The chart below tries to convince us that economists require more social skills than physicians or members of law enforcement:
A part of me is happy to take credit, of course; if, on paper, my job looks like something that requires a diverse skill set, then my wage is more likely to be high. Appearing to be in a desirable intersection of qualities or skills (occupations in the upper right quadrant) is surely beneficial.
But the more honest side of me is wondering whether regressions will end up more robust if I talk to them nicely.
Most of my output is not talking back to me, so most of my language-related skills1 may actually not be as essential as the classification above suggests.
Another reading of the data tells me that members of the police force need to be about as socially skilled as computer scientists. Is that convincing?
Sure, computer scientists will work in teams, and non-cognitive skills will be helpful for many other reasons. But a typical programmer will not avert a disaster thanks to her ability to read somebody's facial expression. People with deadly weapons attached to their belts probably need to be able to read faces accurately a bit more the guardians of our software.
Conclusion: reducing the number of dimensions of people's skills and classifying jobs by two attributes seems pretty hard. I don't have a better dataset, so this is not a criticism of the research referenced above - I am just observing that we get some puzzling results if we rank occupations by the amount of social skills they require.
An interview with David Autor on the limits of the "digital revolution"
Besides Stata and R, of course! ↩