High-frequency trading: what are the risks and the returns for financial markets?

Hi! PARIS is the new Cen­ter on Data Ana­lyt­ics and Arti­fi­cial Intel­li­gence for Sci­ence, Busi­ness and Soci­ety cre­ated by Insti­tut Poly­tech­nique de Par­is (IP Par­is) and HEC Par­is and recently joined by Inria (Centre Inria de Saclay).

The term refers to the pla­cing of buy and sell orders on the fin­an­cial mar­kets over an extremely short time frame, on the order of nano­seconds, gen­er­ally in reac­tion to events. For example, when the Fed decides to cut interest rates, high-fre­quency traders will place orders a nano­second after the mac­roe­co­nom­ic announce­ment is made. 

Obvi­ously, humans can­’t react on this time scale. High-fre­quency trad­ing is char­ac­ter­ised by two phe­nom­ena. Firstly, it is auto­mated trad­ing: it requires the use of algorithms and com­puters to place orders. Secondly, it must react extremely quickly to mar­ket events, wheth­er mac­roe­co­nom­ic announce­ments or changes with­in the mar­ket such as price movements.

If we look at the Par­is Stock Exchange, for many years we had this image of traders shout­ing their orders around the trad­ing floor. This has not been the case since 1986, when the trad­ing floor was replaced by a com­puter RAM. Since then, traders have been car­ry­ing out their trans­ac­tions all over the world via online ter­min­als. This move towards auto­ma­tion began in the late 1960s and ended for equity mar­kets in the early 1990s. This auto­ma­tion led to so-called algorithmic trad­ing and, over the last twenty years, to high-fre­quency trading.

On the equity mar­kets, it’s estim­ated that around two-thirds of trans­ac­tions are car­ried out by high-fre­quency traders. This may seem high, but it’s not all that sur­pris­ing. It’s a bit like super­mar­kets: while they account for a large pro­por­tion of con­sumer trans­ac­tions in France, they are merely inter­me­di­ar­ies. High-fre­quency traders are not the final hold­ers of secur­it­ies, but the prin­cipals. They gen­er­ally only hold the secur­it­ies for a very short time.

It’s true that it has evolved con­sid­er­ably. The traders we often see in the movies, behind their com­puters or on the phone giv­ing their orders, have gen­er­ally been replaced by com­puters and algorithms. Today, most trad­ing algorithms are developed by com­puter sci­ent­ists. They are recruited by spe­cial­ised com­pan­ies, par­tic­u­larly in high-fre­quency trad­ing. These pro­files can provide a def­in­ite com­pet­it­ive advant­age, as they have expert­ise in writ­ing trad­ing-spe­cif­ic codes and have developed expert­ise in optim­ising mar­ket access speed.

While high-fre­quency trad­ing requires an algorithm with the right strategies, it is true that the mar­kets are well aware of them. It’s just that the time scale for put­ting them into prac­tice is much short­er. Optim­ising this speed of access to the mar­ket is there­fore cru­cial, because reac­tion time is essen­tial: the oper­at­or with the shortest reac­tion time has an undeni­able com­pet­it­ive advant­age. These tech­nic­al skills are very spe­cif­ic and gen­er­ally con­cern people who have trained in com­puter sci­ence, for example. If there is one area of the eco­nomy where we can talk about Big Data, it is the fin­an­cial mar­kets. Every day, they gen­er­ate an enorm­ous amount of data in digit­al form. Every trans­ac­tion, every order sub­mis­sion is a new piece of data. For high-fre­quency trad­ing, it is there­fore neces­sary to sub­scribe to data feeds that are sold by the exchanges. This requires highly advanced tech­nic­al expert­ise and tech­no­lo­gic­al invest­ments that are costly both in terms of human cap­it­al and computers.

Investors like you and me don’t have enough cap­it­al to access the oppor­tun­it­ies these investors do when mak­ing these invest­ments. The prin­ciple of high-fre­quency invest­ing is to obtain mod­est profits on each trans­ac­tion. These are very small profits, but because the trans­ac­tions are repeated so many times, the profits are very large very quickly. If the invest­ment is large, risk-tak­ing is min­im­al, because the strategies are well known and reac­tion time is optim­ised to make the most of oppor­tun­it­ies. The unpre­dict­able part is rel­at­ively small. As a res­ult, high-fre­quency trad­ing firms have expan­ded rap­idly over the last ten years. The profits that can be made are now much lower.

There are sev­er­al fam­ous cases of instabil­ity caused both by algorithms and to the det­ri­ment of algorithms, includ­ing the 2010 flash crash in the United States. The stock mar­ket crash las­ted just over 30 minutes and was not dir­ectly due to high-fre­quency trad­ing, but it did cause these oper­at­ors to stop their algorithms, which prob­ably amp­li­fied the crash. And then, as always with algorithms, there can be design flaws, bugs that were not anti­cip­ated. This was the case in 2012 for Knight Cap­it­al, a major US trad­ing firm and one of the four major high-fre­quency trad­ing oper­at­ors. One day, one of their algorithms developed a bug and Knight Cap­it­al quickly recor­ded a loss of 440 mil­lion dol­lars and almost went bankrupt.

One area we still don’t talk about very much is the risk of intru­sion by a plat­form or algorithmic oper­at­or. Ill-inten­tioned hack­ing, for example by ter­ror­ists, could ser­i­ously destabil­ise the market.

We can expect the trad­ing industry to use these tools one day. Some ini­tial thought has already been giv­en to this issue in a field oth­er than the fin­an­cial industry, namely online sales. Reg­u­lat­ors, not­ably in the UK and the US, have shown that there is col­lu­sion between algorithms. Take Amazon’s pri­cing algorithms, for example: they are foun­ded on arti­fi­cial intel­li­gence based on learn­ing first, a self-learn­ing pro­cess for repri­cing products. But they don’t just raise or lower prices depend­ing on the con­text, they find ways to charge non-com­pet­it­ive prices. If they are not coded for this, they dis­cov­er ways of mak­ing profits by cre­at­ing impli­cit collusion. 

Optim­ising the speed of access to the mar­ket is cru­cial, because reac­tion time is everything.

This has not yet become an issue for the fin­an­cial mar­kets, but it soon will, because it is part of the same ques­tion. In 2017, Black­Rock, one of the largest Amer­ic­an asset man­agers, began using arti­fi­cial intel­li­gence tech­niques to fore­cast returns and make port­fo­lio alloc­a­tions. The firm also relies on big data as part of its quant­it­at­ive and tra­di­tion­al invest­ment strategies. As for the invest­ment bank JP Mor­gan, it has already announced the devel­op­ment of a fore­cast­ing tool based on arti­fi­cial intel­li­gence to anti­cip­ate the decisions of cent­ral banks.

Interview by Jean Zeid