According to a study of 7,259 mathematicians who graduated between 1900 and 1960, successful academics did a far better job of mentoring students during the first third of their careers than during the final third of their vocations.
The study, which was conducted by researchers at Northwestern University and published in the June issue of Nature, found that students who were mentored by mathematicians who were in the first third of their career went on to train 29 percent more students than expected, according to an article about the study published in The Christian Science Monitor. By contrast, students trained by mathematicians in the final third of their careers ended up training 31 percent fewer students than expected.
It's not clear whether any parallels could be drawn to the business world and whether, say, entry-level IT workers might be better off hitching their stars to up-and-coming IT professionals or counseled by an IT leader in the sunset years of their career. Or by a mid-career IT manager, for that matter.
This isn't intended as a knock against the Northwestern study, but I don't think you can apply the same logic to the mentoring of IT professionals and the impact it has had on their careers. For starters, the Northwestern study measures the percent of students who were eventually trained by those who were mentored. To measure the effectiveness of mentoring programs in IT, you'd need to examine other performance-related criteria such as advancement, changes in pay grade, etc.
Perhaps a more important gauge for IT professionals are some of the less tangible influences that a mentor has on a person's career, such as guidance around how to treat fellow workers, best practices in handling crises, etc.
Mentorship is very subjective. I bet if you were to select a few hundred IT professionals randomly who are in different stages of their careers, you'd wind up with a potpourri of anecdotes about early, mid and late-career mentors who had a dramatic influence on their careers. Still, it would be interesting to see what the results of such a study would reveal.