How to Disentangle the Impact of HFT on Market Liquidity from Other Factors?
There are a number of challenges for evaluation of the influence of HFT on market quality. First, it is very difficult to disentangle the impact of HFT on the market quality from other technological and regulatory innovations which led to substantial changes in the market structure of many trading venues, e. g., decimalization in U. S. equity markets. Then, it is also difficult to identify HFT traders even if researchers have access to agent-resolved data, which is extremely rare (Kirilenko et al. 2011; Hagstromer and Norden 2013; Benos and Sagade 2012). One explanation suggests that many traders pursue a variety of strategies which both provide and take liquidity depending on market conditions. Then, it is possible that a single trading account represents an omnibus account which may be used by several different agents, e. g., in the case of sponsored market access, provided by financial intermediaries to their clients. Next, much research in this area refers to HFT as a certain homogenous entity, while there is a multitude of trading strategies that have a different impact on the market quality. Sometimes this is due to the above-described difficulties in identifying the HFT traders.
Nevertheless, the positive changes in the market quality are sometimes attributed to HFT as a whole. In this case, as rightly noticed by Tse et al. (2012), it is possible that the positive effects of some HFT strategies (e. g., market-making and statistical arbitrage) outweigh the negative effects produced by other strategies, thereby masking the negative side effects of HFT on market quality. For instance, academic studies examining the phenomenon of HFT consider market-making strategies (Hendershott et al. 2011; Kirilenko et al. 2011), unwittingly spreading their effects on HFT as a whole. It would be better to consider the impact of peculiar trading strategies equipped with HFT technology on the market quality. One can then try to analyze the combined effect of the strategies considered on market quality. Ideally, one should conduct a comparative analysis of market quality with and without HFT (Hendershott et al. 2011). Unfortunately, it seems pretty far from feasibility.
It is reasonable to pose the question how to determine the impact of HFT on market quality when at the times of appearance and increasing use of HFT there were many other significant changes in the market structures which could not but affect the nature of trading in financial instruments. Even if one finds correlation between the increase in HFT and improvement (deterioration) in some market quality, correlation is not necessarily causation. “The challenge is to measure the incremental effect of HFT beyond other changes in equity markets” (Jones 2013). Ideally, one would like to track changes in the market structure which led to an increase in the proportion of HFT in the market, for example, autoquote dissemination on the NYSE in first half of 2003 (Hendershott et al. 2011). Then, one needs to compare the state of the market before and after the changes have occurred. It is advisable in this case to establish a causal link between the increase in the share of HFT and changes in the market structure of some trading venue. What is really important is what metrics will be used to reflect a certain quality of the market. Moreover, the task can be complicated by the fact that the metrics often do not reflect all aspects of some dimension of market quality, especially dealing with market liquidity. In other words, one needs to be more careful in the conclusions and not to make hasty statements in the spirit of “post hoc ergo propter hoc”.
“HFT is not a strategy but a technology” (WFE 2013) that facilitates the implementation of many traditional trading strategies whose effects on the market quality vary considerably. In other words, the nature of the trading strategies is likely to determine the effects of HFT on market quality rather than computerization of these strategies in itself. HFT liquidity providers, in fact, have replaced many of the traditional market makers which became less effective in highly automated order – driven markets. It is obvious that the algorithms are better at monitoring market conditions and adjusting orders than humans, for example, specialists on the NYSE. According to Biais et al. (2010), the machines are more effective because they obviously have no problems with limited attention or concentration required for simultaneously implementing multiple tasks. Undoubtedly, algorithms are much better at detecting and eliminating arbitrage opportunities, reducing, in fact, their lifetime to a few milliseconds (Sorkenmaier and Wagener 2011), to be precise, up to the time delay of the signal (latency) on a given trading platform. Ideally, when assessing the impact of HFT on market quality, one should not consider HFT as a whole but focus on individual trading strategies, using HFT technology. At the same time it would be great to establish whether the use of this high-speed technology exacerbates the problem generated by the strategy, e. g., market manipulation. However, there is a problem herein with the fact that myriad of strategies can be used by market participants, including those who process information on the number of financial instruments and simultaneously trade on multiple trading venues. Even having access to the source code and scripts underlying certain trading strategies it can be difficult to determine their behavior in real markets.
As with any technology, HFT can bring more good than harm, or vice versa. Ideally, appropriate application of technology can enhance market quality which significantly reduces liquidity premium and subsequently the firm’s cost of capital. Therefore, the question of the prohibition looks at least weird and can be even viewed as an attempt to stop scientific and technological progress. However, one should understand how to behave with particular classes of trading strategies using HFT technology to gain a competitive advantage. In other words, it is necessary to highlight key points, i. e., a close attention must be paid to particular trading strategies, while the technology for their application should be considered from the point of view of the possibility of worsening the market quality. However, it is clear that disruptive behavior exists and can be reinforced with HFT technology. Thus, it is important to thoroughly consider these scenarios, stay aware of their consequences and be ready to eliminate or mitigate their negative effects.
HFT strategies can be divided into “good” and “bad” according to their relation to the short-term mispricing (Tse et al. 2012). Those strategies that profit from detecting short-term mispricing and correcting it should be referred to the “good” strategies that improve the market quality. Strategies that profit from the creation of short-term mispricing and its subsequent removal may be considered as “bad” strategies, which have a negative impact on market quality.