By Phil Mackintosh, Senior Economist, Nasdaq
We talked last week about how the new NMS II round lot plan doesn’t actually fix the odd lot problem.
With decimal ticks and odd lots allowed for trading, it turns out that the real market doesn’t care much for convention.
That, in turn, highlights the problems with regulations treating odd lots differently.
Chart 1: Visualizing why round lots with decimal ticks misrepresent how the real market (grey bars) typically looks
Chart 1 combines a lot of data that we know about the market. It shows how:
- Volume accumulates in a V-shape pattern, regardless of tick size (also see Chart 4 here).
- That helps explain why spreads on higher price stocks are wider, as the cost of extra liquidity requires a wider spread (blue bars).
- It also allows us to forecast how much spreads should tighten under the new NMS II dynamic lot In theory a 40-share round lot should be 40% of the 100 share round lot spread (purple bar), until that hits up against tick constraints.
- The chart also shows why, even with smaller round lots, even smaller odd lots will still rest inside the spread (grey bars are inside the purple bars).
This grey “real market” makes it difficult to fix the odd lot problem while decimal ticks and round lots exist. That is true even if each exchange aggregates its odd lots into “virtual” round lots, because that lot is still shown at the worst limit price of all the odd lot orders being aggregated (think the grey bars aggregated to a virtual blue bar).
Fixing this problem might sound like an academic exercise for market structure wonks, but research shows that excessively wide spreads hurt investors with higher costs and hurt issuers with higher costs of capital. And it’s easy to see from the U-shaped spread curve in Chart 2 that spreads are unnecessarily wide for low-priced and high-priced stocks.
Chart 2: The U-shaped spread curve shows that tradability problems exist for high- and low-priced stocks
Why not just eliminate “Round Lots?”
Of course, the easiest way to fix the odd lot problem is to treat all shares the same. In short, get rid of round lots. Given that modern markets have now evolved to trade in fractional shares, this would seem to make some sense.
However, it would create problems of its own. Because of decimal ticks, the NBBO for high-priced stocks could become uneconomically small in terms of the economic exposure and the spread that passive investors could capture.
For example, based on the V-shaped depth curves (Chart 1), we can show how AMZN would trade if round lots didn’t exist (Chart 3):
- AMZN currently trades with a 150-cent spread and 150 share depth. However, there are already many odd lots inside that NBBO. In fact, around 90% of the time, better prices are available in odd lots (blue bars).
- Using the same math and eliminating round lots, the market for AMZN is expected to form at an average two-cent spread on just one share of depth (yellow bars in Chart 3).
For passive investors, a two-cent spread on AMZN represents just 0.06 basis points (bps) in spread captured—hardly enough to make it worth posting a limit order. Similarly, crossing the spread results in a trade worth just over $3,000—hardly enough to make it worth a large mutual fund signaling their order.
This problem was one of the common concerns raised with the SEC’s plan to create an odd BBO.
Chart 3: Smaller lots should lead to tighter spreads in a stock like AMZN
We can use the same V-shaped volume curve math to estimate how the SEC’s proposed new 10-share round lot for AMZN should work. The math shows it is likely to lead to a 16-cent (0.5bps) spread (purple bars) for an average depth of 15 shares. Interestingly, when we looked at the impact of this on odd lots inside the spread last week, we found odd lots are likely to still be inside the AMZN quote more than 50% of the time.
Setting ticks allows us to target an appropriate depth of NBBO
Of course, if we can use this math to work out the spread that a 40-share round lot should trade at, we can also use the same math to work out what spread would be required to form, say, a $10,000 minimum NBBO. For AMZN, the answer is around four cents (small green bar in Chart 3).
Why $10,000?
It’s clear that was on the SEC’s mind when they proposed their dynamic round lots, because their smaller round lots ensure there is a minimum of $10,000 to qualify for the NBBO (See Chart 1 here).
It is also around the same size as the average trade, making the NBBO a benchmark that most trades should “expect” to be able to execute at. We recently found that the average trade size for retail is less than $8,000. Market-wide data also shows that the average trade size remains under $10,000 for most stocks (Chart 4).
It also makes sense when we consider how NBBO also currently protects investors. For example, Rule 605 requires execution quality to be measured vs. the NBBO quote, NMS Rule 611 tries to limit trade throughs, and our own analysis shows very few orders trade through the quote.
Chart 4: The NBBO should reflect at least the average trade value
What happens if you set ticks just right?
As we’ve seen, eliminating round lots fixes the odd lot problem. But for high-priced stocks, it likely makes spreads and spreads and depth much worse for traders.
However we have shown here that widening ticks in high-priced stocks would allow the market to set NBBO to an optimal level, where depth was just high enough to complete a trade, but it also low enough for spreads to be as competitive as possible.
We show how this works in practice below.
- Chart 5a shows the same “real market” as Chart 1.
- However, we color each order to help track what happens when decimal ticks are removed.
- In Chart 5b, we put the stock in a 10-cent tick bucket.
Chart 5a: Quote increments in the real market
Chart 5b: With fewer increments, odd lot limits would have to choose to accumulate at more economic increments
You see the odd lots at one cent (blue), three cents, (orange) five cents (green), seven cents (yellow) and 12 cents (light blue) all decide to accumulate now at the 10-cent tick. Pennying is not possible. Instead, if the 17 share order at the back of the queue want a faster execution, they need to offer 10 cents better, providing more economically significant price improvement to buyers.
Spread costs don’t need to go up
If the ticks are set so the spread is minimized, costs to investors should not go up as spreads should fall, given any sized lot can now form the NBBO.
In the example in Chart 5b, 50 shares are expected to form at each tick. If the cost of a 10-cent tick was too high, a five-cent tick would compress the spread but also see depth at the quote fall to 25 shares.
In either case, it’s expected that less than 100 shares will form the NBBO. But more importantly, we can use tick size to set depth. That allows all shares to be treated the same, in contrast to a dynamic round-lot market.
By the same logic, tick-constrained stocks have too much depth at too wide a spread, so they should be allowed to trade in sub-penny ticks. In doing that, the math says their spread costs and depth would fall proportionately to the tick change. Again, that could be done in a way that optimizes spreads and depth for each stock.
This concept is something that was considered in the consortium of brokers, buy-side, market makers and exchanges that participated in Nasdaq’s intelligent tick proposal.
The group found that a European–style tick-price-ADV grid actually made spreads on some stocks worse. That was especially true for ETFs that can have tight spreads without being liquid.
Instead, the intelligent ticks proposal suggested re-using tick-pilot groups and assigning stocks to buckets based on their current tradability. Chart 6 shows how the buckets (colors) were created based on what spread stocks traded now. Mathematically, no stock has a tick wider than its spread now. In fact, tickers at the bottom of each color zone are close to trading with a one-new-tick spread. Doing this:
- Ensured no stock has a wider tick that its trading spread now.
- Allows the market to set optimal ticks so that minimum depth for an NBBO is maintained.
- Stocks could be “promoted” or “demoted” into other tick buckets as trading changed.
Chart 6: Setting ticks based on market tradability now leads to more competitive quote increments
Even though this was done based on round lots forming the NBBO, we found it led to more consistent queues and spreads across stocks. That makes building algorithms simpler and quotes more competitive.
Allowing this to be done without round lot constraints results in an even more consistent market, with more stocks able to trade with optimal spreads and depth.
Modern markets probably don’t need round lots
Round lots are an artifact of old, human-driven markets. Given modern computer trading systems frequently disregard round lot conventions, it seems inconsistent that regulations designed to modernize markets still want to retain this artifact of the past.
That’s especially true when most countries around the world, taking a data-driven approach, have already eliminated round lot rules and replaced them with intelligent tick regimes.
If we’re really trying to modernize markets, maybe that’s what we should do too.
