Thirty Times Wider, Six Times Narrower

One Spread, Two Units

A trading desk describes a fill in basis points. A quote is published in cents. Those are not two ways of saying the same thing, and the gap between them is the oldest unresolved question in the design of the American price grid.

Rule 605 sits exactly on the fault line. It requires every market center to publish average effective spread and average price improvement in dollars per share, and it never once publishes the price of the share. The unit is the quote's. The economics are not.

So we supplied the missing column. Every symbol in the June 2026 Rule 605 files of the six largest wholesalers was joined to its median prior close for that month. The match covers 2,364 symbols, which is 18.7% of the symbols in those files and 80.7% of the executed shares. Everything below is computed on the matched set, share-weighted, with the six centers pooled and no firm identified.

One exclusion first. Rule 612 permits quoting in hundredths of a cent below a dollar a share, so sub-dollar names face an entirely different grid and cannot sit in the same table as the rest. There are 122 of them and they are not small: 16.3% of the priced shares, at an average price of forty-one cents. They are held out of every figure that follows, and we come back to them at the end.

The Inversion

Rule 605 lets you recover the spread the client actually faced. Price improvement is the distance from the quote. Effective spread over two is the distance from the midpoint. Add them and you have half the quoted spread at the moment the order arrived. Compute it in both units and the result is this.

Exhibit 1

Click to expand

In cents the spread rises thirtyfold from the cheapest band to the priciest. In basis points it falls more than sixfold. The same number, converted, reverses the ranking of the entire market.

Read the columns against each other and the disagreement is total. In the unit a regulator writes rules in, the $250 stock is thirty times the problem. In the unit a portfolio manager is charged in, the $2 stock is nearly seven times the problem. There is no arithmetic that reconciles them, because they are not measuring the same thing. One is a distance and the other is a rate.

Notice also where the basis-point column stops moving. From fifty dollars upward it is 3.9, three times running. Everything interesting happens below twenty-five dollars, and that is not where the cents column would send you looking.

Where the Grid Binds

Double the half spread and you have the full quoted spread, which can be compared directly against the minimum increment those quotes are allowed to move in.

Exhibit 2

Click to expand

In the $1 to $5 band the average quoted spread is 1.35 cents. In the $5 to $10 band it is 1.48 cents. Those quotes are not near the floor in some loose sense. They are sitting on it. A spread cannot be narrower than one tick, so a name whose natural spread would be six tenths of a cent is quoted at one cent, and the queue behind that quote does the rationing that price would otherwise do.

Now look at the right side of the same chart. At $250 and above the quoted spread averages 40.15 cents, which is forty ticks. The grid is not a constraint there in any meaningful sense; it is a rounding convention. The identical rule is binding at one end of the tape and invisible at the other.

This is the observation the Nasdaq working group put in print in December 2019, in a paper called Intelligent Ticks. Its framing was arithmetic and hard to argue with: a one-cent tick is 100 basis points on a one-dollar stock, ten basis points at ten dollars, one basis point at a hundred, and less than a tenth of a basis point on the names that were then trading above a thousand. In every case the tick is the same. What that paper could not do, and what the amended Rule 605 files now allow, is show what the constraint costs the person on the other side of the fill.

What the Client Keeps

Price improvement in cents is not comparable across price bands, for the reason Exhibit 1 just made obvious. The ratio is. Of the half spread that was available, what fraction did the client actually get? Numerator and denominator carry the same units, so the answer is a pure number and the whole cents-versus-basis-points problem drops out.

Exhibit 3

Click to expand

The pattern breaks at ten dollars and it is not subtle. Below it, clients keep 37.5% and 44.6% of the spread available to them. Above it, every band lands between 54% and 63%.

The tape splits almost exactly in half by volume. The cheaper half faces a spread four times wider in proportional terms and keeps 40% of it. The dearer half faces a spread four times narrower and keeps 60%.

Four hundred and thirty-nine symbols carry that first half. This is not a long tail of neglected microcaps; it is a small, extremely heavily traded set of low-priced names, and it accounts for one of every two retail marketable shares executed last month.

One composition point belongs here rather than in a footnote. Exchange traded products return a systematically smaller share of the spread than common stocks in every price band, and they are 43.2% of the sub-$10 volume, so they pull the pooled figure down. Split the two groups apart and the picture moves in both directions at once.

Restricted to common stocks the returned gap narrows, 49.9% against 63.1% rather than 40% against 60%. The proportional spread gap widens, five times rather than four. Anyone reading this as a story purely about equities should use the first and third rows, and the conclusion survives on either cut.

Six Names

Band averages are the right way to establish a pattern and the wrong way to feel one. So here are six real symbols, picked by a rule fixed before the numbers were computed: the common stock with the most executed shares in each price band. Exchange traded products are excluded from that rule, because ranking on raw volume otherwise returns a leveraged fund in every low-priced band, and those are not what most portfolios trade.

The last column is the one that travels. The client cost against the midpoint is the effective spread over two, which is the standard measure, and it restates cleanly as dollars on a million dollars traded. That is a number an investment committee can read without a glossary.

Look at the quoted spread column in the first three rows. 1.01 cents, 1.01 cents, 1.01 cents, on a $4 stock, a $5 stock and a $15 stock. Those quotes are not close to the floor. They are the floor, and the price underneath them varies by more than three times without the spread moving at all.

That is the mechanism stated as plainly as the data allows. The grid does not know what the share costs. A penny on OPEN at $4.47 is 22 basis points of round-trip spread. The same penny on AAL at $14.98 is 7. Same quote, same tick, same wholesalers, three times the proportional cost, and the only thing that changed is the denominator.

Notice also that the returned column does not track the cost column. OPEN returns 64.5% of its spread to the client, better than INTC and better than MRVL. It is still the most expensive name on the list by a factor of six, because 64.5% of a wide proportional spread costs more than 64.7% of a narrow one. A high improvement rate on a tick-constrained stock is not the same thing as a cheap execution, and a report that shows only the rate will not tell you which one you got.

For orientation rather than as a sample, the same calculation on six names most readers will recognise. This list was written down before the numbers were pulled, and nothing about the market as a whole should be inferred from it.

Ford and Pfizer are both quoted at essentially one tick and both return roughly three quarters of that spread, which is about as well as this tape gets. They still cost two to three times what Apple costs per dollar traded. Apple returns less than half its spread and is among the cheapest names here.

An adviser rebalancing ten million dollars pays roughly $4,000 in spread cost doing it in names like OPEN and roughly $340 doing it in names like Apple. Neither figure appears in any report either of them receives.

One caution on reading these rows. They are marketable retail held orders routed to six wholesalers, pooled, for one month. They are not what an institutional or adviser algorithm receives, they are not attributable to any single broker, and they say nothing about how any firm handled any particular order. What they do establish is the resolution that is publicly available when a rule requires the disclosure.

The Objection That Survives

The tempting conclusion is that the tick is doing all of this: a floor holds the quote artificially wide in cheap names, the wholesaler improves inside it by only a fraction of a cent, and the client absorbs the difference. That story is partly right and it is worth being precise about which part.

Rule 605 reports realized spread alongside effective spread. The difference between them is price impact, which is what the market maker gives back to adverse selection after the fill. Splitting the half spread three ways separates what the client gets, what the market maker keeps, and what neither of them keeps.

Exhibit 4

Click to expand

In the $1 to $5 band, of 25.8 basis points of half spread, the client receives 9.8, the market maker retains 5.1, and 10.9 basis points is price impact. The largest single component is not anybody's revenue. It is the cost of being run over, and it is more than seven times the proportional impact of any band above twenty-five dollars.

Cheap stocks are genuinely more dangerous to make markets in. Anyone arguing that the grid alone explains the gap has to explain that column away, and cannot. A wider proportional spread in low-priced names is partly compensation for real risk, and a finer tick does not make the risk go away.

The rest of the column is where the argument actually lives. Retained capture in the cheapest band is 5.1 basis points, against 0.2 to 1.2 in every band from five dollars to two hundred and fifty. Whatever else is true, the cheapest names are also the most profitable per dollar traded, and they are half the volume.

One more honest note. We tested whether the order-size gradient published in our previous piece is really this price effect wearing a disguise. On the unit-free measure, controlling for price directly, the size coefficient falls from -0.079 to -0.063 and stays highly significant, retaining 80% of its magnitude. Both effects are real and they are separable. Neither is a restatement of the other.

Quoting Is Capped. Trading Is Not.

Here is the part that makes the price-improvement industry possible, and it is a single drafting choice.

Rule 612 prohibits a market participant from displaying, ranking or accepting a bid, offer, order or indication of interest in an increment finer than the minimum. It says nothing about the price at which a trade may print. There is no minimum increment for an execution. The Commission's reason was that quoting in trivially better increments lets a participant leapfrog a displayed order for almost nothing, which would discourage anyone from displaying at all. A better execution price harms no displayed quote, so it was never restricted.

The entire off-exchange price improvement business operates in the space between a coarse quoting grid and a fine execution grid. Narrow the quoting grid and you narrow the space.

That is the mechanism behind the whole of Exhibit 3. In the $1 to $5 band the quote averages barely more than one tick wide, so there is two thirds of a cent of room between the quote and the midpoint, and the client receives 0.26 of it. At $250 and above the quote is forty ticks wide, there are twenty cents of room, and the client receives 11.10. The constraint is not the wholesaler's willingness. It is the width of the box.

Which reframes what a finer tick would actually do. It would not hand clients a better price directly. It would narrow the displayed spread, which lowers the benchmark that price improvement is measured against, and it would compress the room the improvement is carved out of. The reported improvement figure would very likely fall. Whether the client ends up better off depends on whether the quote narrows by more than the improvement shrinks, which is exactly the question the reported statistics are not designed to answer.

Where the Money Actually Is

If a coarse grid holds quotes artificially wide, the obvious inference is that whoever quotes them keeps the difference. It is the natural reading of everything above, and it is worth testing directly rather than asserting, because the data does not support it.

Call a name pinned when its quoted spread sits at or under 1.10 cents, essentially the narrowest quote Rule 612 permits at a dollar and above. Then compare pinned and unpinned names inside the same price band, so the basis-point denominator is roughly held still and the only thing varying is whether the quote had room to move.

In all five bands the market maker earns more per dollar traded where the quote is wider than one tick, not where it is pinned against the floor. In the two bands carrying the most tick-constrained volume, $5 to $25, retained capture on pinned names is negative.

That is the reverse of the intuition, and it is not a rounding artefact. A pinned $1 to $5 name returns $189 per million to the liquidity provider. A name in the same price band whose quote is wider than a tick returns $1,287, nearly seven times as much. The pattern repeats at $5 to $10, at $10 to $25, at $25 to $100 and above $100.

The previous section explains why. Rule 612 caps the quote, not the trade. A market maker who thinks a name deserves six tenths of a cent cannot display that, but nothing stops another one from executing inside the displayed quote at a finer increment, and nothing stops a third from beating it. Competition does not stop at the edge of the grid; it moves into the sub-penny space behind it. The floor freezes the display. It does not freeze the economics.

Which relocates the profit pool. It sits in names whose spreads are wide for reasons the grid has nothing to do with: thin books, real volatility, genuine adverse selection. Those are the rows with the large price impact numbers beside the large retained numbers. The $1 to $5 unpinned cohort carries a 46 basis point half spread and gives 16 of it back to price impact, which is a dangerous business being paid for being dangerous, not a rent.

It also changes what the client in a cheap pinned name is actually paying for. Of the 17.42 basis points of half spread on pinned $1 to $5 names, 6.60 comes back as price improvement, 8.92 is price impact, and 1.89 reaches the market maker. The dominant cost is not dealer margin. It is the cost of trading against someone better informed, and no tick regime abolishes that.

Payment for order flow belongs on the same axis, since it is paid out of retained capture rather than alongside it. Scaled to the flow we can price, it runs about 0.36 basis points of notional, roughly $36 per million dollars traded. On a ten million dollar rebalance that is a little over three hundred and fifty dollars of payment, moving from the liquidity provider to the broker, out of a pool the client never sees itemised.

None of that makes the grid harmless. The client in a pinned name still faces a proportional spread several times wider than the client in an expensive one, and that is the finding of this piece. What the evidence does not support is the further claim that the constraint is quietly enriching whoever quotes it. On this month of data it is not. The cost of a coarse grid lands on the client, and it mostly does not land in anyone else's revenue line.

What Arrives in November 2027

The Commission adopted a partial answer on September 18, 2024. For NMS stocks at a dollar or more, the minimum increment becomes half a cent where the time-weighted average quoted spread over a six-month evaluation window is a cent and a half or less, and stays at a cent otherwise. Two increments, reassigned twice a year, on windows running January through March and July through September.

It is not in force. Cboe and Nasdaq petitioned for review, the Commission granted a partial stay in December 2024, and although the D.C. Circuit denied the petition on October 14, 2025, the compliance date has moved twice since. The operative document is an exemptive order issued on June 11, 2026, which relieves compliance with the amended Rule 612 until the first business day of November 2027.

Everything in this article is therefore a measurement of the uniform one-cent grid, taken fifteen months before it changes. That is the useful thing about it. There will be no second chance to observe the baseline.

How much of the tape will the new increment reach? The rule's test is a time-weighted quoted spread, which Rule 605 does not report and we cannot compute. What we can compute is an order-time implied quoted spread, share-weighted. Treating a cent and a half as the cut, and flagging clearly that this is a proxy for a differently constructed statistic rather than the statistic itself:

585 symbols, 26.1% of the priced names at a dollar or more, carry 67.7% of the executed shares. Their half spread averages 10.3 basis points and their clients keep 48.3%.

The order of magnitude is corroborated from an unrelated direction. An exchange operator preparing for implementation put the tick-constrained set at roughly two thousand symbols, about 17% of names and 60% of volume. Different data, different method, same shape: a quarter of the symbols and roughly two thirds of the shares.

Two things follow. The half-penny increment is aimed at very close to the right set, and that set is where clients currently keep the least of their spread. And the change lands on a majority of retail share volume rather than a curiosity at the edge of the tape, so the reported improvement statistics will move for reasons that have nothing to do with anyone's conduct. Anyone comparing 2027 execution quality against 2026 without adjusting for the grid will be comparing two different rulers.

Finally, the names held out at the start. The 122 sub-dollar symbols already quote in hundredths of a cent, the finest grid in the market, and they are 16.3% of the priced shares. Their half spread averages 65.6 basis points and their clients keep 38.4%, which is no better than the cheapest band above a dollar. A finer grid, by itself, did not deliver a better share of the spread. Whatever the case for proportional ticks, it is not that the tick alone determines the outcome.

The honest summary is narrower than either camp usually states it. The grid is absolute, the economics are proportional, and the mismatch is largest and most binding in exactly the half of the tape that carries the most retail shares. That is a measurement, and it holds regardless of what anyone concludes from it.

Method and Sources

Execution figures are computed from the Rule 605 monthly files published by six wholesalers for June 2026, restricted to marketable order types, comprising 395,112 reported cells, 12,643 symbols and 65.5 billion executed shares. Market centers are pooled and no firm-level result is reported. Share price is a June 2026 median prior close per symbol from our own tape and covers 2,364 symbols, 80.7% of executed shares; unmatched symbols are excluded rather than imputed. Half spread is price improvement plus half the effective spread, which recovers the quoted half spread at order receipt. Price impact is effective minus realized spread, halved. All statistics are weighted by executed shares. The legacy Rule 605 tape excludes odd lots and orders of 10,000 shares or more, does not report quoted spread, and blends all originating brokers routed to a given center, so nothing here supports per-broker attribution. The six named symbols are selected by a rule fixed before the figures were computed, the common stock with the most executed shares in each price band, with exchange traded products excluded from that rule; the household names are a list declared in advance and offered for orientation, not as a sample. Cost per million traded is the effective spread over two expressed in dollars. Everything is descriptive; no causal effect is identified. Rule text is quoted from 17 CFR 242.612 and from the relevant SEC releases, including Release 34-101070 (September 18, 2024) and the exemptive order at Release 34-105656 (June 11, 2026). The 2019 tick proposal referenced is Nasdaq's Intelligent Ticks: A Blueprint for a Better Tomorrow.