Portfolio correlation, concentration, and the HHI index
Portfolio correlation, concentration, and the HHI index
A credit portfolio's real risk isn't the sum of its individual deals' risks. It's how those deals move together. Two portfolios can hold identical assets with identical expected losses and still carry completely different odds of a catastrophic quarter, and the difference is correlation and concentration.
How it works
Concentration measures how much of a portfolio's exposure sits in any single deal, originator, deal type, or geography. A portfolio with 40% of its capital in one originator is concentrated, regardless of how good that originator looks today, because a single failure there takes down a large fraction of the whole book at once. Diversification is the direct antidote: spread exposure across many independent positions so no single failure is fatal.
Correlation measures whether positions tend to fail together. Two consumer lending pools originated by different platforms but exposed to the same underlying population, subprime borrowers in one regional economy, aren't really independent; a local recession hits both at once. Correlation is what turns a collection of individually modest risks into a single large one, because correlated defaults cluster in the same period rather than averaging out across time.
The Herfindahl-Hirschman Index (HHI) puts a single number on concentration. Sum the squared percentage shares of each position in the portfolio: a portfolio split evenly across 10 equal positions scores 1,000, while one with a single dominant position scores far higher, approaching 10,000 for a fully concentrated book. The squaring is the key: it penalizes large positions disproportionately, so HHI rises sharply as any single exposure grows relative to the rest. Regulators use HHI to measure market concentration; credit portfolios use the same math to measure exposure concentration.
Why it matters
Concentration and correlation are what drive Unexpected Loss, the tail-risk figure capital reserves are actually sized against, rather than Expected Loss. A well-diversified, low-correlation portfolio can carry the same average expected loss as a concentrated one while being far less likely to blow through its capital cushion in a bad period, because its losses are statistically less likely to all arrive at once.
HHI gives a portfolio manager a running, quantified read on whether diversification is real or illusory. A book can look diversified by number of positions while being dangerously concentrated by originator or deal type, and a rising HHI on any of those dimensions is an early signal that a single failure is becoming disproportionately consequential, before it actually happens.
Where this shows up in Rekord
Rekord tracks concentration limits by originator, deal type, and geography, using HHI-style measures to keep any single dimension from dominating the portfolio's risk. For the underlying loss math these limits protect, see Credit risk math: PD, LGD, EAD, Expected Loss, Unexpected Loss.