Entire fields (e.g., psychometrics) are devoted to finding ways to accurately measure "skills", so it is also disconcerting that they devote all of three sentences to explaining how the cognitive and non-cognitive skills are measured.

The "non-cognitive skills" come from a factor analysis of a hodgepodge of behavioral ratings ("Cheats or tells lies", "Too dependent on others"), none of which seem similar to what I'd call a skill. It is not clear to me why the loadings on these factors vary across years: if you think that not depending on others is a non-cognitive skill, surely it's the same skill in 1997 and 2002.