Here's a touchy matter. How much do you pay a PhD student? I've had work experience before embarking on my program, so I do think the "pay" is peanuts. What else could a reasonable person think? Naturally, as I was applying to a department called "Department of Decision Sciences", it would have been expected that I thought it through. And I did and I do not regret it (much). The money issue still hurts though. I'd like to address this issue briefly in the context of some good research on an analogous matter.
So how much should a PhD student be paid? I write with the assumption that PhD students are not cheap labour. I also write with the assumption that it is highly desirable, from the perspective of the institution PhD students are enrolled in, that PhD students "perform well" to boost the reputation of their school.
Clearly (to me at least) stipends should start low and increase with performance. It certainly should not start high or there will be many who would like to be eternal students (or punt "real work") will come swarming. So how might things increase? Research in contracting informs us that "first-best" outcomes in a "principal-agent R&D" scenario can be implemented using milestone payments. (And if I recall correctly, in one of the articles on the topic, the right thing happens "with probability 1" so we do not need to worry about issues of things being "implemented in Nash Equilibrium".)
So let us work with milestones. I will not talk about issues of risk aversion and paying (the students) a risk premium, and will instead be very concrete. The problem associated with paying PhD students is that their quality is observable only approximately and one cannot contract on effort, both of which combine to generate research outcomes. The intuition behind the claimed effectiveness of milestone payments is that at the achievement of each one, the level of uncertainty goes down, meaning schools need not be so hesitant about handing out rewards as they are more closely approximating "contracting on success".
The obvious milestone is passing the Qualifying Examination (and in big pharma there is FDA approval). At NUS, there is a $500 top-up to the stipend after that stage. Good. But there is nothing after that here. What about other milestones? For instance: first publication assessed by faculty to be of suitable quality, first publication in a "Tier 1" journal, thesis proposal defended, won student paper competition, etc. Prizes are something of "milestones", but with fuzzy requirements. My sense is that the right milestone-top-up structure would support good outcomes for PhD students and their host institutions.