Flash crash report is superficial and disappointing

The joint report by the US CFTC and SEC on the flash crash of May 6, 2010 was released late last week. I found the report quite disappointing and superficial. Five months in the making, the report provides a lot of impressive graphs, but few convincing answers and explanations.

Consider the key finding:

“One key lesson is that under stressed market conditions, the automated execution of a large sell order can trigger extreme price movements, especially if the automated execution algorithm does not take prices into account.” This is referring to a sell order by a mutual fund for 75,000 index futures contracts with a notional value of $4.1 billion that was executed with a target participation rate of 9%. Roughly speaking, this participation rate implies that in any time interval, the algorithm tries to execute a quantity equal to one-tenth of what other traders are executing in the aggregate. The discussion in the report is hopelessly vague about what happened, but it suggests that the crucial problems described in the next point below were triggered when the first 3,300 contracts had been sold. While the report explains at length that 75,000 was a truly large order, it is clear that 3,300 contracts ($180 million) is not an outlandishly large trade. Surely, there must have been episodes in the past of a few thousand contracts being sold very quickly and the report could have provided a comparison of what happened on May 6 with what happened on those dates. And if the problem was market stress, then understanding the nature of that stress is critical.
“Moreover, the interaction between automated execution programs and algorithmic trading strategies can quickly erode liquidity and result in disorderly markets.” The claim is that the large order confused the high frequency traders (HFTs) who increased their trading volume, and this higher volume fooled the mutual fund’s algorithm into selling even faster. As far as I can see, this is pure speculation. I would have liked to see this phenomenon demonstrated using agent based models or some other sound methodology.
“As the events of May 6 demonstrate, especially in times of significant volatility, high trading volume is not necessarily a reliable indicator of market liquidity.” This is referring to the fact that: “ Moreover, compared to the three days prior to May 6, there was an unusually high level of ‘hot potato’ trading volume – due to repeated buying and selling of contracts – among the HFTs, especially during the period between 2:41 p.m. and 2:45 p.m. Specifically, between 2:45:13 and 2:45:27, HFTs traded over 27,000 contracts, which accounted for about 49 percent of the total trading volume, while buying only about 200 additional contracts net.” However, no explanation at all is provided for this – if HFTs were desperately trying to pass around the hot potatoes that they had acquired minutes earlier, why were they buying 200 more hot potatoes?

Even if we were to accept the conclusion of the report that a mutual fund selling $4 billion of index futures in 20 minutes is an adequate explanation of the flash crash in the index markets, there is still the issue of what happened in specific stocks. The index began its recovery at 2:45:28 but the carnage in individual stocks happened a few minutes later at 2:48 or 2:49. The report attributes this to the withdrawal of liquidity by market makers at around 2:45. At this point, the report relies on extensive interviews with market makers and fails to substantiate key assertions with hard facts.

Some portions of the report look more like a journalist’s casual empiricism than the hard analysis that one expects in a fact finding report. For example, on page 66, there is a discussion of data about a single market maker that concludes with a whole string of tainted phrases: “If this example is typical ... it seems that ... This suggests that ... From this example it does not seem that ... ” I can understand all this five weeks after the crash, not five months later.

A few less important quibbles:

I also found some of the graphs difficult to interpret. For example, the charts on the order book are colour coded relative to the mid-price of the stock. Since this price was falling dramatically, it is difficult to see what parts of the order book were actually being eliminated through order execution or order cancellation. That prices can fall to near zero only if the buy side of the order book is exhausted is a tautology. One wants to see how and when the buy orders were cancelled or got executed.
Most of the reported data for individual stocks is at one minute intervals and some at fifteen minute intervals when other analysts have been looking at events on a millisecond time scale. At a one minute time scale, several things happen simultaneously – for example, prices fall and order books shrink – but it is difficult to see what happened first. Causality is hard, but is it too much to ask for at least the sequencing to be described accurately?
The report suggests that the order books of ETFs had less depth far from the mid quote and this led to the disproportionate incidence of broken trades in them. However, in the first part of the report, it is shown that the ETF on the S&P 500 (known as SPY) performed better than the index future itself. This suggests that not all ETFs are similar in the fragility of their order book, and I would have liked to see some exploration of this issue.

In conclusion, the report leaves me disappointed as regards the three critical questions that I asked myself after reading the report:

How far does the report provide confidence to an investor that with the corrective action taken since May 6, 2010, market prices are reliable? I think the report is totally unconvincing on this score.
Does the report provide evidence that the post-Madoff SEC (and CFTC) can analyze a complex situation and arrive at a top quality analysis? Despite the high calibre of resources that have been recruited into the SEC in the last year or two, the quality of the report leaves much to be desired.
Does the report show that regulators have escaped cognitive capture by their regulatees? I am sorely disappointed on this score. The report draws on extensive interviews with traditional equity market makers, high-frequency traders, internalizers, and options market makers. Apparently, nobody thought it fit to interview retail investors to understand how market distortions and data feed delays affected their order placement strategies. For example, Nanex has claimed that the Dow Jones index was delayed by 80 seconds. Were retail investors who follow the Dow Jones thinking that the market index was still falling even when professionals could see that it was recovering? Did this cause panic selling? For a cognitively captured regulator, it is sufficient to report that “Most of the firms we interviewed ... subscribe directly to the proprietary feeds offered by the exchanges. These firms do not generally rely on the consolidated market data to make trading decisions and thus their trading decisions would not have been directly affected by the delay in data in this feed.”

Posted at 2:19 pm IST on Sun, 3 Oct 2010 permanent link

Comments