You are walking down the street, texting to a friend, when suddenly everything freezes. These things happen all the time, you reason, so as annoying as it is you reboot and carry on. A desperate text a short time later comes as a call from your friend to please stop bombarding them! What went wrong? You have no idea. You reboot again and keep walking.
Things like this happen to everyone these days and we’re all used to it. Software glitches. Bugs. Crappy software runs amok in the hands of appliance users.
Now imagine that you are a Wall Street trading firm that handles orders for thousands of clients and this happens to you. Except that this costs $440M in bum trades by the time anyone catches it. That’s exactly what happened to Knight Capital, the company that used to handle 11% of all trading on Wall Street. It’s something that was inevitable in a system that is too big to be useful – and the world is starting to realize how dangerous this is.
It’s not as though this is an isolated event. An unknown event that may have included software glitches was behind JP Morgan’s epic $20B (or so) loss in London. The facebook IPO was famously dogged by glitches that sometimes sold people twice as many shares as they wanted. There was also a “flash crash” in 2010 that sent the Dow down 1,000 points in less than a minute.
What do all these events have in common? The root problem is a trading system that relies on trades that move faster than the blink of an eye – too fast to be done by humans. The computers that look for any tiny advantage don’t hold stocks for long, with estimates of the average holding time on securities running around 22 seconds. It’s a search for a profit of a few percent a year boiling down to a few thousandths a day and a few millionths as each second ticks by. That is how traders make money these days.
In addition to being fast, the computers that make these trades have to be amazingly fault-tolerant. Unlike your buggy li’l smart phone, they are designed with an incredible amount of checking and lengthy protocols that keep them from going nuts. Of course, there’s still a one in a billion chance, or less, that something will eventually go wrong no matter what you do.
What’s amazing is that the people who play the infinitesimal advantages never saw the potential liability of even an infinitesimal chance of failure.
This is a case of a very small number (the probability of failure) being multiplied by a very big number (the cost of that failure), the most common type of problem we face in these highly engineered days. How much is it worth to have a system that reduces the chance of an error? If it goes from one in a million to one in a billion, and the cost of failure is a $1 billion, then such a system is worth roughly $1,000. It’s a simple calculation. Of course, it assumes that you can estimate these probabilities and costs up front with incredible accuracy. If the potential cost is more like $10B and the odds are more like one in a thousand, preventing errors is worth a lot more than you thought at first.
Risk analysis like this is never easy, but it becomes even more dodgy on the margins. It’s common for sophisticated companies like Knight Capital to have it all calculated to incredible precision. The alternative is a more general kind of insurance, a system with checks and balances built into it and perhaps a reserve of cash for “unforeseen events”. That kind of approach is now seen as inefficient and old-fashioned, given how we can calculate and predict failure modes right down to the smallest detail.
In other words, computers are really useful tools for people desperate to fool themselves.
If the Knight Capital event was isolated we could all smirk and go on with our business. But this is a common problem, something that crops up on a roughly monthly basis with some trading firm or investment bank somewhere. The $440M that Knight lost nearly did them in as a company, but it also had the potential to destroy the market for at least a day. The potential harm is shared by everyone.
Flash trading and robotic investment, as a standard, is not good for anyone. One easy solution is a tiny tax on trades, say 0.1% or even less, because as we all know that if you tax something you have less of it. This is one case where the market might benefit greatly from even a tiny tax increase.