Guide

ETF premium and discount to NAV explained

Harbor Capital's tactical desk added $18M to an emerging-market local-currency bond ETF on a Monday when the fund's market price closed 2.4% above its published net asset value (NAV). Underlying bonds had fallen 1.1% on the day, but the ETF traded like a safe haven — retail buy orders stacked on the offer while authorized participants paused creation because settlement friction in three frontier markets made the arbitrage unprofitable after fees. By Friday the premium collapsed to 0.1%, handing new buyers an immediate 2.3% haircut versus fair value. Operations refactored the sleeve: every ETF order now checks indicative NAV (iNAV) in real time, blocks purchases above a +0.75% premium band, and routes large trades through limit orders anchored to mid-iNAV instead of market-on-close prints.

An ETF premium means the exchange-traded share price is above NAV; a discount means it trades below. For liquid U.S. equity index funds, premia and discounts are usually tiny — often a few basis points — because authorized participants (APs) arbitrage gaps via the creation and redemption mechanism. In less liquid, international, fixed-income, or stressed markets, gaps can persist for days and materially change realized returns versus the index. This guide covers NAV and iNAV calculation, the AP arbitrage loop, when discounts become structural, leveraged and commodity ETF quirks, the Harbor Capital sleeve refactor, a technique decision table against mutual funds and closed-end funds, pitfalls, and a production checklist alongside tracking error monitoring.

NAV, iNAV and the premium/discount formula

Net asset value is the per-share value of the ETF's holdings minus liabilities, divided by shares outstanding. For a U.S. equity fund, NAV typically uses closing prices of every stock in the basket, plus cash, accruals, and fee accrual, published once after the 4:00 p.m. ET close.

During trading hours, exchanges and data vendors publish indicative NAV (iNAV) or intraday indicative value (IIV) — a real-time estimate updated every 15 seconds (or faster) using live quotes on the underlying securities. iNAV is what traders use for arbitrage; end-of-day NAV is what accounting and performance reporting use.

The premium or discount percentage is:

Premium/discount (%) = (Market price − NAV) / NAV × 100

A fund at $50.50 per share with NAV $50.00 trades at a +1.0% premium. The same price with NAV $51.00 is a −0.98% discount. Sign matters: buying at a premium and selling after the gap closes loses money even if the index is flat.

Creation and redemption: the arbitrage engine

ETFs are open-ended wrappers with a unique secondary market. APs — large broker-dealers and market makers — can create new shares by delivering a prescribed basket of securities (or cash) to the fund sponsor and receiving ETF shares in 50,000-share blocks (typical creation unit size). They can redeem by returning ETF shares and receiving the basket back.

When the ETF trades above NAV, APs buy the underlying basket cheaply, create new ETF shares, and sell them on the exchange — earning the premium minus transaction costs. When the ETF trades below NAV, they buy discounted ETF shares, redeem for the basket, and sell the holdings. This loop normally keeps ETF market prices within a few basis points of iNAV for liquid products like SPY or VTI.

When premia and discounts get large

Illiquid or hard-to-source underlying

If the underlying bonds or frontier-market equities cannot be bought or sold quickly at quoted prices, APs cannot assemble creation baskets at NAV. During the 2020 liquidity shock, several corporate and municipal bond ETFs traded at double-digit discounts even though NAV itself was debated — the bonds had not traded, so stale NAV met panicked ETF selling. Discounts signaled liquidity stress, not necessarily a free lunch.

Market halts and timing mismatches

International ETFs often have underlying markets closed while U.S. shares trade. News after Asian close but before Europe opens can move the ETF on U.S. sentiment while iNAV still reflects yesterday's foreign closes — producing apparent premia/discounts that mean-revert when local markets reopen.

Commodity and futures-based ETFs

Oil and volatility ETPs can diverge when front-month futures roll, storage economics shift, or issuers suspend creation. Premiums here interact with roll yield and prospectus constraints — not just equity-style arbitrage.

Leveraged and inverse daily-reset products

2x/3x leveraged ETFs reset exposure daily; long-horizon return paths differ from multiples of the index. They can also show wider premia when swap financing costs spike or when issuers cap assets. Treat prem/disc as an execution cost layer on top of volatility decay.

Closed-end funds vs ETFs

Closed-end funds (CEFs) issue a fixed share count with no daily redemption — they can trade at persistent 10–15% discounts to NAV for years because no AP loop exists. Do not apply ETF arbitrage intuition to CEFs; the discount is a separate asset-class phenomenon driven by distribution yield, leverage, and sentiment.

Measuring and monitoring premia in production

Portfolio systems should treat premium/discount as a first-class execution metric, not an afterthought:

Real-time iNAV feed — subscribe to exchange or fund-company IIV; compare to NBBO mid for the ETF.
Historical distribution — plot daily (close − NAV)/NAV for each ticker; flag funds whose 95th percentile exceeds 50 bps.
Creation unit economics — AP profit ≈ premium × creation size minus spread, borrow, and settlement costs; when AP spread widens, expect larger gaps.
Separate tracking error from prem/disc — index replication slippage lives in tracking error; buying at a premium is a one-time execution tax on top.
Corporate actions — dividend ex-dates and fee accruals shift NAV; verify your vendor adjusts intraday estimates.

For large rebalances, use limit orders pegged to iNAV plus a tolerance band rather than market orders during the first and last 30 minutes when spreads widen and premia spike.

Harbor Capital sleeve refactor

After the emerging-market bond premium incident, Harbor Capital implemented four rules across all ETF sleeves:

Premium ceiling on buys — no new purchases if (last − iNAV)/iNAV exceeds +0.75%; queue the order until the gap closes or use a peer ETF.
Discount floor on sells — avoid panic selling below −1.0% unless risk limits force liquidation; use creation-redemption aware brokers when size permits.
Stressed-market protocol — when median bond ETF discount in the sleeve exceeds 2%, shift to mutual fund share classes with end-of-day NAV execution for incremental flows (accepting timing difference).
Attribution line item — monthly reports show index return, tracking error, and prem/disc execution drag separately so PMs cannot hide execution losses inside beta.

Over the next two rebalances, measured prem/disc slippage on the fixed-income ETF book fell from 32 bps to 9 bps annualized — not because markets calmed, but because traders stopped paying voluntary premiums during volatility spikes.

Technique decision table

Goal	Prefer	Avoid when
Tightest prem/disc, intraday liquidity	Large-cap U.S. equity ETF (SPY, IVV, VOO)	Need factor tilt unavailable in mega-cap only
International equity with AP arb	Developed-market ETF with full creation basket	Frontier markets with cash-create only and wide gaps
Fixed income exposure, minimize gap risk	Mutual fund share class for lump-sum flows	Need intraday exit and accept ETF discount risk
Commodity beta via ETF	Physically backed gold ETF (typically tight)	Futures-based ETP in steep contango without roll analysis
Persistent discount as yield play	Closed-end fund with covered-call overlay (know CEF mechanics)	Expecting AP-style mean reversion — CEFs lack it
Large institutional ticket, tight execution	Custom creation with AP desk; limit to iNAV	Retail-size market order at open on illiquid ETF
Tax-loss harvest same exposure	Switch to peer ETF with similar index, check both prem/disc	Harvesting into a fund trading at large premium

Common pitfalls

Assuming all ETFs trade at NAV — only the AP mechanism makes that approximately true; illiquid underlyings break the link.
Using yesterday's NAV during foreign-market gaps — iNAV may be stale; check whether underlying markets are open.
Market orders on bond ETFs in stress — you may sell at a 5% discount that partially reverses next week.
Chasing a premium as momentum — premium often reflects flow imbalance, not alpha; it mean-reverts when APs return.
Confusing NAV uncertainty with discount — if bonds have not traded, published NAV itself may be wrong; the discount is not always mispricing.
Ignoring creation unit minimums — retail cannot arb a 1% gap; only APs can, and they may choose not to.
CEF vs ETF mental model — persistent CEF discounts are normal; do not wait for arbitrage that does not exist.
Double-counting with tracking error — prem/disc is execution timing; replication error is structural — report both.

Production checklist

Subscribe to iNAV/IIV for every ETF in the approved list.
Log (market price − iNAV) at order fill time; attribute slippage monthly.
Set buy premium and sell discount bands per asset class (equity vs bond vs EM).
Plot 252-day prem/disc distribution; flag outliers above 95th percentile.
Document creation unit size and cash-create vs in-kind rules per fund.
During stress, compare ETF discount to comparable mutual fund NAV execution.
Verify dividend ex-dates do not distort intraday prem/disc charts.
For international ETFs, note local market holiday calendar vs U.S. session.
Stress-test sleeves using March 2020 bond ETF discount episode.
Train traders: limit orders pegged to iNAV, not market-on-close in illiquid names.

Key takeaways

ETF premium/discount is (market price − NAV)/NAV — a direct execution cost when buying high or selling low versus fair value.
Authorized participants normally arbitrage gaps via creation and redemption; illiquid underlyings and market stress can suspend that loop.
iNAV is the live trading benchmark; end-of-day NAV is for accounting — use the right one for each decision.
Harbor Capital cut prem/disc slippage from 32 bps to 9 bps by blocking buys above +0.75% premium and attributing execution drag separately.
Closed-end funds are not ETFs — persistent discounts there reflect structure, not a free arbitrage opportunity.