To be clear (and I realize this wasn't quite clear), I wasn't talking about pure monetary ROI, but rather "return" in terms of satisfaction of your preferences.
I.e., a job paying a lot of money is "return" on investment; but so is a job that pays a lot of experience, or satisfaction. Meanwhile, a job that takes a lot of time to do requires a larger "investment" (in terms of the set of BATNA jobs you could otherwise be doing with your skills); but so do jobs that don't take very long, but are more physically demanding/emotionally draining, or where the client keeps jerking you around by changing the requirements. If working the job makes your utility go up (i.e. makes your life better), it's a positive-ROI job. If it makes your life worse, it's a negative-ROI job.
So, let me restate: in some verticals, while the clients do pay a livable time-and-materials wage for the region and experience level... they expect far more labor (and far more highly-skilled labor) for their dollar than that dollar "should" be able to get them, such that the net utility from anyone taking that job is negative, even if they're "making money."
These businesses get by mostly on exploiting young IT ingenues who don't yet realize their worth, and so are willing to give a try to e.g. "program a social-networking app" for $2000; or to e.g. "set up a PXE Linux automatic memory-testing boot env and Windows system-image deployment pipeline" for $300.
(These are both things I naively got myself into when I was younger. Amusingly, I was in over my head on the social-networking app, but managed to complete the hardware qualification+deployment env just fine. And my only major obstacle for the social-networking app, was that the client insisted that it needed to be hosted on some free ISP hosting they had, and thus needed to be written in PHP4. Which I didn't even complain about... it was just a language I wasn't very familiar with, so I tried to learn it while working. Sigh.)
In other words, these companies try to get newbies to do senior-level work, paying only newbie costs. Sometimes, surprisingly, that works! And they survive entirely because it sometimes works.
It's basically the "my nephew knows computers, he can certainly repair mine" mindset applied to corporate planning.
if you do, you're just selling dollar bill for 80c. You may drink growth-kool-aid, or someday-monopoly-hope or VC-subsidizes-business.
In the end, somebody pays for it either from stupidity or hope.