And, moreover, uses some of Milton Friedman’s best work to do so.

First, Joe [Stiglitz] offers a version of the “underconsumption” hypothesis, basically that the rich spend too little of their income. This hypothesis has a long history — but it also has well-known theoretical and empirical problems.

It’s true that at any given point in time the rich have much higher savings rates than the poor. Since Milton Friedman, however, we’ve know that this fact is to an important degree a sort of statistical illusion. Consumer spending tends to reflect expected income over an extended period. If you take a sample of people with high incomes, you will disproportionally include people who are having an especially good year, and will therefore be saving a lot; correspondingly, a sample of people with low incomes will include many having a particularly bad year, and hence living off savings. So the cross-sectional evidence on saving doesn’t tell you that a sustained higher concentration of incomes at the top will lead to higher savings; it really tells you nothing at all about what will happen.

This is from Paul Krugman, “Inequality and Recovery.”

Three comments:
1. By “rich,” it’s clear, in context, that Krugman means “high-income.” Those aren’t the same. You can be high-income and have relatively modest wealth.
2. Krugman states Friedman’s thesis correctly. Friedman laid this out in his 1957 book, Theory of the Consumption Function. It’s Friedman’s first major empirical work and one of his best. It also seems to be broadly accepted among economists. Friedman has an interesting discussion of it on pp. 222-227 of his and Rose Friedman’s book, Two Lucky People. I remember when Armen Alchian, in my first-quarter microecon class at UCLA in the fall of 1972, discussed it. Alchian, a friend and strong admirer of Milton, said, “Friedman is a first-rate economist and a second-rate statistician. And even a second-rate statistician will see that in the consumption/income data, you have regression to the mean.” He said this, by the way, with total admiration.
3. Acceptance of Friedman’s idea has strong implications for Keynesian fiscal policy. It implies that if the government temporarily boosts people’s incomes, either with a temporary tax cut or an increase in transfer payments, people will not spend anywhere near a dollar per dollar of tax cut or transfer payment, but will spend more like 20 to 30 cents per dollar. That strongly undercuts the multiplier part of Keynesian fiscal policy. Does anyone know if Krugman has used this reasoning in his discussion and advocacy of Keynesian fiscal policy? I don’t recall his having done so.