Now, on to my third problem: I think Angus Maddison may be doing things wrong. I realize this is a rather presumptuous thing to say, but I think it's true. Specifically, the assumption that GDP before 1700 was proportional to agricultural productivity seems to me not to be a good one. The reason is that even in a non-industrial society, there is a potentially huge source of GDP increases: trade. Remember, in a world where output is mostly in the form of commodities (i.e. no increasing returns to scale), the old Ricardian theory of trade makes a lot of sense. Stable ancient empires that could act as free trade zones were probably capable of dramatically increasing their per capita GDP beyond the base provided by the productivity of their land. This is the finding of Ian Morris in Why the West Rules For Now. He constructs a "social development index" that includes things like urbanization and military capabilities, and probably correlates with an ancient region's per capita GDP (it is hard to build cities and make war without producing stuff). He finds dramatic changes in this social development index over the course of the Roman Empire; at its height, Rome seems to have been extremely rich, but a couple centuries earlier or later it was desperately poor. Morris corroborates this index with data on shipwrecks, lead poisoning, and other things that would tend to correlate with output. Basically, Rome saw huge fluctuations in per capita GDP. But it is unlikely that Rome's agricultural productivity changed much over this time. Instead, what probably happened was the rise and fall of cross-Mediterranean trade. If trade could make Rome dramatically richer, and its absence could make Rome dramatically poorer, then Maddison's data set is wrong. Just because most people in 100 AD were farmers does not mean that most people were subsistence farmers. And frankly, I'm not sure how people use Maddison's data set without noticing this fact.