I like Oogie's formula too. Mine is simpler - approximate time I think I need for the task, multiplied by 3. Having run engineering teams in a very hectic environment, it worked for me well.
I think, the trick is to have a deadline for each task. If the task is not critical to finish in 7-10 days, put it aside and re-visit in weekly review. If you have a deadline for each task, your boss will not need to ask.
If you discover that it will take longer, re-negotiate the deadline with your boss. I do risk management for projects, but it is overkill for tasks. I also apply "good enough" rule to any task: what is the ultimate outcome I want to achieve and what is the minimum realistic requirements to get it done. I can do research for 3 days or 3 hours. The question is, how can we do it well to ensure quality, but not get carried away?