I attended the second day of the World Research Group's Summit on "Buy Side Tech: High Frequency Trading" (day one summary here). Yesterday was only a half day and I missed the first panel but below is another rundown of what I found to be interesting.Data Centers
I caught the second panel of the day with Adam Honoré of Aite Group, LLC and Kevin McPartland TABB Group. There was a good deal of discussion on the topic of co-location. The NYSE has taken the most political approach as McPartland sees it promising the same latency for all participants located within the Mahwah data center in New Jersey. The NYSE built the 400,000 square foot, 2-story fortress last year to accommodate the burgeoning demand for proximity hosting of high frequency trading firm servers. The NYSE has promised all firms within the complex a 70 microsecond latency which it will accomplish by literally cutting all the wire lengths the same for each rack.
Neither Honoré or McPartland see any game-changing regulation in terms of co-location. While many complain of the unfair advantage this provides in terms of ultra low latency, McPartland pointed out that disallowing exchanges from providing the space around the matching engine would only work to drive up the real estate across the block as proximity will continue to matter in terms of the needed travel distance for data.
While the NYSE has fully dedicated itself to the hardware side of the business with its $250 million investment in Mahwah, other venues such as BATS and Direct Edge do not have such data centers. Honoré saw the NYSE as making this major CAPEX decision in spite of its shareholders because it marks a venture into an entirely unproven business model at this point. This investment requires significant upfront costs and substantial continuing upgrade and maintenance costs. Only time will tell whether this is a viable model. Given current rental rates on space and future expectations, the project should be profitable. Yet, the frenzied race for ultra low latency could slow or even end entirely should a plateau in demand be reached.
Honoré postulated that given the possibilities of co-location, all geographic barriers have been erased and he expects to see high frequency firms arise across the globe that trade US markets.
Muthuswamy on Market Efficiency
The two-day conference ended with a speech by Jayaram Muthuswamy, Professor of Finance at Kent State University, titled "Does High Frequency Trading Make Markets More or Less Efficient?" Muthuswamy's speech was the real reason I attended the second day and while he's a PhD and talked over my head to some degree, I enjoyed the speech nonetheless.
Muthuswamy talked quite a bit about computational complexity. He believes that the winners in the HFT space will ultimately be those that discover and exploit the most arcane relationships and patterns. He touched on the idea of efficiency in the equity markets and believes we have semi-strong efficiency. Strong form efficiency does not exist explicitly demonstrated in takeover situations where a stock often sees a large burst to a higher level and then trades flat at the bid price. In theoretical strong form efficiency, the stock would trade flat and the bid would have no impact on the price.
He briefly touched on the advantages and disadvantages of HFT. Advantages being enhancements in connectivity between markets and assets, lower transaction costs overall, the encouragement of high levels of creativity and more efficient markets as information is assimilated more rapidly than ever before. Disadvantages include glitches such as May 6th, the unfair player advantage where HFT could possibly hurt traditional participants such as "flashing" which most venues have now banned and the exacerbation of volatility.
While his list of advantages seem self-evident for the most part I wonder how Muthuswamy would answer critics that claim May 6th was not necessarily a result of a glitch. The move away from a mandated liquidity provider, namely the specialist system, to liquidity providers that have no obligation and are often directional has many implications. Tradebot and Tradeworx, two of the largest HFT firms, admitted to shutting down their systems prior to the flash crash. In this scenario, we did not see a glitch rather a conscious decision by the human operators of the machines to remove their liquidity provision precisely when it was needed most, a time of increased volatility. It would appear that designated market makers clearly failed in their role to maintain orderly markets. The oft-repeated idea that HFT was not a negative influence because stocks bounced back just as fast as they fell is simply hogwash.
Ultimately, the question for the non-human is: where does the discretionary trader fit into this brave new world? My thoughts on this coming up.