Guide · ETFs and index products

ETF arbitrage on NSE: creation, redemption, and why price tracks NAV

The short answer

An ETF trades near the value of its holdings because its supply is elastic. Authorised participants create new units by delivering the underlying basket to the fund, and redeem units by handing them back for the basket. When the price drifts above the fund's indicative NAV, an AP creates units and sells them, adding supply and pushing the price down. When it drifts below, the AP buys the cheap units and redeems them for the basket. Profit-seeking arbitrage, not fund charity, pins the traded price to the value of what the fund owns.

An exchange-traded fund is two prices at once. There is the price it trades at on NSE, set by supply and demand in the order book, and there is the value of the basket of securities it actually holds, published through the session as an indicative NAV. In a closed-end fund those two numbers are free to wander apart for years. In an ETF they are chained together by a mechanism that lets the share count itself expand and contract on demand. This article is about that machinery: how creation and redemption work, why the arbitrage it enables keeps the price honest, why the fund still does not perfectly match its index, and the specific conditions under which the chain snaps and an ETF trades at a large, persistent premium in India.

Creation and redemption: an ETF's supply is elastic

The defining feature of an ETF is not that it trades on an exchange. It is that its supply of units is not fixed. A small set of large institutions, authorised participants (APs), typically banks or market makers, hold a contractual right to transact directly with the fund in large blocks called creation units. This is the primary market, and it runs alongside the on-screen secondary market where everyone else trades.

To create units, an AP assembles the fund's published basket of underlying securities and delivers it to the fund, receiving a creation unit of new ETF shares in return. To redeem, the AP returns a creation unit to the fund and receives the basket of securities back. Most equity ETFs run this exchange in kind, securities for units and units for securities, rather than in cash. That in-kind design is precisely what separates an ETF from a closed-end fund: a closed-end fund issues a fixed number of shares once and then trades them, so its price can float to a deep premium or discount with nothing to correct it, while an ETF can mint or cancel shares whenever demand justifies it.

In-kind, in both directions. The AP does not send cash and wait for the fund to buy stock. It delivers the exact basket and receives units, or the reverse. This keeps the fund's cash balance small and, crucially, hands the AP a portfolio it can immediately trade against, which is what makes the arbitrage in the next section economic.

Creation versus redemption, and what each does to price and supply
Direction	Trigger	What the AP does	Effect on units	Effect on price
Creation	Price above iNAV (premium)	Buys the basket, delivers it, sells the new units on exchange	Supply expands	Pushed down toward iNAV
Redemption	Price below iNAV (discount)	Buys cheap units, returns them, sells the basket received	Supply contracts	Pulled up toward iNAV

The arbitrage that pins price to iNAV

Elastic supply is only half the story. What actually forces the price back to fair value is a profit motive acting on that elasticity. Through the session the fund publishes an indicative NAV (iNAV), a live estimate of the basket's per-unit value recomputed from the latest prices of the underlying securities. In India, SEBI requires AMCs to disclose iNAV on a near-real-time basis, for equity ETFs roughly every fifteen seconds, on the AMC website or through the exchanges. That published number is the yardstick the arbitrage measures against.

Now watch what an AP does with it. Suppose the ETF is bid up on the exchange and trades above its iNAV, a premium. The AP can buy the underlying basket at the prices that define iNAV, deliver it to create fresh units at that lower value, and immediately sell those units on the exchange at the higher traded price, pocketing the gap. The very act of selling the new units adds supply and pushes the price back down toward iNAV. Run it the other way: if the ETF trades below iNAV, a discount, the AP buys the cheap units on the exchange, redeems them for the more valuable basket, and sells the basket, and the buying pressure lifts the price back up. The correction is not a favour the fund does for holders. It is arbitrage, and it continues only as long as the gap is wide enough to clear the AP's costs.

The correction is self-financing. Each leg of the loop earns the AP the premium or discount it closes, so the incentive to act scales with the size of the mispricing. A one-paisa gap is not worth a round trip; a wide gap is, which is why small deviations are tolerated and large ones are hunted almost instantly, as long as the mechanism is available to run.

The two arbitrage scenarios, illustrative
State	Relation to iNAV	AP's arbitrage	Restoring force
Premium	Price above basket value	Buy basket, create units, sell units	Supply added, price falls to iNAV
Discount	Price below basket value	Buy units, redeem for basket, sell basket	Units removed, price rises to iNAV
At fair value	Price near basket value	Gap too small to clear costs, no action	None needed, price rests near iNAV

Why the ETF still does not perfectly match its index

Arbitrage ties the traded price to the fund's basket. It does nothing to tie the fund's basket to the index it is meant to replicate. That second gap is where tracking sits, and it is worth being precise about, because two different quantities travel under similar names.

The tracking difference is the plain return gap over a period: how far the fund landed from the index, up or down. For a passive fund the single largest and most predictable contributor is the expense ratio, which is deducted from the fund's return but never from the index's. Alongside it sit several frictions. Cash drag: a fund holds a little cash for redemptions and to warehouse dividends before reinvesting, and since the index is treated as fully invested, that cash lags a rising market. Sampling or optimisation: a fund that does not hold every constituent, but a representative subset chosen to mimic the index, carries a replication gap versus full replication. Rebalancing and corporate-action timing: the index assumes changes happen instantaneously at a reference price; the fund must actually trade, at real prices, slightly off the reference. Pulling the other way, securities lending earns the fund extra revenue that can offset costs and narrow the difference.

The tracking error is a different statistic. It is the annualised standard deviation of the daily differences between the fund's return and the index's return, so it measures the consistency of tracking, not its size. A fund can post a small, steady tracking difference and therefore a low tracking error, or a difference that swings around day to day and produces a high tracking error even when the average gap is modest. Difference answers "how far off"; error answers "how reliably close". Conflating the two is one of the most common mistakes in ETF commentary.

Costs drag, one revenue source lifts. The bars are illustrative, but the ranking is not accidental: for a plain passive fund the expense ratio usually dominates the tracking difference, which is why comparing funds on cost is a reasonable first cut. The net of all these, and how steady that net is, are two separate things, difference and error.

Tracking-error source catalogue, and the direction each pushes
Source	What it is	Direction on the difference
Expense ratio	Annual fee deducted from the fund but not the index	Drag
Cash drag	Cash held for redemptions and dividends sits uninvested	Drag in a rising market
Sampling or optimisation	Holding a representative subset rather than every constituent	Either way, adds variability
Rebalancing and timing	Trading real prices around index changes and corporate actions	Either way, adds variability
Securities lending	Revenue from lending held securities to borrowers	Offset, narrows the difference

When the arbitrage breaks: the India premium

Every claim so far rests on one assumption: that an AP can freely create and redeem and hedge the basket. Remove that freedom and the pin is gone. The most instructive Indian episode is the one that hit overseas-investing ETFs. Funds that invest abroad do so under a ceiling set by the Reserve Bank of India: an industry-wide overseas-investment limit for mutual funds, with a separate sub-limit reserved for overseas ETFs. Creating a new unit of an overseas ETF means buying more overseas assets, which consumes room under that ceiling.

In 2022 the industry reached the ceiling. With no headroom left, fund houses had to halt fresh creation of international ETF units, because minting a unit would breach the overseas cap. At that point the left-hand side of the arbitrage loop simply switched off. Indian demand for global exposure kept arriving in the order book, but no AP could create new units to satisfy it, so the only way to buy was to lift the price of the existing, fixed float. The traded price detached from iNAV and stayed at a large, persistent premium, in some cases double-digit percentages above basket value, until the constraint eased. Nothing was mispriced in the sense an AP could correct; the correction mechanism itself was unavailable. This is the cleanest demonstration of the whole article's thesis: the price tracks NAV only because arbitrage can run, and when creation is capped, the tracking guarantee lapses.

Read a premium as a signal about the mechanism, not a bargain or a rip-off. A persistent premium to iNAV is usually telling you that creation is constrained or the underlying is hard to access, so the arbitrage that would close it cannot run. Paying that premium means buying units for more than the basket behind them is worth, and the premium can compress abruptly if creation reopens. A thin, illiquid ETF with little on-screen volume can also show wide gaps for the simpler reason that no AP is actively quoting against a small book.

A live regulatory thread. The reference price ETFs are pinned to for their daily price bands is itself under review. In a consultation paper dated 1 February 2026, SEBI noted that ETF base prices are anchored to the T-2 closing NAV, which builds in a one-day lag and needs manual adjustment for corporate actions, and proposed moving to T-1 references, among them the average iNAV of the last thirty minutes on T-1 or the latest available iNAV. A tighter, fresher reference is intended to let market makers work more efficiently and hold premiums and discounts narrower. It is a proposal at the time of writing, not settled rule, and it is exactly the kind of upstream mechanism change that decides how faithfully price tracks value.

The two layers of ETF liquidity

The overseas episode also corrects a widespread misreading of liquidity. Traders often judge an ETF by its on-screen volume alone and conclude that a low-volume ETF is illiquid and unsafe to trade in size. That reasoning misses half the picture. An ETF's real liquidity has two layers: the visible liquidity in its own order book on NSE, and the liquidity of the underlying basket that authorised participants can reach through the primary market.

Because an AP can create or redeem against the constituents, a market maker can quote size in an ETF far beyond what its own printed volume suggests, sourcing or offloading the basket to back the quote. So a broad-index ETF with modest screen turnover can still absorb a large order cleanly if its constituents are liquid. The corollary is the sharper lesson: the second layer is only as deep as the basket. An overseas-equity ETF whose creation is capped, or a fund over an illiquid or hard-to-access underlying, loses that second layer precisely when it is needed, which is why the same instrument can look tradable on a calm day and gap on a stressed one. On-screen volume is the floor of an ETF's liquidity, not its ceiling, and the basket sets the ceiling.

Reading these two layers correctly, distinguishing what the tape shows from what the basket can bear, is the kind of second-order market-structure judgement that separates a chart-watcher from a desk. That upstream reasoning about liquidity, arbitrage bounds and where a printed price can and cannot be trusted is exactly what the method we teach is built around.

Where the ETF pin holds and where it lapses
Condition	Can the AP arbitrage?	Likely price behaviour
Liquid basket, free creation	Yes, both directions	Price hugs iNAV, small deviations only
Creation capped by a ceiling	Creation blocked	Persistent premium above iNAV until the cap eases
Illiquid or halted underlying	Hedge or basket hard to trade	Wide, unstable gaps to iNAV
Thin on-screen order book	Possible but few active quotes	Wide visible spread despite a tradable basket

Where this sits in the wider toolkit

ETF arbitrage is one member of a family of relative-value ideas that all rest on the same logic: two claims on the same underlying value should converge, and something concrete forces the convergence. Here the two claims are the ETF's price and its basket, and the forcing mechanism is creation and redemption. The same convergence-under-a-bound thinking runs through index construction and the flows around a rebalance, and through statistical relationships between related instruments. If you want to see how the pieces connect, the related reading below moves from the structure of an ETF to the rebalancing flows it must absorb and the statistical convergence trades that share its spirit.

Within the Bharath Shiksha curriculum, the market-structure treatment of ETFs, iNAV and the arbitrage bound sits in the execution and market-microstructure material, while the statistical convergence machinery is developed in the time-series and pairs-trading volumes. The point of teaching it as mechanism rather than tip is durability: the branded product and the exact ceilings change, but creation, redemption and the arbitrage bound do not.

Common questions

Frequently asked questions

What keeps an ETF price close to its NAV?

An arbitrage mechanism, not a rule. An ETF's supply is elastic: authorised participants create new units by delivering the underlying basket to the fund, or redeem units by returning them for the basket. When the traded price rises above the fund's indicative NAV, an AP buys the basket, creates units and sells them, adding supply and pulling the price down. When it trades below, the AP buys the cheap units, redeems them for the basket and sells the basket. Profit-seeking, not fund charity, pins the price to the value of the holdings.

What is the difference between creation and redemption of ETF units?

Creation adds units and expands the fund; redemption removes units and contracts it. In creation, an authorised participant delivers the specified basket of securities, or cash, to the fund and receives a large block of new units called a creation unit. In redemption, the AP returns a creation unit to the fund and receives the basket back. Because units can be created and cancelled on demand, an ETF's share count floats with demand, which is what distinguishes it from a closed-end fund whose share count is fixed.

What is iNAV and how often is it published?

The indicative net asset value, or iNAV, is a live estimate of the per-unit value of an ETF's basket, computed through the session from the latest prices of the underlying securities. In India, SEBI requires AMCs to disclose iNAV on a near-real-time basis, for equity ETFs roughly every 15 seconds, on their website or through the exchanges. It is the reference an authorised participant compares the traded price against to see whether the ETF is at a premium or a discount, and by how much.

Why does an ETF still not perfectly match its index?

Several frictions sit between the fund and the index. The expense ratio is deducted from fund returns but not from the index. Cash held for liquidity and dividends is not fully invested, so it drags in a rising market. A fund that samples or optimises rather than holding every constituent picks up a replication gap. Rebalancing and corporate-action timing differ from the index's instantaneous assumption. Securities-lending revenue works the other way and narrows the gap. The net of these is the tracking difference; how much it wobbles over time is the tracking error.

What is the difference between tracking difference and tracking error?

They answer different questions. Tracking difference is the actual return gap between the fund and its index over a period: how far the fund landed from the benchmark, mostly explained by the expense ratio. Tracking error is the annualised standard deviation of the daily return differences: how consistent that gap is from day to day. A fund can have a small, steady difference and therefore a low tracking error, or a difference that swings around and produces a high tracking error even if the average gap looks modest. Difference is the size; error is the variability.

Can an ETF trade at a large premium or discount to iNAV?

Yes, and it happens whenever the arbitrage is switched off. The correction depends on authorised participants being free to create and redeem and to hedge the basket. If creation is capped, or the underlying is hard to access, an AP cannot manufacture new units to meet demand, so on-exchange buying pushes the price above iNAV and the premium persists. In India this happened when the industry's overseas-investment ceiling was reached and fresh creation of international ETF units was halted, leaving some of those ETFs at double-digit premiums to their basket value.

Why did some India-listed international ETFs trade far above NAV?

Because the supply of units was frozen. The Reserve Bank of India caps how much the mutual fund industry can invest overseas, with a separate sub-limit for overseas ETFs. When those ceilings were reached in 2022, fund houses had to stop creating fresh units, since creating a unit would mean buying more overseas assets. With supply fixed and Indian demand for global exposure still rising, the traded price detached from iNAV and stayed at a large premium until the constraint eased, a textbook case of the creation arbitrage being unavailable rather than a mispricing an AP could correct.

Does low screen volume mean an ETF is illiquid?

Not necessarily. An ETF's real liquidity has two layers: the visible on-screen order book, and the liquidity of the underlying basket that authorised participants can access to create or redeem. Even if on-screen volume looks thin, a market maker can source or offload the basket and quote size against it, so the effective depth can be far greater than the tape suggests. What matters for a large order is the liquidity of the constituents, not just the ETF's own printed volume. The reverse also holds: a liquid-looking ETF over an illiquid basket can gap when the basket is hard to trade.

Is SEBI changing how ETF prices are referenced?

It has proposed to. In a consultation paper dated 1 February 2026, SEBI flagged that ETF price bands are anchored to the T-2 closing NAV, which builds in a one-day lag and forces manual adjustment for corporate actions. It proposed moving to T-1 references, among them the average iNAV of the last 30 minutes on T-1 or the latest available iNAV, to cut that lag. A tighter reference is intended to let market makers operate more efficiently and keep premiums and discounts narrower. It is a proposal at the time of writing, not settled rule.

Where the facts come from

Sources

ETF creation and redemption mechanism. The in-kind exchange of the underlying basket for creation units, the role of authorised participants, and the arbitrage that keeps the traded price near NAV are the standard specification of the product. etf.com
iNAV disclosure and market makers in India. SEBI's mutual fund framework requires AMCs to disclose iNAV on a near-real-time basis, for equity ETFs roughly every 15 seconds, and to appoint at least two market makers to provide two-way quotes. sebi.gov.in
Overseas-investment ceiling and the international-ETF premium. The RBI industry limit for overseas investment by mutual funds, with a separate sub-limit for overseas ETFs, was reached in 2022; fund houses halted fresh creation of international ETF units, and with supply frozen those ETFs traded at large, persistent premiums to iNAV. 1finance.co.in
Tracking difference versus tracking error. Tracking difference is the return gap over a period, driven largely by the expense ratio with cash drag, sampling, rebalancing timing and securities-lending offsets; tracking error is the annualised standard deviation of the daily return differences. morningstar.com
SEBI ETF base-price and price-band review. Consultation paper dated 1 February 2026 flagging the T-2 NAV lag and proposing T-1 references, including a last-30-minute average iNAV, to reduce the pricing lag that widens premiums and discounts. business-standard.com

Educational note. This article explains the market-structure machinery of exchange-traded funds: creation, redemption, the arbitrage that pins price to NAV, tracking error, and the conditions under which the mechanism breaks. It is not a recommendation to trade or invest in any ETF, index product or security, and it is not investment advice. Bharath Shiksha is an educational publisher, not a SEBI-registered investment adviser or research analyst. All figures shown in the diagrams and tables are illustrative and not to scale.