I've just come from a discussion about who is best placed to give estimates on a given piece of work.
At a detailed level I'd always say that the best estimate comes from the person who actually has to do the work as they have the full understanding and this gives them complete buy-in to the , however at a higher level of abstraction (i.e. at overall project level) I'm not so sure.
I'm reminded of chapter 5 in Peopleware which gives the results of an Australian study from 1985 - best link I could find is here.
I'm particularly interested in your focus here - are you answering as a developer, architect, prject manager or whatever?
I'm answering this as both manager (present) and developer (past).
High level estimates should come from the team/project leader, but with input from the developers. They should also be given in the form of a range - most likely to worst case with an indication of the confidence level of each.
There's no way the team lead can know everything about the project in sufficient depth so they will need some input from the developers, but the danger with this is that you get bogged down in the details too early in the process. Equally, individual developers will not have a broad enough knowledge of the project (unless it's really small) to be able to give estimates on everything.
The manager then integrates these estimates and looks for conflicts and synergies to get the "big picture" - after all that's what we're paid for.
As a developer I wouldn't trust a manager who gave estimates without checking with the developers, but equally I wouldn't trust one who just asked the developers and passed that information on without "editing" it in some way.