Creation and Redemption: The Mechanism That Never Makes the Fund Sell

Think of a token a wholesaler will always exchange, in bulk, for one fixed crate of fruit — a set number of mangoes, apples and oranges in fixed proportions, no substitutions. If the token ever starts trading in the market for more than a crate's worth of fruit is actually worth, anyone can buy a wholesale crate, hand it to the wholesaler, collect tokens in return, and sell those tokens at the inflated price — pocketing the gap until enough people do it that the gap closes. If the token trades for less, the same trade runs in reverse: buy cheap tokens, redeem them for a crate, sell the fruit. Nobody has to believe the token is worth anything in particular. The trade only requires that somebody, somewhere, is willing to do the arithmetic and act on it.

That's the entire mechanism keeping an ETF's market price honest — and on the morning of August 24, 2015, the people who normally do that arithmetic couldn't.

What creation and redemption actually do

A new ETF share isn't created by the fund receiving cash and buying stock with it. It's created when an Authorized Participant delivers a full basket of the fund's underlying securities, in fixed proportions, to the fund in exchange for a block of new ETF shares — typically tens of thousands of shares at a time, called a creation unit. Redemption runs the same trade in reverse: the Authorized Participant hands back a creation unit's worth of ETF shares and receives the underlying basket, not cash.

That in-kind exchange is the entire difference from Liquidity Mismatch's mutual funds. A mutual fund honoring a redemption has to raise cash somehow, which usually means selling something out of its own portfolio — the exact mechanic that emptied Franklin Templeton's liquid cushion and forced the Reserve Primary Fund to hold assets it couldn't sell at the price it needed. An ETF honoring a redemption doesn't sell anything at all. It hands over the actual securities, in kind, to the one party positioned to know what to do with them — and the fund's own portfolio never has to transact, or raise cash, or touch its liquid cushion, regardless of how many creation units get redeemed that day.

The Authorized Participant is what makes the token-and-crate arbitrage actually run. Only large, specially contracted market makers and institutional trading firms sign the agreement letting them create and redeem creation units directly with the fund; everyone else trades the ETF's shares on the exchange, at whatever price buyers and sellers agree on. When that price drifts from the value of the underlying basket, the Authorized Participant's arbitrage — buy the cheap side, redeem or create, sell the expensive side — is the only mechanism pulling the two back together. Nobody mandates that it happens. It happens because it's profitable, right up until it isn't.

The day the arbitrage couldn't run

On August 24, 2015, a sharp overnight drop in Asian markets carried into the US open, and the New York Stock Exchange's own circuit-breaker rules — designed to give a volatile stock a pause before it resumes trading — left a meaningful share of S&P 500 constituent stocks not yet trading, sometimes for the first thirty minutes of the session. An Authorized Participant trying to price an ETF basket that morning couldn't get a reliable quote on every stock inside it, because some of those stocks simply hadn't printed a trade yet. Without a dependable basket price, the arbitrage trade an Authorized Participant would normally run — buy the ETF, redeem for the basket, sell the basket, or the reverse — became a bet on prices nobody could see, and the sensible response was to stop making that bet entirely.

With the correcting mechanism sitting out, ETF prices on the exchange were set by whoever was still willing to trade, at whatever price that took. Roughly one in five US-traded ETFs fell 20% or more that morning, some by more than 40%, and reported discounts to net asset value reached as much as 26 percentage points below what the underlying holdings were actually worth. Nothing had happened to the value of the stocks inside those ETFs anywhere close to that scale. What had happened was that the one mechanism keeping the ETF's price tethered to that value had, for half an hour, nobody willing to operate it.

India: the same mechanism, built for a different job

India's ETF market runs on the identical in-kind creation and redemption structure — an Authorized Participant or Large Investor exchanges a basket of securities for creation units directly with the fund, exactly as in the US — but two of its most prominent uses point at a different purpose entirely. The CPSE ETF and Bharat 22 ETF were built as disinvestment vehicles: baskets of equity stakes the government holds — directly in central public sector enterprises for the CPSE ETF, and through a mix of PSU holdings and stakes held via the government-owned SUUTI trust for Bharat 22 — sold down through the ETF wrapper itself, using creation and redemption as the exit mechanism for the state's own holdings rather than a fee-collecting index-tracking product. Bharat Bond ETF, launched in 2019, extended the same in-kind structure to corporate bonds — a basket of eligible PSU bonds, exchanged in kind with the fund by Authorized Participants and Large Investors, per its own scheme documents.

Nothing about the underlying mechanism changes between an index fund tracking the S&P 500 and a fund the Indian government is using to sell down its own holdings in public-sector companies. What changes is why the wrapper exists in the first place — a genuinely different answer to "why build this as an ETF" than the pure arbitrage-driven convenience product the mechanism was originally engineered for in the US.

Why this matters for a Business Analyst

Go back to what Liquidity Mismatch established: a mutual fund's redemption promise is a claim about the fund's own operations, not about how fast its assets can be sold, and the gap between the two is what breaks under stress. A BA who has internalized that lesson and moves to modeling ETF risk the same way will look for the wrong failure. An ETF genuinely doesn't have a Franklin Templeton problem — the fund's portfolio is structurally insulated from ever needing to sell into a falling market to meet redemptions, because redemptions aren't paid in cash the fund has to raise.

What an ETF has instead is a pricing problem with a single point of failure: the market price an ordinary investor sees only stays honest while an Authorized Participant finds it worth their while to keep correcting it. A risk model built around "can the fund meet redemptions" will pass an ETF with flying colors during exactly the kind of event that makes its exchange price meaningless — because the fund was never the thing under stress. The Authorized Participant's willingness to act was, and nothing about a redemption-capacity check was ever built to see that.

Lighthouse Insight

Back to the token and the crate of fruit.

The token never runs out of crates to redeem, because redeeming one doesn't require the wholesaler to sell anything — it just hands over what's already sitting in the warehouse. That's a genuinely different, and genuinely better, answer to the problem that froze Franklin Templeton's six schemes and broke the Reserve Primary Fund's dollar. It buys that improvement by moving the risk somewhere else entirely: onto whether someone is still willing, on any given morning, to walk over and do the exchange. On August 24, 2015, for about thirty minutes, nobody was — and the token's posted price stopped meaning anything at all, right up until somebody found the courage to start doing the arithmetic again.