Competition is a good thing, even when there are “winners”
As part of my weekend reading, I came across a study by Professors Goldstein, Kwan and Philip condemning “HFT” based on two flawed premises. The first is that the only way to measure the impact of HFT firms is during times of volatility and the second is that a goal of financial markets should be all investor types experiencing equality of outcomes for every type of trading action.
The first premise, of looking solely at volatile times is flawed for multiple reasons. First, aggregate spreads and total volume traded, ex ante, are a consistent predictor of trading costs, while volatility in (the U.S.) market has not changed in the High Frequency era. The study ignores the point that today’s order driven markets introduced the concept of order competition, which, along with decimalization, had the effect of lowering spreads and dramatically increasing trading volumes. It is true that faster competitors will be capable of spread capture at a far higher percentage of the time than slower ones (which is a point of this study), BUT, without the competition, volumes would be lower, spreads would widen and those same institutions that lose out to HFT would be worse off. Thus, when fear mongers such as Themis Trading trumpet this study as proving how HFT is detrimental, they ignore the impact of the competition itself. Put simply, while this study proves that firms with superior technology have the ability to both provide and remove liquidity at a profit, it ignores the benefits of the added volume and tighter spreads that gave rise to those profits.
The second premise, that equality of outcomes is the goal, is based on their undocumented claims about “crowding out” of other orders by HFT. They do a great job showing how HFT orders do get better priority and experience better outcomes, but that is irrelevant. Since there is no proof that even a fraction of the trading interest from HFT would be replaced by increased trading from institutions and retail investors if HFT were to be banned, the study is meaningless from a policy perspective.
The other part of this study worth mentioning is that they seem to claim to have “discovered” that order book imbalances predict short term price movements and “prove” HFTs take advantage of that fact faster than other participants. Long before the term “HFT” became part of the lexicon, however, order book imbalance was a primary indicator for short term price movements, which I know from personal experience building algorithms almost 20 years ago. A “study” proving that being faster to react to market data that shows such imbalances is essentially like a study proving that a National Football League team would win a football game against a local Pop Warner team…
That said, their study does prove something important, which derives from a key statement from the world of economics: There is No Such Thing as a Free Lunch. In the context of modern equity markets, this translates into: Institutions should always assume that their trading costs, at a minimum, are ½ of the bid offer spread. The reality is that market intermediaries, whether modern HFT firms, or the specialists or Nasdaq market makers they replaced, have advantages to be able to capture the spread and other participants should expect to pay it for the immediacy that they require. Of course, it needs to be stated that HFT firms operate in a hyper-competitive environment, without any barriers to entry other than cost, while the model that existed previously conferred structural advantages to the market makers.
To put this in historical context, before the modern era in the U.S. equity market started in the late 1990s, with the order handling rules and the evolutions of ECNs, institutions could not even place limit orders to capture spread. (Even Instinet, the first U.S. display system for orders, for a long time did not accept institutional orders directly.) The HFT era was ushered in by that ability, made possible by order driven markets as opposed to the Nasdaq model of dealer quote-driven markets. Nasdaq dealer trading systems in the previous era did accept customer limit orders, but they only executed buy orders when the offer dropped to that price (and vice versa for sell limits). The NYSE order book, meanwhile, was only known to the specialist (and whoever they told when asked). To understand how that worked in practice, one of the most infamous “jokes” about the NYSE in that era was: Q: What is the best way to get the specialist to execute your limit order? A: Try to cancel it!
That said, there is a point to the study that institutions should heed, which is that slower broker dealers or institutional directed orders are unlikely to achieve spread capture without incurring substantial opportunity costs. Unfortunately, as I have pointed out in past commentaries, to measure those costs, institutions must analyze all the unfilled orders routed to the market on their behalf. The SEC could, and should, help, by reforming Rule 606 and creating a template with relevant statistics for analyzing routing behavior.
More nonsense about HFT
I also came across a confused story about HFT and other topics on ValueWalk this weekend. This article blends an assortment of facts and reasonable opinions with misleading statements and unsubstantiated conjecture to blame problems in the equity markets upon automation. Most important, it ignores the dramatic reduction in trading costs (and the corresponding order of magnitude increase in trading volumes) since the “rise of the machines.” Right off the bat, this article equates all algorithms with practices such as “quote stuffing, spoofing, price manipulations” despite these being illegal and constantly monitored by the regulators. This is particularly galling to read since human traders used to “spoof” and “layer” markets as standard operating procedure; I remember watching junior traders being trained on the old Salomon desk, where the senior traders told them “if you get a large order to buy, place your quote just above the offer.”
The article then goes on a screed against central bank manipulation, which as an Austrian monetarist, I am sympathetic to, but I stop short at its next conclusion, that the central banks manipulate markets by machines. Next, it talks about flash crashes, where their twisted logic turns on itself as it blames HFT, as opposed to poorly constructed order books without features like “Limit Up / Limit Down.” This is like blaming the heart pumping for a gunshot victim bleeding to death… HFT participants certainly profit and trade actively in the aftermath of flash events, but the order(s) that cause those events lose money, and there is no evidence that any of them were directed by machines. (NOTE — I do not dispute that the market is more prone to “flash” events, but as I wrote in a paper several years ago, that is more symptomatic of the market penalizing firms from placing many orders away from the NBBO.)
Next, this article points out that trading algorithms are vulnerable to being manipulated (using a “spoofing” case as proof) and uses that to argue that market vulnerability to manipulation is higher. There is some truth to this, as unsophisticated algorithms, from both HF and traditional firms can certainly be tricked into trading badly. Of course, there is no evidence statistically that this is any worse than how badly human traders can be fooled, as anyone who traded in 1987 or during the internet bubble can attest. That said, when it gets to its conclusion, that the market has become more homogenous, and is, therefore vulnerable to a “look out below” event, they may well be correct. That said, it is not automation, per se, that is the danger, but rather a combination of old fashioned greed and the “madness of crowds.”