NAV Arbitrage: Premium and Discount Trading
A premium is a market price above reported net asset value; a discount is below it:
premium_t = price_t / NAV_t - 1
The ratio is simple, but the trade is not. For an open-end ETF, authorized participants may create or redeem shares, so a large executable deviation can invite arbitrage. For a closed-end fund, shares cannot normally be created or redeemed against NAV; a discount can persist for years and may reflect fees, illiquidity, leverage, tax overhang, or investor demand. Treating both instruments as the same strategy is a category error.
The underlying mechanism is described in ETF creation and redemption mechanics. The trading question is whether market price differs from a timely, realizable liquidation value by more than the cost and risk of closing the gap.
Use a tradable reference value
Official NAV is often struck once daily. For US equities this can be close to executable value; for international equities, bonds, bank loans, private assets, or derivatives it may be stale or model-based. Build a fair-value estimate from contemporaneous underlying quotes, FX, accrued income, futures hedges, and expected creation/redemption costs.
| Vehicle | Reported NAV reliability intraday | Typical convergence force |
|---|---|---|
| Broad domestic equity ETF | high | AP creation/redemption |
| International ETF | lower when home market closed | futures/ADR price discovery |
| Bond ETF | model-dependent | dealer basket arbitrage |
| Closed-end fund | varies | activism, tender, liquidation, sentiment |
def premium(price, nav):
return price / nav - 1.0
def executable_etf_edge(etf_bid, basket_ask, creation_fee, hedge_cost):
# Positive means rich ETF before balance-sheet and model-risk buffers.
return etf_bid - basket_ask - creation_fee - hedge_cost
def zscore(series, lookback=126):
return (series - series.rolling(lookback).mean()) / series.rolling(lookback).std()
A historical z-score may help rank opportunities, but it has no structural meaning without a catalyst and a convergence path. In particular, a discount that widened after portfolio quality deteriorated is not necessarily cheap. For a closed-end fund, review leverage, distribution policy, management fee, tax position, shareholder base, and activist ownership before betting on mean reversion.
ETF premium and discount implementation
An AP-style trade requires creation-unit access, basket infrastructure, financing, and operational capacity. A smaller trader may instead hedge an ETF against liquid futures or correlated ETFs, accepting basis risk rather than claiming a pure arbitrage. The edge must cover stock or bond spreads, FX, borrow, fees, settlement, dividend forecast errors, and slippage while the hedge is incomplete.
| Risk | Why it matters | Mitigation |
|---|---|---|
| Stale NAV | apparent discount is not real | estimate intraday fair value |
| Basket illiquidity | creation cannot be priced tightly | require larger buffer |
| AP capacity | arbitrage channel weakens in stress | scale down and diversify routes |
| Hedge mismatch | proxy hedge drifts from basket | bound beta and residual risk |
| Tax/distribution event | NAV moves mechanically | adjust the reference series |
Premiums often widen at the open and close, around geopolitical events, and when underlying markets are shut. Those are precisely the periods with high execution and model risk. Use time-of-day controls and measure opportunities on bid/ask, not midpoints. Backtests based on daily closing prices tend to record a convergence that could never have been captured after spreads.
Closed-end funds are different
For a closed-end fund, discount trading is more like value investing with a catalyst. Potential catalysts include a tender offer, conversion to open-end format, liquidation, fee reduction, distribution-policy change, or sustained portfolio outperformance. The discount can also widen because fund leverage makes NAV volatile or because holders value liquidity more than the underlying assets. A short premium trade can be especially hazardous when borrow is scarce and retail demand is persistent.
Compare funds with similar mandate, leverage, duration, and distribution policy, then decompose returns into NAV change and discount change. This avoids mistakenly attributing a bond rally to discount convergence.
Validation
Test a walk-forward signal with published timestamps for NAV and holdings. Attribute P&L to underlying hedge, premium movement, carry, and execution. Include delisted funds, tender events, borrow failures, and realistic capacity. The useful framework is classical arbitrage: a price relation needs a replicable conversion mechanism before it earns the name arbitrage.
Key takeaways
- Premium/discount is only as good as the valuation timestamp behind NAV.
- ETF deviations can be arbitraged through creation/redemption, subject to costly frictions.
- Closed-end-fund discounts need a catalyst and can remain rationally persistent.
- Trade executable bid/ask edges and separate convergence P&L from underlying exposure.