One of the problem with NASA (and other companies/agencies of the sort) is that projects take a very long time, at least 10 years and usually 20 years. You cannot fire the guy who officially did the estimate at the beginning: he is not on the project anymore, most probably not in the same company, maybe retired, and usually has grown to a position much more powerful.

Underestimating on the contrary leads you to not getting funding for the projects you are responsible for, and that is a nice way to not get a promotion.

As in Python, it is easier to ask for forgiveness than permission.