Most Active Stories
Trading Algorithms Make Stock Market Sensitive To Hoaxes
Originally published on Thu April 25, 2013 4:57 pm
ROBERT SIEGEL, HOST:
That market dip yesterday that Steve mentioned, the result of a hacked Twitter feed, highlights how much the financial industry relies on computer algorithms. U.S. stock markets lost $200 billion in value in just a few minutes. The markets bounced back when the Associated Press made clear there was no explosion at the White House and the tweet was a hoax.
AUDIE CORNISH, HOST:
Last year, malfunctioning software at the Knight Capital Group cost that company $440 million. And don't forget about the infamous flash crash in 2010 when the Dow dove 600 points. Now Scott Patterson joins us to talk about this. He's the author of the book "Dark Pools: The Rise of the Machine Traders and the Rigging of the U.S. Stock Market." Welcome, Scott.
SCOTT PATTERSON: Thanks for having me.
CORNISH: So, Scott, we saw the markets dip here just because of a tweet. I mean, how exactly do these electronic algorithms work?
PATTERSON: OK. So basically, what you have are these algos that are constantly monitoring social networks such as Twitter or Facebook or corporate websites and they're constantly pinging those sites for new information. On Twitter, as I understand it, what they're looking for are trends. So if there's a keyword that keeps popping up that's linked to a company, like company X plus bankruptcy, and that keeps popping up, that might be a sign that something bad is going on.
CORNISH: So when you look at what happened yesterday, is this a situation where the algorithms went haywire? Or did the humans do better than the machine traders?
PATTERSON: No, it's difficult to tell exactly what happened yesterday because there was a period, about 15 seconds after hit, before the market started cascading. And, you know, it is theoretically possible that there were some people who saw it who started selling and that just picked up speed. So I think right now we don't know for sure whether these were computers who detected this and started selling or whether it was people. But when you look at the speed of the decline, there's no question that, eventually, automated trading kicked in.
So many of these firms watch the market itself. And if the market starts moving, they'll start going in the same direction. Like, if there's a big drop, they'll get out of the way. They'll just sell. And that's the exact same thing that we saw in 2010 during the flash crash when a huge amount of selling caused a bunch of firms to pull out of the market, and that triggered a devastating collapse throughout stocks.
CORNISH: So is anyone questioning this as - whether it's a good idea? I mean, should these algorithms be recalculated?
PATTERSON: Well, you know, really, it's the evolution of the market, and I don't know if there's anybody who can do anything to stop it. You know, I, myself, have been talking to regulators since late afternoon yesterday and today about whether there's anything they can do, and it seems like their feeling is that, you know, this is just the market. It's just faster. And what can you do? Can you, you know, tell firms to not sell when something like this happens?
You know, what if, you know, one of these days, you know, something like this is actually true? But it is risky because you have an entirely new news medium rising up which is Twitter, which apparently has some very gaping weaknesses in it if an AP feed can be hacked and you have these rapid fire automated trading firms that can trade millions of dollars in the blink of an eye that are hooked up to it. So it is kind of something that people need to stop and think about but, you know, whether it can be stopped is a very difficult question.
CORNISH: Scott Patterson, thank you so much for speaking with us.
PATTERSON: Thanks for having me.
CORNISH: Scott Patterson, he's a staff reporter with The Wall Street Journal and the author of a book called "Dark Pools: The Rise of the Machine Traders and the Rigging of the U.S. Stock Market." Transcript provided by NPR, Copyright NPR.