What you’re seeing here is the price of shares in BATS, at 11:14 this morning. The white spots are trades: there are 176 of them altogether. They start just below the IPO price of $16, and then just fall lower and lower and lower until the stock is trading for mere pennies. But the key number you want to look at here is not on the y-axis. Instead, it’s the chart report at the very top:
Elapsed Time: 900 Milliseconds
This is what happens when stocks are traded by algorithms rather than humans. The parabolic trajectory of the share price is downright elegant; indeed, if you’re going to crash from $16 to 4 cents within 900 milliseconds, you could hardly do so in a lower-volatility manner. The scary thing here is the sheer speed involved, and the fact that no human intelligence was stopping to think whether these prices made any sense at all.
Of course it’s too early to work out exactly what happened here; a formal statement from BATS talks vaguely about “a software bug”. But the big picture is clear. Most people think there are only two stock exchanges in the US — NYSE and Nasdaq. And indeed those are the only two exchanges where stocks are listed. But there are more than 50 venues, including two different BATS exchanges, where stocks are traded; they all communicate with each other to work out what the best global bid and offer in any stock is at any given time. (This is known as NBBO, for National Best Bid/Offer.)
This fragmentation of trading venues is good for competition, but, as we saw first in the Flash Crash of May 2010 and then again today, when one of those venues encounters problems, very nasty things can happen.
BATS was meant — if everything had gone according to plan — to be the first stock listed on the BATS exchange. They’re not going to try that stunt again in a hurry; as finance professor James Angel told Bloomberg, this was “like seeing an airplane crash on takeoff”. On the maiden flight of a new airline. You can imagine how much appetite anybody would have to fly that airline thereafter.
One obvious similarity between today’s events and what happened in the earlier flash crash is that both involved exchanges declaring “self help” — basically saying that the information coming from some other exchange was so delayed or otherwise unreliable that it couldn’t be used any more as part of the NBBO system. When that happens, you can find order flow sloshing violently around various different exchanges; such moves don’t need to be accompanied by extreme price action, but they make such action much more likely.
There is some good news here. The first bit of good news is that no one was really harmed today: the BATS IPO has been pulled, and the institutions which were trying to sell their shares — foremost among them the estate of Lehman Brothers — will just have to hold on to them for a while longer. And the second bit of good news is that we have a lot of valuable real-world information about exactly how markets fail in today’s high-frequency precincts. I just hope that we’re going to be able to learn from what happened today, and put in measures to prevent it from being much worse next time. Can anybody say Tobin Tax?