Counterparty and exchange risk: the 2022 record

Counterparty risk is the risk that an entity owing something cannot deliver it. In this asset class the obligation is usually a balance shown in an account: a deposit at a trading venue, a lending product paying interest, or a position held with a broker. The 2022 sequence of failures is the most instructive record available, because several large firms failed within months of each other and the mechanisms overlapped almost completely.

What is actually held when a balance appears in an account

A balance at a centralized exchange is generally a claim on that firm rather than an asset segregated in the customer's name, unless a specific legal structure says otherwise. The distinction is invisible in the interface and decisive in a failure. Where deposit insurance exists for money held with a bank partner, it applies to that cash, not to the digital assets themselves, and not to the firm's ability to deliver them. The terms of service usually state what the firm may do with balances, including whether they may be lent or pledged. That document, rather than any dashboard, describes the actual exposure.

The 2022 sequence

In May 2022 the Terra network's algorithmic stablecoin UST lost its peg and collapsed together with the associated token LUNA. The design relied on a mint-and-burn relationship between two assets whose values were mutually reinforcing, so a decline in one undermined the mechanism intended to support the other, a reflexivity examined in the algorithmic stablecoin entry. The losses hit leveraged firms directly and indirectly. In June 2022 the lending platform Celsius suspended withdrawals and subsequently entered bankruptcy. Around the same period the hedge fund Three Arrows Capital failed, and because it had borrowed from many lenders at once, its collapse propagated to firms that had extended credit against collateral that turned out to be insufficient. In November 2022 FTX, then one of the largest trading venues, failed alongside its affiliated trading firm Alameda Research, and entered bankruptcy proceedings. Each event has its own detail, and the shared mechanisms are more useful than the narratives.

The mechanisms that repeated

• Affiliated-token collateral. Borrowing secured by a token that the borrower or a related party issued or largely held. Its market value falls precisely when the borrower is in trouble, so the collateral evaporates exactly when it is needed.
• Commingling. Customer assets pooled with firm assets, so a trading loss at the firm becomes a shortfall in customer balances.
• Rehypothecation. Customer assets pledged as collateral for the firm's own borrowing, which can create several claims on the same units.
• Maturity mismatch. Products redeemable on demand funded positions that were locked, illiquid or long-dated, so a normal rate of withdrawals became impossible to meet.
• Invisible aggregate leverage. No shared credit register existed, so several lenders each extended credit to the same borrower against overlapping collateral without any of them seeing the total.
• Correlated exposure. Nominally independent firms held the same assets, lent to the same counterparties and used the same venues, so one failure produced immediate losses at the others.

Two structural absences made the transmission faster than it would have been in a regulated credit market. There was no lender of last resort, so a firm facing a temporary funding gap had no facility to draw on and had to liquidate positions into a market that was already falling. And there was no resolution regime, meaning no supervised process for transferring client positions to a solvent institution while sorting out the failed firm's own book. In its place came an abrupt halt to withdrawals followed by a general insolvency proceeding, which converts a liquidity problem into a total loss of access for every customer at once.

Why the warning signs were weak

Some indicators did appear before each failure: withdrawal processing that slowed, a venue price persistently away from the rest of the market, unusual movements of assets between wallets, and public balance sheets in which an affiliated token accounted for an implausible share of stated assets. None was conclusive at the time, and several appeared only days ahead. Proof of reserves statements published in the aftermath addressed one half of the problem: they show assets, not liabilities. Without a verified liability figure, a reserve statement cannot demonstrate solvency, and a snapshot can be satisfied with borrowed assets. A reserve attestation against agreed procedures is narrower still than an audit. The workable conclusion is that solvency is not externally verifiable in this market, so exposure to any single firm is a decision made under irreducible uncertainty.

What a failure looks like afterward

An insolvency converts a balance into a bankruptcy claim. Several features of that process are unfamiliar to people whose experience is with brokerage accounts. Whether customers rank as unsecured creditors, and whether the assets are treated as property of the estate, has been contested in different cases and jurisdictions. Claims are commonly valued in fiat terms as of the filing date, so subsequent price movements do not accrue to the claimant. Distributions take years, arrive in tranches, and are usually partial. Claims themselves trade at a discount to face value in a secondary market, which reflects both the time value and the uncertainty of recovery.

What changed and what did not

Several structural responses followed. More firms publish reserve statements, more products use segregated custody or third-party settlement so assets need not sit at a venue during trading, and several jurisdictions accelerated licensing regimes covering custody and client asset segregation. The underlying condition did not change: obligations of a private firm remain unverifiable from outside, leverage remains invisible in aggregate, and leveraged positions concentrated in a few venues can still produce a liquidation cascade that transmits stress between otherwise unrelated participants. On this site, peg deviation and turnover are the closest available proxies for the market stress that accompanies these episodes.

Regulatory questions, taken up next, are largely the unresolved policy response to this record. The incidents record documents each failure, and the stablecoin section covers the Terra mechanism in detail.