The Quietest Risk on the Board
A credit crisis is, almost by definition, the risk that is invisible right up until it is not. This scenario tracks a genuine credit event, high-[yield](/glossary/dividend-yield) spreads widening above 600 [basis points](/glossary/basis-points), a [CDS index](/glossary/credit-default-swap-index) spike, or a major corporate or [sovereign default](/glossary/sovereign-default), that breaks normal correlation structures and forces a [flight to safety](/glossary/risk-on-risk-off). The defining feature of mid-2026 is that none of that is happening. High-[yield](/glossary/dividend-yield) spreads sit near 2.80%, close to the tightest levels of the cycle, and by every visible metric the credit market is calm.
Why Calm Is Not the Same as Safe
The tight spreads are exactly why the scenario stays on the board. A market priced [for](/metrics/fodsp) no stress has no cushion if stress arrives, and the sources of latent risk are well documented: roughly $1.7 trillion of private credit with limited mark-to-market discipline, corporate [leverage](/glossary/leverage) still elevated from the zero-rate era, and commercial-real-estate exposure concentrated at [regional banks](/metrics/kre). Credit crises are non-linear; they do not build gradually in the [spread](/glossary/bid-ask-spread) series, they gap. The compression today is the setup, not the all-clear.
The Transmission Line to Watch
The most likely trigger is not domestic. A [carry unwind](/glossary/fx-carry-unwind) driven by the Bank of [Japan](/metrics/ewj) scenario we track separately is the cleanest transmission mechanism: forced global [deleveraging](/glossary/deleveraging) that pulls [liquidity](/glossary/liquidity) out of credit and widens spreads mechanically, regardless of fundamentals. That linkage is why we watch the yen and the [VIX](/glossary/vix) as credit indicators, not just the [spread](/glossary/bid-ask-spread) itself.
Scenario probabilities are computed using a Bayesian log-odds model with calibrated base rates, z-score evidence weighting on first-differences, cross-metric correlation adjustment, and simultaneous coherence enforcement. Positioning reflects directional expected value under binary resolution assumptions. Full methodology and known limitations →