Felix Salmon passes along this animation that shows the remarkably chaotic growth of high-frequency trading on all of America’s various stock exchanges over the past five years (this includes the three exchanges you’ve heard of plus the dozen others you probably haven’t). The basic idea behind HFT is that humans are taken out of the trading equation entirely. Instead, computer algorithms trade stocks directly, executing millions of trades per second and occasionally going crazy, as they did during the Flash Crash of 2010 and then again a few days ago, when an HFT bug cost Knight Capital $440 million in 30 minutes. Felix wants it to stop:
Back in 2007, I wasn’t a fan of a financial-transactions tax; today, I am. And this chart shows better than anything why my opinion has changed. The stock market is clearly more dangerous than it was in 2007, with much greater tail risk; meanwhile, in return for facing that danger, society as a whole has received precious little utility. Are spreads a tiny bit tighter than they might be otherwise? Perhaps. But that has no effect on stock-market returns for long-term or even medium-term investors.
The stock market today is a war zone, where algobots fight each other over pennies, millions of times a second. Sometimes, the casualties are merely companies like Knight, and few people have much sympathy for them. But inevitably, at some point in the future, significant losses will end up being borne by investors with no direct connection to the HFT world, which is so complex that its potential systemic repercussions are literally unknowable. The potential cost is huge; the short-term benefits are minuscule. Let’s give HFT the funeral it deserves.
I agree. The problem with HFT isn’t that we know it’s dangerous, it’s that we don’t know anything at all. It’s become flatly too complex for even its creators to understand what their creations are doing. Here’s an example. The heart of HFT is speed: even the speed-of-light delay can make a difference, so most HFT shops locate their computers as close to the stock exchanges as possible. Even a few milliseconds can make a difference. At least, that’s what a company called UNX thought until it moved from Burbank to New York:
This is where the story gets, as [Scott] Harrison put it, weird. He explains: “When we got everything set up in New York, the trades were faster, just as we expected. We saved thirty-five milliseconds by moving everything east. All of that went exactly as we planned.”
“But all of a sudden, our trading costs were higher. We were paying more to buy shares, and we were receiving less when we sold. The trading speeds were faster, but the execution was inferior. It was one of the strangest things I’d ever seen. We spent a huge amount of time confirming the results, testing and testing, but they held across the board. No matter what we tried, faster was worse.”
“Finally, we gave up and decided to slow down our computers a little bit, just to see what would happen. We delayed their operation. And when we went back up to sixty-five milliseconds of trade time, we went back to the top of the charts. It was really bizarre. I mean, there we were in the most efficient market in the world, with trillions of dollars changing hands every second, and we’d clearly gotten faster moving to New York. And yet we’d also gotten worse. And then we improved by slowing down. It was the oddest thing. In a world that values speed so much, you could be slower, yet still be better.”
The problem here isn’t that UNX’s move failed, it’s that Harrison still doesn’t know why it failed. Until we do, allowing HFT bots to control our equity markets is just begging for a catastrophe.
And that’s where the transaction tax comes in. HFT works by making tiny amounts of money on a huge number of trades. Even a tiny tax, maybe a quarter of a percent per trade, would make HFT unprofitable and would put our markets back in the hands of human beings. Those human beings will still screw up, but at least there’s a limit to how fast and how badly they can do it.