FSA-sponsored Academic Study Shows HFT Does NOT Reduce Institutional Investor Trading Costs
A January 2013 paper written by Brogaard, Hendershott, Hunt, Latza, Pedace, and Ysusi, titled High Frequency Trading and the Execution Costs of Institutional Investors, attempts to take on a lofty goal: set out to prove that HFT does not harm institutional investors. It looks at TWO data sets – one that tries to capture institutional trading costs, and another that is an attempt at a proxy for HFT activity.
This paper studies whether high-frequency trading (HFT) increases the execution costs of institutional investors. We use technology upgrades that lower the latency of the London Stock Exchange to obtain variation in the level of HFT over time. Following upgrades, the level of HFT increases. Around these shocks to HFT, as far as can be measured, institutional traders’ execution costs remain unchanged. Thus, we find no evidence that these increases in HFT activity impacted institutional execution costs.
The FSA-sponsored work makes many assumptions in its (perceived-to-us) quest to sanitize and champion HFT. We believe it falls short, and not only because in all of its methodology it still does not conclude that HFT lowers institutional trading costs, only that it doesn’t increase them, but because of flaws in its assumptions.
– The paper studies trading in the UK’s 250 largest stocks by capitalization. It does not look at mid and small caps.
– The paper uses FSA TRS data, which only includes HFT activity from registered entities, and not unregistered HFTs. The authors decide who is and who are not HFTs, and they exclude any HFT-oriented activity originating from firms whose primary function is not HFT –i.e. they exclude all banks. Like Knight, Goldman, Barclays, Lehman, Bear, Morgan Stanley, etc. Their data also doesn’t capture HFT activity from unregistered HFTs.
– The paper observed only HFT data only through July 2010. Also in their statistical study they choose to focus only on data prior to May 2010. What major event off the top of your heads, that was a home-run for HFT activity, happened in May 2010 that is excluded from their study? LOL. Also, while data on HFT trades exists after July 2010, the authors do not choose to use it because a requirement was eased for HFT trade reporting, and HFT trade reporting decreased after that. The make an assumption here that HFT trading as a trend still increased, however, which is an assumption that many folks would disagree with. I guess we will never know whether the decrease in HFT activity after July 2010 was responsible for a decrease in institutional trading costs since then.
– The study excludes data on institutional execution costs between July 2007 and June 2009 – the financial crisis – where institutional costs increased.
– And it removes HFT data from the financial crisis of 2008-2009, which we don’t frankly understand why.
Even with a setup like this, the paper still doesn’t conclude that HFT has reduced transaction costs for institutions; it only concludes that it hasn’t increased them. So… Where’s The Beef?