The 40-Point Gap Hiding Inside the S&P 500

Two funds track the same index. In a typical week one sends well over half its volume to dealers off-exchange and the other about a fifth. Most ETF flow never sees a dark pool at all, and the channel it does use is the one nobody measures.

Vanguard's VOO and State Street's SPY both track the S&P 500. In the median week of our sample, 57% of VOO's volume printed to the non-ATS off-exchange channel, the bucket where dealers fill orders against their own inventory. For SPY the figure was 20%. Same index, same holdings, and a gap of roughly 38 points in where the shares actually change hands.

We quote VOO as a median rather than a single level on purpose. Its weekly reading swings from 39% to 83%, and four weeks in our window print off-exchange totals above 90%, which is not a real market outcome for a fund this size. IVV has a week at 109%, meaning the reported off-exchange volume exceeds the entire consolidated tape, which is impossible and tells you plainly that the numerator misbehaves for some symbol-weeks. SPY and SGOV show no such weeks. The volume-weighted figure for VOO is 65%, and we think that one is too high.

Three channels, and most ETF volume avoids the dark

Start with the plumbing, because the headline number gets misread constantly. Every trade in a listed stock or ETF clears through one of three channels. A lit exchange displays the quote in public. An ATS, or dark pool, matches orders without displaying them. The third channel is off-exchange and non-ATS, where a dealer facilitates the trade against inventory or preferenced flow rather than on a public book. Both off-exchange channels report to a FINRA Trade Reporting Facility, so a raw TRF print on its own cannot tell you which one you are looking at. What lets us pull them apart is that FINRA publishes ATS volume separately from non-ATS volume.

Now the part that surprises people who assume ETFs live in dark pools. Across our sample of 34 ETFs, the split was 61.1% lit, 26.2% non-ATS off-exchange, and 12.7% ATS, share-weighted. Only about an eighth of ETF volume touches a dark pool. The rest is either on a public exchange or handled off-exchange by a dealer. ETFs are actually less internalized than single stocks: our 16-name single-stock aggregate ran 51.2% lit, 35.3% non-ATS, 13.5% ATS. The market-wide story is that a small number of specific ETFs print off-exchange far more heavily than the average single stock, and they are not the ones you would guess.

Why VOO and SPY diverge

The first thing to kill is the idea that this is a Vanguard quirk. iShares' IVV, the third big S&P 500 vehicle, sits in the same territory as VOO on every measure we have, and the two lowest-cost buy-and-hold S&P trackers move together throughout the window. SPY sits alone at the other end, with the most lit volume of the group at 61.4% and the most ATS volume at 19.0%. The pattern does not track the issuer. It tracks what the fund is used for.

The leading explanation, and we are labelling this a hypothesis rather than a measured fact, is holder base and order flow. VOO and IVV are the default core holdings of retail and advisor portfolios. That flow arrives as marketable retail orders, and under payment-for-order-flow economics, brokers route much of it to wholesalers who fill it from inventory and report the print off-exchange. SPY, by contrast, is the market's institutional and options-hedging vehicle, and that kind of flow tends to interact on lit exchanges and in ATSs, where the economics that pull retail orders to a wholesaler do not apply.

That is a plausible chain, and the obvious place to check it does not work. Rule 606 order-routing reports are filed by brokers rather than by fund issuers, and they aggregate into three buckets: S&P 500 stocks, non-S&P 500 stocks, and listed options. An ETF tracking the index is not a constituent of it, so VOO and SPY both land in the same non-S&P 500 bucket. No public 606 report can tell them apart. The rule that resolves to the individual symbol is Rule 605, and that is where the rest of this piece goes.

One more channel deserves naming. Authorized participant activity around creations and redemptions leaves prints in the secondary market, and a dealer that fills a customer order from inventory will often offset it through the creation or redemption process. VOO and IVV have absorbed very large primary-market flows over this period, so some part of their off-exchange share reflects that machinery rather than retail order flow alone.

The T-bill fund that broke our explanation

Then there is SGOV, an iShares ultra-short Treasury ETF and one of the largest T-bill wrappers on the market. It printed 67.1% to the non-ATS channel, a hair more than VOO. We assumed this was the counter-example: an institutional cash-management fund that ends up off-exchange for reasons having nothing to do with wholesaler internalization. The execution-quality data says we were half right, and the half we got wrong is the interesting one.

Rule 605 reports resolve to the individual symbol, and the two largest wholesalers, Virtu and Citadel Securities, publish them monthly. Pulling six months of both and matching them against consolidated volume shows that SGOV is internalized more heavily than anything else we looked at. 21.1% of its entire consolidated volume was filled by those two firms alone, against 4.2% for SPY and 6.5% for VOO. Nearly a third of its off-exchange volume runs through them, and they improved on the quote for 87% of those shares, the highest rate in the group. So the fund is not avoiding the wholesalers. It is their single biggest customer in this sample.

What it is not is retail. The order sizes give it away. For VOO, IVV, SPY and QQQ, somewhere between 18% and 22% of wholesaler-executed volume arrives in orders of 2,000 shares or more. For SGOV the figure is 44.5%, and at roughly $100 a share those are orders running from $200,000 to about a million dollars. That is an advisor sweeping client cash or a treasurer parking a balance, not someone buying fifty shares in a brokerage app. The mechanism is the same as the retail case, since a dealer will happily fill an instrument that moves fractions of a cent a day and price-improve to win the flow, but the customer is different. Heavy internalization turns out to be a statement about how predictable the flow is, not about who is sending it.

Execution quality

The same data tightens the VOO and SPY comparison and complicates it at the same time. Price improvement tracks the off-exchange ranking almost perfectly. That is the wholesale-internalization fingerprint, and it lands exactly where the off-exchange shares are highest. But the volume does not line up as neatly. The two big wholesalers account for only 4.7% of IVV's off-exchange volume and 10.0% of VOO's, which means the great bulk of what those two funds do off-exchange happens somewhere other than a wholesaler's covered-order book. IVV is the sharpest version of the puzzle: nearly the highest off-exchange share in the group, the internalization fingerprint on the flow that does run through a wholesaler, and the least wholesaler-executed volume of any fund we measured.

What it means for issuers and traders

For issuers, the number is a distribution map. When two-thirds of a fund's secondary volume clears against dealers rather than on an exchange or in a dark pool, the effective liquidity network is a set of wholesaler and market-maker relationships, not the exchange order book the listing might imply. Two funds on the same index can have completely different distribution footprints, and VOO versus SPY is the proof.

For traders, the split raises an execution-quality question rather than answering one. A high off-exchange rate is not by itself evidence of a worse fill, since a wholesaler can and often does improve on the lit quote, but it changes what you should be measuring. When most of a fund's volume never posts a public quote, a read of displayed liquidity understates what is actually available, and the price-improvement statistics from the internalizing venues become the thing to interrogate rather than exchange depth.

What this shows, and what it does not

Methodology

Channel percentages come from FINRA's weekly OTC Transparency data, cross-referenced against consolidated tape volume. ATS volume is published under FINRA Rule 4552; non-ATS OTC volume is published under Rules 6110 and 6610, per Regulatory Notice 15-48. ETF figures are share-weighted across 34 funds; the single-stock comparison is a 16-name aggregate of the most active large-cap names over the same period. The window runs from December 2025 through early July 2026, with tier-complete data through June 22, 2026. Execution-quality figures come from the monthly Rule 605 filings of Virtu Americas and Citadel Securities, January through June 2026, matched against consolidated volume for the same six months. Rule 605 covered orders are held market and marketable-limit orders below a size threshold. The Virtu file carries both its Virtu Americas and Knight market-center identifiers, and we sum them. One point of vocabulary trips people up: the TRF is a reporting pipe, not a venue. Dark pools and dealer off-exchange trades both post through it, and a bare TRF total cannot tell an internalized fill from a negotiated block or a dark-pool cross. These figures are ours and should be attributed to our analysis, not to FINRA or the SEC.

Sapinover builds intelligence on U.S. equity market structure and the overnight session. This analysis is for informational purposes only and is not investment advice or a solicitation.