Last week Knight Capital Group lost $440 million when it sold all the stocks it accidentally bought because a computer glitch. Well Group ($KCG) lost $440 million – that’s aLmost as much as Zuckerberg loses each day!

When this story first broke, it was a bit different from other “algos gone wild” stories we’ve seen in recent history: there were 140 stocks affected, and they were on both the buy side and the sell side. We can say that the large programming trading desk, and having heard numerous legends of traders who have done exactly the following, that someone at NITE (Knight Capital Group “NITE” throughout this piece – their OTC trading acronym) executed an order over the wrong time period. In other words, maybe the order was supposed to take all day, and they sent it down for 30 minutes by accident.

Anyway I don’t really see why this is such a scandal. As far as it appears a couple of things happened. Knight wrote some bad software that caused their accounts to way over-trade. An equal but opposite flood of liquidity satisfied their (software driven) demand.
Hence the price of the affected stocks remained stable (unlike in the Flash crash when the demand for volume was not met by a supply). The only party that suffered was Knight, which was the party that screwed up and should suffer. Some other (High Frequency Traders) HFT traders, market makers and stat-arb funds made a lot of money. Ordinary investors weren’t affected, and there’s no government bailout required.
Pretty much sounds like the exact way capitalism is supposed to work. The volume numbers on a bunch of stocks were unusually high, but as long as the prices are in line, I don’t see how that qualifies as a negative externality.
As for the issue of “abusive” orders those are largely misunderstood. The “silver bullet” indictment seems to be that a high proportion of quotes are cancelled before execution. The queue matching system already rewards the longest-lived quotes so describing cancellation as costless is erroneous.
The primary reason quotes get cancelled is because market makers are setting their prices off other liquid reference securities. E.g. a market making in MSFT is going to take into account the movement on SPY, E-minis, AAPL, IBM and XLK among others. Any time any of those securities tick it’s new information that potentially causes her to re-evaluate the price she wants to trade at. The market maker might then re-adjusted her bid/ask.
If it’s a less liquid stock she might re-adjust her quote hundreds of times before being filled, but that’s only because the broad market is moving substantially in between trades on the stock. To an outside observer this would look like hundreds of cancelled quotes for a single execution. And they’d call it something ghastly like “phantom liquidity.”
However conceptually it’s nothing more than a single quote whose price is continuously adjusted. It’s a good thing, because it keeps prices more in line with the general market, increasing market efficiency and reducing idiosyncratic volatility.
A better measure is the average potential fill size (the sum of the quantity on all active quotes in a given direction) versus average active inventory. Our heroine above would average a ratio near 1, she might cancel a lot of quotes but her intention is to fill everyone (at a fair price to her given the general market). So she’s not going to quote too large because if filled she would violate her risk controls. Therefore her quote sizes will stay in line with her risk or inventory sizes.
In contrast the true quote stuffing villain would have a ratio far above 1. He’s going to put in massive total quote size in order to scare the market and take advantage using small actual orders that minimize market impact. Thus he should have a very small inventory given his quote sizes, and hence a high ratio.
Obviously this measure isn’t available since
1) we can’t see people’s inventory and
2) even if we could we don’t know which quotes are related to which parties.
However a fair approximation of how this value is changing over time is to compare bid/ask touch quantity against volume. Volume should scale with average inventory size of market-making/HFT traders and touch size should scale with average quote size.
If touch size has increased faster than volume than that’s evidence of more nefarious activity. The opposite is evidence of more beneficial activity.
From the 2006 (basically the beginning of HFT) to today volume has increased substantially more than touch size (which has roughly stayed the same if not decreased in most markets). This is strong evidence that the overwhelming majority of HFT is of the beneficial, not abusive, variety. At least along this dimension.