Wed November 7, 2012
Ask A Banker: High Frequency Trading
Originally published on Sun November 11, 2012 1:58 pm
Hi! I'm back. I was once a banker and now I write for Dealbreaker and answer your questions about the financial world here. You can send questions to email@example.com with "ask a banker" in the subject line, or ask on Twitter (@planetmoney). We've talked before about using financial tools to game the world a bit, and this week's question is in a related line:
Q. Do you think high frequency trading is truly helpful for an efficient market?
Oh, I don't know. Let's start way way way way way back.
Why is there trading? Start with this: If you're a company, you sell stocks or bonds because you want money to do something. You've got a great idea to build a car or an iPhone app or a cure for cancer, and doing that requires building factories or hiring developers or running tests, and doing that requires money, and you've got the idea but not the money, so you go to "the market" — a bunch of people who invest money for themselves or for other people — and you ask for money and those investors give it to you.
Why do they give it to you? Primarily because they like the idea: you convince them that you're actually going to sell cars or apps or cancer cures, and if you do you'll make a lot of money, and if that happens you'll be rolling in money and give them enough money to roll in too.
But secondarily because they're in this thing called "the market." Which means that they can sell the stocks or bonds that they bought for money. Let's focus on stocks for now. A share of stock is "perpetual," meaning that if I bought stock from Facebook in its initial public offering this year, I paid Facebook $38, and they never have to pay me back. Some stocks pay "dividends" - they give you a bit of money every year for owning their stock - but many, including Facebook, don't, and have no plans to.* Basically I've given money to Facebook and that's it.
What am I, an idiot? I mean, kind of? But in one or five or twenty or a million years Facebook will wind up its accounts and its shareholders will participate in some way. If Google buys Facebook for $40 a share, each shareholder will get $40. If Facebook starts paying dividends, each shareholder will get dividends. If Facebook goes bankrupt, each shareholder will get $0.
What that means practically is that right now your share of Facebook is worth something — not $38! but something! — because people are making calculations about what it will be worth in the future, and taking into account the probabilities of it being worth zero or $40 or $1,000 or whatever at some point in the future, and the risks and delays before that happens, and adding that up in some way to get a value today. Which is $21.17, an oddly specific number for all that estimation and uncertainty.
This is amazing.** This is in some sense what finance is; it's transforming the future into the present and the present into the future. Facebook will take over the world and make a zillion dollars, or it will fail and make zero dollars, or any number of intermediate outcomes, and you integrate across those possibilities and you get $21.17. $21.17! To the penny!
We'll talk about high frequency trading in a minute, but to understand the stakes you need to realize that this is magic. And it's magic because there's a system of trading that allows for at least pretty frequent trading. If you couldn't turn Facebook shares into money at some time between the time Facebook was founded and the time it winds up its business, you'd never invest in Facebook. And the more easily you can turn those shares into money, the more appealing it is for you to invest in new ideas.
The thing that lets you quickly transform an investment into money is called "liquidity." A share of Facebook stock is incredibly liquid: if it's trading at $21.17 now, I can sell it in 16.5 seconds for a price that will pretty predictably be between, say, $21.15 and $21.19. I may not want to do that, but it's nice to know I have the option. A house is less liquid; if Zillow tells me my house is worth $250,000, I can probably sell it within six months for a price between $175,000 and $300,000, or something.
Sometimes you want investments, sometimes you want money, and the more easily you can transform the one into the other the happier you are. So liquidity is good, mostly.
So: high frequency trading. High frequency trading is just a super-concentrated form of trading, done so quickly that it is impossible for humans to keep up so it's done by computers instead. High-frequency trading computers constantly submit bids to buy stocks (and offers to sell them). Their goal is to make a profit by immediately reselling a stock for slightly more than they bought it for. If trading provides liquidity, HFT provides super-liquidity.
One reason for the discussion of liquidity above is to make you feel in your bones that liquidity is good and there's some presumption that more liquidity is more good. That presumption is rebuttable, but you need to rebut it. When people say things like "stocks should be for financing companies and you should buy and hold them and not trade them like maniacs," you should be suspicious. How often can you trade? If the ability to trade once a year is good, is once a month better? Once a day? Once a minute? A second?
Despite that bias in favor of liquidity, there are real problems with high frequency trading. Here are three big ones, and one thing to notice about them is that they are basically the problems with all trading, just speeded up.
1. High frequency traders are trying to trick people. A decent chunk of high frequency trading consists of submitting orders that aren't executed. Why does this happen? Essentially to fish out what's going on in the market. High frequency algorithms are designed to, among other things, ferret out whether there's a big order in the market.
If some big, "real money" mutual fund — Vanguard or Fidelity or whoever — buys a million shares of Facebook, that pushes up the price of the stock. If a high-frequency trader figures out that this is happening, they can rush in, buy some Facebook stock, and immediately resell it to the mutual fund at a slightly higher price. This takes money out of the pockets of mutual funds, which tend to hold your retirement savings, and puts it into the pockets of high-frequency trading firms, which tend not to.
But all traders are trying to trick each other. That is the skill of trading: your goal is to buy from people who undervalue a stock and sell to people who overvalue it. The algorithms just do it faster and more dispassionately than humans do, and probably spend as much of their time trying to trick other algorithms as they do trying to trick "real money" investors. And because the algorithms are competing to get business and information, they tend to offer to buy stocks at slightly higher prices and sell them at lower prices than other traders. This is good for ordinary investors.
2. High frequency trading is a waste of resources. It turns out that you can make a lot of money if you are a microsecond faster than the other guy, so you build super fast computers and hire super smart mathematicians all in the pursuit of shaving a few microseconds off the time it takes you to trade a stock. That is, like, a stupid thing for people to devote their energies to. Why not have all those physics Ph.D.s build spaceships to go to Mars or whatever?
But all trading is, in some sense, a stupid thing for people to devote their energies to. Just moving around stocks and money between people is a mostly-pointless activity in itself; if you buy my shares of Facebook stock no new investment happens, no new factories are built, no new diseases are cured. Each individual trade looks pointless from a social perspective. But the ability to buy and sell stocks efficiently makes it easier for new companies to raise money. The trades are mostly pointless, but the system of trading is valuable.
Incidentally, most people think that we do spend too many resources on this whole trading business. There's a wonderful recent paper by Thomas Philippon of NYU, who points out that, while the income of the financial industry has grown from ~5% to ~8% of the American economy since 1980, the value it provides, in terms of liquidity services and financing of businesses, has not grown nearly as fast. As he puts it, the "cost of financial intermediation" has actually gone up since the 1970s, unlike most industries where increasing use of technology has lowered costs. High frequency trading may be a part of this, but it's unlikely to be the main part; it is a symptom of the excessive dedication of resources to trading, rather than a cause.
3. High frequency trading causes instability. This is the scariest of the three problems.
It used to be that humans — called "specialists" — stood around at the New York Stock Exchange making bids and offers to buy and sell stocks. They had a cushy living and you paid them a lot to buy or sell stocks, but there they were, on the Stock Exchange, easy to find. When things got bad — when stocks were plunging — they had a hard time packing up and leaving.
Now the New York Stock Exchange is mainly filled with CNBC correspondents; while the specialists are still around, their function has been largely displaced by computers that charge less for buying and selling stocks. One reason that they can charge less is that they're computers, but another is that, when times get bad, you can't find them. Sometimes they actually get turned off.
When things look dicey and you want to sell your stocks, there's no one to buy them, since all the computers got turned off, so the prices plummet. Specialists made a nice living in exchange for taking the risk of keeping the market orderly when others freaked out; the computers make a slightly less nice living but don't take that risk.
Also, the computers are idiots! If Facebook last traded at $21.17 and I tell my broker "just sell as fast as you can," and he gets to the market and the best bid is $21.10, he'll reluctantly sell. But if the best bid is $0.01, and there's been no bad news about Facebook, he will probably wait a second and see what's up. Some computers will just sell at $0.01. This is an unimaginably dumb example but it actually happened to some stocks (not Facebook).
This, too, is in some ways just a symptom of all trading. One bad thing about liquidity is that it creates volatility: if you can buy and sell constantly, prices will move around a lot. If you can't, they won't.*** Much of the time these price movements are okay: they allow supply and demand to balance, and let everyone know how much things are worth and how much money should be devoted to them. But in extreme cases ,big moves in the prices of financial things, even if they're caused not by economic fundamentals but by stupid computers, can actually cause serious harm. Companies can be destroyed, jobs can be lost, real people can be affected by computers' stupidity.
That's the main worry about high frequency trading: That, while it may marginally increase the efficiency of the liquidity machine that makes it easy to finance good ideas, it also increases the risk of catastrophe. Is that true, and if so, does the risk outweigh the benefit? Possibly, but it's hard to be sure. "Probability of catastrophe" is a hard thing to measure. Until, of course, the catastrophe happens.
* Here is what Facebook has to say about the matter:
We have never declared or paid cash dividends on our capital stock. We currently intend to retain any future earnings for use in the operation of our business and do not intend to declare or pay any cash dividends in the foreseeable future. Any further determination to pay dividends on our capital stock will be at the discretion of our board of directors, subject to applicable laws, and will depend on our financial condition, results of operations, capital requirements, general business conditions, and other factors that our board of directors considers relevant. In addition, the terms of our credit facilities contain restrictions on our ability to pay dividends.
** Perhaps my favorite financial writer, Nassim Taleb, says "that if there is a possible spot in time and space capable of bringing a profit, then the areas surrounding it need to account for that effect." That is the most important thing to understand about finance, probably, and the world, possibly.
Incidentally that quote is from this book, which is great but quite technical; his more accessible books include The Black Swan and Fooled By Randomness, which you should read but which mostly live in the everyday world rather than a space-time-profit continuum. Pity.
Also: this notion of prices today being based on expectations of future prices is very memorably explained in chapter 12, "The State of Long-Term Expectation," of J.M. Keynes's General Theory of Employment, Interest and Money. With a name like that it's gotta be boring but in fact the book is great and the chapter is among the high points of all economic, financial, or really any writing.
*** Think of the 2007-2008 financial crisis: in the past, if house prices dropped, that wasn't a big problem, because what are you going to do, you already own your house, who really cares that it's worth less than it used to be? But with banks trading billions of dollars of mortgages whose value was closely tied to house prices, and with <b>people</b> taking out billions of dollars of adjustable-rate mortgages that they could only afford to pay if house prices went up, houses stopped looking like, just, things you owned, and started looking like "liquid things that trade in the market." And so their prices could bounce around, and those bounces could have negative effects on the real world.