[greenyoda]

"But secondly, it is claimed that he cherry-picks. Economists (we) do that. A lot. And it has to be called out."

The total effect of his cherry picking can't even be discerned from this article. With a lot more work (more than a single blogger probably has time for), it might be shown to be much worse. For example, Piketty chose the U.K., France and Sweden for his list of European countries (averaging them all into "Europe" without adjusting for their populations, as the article pointed out). But why those particular countries? Would the results have been completely different if he would have also included Germany, one of the wealthiest and most populous European countries? We'll probably never know how he came up with his set of countries, and whether he deliberately omitted other countries whose data was available because they would have caused his results to be not significant enough to publish (or even contradicted his thesis).

This reminded me a lot of one of the original studies linking cholesterol to heart disease, which chose a carefully picked subset of European countries even though results were available for several others. Subsequent research showed that including the data from the missing countries would have yielded very different results.

[thanatropism]

Related, useful concept:

https://en.wikipedia.org/wiki/Simpson%27s_paradox

See in particular the Berkeley Gender Bias case if you don't want to hack the math apart.

The specific way you slice and dice data matters.

[Iv]

> But why those particular countries

Because they had coherent archives of several proxies that Piketty uses for a long period of time. He opened the door to people doing the same work he did and contradicting him with other countries, but I would not bash him for not doing the same work on every country on the planet.