There is a new paper out by Maureen O’Hara of Cornell University (she is also the Chairman of the Board of ITG) titled “High Frequency Market Microstructure” . The paper is a call to action for researchers to take a different approach when analyzing high frequency trading and market microstructure issues.  O’Hara says that “the learning models we used in the past are now deficient and that the empirical methods we traditionally employed may no longer be appropriate.”  In other words, O’Hara is challenging the status quo research methods.  She questions  research that uses the consolidated tape and says that the problem is that the tape is out of order due to the many venues reporting to it.  She also questions researchers who have been using the TAQ database and says the “true state of the market may not be visible” to them.

O’Hara recaps many of the current market structure issues including dark pools, predatory trading, colocation, maker/taker pricing, order types, data feeds and fragmentation.  O’Hara makes this statement about the quality of today’s market microstructure research:

“The new high frequency world is thus both complex and constantly evolving. New technology and greater speed lead to new strategies, which lead to new methods of trading, and, in turn, to new market designs. But hidden within this new paradigm are other changes such as the evolving nature of liquidity, the changing character of information and adverse selection, and transformations to the fundamental properties of market data such as buys and sells, quotes and prices. These changes, in my view, are equally important to understand because they challenge the ways researchers have interpreted market data and analyzed market behavior and performance… I believe that the high frequency world has also fundamentally altered some of the basic constructs underlying microstructure research.”

While the paper is an excellent read with some very good thought provoking questions, there is one section of the paper that we think stands out. O’Hara does a study of VWAP trades that were executed with a standard VWAP algorithm from ITG in 2013.  She found:

“The sample size is 243, 772 parent orders. The data clearly show the transition from parent orders to child orders: the algorithm executed 13,468, 847 child trades, meaning that on average each parent order turned into 55.325 child executions. The data also show that the algorithm executes the vast majority of parent orders with passive executions. For the sample as a whole, 65.3% of trades were passive; 21.9% were midpoint trades; and 12.57 % were aggressive.  Thus, a parent order to buy will show up in the data at least two-thirds of the time as “sell” orders – and including the midpoint orders this could be as high as 87%. Less than one in eight executed trades actually cross the spread and thus are the classic “buy” trades of microstructure models. “

 Note: Executed trades are classified as “passive” if the order is buying at or below the bid (or selling at or above the offer); as “aggressive” if the order is buying at or above the offer (or selling at or below the bid); and as “midpoint” if the order is filled at prices within the spread.

 O’Hara wanted to demonstrate that by analyzing the raw data it is often difficult to classify if a trade was initiated as a buy or sell.

But we look at her results a bit differently.  Over 65% of the VWAP algo trades were passive and therefore would be considered adding liquidity.  If we assume most of these child orders interacted with HFT, then we can conclude that HFT does not add liquidity and in fact consumes it.  But let’s take this one step further, why are these small child orders sitting on the book waiting to get passively executed?  Why are they not immediately just crossing the spread to get a fill? It could be because the VWAP algo doesn’t want to cross the spread and incur the “take” fee.  This means that O’Hara’s evidence may be demonstrating that a more sophisticated algo could be spotting the VWAP algo and walking it up to the price they want and then executing the trade by crossing the spread and extracting the liquidity that the VWAP algo is providing.  This would mean that the VWAP algo is causing the price of the stock to go higher (assuming a buy) than if it had just crossed the spread immediately.

Overall, we believe the O’Hara paper is the beginning of the next step in market microstructure research.  We applaud her for questioning the norms and hope others follow.