Based on a set of thoughtful adjustments that appear to me clearly motivated by a scientific desire to get it right, rather than by any ideological or other axe to grind, Zwick et al find that the top 0.1%’s wealth share is somewhat smaller, and has grown less substantially in the last 50 years, than Saez and Zucman find. But I’d say that the basic outlines of the story – rising high-end wealth inequality- don’t change in any fundamental way.
One implication that could in principle matter, however, pertains to Zwick et al’s “mechanical tax revenue calculation” for the proposed wealth tax. As they emphasize, this is not an actual revenue estimate, as it omits behavioral responses and enforcement issues. It is based on asking, if all the actual wealth above the proposal’s threshold amounts were actually taxed at the indicated rates, how much revenue would be raised? So it offers an upper bound that is likely substantially to exceed the actual revenue yield if the wealth tax were enacted. Zwick et al find this amount to be significantly lower than Saez and Zucman had concluded in their letter to Senator Warren (which was self-presented as a revenue estimate, albeit based on a mechanical calculation under their figures plus a downward adjustment for tax planning effects et al).
Certainly the most interesting, and perhaps the most important, of the ways in which Zwick et al compute top wealth shares differently than Saez-Zucman pertains to capitalization rates. Suppose (as is in fact the case here) that one is using income tax data to infer wealth holdings. To illustrate with a very simple example, suppose that IRS data show I am reporting $5/year of interest income. The wealth question is what wealth holdings (e.g., a bond that I might own) are generating this income flow.Under a capitalization approach, if we assume that my savings earn 5% a year, it must be a $100 bond, so we include that amount in an estimate of my wealth. But if we assume that I earn only 4% a year, then it’s presumed to be a $125 bond. In short, the lower the presumed interest rate, the more wealth I must have to generate the observed income flow.
I gather that prior work (such as Saez-Zucman) assumed uniform capitalization rates for lower- and higher-income individuals. But Zwick et al note evidence suggesting that higher-income individuals earn higher rates of return than lower-income individuals. So they apply heterogeneous capitalization rates. In this example (retaining or increasing its hyper-simplification for expositional purposes), we might assume that it had to reflect $125 of savings if in the hands of a non-plutocrat, but only $100 if in the hands of a plutocrat.
Unsurprisingly, this adjustment reduces the estimate of top wealth shares – although Zwick et al also make adjustments that increase it. (Again, this is a fair-minded effort, however one comes out on all of the estimating issues.)
Why would higher-wealth individuals earn higher rates of return than lower-wealth individuals? There are a number of possible reasons. Savings by the latter are very substantially placed in checking and savings accounts in banks, which offer liquidity but very low returns. The upper-tier folks are far more invested in bonds that pay higher rates but that are riskier – at a minimum, by reason of not being federally insured. (They’re also far more in the stock market, but this is in a different computational bucket as it doesn’t yield reported interest income.) They may have greater risk tolerance, less need to sacrifice expected returns for liquidity, greater access to information and opportunities that permit them to find higher-yielding investments, and so forth.
But here’s a peculiar aspect to the adjustment that calls for rumination outside the four corners of the computational debate itself. One of the aspects of the wealthy’s superior position in our society is that they can earn greater returns. So it’s peculiar and paradoxical, however logically consistent with the underlying computational enterprise, to say: Because they can earn a higher rate of return than the rest of us, therefore we will lower our estimate of how much high-end inequality there is. Isn’t that disparity a part of the broader story, rather than an indication that things are less askew than we thought?
But I will place reflections on that question in a separate blog post, to follow shortly.