The question of what to do, if anything, about high-frequency trading has been much in the news in recent weeks, largely as a result of Michael Lewis’s new book on the subject (which I haven’t read yet). It is a subject that I’ve thought about a fair amount and about which I remain somewhat agnostic. Perhaps you might find it interesting why, despite much consideration, I have no strong feelings on the subject.

As a starting point, of course, there should be a strong presumption in such matters that if all the transactions are consensual and there are not externalities, then the activity should be legal. “Price discovery,” broadly defined, is a socially-valuable activity in the abstract that increases social wealth. And at first glance there doesn’t seem to be anything inherently wrong with what these guys are doing.

My initial positive instinct toward these guys was strengthened a few weeks ago when Goldman Sachs’s president came out in favor of restrictions on HFT. It isn’t fail-safe, but in general I’ve found it to be a pretty reliably use a rule of thumb that any policy that Goldman Sachs supports (especially issues on which Goldman has a direct financial stake) is probably driven almost exclusively by rent-seeking to benefit Goldman, not the public interest. So, everything being equal, Goldman taking a public position on one side of an issue generally makes me more inclined to discount the logic of that position.

But yet I remain agnostic; here’s why: In economics, there are certain economic activities that although they produce wealth for the parties undertaking the activity, they produce no social wealth. These are activities where parties invest real resources for a purely distributive gain (increase their size of the slice of the pie) rather than increasing overall social wealth (increasing the size of the pie). In fact, because it uses economic resources to bring about the transfer, the net effect is to shrink the size of the pie overall, because resources are diverted from higher-valued uses to lower-valued uses.

When this happens, it is a form of rent-seeking. Rent-seeking is usually used to describe this phenomenon in the specific context of government policy — for example, an industry  (say the steel industry) hires lobbyists and undertakes other similar useless activity to pass a steel tariff that increases the wealth of the steel industry, but simply by transferring wealth from consumers of steel products.

But the phenomenon extends outside of government rent-seeking. As Gordon Tullock noted in his justifiably-famous article “The Welfare Costs of Tariffs, Monopoly, and Theft” the same phenomenon can occur outside government, using the example of theft. The social cost of theft is not merely the redistribution of my TV to the thief, it is all the social costs surrounding that activity — costs in an opportunity cost sense. So, for example, it is the cost of the thief spending time casing my house waiting for me to leave so he can break in, rather than going to work at 7-Eleven or Kroger. It is the skills and activity by the fences of stolen property, who conduct business designed to conceal the theft. The burglar alarm makers, the lock makers, the raw materials for those things, the police who work so stop this redistributive activity, etc.

The point is that although these are voluntary transactions — buying locks and burglar alarms — in the end they are occasioned by the fact that what underlies all this is theft, a purely redistributive and not wealth-enhancing activity. (There is actually a larger point here about Keynesian economics and the “seen and unseen” that I will save for another day but that some readers will recognize). In a similar vein, the economic argument against legalizing blackmail (leaving aside the moral argument for the time being) is that legalizing blackmail would provide people with incentives to invest their time and energy into collecting information with no social value — after all, the whole economic logic of blackmail (from the perspective of the blackmailer) is to essentially collect information and sell it back to the blackmail victim. While the blackmail transaction is voluntary (the victim pays the blackmailer not to publish the information) that does not mean that the activity is social valuable. One could apply the same logic to a host of transactions, including professional gambling (as opposed to recreational gambling) and the like.

Now, to be sure, there are elements of rent-seeking in ordinary competitive activity — when Apple makes a better iPhone, in part they expect to make money by redistributing Android phone users to Apple. But they compete by making a better product at a better price — so the end result is to increase social wealth, despite elements of rent-seeking.

Ok, so with that wind-up, here’s the central point for HFT: Is HFT like the thief or like Apple?

To put it another way, HFT spends huge (unbelievably huge, actually) amounts of money to make trades faster. Is this purely (or largely) to make a distributive gain to itself (rent-seeking) or does this increase social wealth by increasing the efficiency of markets? Clearly, across some margin, there is no doubt that faster processing of orders would increase market efficiency — and there seems to be little doubt that electronic trading and other innovations over the past decades have increased the efficiency of markets by increasing the speed of trading. But HFT, as I understand it, seems to turn on milliseconds of trading speed — it is hard at first glance to see how that can really contribute to increasing market efficiency, or at least increasing market efficiency to a sufficiently great degree to justify the amount of social resources (human, physical, and financial capital) put into making trades a millisecond faster. My friend Brian Mannix has written a paper arguing along those lines. Thus, despite the fact that Goldman Sachs favors restrictions on HFT, it might nevertheless make sense from a social perspective.

Having said all that, I was given a pause when I read this article on Bloomberg, which presents the structure of an argument for why HFT is not merely rent-seeking but could be efficiency-enhancing. This is the first time I’ve seen this particular argument, and so perhaps there is a flaw in it that will be developed by others (links to intelligent critiques or elaborations are welcomed):

Then along come some smart young whippersnappers who replace gut instinct with statistical analysis. Instead of relying on some equity trader’s ample gut to guess what orders are likely to move the market, you can use a computer to figure it out, and then automate how you trade. If you build your computer right, you won’t be blindsided by informed orders that go against you, like a lot of mortgage-backed securities traders were.
And this means that you can quote a much tighter spread: By being smarter than the competition, you can also be leaner and more efficient. You might quote Microsoft at $39.97 bid and $39.98 offered on every exchange, knowing that as soon as someone hits your bid on one exchange, you can instantly move your market down to $39.96/$39.97 everywhere else. This reduces your risk of being picked off by trading with informed traders, which lets you make a profit even on much narrower spreads. You can charge less to trade by being more informed.
There are two ways of characterizing high frequency trading. In one, HFTs are front-running big investors, rigging the game against them and making the stock market illusory. In the other, HFTs are reacting instantly to demand, avoiding being picked off by informed investors and making the stock market more efficient.7

And what about Goldman’s interest in all this?

In my alternative Michael Lewis story, the smart young whippersnappers build high-frequency trading firms that undercut big banks’ gut-instinct-driven market making with tighter spreads and cheaper trading costs. Big HFTs like Knight/Getco and Virtu trade vast volumes of stock while still taking in much less money than the traditional market makers: $688 million and $623 million in 2013 market-making revenue, respectively, for Knight and Virtu, versus $2.6 billion in equities revenue for Goldman Sachs and $4.8 billion for J.P. Morgan. Even RBC made 594 million Canadian dollars trading equities last year. The high-frequency traders make money more consistently than the old-school traders, but they also make less of it.


Anyway, that’s how I see the issue — and why I’m on the fence.

What does this mean about regulation? To me, this means first that the state of knowledge is sufficiently ambiguous that at the current time I’m not persuaded that there is a case for regulation. But second, it also suggests a case for experimentation — which is to say that there may be a market for different markets, some of which permit HFT and some that don’t. If HFT increases market efficiency, then you’d expect parties to gravitate to those markets. If not, then you’d expect the opposite.

That conclusion, of course, is just an application of Hayek’s idea of competition as a discovery procedure — the value of competition among markets here would be to discover through competition what is otherwise difficult or impossible to determine through a prior reasoning. For an elaboration of Hayek’s logic of competition as a discovery procedure in this sort of context, you can check out the first section of this article of mine from a few years ago. It is through the dynamic discovery procedure of competition that we find out the answer to the question whose answer is ambiguous as an a priori matter. Even if one concludes that HFT is likely not wealth-enhancing, it does not necessarily follow that it should be banned, if there is competition among exchanges.