For both undergrad and grad school, I attended universities in the top portion of the various rankings, but earlier this fall I had been dating a PhD student in the humanities at a local university. When we first started dating, I decided to see out of curiosity where his program ranked in his field, and could not find the number. His program doesn't fund its students, give opportunities for TA'ing or RA'ing, or even give tuition remission, so he pays tuition and also works virtually full time. It's a world that I was totally unfamiliar with, and my friends are also unfamiliar with it, so I actually had to explain to my friends more than once that he not only doesn't get paid, but actually pays tuition.
Despite working full time, he hopes to finish his program in only a few years, and I'm sure he will. I'm certainly more impressed with his work ethic than my school mates'; even though he has less time to do his work, it looks to me like he makes the most of his time: definitely, he doesn't watch the paint dry or read phdcomics or pursue the other hobbies of graduate students.
It originally didn't make any sense to me why someone would pursue a PhD in the humanities at an obscure school; since top-rated humanities PhDs end up leaving academia, how could someone at a less-well-rated school expect to work anywhere above high school? A survey of employment in history from a few years ago finds that PhD programs are a relatively closed system, with professors at elite schools coming from elite schools, and those at non-elite schools coming from non-elite schools. We know there aren't enough jobs for all the PhD programs' graduates, though, and indeed about 50% of graduates of elite schools leave academia, but 25% of graduates of bottom-ranked schools stay in academia. In other words, graduates of elite schools could have probably obtained academic employment, but they chose to leave their fields rather than work at non-elite schools.
All graduate students lose income due to going to graduate school, and it's usually thought that they choose to go nonetheless because they are so interested in their subject that being able to pursue this interest compensates for the loss of income. If this is true, we wouldn't expect them to leave their fields if they have the option of academic employment. I have a few hypotheses for this:
1. You can eat prestige. Grad students at elite institutions receive compensation not just in the form of the academic basics, such as access to libraries and a scholarly community, but also from the prestige, facilities, resources, social networks, and other features unique to elite schools. Jobs which allow them to be academics, without access to the unique features of elite schools, do not compensate them enough: they instead choose jobs which pay more, and which may also offer them some of the features of graduate school, such as prestige and social network access. Grad students at non-elite schools have already displayed a preference for the academic basics over income.
Universities, at least my alma maters, like to perpetuate this story of falling in love with an academic subject, and these type of decisions --- leaving academia rather than take a suboptimal academic job --- suggest that people choose to go to graduate school for reasons other than their love for the subject. Or at least that this love is fickle.
2. Time management and perfectionism: many grad students at elite institutions are perfectionists to start out with, and their institutions may also encourage this trait by setting up the norm of the scholar without external distractions as the best conditions for good academic work, and may actually even give students funding to enable them to pursue their work without outside work. Grad students at non-elite institutions work more in addition to their studies, and learn throughout graduate school how to divide their time between income-earning and scholarly activity. Alternatively, grad students at elite institutions love their subject, but simultaneously are such perfectionists that they don't see a way to study it under non-ideal conditions.
3. Social norms and role model availability. Grad students use their own professors as models. Grad students at elite schools have faculty models who put much more work into research than teaching, and put a great deal of work into only 1-2 courses per semester and delegate a great deal to teaching and research assistants. Faculty at elite schools with low teaching loads and lots of assistants may be many times more productive than faculty at non-elite schools with higher teaching loads and few/no RAs and TAs. An academic job with high teaching load and little assistance may differ radically from the conception of a grad student from an elite school. The elite school PhD may feel like a job in the non-academic sector is more similar to their idea of a faculty job than an actual faculty job at a non-elite school. They also may have few role models who could illustrate the role of non-elite faculty for them.
4. Changing preferences. Grad students at elite schools could leave academia because their preferences have changed: many enter immediately after college, before learning their preferences for money. As they get older and think about children and other things that cost money, they may realize that their preference for money is higher than their preferences for academia, so we might expect a large fraction of graduates of elite programs in the humanities to leave academia no matter what the academic labor market looked like.