A draft academic paper casts doubt upon the
role played by UK trader Navinder Sarao in the May 2010 Flash Crash, and
instead suggests weakness in the trade reporting infrastructure may have played
a significant part.
The draft paper, The Flash Crash: A New Deconstruction was written by Eric Aldrich
and Gregory Laughlin from the University of California, along with Joseph
Grundfest from Stanford University, and uses millisecond analysis of the events
of May 6, 2010, both in terms of executed trades and, crucially, in terms of
Noting that the joint CFTC-SEC report on
the events surrounding the Flash Crash is “consistent with the data”, the
authors also states, “While
assertions relating to causation of the Flash Crash must be accompanied by
significant disclaimers, we suggest that it is highly unlikely that, as alleged
by the United States Government, Navinder Sarao’s spoofing orders, even if
illegal, could have caused the Flash Crash, or that the crash was a fore-
seeable consequence of his spoofing activity.”
As part of the charge
against Sarao, the Commodity Futures Trading Commission (CFTC) and Department
of Justice (DoJ) argue that he placed a huge amount of orders to sell futures
at out of the market levels, and this created an order book imbalance which was
picked up by traders who then sold aggressively, triggering the Flash Crash.
The authors dispute
this, however, noting that commentators are sceptical that his relatively
small-scale trading could have such an impact. “Indeed, if Sarao’s relatively
small-scale trading could in fact generate the large-scale effects asserted by
the government, modern equity market structures could be viewed as alarmingly
fragile,” they suggest.
The paper further
argues that the Flash Crash is sufficiently explained as the result of the
confluence of the unsettled market conditions that prevailed in the hours
leading up to the Flash Crash combined with the size and execution strategy of trades
by asset manager Waddell & Reed.
Waddell & Reed
used an algorithm to execute trades in the E-Mini S&P 500 contracts but the
algorithm had no time or price parameters, merely it was limited to being 9% of
market volume from the previous minute. As liquidity disappeared, the algorithm
inadvertently became much larger than 9% of volume, thus is exacerbated an
already nervous market, leading to a crash in prices.
model used by the authors suggests that the probability of another Flash Crash
is directly related to the imbalance in the ratio of high frequency, liquidity
consuming, traders compared to the amount of fundamental, non-HFT traders
The paper argues that
fundamental traders would have paid no attention to Sarao’s orders because they
did not “display any useful information” as they were so far from the market.
As the fundamental traders exited the market, however, machine-based traders
would have interrogated the order book further and found his orders. Sensing a
huge order book imbalance, the HFTs then “rationally withdrew liquidity” which
once again exacerbated the problem.
The authors note,
however, that Sarao’s orders were often between levels four-to-seven in the
order book and their analysis argues that at these levels the order imbalance
would have a 0.0135 basis point impact – “drastically smaller than the 500
point loss that occurred”.
A new area of analysis
used by the authors, however, offers a different, or additional reason for the
events of May 6, 2010, saying, “Our analysis of message traffic also leads to
the discovery of previously unobserved anomalies in information flows during
the Flash Crash.”
They note there was a
breakdown in the usually highly correlated price links between the underlying
Nasdaq equities market and the CME Group futures market. Communications during
this period were fully operational but there was a break down in the
reiterates that HFT firms had exited the market, for as the authors note,
arbitrage between the futures and underlying contracts is at the heart of many
of their strategies. These firms were absent because, due to some of the
aforementioned events, they had hit internal risk thresholds which triggered
their exit from the market. “It appears plausible that the ensuing, and
extended period of outright arbitrage opportunities arose because an
insufficient number of participants had the margin to take advantage of these
opportunities,” the paper states. “Several minutes were required for new
entrants to the market to arrive, eventually allowing disciplined,
cross-exchange pricing efficiency to resume.”
The authors discover
that there were delays in reporting off exchange transactions to the
Consolidated Tape System (CTS) as well as in quotations on the NYSE and the
FINRA trade reporting facility. Further the delayed transactions were not being
reported in a uniform fashion, rather they caused the CTS to oscillate
“rapidly” between the market consensus price and the reported trade price.
confusion, the authors find that for roughly three-and-a-half minutes, the
delayed off-exchange prices were not labeled as such and “could have been
interpreted as live, marketable, prices when, in fact, they reflected stale
prices no longer available in the market”.
None of this broke the
rules on reporting trades as they were corrected within the allowed window of
time and reported correctly, but in such market conditions it could be that the
damage was already done. To highlight how much of a problem this was on May 6,
2010, the authors report that in a 16 month period only seven days had
witnessed price deviations and on only one of those had there been a pricing
error in the market.
In contrast they find
that 878 deviations occurred in just over 11 minutes on the day of the Flash
Crash. “If this data feed issue explains either the price decline or recovery,
it could call into question the dominant narrative regarding the cause of the
Flash Crash and common perceptions regarding the role of the CME halt in stimulating
the market’s recovery,” the authors suggest.
The authors are keen
to stress that their analysis, “does not negate that Sarao engaged in illegal,
manipulative conduct. Nor do we contest the allegation that Sarao’s algorithms
could have, on occasion, caused “artificial” prices and volatility to appear in
the market, as alleged by the Government. Our challenge is, instead, to the
inference that Sarao’s illegal conduct was a material contributing cause of the
Flash Crash, that he intended to cause the Flash Crash, or that the Flash Crash
was even a foreseeable result of his illegal trading activities.
distinction, Mr. Sarao’s conduct may allow him to be incarcerated, enjoined,
and fined,” they observe.
The authors also
suggest that deeper analysis of the role played by data feeds could lead to “reconsideration” of the role played
by CME Group.
The Merc’s five second
trading halt on May 6, 2010 is “commonly credited with creating an opportunity
for the market to recover…However our data establish that the market continued
to decline after the CME halt, and that the recovery did not begin until after
a halt in the FINRA data feed that began 15 seconds after the end of the CME
halt and lasted for an additional 5 seconds.
“This pattern is
consistent with the possibility that the CME halt gave the FINRA feed an
opportunity to “catch up” so as to stop delivering stale and confusing prices,
and that only after the confusion was resolved could a recovery take place,”
the paper continues. “If so, the CME halt might be better described as a
necessary condition for the correction of the FINRA feed, but an insufficient
condition for the market’s recovery. In contrast, the correction of the FINRA
feed, for whatever reason, would constitute a necessary and sufficient
condition for the market’s recovery.”
Whilst stressing that their
analysis of the role of data in the Flash Crash is continuing, the authors
warns that, “The policy implications of this explanation, if correct, are
simple and straightforward: pay attention to data feeds.”
Colin_lambert@profit-loss.com Twitter @lamboPnL