6B.9 How ETFs trade: market makers, creation units, iNAV, premiums

Motivation

An ETF has two prices: the NAV of its holdings, and the price at which its units trade on the exchange. The second is held near the first by an arbitrage that only some participants can perform. When that arbitrage works, an ETF is the cheapest way to hold an index; when it stops, retail buyers can pay 20% more than the assets are worth.

6B.9.1The machinery#

Exchange trading. ETF units trade on NSE and BSE in the normal order book (M01) and settle T+1 into a demat account. The investor pays brokerage, the bid–ask spread, stamp duty on purchases, and exchange charges, and there is no exit load.
Market makers. The AMC appoints exchange members as market makers who quote two-way prices in the ETF and may deal with the AMC directly in creation-unit size.
Creation units. The AMC issues and redeems units only in large blocks (a creation unit, for example tens of thousands of units). A creation is usually in kind: the market maker delivers the index basket plus a cash component and receives new ETF units; a redemption is the reverse. Some ETFs (debt, international, commodity) create in cash. The AMC charges a transaction fee.
Direct dealing. Since 1 Nov 2022, investors other than market makers may create or redeem directly with the AMC only for amounts above ₹25 cr; smaller investors buy and sell on the exchange (SEBI circular of 23 May 2022, applicability deferred to 1 Nov 2022 by the addendum of 28 Jul 2022)¹.
iNAV. The indicative NAV is computed during the day from the live prices of the basket (and the exchange rate for international ETFs) and is published on the exchange and AMC websites. It is the fair-value benchmark for the traded price.
Price bands. A SEBI circular of 15 Jun 2026 replaced the old flat ±20% band around the NAV of two days earlier with a base price equal to the previous day's closing price (the last 30 minutes' VWAP), an initial band of ±10% for equity and debt ETFs that can widen in steps to ±20%, and an initial ±6% for gold and silver ETFs that widens in 3% steps, plus a pre-open call auction for commodity ETFs; implementation was extended to September 2026 by a circular of 28 Aug 2026, and SEBI proposed moving the base to the previous day's closing NAV from 1 Apr 2027².
Gold and silver valuation. From 1 Apr 2026 the gold and silver held by mutual funds are valued at domestic spot prices polled by Indian exchanges instead of the LBMA benchmark (SEBI circular of 26 Feb 2026)³.

6B.9.2The no-arbitrage band#

Proposition 6B.9.1premium band

Let the iNAV be $I$ and the ETF's mid price $P$ . A market maker can (i) create: buy the basket at a proportional cost $\kappa_b$ above $I$ , pay the creation fee $\varphi_c$ , carry the position for the settlement lag at cost $h$ (financing and price risk), and sell the new units at the bid, a half-spread $s/2$ below $P$ ; or (ii) redeem: buy units at the ask, a half-spread above $P$ , pay the redemption fee $\varphi_r$ and carry $h$ , and sell the basket at a cost $\kappa_s$ below $I$ . If both routes are open, absence of profitable arbitrage requires, to first order in the costs,

\begin{aligned} I\,(1 - C_r) \;\le\; &P \;\le\; I\,(1 + C_c),\\ C_c &= \kappa_b + \varphi_c + h + s/2,\\ C_r &= \kappa_s + \varphi_r + h + s/2. \end{aligned} \tag{6B.26}

If creations are impossible, there is no upper bound; if redemptions are impossible, there is no lower bound.

Proof

Creation per unit of value costs $I(1 + \kappa_b + \varphi_c + h)$ and yields $P(1 - s/2)$ ; it is profitable iff $P/I > (1 + \kappa_b + \varphi_c + h)/(1 - s/2) = 1 + C_c + O(\text{cost}^2)$ . Redemption costs $P(1 + s/2)$ plus fees and carry and yields $I(1 - \kappa_s)$ ; it is profitable iff $P/I < (1 - \kappa_s - \varphi_r - h)/(1 + s/2) = 1 - C_r + O(\text{cost}^2)$ . Competing market makers act whenever either inequality holds, issuing units (which pushes $P$ down) or retiring them (which pushes $P$ up), until neither holds. If a route is closed, its inequality imposes nothing.∎

The band is a bound on arbitrage, not a prediction of where $P$ sits: inside it, the price is set by order flow and by the market maker's inventory. Petajisto (2017) found that US ETF prices stay close to NAV for liquid domestic equity ETFs, while international and illiquid ETFs show wider and more persistent deviations, as (6B.26) predicts. The arbitrage also runs the other way: Ben-David, Franzoni and Moussawi (2018) show that creation and redemption transmit demand shocks in ETF units to the underlying stocks, so stocks with more ETF ownership are more volatile.

Worked example 6B.12a Nifty 50 ETF

iNAV ₹270.00. A market maker creating units pays STT of 0.1% on buying the basket, stamp duty of 0.015% and exchange charges of about 0.003%, and expects impact of 0.03%: $\kappa_b = 0.148\%$ . Take a creation fee of 0.02%, a carry of 0.01% and an ETF half-spread of 0.03% (illustrative): $C_c = 0.208\%$ . For redemption: STT of 0.1% on selling the basket, exchange charges 0.003%, impact 0.03% ( $\kappa_s = 0.133\%$ ), a redemption fee of 0.02%, carry 0.01%, half-spread 0.03% and stamp duty of 0.015% on buying the ETF units: $C_r = 0.208\%$ . The band is ₹269.44 to ₹270.56. STT alone makes the Indian band for equity ETFs at least 0.1% wide on each side, wider than for a US equity ETF.

6B.9.3When the band breaks#

International ETFs and the overseas limit Indian mutual funds may invest abroad within an industry-wide limit of US$7 bn, with US$1 bn per AMC, and a separate industry limit of US$1 bn for investment through overseas ETFs⁴. When the industry approached the limit, SEBI asked AMCs to stop fresh subscriptions to overseas schemes from 1 Feb 2022 (a letter of 28 Jan 2022); a letter of 17 Jun 2022 allowed subscriptions to resume up to each AMC's 1 Feb 2022 level⁵; and in March 2024 AMCs were told to stop subscriptions to schemes investing in overseas ETFs from 1 Apr 2024 as that limit neared⁶. No increase in the limits had been announced by 6 Nov 2026 [S]. With creations stopped, the upper edge of (6B.26) disappears while redemption still bounds the price from below. The result was persistent premiums: a Nasdaq-100 ETF traded at ₹214.5 against an iNAV of ₹181.4 on 26 Dec 2024, a premium of 18.2%⁷; a NYSE FANG+ ETF at about 22.5% in April 2026⁸. A buyer at an 18% premium loses 18% on the day creations reopen, whatever the index does. NSE issued a press release on 25 Sep 2026 cautioning investors about international ETFs trading at substantial premiums⁹.

The silver ETF premium of October 2025 In early October 2025 a shortage of physical silver in London and in India coincided with strong Indian festive demand. The domestic premium of physical silver over the international price rose from about 0.5% in early September to 5.7% at the close on 9 Oct, with intraday spikes of 12% (an AMC note as reported by Business Standard, 19 Oct 2025), and silver ETFs traded 10% or more above iNAV¹⁰. Market makers could create only by buying physical silver of the required standard at the domestic premium, so $\kappa_b$ in (6B.26) jumped while the iNAV (then based on the international benchmark) did not. From 10 to 14 Oct 2025 at least five AMCs suspended lump-sum purchases and switch-ins into their silver ETF fund of funds, because the FoFs would have had to buy ETF units at the premium¹¹. By 17 Oct the ETFs traded at par or at discounts of up to about 3%, roughly 10% below their peaks¹². The moral is (6B.26) read backwards: the premium measured the cost of creation, and buyers paid it.

Thin ETFs Many Indian ETFs trade a few lakh rupees a day. With one market maker quoting wide, the half-spread $s/2$ alone can be 0.5–1%, and the traded price can drift from iNAV within a wide band. The NAV-tracking quality of the fund is irrelevant to an investor who pays 1% each way to get in and out.

Practical rules

Compare the order price with the live iNAV; use a limit order at or near it, never a market order.
Avoid the first and last minutes of the session, when iNAV is stale (the basket has not opened, or the close is being set) and spreads are widest.
For international and commodity ETFs, check whether creations are open before buying.
Prefer ETFs with tight quoted spreads and real depth, not just low expense ratios.

6B.9.4Choosing an index fund or an ETF#

Proposition 6B.9.2crossover horizon

An ETF has a one-off round-trip cost $k_E$ (spread, brokerage, stamp duty, expected premium paid) and annual cost $c_E$ ; an index fund on the same index has $k_I$ (stamp duty, exit load if any) and $c_I > c_E$ , with $k_E > k_I$ . With gross return $r$ , the ETF leaves more wealth after $T$ years if and only if $T > T^*$ , where

T^* = \frac{\ln\big[(1 - k_I)/(1 - k_E)\big]}{\ln\big[(1 + r - c_E)/(1 + r - c_I)\big]} \;\approx\; \frac{k_E - k_I}{c_I - c_E}. \tag{6B.27}

Proof

Wealth per rupee is $(1 - k_j)(1 + r - c_j)^T$ . The ETF wins iff $(1 - k_E)(1 + r - c_E)^T > (1 - k_I)(1 + r - c_I)^T$ ; take logs and divide by the positive $\ln[(1 + r - c_E)/(1 + r - c_I)]$ . The approximation uses $\ln(1 + x) \approx x$ for both logarithms and drops the $1 + r$ in the denominator.∎

Worked example 6B.13

An ETF at $c_E = 0.05\%$ with a round-trip cost of $k_E = 0.25\%$ against an index fund at $c_I = 0.20\%$ with $k_I = 0.006\%$ (stamp duty and STT): $T^* \approx 0.244/0.15 = 1.63$ years by the approximation, and 1.82 years by the exact formula with $r = 12\%$ . For a lump sum held for decades the ETF wins. For a SIP of ₹5,000 a month bought through a brokerage plan with a flat ₹20 per order, the brokerage alone is 0.4% of each instalment; then $k_E$ for each instalment exceeds 0.4%, and if many instalments are sold within a few years the index fund wins. The answer depends on ticket size, holding period and liquidity, which (6B.27) makes explicit.

Practitioner notes

Large, liquid Nifty 50 and Sensex ETFs are cheap to hold and cheap to trade; small-cap, sectoral and factor ETFs often are not.
An index fund is the ETF's NAV without the spread, the premium and the order: for regular small purchases it is usually the cleaner instrument.
See it live: listed ETFs with their prices are at https://oicharts.com/etfs (gold, silver, equity and debt groups); compare the traded price with the iNAV on the exchange or AMC page before any order.

Exercisem06b_etf.pyOpen in editor →