London, 7 March 2011

Computer-driven trading has grown exponentially in commodities as techniques pioneered in fixed income, currency and equity markets have been applied to trading oil, natural gas and corn. Automated or algorithmic trading accounted for over 30% of all trades executed in crude oil in the fourth quarter, according to CME Group, and a similar proportion of completed trades in other energy contracts, as well as agricultural contracts such as corn. [download below]

Automated trading systems (ATS) accounted for a smaller share of completed trades than in interest rates (42 percent), equity (51 percent) and foreign currency (58 percent) derivatives but have been catching up fast and can be expected to expand further in the next few years.

Block trades, where a large number of contracts are executed at a single price for major counterparties, and trades for forward dates and spreads with limited liquidity are still dominated by human market makers and likely to remain so.

But algorithmic systems are gradually taking over routine market-making and liquidity-providing functions previously performed by floor traders and humans sitting at screens, at least for trades that do not have special size requirements or other forms of complexity.

Supporters argue algo trading has brought clear advantages in terms of narrower bid-offer spreads and a clearer audit trail. However explosive growth has made it a lightning rod for criticism.

SURVIVAL OF THE FITTEST

Early complaints centred on the perceived unfairness of renting space in exchange buildings to firms with algo systems to enable them to co-locate their servers with those operated by the exchange, increasing trading speed and reducing delays in decision-making and execution (" latency").

More generally, there have been concerns algo trading and co-location have triggered a technological arms race -- favouring firms that can afford to invest in expensive systems and financial engineering over other market participants.

Much of this is nostalgia -- the complaints of those history is leaving behind. Exchanges have never afforded equal access or a level playing field to everyone. Members have always enjoyed inbuilt advantages in terms of greater proximity and ability to execute orders faster than nonmembers.

Renting space to algo servers is no different from renting memberships and seats on the old trading floors. Early adopters may have been able to extract temporary and "unfair" first-mover advantages, but such benefits have likely been eroded over time. Exchanges have started to put in place more transparency and equal access rules for co-locaters, and regulators have started to require them.

Computer-driven systems are expensive to design and implement, so they raise barriers to entry of new firms. But cost barriers should diminish over time as the technology becomes more widely available. Critics understate just how high the barriers were in the good old days of pit trading.

There are concerns about specific practices employed by some systems -- including pinging the market with orders that are good for a only a few microseconds to probe the nature of the order book, with most cancelled unfilled, and sometimes as few as one in 30 actually executed.

More insidious are systems that allow algo programmes to glimpse orders for micro seconds before they are seen by the wider market. While market-makers have always had some preferential knowledge of the order book, this verges on abuse and should be outlawed.

QUALITY OF LIQUIDITY

The real concerns centre on the quality of the liquidity algos provide and whether the speed at which trading is executed threatens to outstrip safeguards established to prevent crashes. For critics, algo trading gives a false impression of liquidity. Computers supply plentiful liquidity in ordinary conditions, when it is least needed, but withdraw from trading when volatility rises, creating a liquidity hole and increasing the risk of flash crashes or price spikes.

To be fair, human market-makers also withdraw liquidity in a crisis. Even if market-makers continue to quote two-way prices, as many must, spreads may be so wide and quoted volumes so small liquidity effectively vanishes. The question is whether algo systems are more prone to withdraw liquidity than human traders, and if they are more prone to the type of herding behaviour that makes crashes and spikes more likely.

Evidence is mixed. Flash crashes and spikes do appear to have become more common recently. Some have been attributed to herding behaviour among the computer-trading community. One response has been heightened interest in circuit breakers in all markets, helping prevent machines trading with one another in a self-reinforcing crisis and giving time for human intervention by traders, exchanges and regulators.

The fact algorithmic trading systems have to be preprogrammed (limiting their flexibility during a sudden surge in volatility compared with human counterparts who may form more diverse views) and according to statistical regularities (which means they may all respond and try to trade the same way) is a source of concern.

In his book "The Quants," Scott Patterson argues convincingly that herd behaviour among the big computer-driven trading systems contributed to the flash crash in U.S. equity markets in August 2007. They were all programmed to exploit the same statistical regularities so they all ended up holding the same positions.

TRUST -- BUT VERIFY

In an interview with Reuters, CME Chief Executive Craig Donohue described concerns about high-frequency and algo trading as generally "specious".

Asked about the quality of liquidity, he argued "I don't think you can place a value judgment on it. Even a crowd follower or scalper contributes to liquidity in some way, perhaps not as valuably as someone who is lifting bids or taking offers, but they still have a role to play.

"The real question is have we put in place the systems, the protections, the protocols, the speed bumps, the rules and processes that allow us to maintain orderly markets and prevent meltdowns? I think speaking for the CME Group, it's clear that we have.

"Donohue dismissed liquidity concerns too lightly; they are certainly not specious. Like the other exchanges, CME has a vested interest, since algo trading has been a major contributor to volume growth.

Time and again markets have shown that in periods of rapid innovation there is a tendency for trading and investment techniques to run ahead of the appropriate regulatory framework, resulting in a crisis and then reform.

So rather than closing down debate (as exchanges and algo firms have sometimes seemed to want) or indulging in conspiracy theories, the market would benefit from an extended and open debate about how to harness the benefits of automated trading while minimising the possible costs.

The first requirement is for more transparency about the proportion of trades done by automated systems so their impact can be understood. So far, all research has been performed by those with a vested interest. The market has remained almost completely opaque to regulators and the wider public.The U.S. Commodity Futures Trading Commission last summer revived its Technology Advisory Committee to study computer-driven trading and update the regulatory framework.

In the past six months, CME has begun publishing quarterly data on algo trading volumes for a handful of flagship contracts as well as broad product categories. It has provided a welcome indication about the scale of algo trading,

but far from the level of detail that would enable outside observers to reach conclusions about the impact on liquidity. Algorithmic trading is here to stay. On balance it probably provides significant benefits. In any event it can no more be rolled back than cottage weavers could return to the putting out system of clothing manufacture by Luddite machine breaking.

But a series of trading disruptions in markets as diverse as equities, oil and cocoa indicate that innovation has outstripped the regulatory framework. Rules need to be updated to ensure it does not increase the risk of flash crashes and spikes.

Ends --

By John Kemp, Reuters market analyst – for Commodities Now.

The views expressed here are his own.

Attachments:

Algo_Trading_Update_Q4.pdf

48 Kb