Something strange is happening at the world’s biggest stock exchanges. Share prices are rising, the sums of cash being invested are getting larger, but the average value of trades — the “bargain size”, in the jargon — is getting steadily smaller. Investors are trading smaller blocks of stock more frequently.
The average trade size on SETS, the London Stock Exchange’s electronic order book, has fallen by two thirds over the past five years, from US$105,000 to US$35,000. Over the same period, the annual volume has multiplied more than fivefold, from 8.6 million trades to 46.9 million.
The immediate explanation for this atomisation is that electronic markets — the screen-based SETS replaced the telephone-based SEAQ quote-driven system in 1997 — are unsuited to trading large parcels of shares. The hidden cause is that big investment banks increasingly are adopting algorithmic trading, the latest application of computer science and quantitative finance to stockbroking.
Its explicit aim is stealth — the harnessing of computing power to trade large volumes of shares rapidly and automatically in a way that is imperceptible to the wider market.
The best way to remain anonymous is to slice up large orders into many small ones and drip-feed them into the market at intervals governed by a software program.The most frequently used analogy is of dropping hundreds of pebbles into a lake rather than a boulder, creating faint ripples rather than a big splash.
“We want to make ourselves as invisible as possible,” Peter Sheridan, head of algorithmic trading for Europe at Goldman Sachs, said. Some algorithmic trading programs will buy back quantities of the shares that they have just sold — or vice versa — to cover their tracks. Such techniques, pioneered on Wall Street, have been embraced in the City. Goldman, Credit Suisse First Boston, HSBC and Morgan Stanley all have dedicated algorithmic trading teams in London.
Tabb Group, the financial consultancy, estimates that 11% of American share trading this year was handled by algorithms and predicts compound annual growth of 34% over the next two years. Estimates for London suggest that, while the equivalent figure may be 5% or less, the rate of growth should at least match that in the United States.
Sheridan highlights two reasons for algorithmic trading. First, fund managers must focus on transaction costs after the Myners report for pension funds said that stockbrokers should “unbundle” the cost of trading from that of investment research within their fee structures. Secondly, the European Union’s Markets in Financial Instruments Directive (MiFID) is due to come in next year and its “best execution” tenet means that stockbrokers must show that they have achieved the best price for their clients.
Algorithmic trading seems to win on both counts. The commission is lower than for conventional trading and anonymity can make the implicit cost of trading lower; investors can secure a better price when buying or selling, because the market has not moved against them. A high degree of automation means that it can monitor multiple markets — national and regional exchanges and “alternative” trading systems such as Instinet — assisting compliance with MiFID.
With fund managers pushing fees sharply lower, algorithmic trading also enables stockbrokers to improve productivity by using fewer traders to handle greater volumes. However, the greatest beneficiaries of algorithmic trading have been stock exchanges themselves. “It has given them a massive boost”, Roberto Rivero, director of Intelligent Growth, a financial markets consultancy, said.
He added: “It has multiplied the number of orders sent to an exchange, which, given their fee structure, has increased revenues … And it has fuelled a demand for historical tick-by-tick share price data, which until recently only exchanges bothered to keep.”
Algorithmic trading could also prove a threat. Exchanges charge for information on a per-seat basis. “If one machine replaces many traders, the exchanges will lose revenues under their current charging model,” according to Rivero. There is also the threat that machines, less constrained by habit than human traders, will seek alternative markets.
Follow us on
