ETF Liquidity Provision: Authorized Participants and Market Makers

Authorized participants and lead market makers create, redeem and quote ETF shares. Here is how their incentives shape spreads, premiums to NAV and tracking error.

2026-11-06 · 10 min read

Key takeaways

  • ETF liquidity comes from two markets: the exchange tape and the off-exchange primary market
  • Contracted AP counts overstate activity, because a handful of firms do most creations and redemptions
  • Lead market maker spread ceilings and depth minimums are tiered by volume and price bucket
  • Spreads are a one-off execution cost, while tracking error is a separate, ongoing deviation from the index

Two markets behind one ticker

An ETF trades like a share, but its price is anchored by two mechanisms running at the same time. On the secondary market, buyers and sellers meet on an exchange. On the primary market, a small set of institutions create and redeem shares directly with the fund. The second market is invisible during most of the trading day, yet it is what keeps the first one honest.

Scale first, because it frames everything else. US-listed ETFs now hold assets measured in the trillions of dollars across thousands of funds, with bond ETFs a substantial and growing share, according to the ICI's monthly ETF statistical release. Almost all of that value sits in funds whose market price is meant to track a published basket — and the tracking is enforced by firms that earn a spread for doing it.

The primary market opens only to authorized participants

The framework is SEC Rule 6c-11, adopted in 2019 and effective from 23 December 2019, which lets qualifying ETFs operate without fund-by-fund exemptive orders and defines an authorized participant (AP) as a member or participant of a registered clearing agency with a written agreement with the fund or a service provider, allowing it to place orders for the purchase and redemption of creation units (SEC final rule 33-10695).

Only APs transact with the fund, and only in creation units — blocks far larger than a typical retail order. On a creation, the AP delivers a basket of securities plus a cash balancing amount and receives newly issued shares it can sell on the exchange. On a redemption, it collects shares in the market, delivers them to the fund and receives the basket back. Baskets may be in kind, in cash or a mix of both, and custom baskets are allowed under conditions set out in the rule (SEC final rule 33-10695).

Two details shape how you should read the data. First, US APs are self-clearing broker-dealers and members of the clearing infrastructure, so institutional investors that do not self-clear cannot reach the primary market directly and must go through an AP as agent. Reported primary-market volume therefore aggregates a broker's own book with its clients' flow. Second, APs are not paid by the fund: they "do not receive compensation from an ETF or its sponsor and have no legal obligation to create or redeem the ETF's shares", as the ICI's ETF FAQ puts it. They earn from dealing.

Contracts are not the same as activity

The number of AP agreements a fund has measures legal capacity, not who is actually arbitraging. In 2024, 65 firms held registered AP agreements with ETF sponsors and 43 of them created or redeemed shares, per the ICI's ETF FAQ. Fund-level counts are higher than firm-level counts: a 2015 ICI survey of ETF sponsors, covering sponsors that managed roughly 90% of industry assets as of November 2014, found an average of 34 AP agreements per ETF and a median of 36.

Registered connections and active ones are different populations. Using SEC Form N-CEN filings, Gorbatikov and Sikorskaya document that as of 2019 the median US equity ETF was connected to 22 of 50 APs, and only four of those connections were active (Section 4.2.1). Concentration is sharper still in the same filing data: 15 of 50 APs accounted for 98% of creation and redemption volume and the top three generated half of it, while for over 65% of funds a single AP executed more than half of all creations and redemptions — and for 11% of funds, all of them.

Non-US markets show the same shape at a smaller scale. BlackRock reported that its Australian-domiciled iShares ETFs had on average 10 contracted and six active APs in the year to July 2025, with the busiest single AP handling 20% of gross creation and redemption activity (iShares Australia). Disclosure routes differ by jurisdiction: in the US, Form N-CEN reporting plus the prospectus; in the EU, UCITS documentation and the PRIIPs KID; in Australia, ASX and issuer disclosures.

Lead market makers quote the tape

Retail and most institutional orders never meet an AP. They trade with market makers, whose job is to quote continuously so an investor can transact without waiting for a counterparty with the opposite view.

On NYSE Arca, primarily listed exchange-traded products are supported by a Lead Market Maker (LMM) programme that pairs quoting obligations with fee economics. LMMs must maintain continuous two-sided quotes within the parameters of Arca Rule 7.23 and meet performance metrics covering quotes at the inside, minimum displayed size, maximum quoted spread and participation in the opening and closing auctions (NYSE Arca LMM fact sheet). The inside-quote ceilings are tiered by the product's monthly average daily share volume and by price (NYSE Arca LMM Requirements):

ETP daily share volume Inside spread ceiling (price $1–4.99 / $5–14.99 / $15–49.99 / $50+) Minimum depth
0–4,999 15.00% / 6.00% / 5.00% / 4.00% 700 / 400 / 300 / 200 shares
5,000–24,999 7.00% / 3.00% / 2.00% / 1.50% 700 / 400 / 300 / 200 shares
25,000–74,999 5.00% / 1.50% / 1.00% / 0.70% 700 / 400 / 300 / 200 shares
75,000–199,999 3.00% / 1.00% / 0.50% / 0.30% 700 / 400 / 300 / 200 shares
200,000–499,999 2.00% / 0.60% / 0.30% / 0.20% 700 / 400 / 300 / 200 shares
500,000 or more 1.00% / 0.30% / 0.20% / 0.10% 2,000 / 1,000 / 500 / 300 shares

Read the table as a map of what the exchange expects liquidity to cost. A thin, low-priced ETP may show a 15.00% inside spread in the lowest price bucket; an ETP trading 500,000 shares or more a day must keep the same quote inside 0.10% in the top price bucket. Depth minimums run from 200 to 700 shares in the first five volume tiers and from 300 to 2,000 shares in the most active tier.

In return, LMMs receive fee credits: rebates of $0.0045, $0.0040 or $0.0033 per share for posting liquidity, depending on the product's consolidated average daily volume, against a $0.0029 per share fee for removing LMM-allocated liquidity (LMM fact sheet). Two implications follow. Obligations are floors and can be relaxed in fast, one-sided markets, so a displayed quote is not a promise. And part of the quoting you see is paid for, through fee credits or issuer incentive programmes, which matters before reading an unusually narrow quote as purely organic.

Three roles sit behind one ticker:

  • Authorized participant — primary market only, at NAV, in creation units; earns from arbitrage and from agency fees.
  • Lead or designated market maker — secondary market quoting, depth and auction participation; paid by spread, size and exchange fee credits.
  • ETF sponsor — sets the basket, publishes NAV, discloses creation and redemption data; paid by the management fee.

Where spreads and tracking error come from

Market makers price immediacy. The quoted spread compensates for volatility in the basket, hedging cost, adverse selection and tick size, and the effective spread you pay can differ from the quoted one — a tight display in a thin book is not the same as a good fill. Bid-ask spread and market depth covers that distinction in detail.

Tracking error is a different quantity. Tracking difference is the annual return gap between fund and index; tracking error is the volatility of that gap. An MSCI analysis of US bond ETFs found median annual tracking error of 1 to 6 basis points in government (non-mortgage) sectors, 9 to 14 basis points in MBS, broad investment-grade and short/intermediate credit, 39 basis points in long investment-grade credit and 67 basis points in high yield, where individual funds ranged from 46 to 88 basis points. For equity the numbers can be smaller: an analysis of five large S&P 500 index funds found average annual tracking error of about 2 basis points measured before fund expenses (Morningstar). Both are samples of specific funds over specific windows, not expectations for any fund.

AP and market maker behaviour shows up most directly in the premium or discount to NAV. Persistent premiums point to creation friction; persistent discounts point to redemption friction or hedging costs in the underlying market. Friction has familiar sources:

  • Creation or redemption suspended, by capacity limits, regulatory constraints or fund policy.
  • Cash creations where the underlying market settles differently from the ETF, pushing transaction costs into the fund.
  • Illiquid underlying assets, as in corporate bond ETFs, where the priced basket is an estimate rather than a live market.
  • Closed foreign markets: an international equity ETF can trade while its basket is shut, so the published NAV is stale and part of the apparent premium or discount is a valuation artefact.

Network breadth feeds directly into pricing. The same paper finds that a one standard deviation increase in the number of registered APs is associated with 15% lower average daily mispricing, with the effect concentrated on days of market stress. More independent arbitrageurs, faster gap closure.

A practical monitoring checklist

  • Spread in basis points at your order size, time-weighted rather than the best displayed quote.
  • Median premium or discount to NAV over one to three months, and how often it breaches your tolerance.
  • Number of APs that actually transacted, where fund or issuer disclosure provides it.
  • Creation unit size against turnover. The prospectus states the creation unit, and qualifying ETFs disclose the basket they will accept, so you can divide the creation unit's share count by average daily volume in shares to see how many primary-market events a full day of trading implies. Creation unit sizes differ widely by fund and asset class, so there is no single "normal" figure — the ratio is the point. A creation unit that is large relative to turnover means each create or redeem moves the arbitrage mechanism more, and small flows can be absorbed without any primary-market activity at all.
  • Basket composition (in kind, cash or custom) and any disclosed suspension of creations.
  • Tracking difference and tracking error from the factsheet, read alongside the TER rather than instead of it. See tracking difference vs tracking error and creation, redemption and NAV divergence.
  • Hedging and valuation conventions: share class currency, hedging overlay, foreign market hours.

Risks and costs

Supplying liquidity is a business with capacity limits and incentives, and both land on investors.

Concentration is quantifiable, not theoretical. Where 15 APs produce 98% of primary-market volume and the top three produce half of it, while most funds have only about four active AP connections (Gorbatikov and Sikorskaya), a single firm stepping back affects many funds at once. Quoting obligations do not prevent market makers from widening or stepping away when inventory risk and hedging costs spike, and fee credits and incentive programmes create an economic link between exchanges, issuers and the firms quoting a product.

Trading costs are not confined to the spread. Premiums and discounts to NAV, market impact and the gap between the fund's price and the value of its underlying assets at execution all land on the investor. A narrow spread on a fund whose creation mechanism is strained can still mean transacting away from fair value.

Tracking error carries costs beyond the TER: transaction costs inside the fund, cash drag, sampling when full replication is impractical, dividend and withholding-tax timing, securities lending arrangements and index rebalancing. Funds that lend securities may offset costs with revenue but take on counterparty and collateral exposure. Jurisdiction matters throughout. In the US, Rule 6c-11 governs primary-market mechanics, and an in-kind redemption by a regulated investment company generally does not trigger fund-level gain recognition under Section 852(b)(6) — a treatment the IRS continues to examine and has recently addressed in guidance on ETF creation and redemption structures (Rev. Rul. 2026-20). Outside the US, this logic does not travel: in the EU, UCITS rules shape concentration and counterparty limits while the tax treatment of creations, redemptions and distributions is set by each member state, and in Australia the tax outcome of in-kind activity follows domestic rules rather than the US model. Investors should confirm their own position with local tax guidance. Nothing described above is a recommendation, and the instruments we follow sit on our research watchlist only.

Bottom line

ETF liquidity is manufactured in two places. APs arbitrage the primary market against the exchange price, which is why a fund can trade millions of shares a day while only a handful of firms ever touch it directly. Market makers quote the secondary market, and their obligations — spread ceilings, depth minimums, auction participation — shape what you pay. For execution quality, measure the spread in basis points and the premium or discount to NAV. For holding-period performance, measure tracking difference and tracking error, and remember that both are driven by the fund's costs and replication choices as much as by the firms supplying liquidity.

This article is for information and education only. It is not investment advice and does not take into account your personal situation. Read the full disclaimer.

Follow the shortlist

Comparing ETFs on cost, liquidity and structure? Our global ETF and stock research watchlist is tracked daily. Premium members get the full list, the reasoning behind every entry and updates when entries change.

See premium access

Information and education only, not investment advice.

Related