[jonstewart]

Yeah, it'd be nice to see related stats. In the absence of them, I can only speak for the interviewing and hiring I've done, i.e., my own bias. That bias generally is that with a Ph.D.-no-work-experience-hire, I'll have just hired someone who's extraordinarily knowledgeable and capable in some aspects of job performance (research, algorithm analysis, etc.) and almost helpless in others (teamwork, version control, reading code they didn't write, writing code to be maintainable over time, letting well enough alone). Such wild disparity in skill levels can be hard to manage.

(I do think colleges are generally getting better at making sure students have some of these skills, and GitHub's popularity has also helped. So, hopefully this bias becomes obsolete.)

Someone who starts down the Ph.D. path, then says, "you know what? that's not for me right now", has just demonstrated some uncommon wisdom and self-awareness.