The NYSE floor closure caused by the COVID-19 virus resulted in a substantial change to the market structure of NYSE stocks literally overnight. The closure removed the face-to-face interactions among DMMs and floor brokers, and prevented off-floor traders access to special floor broker order types, such as the D-Order. This event raised questions about whether the floor actually mattered in a world increasingly dominated by electronic trading and whether the closure would mark the beginning of the end for NYSE floor trading. (See, for example, our earlier blog post here).
The Academic Evidence
Two academic studies released recently address empirically whether the NYSE adds value by comparing market quality before and during the closure. The first paper compared how measures like spread, volatility, and “pricing errors” were impacted by the floor closure. Specifically, the authors compare changes in NYSE market quality to that of comparable Nasdaq stocks, effectively treating similar Nasdaq stocks as a control group. Having some type of control group is key here since simply looking at how NYSE market quality changed when the floor closed would capture not only the effect of the floor closure, but also the changing market conditions brought about by the virus itself, e.g., the concurrent increase in market volatility.
The authors find that NYSE spreads rose by 11% following the floor closure, relative to the change on similar Nasdaq stocks. They find similar relative degradation in price impact and volatility as well. They also show that the relative degradation in market quality is greatest in the first half-hour of the day, when volatility and information asymmetry are high. Such a finding suggests that the floor helps to dampen some of this volatility and potentially enhance price discovery, a key finding since proponents of the floor often argue that the floor is most relevant around the open and close. The authors conclude that, even in a day where electronic trading has become the dominate form of trading, the NYSE floor still adds value. And, as importantly, this value does not simply stem from unique order types, like D-Orders, that are available only via NYSE floor brokers.
The second paper focuses primarily on the closing auction, where the floor closure would be most pronounced, in large part due to the unavailability of D-orders during this time.  The authors look at how the indicative pricing from the imbalance feeds is impacted by the floor closure. Specifically, they look at how well the indicative prices just prior to 3:55PM are at predicting the closing price when the floor is open relative to when it is closed. The notion here is that if the NYSE floor adds value, the indicative prices should provide investors with a better forecast of the closing auction price than when the floor is open.
The authors find the NYSE indicative prices from the imbalance feed become better predictors when the floor is closed. Without the floor broker D-orders being included in the imbalance feed en masse at 3:55 PM, the indicative price just before 3:55PM becomes significantly better at forecasting the closing price. While this insight is useful in understanding the value of the indicative prices generally, it is a somewhat mechanical finding. When the market is open, D-Orders are allowed to be entered almost right up to the close. And information on a large portion of these orders is not presented to the broader community until 3:55PM. Therefore, when the floor is open, the market design permits a significant flow of new information regarding the closing auction to enter the market after 3:55PM. When the floor is closed, on the other hand, information flow after 3:55PM is much more limited. Traders wanting to participate in the closing auction must enter their orders prior to the 3:50PM cutoff time, resulting in less incremental information occurring after 3:55PM. Given that the D-Orders can comprise a substantial fraction closing auction volume, it is not surprising that indicative prices are more informative when the floor is closed since relatively less new information can arrive at/after 3:55PM. Consequently, it is unclear whether the market is better off if traders are required to submit orders earlier, thereby resulting in a better forecast of the closing price. Or whether market quality is enhanced by allowing late arriving traders the opportunity to participate in the auction via D-Orders.
In effect, the authors do not address what most practitioners would view as the real question related to the close, namely how the closing auction price itself is impacted by the floor closure. Specifically, the important issue at the close is whether the market can draw in enough offsetting liquidity so that the closing auction price is not distorted by the large inflow of orders benchmarked to the closing price. Using pre-auction market prices (e.g., prices before any imbalance information is released) and next-day prices as benchmarks instead would provide insights on whether the closing prices were printed further away from the “true” price. For example, if the magnitude of the overnight price reversals (relative to the control group) stayed the same or lessened during the floor closure, this would suggest that floor traders do not play a consequential role in the closing auction. But by simply looking at how indicative prices fair relative to closing prices, the net effect is less clear since both the indicative price and the closing prices are likely affected by the floor closure. Put differently, the goal of the closing auction process – the order types, timing of order submission and information releases, etc. – is to provide an efficient closing price. So, the most direct way to measure the impact of a floor closure on the closing auction would be to evaluate its effect on the closing price.
A Big Unanswered Question
Perhaps the biggest question left out of both papers (though touched on briefly in the former paper) is how the opening price itself is impacted by the floor closure.The closing auction is more about liquidity discovery and less about price discovery, because, with 6.5 hours of trading already having occurred, the price uncertainty is less of an issue than around the open. What the floor is grappling with most at the end of the day is how to provide liquidity to the inflow of incoming on-close orders, which could lead to price dislocation. The opening auction, on the other hand, comes after a long market closure, which makes price discovery a key objective of the opening auction.Neither study has addressed this question, and I have not seen any analysis that has looked into this (if anyone has, please let me know). This is particularly unfortunate since the human interactions around the open, when price uncertainty is relatively high, has been one of the frequently mentioned benefits of floor trading. Hopefully, some enterprising academics or practitioners will provide some insights on this question since, without it, we are missing a critical piece of it.
The author is the Founder and President of The Bacidore Group, LLC.
 See “Does Floor Trading Matter?” by Brogaard, Ringgenberg, and Roesch (2020).
 See “Vestigial Tails: Floor Brokers at the Close in Modern Electronic Markets”, by Hu and Murphy (2020). See also Webster’s Dictionary for the definition of “Vestigial Tail”.