Quotulatiousness

July 7, 2013

Trying to prevent another “flash crash”

Filed under: Business, Economics, Technology — Tags: , , , — Nicholas @ 10:57

Tim Harford discusses high speed trading and its potential problems:

“High-frequency trading” is a rich environment of algorithms, of predators and prey, all trying to make money by trading financial products at tremendous speed. But the basic proposition is simple to state. When the price of a share rises in New York, the price of related contracts will rise in Chicago just as soon as the news arrives. But if everyone else gets the news on the regular cable, and you’re renting space on the faster cable, you can see into everyone else’s future by (say) 0.7 milliseconds, plenty of time to buy soon-to-rise assets and then, less than a thousandth of a second later, to sell them again.

You don’t have to be a socialist to find this kind of thing discomfiting. There are three concerns. The first is that scarce resources are being spent on high-speed connections that have no social value in what is at best a zero-sum game. The second is that high-frequency traders may be making money at the expense of fundamental investors. The third problem is that such trading appears to introduce systemic risks. The “flash crash” of May 2010 is still poorly understood, which should ring alarm bells — especially since the need for speed means most high-frequency algorithms are simple and therefore stupid.

What, then, should be done? Rather than trying to slow down the algorithms, why not slow down the market? Most financial exchange markets run continuously, effectively assuming that traders can react instantaneously, withdrawing out-of-date offers and replacing them with up-to-the-picosecond prices. It’s this flawed premise — that all trades could be instantaneous — that means that no matter how fast the computers get, there will always be an incentive to go faster still.

A simple way for an exchange to improve matters would be to run an auction once a second, batching together all the offers to buy and sell that have been submitted during that second. Unsuccessful bids and asks would be published and would remain on the books for the next auction, unless withdrawn. One auction a second ought to be enough for anyone; it would deliver a stream of well-behaved data to regulators — currently unable to figure out what is going on — and it is plenty of time for a computer to weigh its options.

No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.

Powered by WordPress