Skip to main content
Status of this page: P0 draft. Mechanics are standard; HK-specific parameters are flagged under “Points to verify”. For the operational side see ETF creation and redemption and HK market maker rules.

1. Definitions

  • Premium/discount (bps) = (P / NAV - 1) x 10,000, where P is the ETF price and NAV is the reference value at the same timestamp.
  • iNAV (intraday indicative NAV) is the basket value computed from live or last-known constituent prices, converted to ETF currency. It is an estimate, not a tradable price.
  • Example: P = 20.04, NAV = 20.00 gives a premium of 20.0 bps.

2. Why a premium or discount can persist

The arbitrage band is set by the cost of the creation/redemption (C/R) loop, not by zero:
  • Upper bound: premium cannot exceed the cost of buying the basket, creating, and selling the ETF (basket spread + C/R fee + financing/hedge cost + risk premium).
  • Lower bound: symmetric for redemption, usually wider if constituents are hard to short or borrow.
  • Inside the band, the premium is a noisy state variable, not an opportunity.
Band width drivers: constituent liquidity and spread, borrow availability, creation unit size, settlement cycle mismatch, FX hedge cost, and the cutoff time for C/R requests.

3. Stale pricing and non-synchronous markets

When the ETF trades while constituents are closed or illiquid (cross-time-zone ETFs, A-share and bond ETFs), measured premium reflects information arriving at the ETF first, not mispricing. Petajisto (2017) documents that premiums for international and fixed-income ETFs are larger and more persistent than for domestic equity ETFs, and that part of the gap is stale NAV. Practical consequences:
  • Measure premium against a fair-value NAV built from futures, ADRs, or factor models, not the last official close.
  • Treat the iNAV update interval and its FX source as model inputs.
  • Separate the informational premium (ETF leads) from the liquidity premium (ETF absorbs flow).

4. A simple model

Let x_t = log(P_t / FV_t) where FV is fair-value NAV. A common reduced form is a mean-reverting process with a no-trade band:
  • Inside the band: dx = -kappa x dt + sigma dW.
  • Outside the band: reversion is accelerated by the arbitrageur flow (regime-switching or threshold AR).
Estimate kappa and the band edges from tick data, using the same timestamp alignment for ETF quotes and basket prices. Report half-life = ln 2 / kappa. The result is only meaningful if the NAV timestamp is aligned with the quote timestamp.

5. Signals and filters used in practice

6. Hedging and P&L decomposition

A premium trade (sell ETF, buy basket) earns the premium minus: basket crossing cost, ETF crossing cost, financing, tracking error until C/R settles, and residual delta from non-hedged constituents. Decompose realized P&L into (a) premium capture, (b) hedge slippage, (c) financing and fees, (d) basis between futures hedge and basket. Link: Delta-1 hedging.

Points to verify

  • HK ETP C/R cutoff times, minimum creation unit sizes, and cash-component rules per product.
  • Whether a given product’s market-maker obligations (spread, size, uptime) are expressed in premium terms.
  • Petajisto’s magnitudes are from US data; do not apply them to HK without local estimation.

Sources

  • Petajisto, A. (2017). Inefficiencies in the Pricing of Exchange-Traded Funds. Financial Analysts Journal.
  • Internal pages linked above.