My temporal mind has been whipsawed lately, as having just finished Piketty’s excellent tome on wealth inequality over the centuries—a book where a decade is a flicker of time—I’m turning to Michael Lewis’ book on high-speed traders cashing in on arbitrage opportunities measured in milliseconds.
In case you’re wondering the latter is all about, I’m thoroughly enjoying this paper by Budish et al from which the figure below is drawn. I think it provides a great way of seeing what’s going on here.
The figure tracks the prices of two securities, labeled ES and SPY, both of which track the S&P 500, and it tracks them over four diminishing time horizons. Clearly, over a typical trading day these two securities move in lockstep, as you’d expect. When equity prices are thusly correlated, there’s no obvious chance to make some money by selling one high and buying the other low. That tight correlation is still there on an hourly basis and even over the course of a minute.
But once you get to milliseconds, the correlation breaks down. Somewhere between 1:51 and 1:52, more precisely between the 39th and 40th second therein—even more precisely, I’d say around the 558th millisecond—you could sell ES high and buy SPY low.
Perhaps it’s not so surprising that correlations break down at millisecond intervals—kinda like the first time you looked at a leaf under a microscope, it looked nothing like a leaf. But the fact that technology allows high-frequency traders to take advantage of such arbitrage opportunities is what’s germane, and pretty amazing, here, as Lewis writes about so compellingly (and I’ve just started his book).
OK, so what’s the problem? Well, there are many. First, this type of trading ain’t exactly a model of your grandad’s stock market, the purpose of which was to allocate excess savings to its most productive use. It’s investing lots in hardware, software, and human innovation by smarter than average people to uncover millisecond price differences that have nothing to do with added economic efficiency in any way that at least I can recognize.
Moreover, according to Budish et al:
We calculate that there are on average about 800 such arbitrage opportunities per day in ES-SPY, worth on the order of $75 million per year. And, of course, ES-SPY is just the tip of the iceberg [I’m not sure why there aren’t many more than 800–JB]. While we hesitate to put a precise estimate on the total prize at stake in the arms race, back-of-the-envelope extrapolation from our ES-SPY estimates to the universe of trading opportunities very similar to ES-SPY – let alone to trading opportunities that exploit more subtle pricing relationships – suggests that the annual sums at stake are in the billions.
These authors point out ways in which such HFT actually reduces market liquidity (relative to a world without HFT) by exposing “liquidity providers”—traders selling stocks—to high-speed front-runners who can “pick off” stale quotes, like the elevated ES price in the millisecond panel above, before the sellers can adjust.
I’m working on something longer about this that I think my readers will find quite interesting. It’s not just technology that allows this to occur. It’s also lax regulation, and there are good, smart people thinking and acting on ways to stop it.
More to come.
Source: Budish et al.