Growth in High-Frequency Trading

In trying to examine the background and underlying causes of financial market events such as the disruption in equity trading on August 1, or the “flash crash” of May 6, 2010, the question of whether we are looking at some fundamental change in the way the markets work is frequently raised.  Trading volumes have certainly increased over the past few years, but firms engaged in high-frequency trading tend to be rather reticent about the exact nature and extent of their operations.

In one of my posts about the 2010 flash crash, I mentioned an analysis carried out by Nanex, a firm that specializes in systems and software to distribute and process market data, giving them a particularly good view of the overall market.

Our business is supplying a real-time data feed comprising trade and quote data for all US equity, option, and futures exchanges.

I have just come across a rather striking demonstration, put together by Nanex, of the growth of high-frequency trading (HFT) activity in US markets over the period January 2007 through January 2012.    The top graph on the page is an animated GIF graphic, which shows daily HFT activity for the various US exchanges (denoted by different colors).  At the beginning of the period, in 2007, the lines are relatively flat.  As time progresses, one begins to see more activity, first at the beginning and end of the trading day, then spreading throughout the day. Assuming Nanex’s data are correct (and I have no reason to doubt that), there has been a very substantial increase in HFT activity over the last few years.

The graphs on the lower part of the page show a more interesting aspect of this growth.  The graph on the left shows the growth, over time, of quote activity (that is, the posting of bid and offer prices for securities).  The graph on the right shows actual trades over the same period,  It is clear that most of the activity growth is in quotes, rather than trades.   There was some suspicion, in the 2010 flash crash, that some market participants might have engaged in “quote stuffing”, the generation of spurious quotes that would effectively clog up competitors’ systems.   There is certainly nothing in this evidence to disprove that suspicion.

Nanex also points out that spurious quote messages, like spam, are close to free for the sender, but not for other market participants.  As noted above, the large increase in quote activity does not correspond to an increase in actual trades, meaning that the cost per trade of processing quote messages has gone up substantially.

We think that a 10-fold increase in costs without any benefits would be considered “detrimental” by most business people. We think the regulators would agree with us as well.

This analysis, together with other evidence, suggests that HFT activity is not as benign as some proponents would have us believe.   As Prof. Ed Felten at Princeton observed back in 2010, it can be difficult to disentangle just what happened after the fact, but I’ll try to fit some of the pieces together in future posts.

Speedy Trading — Again

On Wednesday, August 1, traders on the New York Stock Exchange [NYSE] and other US equity exchanges had a more than usually interesting morning.   Shortly after the market opened, unusually high trading volumes and large price swings occurred in more than 100 stocks, including well-known names like Coca-Cola, IBM, and McDonalds, as well as smaller firms like Wizzard Software, Kronos Worldwide, and Trinity Industries.  The wild price gyrations in some stocks were reminiscent of the “flash crash” of May 6, 2010 (which I wrote about several times here), and of the more recent problems connected with Facebook’s initial public offering [IPO] in May of this year.  As in the 2010 incident, the problem appears to have been connected to am automated trading system; according to a report in the New York Times,

An automated stock trading program suddenly flooded the market with millions of trades Wednesday morning, spreading turmoil across Wall Street and drawing renewed attention to the fragility and instability of the nation’s stock markets.

The problem appears to have originated in the market-making unit of Knight Capital Group, a New Jersey broker that executes “algorithmic” trades on behalf of other firms.  According to a report from Bloomberg News Service, Knight said the problem was due to a “technology issue”:

Knight’s initial review indicated a “technology issue” occurred in the company’s market-making unit related to the routing of shares to the NYSE, according to an e-mail from spokeswoman Kara Fitzsimmons.

Trading in several issues was suspended for a time; after reviewing trades in over 100 stocks, the NYSE announced that it would void trades in six issues that exhibited particularly wide price swings.

Trades that occurred 30 percent above or below the opening price in Wizard Software Corp., China Cord Blood Corp. (CO), Reaves Utility Income Fund, E-House China Holdings Ltd., American Reprographics Co. and Quicksilver Resources Inc. will be voided, according to a statement on the NYSE website.

Knight Capital itself was badly affected; according to a report at the Washington Post, the “trading glitch” will cost the firm $440 million.

A technical problem that briefly threw dozens of stocks into chaos Wednesday will cost Knight Capital Group $440 million, the trading firm said Thursday. Knight’s stock plunged for a second day, erasing 70 percent of its value from two days ago.

That’s a hell of a glitch.  The situation is also embarrassing for the Knight Capital CEO, Thomas Joyce, who has been a vocal critic of NASDAQ for the problems it experienced during the Facebook IPO.

Apart from the immediate disruption and costs associated with these incidents, there is a concern that they are eroding ordinary investors’ confidence in equity markets, and that they may be a warning that those markets have lost sight of their primary purpose.  I’ll talk about some of those issues in future posts.