The great egalitarian political philosopher John Rawls wrote that he preferred the idea of a consumption tax to an income tax "since it imposes a levy according to how much a person takes out of the common store of goods and not according to how much he contributes." In response to a very similar argument from Scott Sumner, a smart Steve Roth post replies that financial saving is not the same as saving real resources:
This makes absolutely no sense. If you forego a massage this week, or wait a few years to get your house painted, is the labor for that massage or paint job “saved”? How about this year’s sunlight — the ultimate source of that labor power? Can you use it next week, or next year? Understand: services comprise 80% of U.S. GDP. And that’s before you even think about Apple and similar, with their just-in-time, on-demand supply chains — when you buy it, and only when you buy it, they produce it.
If you don’t buy it, it doesn’t get produced.
I think that to understand the Sumner/Rawls view you have to remember that both are assuming that the economy is always operating at full employment. Rawls doesn't specifically say anything about this, but it's the only way to make his viewpoint make sense. Sumner writes extensively about business-cycle issues, however. One of his main themes is that a competent central bank can always guarantee full employment and that it always should guarantee full employment in part because the full employment macroeconomy of steady national GDP growth is one in which all these nifty neoclassical ideas actually work. So the way the story goes is that the people thrown out of work when you switch from consuming to saving will be reemployed in the production of capital goods. We all become thriftier, stop dining out so much and start cooking at home, and all those unemployment cashiers and waitresses get jobs building houses and manufacturing tractors. Thus society's stock of capital goods does in fact increase through a big shift toward a higher savings rate.
Something to note, however, is that investment is by far the most cyclical element of the economy. A recession, in other words, is basically a sharp shortfall of investment. That's why "normally" interest rate cuts can cure recessions (low rates make investment cheaper) and also more fundamentally why expectations matter so much. The upshot of this is that if we were living in SumnerWorld and had no recessions, that alone would massively increase the long-term capital stock regardless of the tax rate. The main benefit of the Sumner/Rawls program of full employment and consumption taxation, in other words, comes from the full employment. But I do think the theory is valid.
The much bigger problem with the entire line of argument focused on the accumulation of physical capital is that it's left the production of human beings off the table (the Chamley/Judd theorem on optimal taxation, for example, assumes that people are never born and never die but just exist forever in delightful equilibrium). In practice, one of the main ways by which a middle-class American family is likely to exhibit thrift is by having one child rather than two or two children rather than three. Now obviously it's true that a tax code aimed at minimizing the number of future people will succeed in generating a high per capital level of capital. But it's a bit perverse to characterize this as a future-oriented policy framework. Yet according to NIPA such things as folic acid for pregnant women, vaccines for infants, school supplies, and the resetting of broken limbs after an accident on the monkey bars count as frivolous consumption in contrast to the prudent savings involved in the acquisition of a ski house. That kids-related problem with consumption taxation can be addressed in principle, as can the problem that an awful lot of "capital" in the actual economy is really rent (land, patents, etc.), but working it out in practice is hard, just like achieving full employment in practice is hard.