The way that the success or failure of a graduate student's scientific career is in the hands of a very small number of people without oversight has no parallel

The way that the success or failure of a graduate student's scientific career is in the hands of a very small number of people without oversight has no parallel in private industry. If you're lucky and spot trouble early on you can maybe get a new project/advisor, but you can't change graduate programs. You're committed for 4-6 (or 7 or 8) years. Contrast this with industry, where you can leave an abusive job after a year or two and still have a very productive career. Even in the worst case that an evil manager fires you from your first job, you can get another. You probably will. In academia, if your halo ever loses its shine you can never get it back. There are no do-overs or next-times when it comes to your scientific reputation. Also, don't underestimate the importance of money. Materially speaking, the opportunity cost amounts to many hundreds of thousands of dollars. And in terms of psychology, only people who've done it know how much it sucks to still be making $25k/yr in your late 20's. It's worse than ever now that we have Facebook as a mirror to show us where we could have been if we'd chosen industry instead.