How to Evaluate the Impact of HFT on Particular Aspects of Market Liquidity?
In general, market liquidity can be defined as ability to trade when you want to trade (Harris 2002). To be more specific, a liquid market can be described as a market where participants can rapidly execute large-volume transactions with a small impact on prices (BIS 1999). Even the last definition is not precise enough, since it’s not clear what the following expressions mean: “rapidly execute”, “large – volume transactions” and “small impact”. In order to somehow evaluate such elusive characteristic of market quality, Kyle’s approach is usually applied in market microstructure research (Kyle 1985). Its key idea is to consider separately three different aspects of market liquidity: tightness, depth, and resiliency. Tightness is the cost of opening and closing a position over a short period of time. It is well characterized by the bid-ask spread. Depth denotes the volume of incoming order required to change the price a given amount or the total amount of orders in limit order book. Resiliency refers to the speed with which market recovers from a random, uninformative shock. Next, we consider the impact of HFT on each of these aspects of market liquidity.
Tightness. High-frequency traders (HFTs) have largely replaced traditional market makers because they are able to post more competitive quotes, thereby providing tighter bid-ask spread. In market microstructure research the bid-ask spread is usually decomposed into three following components: order-processing costs, asymmetric information costs and inventory-carrying costs (Huang and Stoll 1997). Technological and regulatory changes gave to these “new” market makers (Menkveld 2013) an advantage over the traditional ones. Firstly, they can intermediate trades at lower costs. This is partly from the automation of trading process that has brought lower costs comparing with manual execution of trades. Therefore, they have smaller order-processing component of their bid-ask spread. However, they obviously have to incur additional costs of developing, testing and maintaining algorithms as well as make investments in hardware and software to implement them. Perhaps, there are some economies of scale in this case. Recently, HFTs most likely had to incur serious start-up costs which were likely reimbursed afterwards by profits from market-making due to speed advantage over traditional market makers. It is possible, that this suggestion is supported by researchers’ findings on increases in realized spreads and other measures of liquidity supplier revenues (Hendershott et al. 2011).
Secondly, due to automation of access to markets they can react more quickly to any new information about multiple financial instruments. Furthermore, computerized trading strategies have no problems with concentration or tiredness when monitoring market conditions. So, they can make timely response to a relevant event, if any. Thus, they reduce their exposure to the risk of being picked off by informed traders that, in turn, reflects in smaller asymmetric information component.
Thirdly, HFTs are also more efficient in inventory management due to holding relatively small positions, keeping them for a very short period of time, not carrying inventories overnight and having more diversified portfolio because of trading more financial instruments. Thus, they reduce their exposure to the inventory risk that, in turn, reflects in smaller inventory-carrying component.
Therefore, all of these effects lead to narrower bid-ask spreads. Many studies (Angel et al. 2010; Hasbrouck and Saar 2013; Hendershott et al. 2011; Kirilenko et al. 2011) support the narrowing of bid-ask spreads up to the size of the minimum price increment (tick). Obviously, it witnesses improving of tightness. It turns out that market participants, arranging small-volume transactions (i. e., not exceeding the quoted depth), have benefited as their transaction costs have dramatically reduced.
Depth. And what has happened with profitability of market participants making large-volume transactions, i. e., exceeding the quoted depth? It seems that this question cannot be answered definitely. On the one hand, they also have benefited from lower bid-ask spreads, thereby reducing the value of their implicit transaction costs as for the volume not exceeding the quoted depth. On the other hand, some market participants indicate a decrease in the depth of the market, linking this phenomenon primarily with a decrease in tick size (e. g., decimalization in U. S. markets). Under these new conditions it is much easier to rearrange limit orders to advance in the queue. This process resembles leapfrog, as Larry Harris describes it (Harris 2002). In this case, implementing front-running strategies becomes cheaper, since one more tick (i. e., one cent or penny nowadays) does not significantly increase the costs of execution of trading strategies. It turns out that it makes little sense to put a large amount of limit orders as faster market participants can easily stand ahead in order to benefit from this situation (see “quote matching” for more information on this type of front-running). As a result, many market participants reflect their trading intentions less intensively in their limit orders, thereby increasing the amount of hidden liquidity, hanging over the market.
For the sake of justice it should be noted that before the proliferation of HFT submission of large-volume limit orders often affected adversely on the financial performance of their initiators. Thus, they also could become a victim of one of the front-running strategies. The difference with the past is the following: under the conditions of small tick sizes in markets pursuing such “parasitic” (Harris 2002) strategies has become much less expensive, i. e., front-running has become more feasible. Moreover, some market participants have a priori advantage in speed of access to the market, which affects the distribution of the balance of power between the market participants. In other words, without the same technology withstanding front-running and protecting the value of the embedded option in limit order from extraction is much more difficult. Submitting a limit order, especially with large volume, this market participant provides to other traders, in fact, a free option (Copeland and Galai 1983). In the event that it becomes “in-the-money”, HFT traders having technological advantages in speed of access to the market will be able to extract its value faster than the participant will be able to cancel this limit order. In such circumstances, limit order submission is a luxury. Thus, the displayed depth of the market likely has most likely deteriorated, but first of all it depends on the tick size, which determines the profitability of the front-running strategy.
Tightness improvement together with depth deterioration has an ambiguous effect on the implicit transaction costs of the market participants. On the one hand, it decreases the value of the bid-ask spread. On the other hand, it increases the costs of market impact. The final result will depend on the volume distribution between the components of implicit transaction costs. In order to have a positive result the additional gain referred to the volume below the quoted depth have to exceed the additional loss from walking the book. So, the market participant faces trade-off and needs to choose the volume to balance the bid-ask spread with costs of market impact.
Market impact depends on the structure of the limit order book, particularly on the distribution of volumes among price levels and the presence of gaps in the book. By the way, it is possible that the rest of the amount will be executed at prices that would have been inside the market when compared to the previous bid-ask spread, or worse just one penny (by reducing the tick size). In general, one needs to evaluate the entire magnitude of the implicit transaction costs for different levels of volume. It might be supposed herein that upon reaching a certain level transaction costs will increase, while remaining at a lower level in the new environment, and then they will exceed the previous total implicit transaction costs after passing that level.
It should be considered whether such volumes were traded in the past, i. e., before changes in market structure induced by HFT. Possibly, it wasn’t so feasible. Then, there is no question herein. Before the era of automated trading block traders used the services of intermediaries from upstairs market, i. e., the services of so-called block broker/dealers, or acted in the market through a single or multiple floor traders, who “quietly work the order”. It seems very plausible that the emergence of algorithmic trading (especially after the seminal work of Almgren and Chriss 2000) led to the switch of block traders from over-the-counter markets to organized markets where securities were primarily listed. In this case, the performance of block traders would be increased as it became possible to make transactions more quickly and at lower cost in comparison with those arranged at the discretion of the floor traders. In this case, there were eliminated any conflicts of interest associated with the use by floor brokers information about the positions of block traders to pursue their own interests to the detriment of the interests of their clients which is the breach of fiduciary duty. Can it be said that some of the increase in the total volume of trade associated with the appearance of algorithmic trading relates to institutional investors trading in large volumes (block traders)? Some researchers share this opinion, e. g., (Jorion 2007). However, the task of identifying block traders is extremely difficult if one has no access to agent-resolved data, because it requires integrating the small orders (“child” orders) in single “meta-order” (“parent” order). Although it is possible that in the data there are certain patterns, reflecting a presence of order-splitting strategy.
However, this has decreased the apparent depth of the limit order book, as shown above. Perhaps for some market participants it has become more difficult to sell their volumes, despite lowering bid-ask spreads. It is still necessary to check what outweighs: gain from narrowing bid-ask spread or losses from the reduction in the depth of the best available orders. Nevertheless, it is conceivable that most of the volumes which became relatively more expensive to trade were hardly traded before. It is possible that some of the volumes were executed for the next several price levels of limit orders, and now it takes more price levels, i. e., one has to walk deeper the book. However, the difference between two adjacent levels of ticks has most likely decreased. How to unravel this tangle: the decimalization led to an increase in algorithmic trading (it has become easier to rearrange the best limit orders, thus reducing the value of time priority as an order precedence rule). It is unlikely that anyone at once traded volumes of more than 5 % of the daily volume, a famous ad-hoc rule used by traders to determine market liquidity. However, it needs empirical checking. At the same time there are dark pools which are rather effective substitutes of upstairs markets. However, trading at dark pools, as a rule, is not conducted continuously. One might pose the question whether large volumes participate in the price discovery given that they are directed to the dark pools, where the price is usually taken from other trading venues. In other words, does this practice lead to the fact that not all the information is reflected in the price? Do we need then these dark pools? But in the upstairs markets the usual practice was almost the same. It would be necessary to compare the two regimes, which is practically impossible because of the lack of necessary data. Is it worth so worrying about reducing the depth of the market after all? We compare the current state of markets with what it was before, or with what we would like to have? Moreover, all the consequences of this “ideal” world we can hardly evaluate. And then there is the hidden liquidity, which is simply not reflected in the limit order book (partly also due to the front-runners), but always ready to join in the action (fundamental buyers and sellers in the terminology of Kirilenko et al. 2011).
However, there are studies that claim that the depth of the market has increased (Angel et al. 2010; Hasbrouck and Saar 2013). Can it be contributed to quote stuffing, at least partly? Submission and immediate cancellation of orders located deep inside the book really does not increase the depth of the market, unless the random order with large volume is executed during the lifetime of this fleeting order. However, not every methodology for measuring the liquidity of the market will be able to properly take into account (more precisely, to exclude from consideration) this effect. Most likely, the market depth metrics, taken on various time frames, would demonstrate an increase in depth even after aggregating.
Resiliency. Even if the depth has decreased, then the slower execution of the order (implying a splitting of the total order into smaller parts) may reduce market impact costs subject to a decrease in time of market resiliency. There is evidence that the algorithmic liquidity providers closely monitor the situation with the anomalous expansion of the bid-ask spread. In this case HFTs promote liquidity replenishment due to speed advantage over other market participants (Brogaard 2010). Furthermore, this effect can even be used to detect HFTs, i. e., detecting who submit limit orders (which tighten the bid-ask spread) immediately after execution of market or marketable limit orders which led to widening of the bid-ask spread.
However, under certain conditions (e. g., during the Flash Crash large amounts were systematically dumped to one side of the market) liquidity providers become liquidity takers (Kirilenko et al. 2011). It can be brought about by pursuing their intentions to keep the level of inventory in the area of the preset target which leads to a further increase in the bid-ask spreads and a sharp price movement in an unfavorable direction. Put forward the assumption that up to a certain level of the bid-ask spread new liquidity providers contribute to the resiliency of the market, and above this level there comes a realization of systemic risk. So, liquidity providers face another trade-off. However, under normal market conditions the time of resiliency likely decreases considerably, thus compensating to a certain extent the decrease of the market depth.
Thus, under these new conditions order-splitting has become even more meaningful. For those who want to immediately sell significant volumes there are different dark pools. Rather, in terms of fragmented liquidity there will be some combination of rational order splitting and the use of dark pools. The answer to the question what and when to use will depend on the current market conditions, the rules of engagement into dark pools and pricing rules. Moreover, in order to improve the efficiency of this strategy one needs to split the block not only in time but also in space (across different trading venues) using smart order routing technology.
Thus, market microstructure theory identifies several positive and negative effects on market liquidity which could be produced by HFTs. Ultimately, determining net effect is an empirical question given that methodological choices are reasonable enough to reflect multi-faceted nature of market liquidity.
Almgren, R., & Chriss, N. (2000). Optimal execution of portfolio transactions. Journal of Risk, 3(2), 5-39.
Angel, J., Harris, L., & Spatt, C. (2010). Equity trading in the 21st century. Working paper. Carnegie Mellon.
Committee on the Global Financial System. (1999). Market liquidity: Research findings and selected policy implications. Bank for International Settlements – Monetary and Economic Department.
Biais, B., Hombert, J., & Weill, P. (2010). Trading and liquidity with limited cognition. Working paper. Toulouse University, IDEI.
Copeland, T.. & Galai, D. (1983). Information effects on the bid-ask spread. The Journal of Finance, 38, 1457-1469.
Hagstromer, B., & Norden, L. (2013). The diversity of high frequency traders. Journal of Financial Markets, 16, 741-770.
Harris, L. (2002). Trading and exchanges: Market microstructure for practitioners (p. 656). New York: Oxford University Press.
Hasbrouck, J., & Saar, G. (2013). Low-latency trading. Journal of Financial Markets, 16, 646-679.
Hendershott, T., Jones, C., & Menkveld, A. (2011). Does algorithmic trading improve liquidity? The Journal of Finance, 66, 1-33.
Huang, R., & Stoll, H. (1997). The components of the bid-ask spread: A general approach. Review of Financial Studies, 10, 995-1034.
Jones, C. (2013). What do we know about high-frequency trading? Working paper. Columbia Business School.
Jorion, P. (2007). Value-at-risk: The new benchmark for managing financial risk (3rd ed., p. 602). New York: McGraw-Hill.
Kirilenko, A., Kyle, A., Samadi, M., & Tuzun, T. (2011). The flash crash: The impact of high frequency trading on an electronic market. Working paper. University of Maryland and CFTC.
Kyle, A. (1985). Continuous auctions and insider trading. Econometrica, 53, 1315-1336.
Menkveld, A. (2013). High frequency trading and the new-market makers. Journal of Financial Markets, 16, 712-740.
Sorkenmaier, A., & Wagener, M. (2011). Do we need a European “National Market System"? Competition, arbitrage, and suboptimal executions. Working paper. Karlsruhe Institute of Technology.
Technical Committee of the International Organization of Securities Commissions. (2011). Regulatory issues raised by the impact of technological changes on market integrity and efficiency (60 pp.).
Tse, J., Lin, X., & Vincent, D. (2012). High frequency trading – The good, the bad, and the regulation. Credit Suisse. AES Analysis. Market Commentary, 5 December 2012. https://edge. credit-suisse. com/edge/Public/Bulletin/Servefile. aspx? FileID=23284&m=1815212669
World Federation of Exchanges. (2013). Understanding High Frequency Trading (HFT) (5 pp.).