While the merits of high-frequency trading (HFT) have been espoused many times and in many venues, one research paper argues all the good these market participants do is countered by a single trading tactic they employ – tens of thousands of times per second.
Drum roll, please?
And that tactic is the rapid submission and subsequent cancellation of limit orders, according to new research paper by Alvaro Cartea, Richard Payne, Jose Penalva, and Mikel Tapia. The paper, “Ultra-Fast Activity and Market Quality,” available on the Social Science Research Network’s website, said that “rapidly placing and cancelling limit orders significantly reduces liquidity and increases buy/sell spreads.”
HFT firms and their proponents often counter this arguing that they improve price discovery and supply market liquidity.
And while the paper acknowledges these claims have merit and have at least some backing in academic research, HFT firms “use lots of different tactics at their disposal, and looking at the firm-wide impact may be too broad to identify trading strategies that reduce market quality.”
To arrive at their conclusions, the authors looked order and trade messages sent to NASDAQ between March 2005-2013, to build a measure of ultra-fast activity (UFA) which captures the times when trading algorithms are most active. The results suggest, quite consistently, that UFA is associated with lower liquidity in stock markets. An increase in UFA leads to greater quoted and effective spreads and lower depth posted in the limit order book.
For example, in 2013 they found that a 1 standard deviation increase in our measure of UFA is associated with a 0.13 standard deviation increase in the quoted spread. UFA causes similar increases in effective spreads and also reduces quoted depth. Finally, the effect of UFA is also economically significant.
And again, in March 2013 the effect of a one standard deviation in UFA generated on average an increase of between 3 and 6 percent in the quoted spread and effective spreads, as well as a drop of between 3 and 4 percent for depth measured close to the best bid and ask prices.
A one standard deviation jump in ultra-fast activity costs traders 30 bps, they noted.
“Thus, our analysis of ultra-fast quoting activity suggests that it has a consistent and strong negative association with liquidity. The difference between our results and those who study HFT and show that it improves liquidity obviously lies in the difference in our respective measurements of high frequency or low latency activity.”
The paper specifically looks at the effect of limit orders that are posted and then cancelled within 100 ms, a measure they call PC100. This short time frame eliminates human traders but doesn’t necessarily throw out algo traders, which is why they call it ultra-fast activity instead of high frequency trading, but the researchers do expect that they are mostly observing HFT firms. Comparing NASDAQ data every March from 2005 – 2013, corrected for changes in market liquidity, and found a significant relationship between PC100 and bid/ask spreads. For every one standard deviation increase in PC100, spreads go up by 0.1345 standard deviations. In more concrete terms, that amounts to 30bps on the dollar.
The paper’s authors said that they aren’t calling for this hyper-fast submission and cancellation of orders tactic be banned, or for regulators and governments to step in. But rather, they do say that a more careful and nuanced view of equity market activity and the context in which such activity takes place should be examined.
