Share of the spread Robinhood customers paid on S&P 500 market orders in August 2026, against Schwab customers on the same stocks and order sizes. Both brokers take payment for order flow.
Extra paid over 18 months of recovered filings, compared with the prices Citadel charged once disclosure was coming. The old reports cover only part of its orders, so the true figure is higher.
Citadel's share of the spread on S&P 500 stocks, the category the new summary reports break out, against everything else, on the same retail orders. In August it lost an estimated $10.7M on S&P 500 orders and made $21.3M on the rest.
What customers of 8 of the 10 other brokers paid beyond Schwab-level execution on the same orders. Schwab takes payment for order flow too, so this is not an unreachable bar.
Best execution is the law. The filings show it isn't happening.
Every broker owes its customers best execution. FINRA Rule 5310 requires "reasonable diligence to ascertain the best market for the subject security and buy or sell in such market so that the resultant price to the customer is as favorable as possible under prevailing market conditions," and requires firms to review execution quality regularly and rigorously against competing markets. The first comparable reports show what those reviews have been missing.
- Better prices were on offer, and brokers didn't take them. On the same stocks and order sizes, G1 gave better prices than Citadel and Virtu. Citadel and Virtu still filled 55% of the market-order dollars handled by the seven wholesalers studied. A review that compares competing markets, as Rule 5310 requires, should have moved that flow.
- Better execution was possible all along. Robinhood's own figures went from 37–41% of the spread to 24.5% once comparable reporting was coming. The market did not simply get easier: over the same stretch Virtu's own figure for S&P 500 market orders went from 38% to 42%. What changed is that someone would be checking. FINRA has already fined Robinhood under Rule 5310 for reviewing only the firms that paid it for orders, without reasonably considering the prices other markets offered.
- Customers who send the easiest orders still pay more. Robinhood's orders are among the least costly to fill of any large broker's, yet its customers paid more than customers of Schwab, Fidelity, Morgan Stanley / E*Trade, Merrill and Vanguard. Robinhood is paid 12.35% of the quoted spread on those orders, so its payment depends on the size of the spread, not on how good a price its customers get.
- Averages can be gamed, so order-by-order review matters. Citadel loses money on S&P 500 orders, the category the new summary reports break out on its own, and earns it back on everything else. A review of averages rewards that. Reviewing actual prices order by order does not, which is why We the Investors urged FINRA to keep and extend order-by-order review (our FINRA comment).
- The displayed price is the only floor retail orders have. Today wholesalers fill only 0.1–0.8% of S&P 500 market-order dollars, and 0.5–1.8% for other stocks, at a price worse than the best displayed quote when the order arrives. Rule 611 is the rule that protects that price. The SEC's own analysis of removing it says wholesalers "may begin to trade through displayed round-lot quotes," "resulting in worse prices" (our Rule 611 comment).
Where the money goes
Robinhood collected $4.2B in payments for its customers' stock and options orders from 2022 Q3 to 2026 Q2, according to its own Rule 606 reports. Those payments come out of the same spread its customers pay. The other tabs show the detail: Robinhood against other brokers, what changed before the rules, and how the wholesalers price.
Robinhood sends some of the easiest orders of any large broker and gets one of the worse deals for them. Comparing the same stocks, order sizes and order types, weighted by Robinhood's own order mix, Robinhood's customers paid more than customers of Schwab, Fidelity, Morgan Stanley / E*Trade, Merrill and Vanguard. They paid less than customers of Interactive Brokers, Webull, SoFi, Public and tastytrade.
Robinhood says comparisons mislead because its customers send different orders. The filings say the opposite. Five minutes after the trade, Robinhood's market orders have moved prices less than those of every large broker: 0.48 basis points, against 1.29 for Schwab and 1.45 for Vanguard. Its customers still pay 2.88 basis points of every dollar traded. About 1.50 goes back to Robinhood as payment for order flow and about 0.90 stays with the wholesaler.
What Robinhood's brief says
"Rule 605 and the Limits of Efficacy" was published by Robinhood's chief economic advisor on September 28, 2026, days before the first broker reports were due. It argues that the reports show what happened, not what would have happened with different routing, and that they cannot reveal whether one broker's customers send harder orders. It cites Glosten and Milgrom (1985) and work finding 43–114% more price impact from retail orders on exchanges than from retail orders filled by wholesalers.
The point about order mix is testable. The detailed reports break every broker's orders down by stock, order size, lot type and order type, so brokers can be compared on the same orders. That is what follows.
Same stocks, same order sizes: who gives up more of the spread?
Each row pairs Robinhood with one broker on the cells both reported: stock × order size × lot type (fractional, odd, round). Both sides are weighted by Robinhood's own dollar volume, so Robinhood's order mix is held fixed.
Robinhood customers' extra cost, or saving, against each broker in one month
For each matched cell: Robinhood's dollar volume × (Robinhood's effective spread − the other broker's) ÷ 2, summed over market orders under $200k ($110B–$135B of Robinhood volume, depending on overlap). Positive means Robinhood customers paid more.
Fidelity and Vanguard report filling big-company orders better than the midpoint on average, which is unusual. Leaving them out, Robinhood still trails Schwab, Morgan Stanley / E*Trade and Merrill by $4.0M–$11.5M a month.
The easiest flow, the biggest cut
Every basis point a customer pays relative to the midpoint ends up in one of three places. Adverse selection is the market maker's real cost: how far prices move against it after the trade. Payment for order flow goes back to the broker. Wholesaler margin is what is left. Each broker is measured on the orders it shares with Robinhood; Robinhood's row uses its pairing with Schwab, which covers 99.9% of its volume.
Basis points of the dollars traded, relative to the midpoint when the order arrived. Payments are modeled from each broker's Q2 2026 Rule 606 terms: Robinhood 12.35% of the quoted spread, Webull 13%, per-share rates for the rest, none for Fidelity and Vanguard, and none from third parties for Merrill, which routes to its own affiliate. Interactive Brokers is left blank because one report mixes its Lite accounts (paid 12% of the spread) and Pro accounts (not paid). The model puts Robinhood's August market-order payments at $20.2M; its Rule 606 reports show $13.8M–$29.7M a month for market orders in Q2 2026 (including extended hours and larger orders).
Whose orders are hardest to fill?
How far prices moved after the trade (basis points), Robinhood / other broker, on the same orders. Lower means easier orders. Shaded where Robinhood is lower.
| vs | 1 sec | 15 sec | 1 min | 5 min |
|---|
A smaller, mixed picture
Marketable limit orders are a small part of Robinhood's flow ($8B a month against $135B in market orders). Here Robinhood's cost (basis points, matched) beats Schwab, Morgan Stanley / E*Trade, Merrill, Interactive Brokers, Webull and tastytrade, and trails Fidelity and Vanguard.
| vs | Robinhood | Other | Coverage |
|---|
The headline numbers each broker published
Unadjusted figures from each broker's own summary report, market orders. They tell the same story as the matched comparison.
| Broker | Avg order | Share of spread paid | Eff. spread (bps) | % price improved | At or better than quote |
|---|
Payments from wholesalers, by quarter
Net payments each broker reported receiving for routing customer orders, summed across venues and order types.
Robinhood's stock-order payments reached 85% of Schwab's in Q2 2026 ($141.8M vs $166.6M). Its options payments ($356.6M) now exceed Schwab's ($331.4M). Schwab includes TD Ameritrade until its accounts moved in 2024. E*Trade's filings for late 2023 and early 2024 are missing.
Payment terms on marketable stock orders
| Broker | Terms |
|---|---|
| Robinhood | 12.35% of the quoted spread |
| Webull | 13% of the quoted spread |
| Interactive Brokers (Lite) | 12% of the quoted spread |
| Public | about $0.0017 per share (via Apex) |
| SoFi | about $0.0018 per share (via Apex) |
| tastytrade | about $0.0015 per share |
| Morgan Stanley / E*Trade | about $0.0011 per share (self-directed accounts) |
| Schwab | about $0.0010 per share, same for every market maker |
| Fidelity, Vanguard | No payment for order flow from wholesalers on stock orders |
| Merrill | No third-party payments; routes to its affiliate, BofA Securities |
Per-share figures are net payments per share on S&P 500 market orders, computed from each broker's reported totals. A payment set as a share of the spread grows with the size of the spread, whatever price the customer gets.
Execution improved when disclosure was coming, which means it could have been better all along. Robinhood's own reported figures sat at 37–41% of the spread through 2023 and 2024, then came down to 30.4% in Q1 2025, 32.2% in Q4 2025, 28.1% in Q1 2026 and 24.5% in 2026 Q2. Its reported savings per 100 shares went from $2.21 at the end of 2024 to $6.11, 2.8× as much.
Robinhood's own numbers, before and after the rule
Share of the spread paid, as Robinhood published it each quarter (computed for Robinhood by S3 Matching Technologies). Recovered from archived copies of Robinhood's execution-quality page. Robinhood changed the orders covered in 2021: early figures cover orders up to 1,999 shares, later ones orders of 1–99 shares.
Archived copies show Robinhood's page kept its Q1 2025 figures from April 2025 to January 2026; Q2 and Q3 2025 never appeared on it. Quarters with no figures in this series: 2023 Q3, 2025 Q2, 2025 Q3.
$153M in extra costs, using only Robinhood's own figures
Robinhood is paid 12.35% of the quoted spread on these orders, so its reported payments reveal the total spread its customers' orders faced each quarter. Customers' cost is that spread × the share they paid ÷ 2. The bars show the extra cost in each quarter compared with the share Robinhood delivered in 2026 Q2.
Conservative choices: the 12.35% rate overstates what Robinhood earns on extended-hours orders, which understates the spread base; improvements are applied in percentage points rather than proportionally; unpublished quarters are left out. Check: the spread base this method gives for Q2 2026 is within a factor of 1.3 of the one measured directly in the August 2026 detailed report.
Citadel sharpened its prices on the stocks that get compared
Share of the spread paid at each wholesaler, S&P 500 stocks vs everything else, from their monthly reports. The amended rule was proposed in December 2022, adopted in March 2024, and took effect in August 2026 after an extension announced in September 2025.
Citadel priced S&P 500 orders about 8 points better than other stocks in 2024 and about 19 points better in 2026. Its S&P 500 figure went from 41% to 30% while Virtu's went from 38% to 42%. Only 4 months of Citadel's 2024 filings could be recovered.
$66M over 18 months of recovered filings since the proposal
Each month: what Citadel's customers paid on S&P 500 market orders, minus what they would have paid at the share of the spread Citadel charged in 2026 (27.0%). About $3.7M a month. The old reports cover round-lot orders of 100–9,999 shares only, so this understates the total.
Two firms handle most retail orders, and on the same stocks they do not offer the best prices. Citadel Securities and Virtu filled 55% of the market-order dollars and 61% of the marketable-limit dollars handled by the 7 wholesalers in this study. On an identical basket of retail orders, G1, a much smaller firm, gave better prices than both. Citadel ranks number 1 of 7 on S&P 500 stocks and number 5 on everything else; its overall rank of 2 rests entirely on its S&P 500 pricing.
A loss leader on the stocks that get compared
For each wholesaler and group of stocks: what customers paid relative to the midpoint, minus adverse selection after five minutes, minus estimated payments to brokers, leaves the wholesaler's margin. Citadel is the only one that loses money on S&P 500 orders. On those orders its customers paid -0.09 basis points (below zero means fills better than the midpoint); Citadel then bore 0.26 in adverse selection and about 0.08 in payments, a margin of -0.44.
Payments are estimated from the brokers in this study: each broker's modeled payments × the share of its market orders its 606 sends to each wholesaler. Payments from other brokers are not counted, so margins are if anything overstated. Citadel reports filling S&P 500 orders of $20k and up better than the midpoint on average; these filings are unaudited.
Four years of routing to the same two firms
Share of each broker's routed S&P 500 orders sent to Citadel Securities and to Virtu. Robinhood sent 59% in Q2 2026.
Monthly percentages averaged by quarter. Public routed through Apex and named no wholesaler before 2025.
Best where everyone looks, middling where no one does
Every wholesaler priced on the same cells (stock × size × lot type), weighted by the pooled dollar volume of the retail brokers in this study. Only cells all 7 wholesalers reported are used (99% of the basket). Lower is better.
| Wholesaler | Share of $ | Share of spread paid | Cost (bps) | Kept after 1 min | Kept after 5 min | Adverse selection, 5 min |
|---|
Extra cost vs G1's prices on the same orders
Each wholesaler's own dollar volume × (its effective spread − G1's) ÷ 2, over cells both reported. Market plus marketable limit orders.
G1 is the benchmark because it gave the best prices on the full basket, so those prices were available. Against HRT, Citadel and Virtu cost $26M. Excluding S&P 500 stocks, the gap to G1 is $45M.
What a working market would look like
- Share does not follow price. In a competitive market the best-priced firms would win orders. G1 (15.9%) and HRT (15.7%) beat Citadel on everything outside the S&P 500 and beat Virtu everywhere, yet hold 32% between them.
- Prices differ by broker, not by order. Robinhood and Schwab route largely to the same firms. Robinhood's orders are easier to fill, yet get worse prices, and Robinhood is paid a share of the spread. Academic work found the same: wholesalers charge different brokers different prices for the same trades.
- Benchmarks invite gaming. Summary reports split S&P 500 stocks from the rest. The largest wholesaler looks best in the S&P 500 category, and loses money there doing it.
Every broker and wholesaler, by order size
Share of the spread paid, by order size, as each firm reported it. Darker is worse. Outlined cells report fills better than the midpoint on average.
What did each firm charge on one stock?
Pick one of the stocks retail investors traded most. Each bar is the share of the spread that firm's customers paid on that stock (lower is better), weighted by the same retail order-size mix for every firm. Hover over or tap a bar for the cost in basis points. Firms are shown where they reported at least half of that mix.
How we did this
- Every number on this page is computed by code from public filings. The pipeline (fetch, parse, compute, check, render) refuses to publish if any figure in the text was typed by hand.
- Checks. All 72 headline figures in the summary reports of the 18 firms studied are rebuilt from their own detailed reports, with a median difference of 0.16 percentage points. A second, independent implementation reproduces the key comparisons exactly. Quoted spreads in the old-format reports are rebuilt from their price-improvement fields, an identity confirmed on every new-format report tested.
- Share of the spread paid (E/Q) is the effective spread divided by the quoted spread, computed as the reports' own summary files do: the sum over orders of shares × effective spread, over the sum of shares × quoted spread.
- Brokers: Robinhood Securities, Schwab, Fidelity, Morgan Stanley (E*Trade), Merrill, Vanguard, Interactive Brokers, Webull, SoFi, Public and tastytrade, August 2026 detailed and summary Rule 605 reports. Wholesalers: Citadel Securities, Virtu, Jane Street, G1, HRT, Two Sigma and UBS.
- History: old-format monthly Rule 605 reports from each wholesaler, the timestored.com archive (checked month by month against the firms' own files over 52 overlapping firm-months; they match closely), and the Internet Archive for older Citadel files and Robinhood's execution-quality page.
- Payments: quarterly Rule 606 reports from the brokers and FINRA's 606 repository.
- Limits. One month of new-format data. Reports are self-published and unaudited. The S&P 500 list is current membership. Old and new report formats are not comparable with each other.
- Prior work: Schwarz, Barber, Huang, Jorion and Odean, "The 'Actual Retail Price' of Equity Trades".
Disclaimer
- Not advice. This page is educational and advocacy material from We the Investors. It is not investment, legal, tax or accounting advice, and it is not a recommendation to open, keep or close an account with any firm or to buy or sell any security.
- Source data. Every figure is calculated from public regulatory filings: broker and market-maker reports under SEC Rules 605 and 606, FINRA's Rule 606 repository, and archived copies of firms' own published statistics. These reports are prepared and published by the firms themselves. They are not audited by the SEC or FINRA, may contain errors, and may be corrected or restated. We have not independently verified them, and errors in the filings carry through to our results.
- Estimates, not findings. "Extra cost" figures compare what customers paid with a stated benchmark: another firm's execution of the same stocks and order sizes, or a firm's own later execution. They depend on the methods and assumptions described on the Method tab, including annualizing a single month and modeling payment for order flow from disclosed terms. They are estimates, not measurements of any individual's losses, damages or entitlement to compensation.
- No finding of wrongdoing. Nothing here is a finding or allegation that any firm violated any law or rule. Descriptions of firms' execution reflect what their filings show, read through the methods we describe. Characterizations and conclusions are the opinions of We the Investors.
- Your results differ. Execution quality varies by stock, order size, order type, time of day and market conditions. Past execution quality does not predict future execution quality. Figures for a broker or market maker are averages and may not reflect your own orders.
- No affiliation or endorsement. Firm names and trademarks belong to their owners. Mentioning a firm does not imply its endorsement of this analysis or our endorsement of it.
- Corrections. If you believe a figure or statement is wrong, write to press@wetheinvestors.org. We will review it and correct the page where warranted. The method, code and checks behind every figure can be shared on request.