It would be interesting to see these results cross-checked, by applying the methods from one job to another - for example by applying the methods used for law and engineering (Cross-country regression of GDP on <job> per capita) to the other fields listed.

Personally I find the results presented surprising - especially the claim of research being worth 25% of GDP, which seems very high to me.