The same data, the same map, different stories

Nice, but most visualizations don't take in account the scaling effects present in the data. Most social science systems are wildly non-linear.

IMHO, the best way is to try to model the data, then calculate the ratio between the predicted and the observed (real) data point. From there, you would have an index that account for the scaling in the system. The problem is that the model needs a good number of points and the data should not be too heteroscedastic. It's the hot new thing in scientometrics...

That is a nice post indeed, simple and to the point. I also enjoyed http://www.excelcharts.com/blog/creative-pie-charts-better-i... but that is a different genre of visualisation.