Robert Barro has been pushing recently the old idea that expectations of catastrophic events with low probability can matter substantially in a economy. For example, he demonstrates that this could explain the equity premium puzzle. What matters here is how to define a catastrophic event.
Initially, Barro defined a catastrophic event as a reduction of GDP by 15% from a previous peak. In more recent studies, he lowered that to 10% and is now looking at agregate consumption instead of GDP, which makes sense, as this is what really matters to people.
This makes me wonder: in recent years data has shown that economic fluctuations have considerably lost in amplitude (the "Great Moderation"). The origin of this is not clear, but one consequence may have been that people have let their guards down with respect to risk, i. e., they discounted more the likelihood of downturns or catastrophic events. This would be consistent with the reduction in the equity premium, and the rise in sub-prime loans.