It can even go a step further. Recently, one bulge bracket bank admitted in a New York conference that it is struggling to keep up with the demands of today’s market, calling the challenges overwhelming. It admitted to buying a product directly from an HFT house and thus facing a very visible relationship conflict. Unfortunately for a bank, though we can speculate on the quality of the product purchased compared with that used in-house by the HFT firm, it would be interesting to know how far the bank trailed the field and what specifically forced its hand.
What is HFT?
Generally speaking, HFT houses are proprietary trading firms that hold few, if any, overnight positions. HFT are fully automated with high spends on technology and are highly latency (speed) sensitive. Cross technique risk adjusted returns are abnormally high, with Sharpe ratios often in the order of nine or double digit. Holding periods are at the extreme short end of the curve, operating in a time frame ranging from milli seconds to a few hours. Well known names in the HFT space would include Getco, Infinium and Optiver.
In terms of market share, HFT accounts for approximately 60% of US secondary market equity trading and about an average of 35% of total pan European trading (with considerable variation between stocks and countries; see Fig. 1.1). During the volatile days of August, HFT was reported to be 75% of US equity trading making net profits of $60 million in US stock markets on 8 August.
Widely accepted global net profits are in the multi billions of dollars with TABB Group noting 2008 fiscal year estimates of $8-$20 billion net profit for HFT in the US alone. In addition, recent equity volatility is likely to advance year on year profit in 2011.
Broadly speaking, most techniques fall under the two umbrellas of statistical arbitrage and market making (a technical description, not the role of an obligated market maker). Sub categories would include latency and rebate arbitrage as well as directional trading. Beneath these top lines, processes from artificial intelligence, machine learning, physics and mathematics are widely deployed in each umbrella. Similarly, stochastics and GARCH and a high number of algorithms adapted from voice recognition and defence technology to enable pattern recognition sequences in both linear and nonlinear conditions are also deployed, an example being the Baum-Welch algorithm.
Why the controversy?
One issue is whether a two tiered market exists between firms with higher and lower technological resources. There are also questions about whether market manipulation and market abuse have occurred. This piece, however, will look at three other areas of HFT, with a particular reliance on the importance of objective evidence, as opposed to subjective opinion. The three points we will cover are price discovery, volatility/stability and liquidity/volume.
Across all three areas, the academic literature commenced relatively recently and certainly up to the first quarter of 2010, little rounded academic work existed. Such was the shortage that a doctoral candidate’s work was widely used as a benchmark study by investment banks, HFT and in some cases regulators. Subsequently, this piece has been shown to hold material flaws and holds little credibility today. Some papers were commissioned by exchanges who gain commercially from HFT business (e.g. Gomber/Deutsche Boerse). Not only was credibility compromised, but more importantly, in many cases, so was the methodology. As time progressed more universities and researchers, un-conflicted by association, were able to conduct research into the area and the position began to shift.
In effect, this shift in output occurred as a more genuine understanding developed. What’s more, as research standards improve, simplistic assumptions like HFT are “liquidity providers” or “dampen volatility” or “decrease bid-ask spreads” have become increasingly less credible.
Recently, many observers have assumed that price discovery has improved. But evidence to support this is limited. For example, HFT market makers need not just learn passively from observed order flow, but can also strategically set quotes to induce the revelation of information, potentially distorting short and longer term formation. Even the capital asset pricing model has been reviewed. CAPM assumes a frictionless world where price discovery is trivial and risk and return are the two components, yet CAPM ignores the impacts of liquidity, both breadth and depth.
Liquidity transience is also key. With holding periods trending downwards, largely due to the weighted impact of HFT, alterations can occur to the inter-temporal structural changes in stock trading volume and price dynamics. As Zhang stated in 2010, this could “have broad implications for studies that assume volatility, trading volume, or price discovery to be stationary over time (no structural changes are allowed in the classic Fama-MacBeth approach).”
It is rarely doubted that HFT tightens the spread at the first quote on the book, however, questions remain about HFT’s impact on market depth. Frequently the size of the bargain at the top ofthe book is, despite being the best price, the smallest in quantity. Thus 200 shares with a marginal price improvement over a 5000 share order second in line opens the question of HFT’s impact on absorption and thus price discovery.
Zhang’s seminal study goes on to find that in fact, over the longer term (quarterly periods), HFT hinders price discovery. This finding represented the first trend shift away from other studies which confirmed a positive impact on price discovery, for example, the 2009 study from Hendershott and Riordan. To further elaborate, there is a general view that increased trading activity leads to improved bid ask spreads, and thus, improved price discovery. However, this view tends to overlook the impact of noise trading on the market.
As early as 1993 Campbell, Grossman and Wang conducted studies into this area, finding that noise traders lead prices away from fundamentals, agitating prices into temporary swings and reversals which would distort the discovery of a genuine price. The relationship between over- and undershooting the price was further shown by Schwartz and Francioni in a 2004 study when they noted “price discovery is inaccurate when new equilibrium values are not instantaneously achieved.” Today, these factors still underpin the market. What’s changed is that HFT dominates secondary equity market trading.
The self-defined HFT technique of market making, deployment of flickering quotes spread across multiple venues, as well as more passive aggressive forms of noise trading, has led to concepts such as “artificial liquidity” and “disappearing liquidity” entering the lexicon of methods to describe HFT. In turn, they have raised questions over HFT’s role in effecting price discovery.
Proponents of HFT have long argued that HFT’s liquidity provision into the market has the impact of dampening volatility. Some point to the fact that HFT ends the day flat and so cannot impact volatility. However, this ignores what occurs intraday.
Given that stock correlation is positively correlated to volatility Andrew Haldane, Head of Financial Stability at the Bank of England, in his July 2011 study found that “intraday volatility has risen most in those markets open to HFT.” Haldane also noted that “HFT algorithms tend to amplify cross stock correlation in the face of a rise in volatility”, in effect, stating that in volatile times HFT accentuates volatility given the positive correlation between the two variables. Separately Biais and Wooley found that “(HFT) algorithmic trading could create the scope for systematic risk”, potentially opening the debate as to whether or not HFT techniques have the ability to shudder the market before moving it into explosive and potentially contagious scenarios. This concept of HFT adding a volatility layer, over and above the pre-existing fundamental volatility, is increasingly embedded in the body of credible research.
Protter and Jarrow earlier this year tackled the relationship between the spread and volatility head on by creating a model with perfect liquidity and a zero bid/ask spread. The authors found that under these conditions HFT increases market volatility. This finding re-enforced Zhang’s work which uncovered that HFT is positively correlated with stock price volatility. Crucially, Zhang controlled for fundamental volatility and other exogenous factors; the author expanded by discussing HFT institutional order detection mechanisms and refers to them as “a practice that pushes the stock price up (down) if institutional investors have large buy (sell) orders, thereby increasing stock price volatility.”
Over recent years, the increases in trading volume are largely attributed to HFT, with that generally being seen as a positive outcome by many in the market, particularly investment banks and execution venues. However, in line with more recent studies into HFT, Dichev, Huang and Zhou reported in 2011 that “there is a reliable and economically substantial positive relation between volume of trading and stock volatility.” The authors conclude that “trading-induced volatility accounts for about a quarter of total observed stock volatility” and that high volume stock trading “injects an economically substantial layer of volatility above and beyond that based on fundamentals.”
Liquidity and volume
A final point to note is about the differences between liquidity and volume. Most have said they are one and the same, and in the more traditional sense of understanding they often are. However, in the more contemporary understanding they are not. As the Securities and Exchange Commission and Commodity Futures Trading Commission noted in their joint report into the Flash Crash of 6 May 2010, “high trading volume is not necessarily a reliable indicator of market liquidity.”
This relates to the concept now referred to as “disappearing liquidity”, where there is a marked imbalance between executable liquidity and net executed volume. This is not always well understood. For example, some execution venues offer members “hit rate” scores as evidence of the benefit of interacting with HFT. From an HFT perspective, the hit rate is the number of times the short term prediction method was correct and within an accepted confidence level. As such, it is no surprise that slower/less sophisticated traders have higher “hit rates” from both the venue and HFT perspective. Whilst what a client “hits” does matter, what a client “misses” is crucial to understanding the real costs of interacting with HFT.
The International Organisation of Securities Commissions took this a step further in their recent HFT consultation document, noting that liquidity could be defined as the ability to “trade in large size quickly, at low cost and when market participants want.” HFT firms are frequently attempting to redefine themselves as “liquidity providers” and yet, HFT does not facilitate the ability to trade in large size quickly, having been shown to increase the total costs of trading and, due to HFT’s small quote size and disappearing liquidity, struggle to assist the institutional investor in allowing them to trade when they want. Larry Tabb, CEO of TABB Group exemplified this latter point in discussing recent volatility in the British press when he said: “HFT is indirectly to blame by removing vast swathes of liquidity from the market.” Some buy side dealers are now referring to HFT as artificial liquidity providers to emphasize the point.
Given these sensitivities and shifts in evidence based opinion, those supportive of HFT, perhaps for commercial gain and also those against HFT, may be best placed to put aside any inherent bias and focus on evidence based assessment for the wider good of the long term market. Recent pro HFT commentary has, unfortunately, been verging on the more extreme end of the subjective curve, often off the back of out of date research. Regardless, what is objective should eventually win out. Indeed, that trend seems to have gained momentum in recent months as more evidence is gathered and more intellect is deployed. The progress of this move to put a finer understanding of HFT into a wider context has much to offer the industry now and in the future.
Stuart Baden Powell is Head of European Electronic Trading Strategy at RBC Capital Markets. He sits on the Securities Trading Committee for the Association for Financial Markets in Europe and is an associate editor of the Journal of Trading.
High Frequency Trading
STUART BADEN POWELL, HEAD OF EUROPEAN ELECTRONIC TRADING STRATEGY, RBC CAPITAL MARKETS
Originally published in the September 2011 issue