Primary vs Secondary Market Structure

The two markets, precisely

The secondary market is everything you can see: exchange order books, market-maker quotes, off-exchange and over-the-counter trades between institutions, and, in Europe especially, a large volume of bilateral and RFQ dealing that never appears on a lit exchange tape (Chapter 92). No new shares are created here; existing ones change hands.

The primary market is the wholesale layer of Chapter 31: APs delivering baskets for creation units, or returning shares for baskets, at NAV, once per day per fund per AP, with cash variants where in-kind is impractical.

The interface between them is arbitrage. When secondary demand outruns supply and the price ticks above fair value, the primary window converts that demand into new shares backed by new assets. When it recedes, redemption removes them.

The ratio, and why it matters

Industry data consistently shows secondary-market trading dominating primary flows by a wide margin: across US ETFs, about five dollars trade in the secondary market for every dollar of creations and redemptions (ICI, 2024 data), and the ratio runs from roughly 5:1 for high-yield bond ETFs to nearly 20:1 for emerging-market equity ETFs. Three consequences follow, and each resolves a common misconception.

1. ETFs trade far more than they transact. A fund can turn over billions in a day while the manager buys and sells almost nothing, because most of that volume was investors trading with each other. The frequently-heard claim that "ETFs now dominate trading, so they must dominate markets" conflates the two layers. The number that matters for market impact is primary-market flow, not headline volume.

2. The wrapper is a shock absorber, not an amplifier: usually. In a selling wave, sellers first meet buyers in the secondary market. Only the net imbalance reaches the primary window and becomes actual sales of underlying securities. The ETF acts as a netting layer between opposing flows, which is precisely the opposite of the "ETFs force selling" story. The honest caveat, developed in Chapter 87: in a one-directional panic, netting has less to net, and the ratio compresses toward the primary market.

3. Liquidity is layered. The ETF's real depth is the sum of three sources: displayed quotes, undisplayed market-maker capacity, and the underlying basket's liquidity accessed through creation/redemption. Chapter 71 makes this its central argument; here it is a structural fact about where the layers sit.

Who trades where

The two markets have different populations, and this shapes everything:

• Retail lives entirely in the secondary market, in small sizes, absorbed by standing quotes (Chapter 58).
• Institutions span both: smaller orders on exchange, larger ones through RFQ and risk pricing (Chapter 72), and the largest via creation/redemption directly, where the cost of entry is the basket's trading cost rather than the ETF's spread.
• Market makers operate in both continuously, quoting in the secondary market and using the primary window to manage inventory (Chapter 32).
• The issuer touches only the primary market, and only in response to AP orders. This is worth stating plainly: the fund manager does not trade in response to your purchase. They respond to net primary-market flow, which is a different and much smaller thing.

Where the structure differs by market

The US has a consolidated tape: roughly half of all volume now trades off-exchange — RFQ, dealer internalisation, dark venues — but every one of those trades must still print to the tape, which is why US ETF volume statistics look large and are broadly complete.

Europe is the structural outlier (Chapter 92): the same fund lists on multiple exchanges, a large share of trading happens OTC or via RFQ platforms, and historically much of it went unreported until MiFID II improved transparency. Consequence: European ETF volumes are systematically understated on any single venue, which makes "low volume means illiquid" an even worse heuristic in Europe than elsewhere.

Fixed income deserves its own note (Chapter 82): the secondary market for a bond ETF is frequently more liquid than the market for its constituent bonds, which produces the price-discovery phenomenon that made March 2020 look scandalous to people reading NAV instead of price (Chapters 33, 84).

The stress question, stated structurally

The primary-secondary ratio is the mechanism underneath the entire systemic-risk debate, so it deserves a clear statement here before Chapter 87 argues it out:

The reassuring reading. Netting is real and large. In every stress episode so far, the secondary market absorbed the majority of trading, primary flows rose but stayed manageable, and the machinery kept functioning at wider prices (Chapters 32, 84).

The concerning reading. Netting depends on two-sided flow. In a genuine one-way panic, the ratio compresses: more of the selling reaches the primary window, more underlying securities must actually be sold, and if those underlyings are illiquid (high yield, EM debt, small caps), the ETF becomes a transmission channel rather than a buffer.

What the evidence says so far. The compression has happened, been measured, and stayed within the system's capacity, including in March 2020's bond market seizure. That is a track record rather than a guarantee, and the honest position, which Chapter 87 develops, is that the structure has been tested repeatedly and has bent without breaking, while the untested scenario always remains one degree worse than the last one.

A worked example

A €40bn European equity ETF on an ordinary Tuesday:

• Secondary market: €310m of shares change hands across four exchanges plus OTC and RFQ channels. Thousands of investors, dozens of institutions, several market makers.
• Net imbalance: buyers exceed sellers by €28m.
• Primary market: one AP creates €25m of new shares in the afternoon, having accumulated the basket during the session; another redeems €3m for an exiting client.
• Underlying market impact: roughly €25m of stock purchases across 1,400 companies, in a market that trades hundreds of billions daily. Undetectable.

Headline volume: €310m. Actual footprint on the underlying market: €25m, about 8%. The other 92% was investors trading with each other, which is what the wrapper is for.