NEW: Please also take a look at our follow-up post on Nasdaq Intelligent Ticks, where we discuss the five major reasons against the proposal (available HERE).
Having written my dissertation and a few other academic papers on optimal tick size back in the day, I was intrigued when I heard the Nasdaq had put out a proposal for a more flexible tick size structure (https://www.nasdaq.com/docs/2019/12/16/Intelligent-Ticks.pdf). I had presumed it would be akin to having the tick size increase with price, as other venues do, thereby holding the relative tick (i.e., the tick as a percent of price) roughly constant. Instead, they are proposing that the tick of a stock be set as the duration-weighted average spread rounded to the nearest proposed tick. For example, a stock with a 5.4 cent average spread would see its tick set to 5 cents. A stock with a 12.2 cent average spread would see its spread rounded to 10 cents, etc.
Of course, as a former head of algo trading, I instinctively broke into a cold sweat when I thought about introducing yet another complexity into the trading process, but that eventually subsided when I remembered that I am no longer on the hook for these types of changes. But as my pulse fell back to its normal level, I was left with the following questions and observations.
Initial Observations
Observation 1
The tick is proposed to be set to the level the market already determined. Why do we need a constraint added if the market found the “right one” anyway – especially one which may become stale over time?
It would seem the point of finding the “optimal tick size” would be to solve some specific problem with the current tick regime, like the issues put forth by academics and practitioners for an economically meaningful tick that is not too wide (to prevent excessively wide spreads) but sufficiently high as to be economically meaningful (to prevent “pennying” on high priced stocks, excessive price changes, etc.).[1] For example, a 1 cent tick on a low priced stock may be too restrictive (e.g., see Sirius XM). But for very high-priced stocks, one cent may be too low.
So, while Nasdaq does propose making the spread smaller for stocks where the one-cent tick currently binds, I had thought they would also argue that a high-priced stock with, say, a one-cent tick is not economically meaningful. But it turns out they propose to increase the tick only for stocks where the spread is already at or above 2 cents.
Observation 2
Nasdaq proposes that the tick be set based on historical spread. While the initial tick may make sense and long-term increases may also be captured, it could take a long time before the average spread falls sufficiently to warrant a change given the tick binds the stock from below. And in fact, as written, the proposal appears to inadvertently prevent ticks from ever decreasing once they have been initially set.
In other words, suppose a stock has a spread of 6.5 cents and is assigned to the 5-cent bucket. But over time, the stock becomes more liquid, such that its “true” (unconstrained) spread is well below 5 cents. In a world of 1 cent ticks, the data would show this increased liquidity in a lower spread very quickly, as the historical spread would eventually reach the new lower equilibrium value. But with a binding 5-cent tick, it could take significantly longer before the average tick falls close enough to 5 cents to be shifted downward.
In fact, given the specifics in the proposal, once a stock is put into a bucket, there is no way for it to move to a lower tick size bucket. To see why, consider a stock currently in this 5-cent tick bucket. Because the tick is 5 cents, the stock’s spread can never fall below 5 cents – which means its average spread can never fall below five cents either. But if its average can’t fall below 5 cents, it can never be assigned to a lower tick bucket! (For Boston clients, this means “nevah evah”).
Observation 3
The tick is based on the duration-weighted average. This will increase the likelihood that the spread will bind at the end of the day, when the “true” spread should be tighter. Consequently, the proposal may actually increase the depth and time-to-fill relative to the existing regime, an issue that the proposal uses to motivate tick size reductions.
To avoid this, perhaps the tick could be based on the average spread around the close, when spreads are typically at their lowest point of the day, instead of on the overall average, which includes periods around the open where spreads can be 5-10 times higher on average than the remainder of the day.
Similarly, the use of a long-term average ignores the interday variation. On days when liquidity is high or volatility abnormally low, the true spread would likely fall below its historical average. But the tick set under this regime would reflect the average volume and volatility of the past, which may be quite different from the current trading environment.
Observation 4
Setting the tick based on its average spread may result in the average spread rising significantly above its current level
This is perhaps the most economically harmful aspect of the proposal. The proposal sets the tick by rounding down the average spread to its nearest possible tick, as noted above. But this doesn’t mean liquidity providers will round down. In fact, they will likely round up to avoid losses, thereby leading to wider spreads. For example, suppose one trader wanted to bid $19.98 and the other $20.03 (a five cent spread), neither would round to the nearest tick ($20.00).[2] Rather, the bidder would round down to $19.95 and the seller would round up to $20.05, resulting in a 10-cent spread! This not only increases cost, but would also increase the average spread, which in turn makes it even harder for stocks to see their tick size reduced should their underlying liquidity justify it.
Conclusion
I think the Nasdaq’s push to improve markets should be applauded, and I appreciate them making their proposal public to allow people like me to comment and provide hopefully useful feedback. I suspect if the tick size regime is ever changed in the U.S., it will move toward one based more on the model of price-based tick sizes (or some variant) given its conceptual and operational simplicity and its focus on the relative tick as the primary input, even though such an approach captures only one determinant of the optimal tick (i.e., price). But the Nasdaq proposal serves as a great starting point for this discussion.
NEW: Please take a look at our follow-up post on Nasdaq Intelligent Ticks, where we discuss the five major reasons against the proposal (available HERE).
The author is the Founder and President of The Bacidore Group, LLC. For more information on how the Bacidore Group can help improve trading performance as well as measure that performance, please feel free to contact us at info@bacidore.com or via our webpage www.bacidore.com.
For an overview of our other services, please click HERE.
And please check out our other blog posts available HERE.
And if you'd like to receive notification of new posts, please join our mailing list below.
Copyright 2019, The Bacidore Group, LLC. All Rights Reserved.
Footnotes
[1] See, for example, Professor (and former SEC Chief Economist) Larry Harris’s papers on “pennying” and tick size generally, Professor Jim Angel’s writing on optimal tick sizes, , etc.
[2] Of course, they couldn’t both do this because it would lock the market. But this is for illustrative purposes only.
Comments