I think that it is possible to give quick and accurate estimations for in-my-area-of-expertise tasks.
I agree that we cannot give considered estimations in the "desperately understaffed" environment where everybody tries to keep his head above the moving sands around. ;-)
Yep. And, just to expand, in a multi-multi-multitasking environment, realistic estimates and intuitive estimates diverge wildly.
For example, there's Gerald Weinberg's rule of thumb that says that if you have five programming projects, you're losing 75% of your time in multitasking, and each project is getting five percent of actual time.
So in a forty hour week, each project will get the equivalent of two hours--IF you only have five projects, and how many people have ony five projects?
Of course, you could work more than forty hours, and many people do, but if you work eighty hours, odds are that each project will get, oh, two and a half hours of work-equivalent, because productivity drops off a cliff after roughly the sixth hour of work each day. In fact, they might go down to one and a half hours of work-equivalent, because true productivity, not just productivity per hour, drops with too many hours. There's evidence that people get more done in a thirty hour week than in a forty hour week.
If just creating that estimate will call for four hours of solid work, then according to that rule of thumb, that estimate will take that project's allotment for two full weeks. And that's the ONLY thing that will get done for that project. And then the estimate itself should account for the project getting one day's worth of work per month.
That sounds...ridiculous. Simply ridiculous. Just ludicrous.
And then it's two weeks later, and you've just barely gotten the estimate done, and you realize it's not ridiculous, and you want to crawl under your desk and sleep the sleep of the depressed for the rest of the quarter.