Reader question: An ETF trades all day on an exchange like a stock, while its net asset value is calculated from a portfolio of securities. Why does the ETF’s market price usually remain near NAV instead of drifting arbitrarily far away?
The key mechanism is the ETF’s creation-and-redemption process. Authorized participants (APs) can transact directly with the ETF in large blocks called creation units. When ETF shares trade rich relative to the value of the underlying basket, an AP may be able to assemble the required basket, deliver it to the fund and receive new ETF shares that can be sold in the secondary market. When ETF shares trade cheap, an AP may buy ETF shares, redeem a creation unit and receive basket assets. Competition around those opportunities tends to connect ETF share prices with underlying portfolio value.
That mechanism is not a frictionless mathematical identity. It depends on basket design, transaction costs, securities liquidity, settlement, taxes, AP balance-sheet capacity and market conditions. The algorithm therefore needs both an arbitrage signal and a feasibility test.
What this page owns — and what it does not
This page owns:
ETF portfolio + published basket + ETF market price + NAV + AP transaction costs → create/redeem decision economics → primary-market share creation or cancellation.
It does not replace exchange order matching, securities settlement, or index construction. Those are adjacent layers.
This is public ETF mechanism education, not a recommendation to buy, sell or arbitrage an ETF.
Secondary market versus primary market
Ordinary investors usually trade ETF shares with other market participants on an exchange.
The ETF itself generally does not issue one share every time an investor clicks “buy”. Instead, the fund creates or redeems large blocks through APs.
This creates two linked markets:
- Secondary market: ETF shares trade between investors;
- Primary market: APs exchange creation units with the fund.
The creation/redemption algorithm connects them.
Authorized participants are special counterparties
Only an authorized participant with the required agreement can place creation or redemption orders directly with the fund.
APs are typically large financial institutions that can move securities, cash and ETF shares through clearing and settlement infrastructure.
An ordinary investor can benefit indirectly from AP competition without having direct creation-unit access.
Creation units are large blocks
An ETF defines a creation-unit size under its operational documents. An AP that satisfies the creation terms receives a block of ETF shares rather than a handful of retail shares.
Conceptually:
Basket Assets + Cash Component + Fees → Creation Unit Shares.
For redemption:
Creation Unit Shares → Basket Assets + Cash Component − Fees.
The exact composition can be in-kind, cash or a mixture depending on the ETF and the accepted basket.
What is the basket?
The basket is the set of securities, assets and/or cash used in the create/redeem transaction.
Rule 6c-11 requires an ETF to adopt written policies and procedures governing basket construction and acceptance. The rule also permits custom baskets if the ETF has detailed parameters and review controls designed to protect the interests of the ETF and shareholders.
A basket therefore is not simply “whatever the AP wants to deliver.”
Standard basket versus custom basket
A standard basket can closely resemble the fund portfolio or a representative slice of it.
A custom basket can deviate from that ordinary construction, for example by:
- using a non-representative selection of portfolio holdings;
- changing quantities for liquidity or tax reasons;
- substituting cash for a particular security;
- handling restricted or hard-to-transfer instruments.
The SEC requires governance around these deviations because an unfair basket can transfer costs or adverse securities to remaining shareholders.
NAV and market price
For ETF net assets A, liabilities L and shares outstanding N:
NAV per share = (A − L) / N.
The ETF’s exchange price P can differ from NAV.
Define the premium/discount:
PremiumDiscount = P / NAV − 1.
If NAV = $100 and ETF price = $100.50:
premium = 0.5%.
If price = $99.50:
discount = −0.5%.
Premium arbitrage logic
Suppose the ETF trades at a premium.
A stylised AP trade is:
- buy or source the basket assets;
- deliver the basket to the ETF;
- receive creation-unit shares;
- sell ETF shares in the secondary market.
The gross theoretical edge per creation unit is approximately:
ETF Sale Proceeds − Basket Acquisition Cost.
But the actual decision must subtract:
- bid-ask spreads;
- brokerage and exchange costs;
- creation transaction fees;
- financing costs;
- hedging costs;
- settlement and borrow costs;
- tax effects;
- execution slippage.
Discount arbitrage logic
If ETF shares trade below basket value, a stylised AP can:
- buy ETF shares;
- assemble enough shares for a creation unit;
- redeem the creation unit;
- sell or retain the delivered basket securities.
Gross edge:
Basket Liquidation Value − ETF Acquisition Cost.
Again, only the net edge after costs matters.
The true arbitrage inequality
A simple decision rule is:
|Market Price − Replication Value| > Total Execution and Funding Costs.
The threshold is dynamic. A 20bp premium can be actionable in a liquid equity ETF but uneconomic in a bond ETF holding wide-spread securities.
Why creation can move the ETF price downward
If a premium creates profitable new-share supply:
- APs create ETF shares;
- they sell those shares into the secondary market;
- share supply rises;
- selling pressure tends to reduce the ETF price relative to NAV.
At the same time, AP purchases of underlying basket assets can support those underlying prices.
The two-market trade tends to compress the price gap.
Why redemption can move the ETF price upward
If a discount creates profitable redemptions:
- APs buy ETF shares;
- secondary-market demand rises;
- shares are redeemed and removed from circulation;
- basket assets are received by the AP.
This tends to reduce the discount, subject to costs and market conditions.
A 50bp premium does not guarantee a 50bp arbitrage profit
Suppose:
- ETF NAV = 100;
- ETF price = 100.50;
- basket acquisition spread/slippage = 0.20;
- creation fee and operational cost = 0.08;
- hedging/financing cost = 0.12;
- ETF selling slippage = 0.12.
Total costs = 0.52.
The apparent 0.50 premium is not profitable after costs.
This is why observed premiums and discounts can persist inside a no-arbitrage band.
In-kind creation and redemption
Many ETFs exchange securities rather than cash for creation units.
In-kind transactions can:
- reduce the fund’s need to trade securities itself;
- shift some transaction costs to the AP conducting the transaction;
- reduce taxable realisation in some U.S. ETF structures;
- help the fund manage portfolio composition.
But they also require operational capability to move the component securities.
Cash creations and redemptions
Some ETFs use cash when in-kind delivery is impractical.
The fund may then have to trade the underlying securities itself.
That can create transaction costs and market impact, which is why funds can impose transaction fees or cash-adjustment amounts designed to reduce dilution of existing shareholders.
The balancing amount
A basket may not equal the creation unit’s NAV perfectly because of accrued income, liabilities, rounding or excluded positions.
A balancing cash component can reconcile the difference:
Cash Component = Creation Unit NAV − Market Value of In-Kind Basket.
Depending on sign, cash can flow from AP to fund or fund to AP.
Custom baskets require governance
Rule 6c-11 allows flexibility but requires detailed written parameters for custom baskets and specified review responsibility.
The governance question is:
Does this deviation serve the ETF and shareholders rather than create an unfair transfer?
Examples of potential weak links include giving one AP unusually favorable securities, repeatedly pushing less-liquid assets onto the fund, or using basket deviations without documented rationale.
Representative baskets and sampling ETFs
An ETF does not always hold every constituent of an index.
A sampled fixed-income ETF can hold a representative subset.
The creation basket can therefore be built to preserve risk characteristics such as:
- duration;
- sector exposure;
- credit rating;
- country exposure;
- coupon/maturity buckets.
This connects to bond-index algorithms but solves a different problem: basket transfer rather than benchmark definition.
AP and market maker are different roles
A market maker posts bids and offers on an exchange.
An authorized participant can create and redeem ETF shares directly with the fund.
One firm can perform both roles, but the roles are conceptually distinct.
A system that assumes every market maker can submit creation orders can route transactions incorrectly.
AP concentration risk
SEC ETF disclosures routinely warn that a fund can have a limited number of APs and that no AP is obligated to create or redeem.
If APs withdraw during stress:
- the arbitrage channel can weaken;
- bid-ask spreads can widen;
- premiums/discounts can increase;
- trading can become less efficient.
The mechanism therefore depends on economic incentives and operational capacity, not just legal eligibility.
Bond ETFs during market stress
In stressed fixed-income markets, underlying cash-bond prices can be stale or difficult to observe while ETF shares continue to trade frequently.
An ETF discount to published NAV can therefore reflect:
- true dislocation;
- stale underlying marks;
- liquidity compensation;
- anticipated transaction costs;
- market makers’ risk limits.
A large displayed discount is not automatically free arbitrage.
Settlement risk sits downstream
A profitable theoretical creation can still fail operationally if basket assets are not delivered, creation orders miss cutoffs, or settlement instructions break.
This connects to securities settlement algorithms.
Arbitrage logic without settlement feasibility is incomplete.
Inputs and outputs
An ETF create/redeem engine can require:
- ETF identifier;
- creation-unit size;
- published basket and cash component;
- custom-basket policy and approval status;
- ETF market bid/ask;
- NAV or estimated basket value;
- component security bids/asks;
- creation/redemption fees;
- financing and borrow costs;
- order cutoffs;
- settlement dates and calendars;
- AP authorization state.
Outputs can include:
- premium/discount;
- replication cost;
- net create/redeem economics;
- creation/redemption order size;
- basket validation result;
- cash balancing amount;
- settlement instructions;
- diagnostic rejection reasons.
Evidence polarity: what supports confidence?
Evidence for a reliable mechanism includes daily basket disclosure where applicable, NAV and market prices that reconcile, AP orders tied to creation-unit sizes, custom baskets approved under documented controls, transaction fees matching operational documents, and create/redeem activity increasing when economically meaningful premiums/discounts appear.
Evidence against confidence includes creations submitted by non-AP entities, custom baskets with no policy rationale, premiums measured from stale NAVs, transaction costs omitted from arbitrage tests, or a supposed arbitrage that cannot settle the required basket.
Counterexample: premium without profitable creation
A 30bp premium can coexist with 50bp of basket spreads, funding and fees.
The ETF is technically above NAV while creation remains uneconomic.
Counterexample: discount caused by stale NAV
If underlying bond marks are stale and ETF shares incorporate fresher market information, the ETF can appear discounted even when the ETF price is closer to the executable value of the portfolio.
Counterexample: AP exit can widen the arbitrage band
If balance-sheet constraints reduce the number of active APs, the remaining firms can demand a larger premium or discount before creating or redeeming.
The arbitrage relationship becomes weaker without being eliminated.
Counterexample: basket arithmetic can be right while shareholder economics are wrong
A custom basket can sum to the correct NAV but still transfer unusually illiquid or costly assets to the fund.
That is why governance tests go beyond arithmetic equality.
Weak links in implementation
NAV timing mismatch. ETF price is compared with a NAV from a different valuation time.
basket staleness. Yesterday’s basket is used for today’s creation order.
AP authorization drift. An expired or inactive AP agreement remains enabled.
transaction-cost omission. Gross premium/discount is mistaken for profit.
custom-basket control failure. Deviations bypass the required review process.
cash-component sign error. AP and fund cash directions are reversed.
settlement-calendar error. basket assets and ETF shares settle on incompatible dates.
market-maker/AP conflation. an exchange liquidity provider is incorrectly given primary-market access.
Diagnostics: how to test the engine
- NAV identity test: independently calculate portfolio net assets divided by shares outstanding.
- premium/discount test: compare ETF market price with synchronized NAV/basket value.
- creation-unit test: reject non-multiple order sizes where required.
- cash-balancing test: basket value plus cash equals creation-unit NAV within defined tolerances.
- cost-band test: arbitrage signal must clear realistic trading and funding costs.
- custom-basket test: non-standard baskets require policy-compliant review.
- AP-state test: only active authorized participants can submit primary-market orders.
- stress test: widen underlying security spreads and verify the arbitrage threshold widens.
- settlement test: basket and ETF-share instructions reconcile through the required clearing path.
- AP-outage test: remove one or more APs and measure the impact on executable arbitrage capacity.
What would falsify confidence?
Confidence should be withdrawn if creation/redemption orders cannot be reconciled to baskets and shares; if custom baskets lack documented controls; if premiums/discounts use unsynchronized values; if AP authorization cannot be verified; or if apparent arbitrage profits disappear after realistic transaction and settlement costs are included.
Alternatives and limits
Open-end mutual funds generally transact with investors directly at NAV rather than through exchange-market arbitrage. Closed-end funds do not have the same routine creation/redemption mechanism and can therefore trade at persistent premiums or discounts.
The ETF mechanism reduces price/NAV dislocations but does not promise zero premium, zero discount or unlimited liquidity.
Verification and update triggers
Preserve Rule 6c-11 version, AP agreements, creation-unit size, basket files, custom-basket approvals, NAV timestamps, market-price timestamps, transaction-fee schedules and settlement rules. Revalidate after SEC amendments, fund prospectus changes, AP additions/removals, basket-process changes or persistent premium/discount episodes.
Primary and high-quality references
- U.S. Securities and Exchange Commission, Exchange-Traded Funds: A Small Entity Compliance Guide.
- Electronic Code of Federal Regulations, 17 CFR §270.6c-11 — Exchange-traded funds.
- U.S. Securities and Exchange Commission, Rule 6c-11 adopting release and ETF rulemaking materials on baskets, disclosure and arbitrage.
- International Organization of Securities Commissions, ETF good-practice material on primary-market liquidity, APs and arbitrage mechanisms.
Educational boundary: This article explains public ETF creation/redemption mechanics. It does not recommend an ETF, arbitrage strategy or trading action and does not provide personalized financial advice.