Bob Martin's new book "Clean Agile" covers this topic well. The problem isn't dates or deadlines, it's arbitrary dates or deadlines that are not backed up by data that indicates whether or not they are achievable. Often we create estimates with no prior knowledge of the team composition and particularly their ability to apply their skills to the problem at hand. The only way to make good estimates is to give them work, let them start doing it, and observe their progress. You then can manipulate the scope of your project to determine what you want done (the highest value things) and estimate when you can get those things. The alternative levers (add a bunch of people, sacrifice quality) end up not working in software projects and become the "value destroying mistakes."