Combining data with political analysis

I just came across Duncan Green´s entry on inequality from a couple of years ago.  I like his argument that we need more on the politics of inequality reduction and his research project proposal (has he or his group advanced on it?  Does anybody knows?)  Yet one thing he does not acknowledge is the need to have more interaction between data and politics to make sure we are discussing real life.  His example of Brazil is paradigmatic in this case: Brazil in the 2000s would probably be a cae study in his inequality project (it has been a case study of inequality reduction in many other studies and books, including Why Nations Fail)... but maybe this would be a mistake.  I am afraid that unless we know more about the rich, we will know relatively little about inequality... but it is just too hard to get this information. What is the best way forward?