We Feel Your Pain. Come Have a Wee Cuddle…


A little over a week ago, I read an article on Bloomberg by Matt Levine titled People Are Worried About Bond Market Liquidity. (I read Matt Levine regularly, despite my tendency to disagree with much of his flippancy; frankly he is the best financial writer in media). The article pointed out Pimco’s worrying about Treasury Market Flash Crashes:


“The re-regulated, better capitalized global banking system allocates little of its balance sheet to making markets, resulting in greater likelihood of flash crashes, air pockets and trading volatility.”


And Levine argues that the concern is much ado about nothing:


“People hate flash crashes, and obviously they cause some people to lose money, but they have always struck me as sort of non-systemic, a technical glitch rather than a major fear. A sharp permanent drop in asset prices is scary. A sharp temporary drop in asset prices is kind of funny, honestly.


We scratched our heads a little, because we didn’t think it was funny. We were surprised that it was lost on Mr. Levine how microstructure shocks create real expense for institutional investors. And that is the topic of this Friday’s note.


Do you remember the August 2012 Knightmare on Wall Street, where Knight lost $440 million in a short period due to a glitch in its programming? Obviously Knight paid a very steep price, and they had to be rescued by merging with the high frequency trading firm, Getco. In the aftermath of that gigantic error,  “Knight took the blame for the mistake, and didn’t hurt customers.”


He was wrong. And Matt Levine is wrong too. These market structure crashes and glitches are not funny, and not all that temporary, and they don’t solely affect the Glitch-or! Check out this academic study out of the University of Cincinnati, titled:


The Rogue Algorithm and its Discontents – Evidence From a Major Trading Glitch


The author, Mehmet Saglam studied institutional customer trades given to it by a major bank – VWAP and POV trades – from before, during, and after the Knight Glitch. He examined trading costs – VWAP and Implementation Shortfall (IS), and found that the Knight error cost institutional investors $100 million!


Some highlights from the study:


  • The estimated Implementation Shortfall (IS) cost increase due to the glitch is about 18 basis points. Historically that costs is 3 basis points. This is a six-fold increase.


  • The estimated VWAP cost increase due to the glitch is about 6.7 basis points. Historically that costs is 1.2 basis points. This is a also a 5-6 fold increase.



  • The cost increase is akin to an institutional trader trading at a 40% participation rate, as opposed to a 10% participation rate (which is the “aggressive” rate that Waddell & Reed used to trade S&P futures on May 6th 2010).


  • The additional institutional expense to trade was incurred not only during the glitch, or on the day of the glitch; the elevated costs persisted for a week after the glitch!



So what’s the takeaway? Flash crashes, Mini-flash crashes, and glitches are everybody’s problem. They affect trading costs for long-term investors not only while they are happening, but for a period of time afterwards as well. We noticed anecdotally how spreads were wider in stocks, and especially ETFs, for several days after the Knightmare. This academic study looked at actual institutional orders and quantified/proved it.