For many organizations, we need to be able to forecast. But, as Ron Jeffries says:
“Yes, estimation is fraught. It is inaccurate, and politically dangerous. But we do have some knowledge and the project deserves to have it.”
Estimates are not evil in and of themselves. But the results are often used for evil. Dilbert summarizes the traditional approach and the way we feel about it. Problems include:
Estimates often become commitments and / or targets. People tend to treat these numbers as factual true data points instead of the probabilistic statements they are, with poor results.
Estimates take a long time. Because we know the estimate is about to become a commitment, we almost do a complete design of the system in order to come up with an estimate.
Estimates are wrong. Even after spending that amount of time on it, the estimates are wrong. And you'll find that the more time you spend on an estimate, the worse it becomes, mainly because you are building your estimate on assumption over assumption over assumption.
Estimates are done by one group, or worse, by a single person. And so do not reflect the total view of what is required to do the work.
The agile approach to estimation is aimed at improving these outcomes.