CommunityFDL Main Blog

Official Report on Flash Crash: Nothing to See Here

May 6, 2010 flash crash (wiki)

The CFTC-SEC report (.pdf) on the flash crash of May 6 is out, and concludes that it was all a big accident. All we need is a few tweaks to nearly perfect markets. It flatly rejects the Nanex report, and according to the owner of Nanex, did not make use of their data.

Two weeks ago, someone tapped business reporter Graham Bowley at the New York Times to write a puff piece on the guy doing the report, Gregg Berman. Market participants, that crowd of banksters who cause these crashes, wanted this report to show that the markets are just fine, probably because, as Bowley points out, since May 6, investors have pulled money out of mutual funds every week. Those same market participants found Berman “refreshing – and reassuring”.

“He had a sense that actions in markets provoke reactions and they are done for reasons, benign or not benign,” said Frank Hatheway, chief economist at Nasdaq, one of the exchanges that provided data for the inquiry.

Indeed, the report is written in the language of intention. It treats the computers doing High Frequency Trading and selling securities as if they were human beings acting for specific reasons. For example, the report says that the cause of the problems was the decision of one investor to sell 75,000 futures contracts related to the S&P 500, the highest single trade of the year in that security, worth $4.1 billion. The seller, which other articles identify as Waddell & Reed, used a computer algorithm to sell this huge position. From the report at 15:

At this time, HFTs were competing with [Waddell & Reed’s algorithm] for the liquidity expected to be provided by Fundamental Buyers who would hold their positions, or by Opportunistic Buyers who would trade based on their ability to hedge their positions in the equity markets.

Berman interviewed major players in the flash crash.

Each interview was conducted in two parts. In the first half, firms outlined their trading strategies and business models. In the second half, we discussed their actions on May 6, paying special attention to what caused each firm to act in a particular way.

Berman explains that the computers check incoming data, and stop trading if certain parameters exceed specified limits. At that point, humans are brought in. The flash crash was about 10 minutes start to finish. The idea that a human being acted intentionally in that short time frame is hard to accept. Firms reported that they returned to the market within a few seconds to several hours. P. 38. People acting in a few seconds are not thinking, they are following preset procedures or they are idly flicking switches. . . .

Eric Hunsader, the owner of Nanex, spoke to the investigators. He argues that the problem was data delays caused by sending enormous numbers of quotes valid for millisecons to an exchange, a practice he calls “quote-stuffing”. The report dismisses that theory without naming Nanex, and leaves open the possibility that this is a problem independent of the flash crash. P. 79. The report says that withdrawal of liquidity was the cause of the crash, and accepts the explanations for withdrawal of liquidity given by market participants.

Reuters reports that HFT firms love the report. The article quotes Hunsader saying that the report ignores his data. Nanex looks at trading in millisecond intervals, while the report deals with one second intervals. This is too long an interval to see the impact, Hunsader says. I think he may be right. When quotes and trades happen as rapidly as co-located computers can make them, it is silly to think that any kind of human intention is involved. I think there are a large number of interactions that can cause unanticipated outcomes. That makes me think that the Nanex approach makes more sense. It is driven by data on actions and results, not by attempts to understand the motivations of computers and their code writers.

I also think it is perfectly possible to experiment to see if you can make a profit off such outcomes. Hunsader identified another mini-flash crash on April 28, which he says followed the same pattern. His data is here, and it does look similar.

Perhaps the most trenchant analysis of this report comes from It describes a recent mini-flash crash in the stock of Progress Energy, and says that retail investors are getting hurt by these crashes. The machines have won the investing battlefield, and the humans simply cannot get a decent quote when the machines run amok. They ask if investors are getting ready to desert the stock market.

I doubt that the report will ease those concerns.

UPDATE: More now available on our corporate citizens’ liquidity fetish.

Previous post

Are we about to see a Financial Sector Celebrity Death Match Between the Title and Mortgage Industries?

Next post

War Everlasting



I read a lot of books.