"I think that we have to be constantly asking ourselves, 'How do we calculate the risk?' And sometimes we don't calculate it correctly; we either overstate it or understate it."