STARTING TOO YOUNG? Voluntary Xchange finds the critical determinant of Ph.D. completion.

When I directed a Ph.D. program, I went back and got all our files on every student who'd ever been in the program, and did some regressions of whether or not they graduated (I also looked at other dependent variables that interested me because of the way our program was evaluated by my administration).

The message was that virtually nothing had any predictive power - GRE scores, GPA, degrees, majors, native language, and so on.

But, there was one variable, and its square, that were highly significant. In fact, with these two in the regression, no other variable was even close to being significant. The variable was calender age (at the start of the first year in graduate school). Age was positively related to finishing, and the square of age was negatively related to finishing. The fitted values of the regression suggested a probability of graduating with a Ph.D. that peaked quite strongly if you were 28 when you entered the Ph.D. program. The "sweet spot" around that peak was fairly small - about 3 years either way.

I don't seriously think that age was the true determinant; I'm quite sure it was proxying for something else.

For what it's worth, both Professor Tufte and Professor Karlson entered Ph.D. programs at the age of 21 (in my case, just turned 21.) I finished in 4 1/2 years.