March 2020: Bond ETFs Under Fire

What happened

The setup. As the pandemic escalated, a dash for cash hit every risk asset simultaneously. Credit spreads gapped wider; investment-grade and high-yield markets faced enormous redemption pressure across all fund types.

The seizure. Corporate bond trading became extraordinarily difficult. Dealers, constrained by post-2008 balance-sheet rules and their own risk limits, reduced quoting sharply. Many bonds simply did not trade for days; those that did traded at prices far from prior marks. Even the Treasury market, the world's deepest, experienced severe dislocation, requiring central bank intervention in its own right.

The ETFs. Bond ETFs continued quoting and trading throughout, in size, with spreads several times normal. Their prices fell faster and further than their published NAVs, producing the discounts that generated the headlines: mid-single-digits in investment grade, larger in high yield.

The intervention. On 23 March, the Federal Reserve announced credit facilities that would, unprecedentedly, purchase corporate bond ETFs alongside individual bonds. Markets stabilised rapidly, and the facilities ultimately deployed far less than authorised.

The diagnosis: whose price was wrong?

This is the analytical core of the episode, and Chapter 82 built the machinery for it.

NAV was an estimate from vanishing inputs. Evaluated pricing infers a bond's value from comparable bonds, dealer quotes and issuer curves. When comparables stop trading and dealers stop quoting, the inputs become stale, and the estimate drifts toward yesterday's reality. In a fast-moving crisis, NAV was describing a world that had already ended.

The ETF was a transaction record. Thousands of trades per hour between willing buyers and sellers, with real money, in real time. Whatever that price was, it was the price at which credit exposure could actually be transferred that afternoon: which is the operative definition of a market price.

The confirmation came from the Fed. When the central bank began buying actual corporate bonds, transaction prices printed close to where ETFs had been trading, not where NAVs had been marked. NAVs then adjusted downward toward the ETF prices, and the "discounts" closed largely without the ETFs rallying.

The academic and official verdict. Federal Reserve, BIS and academic analyses converged on the same reading: bond ETFs provided price discovery when the underlying market could not, and their discounts reflected NAV staleness plus a genuine liquidity premium for immediacy, rather than a failure of the arbitrage mechanism.

What the machinery actually did

Four operational findings, each answering a specific fear:

Creation and redemption kept working. Primary market activity rose to record levels rather than freezing. APs continued transacting, and custom baskets (Chapter 70) proved essential: funds could deliver what could actually be delivered instead of demanding the impossible.

Redemptions were met, largely without fire sales. Because most selling met buyers in the secondary market (Chapter 68's netting), only net imbalances reached the primary window. Several analyses concluded bond ETFs reduced pressure on the underlying market by absorbing flow that would otherwise have become forced bond sales.

The discount was a shock absorber. Investors wanting immediate liquidity paid for it through the discount. That is the price mechanism functioning: immediacy is scarce in a crisis and should cost something. Mutual funds, by contrast, transacted at stale NAVs, meaning redeeming investors took value from those who stayed: the opposite distribution of costs.

Spreads widened but quotes persisted. Market makers charged crisis prices for crisis uncertainty and did not disappear (Chapter 32). Wider-but-continuous pricing compares favourably to the alternative on display in the bond market itself, which was frequently no price at all.

The asterisk

Here is what the industry's retelling usually omits, and this book's standards require:

The test ended with unprecedented intervention. The Fed announced it would buy corporate credit, including ETFs, and the market turned within days. We do not know how the machinery would have performed in a further two weeks of seizure without that backstop. The mechanism held under the conditions that occurred; those conditions included a central bank arriving.

The Fed's ETF purchases were themselves an endorsement with a complication. Buying ETFs was the fastest available route to broad credit exposure, which is a striking testament to the wrapper's utility. It also means the most-cited evidence of resilience is entangled with official support: the wrapper performed well in a market that knew help was coming.

High yield behaved worse than investment grade, as theory predicts: the less liquid the underlying, the larger the dislocation. The relationship between underlying liquidity and ETF discount held exactly as Chapter 71's ceiling argument implies.

Nothing here settles the systemic question. Three survivals are a track record, not a proof, and each stress episode has been different in character. Chapter 87 takes the debate up properly rather than treating this chapter as its conclusion.

What changed afterwards

• Institutional adoption accelerated sharply. Investors who had regarded bond ETFs as retail instruments watched them provide liquidity and pricing when nothing else did, and allocated accordingly. Bond ETF assets grew strongly in the following years.
• The price-discovery role became mainstream. Practitioners now routinely reference ETF prices as real-time credit indicators, a role that would have been dismissed a decade earlier.
• Regulatory attention shifted from "are ETFs fragile" toward the fund-liquidity questions affecting open-ended mutual funds, where the stale-NAV redemption dynamic transfers value from stayers to leavers. Several jurisdictions have since worked on swing pricing and liquidity management tools for exactly that reason.
• Evaluated pricing practices came under closer scrutiny, since the episode demonstrated how far estimates can drift from transactable reality.

A worked example

An insurance company needed to raise €200m from credit exposure on 18 March 2020.

• Route one, individual bonds: its dealers quoted a handful of its holdings, in small size, at prices 4–8 points below prior marks; several positions received no bid at all. Estimated time to liquidate: days, with uncertain execution.
• Route two, the bond ETF it also held: sold €200m in a single session, at a 3.4% discount to the published NAV and a spread roughly five times normal.

The ETF was the more expensive-looking option on paper and the only one that actually worked. Two weeks later, when NAVs had caught down to reality, the "discount" it paid turned out to be close to the market's true clearing level all along. What it purchased for that cost was certainty of execution on the day it needed it, which is precisely what liquidity means and precisely what was scarce.