How does default correlation affect the value of different CDO tranches, and why do equity and senior tranches respond in opposite directions?

Default correlation is the single most important parameter in CDO pricing, and its effect on different tranches is indeed counterintuitive until you understand the loss distribution mechanics.\n\nThe Key Insight:\n\nHigher default correlation doesn't change the expected number of defaults, but it changes the distribution of defaults. It makes extreme outcomes (very few defaults OR very many defaults) more likely, while reducing the probability of moderate default outcomes.\n\n`mermaid\ngraph TD\n A[\"Default Correlation\"] --> B{\"Increase Correlation\"}\n B --> C[\"More probability mass
in the tails\"]\n B --> D[\"Less probability mass
in the middle\"]\n C --> E[\"More likely: 0 defaults
(equity tranche safe)\"]\n C --> F[\"More likely: many defaults
(senior tranche hit)\"]\n D --> G[\"Less likely: moderate defaults
(only equity tranche hit)\"]\n E --> H[\"Equity tranche value RISES\"]\n F --> I[\"Senior tranche value FALLS\"]\n G --> H\n`\n\nNumerical Example:\n\nHarbourside CDO: 100 reference names, each with 3% annual default probability.\nCapital structure: Equity 0-5%, Mezzanine 5-15%, Senior 15-30%, Super Senior 30-100%.\n\nLow Correlation (rho = 0.10):\nDefaults cluster around the expected value (3 defaults). Distribution is tight.\n- P(0-5 defaults) = 82% -> equity tranche absorbs losses most of the time\n- P(>15 defaults) = 0.1% -> senior tranche almost never touched\n- Equity tranche expected loss: high (~60% of notional)\n- Senior tranche expected loss: negligible (~0.05%)\n\nHigh Correlation (rho = 0.50):\nDefaults are either very few or very many. Distribution is bimodal.\n- P(0-1 defaults) = 35% -> equity tranche survives entirely more often\n- P(>15 defaults) = 8% -> senior tranche gets hit meaningfully\n- Equity tranche expected loss: lower (~40% of notional)\n- Senior tranche expected loss: higher (~3.5%)\n\nImpact Summary:\n\n| Tranche | Low Correlation | High Correlation | Effect |\n|---|---|---|---|\n| Equity (0-5%) | Expected loss 60% | Expected loss 40% | Benefits (value rises) |\n| Mezzanine (5-15%) | Expected loss 8% | Expected loss 10% | Ambiguous (depends on attachment) |\n| Senior (15-30%) | Expected loss 0.05% | Expected loss 3.5% | Harmed (value falls) |\n| Super Senior (30-100%) | Expected loss ~0% | Expected loss 0.8% | Harmed (value falls) |\n\nTrading Implication -- Correlation Trading:\n\nPre-2008 correlation desks at Thornhill Capital and similar banks ran \"correlation books\" by:\n- Buying equity tranche protection (short correlation)\n- Selling senior tranche protection (long correlation)\n- Net position: short correlation -- profits when correlation decreases\n\nThe reverse (\"long correlation\") trade benefits when systemic risk increases.\n\nWhy This Matters for Risk Management:\nCorrelation is notoriously hard to estimate and unstable through time. During the 2008 crisis, correlations spiked dramatically, devastating senior tranche holders who had assumed near-zero loss probabilities based on historical correlations.\n\nStudy structured credit risk in our FRM resources.