An ETF is a basket of securities wrapped in a single tradeable share. That share has 

In a frictionless world these two are identical. In the real world they drift apart — by a few basis points normally, by whole percentage points under stress. 

 is the family of trades that profits from forcing those two numbers back together, or from front-running the mechanical buying and selling that index rules require.

Here's the crucial contrast with merger arbitrage, since you've just read that post: in merger arb, convergence is 

 — the deal might break, which is why you get paid a fat spread. In index/ETF arb, convergence is 

 — it's guaranteed by the ETF's plumbing or by a futures expiry date. That near-certainty is wonderful and terrible at the same time: wonderful because you rarely take a catastrophic loss, terrible because everyone can see the same riskless gap, so competition crushes the margin to almost nothing. The money is made on 

, not on being right about an uncertain event.

The engine: creation, redemption, and the authorized participant

You cannot understand ETF arbitrage without understanding the 

. This is the single most important plumbing in the entire ETF universe.

A special class of large institutions — 

 (APs), typically market-making desks — have the right to exchange ETF units for the underlying basket directly with the fund, in large blocks (a "creation unit"). This creates a riskless loop:

This arbitrage runs all day, every day, in the background. It is 

 an ETF tracks its index — not magic, just APs continuously closing a gap for profit. Keep this loop in mind, because 

every interesting ETF arbitrage story is a story about this loop working beautifully — or breaking down.

Form A — Premium / discount (NAV) arbitrage

The plain-vanilla version: when an ETF's market price diverges from its NAV, you trade the gap and let the creation/redemption loop pull it shut.

In a deep, liquid ETF on a domestic index, this gap is tiny and gone in seconds — pure high-frequency territory. It becomes a 

 That's exactly what happened in the most instructive Indian example of the decade.

Real example: Motilal Oswal Nasdaq 100 ETF (when arbitrage broke)

Indian mutual funds can only invest abroad within hard regulatory ceilings. Under SEBI's framework, the industry's overseas investments are capped at US $7 billion overall, with a separate US $1 billion limit for overseas ETFs and a US $300 million sub-limit per fund house. These are 

, not per-fund allowances — every dollar any AMC sends abroad eats into the same cap.

In early 2022 the broader US $7 billion limit was nearly breached, and from 2 February 2022 fund houses were told to halt fresh subscriptions into overseas schemes. Motilal Oswal had already hit its own ceiling in January. Later, as the separate US $1 billion overseas-ETF cap filled up, SEBI halted fresh mutual-fund investment into overseas ETFs effective 1 April 2024.

Here's the arbitrage consequence. With the cap hit, 

the AMC could no longer create new ETF units

 — because creating them would require buying more Nasdaq-100 stocks abroad, which the cap forbids. The creation half of the loop was switched off. Indian investors still wanted Nasdaq exposure, demand kept pushing the ETF's market price up, and 

with no AP able to create units to meet that demand, the price detached from NAV and stayed detached.

. At times the Motilal Oswal Nasdaq 100 ETF traded at a premium to NAV in the range of roughly 12–18% — meaning, as one analysis put it, paying around ₹214 for holdings actually worth about ₹181. And the danger was asymmetric: the premium could evaporate the moment the regulatory limit eased and unit creation resumed, handing recent buyers an instant unrealised loss.

only as long as the arbitrage loop is open.

 The premium/discount is a direct, tradeable readout of how freely APs can create or redeem. When you see a fat, sticky premium, you are not looking at a free lunch — you are looking at a 

, and the real question is what reopens it and when. (The same lesson played out globally in March 2020, when several fixed-income ETFs traded at steep discounts to stale bond NAVs because the underlying bonds had stopped trading — the ETF was arguably pricing reality better than the NAV was.)

Form B — Index reconstitution (inclusion/deletion) arbitrage

Indices are rule-based. When a stock is added to a major index, 

every passive fund tracking that index must buy it

 — on or near a specific date, in size, regardless of price. That forced, price-insensitive demand is predictable, and predictable flow is tradeable. Buy ahead of the index funds; sell into their buying.

Real example: Tesla joins the S&P 500 (December 2020)

The cleanest reconstitution trade in living memory.

S&P announced on 16 November 2020 that Tesla would join the S&P 500, with inclusion effective at the open on 21 December 2020. Because the S&P 500 is market-cap weighted and Tesla was enormous, the mechanical rebalance was gigantic: index funds had to buy roughly US $51 billion of Tesla and sell a similar amount across the other index members, with Tesla taking about a 1% index weight.

Everyone knew this buying was coming, so everyone tried to get in front of it. From the announcement into the 18 December inclusion print, front-running flow drove Tesla roughly 70% higher, peaking right at the inclusion. The liquidity providers had done their job: hedge funds and market makers accumulated the shares in advance precisely so they could deliver them to the index funds at the inclusion print.

And then the tell-tale reversal: once the forced buying was done, the front-running money exited and Tesla drifted sideways-to-down for weeks. Buy the rumour, sell the print.

 — they must own the added name at whatever inflated price the rebalance creates, and dump deleted names into forced selling. Arbitrageurs are paid to provide that liquidity. But this "index effect" has been 

 over the years as markets anticipate it and as many additions are now "migrations" (a stock leaving the S&P MidCap as it joins the S&P 500 nets the forced flows against each other). The edge is real but decaying — which is itself a lesson about how arbitrage competes itself away.

Form C — Index futures vs. cash (basis) arbitrage

The oldest, purest index arbitrage: an index 

 it represents must converge at expiry. Between now and then, they can diverge from fair value, and that gap — the 

Fair Futures = Spot × (1 + (r − d) × T/365)


 is the dividend yield on the basket, and 

 is days to expiry. The intuition: buying the basket today and holding it costs you interest but pays you dividends; the futures price must reflect that net carry.

Fair Futures = 23,500 × (1 + (0.07 − 0.012) × 30/365)
             = 23,500 × (1 + 0.058 × 0.0822)
             = 23,500 × 1.00477
             ≈ 23,612   →  fair basis ≈ +112 points


Now suppose the futures are actually trading at 

 (all 50 Nifty stocks in weight, or a tight tracking proxy) at spot, and 

 at 23,660. You've locked the ~48-point mispricing. At expiry, the future converges to spot, the basis collapses to zero, and you unwind — capturing the gap regardless of where the market went.

: short the basket (via the Securities Lending & Borrowing window) and buy the cheap future.

 (Indian index futures are cash-settled to the spot index). There's no "deal break" risk on the convergence itself. Your risks are different — financing-rate moves, dividend estimation error, execution slippage on a 50-stock basket, and the cost of carrying the position. This is why it's a 

 trade, the structural mirror image of merger arb's fat-margin, low-certainty profile.

The quant perspective: how the money is actually made

If convergence is near-certain and the gaps are tiny, where's the edge? Not in 

executing it faster and cheaper than anyone else, at scale.

liquidity-provision and microstructure business.

1. Fair-value engines running in real time

The desk computes a live fair value for every instrument — ETF NAV from the live basket, futures fair value from the carry model, the implied value of every index constituent — and watches for any instrument straying outside a no-arbitrage band. The models are simple; the 

infrastructure to evaluate them on every tick across thousands of names

2. Modelling the premium/discount as a signal

Persistent premiums/discounts aren't noise — they encode information about creation friction, borrow availability, FX limits, and underlying liquidity. Quants model the 

 of an ETF's premium and trade mean-reversion 

, while flagging regimes (like the Motilal cap) where the loop is jammed and the "arb" is a trap.

3. Predicting reconstitution before it's announced

The richest reconstitution money is made by 

forecasting index changes ahead of the official announcement.

 Index methodologies are public rulebooks — market-cap thresholds, float requirements, profitability tests, ranking dates. Quants run the rulebook on live data to predict adds/deletes weeks early, then build the position before the crowd. After the announcement the edge is mostly gone.

When you must trade a 50-stock basket or unload into an inclusion print, 

 is your dominant cost. Quants model impact as a function of size, volume, and volatility, then schedule the order (VWAP/TWAP, participation algos, dark venues) to minimise it. On thin-margin arb, 

execution quality is the difference between profit and loss

 — there's no fat spread to hide sloppy fills.

Basis arbitrage lives or dies on accurate inputs: the right financing curve and a precise 

 for the index over the contract's life. Misjudging the dividend stream by a few points wipes out the basis you were trying to capture. This is genuine quantitative estimation, not arithmetic.

6. Risk: small, frequent, and occasionally not

The return profile is the inverse of merger arb: 

 punctuated by rare blow-ups when a structural assumption fails — a borrow disappears mid-trade, financing spikes, a creation loop jams, or a basket can't be traded (March 2020 bond ETFs). Risk management is about 

leverage discipline and stress-testing the plumbing

, because the strategy's thin margins tempt heavy leverage, and leverage is exactly what turns a rare jam into a real loss.

The textbook trades assume cheap shorting, cheap financing, and frictionless basket execution. In India, frictions decide whether the trade survives:

The practical takeaway: clean index/ETF arbitrage is largely an 

institutional, low-latency, low-cost-of-capital

 business. For individuals, the more realistic engagement is 

 — e.g., never buying an international ETF at a 15% premium — rather than trying to capture the gap.

When the arbitrage breaks (limits to arbitrage)

The recurring theme of this whole post is that index/ETF arb works 

until the mechanism that guarantees convergence is obstructed

 mechanism is obstructed — and whether it will reopen — is the real analytical skill. A gap is only an opportunity if the force that closes it is still operating.

Written for FessorPro. Educational content only — not investment advice. Regulatory limits and transaction-cost specifics described are India-specific and as of writing; ETF premiums/discounts can change rapidly. Always read scheme documents and do your own research.