The noise of money

A thunderbolt struck American stock markets on 6 May 2010. But it was no conventional lightning strike that wreaked such devastation. It was a ‘flash crash,’ a lightning-fast plummeting of stock prices, which temporarily wiped hundreds of billions of dollars off the screens of traders: more than the annual national output of Belgium. Although most stock prices bounced back quickly – stock exchange money generally does not vanish into thin air; it just changes location – the lightning strike caused people around the world to realize how vulnerable securities markets have become as a result of the forward march of computer-based stock exchange trading.

Enlarge

In the search for the guilty parties, suspicion soon fell on a relatively new practice at the time: high-frequency trading (HFT). LMU researcher Ryan Riordan investigates what effects the purchase and sale of securities within fractions of a second have on financial markets.

A native of Canada, Riordan remembers well the flash crash of 2010. Two months previously, the US technology stock exchange Nasdaq had given him a dataset on trading orders, which he was in the process of evaluating in May 2010.

“Getting access to such material in the first place makes up about a third of our working hours as researchers,” explains the financial scientist with an eyeroll. But his efforts to procure and study the Nasdaq dataset were worth it. He discovered that some of the assumptions he had had about high-frequency trading were wrong.

It took about four years for the corresponding essay to be published. If stock markets have become the epitome of speed and dynamism, then the wheels of financial science often turn slowly. But the thoroughness more than justifies the wait, asserts Riordan. In this spirit, the head of the LMU Institute for Financial Innovation and Technology cautions HFT skeptics against being in their judgments precisely what they accuse high-frequency trading of being: too fast.

Algorithms fan digital flames

There is no doubt that the automatic buying and selling of securities played an important role during the flash crash of 2010 and in similar stock market shocks. After all, computer programs, which buy stocks en masse and rapidly offload them again according to certain algorithms, can accentuate price movements. That being said, it has since been demonstrated in the case of the flash crash in 2010 that it was an individual who set off the avalanche – and did so deliberately. In April 2015, a private financial trader from London was identified as the main culprit. He later confessed to placing bogus sell orders in order to benefit from lower prices.

Because it was clearly one individual and not an out-of-control algorithm that triggered the 2010 flash crash, Ryan Riordan does not see this case as an indictment of high-frequency trading. For the LMU researcher, high-frequency trading is the logical continuation of a development that began at the end of the 1990s.

At that time, banks and stock exchange operators worldwide were facing new competition in securities trading. “They’d had a monopoly in this area, and this was no longer tenable,” explains Riordan. From the turn of the millennium, securities trading in the United States, Europe, and Asia opened up to companies and private individuals. At the same time, orders were no longer transacted by telephone or by hollers on the trading floor. Soon the technique prevailed that has become standard today, with computerized data exchange bringing buyers and sellers together. Meanwhile, securities traders realized that if you could place buy and sell orders at a rate faster than the reaction times of humans, you could make a lot of money. High-frequency trading was born.

Today, a large proportion of orders on international exchanges are transacted using HFT. A handful of companies little known to the general public have made this business their own. But in addition to these specialists in ultra-fast trading, large financial institutions are also in on the action. They can all benefit in different ways from high-frequency trading.

Striking in under a thousandth of a second

When price differences arise on different securities markets, this can be exploited. Take foreign exchange trading, for example, where prices today are globally synchronized virtually in real time. But if, say, the US dollar is a smidgen cheaper at one trading venue than at another for even just a fraction of a second, a high-frequency trader can exploit this price difference before a human would even notice it. This is possible because HFT is capable of executing trades at mindboggling speeds – under a thousandth of a second in some cases. For years, high-frequency traders have been locating their computers directly at the stock exchange so as to minimize the electronic transmission times of their orders.

As well as exploiting the tiniest fleeting price differences, high-frequency traders can structure the algorithms by which computers automatically trade securities such that they recognize and respond to market movements extremely early. When a major shareholder in a company offloads a lot of securities, for example, the price usually falls, even if often just for a brief moment. In such cases, HFT algorithms can strike much faster than human traders.

A procedure that was inconceivable before the advent of computer trading is ‘pinging.’ This term comes from the world of submarine technology: With a noise, the ‘ping,’ and its echo, submarines are able to locate ships or other submarines. High-frequency traders employ pinging to rapidly send orders and immediately cancel them again. If, say, a large investor is planning to obtain a larger position in a certain stock, pinging can be used to discover what the highest price is that this investor is willing to pay. A company that is engaged in HFT can benefit accordingly.

Riordan makes an important distinction here. There are illegal practices that use HFT techniques to manipulate market movements, certain varieties of pinging among them. And then we have the kinds of HFT that are regulated by the official bodies of the various international trading centers. Riordan thinks that the latter have firmly ensconced themselves in securities trading and are here to stay.

Ever since high-frequency trading was invented, critics have been warning that it distorts the prices of securities, because people do not make the buying and selling decisions, but computer algorithms. Following extensive investigations, however, Ryan Riordan has arrived at a clear conclusion: High-frequency trading can help discover the price that best reflects the interests of buyers and sellers.

A key metric in securities trading is the bid-ask spread, or ‘spread’ for short. It quantifies the distance between the bid price and the ask price. If this gap is wide, the market will have no consistent handle on the appropriate price for a security. In this context, experts refer to the ‘noise’ that overlays the actual market price.

Riordan wanted to know what proportion of a price movement is made up of ‘noise’ and what role high-frequency trading plays in this. “It took a very long time before we could calculate this,” he recounts. “You see, it is very difficult even to develop an algorithm that can distinguish between the proportion of a price movement made up by information – for example, about the business performance of a company – and the proportion made up by noise.”

His results were clear and unambiguous and not what he was expecting. High-frequency trading significantly reduces noise on the stock market. Riordan therefore concludes that HFT plays an important positive role in reducing arbitrary distances between ask and bid prices.

A warning against out-of-control algorithms

When considering the future of computer-based securities trading, however, Riordan has concerns. Although he is convinced that controls introduced in recent years to regulate HFT are largely working, developments are so rapid that it is becoming increasingly difficult for humans to control the machines. “20 years ago, the programs were still very simple,” he says. “Now we’re entering territory where algorithms act in ways their programmers hadn’t actually envisaged.” Situations could arise, for instance, where the computer program always pushes prices down immediately when somebody wants to sell a security and always pushes them up when somebody shows an interest in buying. And that would not always work in favor of adequate pricing on the securities market.

And because high-frequency trading reacts blindly, as it were, to business news and political developments, the potential for damage from false reports will continually increase, cautions Riordan. If fake news is fed into the securities market, there is the risk that prices will leap up or down in an irrational manner. Over time, this could lead to general uncertainty in securities markets – with far-reaching consequences: “Where there is uncertainty, the human reaction is first and foremost to do nothing. But in Europe, and specifically in Germany, we currently have the problem that many people are already not engaging with securities and that not enough people are investing in stocks.”

Text: Nikolaus Nützel

Prof. Ryan Riordan is head of the Institute for Financial Innovation and Technology at LMU’s School of Management. Riordan studied business at Carleton University in Ottawa, Canada. He obtained his doctorate at Karlsruhe Institute of Technology (KIT) with a thesis on the economics of algorithmic trading. He was assistant professor at the University of Ontario Institute of Technology, Oshawa, Canada and then full professor in the Smith School of Business at Queen’s University, Kingston, Canada, where he was also research director of the Institute for Sustainable Finance. He was appointed professor at LMU in 2022.

Read more articles from "EINSICHTEN. Das Forschungsmagazin" in the online section and never miss an issue by activating the e-paper alert.
Or subscribe to the free print version of EINSICHTEN (in German).