Crashes, defaults and such like should not be referred to as "tail events". I prefer "large loss events" or "large events". Even "rare events" is much better, although it brings out into the open the hidden assumption that events of this sort are rare, which could be controversial in some instances.
However, "tail events" is much worse. It makes the implicit claim that not only are these events rare (which is a factual and falsifiable statement), but that there is some probability distribution whose tails describe them. This is a statement that is in the worst scientific standing possible; it is not falsifiable by experiment, but there is good reason to believe that it is nonetheless false. The conditional likelihood of something like World War 2, the invention of the Internet or the 1970s oil shocks is not something that is determined by a single underlying data-generating process, and the fact that you can massage the data to kindasorta fit some kind of power law does not change this. Even though I am not as much of a fan of Paul Davidson as I was when I started this blog, some things are nonergodic; there are "crucial experiments" which change the nature of the world. There is no immutable underlying reality for these large events to be "tail
events" of.
In any case, it isn't a sensible way to think about big events as outliers from the standard mean-variance paradigm. They're historical, unique events with specific causes and effects. As I regularly find myself having to remind cadet risk managers with newly-minted PhDs in financial econometrics, the Great Depression did actually happen; it wasn't just a particularly innaccurate observation of the underlying 4% rate of return on equities.
<http://d-squareddigest.blogspot.com/2006/09/tail-events-phrase-considered-harmful.html>