Just kidding... Robert Murphy's recent Econlib article on income and capital was quite good. The title is just a reference to a recent Noah Smith article*. This is not to say I am 100% convinced by Murphy.
Bob starts off by explaining the value of capital as the capitalized value of a stream of income generated from that capital, and then also introduces Hayek's alternative view from Pure Theory of Capital. I had a little trouble with the sharp distinction Bob was trying to draw here, with the former as being "static" and the latter as being "dynamic", united only when capital gains are considered. Capital gains follow so fundamentally and naturally from a basic capitalization approach that I think this treatment of capitalization as "static" is pretty weak. I think of the real contribution of Pure Theory of Capital as being the development of the idea of roundaboutness, but in fairness I've only read parts of the book.
I would not worry so much about the capitalization/Hayek distinction though. Indeed it seemed like just a nice excuse to get Hayek and Mises in there. The important thing is that Bob explains capitalization and capital gains to his readers because these are critical for understanding the new Saez-Zucman work and the potential trouble with using tax data.
So in a nutshell the Saez-Zucman study: (1.) uses the capitalization method to estimate wealth distributions from income data, and (2.) diverges from direct tax measures of wealth:
Bob argues that income tax changes in 1986 reducing top marginal income tax rates incentivized many people to reorganize their businesses as S corporations that would show up in the income tax data. Several years ago Scott Winship shared a chart that I think originally came from Alan Reynolds illustrating the point:
Notice, though that, the big jump you see in Winship's series happens between 1986 and 1988, for obvious reasons. After that there seems to be low, steady growth in the contribution of "entrepreneurial income" but the major contribution to personal incomes reported for tax purposes has already been made. That does not seem to explain the trend increase in the S-Z IRS data or the SCF (unless I'm misunderstanding something about how business income is reported and whether some of that comes under the wage increase - which is entirely possible). In other words, the big pre-1986/post-1986 change seems to be a level change and a composition change, but what Bob is trying to explain here is a trend change.
It's also worth noting the big differences between the rich and the super-rich (or more like the super-rich and the super-duper-rich). Most of the changes have happened at the very top, so it's important to think about whether Winship's income numbers, Saez-Zucman's wealth numbers, the SCF's wealth numbers, and Kopczuk-Saez's wealth numbers are all dealing with the same population. The sampling frames are quite different, and given the small share of the population we're talking about small differences in the population can make the difference between a stark U-shape and a flat line. In other words, the contradiction here may or may not even be contradictory.
Ultimately, something that the Zucman-Saez number has in its favor is that it lines up with the increase in the SCF over time. The SCF is designed to really get at the top earners and it's direct survey data that does not require the substantial imputations necessary for using estate tax data. The SCF is also nice because it reports both income and wealth (and it even counts the tax-deferred wealth that Bob raises as a problem for income tax measures). In theory, you could apply the Saez-Zucman capitalization method to the SCF and compare the SCF capitalization estimate of the top 1% share to the SCF reported wealth estimate of the top 1% share. Maybe someone has already done this (maybe Saez-Zucman have already done this and I haven't looked closely enough!) but that would be a very nice test of the validity of all this.
So we still have the estate tax data to explain. Needless to say I am neither old nor rich (although I am feeling a little more of both every year) so I don't feel like I have a good grasp of what goes into estate planning that might be relevant here. But it is worth noting that in the last thirty years estate tax rates and exemptions haven't exactly been stable either. The best explanation I can come up with is that a person's estate is going to be a lagging indicator of wealth changes for the obvious reason that dead people amass their wealth over several decades. Short-term effects like financial crises may have big immediate impacts on estates, but a legitimate increase in wealth inequality starting in the mid-1980s might take some time to show up.
Those are my thoughts - I have no strong answers but my prior is that personal income tax and SCF data are more direct measures and estate tax data is less direct, so when the former two agree with each other we ought to take note. I also think that Bob's explanation is more of a level effect than a trend effect (unless I'm misunderstanding something), so I'm not sure how much it really explains. But the article is a really nice introduction to the issues and to the Saez-Zucman controversy.
* - Which, in case you care about my opinion on these things, mirrored quite closely the criticisms that the GMU Austrians have made of the exact same people that Noah was criticizing. For some reason, though, even the GMU Austrians didn't seem happy with his article. I guess it's an "only we are allowed to criticize our own" thing.