Fixed Income ETF Microstructure

The underlying problem

An equity index fund tracks instruments that trade continuously on lit exchanges. A bond index fund does not, and the differences are structural:

• Enormous universes. A global aggregate index can contain 25,000–30,000 individual bonds; a single issuer may have dozens of outstanding lines with different coupons and maturities.
• Most do not trade daily. A large share of investment-grade corporate bonds go days or weeks without a transaction.
• Trading is dealer-intermediated and opaque, historically over the counter, with prices negotiated rather than posted.
• New issues arrive constantly and index composition churns as bonds mature, are called or are downgraded (Chapter 43).

Wrapping that in a continuously-traded instrument is not a cosmetic improvement. It is a genuine liquidity transformation, and the honest analysis has to hold both its benefits and its risks simultaneously.

Evaluated pricing, and why NAV is an estimate

Because most bonds have no trade to price from, fund NAVs use evaluated prices from pricing services: model-based estimates built from comparable bonds, dealer quotes, matrix pricing and issuer curves. These are professional, widely used and usually good. They are also, by construction, backward-looking and smoothing — they update as inputs update, and in fast markets the inputs lag.

This single fact explains a family of phenomena covered elsewhere in this book:

• Premiums and discounts breathe wider in bond ETFs than in equity ones (Chapter 33), partly because NAV itself is an estimate.
• The ETF frequently leads. With thousands of live transactions per hour, the ETF's price incorporates new information faster than the evaluated NAV does. The tail is not wagging the dog; the dog is being described by a slow observer.
• In stress, the gap becomes dramatic. March 2020's headline discounts of 3–6% were largely the distance between a live traded price and a stale estimate: the subject of Chapter 84's full case study.

The practical rule follows directly and applies to any investor holding bond ETFs: in disorderly markets, trust the traded price over the NAV, and never assess a bond ETF's "fair value" from a NAV struck hours ago.

The creation and redemption adaptation

Chapter 31's clean in-kind mechanism cannot run unmodified when APs cannot source the instruments. The adaptations:

Sampled baskets. The creation basket is a few hundred liquid bonds matched to the portfolio's duration, credit and sector profile, not a pro-rata slice (Chapter 69).

Custom baskets, negotiated per transaction. The AP and issuer agree what can actually be delivered, letting the fund acquire what it needs and dispose of what it is overweight (Chapter 70). This is where the flexibility is genuinely load-bearing: without it, bond ETF creations would be far more cash-dependent.

Cash creations with transaction-fee adjustments. Where in-kind is impractical, the AP delivers cash and the fund trades, with a fee schedule ensuring the flowing investor bears the cost rather than continuing holders (Chapter 31's principle preserved).

Consequences to carry forward: bond ETFs realize more internal gains than equity ETFs, so the Chapter 54 tax shield is weaker; tracking is looser; and the arbitrage band is wider because the round trip through the basket genuinely costs more.

How market makers actually hedge

The critical operational fact: a market maker quoting a corporate bond ETF is usually not hedging with the underlying bonds. Instead:

• Credit derivatives (index CDS) and interest-rate futures hedge the two main risk factors, spread and duration.
• Other bond ETFs serve as hedges for correlated exposure.
• Residual basis risk is priced into the spread: the hedge is imperfect, and the imperfection is what the investor pays for.

This is why bond ETF spreads widen when credit derivative markets are disorderly, and why the ecosystem's health depends on liquid hedging instruments existing at all. It also explains the counterintuitive observation from Chapter 71: the ETF can be more liquid than its own holdings, because it aggregates diffuse demand into one instrument that can be hedged in size.

Price discovery: the genuine achievement

Multiple studies, and the direct experience of 2020, support a claim that would have sounded absurd fifteen years ago: bond ETFs are now a leading source of price information for the bond market. The reasoning is mechanical. The ETF trades thousands of times an hour; its constituents may not trade at all. When conditions change, the ETF's price moves first because it is the only instrument where opinion can be expressed immediately.

Regulators and central banks have accepted this in practice as well as in analysis: the Federal Reserve's 2020 corporate credit facilities purchased bond ETFs precisely because they offered immediate, executable access to broad credit exposure when individual bonds could not be sourced at scale.

The criticism, stated at full strength

This is a liquidity transformation, and liquidity transformations have a history in finance. The serious version of the concern (developed fully in Chapter 87):

The mismatch is real. Investors can sell an ETF instantly; the underlying bonds cannot all be sold instantly. In a sufficiently one-directional flight, secondary-market netting (Chapter 68) does less work, redemptions reach the primary market, and the fund must transact in a market that has stopped functioning.

The counter-evidence is also real. In every stress episode so far, including the most severe bond-market seizure in decades, the mechanism held: redemptions were met, largely in kind or with fee-adjusted cash; discounts appeared and closed; and the ETF continued providing prices while the underlying market went quiet. Several analyses concluded the wrapper reduced pressure on the underlying by absorbing flow in the secondary market.

What remains genuinely open. Every test so far has ended with policy intervention in the broader credit market. Whether the machinery holds in a comparable seizure without a central bank arriving is untested, and honesty requires saying so rather than treating three survivals as proof.

A worked example

A €2bn EUR corporate bond ETF on a stressed Thursday.

• Underlying market: roughly 12% of the fund's 1,600 holdings trade at all that day. Dealers quote wide and in small size.
• The ETF: 6,400 trades, €140m of volume, continuous two-sided quotes at a spread of 0.35% (up from 0.09%).
• The apparent discount: the ETF closes 1.8% below the evaluated NAV — because the ETF price reflects Thursday's sentiment and the NAV reflects estimates built partly from Tuesday's inputs.
• The primary market: net redemptions of €40m, met with a custom basket of the fund's most liquid overweights, so the fund itself sells almost nothing into the disorder.
• Two days later: pricing services update, NAV falls toward where the ETF traded, and the "discount" largely disappears without the ETF price moving much.

Everything in that sequence is the machinery working. It is also, from the outside and in a headline, indistinguishable from a fund trading at a scary discount: which is why this chapter exists before the case study that follows.