The data is derived empirically; that r > g has been the historical norm since the industrial revolution, with the notable exception of the period ~1910-1970.
If that fact changes then the entire premise of the book is wrong. But the thesis of the book is grounded in historical evidence (this is how the world works), not any particular theory that answers "why."
r > g has been the historical norm since the industrial revolution, with the notable exception of the period ~1910-1970
Actually, the data says quite a bit more than that. I'm looking at the chart shown in Krugman's review (Figure 1, a bit more than halfway down the page):
First, this chart says that r > g has been the historical norm since antiquity, not just since the industrial revolution. The first data point on the chart is for years "0 - 1000". I'm assuming that Piketty's book gives some explanation of how he figured out the rate of return on capital during the Dark Ages. :)
Second, the chart says that the difference between r and g decreased all through the period studied, up to the "1913 - 1950" data point, when r < g became true for the first time. Then, up through the "1950 - 2012" data point, r and g both rose sharply, but r < g was still true, and the difference between the two was more or less the same.
Third, the chart says that, for the rest of the 21st century, not only will r > g be true again, but r will continue to rise while g will fall sharply--which has never happened before in history. So I think it's wrong to claim that Piketty's prediction for the 21st century is just a prediction of things "returning to the historical norm". It's not--it's a prediction of a drastic change from the historical norm.
If that fact changes then the entire premise of the book is wrong. But the thesis of the book is grounded in historical evidence (this is how the world works), not any particular theory that answers "why."