> ## Documentation Index
> Fetch the complete documentation index at: https://raymondchang.me/llms.txt
> Use this file to discover all available pages before exploring further.

# ETF arbitrage and premium/discount models

> How to model ETF premium/discount, iNAV staleness, and the cost band that bounds arbitrage.

> [!NOTE]
>
> **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](https://raymondchang.me/equities-derivatives/global/etf-creation-redemption.md) and [HK market maker rules](https://raymondchang.me/equities-derivatives/hk/cash-etf/market-maker-rules.md).

## 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

| Signal | Use | Main failure mode |
| - | - | - |
| Premium z-score vs fair-value NAV | Entry | Stale NAV makes z-score look large |
| Premium vs C/R cost band | Execution decision | Band estimated with wrong borrow cost |
| Flow (creations/redemptions, AUM change) | Persistence | Reported with a lag |
| Constituent spread and depth | Capacity | Depth at top of book overstates size |

## 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](https://raymondchang.me/equities-derivatives/global/delta-1-hedging.md).

## 7. Mock interview links

* [ETF market making](https://raymondchang.me/equities-derivatives/mock/etf-market-making.md)
* [Quant mock interview](https://raymondchang.me/quantitative-research/mock-interview.md)

## 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.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.