What is contagion in financial crises?

The short answer

The textbook definition of financial contagion focuses on direct linkages: when one institution fails, its counterparties suffer losses, which can in turn threaten their own solvency. Lehman Brothers in 2008 illustrated this — its derivatives book and short-term funding obligations transmitted shocks to AIG, money market funds, and bank dealers globally.

A second classical channel is information cascade: when a bank failure reveals industry-wide weaknesses, depositors and creditors of similar institutions reassess their exposure. The 1930s US bank failures spread through this mechanism in regional waves.

The 2023 episode introduced a third channel that doesn’t require direct linkages or fundamental information revelation. Markets identified Signature Bank and First Republic as having SVB-like profiles — high uninsured deposits, regional concentration, duration-risk exposure — and treated them as similarly fragile. Contagion operated through pattern recognition, not balance-sheet exposure.

→ New to systemic risk? Systemic fragilities pillar

What the data shows

Major contagion episodes by channel (Federal Reserve, BIS, IMF):

• 1998 LTCM: counterparty channel. Long-Term Capital Management’s leveraged positions threatened 14 major banks via direct exposures, requiring a Fed-coordinated rescue of $3.6bn
• 2008 Lehman: counterparty + information channel. The bankruptcy on September 15 froze short-term funding markets globally, triggering money-market fund losses and AIG’s $182bn rescue within weeks
• 2010-2012 Eurozone: sovereign-bank doom loop. Greek default risk transmitted to Italian and Spanish banks holding sovereign debt, forcing ECB intervention via OMT in 2012
• March 2023 US regional banks: perception channel. Signature ($110bn assets) closed March 12 within 48 hours of SVB; First Republic (~$229bn) sold to JPMorgan May 1, 2023 after weeks of accelerating outflows
• March 2023 Credit Suisse: international perception spillover. Swiss authorities forced UBS merger March 19, 2023, with AT1 holders wiped out — partly attributed to global stress narrative emerging from US events

The exception that complicates the pattern: not all banks with similar profiles to SVB experienced runs. PacWest, Western Alliance, and others stabilized after initial pressure, indicating that perception contagion is contingent rather than mechanical.

→ Dataset: Credit spreads dataset

Why it happens — the macro mechanism

Contagion operates through three distinct channels that often combine in real episodes.

The counterparty channel. Direct exposures through derivatives, repo, and unsecured lending mean that one institution’s failure imposes losses on others. The 2008 Lehman case illustrates this — its $639bn balance sheet was interconnected with the entire global banking system. Modern central clearing reduces but does not eliminate this channel.

The information cascade channel. A failure can reveal industry-wide problems. The 1930s US deposit runs spread through this channel — depositors at solvent banks reassessed risk based on the failure of similar institutions in their region. Bagehot’s 1873 classic Lombard Street already described this mechanism. Bank runs explained details the underlying coordination logic.

Bridging point: the third channel emerged distinctively in 2023 and is harder to model than the first two.

The perception contagion channel. Markets identify a shared fragility pattern across institutions — high uninsured deposits, duration mismatch, sectoral concentration — and treat them as a single risk class. Signature and First Republic had no direct financial relationship with SVB but shared structural attributes that made them perceived equivalents in a stress narrative. Contrary to the textbook framing where contagion requires fundamental linkages, perception contagion can spread through pattern matching alone. This is the angle distinctive that classical contagion models did not capture.

Synthesis by regime: in the pre-2008 regime, contagion was primarily counterparty-based, and risk was reduced by limiting individual exposures. Post-2008, central clearing and capital buffers reduced direct counterparty risk while regulatory transparency increased information cascade risk. The 2023 perception contagion regime is enabled by social media coordination and detailed public balance-sheet disclosures — markets now identify peer institutions almost instantly when a stress event hits.

The 2023 contagion mechanism does not need counterparty links — it needs only a recognizable fragility pattern that markets can apply across institutions in minutes.

→ Framework: Systemic fragilities

What it means for different economic actors

Bank shareholders. Equity prices in 2023 moved on perceived peer-group attributes rather than individual-bank fundamentals during the stress weeks. Identifying a bank’s pattern signature — uninsured deposit ratio, HTM losses, sectoral focus — became more relevant than its standalone metrics for short-horizon trading.

Bank creditors and uninsured depositors. The realization that perception contagion can hit any bank with a recognizable fragility profile changed treasury management practices. Diversification across banks with different profile signatures (universal vs regional, retail vs corporate, etc.) became a more explicit risk-management criterion.

Regulators. The 2023 episode demonstrated that supervisory frameworks designed for counterparty contagion were insufficient against perception-based spread. Stress tests have begun incorporating cross-bank perception scenarios, and resolution playbooks now address rapid spillover within days rather than weeks.

A common error is to assume that institutions with no direct linkages are protected from contagion. The 2023 evidence is that pattern matching by markets, social-media-amplified narratives, and uninsured deposit mobility can transmit stress across institutions whose only commonality is a structural profile.

Practical observation

Go deeper

Related questions

• How did the 2023 regional bank crisis differ from 2008?
• What causes a bank run and why are they self-fulfilling?
• Why did SVB fail so quickly in March 2023?
• What are too-big-to-fail banks and the moral hazard?

Frequently asked questions

How does contagion differ from a general market sell-off?

A general sell-off reflects broad risk reassessment driven by macro factors. Contagion is more specific — it transmits stress from a defined source institution or asset class to others through identifiable channels. The 2008 Lehman week showed both: the initial bankruptcy was contagion through counterparty channels, while the subsequent market panic was a general flight to safety. These episodes are arranged by regime over time in the timeline linking each crisis to its regime. The two phenomena often combine but operate through different mechanisms.

What is the angle distinctive about perception contagion in 2023?

Classical contagion models — Allen-Gale 2000, Diamond-Rajan 2005, Acemoglu-Ozdaglar-Tahbaz-Salehi 2015 — focus on network linkages between institutions. The 2023 episode showed that contagion can spread without any network linkage, purely through pattern matching by sophisticated market participants. Signature and First Republic had no balance-sheet exposure to SVB; what they shared was an uninsured-deposit profile that markets recognized within hours. This requires updating the theoretical framework to incorporate observational similarity as a contagion channel.

Has financial regulation reduced contagion risk over time?

For some channels, yes. Central clearing of derivatives reduced counterparty contagion. Higher capital and liquidity requirements made individual banks more resilient. However, the 2023 episode demonstrated that new contagion channels can emerge from changes in market structure — social-media coordination, public balance-sheet disclosure, and uninsured-deposit mobility have created perception-contagion paths that did not exist a decade ago. Regulation has reduced some risks while structural change has introduced others.

Last updated — 2 June 2026