Market Structure Debates

Three kinds of claim, and only two of them are arguments about facts

Market-structure debates are loud because the same words are used for different kinds of claim. Sorting them is the skill this lesson teaches, and it is what turns an opinion into an analysis. An **empirical claim** is about what happened, and it can be measured: the average effective spread retail orders paid, how much of a company’s flow went to one wholesaler, how much a broker received in payment for order flow. These claims are decided by data, and the data is mostly in the disclosures of M22 — which is why that lesson came first. A **mechanism claim** is about why: internalisation concentrates flow, which may or may not reduce the competition that a public auction would create. Mechanism claims are testable in principle and hard in practice, because the alternative world never happened. And a **value judgment** is about who should bear a cost and what counts as fair: whether it is acceptable for a broker to be paid for routing, whether liquidity should be segmented by order type at all. No dataset settles a value judgment, and pretending otherwise is the most common dishonest move in this debate. Two further habits make the analysis honest. The first is **incidence**: when a policy changes, ask who actually pays, remembering that a rule aimed at firms lands on whoever cannot pass it on. A ban on payment for order flow reduces broker revenue; whether that shows up as commissions, worse fills or lower marketing spending is an empirical question, and the answer differs by broker. The second is the **conflict-is-not-harm** distinction from M22: a conflict of interest is a fact about incentives, and harm is a fact about outcomes. Payment for order flow is a genuine conflict — your broker is paid by the venue that fills your order — and whether it costs you anything is a measurable question. Conflating the two produces arguments that feel conclusive and are not. Sorting a typical claim — "Retail orders get 2 cents better than the NBBO on average.": Empirical — measure it in 605 data · "That is because wholesalers compete for the flow.": Mechanism — testable, but the counterfactual is unavailable ← · "Nobody should profit from routing my order.": Value judgment — a choice about who bears a cost

The five live debates

**1. Payment for order flow.** The case to ban it rests on the conflict (your broker is compensated by the venue), on concentration (a small number of wholesalers handle most retail flow, which may weaken competition for it), and on the fact that improvement is not the same as perfect competition. The case to keep it rests on measured outcomes: retail marketable orders generally receive price improvement relative to the displayed quote, spreads for retail size are tight, and the economics that pay the wholesaler come partly out of a spread the customer was never going to keep anyway. The unresolved core is the counterfactual — how much of that improvement would survive if the practice were prohibited — and no disclosure answers it. **2. Tick sizes and access fees.** A minimum increment is a constraint on competition: with a one-cent tick, nobody can quote a penny better and the spread cannot go below a cent. Smaller ticks narrow the spread and reduce the profit from queue position, at the cost of thinner displayed depth and less incentive to quote size, which hurts larger orders. Access fees are the same problem from the venue side: a cap on what an exchange may charge a firm that removes liquidity changes which venue is cheapest to trade on, which changes where orders go. The 2024 amendments — a half-penny tick for the most liquid names and a 10-mil access fee cap, with compliance relief to the first business day of November 2026 — are that trade-off legislated, which is why the debate did not end when the rule was adopted. **3. Order competition.** The 2022 proposal would have required many retail orders to pass through a competitive auction before being filled internally, on the theory that an auction generates more price competition than a wholesaler facing only its own inventory. The objection is practical: auctions take time, and time is exactly what a marketable order is paying to avoid; a customer whose order is exposed for a few hundred milliseconds may get a better price or may get a worse one after the market has moved. Both sides are arguing about the same mechanism with evidence that cannot include the counterfactual, which is why this debate has been the most contested of the five. **4. Dark trading and the tape.** Trading off a lit exchange protects a large order from being front-run, which lowers its impact — a real benefit. It also moves price discovery off the lit market: if most trading happens without displayed quotes, then the public prices that funds and indices are measured against are computed from a shrinking share of the market. The consolidated tape debate is the same argument about information: who is entitled to see the best prices, how fast, and at what cost. The 2020 Market Data Infrastructure rule rewrote how the tape works and added odd-lot information to the top of book — a step toward a fuller picture of the market that is itself contested by those who think the display should remain the definition of the market. **5. Trading hours.** Extending the session toward round-the-clock trading serves a global investor who wants to trade U.S. names in their own day, and it fragments liquidity into thinner hours with wider spreads. It collides with mechanics that assume a daily rhythm: most corporate actions and many announcements are scheduled around a close, settlement still runs on a business-day clock (M13), and the closing auction — the deepest, cheapest liquidity of the day (M11) — has no natural equivalent at three in the morning. The honest framing is not "should the market be open more" but "which participants are better served by a thinner market at an hour they chose". Each of the five has the same shape: a real cost on each side, a metric that measures one of them better than the other, and a value judgment about which cost should be minimized. That is why mastery here is not knowing the answer but being able to state the trade and the uncertainty.

What the evidence does and does not support

Some conclusions are well supported. **Retail marketable orders are executed at prices that generally beat the displayed quote**, and the disclosures that show it are audited and public. **The routing of retail flow is highly concentrated** in a few wholesalers, which is a fact from 606 reports rather than a claim. **Institutional execution is a different product**: larger orders walk the book, pay measured impact and are the reason desks exist at all (M17). And **the two tiers are segmented by order type**, so the retail experience of the market and the institutional experience of it are not the same market. Other conclusions are not supported by the evidence available. That retail investors are **harmed** by payment for order flow: not established, because nobody can run the counterfactual, and the observable metrics point the other way. That the practice is **harmless**: also not established, for the same reason — and a conflict of interest that cannot be shown to cost money today is still a conflict that shapes which venues exist and who pays to attract flow. That the **tick change will help or hurt investors on net**: depends entirely on what they trade and how they trade it, which is the definition of a question that needs a metric rather than a slogan. And that any of these debates is a matter of **good versus evil**: the participants are firms with incentives, and naming an incentive is not the same as proving an outcome. The mastery standard is a claim that can be checked and a judgment that is labelled. A sentence like "retail effective spreads improved by about a cent over the period in which this practice grew, and I still think a broker should not be paid for its customers’ orders" is honest: the first half is a finding with a metric and a period, the second half is a position, and a reader knows which is which. An argument that merges them is not more persuasive — it is just harder to falsify, which is the whole trick being avoided here. Beware the argument that cites a large number without a denominator. "Billions paid for order flow" is a fact about the industry; the question about a customer is basis points on a trade, and those are different conversations.

Sources

• The order competition proposal: auctions before internalisationSEC — Order Competition Rule proposal (December 2022)
• Tick size and access fee amendments, and their compliance reliefSEC — tick size and access fee cap amendments (2024) and subsequent compliance relief
• Market data infrastructure: the consolidated tape and odd lotsSEC — Market Data Infrastructure Rule (2020)
• The January 2021 episode and what it revealed about routing and settlementSEC — Staff Report on Equity and Options Market Structure Conditions in Early 2021
• Payment for order flow, internalisation and retail execution qualityFINRA / academic literature on wholesaler economics and retail price improvement
• Extended and overnight trading proposalsExchange rule filings on extended trading hours