I sent an email to Professor Krantz mentioning the present discussion and asking him whether his numbers are "supported by published or rigorously gathered data, or your own observations, or are simply meant rhetorically". Here is his response, posted here with his permission.
Summary: he does not appear to have data at hand to support those numbers, but he believes they are accurate.
Your question is a good one.
Certainly my statement is well supported by my own personal
observation---after forty years in the business. But I have seen
sources even recently that say pretty much the same thing. I'm sorry
that I can't say what those sources are.
You can think about the matter this way. The vast majority of
academic jobs in this country are at what we call comprehensive
schools. And those are places where teaching is the thing.
Generally speaking, people don't do much research there. They have to
publish a paper or two to get tenure, and those papers tend to be
fairly close to the thesis. But then that's about it.
A lot of other people get jobs at the National Security Agency, or Los
Alamos, or another government think tank. Generally speaking,
publishing is not the thing there either. Sometimes people publish in
special classfied government journals.
Other people get jobs at Microsoft or Hewlett-Packard or what have
you. And publishing is not the thing there either.
Those of us who are lucky enough to be in math departments at true
research universities are definitely in the small minority.
I believe that, at the time I wrote those words, I consulted some
people at the American Mathematical Society and they had data to
support what I was saying.
Feel free to quote me to others if you wish.