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

# Index and basis arbitrage models

> Fair value, implied financing, and cost bands for index futures basis trades.

> [!NOTE]
>
> **Status of this page:** P0 draft. Desk-level mechanics are on [Index arbitrage](https://raymondchang.me/equities-derivatives/global/index-arbitrage.md) and the [mock interview](https://raymondchang.me/equities-derivatives/mock/index-arbitrage.md); this page covers the modelling layer.

## 1. Fair value

With discrete dividends, cost-of-carry fair value of an index future is

`F* = (S - PV(D)) x (1 + r x d / 365)`

where S is spot, D is the dividends expected before expiry, r is the funding rate, and d is days to expiry. Use the continuously compounded form if the desk convention requires it.

**Worked example (verified numerically).** S = 20,000, dividends = 60 points, r = 4%, d = 30 days, F = 20,040.

* Fair value = (20,000 - 60) x (1 + 0.04 x 30/365) = 20,005.56.
* Mispricing = 20,040 - 20,005.56 = 34.4 points.
* Implied financing rate = (F / (S - D) - 1) x 365 / d = 6.10%.

The implied financing rate is the cleaner quantity to monitor: it can be compared directly with the desk's funding curve and with other expiries.

## 2. Cost band

A trade is only attractive when |F - F\*| exceeds the all-in round-trip cost:

* Futures spread and fees, basket crossing cost, stamp duty or transaction levies where applicable.
* Borrow cost if the cash leg is short.
* Funding spread over the risk-free rate.
* Dividend uncertainty (ex-date and amount) and index-weight rebalancing risk.
* Tracking error of the replicating basket (full replication vs sampled).

Band asymmetry is normal: short-stock legs are costlier than long-stock legs.

## 3. Modelling the basis

1. Convert futures prices into implied financing (or implied dividend) so that different expiries are comparable.
2. Fit a term structure of implied financing. Deviations of a single expiry from the curve are calendar-spread signals.
3. Model the residual as mean-reverting with a half-life estimated out of sample. See [stat arb](https://raymondchang.me/quantitative-research/strategies/stat-arb-pairs.md) for the estimator.
4. Treat dividend forecasts as a separate model: the implied dividend is the main source of apparent mispricing around the dividend season.

## 4. Execution structure

* **Basket vs futures** requires simultaneous execution; leg risk is the dominant risk.
* **Closing-auction execution** lets the cash leg trade at the official close, which matters for index-tracking flows. See [auction mechanisms](https://raymondchang.me/quantitative-research/microstructure/auction-mechanisms.md).
* **Roll trades** (expiry to expiry) carry less cash-leg risk but depend on the calendar spread.

## 5. Risks

* Dividend surprise or ex-date shift.
* Index rebalancing and constituent changes.
* Borrow recall.
* Margin and funding changes on the futures leg.
* Settlement price definition (average of final-day quotes vs closing price).

## Points to verify

* HKEX settlement-price definition for HSI and HSCEI futures and the current margin rules.
* Stamp duty treatment per leg for the specific instruments used.
* Day-count convention used by the desk funding curve (365 vs 360).

## Sources

* Internal pages linked above. The fair-value formula is the standard cost-of-carry relationship; the numbers were checked by calculation.


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