Richard Florida weighs in.

The returns to analytical skill rise consistently across the skill distribution; moving from the 25th to the 75th percentile increases earnings by more than $25,000. The same basic pattern holds for social intelligence skills, like teamwork, communication, people management, and so forth; moving from the 25th to the 75th percentile increases wages by nearly $35,000. But physical skill is different, registering decreasing returns. Moving from the 25th to the 75th percentile lowers earnings by $13,600.

Thanks to Alex Tabarrok for the pointer.

My interpretation of the graphs is that the correlation between occupational wages and occupational requirements for analytical skills is strongly positive, as is the correlation between occupational wages and occupational requirements for social skills. However, the correlation between occupational wages and occupational requirements for physical skills is weaker and is negative.

This is an instance in which bivariate correlations are difficult to interpret. I think the dataset cries out for a multiple regression of occupational wages on the three skill components. I would hope that with multiple regression the correlation between wages and required physical skills would be positive, controlling for everything else. The negative correlation would be difficult to explain on an all-else-equal basis.

It would be interesting also to see how these relationships have changed over time.