Using correlation to model op risk losses may be unsafe – study

Techniques for linking economic factors and bank losses produce varying – and sometimes contradictory – results

Banks cannot rely on assumptions that operational risks are correlated with economic factors, casting doubt on the loss projections used in bank stress tests, new research suggests.

Operational risk predictions are a key part of regulatory stress-testing, but “wide error margins” in the correlations mean the forecasts may be flawed, according to a paper by Peter Mitic, head of operational risk methodology for UK at Santander. The paper was published in the Journal of Operational Risk in June.