Where is the Value in High Frequency Trading

Electronic copy available at: http://ssrn.com/abstract=1712765 Where is the Value in High Frequency Trading? ∗ Álvaro Cartea† and José Penalva‡ February 28, 2011 Abstract We analyze the impact of high frequency trading in financial markets based on a model with three types of traders: liquidity traders, market makers, and high frequency traders. Our four main findings are: i) The price impact of the liquidity trades is higher in the presence of the high frequency trader and is increasing with the size of the trade. In par- ticular, we show that the high frequency trader reduces (increases) the prices that liquidity traders receive when selling (buying) their equity holdings. ii) Although market makers also lose revenue to the high frequency trader in every trade, they are compensated for these losses by a higher liquidity discount. iii) High frequency trading increases the volatil- ity of prices. iv) The volume of trades doubles as the high frequency trader intermediates all trades between the liquidity traders and market makers. This additional volume is a consequence of trades which are carefully tailored for surplus extraction and are neither driven by fundamentals nor is it noise trading. In equilibrium, high frequency trading and traditional market making coexist as competition drives down the profits for new high fre- quency traders while the presence of high frequency traders does not drive out traditional market makers. ∗We would like to thank Harrison Hong for his valuable comments and discussions. We are also grateful to Andrés Almazán, Gene Amromin, Michael Brennan, Pete Kyle and Eduardo Schwartz for their comments. We also thank seminar participants at CEMFI. The usual caveat applies. We welcome comments, including references we have inadvertently missed. The views expressed in this paper are those of the authors and do not necessarily reflect those of the Banco de España. †alvaro.cartea@uc3m.es, Universidad Carlos III de Madrid. ‡jpenalva@emp.uc3m.es, Universidad Carlos III de Madrid and Banco de España. 1 Electronic copy available at: http://ssrn.com/abstract=1712765 Keywords: High frequency traders, high frequency trading, flash trading, liquidity traders, institutional investors, market microstructure 1 Introduction Around 1970, Carver Mead coined the term “Moore’s law” in reference to Moore’s statement that transistor counts would double every year. There is some debate over whether this “law” is empirically valid but there is no discussion that the last forty years have seen an explosive growth in the power and performance of computers. Financial markets have not been immune to this technological advance, it may even be one of the places where the limits of computing power are tested every day. This computing power is harnessed to spot trends and exploit profit opportunities in and across financial markets. Its influence is so large that it has given rise to a new class of trading strategies sometimes called algorithmic trading and others high frequency trading. We prefer to use algorithmic trading (AT) as the generic term that refers to strategies that use computers to automate trading decisions, and restrict the term high frequency (HF) trading to refer to the subset of AT trading strategies that are characterized by their reliance on speed differences relative to other traders to make profits and also by the objective to hold essentially no asset inventories for more than a very short period of time.1 The advent of AT has changed the trading landscape and the impact of their activities is at the core of many regulatory and financial discussions. The explosion in volume of transactions we have witnessed in the last decade, and the speed at which trades are taking place, is highly suggestive that AT is very much in use and that these strategies are not being driven out of the market as a result of losses in their trading activities. Indeed, different sources estimate that annual profits from AT trading are between $3 and $21 billion (Brogaard [2010] and Kearns et al. [2010]). These strategies have supporters and detractors: on one side we find trading houses and hedge funds who vigorously defend their great social value, whilst being elusive about the profits they make from their use; and on the other hand there are trading 1This definition is consistent with the one used in Kyle [Flash Crash p3]: “We find that on May 6, the 16 trading accounts that we classify as HFTs traded over 1,455,000 contracts, accounting for almost a third of total trading volume on that day. Yet, net holdings of HFTs fluctuated around zero so rapidly that they rarely held more than 3,000 contracts long or short on that day.” 2 houses that denounce high frequency traders (HFTs) as a threat to the financial system (and their bottom line). Although AT in general and HF trading in particular have been in the market supervisors’ spotlight for quite some time and efforts to understand the consequences of HF trading have stepped up since the ‘Flash Crash’ in May 6 2010 (SEC [2010], Commission et al. [2010], Kirilenko et al. [2010], and Easley et al. [2011]) there is little academic work that addresses the role of these trading strategies. The objective of this paper is to provide a framework with which to analyze the issues surrounding HF trading, their widespread use, and their value to different market participants. To analyze the impact of HF trading in financial markets we develop a model with three types of traders: liquidity traders (LTs), market makers (MMs), and HFTs. In this model LTs experience a liquidity shock and come to the market to unwind their positions which are temporarily held by the MMs in exchange for a liquidity discount. HFTs mediate between LTs and MMs. HFT mediation instantaneous, buying from one and selling to the other while holding no inventory over time. HFTs, because of their information processing and execution speed, make profits from this intermediation by extracting trading surplus. The same model without HFTs, which corresponds to that of Grossman and Miller [1988], serves as benchmark to analyze the impact of HFTs. Naturally, we find that HFT’s additional intermediation increases the volume of trade sub- stantially (it doubles). The additional volume is neither driven by fundamentals (only the original trades, without the HFT, are driven by fundamentals) nor is it noise trading. Far from it, the extra volume is a consequence of trades which are carefully tailored for surplus extraction. Moreover, HF trading strategies introduce “microstructure noise”: in order to profit from inter- mediation HFTs buy shares from one trader at a cheap price and sell it more dearly to another trader, generating price dispersion where before there was only a single price. These properties, which are built into the model, closely correspond to observed behavior (e.g. Kirilenko et al. [2010]). Our main findings are: (i) the presence of HFTs exacerbates the price impact of the initial liquidity trades that generate a temporary order imbalance, imposing a double burden on liquidity demanders: the direct cost from the trading surplus extracted by the HFT, and the 3 indirect cost of a greater price impact; (ii) furthermore, this effect is increasing in the size of the liquidity need, consistent with the results in Zhang [2010]; (iii) the higher initial price impact arises as traders anticipate the future additional trading costs from the presence of HFTs, which generates an increase in the liquidity discount. Thus, MMs suffer countervailing effects from HFTs: increased trading costs from HFT surplus extraction versus increased expected returns from higher liquidity discounts. In our model these two effects cancel each other, leaving expected profits for MMs unchanged. (iv) Standard measures of market liquidity may lead to erroneous conclusions: in our model, HFTs do not increase liquidity and yet we observe increased trading volumes. In fact, liquidity traders face overall lower sales revenue and higher costs of purchase, suggesting that liquidity is better measured through total cost of trade execution. (v) Finally, we consider competition between HFTs and the decision for MMs to become HFTs. We find that profits from HF trading attract entrants who are willing to invest in acquiring the skills necessary to compete for these profits, but that competition is limited. As the number of HFTs increases, the expected profits of HF trading falls until the expected skills of an entrant (relative to those of existing HFTs) are insufficient to generate enough profits to cover the initial investments required to become an HFT. Thus, in equilibrium traditional MMs with low expected skills as HFTs will continue in their traditional role, coexisting with others acting as skilled and profitable HFTs. Our analysis focuses on the effect of HFT’s surplus extraction on trades initiated by liquidity needs that generate temporary trading imbalances. Nevertheless, our analysis of the role and effect of HFTs also applies to a broader set of circumstances. In particular it applies to trading by mutual fund managers, hedge funds, insurance companies and other large investors, and trading motivated not only by immediate liquidity needs, but also trading to build up or unwind an asset position, for hedging, etc. Two contemporaneous empirical papers lend strong support to the stylized features that our theoretical model captures as well as the implications concerning the impact that HF trading has on financial markets. The recent work of Zhang [2010] firmly concludes that HF trading increases stock price volatility and that this positive correlation between volatility and HF trading “is stronger for stocks with high institutional holdings, a result consistent with the view 4 that high-frequency traders often take advantage of large trades by institutional investors”. Kirilenko et al. [2010] study the impact of HF trading during the Flash Crash on May 6 2010. Their findings about the activities of HFTs also provide strong support for the theoretical description we use to include HFTs as pure surplus extractors in our theoretical model. They find that HFTs have among all types of traders the highest price impact and that “HFTs are able to buy right as the prices are about to increase. HFTs then turn around and begin selling 10 to 20 seconds after a price increase.” Moreover, they find that “The Intermediaries sell when the immediate prices are rising, and buy if the prices 3-9 seconds before were rising. These regression results suggest that, possibly due to their slower speed or inability to anticipate possible changes in prices, Intermediaries buy when the prices are already falling and sell when the prices are already rising.” These findings strongly support our assumption that HFTs (due to their speed advantage) can for the most part effectively anticipate and react to price changes as a key part in their strategies for surplus extraction. Before delving into our analysis of HF trading, in Section 2 we provide a brief overview of HF trading and HFTs, what HFTs could be doing, and what is it about trading speed that is so profitable for some and damaging for others. After this quick overview, in Sections 3 and 4 we develop our framework and analysis with a single HFT, and use the model to discuss the main issues raised by the presence of HFTs, respectively. In Section 5 we introduce competition amongst HFTs and the decision of an MM who considers setting up an HF trading desk. In Section 6 we conclude and discuss some key features about HFTs that require further research (and quality data). 2 Trading Algorithms, High Frequency Traders, and Financial Markets 2.1 Financial Market Developments Over the last years all major exchanges have revamped their systems to give way to the new era of computerized trading. Speed of trading and volume figures speak for themselves. In 5 the SEC’s report on “Findings regarding the market events of may 6, 2010” (SEC [2010]) we read that NYSE’s average speed of execution for small, immediately executable orders was 10.1 seconds in January 2005, compared to 0.7 seconds in October 2009. Also, consolidated average daily share volume in NYSE-listed stocks was 2.1 billion shares in 2005, compared to 5.9 billion shares in January through October 2009. Consolidated average daily trades in NYSE-listed stocks was 2.9 million trades in 2005, compared to 22.1 million trades in January through October 2009. Consolidated average trade size in NYSE-listed stocks was 724 shares in 2005, compared to 268 shares in January through October 2009. Other important metrics that intend to capture market efficiency and information transmis- sion have also undergone considerable changes as a result of modifications of market rules and the prominent role that computing has taken in financial markets. For example, Chordia et al. [2010] focus on comparisons of pre- and post-decimal trading in NYSE-listed stocks (subperi- ods from 1993-2000 and 2001-2008). Some of their findings are that average effective spreads decreased significantly (from $0.1022 to $0.0223 cents for small trades (<$10,000) and from $0.1069 to $0.0267 for large trades (>$10,000)), while average depth available at the inside bid and offer declined significantly (from 11,130 shares to 2,797 shares). From 1993-2000 the mean trade size is $82,900 and from 2001-2008 $36,400 while the mean number of transactions is 1,136 and 14,779 respectively. 2.2 What characterizes Algorithmic and HF Trading We adopt Hendershott et al. [2010]’s definition of AT: “the use of computer algorithms to automatically make certain trading decisions, submit orders, and manage those orders after submission”. We distinguish HF trading as a subset of AT. An HF trading strategy is an AT strategy that is based on exploiting greater processing and execution speed to obtain trad- ing profits while holding essentially no asset inventory over a very short time span—usually measured in seconds, mostly less than a few minutes, and certainly less than a day.2 One sometimes finds these strategies described as latency arbitrage. HFTs are proprietary firms and 2In their study of the ‘Flash Crash’, Kirilenko et al. [2010] find that, holding prices constant, HFTs reduce half their net holdings in 115 seconds. 6 proprietary trading desks in investment banks, hedge funds, etc, that based on these strategies have the ability to generate large amounts of trades over short periods of time, Cvitanić and Kirilenko [2010]. There are other AT strategies that are in use for other purposes. For example, there are AT liquidity strategies which strategically post and cancel orders in the order book to exploit widening spreads, or AT strategies designed to execute large orders with the smallest price impact. Our analysis focuses exclusively on HF trading strategies, which we believe are the ones most critics of AT have in mind. Making the distinction between HF trading and AT is important because it highlights the substantial difficulty one encounters when measuring the impact that HFTs have on markets according to metrics such as volume, spreads, and liquidity. For example, estimates of HF trad- ing volume in equity markets vary widely depending on the year or how they are calculated, but they are typically between 50% and 77% of total volume, see SEC [2010] and Brogaard [2010]—although how much is actual HF trading versus generic AT is unclear. Also, Hender- shott et al. [2010] find that for large-cap stocks AT improves liquidity and narrows effective spreads. They also find that AT increases realized spreads which indicates that revenue to liquidity suppliers has increased with AT, but it is difficult to infer how much of these effects are due to AT that is not HF trading. Similar identification problems are present in another recent study, Brogaard [2010], which finds that HFTs contribute to price discovery and reduce volatility. Thus, this identification problem, as well as possible collateral effects on other AT strategies, have to be taken into account in any regulatory implications one may draw from our analysis, as we focus exclusively on HF trading. As per our definition, paramount to the activities of HF traders is the speed at which they can: access and process market information; generate, route, cancel, and execute orders; and, position orders at the front of the queue in the trading book to avoid having stale quotes in the market. Their speed or low latency is mainly due to two key ingredients: capacity (software and hardware), and co-location. Co-location allows HFTs to place their servers in close physical proximity to the matching engines of the exchanges. Surprisingly, being near the exchanges can shave the speed of reaction by a sufficient number of fractions of a second to provide HFTs 7 a valuable edge when trading in the market—to the extent that they are willing to pay millions of dollars for this service. Perhaps the most revealing behavior of HFTs is how they make use of cancelations to poke the market and extract valuable information. For instance, the strategy known as ‘pinging’ is based on submitting immediate-or-cancel orders which are used by HFTs to search for and access all types of undisplayed liquidity SEC [2010].3 Another strategy, known as ‘spoofing’, consists of sending out a large amount of orders over a short period of time before immediately canceling most of them so that only a few are executed. This burst of activity is expected to trigger other algorithms to join the race and start buying or selling (and slow down information flows to other market participants). 2.3 How is it that HFTs could be making money? Here we provide four stylized examples that show how HFTs could be exploiting their speed advantage by posting, executing, and canceling orders to position their orders at the front of the queue and intermediate in market transactions for a negligible period of time. Although the first three examples outline different strategies used by HFTs, the underlying feature common to all three is the ability that HFTs have to extract trading surplus. Example 1. One way in which HFTs can extract surplus is by exploiting their speed to alter market conditions in a way that encourages buyers to accept a slightly higher price and sellers a slightly lower one. The idea is relatively simple and works in a setting where liquidity traders split their trades in small packages and MMs do not have large outstanding offers in the books. Suppose a trader (LT) needs liquidity and wants to sell a block of shares. As the first shares come into the system (say at the best buy price of $5.50 per share), the HFT cancels her outstanding posted buy offers that have not been executed. She then posts additional sell offers, adding to the increased selling pressure, in order to help clear the remaining posted buy orders in the book. Once the book is clear, the HFT quickly reposts a significant number of offers at lower prices (say $5.47) so that she is first in the buying queue. This is only possible if 3There are circumstances when orders are placed close to best buy or sell with no intention to trade, this is known as book layering, and up to 90% of these orders are immediately canceled, see SEC [2010]. 8 she can move quickly enough that by the time the MM reacts to the increased selling pressure, new posted offers by the MM sit behind those posted by the HFT at $5.47 per share. The LT finds that the market around $5.50 has dried up and can only sell at $5.47. These shares are bought by the HFT, who is at the front of the queue. Also, having posted substantial orders at