capital markets and the Economic Neoclassical Theory
3.1 The Economic Neoclassical Theory and its relation to HFT
3.2 The effect of HFT on liquidity
3.3 The effect of HFT on price discovery
3.4 The effect of HFT on volatility
4 The regulation and supervision of HFT
4.1 Manipulation and the importance of regulation and supervision of HFT
4.2 The Flash Crash
4.3 General regulatory measures
4.3.1 Unfiltered/Naked Sponsored Access
4.3.2 Flash Orders
4.4 The regulatory response to the Flash Crash
4.4.1 Circuit breakers
4.4.2 Stub quotes and erroneous trades
4.5 Other possible regulatory measures and their effect on volatility
4.5.1 Tobin tax
4.5.2 Short sale constraints
5 Conclusion
References
Table of figures
Abbildung in dieser Leseprobe nicht enthalten
In the last four decades, technological progress led to an electrification of stock trading systems. Traders were enabled to place their orders, which were later processed via electronic networks, with the help of computers. Soon they realized that the profitability of trading strategies could be increased by employing computer algorithms to trade autonomously, reducing time needed to analyze information, publish quotes as well as trigger and process trades. This led to the implementation of Algorithmic Trading (AT). High Frequency Trading (HFT) is a subset of AT, at which financial instruments are traded by algorithms at very high speed.
The past has shown that negative developments on capital markets are intensified by HFT. Andrei Kirilenko explains in his work “The Flash Crash: The Impact of High Fre- quency Trading on an Electronic Market” that HFT did not trigger the Flash Crash but intensified the volatility that resulted from the event. Also on the 19th of October 1987, “Black Monday”, the increasing computerization of stock trading processes led to a significant price drop. As a consequence, the high and still growing market share of HFT leads to an increase in risk that a simple correction turns into a serious drop in prices causing market instability. Theoretically HFT should increase efficiency in fi- nancial markets. However, due to the empirical observation mentioned above, it seems that HFT takes effect the other way round. It seems that, at least under cer- tain circumstances, HFT enlarges volatility. This cannot be explained by the econo- mic neoclassical theory. This problem is discussed in a lot of literature in which se- veral different approaches have been made to explain it.
The aim of this paper is to discuss why HFT cannot be fully explained by the neoclassical theory of economics. Therefore, the controversial positions in literature will be presented and discussed. Primarily, its negative influence on volatility seems to contravene the modern finance. Furthermore, in the course of this work it will be illustrated that, by employing strict regulation of financial markets, this negative impact cannot be reduced to a sufficient extent in order for HFT to be characterized as market optimizing, according to the neoclassical theory of economics.
As a result of this objective, the following research questions arise:1) Is there research that perfectly describes the effect of HFT on capital mar- kets or is it, based on a comparison of different researchers’ findings, ne- cessary to combine these findings in order to ensure a realistic description of the matter?
2) Can the negative effect of HFT on market quality be reduced, by employing strict regulation of financial markets, to a sufficient extent in order for HFT to be explained by the neoclassical theory of economics?
The scientific method is comprised of literature research. For this paper mainly working papers and professional articles have been used as references.
In the first part of this thesis, the basics of HFT are explained, with reference to its history, the presence of HFT in current markets as well as its users and their strate- gies.
The following chapter deals with the effect of HFT on capital markets and its capability of optimizing financial markets according to the neoclassical theory of economics. Findings on the effect of HFT on liquidity, price discovery and volatility in financial markets are examined.
The importance of regulation and supervision of HFT as well as regulatory measures and their influence on market quality are topics of the third section of this paper. Also, the “Flash Crash of May 06, 2010”, an incident impelling the regulation of HFT, is il- lustrated in detail.
A conclusion is shown in the last part of the paper.
Within the last 40 years, a drastic change in trading processes of securities and other financial products has taken place. Technological development has led to an electrifi- cation of trading systems, which extensively replaced physical trading floors (see Gomber et al. 2011, p. 8). Lee defines ETSs as exchange-systems for financial ins- truments, in which buyers and sellers are brought together (see Lee 1998, p. 282).
The first computer-assisted trading system - today it is known as NASDAQ - was put into operation by the NASD in the USA, in 1971. Since then, market participants have continuously aimed to improve trading processes (see Gomber et al. 2011, p. 8). Today almost all markets are electronic. According to Jain, the leading exchanges in 101 of 120 sample countries have electronic trading, 85 of them act fully electronically, without floor trading (see Jain 2005, p. 2965).
Electronic trading systems execute their trades with the help of computer algorithms. These use direct market access in order to autonomously make investment decisions, based on observed and analyzed real-time market information, submit orders and manage them after submission without human intervention of any kind. AT can be defined as the use of such means for trading (see Hendershott/Jones/Menkveld 2011, p. 1; see Gomber et al. 2011, p. 14).
The algorithms can be programmed in various ways, depending on the trading stra- tegies the market participants aim to execute with the help of AT. Algorithms for AT are generally characterized by holding periods of up to several weeks and months as well as the aim to achieve a particular benchmark, among other things. Some strate- gies, however, may require algorithms that have different features for their execution. Such might be applicable to HFT, a subgroup of AT (see Gomber et al. 2011, p. 14).
As with AT, there is not only one correct definition of HFT. Basically, algorithms used by HFTs and those necessary for general AT, have common characteristics. Both are used by professional traders to observe market data and initiate and manage trades by using direct market access (see Gomber et al. 2011, p. 15). However, there are certain differences. Unlike ATs, HFTs usually do several thousands of trades per day (see Hasbrouck/Saar 2010, p. 1) on their own accounts (see Hasbrouck/Saar 2010, p. 13). The holding periods of financial instruments traded that way amount to several milliseconds (see Authority for the Financial Markets 2010, p. 8). Algorithms that exe- cute HFT-strategies generally update, cancel and delete their orders frequently, for the purpose of extracting small profits per trade. Within the last four years, the ratio of orders to completed trades has changed from 20:1 to more than 50:1 (see Watson 2011, p. 59). Focusing on highly liquid financial instruments helps to achieve a flat position at the end of the day, meaning that holding positions overnight is avoided (see Gomber et al. 2011, p. 15).
The market shares of AT and HFT are difficult to measure, as both human traders and trading computers submit their orders electronically. Most algorithms send their orders electronically to an exchange, which is why the rate of electronic message traffic is commonly used as a proxy for the amount of AT - this is also applicable to HFT - on the market (see Hendershott/Jones/Menkveld 2011, p. 6). Figure 1 illus- trates the best estimates from several institutions of the market share of HFT.
Abbildung in dieser Leseprobe nicht enthalten
Figure 1: HFT Market shares from industry and academic studies (Gomber et al. 2011, p. 73).
The results show a growing market share of HFT over the last five years. An estimate of WM Gruppe, Börse-Zeitung, which is not illustrated in Figure 1, indicates that, in 2011, HFT-activity amounted to 70% of the volume traded in securities in the US. In the same year, 40% of the stock trades of Deutsche Börse AG were executed by HFTs, and the trend is still growing. Eurex, one of the world’s leading derivatives ex-changes, is expecting a growth rate of 20% per year in the field of automated trading,due to expected increases in the efficiency of arbitrage-trading, faster marketmaking, rationalization and further automation of trading, among other things (see Gomber 2011, p. 19).
HFTs need a comprehensive IT-infrastructure in order to be able to execute their trades within milliseconds or microseconds. The lower the time needed to receive market information, make investment decisions and place an order at an exchange, the more profitable is the system for its user. This is why HFTs strive to minimize this so called latency period (see Hasbrouck/Saar 2010, p. 1). HFTs try to get the ad- vantage over other traders by employing high performance supercomputers and ob- taining information faster than others, for instance through direct connections of their servers to exchanges (see Angel/Harris/Spatt 2010, p. 38). They continuously aim to optimize their processes in terms of software as well as hardware.
Depending on their strategies (see chapter “2.4 Strategies of HF-Traders”), HFTs use different software, including standard trading programs available on the market as well as individually programmed and customized software. Every piece of software consists of a front-end, the user interface that shows the information the HFT needs, and a back-end, the actual algorithm. Such algorithms are the core of the software and a prerequisite for successful HFT, as they obtain and analyze market information and make investment decisions based on these data. The fewer and simpler the re- quired information is, the faster this process can be executed (see Gomolka 2011, p. 198). This results in lower latency and higher profitability of the system.
Besides software, the quality of hardware has a big influence on an HFT’s success. Modern technology makes it possible for electronic signals to travel at the speed of light (see Kumar et al. 2011, p. 4). According to Yan Ohayon, this fastest possible way of transmitting electric signals actually slows the orders of HFTs down, due to HFTs’ enormous capacity in terms of speed which is created by the software (see Watson 2011, p. 56). For the fastest possible data processing and execution of their international trading activity, HFTs need a WAN, a network connecting multiple ser-vers (see Zadorozhny/Raschid/Gal 2008, p. 1). These servers are located as close as possible to the servers of the exchanges the HFT operates on (“co-location”), transferring the information that is generated by the exchanges’ servers to a control center that is ideally located half way between the respective trading venues. As such “co-locations” are very valuable to HFTs, exchanges rent them to the traders (Sietmann 2011, p. 152).
Figure 2 illustrates the development of execution times of small market orders for stocks listed on NYSE and NASDAQ (in seconds), from 2001 to 2009. It is recognizable that further automation of trading processes leads to an increase in execution speed (see Angel/Harris/Spatt 2010, p. 22). Today, according to Kumar et al. (2011), execution times of one millisecond and less are available at numerous American trading venues (see Kumar et al. 2011, p. 4).
Abbildung in dieser Leseprobe nicht enthalten
Figure 2: Market Order Execution Speed. Rule 605 data from Thomson for all eligible market orders (100-9999 shares) (see Angel/Harris/Spatt 2010, p. 22).
AT can be separated into two branches with different characteristics: (1) Agency tra- ding: The automated execution of trades in the course of which a trader invests his clients’ capital. The purpose of such deals is to profit from an investment rather than trading itself. (2) Proprietary trading: A method of using strategies for the purpose of pattern detection in order to generate profits from trading itself (see Hasbrouck/Saar 2010, p. 12).
As illustrated in chapter “2.1.2 Definition and characteristics of HFT”, HFT is charac- terized by the investment of the trader’s own capital. As a consequence, agency traders, such as investment banks, can only employ HFT to maximize the profits they earn from their core business, the service they get paid for by their clients. Proprie- tary trading firms, however, are meant to generate profits from the trading of financial instruments, using their own capital (see Hasbrouck/Saar 2010, p. 12). These com- panies represent a fraction of approximately 2% of the 20.000 hedge funds, mutual funds and brokers operating in US-equities. Businesses like RGM usually do not em- ploy more than 120 scientists and software engineers for the development of tech- nology that automatically monitors market information and trades based on these observed data. Doing that, without taking into account the real value of a company, currency or commodity, this economic sector generates an annual profit of several billion USD - USD 7.2 billion during the economic and financial crisis, in 2009 (see Foroohar 2010). The most important of the strategies HFTs pursue, when trading based on monitored and examined market data, are explained in chapter “2.4 Strate- gies of HFTs” of this paper.
HFT itself is not a strategy. The strategies used by HFTs are often traditional and well known to the market. What makes them so efficient is the use of modern technology to employ them. Depending on their business models, HFTs have various different strategies (see Gomber et al. 2011, p. 24). In the course of this chapter, three well known and frequently used strategies are introduced: Liquidity Provision (Market Making), Liquidity Detection and Arbitrage.
By submitting buy- and sell-limit orders at the same time, HFTs provide liquidity to market participants who want to trade. This way they create “artificial liquidity”, liquidi- ty that results from contractual obligations instead of trading interest (see Gom- ber/Lutat 2007, p. 11). Such market makers usually profit from (1) buying at the bid price and selling at the ask price and (2) receiving liquidity rebates from trading ven-ues (see Gomber et al. 2011, p. 24).
Buying at a certain price and selling at a higher price, as described in (1), is an HFT- strategy called “spread capturing”. If this principle is followed in the majority of trades executed by HFTs, it can contribute significantly to the revenues of these firms (see Gomber et al. 2011, p. 26). (2) Outlines a strategy, employed by HFTs, that allows them to generate revenues from trading fee discounts and liquidity rebates they re- ceive from exchanges. Many trading venues charge lower fees and offer incentives to liquidity providers while making trading more expensive for liquidity demanders (see Gomber/Lutat 2007, pp. 10-12). They do that in order to attract more of such passive traders (liquidity providers), and further to make it affordable to submit more limit orders to always reflect the total market information. Thus they narrow the bid- ask-spread (see Gomber et al. 2011, p. 26) and improve market quality. HFTs can, but are not obliged to, pursue a market-making-strategy. However, such strategies are frequently adopted by HFTs (see Gomber et al. 2011, p. 17), as they can be related to certain benefits, such as maker fees (see Authority for the Financial Markets 2010, p. 15).
HFTs employing liquidity detection strategies use ultra-high speed technology in or- der to analyze the market activity of other traders for the purpose of profiting from information they have (see Gomber et al. 2011, p. 28). HFTs “sniff out” the market in order to find out if a large order is placed. That would be an indicator that a market participant has valuable information based on which he wants to trade. This can be detected by simply posting an immediate-or-cancel order, priced at the bid-ask- midpoint, in the market. If the order is executed, this might indicate that a large order is posted (see Reynolds 2011, p. 24). The ability of trading much faster than other market participants based on this data, gives HFTs an enormous advantage (see Gomber et al. 2011, p. 28). For this reason, such strategies can be profitable for HFTs, but they often are a matter of concern for institutional investors and other non- HFTs. Jones (2013) describes this problem with the help of an example: If an institu- tional investor is buying shares, HFTs might be able to deduce that from information detected on the market. As a consequence, they can drive the price of these shares up and sell them at a higher price, probably even to the institutional investor who ini- tially posted the buy-order (see Jones 2013, p. 9). This means that the profit of HFTs
[...]
