When the Intelligent Ticks Proposal was first released, I put my initial thoughts out in a previous blog post.[1] Now that the proposal has been out for a while and Nasdaq has re-circulated it, I thought it would be useful to revisit the topic to clarify and extend my initial thoughts.
Specifically, in this post, I "tick off" 5 key points that argue against the Intelligent Tick Proposal:
Most of the problems with the existing tick regime noted by Nasdaq could be solved by making the tick a simple function of price level.
Increasing the tick based on the average spread will result in an increase in the average spread. The only question is by how much.
Even if Intelligent Ticks was significantly better than a simple price-based approach, why use an average to define a minimum? The minimum tick is meant to define the smallest spread allowable, so wouldn’t it be more reasonable to set the minimum tick based on some estimate of the minimum spread?
The tick size is outdated and in need of revisiting. But is such a complex and opaque tick structure the most efficient way of doing it?
The Intelligent Ticks regime does not allow ticks to adjust immediately when the price of the stock moves, even though price is perhaps the biggest determinant of optimal tick. A simple price-based tick regime, by contrast, could respond immediately.
The Details
Most of the problems with the existing tick regime could be solved by making the tick a simple function of price level.
As is shown in Chart 1 of their proposal, price level is the key driver of inflated spreads, a fact well-known among academics and practitioners. This is also implicit in their chart of inverted venue usage, as inverted usage increases in the degree to which the spread is constrained by the tick. This suggests that simply making the tick a function of price level would solve much of the problem. Indeed, this is the approach followed by all the venues currently using an “intelligent tick” regime, as the Nasdaq points out. None of them use anything like the complex and opaque method proposed by Nasdaq. So, why not just make tick a function of price?
Increasing the tick based on the average spread will result in an increase in the average spread. The only question is by how much.
The Nasdaq indicates that they do not expect costs to increase for stocks where the tick would be increased. For this to happen, the higher tick would need to never bind. But the issue here is that the way that Nasdaq proposes setting the increments will almost certainly result in the tick being binding for at least some of the day. And for stocks near the lower bound of the category, it could be sizeable fractions of the day, leading to higher trading costs.
To see why using the average as the basis for setting the minimum spread (i.e., tick) is so problematic, consider the following simple example. Suppose the average spread on stock XYZ is 3 cents, but it takes on 5 different values over the day: 1, 2, 3, 4, and 5 cents. Assume for simplicity that these values occur in equal proportions to make the math clean. If we were to set the tick to the average of 3 cents, the 1- and 2-cent spreads are no longer allowed and would need to be rounded up to 3 cents. Now the five possible spreads going forward would be 3, 3, 3, 4, and 5 cents, and so the new average spread under Intelligent Ticks for this stock would be 3.66 cents, a 20% increase – even though the underlying distribution of the "true spread" didn't change.[2]
Another way to see this is, suppose the Nasdaq increased the tick to 2 cents on a stock with an average spread only slightly above 2 cents, as per their methodology. If Nasdaq's assumption of "no change in spread" were correct, the stock's average spread would remain unchanged, with the average spread now being only slightly above the new (higher) 2-cent tick. But now this stock is in the exact same situation as a stock whose average spread is currently at or near 1 cent, as both the 1-cent and 2-cent spread stocks have average spreads that are very close to their respective ticks. In their proposal, Nasdaq argues that reducing the tick by half for stocks with a 1-cent average spread would cause a reduction in their average spreads. But wouldn't that imply that our 2-cent stock above would also be better off if its tick were halved as well, back to its original 1-cent level ? Clearly, it is inconsistent to assume that increasing a tick to a level near to the average spread would have no consequence, while at the same time assuming that, should a stock find itself in that position after the change, reducing the tick would result in a reduction in spreads. (Though if it were true, an even "Intelligenter" proposal would be for Nasdaq to increase the tick on all stocks to just below their average, which would have no effect on spreads, and then immediately reduce them back to their initial level, which would reduce spreads. Then repeat until spreads hit negative infinity, or wherever they want them to be!)
Nevertheless, whether this matters or not could easily be checked by doing the same thought experiment I did above, but on actual data. Specifically, using the historical data on bid-ask spreads used in their proposal, they would simply need to round the spreads to the nearest new tick using their Intelligent Tick methodology. Of course, for spreads smaller than the new tick, the spread would need to be rounded up (since the spread cannot be zero). Such a simple calculation could show whether or not the issue I describe above is economically meaningful, at least given the current state of the market. I suspect it is, but only the data knows for sure.
Even if Intelligent Ticks was significantly better than a simple price-based approach, why use an average to define a minimum?
The minimum tick is meant to define the smallest spread allowable, so wouldn’t it be more reasonable to set the minimum tick based on some estimate of the minimum acceptable spread (i.e., a value where the tick is a meaningful percent of price)? Because the average spread is always above the minimum (unless the spread were constant), we know that we are setting the tick to a level that will constrain at least some of the spreads observed during the day.
While my numerical example above aims to make this point, a more intuitive way to characterize the issue is that the Nasdaq approach is like saying it’s safe to give everyone in the house a beer because the average age of people in the house is over the minimum drinking age of 21. But it’s not the average age that matters – it’s the distribution of ages that matters. This is precisely why we should think of the optimal tick in terms of the optimal minimum spread we are aiming to achieve. Just like in my example – if you were to use the minimum age in the house instead of the average age, you won't end up handing out ice cold Budweisers to 8-year olds!
Is Intelligent Ticks really the most efficient way to address the problem?
The Intelligent Tick regime is significantly more complex and its resulting pricing structure more opaque than the other “intelligent tick” regimes Nasdaq cites (e.g., Europe, Hong Kong, and Japan). For this to be implemented, average spreads need to be computed, a file generated and distributed, data uploaded to various trading systems, etc. across all exchanges, brokers, trading desks, technology providers, etc. every day. All of which has to be implemented, tested, and monitored. And this is only after we agree on the appropriate methodology.
For example, the proposal needs to address how to effectively handle stock splits, new listings, acquisitions, and other practical considerations. Even the averaging methodology needs to be evaluated since different methods give different values because spreads vary across the day. Volume- or traded-weighted averages would give more weight to the open and close than time-weighted, for example. And when we average across days, does it make sense to use a simple average, which gives the same weight to spreads that occurred months ago? Or should we use some type of exponential-weighting (and if so, what’s the half life??).
Lastly, even if all these technical details are worked out, the resulting tick structure would be opaque. PMs and traders, including less sophisticated retail traders, would somehow need to figure out whether a stock has a tick of 1 cent, 5 cents, 10 cents, etc. (But first they need to be educated that this regime even exists!). And even if they remember that stock ABC has a tick of 5 cents and stock XYZ has a tick of 10 cents, these may change over time, sometimes overnight.
And lastly, one final issue that shows a significant weakness of Intelligent Ticks over a simple price-based approach…
The Intelligent Ticks regime does not allow ticks to adjust quickly when the price of the stock moves significantly over a short-period of time. A simple price-based tick regime would respond immediately.
And even worse, for steep reductions in price, the fact that the tick is constrained by the increased tick means it could be months before the average spread falls enough for the tick to be reduced. By contrast, a simple price-based tick regime, like those used in other markets, could respond immediately.
Conclusion
In short, I think the Nasdaq raises some good points about the tick being outdated and in need of an update, especially given that so many stocks now trade at high price levels and markets are now much more automated. But I don’t see why we need to resort to such a complex approach that suffers from its own set of problems. The mechanism doesn’t even adapt quickly to significant price-level changes and may increase the cost on many stocks inadvertently.
Perhaps, a more sensible way forward (“Sensible Ticks Proposal”?) would be to use Nasdaq’s analysis to help inform a simple price-based tick regime, similar in spirit to that used in the markets they reference. Of course, part of me hates criticizing something because it’s complex. I feel like the Emperor in Amadeus – played by Mr. Rooney from Ferris Bueller – who criticizes Mozart's music for having “too many notes”.[3] But in this case, I think the market would be better served with some simple power chords, then a full symphony.
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 2020, The Bacidore Group, LLC. All Rights Reserved.
Footnotes
[1] See https://www.bacidore.com/post/the-nasdaq-tick-size-proposal-some-initial-thoughts.
[2] But since the 4- and 5-cent spreads are not possible, they would be rounded too. But in general, this would not change the argument above, though in a prior post, I argue why these spreads are more likely to be rounded up to higher values than down.
Comments