Independent risks are manageable. Correlated risks can be catastrophic. The difference is structural, and most risk systems do not account for it.
The Correlation Problem
Risk management builds on the principle of diversification: the distribution of exposure across multiple independent risks reduces the probability that all risks materialise simultaneously. The portfolio of independent risks is more predictable than any individual risk, because the law of large numbers produces a distribution of outcomes that converges on the expected value as the number of independent risks increases. This principle works when the risks are genuinely independent — when the occurrence of one does not affect the probability of the others.
Correlated risks violate this principle. When risks share underlying drivers, the materialisation of one increases the probability of the others — not through mechanical linkage but through the same underlying condition affecting multiple risks simultaneously. The financial institution whose loan portfolio is diversified across many borrowers in the same geographic region has not diversified away correlation risk: a regional economic shock affects all borrowers simultaneously, producing losses that the diversification calculation assumed were independent but were not.
Where Correlation Risk Concentrates
Correlation risk concentrates wherever multiple apparently independent risks share an underlying driver that has not been identified as a common factor. The most consequential correlation risk structures are those where the correlation is invisible under normal conditions — where the risks genuinely are independent when conditions are stable and only become correlated when conditions are stressed. This stress-conditional correlation is the most dangerous form because it produces the worst correlation precisely at the moment when the risk management benefit of diversification is most needed.
The 2008 financial crisis was a stress-conditional correlation event: mortgage-backed securities whose underlying loans appeared to be diversified across different borrowers and geographies became highly correlated when the condition that had supported the entire housing market — the assumption that housing prices would continue to rise — reversed simultaneously across all markets. The diversification that appeared to exist under the prior condition did not exist under the stressed condition, which was the condition that mattered for the risk management calculation.
Diversification reduces the expected loss from independent risks. It provides no protection against correlated risks materialising simultaneously. The risk system that does not identify correlation structure has not managed the risk — it has accounted for it incorrectly, which is worse than not accounting for it at all.
Discussion