“HFT Groundhog Day”


The HFT public relations machine is out in full force.  They are currently busy pushing a book which tries to counter Michael Lewis’ “Flash Boys (we’ll have more on this later this week) and a new academic paper written by one of their favorite authors, Professor Jonathan Brogaard from the University of Washington.  The paper titled “High Frequency Trading and Extreme Price Movements”  is likely to be one which is often cited by the pro-HFT folks.

As you know by now, before we read a new academic paper, we always like to check the bibliography, the source of data and if there was any financial support given to the authors.  This paper gets red flags on each of these three issues.

– The bibliography includes many of the same papers that pro-HFT papers usually cite from including papers from Hendershott, Brogaard, Menkveld and Jones.

– The data comes from the questionable NASDAQ HFT data set and is limited to one exchange.

– The authors did receive financial support from something known as the ARC grant 09/14-025.

The paper makes this incredible finding:

“We find that during extreme price movements high-frequency traders act as net liquidity suppliers, while non-high-frequency traders act as net liquidity demanders.   Our evidence is consistent with HFT performing a stabilizing function in modern markets.”

Whoa, did we just read that right?  HFT acts as liquidity suppliers during extreme price movements? This would be contrary to the findings of the Joint SEC/CFTC Flash Crash Committee and contrary to some of the best sourced papers written by the independent research analyst program of the CFTC. Let’s take a closer look at some of Brogaard’s findings:

-Some market observers claim that HFTs only supply liquidity in normal times and withdraw liquidity in times of stress. We observe the opposite; normally, HFTs as a group demand more liquidity than they supply. Nevertheless, during price jumps HFTs supply more liquidity than they demand, compensating for the net liquidity demand from nHFTs.

 – Neither HFTs nor nHFTs act as net liquidity providers during the systematic jumps (systematic jumps are instances when at least two stocks undergo an extreme price movement in the same direction during the same time interval).

Brogaard goes out of his way in this paper to address “the critics” and their accusations about disappearing HFT liquidity.  But we think Professor Brogaard should have spent more time focusing on the lack of good data in his paper.  His data has many deficiencies that we cite below.

Brogaard’s Data Shortcomings:

 Brogaard only studies the largest 40 stocks by market capitalization (20 NASDAQ listed and 20 NYSE listed stocks).  Supplying liquidity in a stock that trades 50 million shares a day is much different than supplying liquidity in a stock that trades 50,000 shares per day.

– Brogaard uses five year old data (from 2008 and 2009) derived from the NASDAQ HFT data set. Technology as well as market structure has changed dramatically over the past 5 years and producing conclusions based on five year old data is suspect at best.

– Brogaard’s data is for trades that only occur on NASDAQ.  Why would Brogaard rely on only one exchanges data when there are 11 stock exchanges and over 40 ATS’s?  Brogaard studied a subset of data which only represents approximately 20% of all volume.

– The NASDAQ HFT data set that Brogaard relies on for his HFT liquidity analysis only includes 26 proprietary trading HFT firms.  Investment banks that have HFT arms are not included in his study.  Smaller HFT firms that were not identified by NASDAQ are also not included in this study.

Brogaard studies only one asset class.  While stocks are what we trade and study, many HFT firms use strategies that employ multiple asset classes including commodities, bonds and currencies. Studying only one asset class at a time only gives you a sliver of information about what HFT firms are actually doing.  For example, some HFT firms may be adding liquidity in the stock market while at the same time consuming liquidity in the commodities market.

Based on the incomplete and limited data, we find that Professor Brogaard’s paper lacks credibility and should be discounted.  We also find it very sad that Professor Brogaard and company keep recycling and analyzing the same old NASDAQ data set to try and support their HFT friends.  It is like HFT Groundhog Day.  Somebody please let us know when the time loop ends.