Earlier this year, we took a deep dive into dark trades. We found that not only do more trades occur off-exchange for smaller stocks, but also the proportion of trades that occur at non-decimal prices also increases (green in Chart 1).
Chart 1: Proportion of all trades that are off-exchange, by price type
How do all those sub-decimal trades occur?
NMS rule 612 only requires limit orders to be in decimal increments. It says nothing about trades.
That means that orders that cross the spread and should instantly trade can be filled by a broker, off-exchange, at any price (in hundredths of a cent). However, other NMS rules require those fills to be no worse than the NBBO. As a result, sub-decimal fills usually receive “price improvement,” or a better price than the NBBO. .
Because that also results in less spread capture for the broker filling the trade, those brokers (wholesalers) typically focus on smaller, most commonly retail, orders that are also less likely to cause them adverse selection.
Today we take a deeper dive into those sub-penny fills by comparing the trade prices to the NBBO at the same time.
What we find is that for tick constrained stocks, 16% of sub-decimal fills are just 1/100th cent better than the NBBO (dark green bars in Chart 2), and 44% of all trades are improved by less than 1/10th of a cent (dark green line).
However, not all stocks are tick constrained. In fact, we recently found that around 30% of stocks trade at wider spreads, with many trading more than 25 cents wide. Not surprisingly, price improvement on those stocks (in cents) is much better. Only 11% of sub-decimal fills are 1/100th cent better than the NBBO (light green bars), and around 30% of all trades are improved by less than 1/10th of a cent (light green line), with roughly 35% of all those orders being improved by more than a whole cent versus the NBBO.
Chart 2: What prices sub-decimal fills occur at in cents (for tick constrained and wider spread stocks)
What does a “30 EQ” mean?
Importantly for investors, NMS Rule 605 requires all market centers to calculate execution quality metrics for orders they receive. That helps to show the proportion of orders being price improved as well as the average level of price improvement across different trade sizes. It also provides transparency into whether any “bad” fills seem to be occurring, showing up as price disimproved executions.
The industry often uses this data to compute a simplified “EQ score.” It represents the effective spread divided by quoted spread or NBBO (Chart 3 below) for all stocks traded.
Data shows that retail investors, on average, capture 30% of the half spread. In other words, they capture an average of about 15% of the total spread.
Chart 3: How EQ metrics work
Looking at PI as a percent of spreads
To fix the problem of comparing tick constrained and wide-spread stocks, seen in Chart 2, we can look at price improvement as a percentage of spreads, like 605 does.
Replotting the data as a percent of the spread (Chart 4), we see that the dark green line and the light green line essentially converge. In other words, the proportion of spread captured by the wider spread stocks is actually fairly similar to the spread captured by tick constrained stocks.
That means that although you see larger price improvement (in cents) on stocks with wider spreads, the savings (in %) are roughly the same.
Chart 4: Price Improvement as a percentage of spread (for tick constrained and wider spread stocks)
Does trade size matter?
The other important factor in 605 is trade size. Stocks with larger shares to trade are grouped, as we show in the color changes in Chart 5 below.
There is plenty of established research that shows that larger trades cost more to trade, increasing shortfall on an order. So it’s no surprise that we see price improvement fall as size increases. That’s true whether we look at impact as a percent of spread or in cents.
But it is a little surprising to see that the blue line, in particular, is not smooth. Interestingly, it seems like for:
- Small orders (left side), there seems to be “jumps” in the curve as it crosses each bucket cut-off. That may mean there is some cross-subsidization within buckets, even though that’s not clear from the cents of price improvement data.
- Larger orders (right side) see V-shaped spikes that dip at “very round” share quantities, meaning they get less PI than “less-round” quantities. It’s possible very-round share quantities are indicative of “more informed” investors – and are priced accordingly.
We’d also highlight that only a fraction of all sub-decimal trades occurs above 2,500 shares.
Chart 5: Price improvement falls as trade size increases (in cents and percent of spread)
Market-wide averages also blend a number of competing factors which blur the results in Chart 5.
- High-priced stocks create much larger trades (and so should cost more).
- But they are also usually among the largest market cap and most liquid stocks (and so should cost less, especially for larger trade sizes).
For that reason, splitting the results from the blue line above by market cap is interesting (Chart 6). It shows that the larger-cap stocks (dark blue line) get slightly better price improvement (in % of spread), especially for smaller orders (top left side of the chart), even though their spreads are usually much tighter (in basis points). However, that may be because of odd-lots available inside the NBBO.
Also interesting is that despite the downward sloping line generally, there is little difference in costs across market cap for trades of the same number of shares. Especially when we consider stock prices in a nano-cap are much smaller than a mega-cap, meaning the value traded is significantly lower as the lines fade from blue to yellow.
It would be interesting if our recent study, suggesting 605 might work better in dollars, would change this result. Especially when a cap for covered orders at (say) $200,000 would allow not only for odd lot orders to be included in 605 metrics but also encourage orders larger than the typical NBBO size to be worked, rather than included in EQ metrics. That may even help them interact more with institutional investors.
Chart 6: Price improvement across trade size broken out by market cap
Why is this important?
Sub-decimal orders are a unique subset of trades on the SIP for NMS stocks.
By taking a deeper dive into this data, we can see just how much spread those orders typically capture. Consistent with theory, the data shows that generally, larger and harder-to-trade orders cost more (spread capture falls).
Most sub-decimal fills are also mostly for small value trades, which supports our suggestion that notional 605, with a cap on covered orders that is closer to typical NBBO size, makes more sense than the current 10,000 shares.
Phil Mackintosh, Nasdaq Chief Economist, has 30 years of experience in the Finance industry, including roles on the sell-side, buy-side and at accounting firms, which included managing trading, research and risk teams. He is an expert in index construction and ETF trading and has published extensive research on trading, ETFs and market structure.