New Paper Sheds Light On Mini Flash Crashes

We found another academic paper which has broken from the mold from the typical “tobacco professor” style papers that we normally read relating to HFT.  The paper titled  “High Frequency Trading and Mini Flash Crashes” was written by Golub, Keane and Poon and makes some very bold statements about mini-flash crashes.  They state:

“We find that Mini Flash Crashes are the result of regulatory framework and market fragmentation; in particular due to aggressive use of Intermarket Sweep Orders and Regulation NMS protecting only the Top of the Book.  We find that Mini Flash Crashes have an adverse impact on market liquidity resulting in wider spread, increased number of locked and crossed NBBO quotes and decrease in quoted volume.”

The data set that the authors work with are the periods from September 2008 to November 2008 and the month of May 2010 (they picked these months because they were the most volatile over the past few years).  There were 5,140 mini flash crashes in these four months which represents 28% of all mini flash crashes from 2006-2011.  Before we get into the rest of the paper, think about these numbers again.  How many times have we been told by exchange officials and regulators that the stock market performed extremely well during the financial crisis of 2008?  There were 4,649 mini flash crashes from September to November of 2008.  These were possibly the cause of some of the gut wrenching ups and downs during that period.  Maybe the stock market didn’t perform as well as some have claimed.

Back to the paper, the authors find that most mini flash crashes were caused by aggressive use of ISO (inter market sweep) orders.  They say ISO’s are dangerous because they sweep only top of books from the various market centers:

In many cases there is a significant amount of liquidity on the other exchanges during the time frame of interest, available deeper in the order book. However, as the Order Protection rule only protects the Top of the Book, trades can occur at vastly inferior prices resulting in Mini Flash Crashes in a particular exchange.”

And who do they blame for these crashes?

“Given the speed and the magnitude of the crashes, it appears likely that Mini Flash Crashes are caused by HFT activity….Today’s traders operate in a tiered market environment where co-location and di- rect data feeds have made it possible for some participants to obtain tremendous speed advantage, rapidly submitting orders and instantaneously cancelling them.”

The authors conclude with a few recommendations:

Mini Flash Crashes have become the consequence of the fragmented liquidity that is indicative of the current market structure with its myriad of trading venues.  Mini Flash Crashes demonstrate that market participants require a better understanding of how their orders take liquidity. The visible liquidity at the time an order was placed does not necessarily remain when the order is executed – but rather the converse.”

Extending the Order Protection rule to the Depth of the Book may be a quick fix to the Mini Flash Crashes studied here. However, such a requirement for market participants to comply with the same trade-through protection for the entire Depth of the Book quotations would be extremely costly and complex.  Nonetheless, the aggressive use of ISOs resulting in Mini Flash Crashes should not be tolerated.”

“We suggest that regulators pay particular attention to ISO-initiated Mini Flash Crashes and take appropriate actions against market participants responsible for such Mini Flash Crashes. We also suggest regulators should consider the banning of ISOs.”

So, do mini flash crashes really matter? Do they cost investors any money?  Many advocates of the current market structure claim that nobody really gets hurt since the stock prices revert back to their true value usually within minutes.  But let’s look at a recent example of a mini flash crash.

On 12/4/12, First Republic Bank (FRC) opened at $32.69.  At 9:31am, one minute later, it  dropped 3% to $31.42 on no news.  FRC then rallied back to $32.70 in the next minute.  Volume spiked to 250,000 shares during this short time period.  What’s interesting is that the stock crashed just as it broke through the 200 day average.  Surely there were lots of stop losses that were triggered at that level,  Of course, all these trades will stand since a clearly erroneous trade  for a stock priced over $25 needs to be at least 5% away.

Did the investors who placed prudent stop loss orders get harmed by this mini flash crash? You bet they did.  Now, the question becomes, was the FRC mini flash crash an intentional event that was started to flush out the stop loss orders?  We’ll leave it to the regulators to answer that question.