Despite their reputation for over optimism, most banks appear to be far more pessimistic about their likely losses than turns out to be the case, reports The Telegraph.
A study of disclosures lenders have been required to make since the financial crisis shows most are more conservative about their likelihood of losing money than they need to be.
UK, European and Asian banks, on average, forecast losses nearly 30pc higher than those they actually faced, the survey by analysts at Barclays found.
Lloyds and HSBC predicted a default rate on their lending portfolios more than 50pc above what they actually experienced. Barclays was found to have been too pessimistic, particularly with assets in its investment bank where it forecast a default rate 78pc higher than in reality.
Royal Bank of Scotland appears to be one of the few banks that was too optimistic, recording a default rate more than double its predictions.
Asian banks appear to be the most conservative, seeing losses way above the actual level. In the case of Bank of China Hong Kong, the lender’s predicted default rate on its loan book was 76pc higher than the actual rate.
The Barclays analysts said the disconnect between predicted losses and default rates and what occurred was worrying given the implications for banks’ calculations of their risk-weighted assets (RWA).
“Most of the time banks’ PDs [predicted defaults] are lower than forecast, suggesting a degree of conservatism,” the analysts said. “The forecasting ‘errors’ can be massive, which raises questions over both their predictability and hence meaningfulness of the resulting RWAs. The banks don’t seem to be that good at forecasting their future, though definitive conclusions are a step too far.”
The research is based on the “Pillar III” disclosures banks have begun publishing since the crisis that contains detailed projections of what losses they expect to face. The figures are only available for European and Asian banks as US firms have yet to sign up to the regime. “Detailed Pillar 3 documents might run to over 100 pages but typically give tantilising partial answers to the question of forecasting accuracy,” said Barclays.