Separately, and along with my colleagues Aaron Brown, Michael Mendelson and Hitesh Mittal, I have written several articles on high-frequency trading. We believe HFT has lowered investors’ trading costs and that many of the attacks on it are misguided. We have expressed concern that our fragmented, domestic trading infrastructure is technologically risky and due for a hard look, separate from allegations about HFT. Finally, we’ve noted the obvious: that while defending HFT broadly, we can't and wouldn't vouch that each HFT trader acts lawfully and ethically (nor would we do so for each Good Humor ice cream salesman).
Bloomberg View contributor Noah Smith recently wrote on HFT. I haven't responded to other articles on HFT, but Smith focuses directly on our writings and his overall conclusion is very similar to ours, though far more equivocal. So, I will make an exception and respond to Smith -- rather than to the legion of rather nutty pieces out there -- out of respect, not ire.
First things first. Smith calls me a “hedge fund magnate.” Although being a magnate sounds kind of cool, my firm -- AQR Capital Management -- is more of a diversified-investment manager than a hedge fund. We manage a lot of traditional funds as well as hedge funds. In the hedge-fund space, AQR is known for demystifying strategies and rationalizing fees (in other words, lowering them), so the whole label is a little off. Still, it’s more interesting and punchy than “the senior guy at a midsize investment manager” and clearly “magnate” is better for me than “plutocrat,” “oligarch” or “fat cat,” so maybe I should feel lucky.
Smith’s lead point is that I claim HFTs don't front-run because I used an outdated definition of the term. That is, I take “front-running” to mean using information you’ve been entrusted with and promised not to use to the disadvantage of whoever entrusted that information to you (usually a paying client). Smith says the definition must be broadened to encompass “order anticipation with speed advantages.” Therefore, he says, it is legitimate to say that HFTs engage in front-running.
He’s wrong. First, you can’t just change the definition of a word used to describe a crime and apply it to something that’s both legal and ethical. Perhaps all of us should have used the more straightforward term “guessing,” because that is all “order anticipation” means. Traders have always made educated guesses about who is going to buy and sell what, and they’ve always tried to do it faster than the other guy. They did this long before HFT. In fact, what are traders doing if not trying to make educated guesses about who plans to buy and sell what and then act before other traders? If these traders -- HFT or not -- are using legal, public information, this is exactly what they’re supposed to be doing. To say “we’ll call that front-running as we’ve broadened the term” is ridiculous.
Smith goes on to question whether HFT has lowered the so-called all-in costs for investors. He concedes that bid-ask spreads have come down, but rightly notes that costs must also include market impact, meaning the act of buying or selling enough shares to move the price before you have bought or sold all you want. We did note in a Wall Street Journal op-ed article that while we can only speak for ourselves, our all-in costs have come way down, and that we find it logical that HFT has played a big role in the decline. Next, we’d note that for small, self-directed investors, only the bid-ask spread is relevant because their orders can be filled without affecting the price. For the little guys, it’s a lock that their trading costs are cheaper than before HFT. Smith, to his credit, agrees with this assessment.
But Smith is creative! He advances the untestable hypothesis that perhaps fewer people are bothering to become informed traders because HFTs will just take their profits, and as a result markets are less efficient. As we have pointed out, HFTs shouldn't be judged against an unattainable nirvana of zero bid-ask spreads and infinite liquidity, but against the system that preceded it. Before the advent of HFTs, human market-makers engaged in order anticipation and moved prices away from large traders all while earning much fatter profits.1 Any informed trader discouraged because potential profits are syphoned away by HFTs would have been even more discouraged under the old trading regime.
Now here’s the rare point where we agree with Smith. He cites the staggering cost (italics Smith's) associated with HFT. We would note that HFTs are highly competitive with one another, delivering their services at much lower cost to investors than traditional market-makers who operated under a more monopolistic system. The profits HFTs make, and the costs they incur, have been wildly exaggerated in service of a pre-existing thesis. That’s not Smith’s doing, but by repeating this myth he’s just getting it wrong. The things he cites -- computing cycles, fiber-optic cables (and their sometimes accompanying tunnels), and the efforts of smart people -- are costly but still cheap compared with the monopoly rents investors used to pay. That’s what HFT has moved us away from (italics mine).
A common argument that parallels Smith’s is that HFTs are engaged in an unproductive arms race, spending too much to eke out that last ounce of speed. This echoes an old fight between capitalism and its ancient enemies. You could apply this to a legion of examples just by taking a stroll down the aisle of a U.S. supermarket. There you will find hundreds of varieties of breakfast cereal. Clearly, that’s a wasteful spoons race and a central planner could do better. This has been tried several times, and the central planners always seemed to decide on only three types of cereal, all tasting like borscht.
Let’s imagine the HFTs agree to stop the arms race. We can easily imagine this as an optimal strategy, if they are spending so much to deliver so little for investors. But it’s also a strategy that reeks of monopoly, or more precisely, oligopoly. Furthermore, it’s the argument monopolists always make -- that they are really only trying to create efficiencies and eliminate waste for the customer. Ironically, the same people critiquing the HFT arms race are often the ones who would be the first (and in some cases be right) to attack monopolies. The current competitive market-based solution is delivering the product, meaning liquidity for investors, better and cheaper than ever. Moving away from this competitive landscape would be an invitation for incompetent central planning or rapacious monopolistic practices (you know, like we used to have). The current system works and has avoided both dangerous extremes.
Smith goes on to the last refuge of the desperate, asking “what is the social benefit?” Watch out when people say this because they’re about to use it as a justification for telling somebody else what to do (OK, Smith doesn’t go that far and only suggests we think about it.) First, we don’t have to demonstrate a social benefit for a legal and ethical activity, just no large external social cost. Second, there happens to be a clear social benefit: lower expenses for investors. This means more money in the pocket of retirees and other investors, and less paid to middlemen. It doesn’t get much more direct than that.
Smith concludes with the observation that HFTs should be presumed innocent until proven guilty, but fails to note that we already have overwhelming evidence of the former. It may seem rational and wise to urge slow judgments, as Smith does, but not when the jury is already in, and the arguments for the prosecution have already been demolished.
1 There is a theoretical possibility that HFTs have lowered all-in costs a lot, but gotten so much better at detecting and avoiding informed traders, and that could lead to Smith’s theoretical outcome. But there’s no evidence for this, or a clear test for it, and what we do know is that costs have come way down and market-makers have always attempted to avoid trading with those who know a lot more than they do about the short-term.
To contact the author of this article: Clifford Asness at Cliff.Asness@aqr.com.
To contact the editor responsible for this article: James Greiff at firstname.lastname@example.org.