This is a lot more satisfying than just r > g, which although seeming to be well supported in a rather large book somehow is dissatisfying. Specifically, I was looking for some kind of agent model like what this author is presenting here. I've often thought about the "it's expensive to be poor" adage. What's interesting is he takes in this risk preference idea, in which you naturally expect the poor to have to be ced…
I don't find the model itself especially compelling as a broad model of the economy myself, though it's nicely written exposition of how variance can wipe people out and largely agrees with my biases. The fact that a curve with three arbitrarily-set parameters can be made to fit empirical Lorenz curves doesn't particularly impress me, and the concept of rich people enriching themselves through wiping out poor people in zero sum trades as opposed to having the ability to transact with a lot more people in positive sum trades doesn't ring especially true of a modern economy.