> get paid well and work on what they're interested in This is a common reason in pure math and theoretical computer science. There are not a whole lot of plac

> get paid well and work on what they're interested in This is a common reason in pure math and theoretical computer science. There are not a whole lot of places that will give you a good job to do basic research. Tenured professor is another one, but hard to come by, and has a different set of tradeoffs. If you do certain kinds of applied research (e.g. machine translation) you can do it at Google, but Google is not tripping over itself to hire number theorists to do actual number-theory work (maybe they would hire a number theorist, but to do something else).