Somebody delivering the same service, but including HTML/CSS, could easily charge $150 and have more customers than they could possibly handle.
That said, $15-20 is comfortably in that "If it doesn't work out, who cares" bracket. The problem will be that for many people, time is far more crucial than the cost. If I need a turnaround of 3-4 days, and the queue is 7+ days, it won't matter if it's $15 or 15 cents.
Don't compete on the same turf as elance types. I've used PSD2XHTML type services for $150-250 before and the main reasons I stopped was that the product wasn't really done to our house style and they couldn't hit deadlines. (Now I have staff that do these jobs for me.)
In any case I like the idea behind the $15 blind design since designers tend to suffer from "I like it but..." with indefinite tweaks and last minute changes. I wouldn't be surprised if customers end up going through several iterations...which would still be inexpensive in this case.
Not that he needs to use any of the $15 work directly in his portfolio -- but if he hashes out lots of designs in a creative way, he's going to end up with some obvious winners that he can polish up.