Crash hazard of high-speed trades

It took the firm 47 minutes to realise it was bankrupt. Such was the world of high-frequency trading eight years ago, which was the technological dark ages in that it could take as long as a second to execute a trade. These days, trading times are measured in microseconds, the fastest being done in under 10 millionths of a second. So, were a similar mistake to be made today, the firm would go bust in the blink of an eye - in which time, incidentally, it would have carried out 30,000 trades.

Imagining this kind of speed is almost impossible but if people lived and worked at the pitch at which these computers operate, it would mean the end of the weekly trip to the supermarket. You could buy an entire lifetime of shopping for a family of four in one trip to Sainsbury's - and do so in less than a second. Mind you, getting it all into the car boot might take a bit longer.

The volume of shares trades in the US has quadrupled in a decade, meaning that the average time a share is held has shrunk from eight months at the beginning of this century to two months today. That hardly seems like investment. In 1945, it was around four years. Luddites like me instinctively think this pace and volume of share trading can't be right, or at least that it must have some uncomfortable side-effects.

It's a bit limp to say that somehow it's not natural - that's like saying that if God had meant us to fly, he wouldn't have given us the railways - but it does seem high-speed trading takes us into unknown territory where all sorts of strange things might happen.

And that is just the point because they already have, as in the American flash crash of just a year ago. That was when markets went mad for no apparent reason and $1 trillion of value was wiped off shares in about half an hour - although some of them actually went up. Shares in Accenture dropped by 99% from $40 to $0.01. At the same time, Sotheby's shares increased 3000-fold, from $34 to $99,999.99. Then, just as quickly, things went back to "normal".

Nothing like this had ever happened before but no one could pin it on high-frequency trading. No one, that is, until today. Overnight in Beijing, Andrew Haldane, the director of financial stability at the Bank of England, delivered another of those deeply thoughtful papers that help us understand what is happening in markets.

Last night's subject was the impact of high-frequency trading, and his suggested conclusions are far from comfortable. The good news is that in normal times high-frequency trading adds to market liquidity but seems also to increase volatility and correlations between markets. In bad times - meaning when there is a shock of some kind - the liquidity dries up, the algorithms feed on each other in a gigantic game of pass the parcel, and a flash crash can result.

During the actual flash crash, 27,000 contracts in one S&P 500 futures changed hands in 14 seconds, but only 200 were actually purchased and held. Haldane also suggested that the flash crash was not a unique event - it is the kind of thing you would expect to happen again and again.

He makes a further telling point. Such is the power and strength of high-frequency traders and the forces they unleash that normal buyers and sellers take great risks being in the same place as them because they are sure to get squashed when things run amok.

The implications of that are pretty serious. One of the first discoveries in banking 500 years ago, Gresham's Law, was that bad money drives out good. Now we have the modern equivalent: bad share-trading drives out good - or to put it less emotively, those who seek to hold shares for a fraction of a second are making life impossible for those who want to invest.

The maths behind this has no place in a family newspaper - we will leave that to the News of the World - but the prose tells the story in simple and highly convincing language, and raises an uncomfortable question.

High-speed trading has come from nowhere to become the dominant form of market activity in less than a decade. Speeds of 10 microseconds may well soon give way to speeds in nanoseconds - a billionths rather than a millionths of a second.

But if this trading is the potential source of systemic risk - which it certainly appears to be, in other words if it has the potential to impoverish us all by bringing the system crashing down - should regulators sit idly by with fingers crossed, as they do now, and let the people get on with it? Or should they do something about it?

The answer surely is that they should do something. As Haldane says, grit in the wheels of the market, like grit on the roads could help forestall the next crash.