In a recent op-ed ("High-frequency traders can't front-run anyone"), I referred to conclusions about front-running in the stock market by Michael Wellman at the University of Michigan as "blatantly false." Professor Wellman wrote a rebuttal specifically addressing HFT's use of "latency arbitrage," a term that means very different things to different people, to "effectively" front-run other investors. Luckily, he gave a specific example of what he means by latency arbitrage, so we can easily dismantle this alleged problem with our equity market.
In his rebuttal, Wellman admits that there is no front-running, as legally defined. Rather, he spends approximately 80 percent of his rebuttal claiming that "it does not matter much what we call it," but that HFTs are "in effect jumping ahead of incoming orders." I'll address the "jumping ahead" issue in a moment, but it is important to understand that word choice matters. I referred to Wellman's paper as "blatantly false." I could have used more loaded terms, like "fraudulent" or "libelous," but I didn't. Those terms mean something different. Like fraud and libel, "front-running" carries meaning. And the meaning has very real legal consequences associated with it.
If Wellman is worried about a flaw in market structure that's being exploited by some players at the expense of others, then let's call it an inefficiency that should be fixed and get to the meat of it. But to argue that "what we call it doesn't matter much" shows a gross lack of appreciation for dealing with the fallout of frivolous lawsuits, which are real and which are happening now, thanks to the likes of Michael Lewis and Michael Wellman.
Read MoreWhat is HFT?
As to the substance of his argument, Wellman gives this example of latency arbitrage: "Suppose an order to sell exists in exchange A at $100.00, but the outdated National Best Bid and Offer (NBBO) ask quote is $100.02. If an order to buy at $100.01 comes into exchange B (and stays there based on the outdated NBBO), the HFT can arbitrage between the two markets, buying in A and selling in B, pocketing $0.01."
Wellman has used a bunch of words to describe a "crossed market," where the best bid is above the best offer. I asked my brother Manoj Narang, CEO of the high-frequency trading firm we co-founded together, Tradeworx, to help provide some analysis and data on this subject.
It should be noted that we already have a regulation — Rule 610 (d) — that is designed to prevent locked (where bid and ask are the same) and crossed markets, and it does that job about as well as it is possible for a law to do it. Does Wellman want a second regulation to prevent it again? Or would he prefer that the crossed market persist forever? If a crossed market exists, it is an arbitrage opportunity, as he has rightly understood. The reality is that somebody will take that arbitrage out of the market, and that somebody will pretty much always be the fastest player.
There is nothing wrong with somebody taking out an arbitrage in the market. Look at it another way. There is no regulation that keeps the price of IVV in line with SPY. Yet, the prices are basically always in sync because of arbitrage. This is the point of arbitrage, and it has tremendous value for the market in terms of pricing efficiency.
That said, crossed markets are extremely rare. Based on Tradeworx's direct feed data, crossed markets occur 0.01 percent of the time. You are as likely to be struck by lightning (NOAA.gov) as you are to find a crossed market. The Iranian national soccer team is five times more likely to win the 2014 World Cup (at the time of this writing, according to skybet.com) than a crossed market is to occur. The reason is that it requires a several independently unlikely events to occur effectively simultaneously.
The first unlikely assumption in Wellman's scenario is that Exchange B has to be using an outdated quote (presumably from the SIP) in order to allow the buy order to post. But most exchanges use direct feeds, not the SIP, for this purpose. Thus, it is highly likely that B knows about A's quote, and therefore won't allow the buy order to post as a result, due to it violating Rule 610 (d).
The second unlikely part of this story is that, even if Exchange B is using the SIP, the offer of $100.00 and the bid of $100.01 both must be placed within a millisecond of each other. Why? Because that's how little the latency differential is between the SIP and direct feeds. If it takes longer for the second quote to appear than it does for the SIP to catch up (again, about one millisecond), then the ban on locked markets kicks in and renders it impossible for this crossed market to exist.
Third, not only does the exchange have to be using a slow feed in this example, but the trader placing the $100.01 bid has to be using a slow feed also. But more than 95 percent of passive orders are placed by HFTs, using direct feeds. Even if the sender of the order isn't using a direct feed, they place orders through a broker that does route using direct feeds. In this case, the broker would have routed this $100.01 bid to exchange A and lifted the $100.00 offer. Why? Because even after take fees, it would have resulted in an immediate fill at $100.003, versus posting passively at $100.01 in the hope that it gets filled for an effective price of $100.007, after receiving the exchange rebate for liquidity provision.
Fourth, for his setup to exist, it appears to require a $0.03 spread. By Tradeworx's count, less than 10 percent of U.S. equity volume is traded at a spread of $0.03 or wider, and in fact, about 83 percent of the volume (at median) is done at a $0.01 or smaller spread. Without the $0.03 spread on the NBBO, there wouldn't be room to play around inside the bid/offer spread with crossed markets in this entirely hypothetical way. Oh, and in order for this trade to be worth more than $0.01 per share (total, before costs, see below), we'd have to see this already-unlikely combination of scenarios play out repeatedly in sub-one-millisecond intervals.
Adding to the point, it is not especially profitable to pursue crossed-market arbitrage. The actual reward (per share) is 1 cent, minus 0.6 cents of take fees (the HFT would have to take liquidity on both his buy and sell orders), and another 0.1 cents of SEC fees, brokerage fees, trading-activity fees, etc. The net is around 0.3 cents per share. In exchange, you have to shoulder the risk that you will only get one side of the trade done. This is a likely enough scenario, because there are plenty of very fast players out there, and nobody is deterministically faster than everyone else at all exchanges. And for it to have any real economic value, you'd have to believe this extremely unlikely scenario plays out extremely frequently. If that's where you stand, there's this foreign prince who's dying to meet you ... he's got an amazing offer.
By all means, let's fix what's broken in the markets: The ban on locked markets should go. The SIP should be sped up. Special deals for unfair access should be prosecuted. Off-exchange transactions of exchange-traded instruments and payment for order flow should be scrutinized. A flattening of the tiers for providing liquidity also deserves consideration. There's plenty to fix and to analyze, and from what I can tell, the "new-look" SEC is doing exactly that. But that doesn't imply that the market is unfair, much less rigged. Irresponsible, factually false claims like these — especially shouted loudly by folks who carry lots of credibility with the public and politicians — have real repercussions. The only ones who'll make out are the lawyers. If Dr. Wellman and his ilk are so concerned about negative utility, they should consider this fact before misleading others.
Rishi K Narang is the founding principal of T2AM LLC, a hedge-fund advisory firm located in Los Angeles. He was also co-founder, with his brother, Manoj Narang, of the high-frequency trading firm Tradeworx. He is also co-author of the new book, "The Truth About High-Frequency Trading: What It Is, How Does It Work, and Is It a Problem?" Follow him on Twitter at @rishiknarang.
The regulation the author is referring to is Rule 610 (d).