Tax policy colloquium, week 8: Gabriel Zucman's The Rise of Income and Wealth Inequality from America: Evidence from Distributional Accounts, part 2

 Carrying on with respect to the paper by Gabriel Zucman that we discussed at yesterday’s colloquium, here are some thoughts on particular issues raised by the aim of measuring material inequality:

1) Future Social Security benefits - Should these be included in distributional tables, at their expected present value? Does it matter, for this purpose, if they’re funded or unfunded?

I’d say it depends on the reasons for concern about inequality that motivate a particular measure, or use thereof. For example, if you’re trying to measure the relative expected lifetime material resources available to people at different vertical levels, they absolutely should be included. On the other hand, if you’re interested in a snapshot of the present, illiquidity and myopia may render them considerably less relevant, at least with respect to people who are not close to retirement. If you’re concerned about social gradient ills, top-down consumption cascades, or unequal political influence, their expected value may likewise have little relevance.

The particular measure that the paper works on developing aims at allocating all current-year national income to one individual or another. This is a worthwhile task, but consider what it misses in the following hypothetical.

Suppose that Congress today votes to have me given a $10 billion bond, to be cashed for that face amount in 2030. Suppose my property rights to it are so clear that it’s certain to be paid, and hence is worth $10 billion today (minus the loss of present value from my having to wait ten years for it). Suppose even that I’m forbidden to sell it, and can’t expressly borrow against it (i.e., by conveying a security interest to a lender). This is not part of national income – although it would appear in my Haig-Simons income, and perhaps even potentially in my taxable income (if not deemed a grateful nation’s gift). But it clearly makes me better-off, and is going to change how I live starting today.

The point is that future benefits, funded or not, may matter to welfare and behavior today – or may not, depending on the particulars of liquidity, one’s operative time horizon, etc.

2) Consumer durables - The rental value of consumer durables is excluded from the main income measures in the paper we discussed yesterday, to advance comparability with other countries’ existing measures. But the value of home ownership may effectively be included via real estate’s inclusion among assets that have imputed returns. There’s a strong case that this stuff should more generally be included, at least to get an accurate domestic picture (leaving aside the value of cross-border comparability of measures). And it might compress things a bit given that, for people at the very top, the value of consumer durables may be a smaller percentage of economic income than it is for people more towards, say, the upper middle.

3) Capitalization - One of the biggest disagreements between experts who prepare these measures (including, e.g., Zucman, and our earlier years’ colloquium guests Gerald Auten and Eric Zwick) pertains to the capitalization rates that one needs to use in navigating back and forth between income and wealth measures. Suppose, for example, that you’re using a 2% capitalization rate. Then $2 of current year capital income implies an asset worth $100, while an asset worth $100 implies $2 of current year income. When a study finds less high-end wealth concentration (and recent increase therein) than Saez and Zucman find, this often reflects the influence of assuming that higher discount and capitalization rates apply to higher-income individuals, on the ground that they are able to earn higher returns.

To illustrate, suppose Ms. Middle has $2 of capital income on her tax return, while Ms. Upper has $6. Applying a uniform capitalization rate of 2%, Middle is deemed to have $100 in capital assets, and Upper $300. But suppose instead that Upper earns a 3% return. Then, while we still infer $100 for Middle, we now infer only $200, not $300, for Upper.

Although I understand the logic, I’ve always been a bit bemused by this sort of adjustment. Personally, all else equal, I’d rather earn 3% than 2%, and would deem myself better-off if I were doing so. 

The result is also impossible in perfect capital markets where there are no risk differences between assets and also no admixture between labor and capital income. Thus, while I perfectly well accept that it may indeed be true in our world, I’m not convinced that thus lowering Upper’s measured wealth, and thus showing the society to be more equal in the scenario where she earns a higher rate of return, I would tend to think of this differently.

Should we say that she has an implicit asset that raises her return from 2% to 3%? Or that it’s human capital / conceptually labor income? If it reflects her accepting greater risk (e.g., of a black swan collapse of the stock market), does that downside risk matter, for any of the purposes for which we measure inequality, if people are pretty much ignoring it and ex post the risk never is realized?

4) Value of human capital - Many people dislike the term “human capital” for one reason or another. But, semantics notwithstanding expected future earnings can raise current wellbeing and affect current behavior, in some circumstances no less than tangible or financial wealth.

Consider James Dolan, who as the Madison Square Garden Company’s (and thus effectively the New York Knicks’) CEO earns $54 million per year. He has job security since he is the owner, and he evidently loves his job, despite his being only a Daniel Snyder away from being indisputably the worst major sports franchise owner in North America. Aren’t his expected future earnings part of his current financial status?

Admittedly this example ignores the fact that Dolan is effectively paying his own salary via his ownership interest. But if a CEO without a similar ownership stake shared not only his current salary but his future salary expectations, it would understate his current true position to ignore the present value of expected future earnings.

America’s rising high-end inequality in recent decades has largely been fueled by changes in labor income. I think we risk under-measuring it when we don’t include this sort of thing.

Should everyone’s expected future labor income be included? After all, a 60-year old janitor, earning $20,000 per year, also has an expected present value that one can compute. But the grounds for inclusion are clearer when the individual (a) likes the work and hence wouldn’t retire if that were feasible, and (b) has greater rather than lesser access to capital markets to realize this value currently, even if not by formally borrowing against it.

All this doesn’t mean so much that human capital should ever be included in these measures, as that we should keep in mind that excluding it may throw off the wealth measure’s capacity to illuminate actual high-end inequality (and its various potential derivative consequences).

(5) Individual versus household measures - In measuring high-end inequality, it can make a big difference how one treats, say, spouses. Does one amalgamate them in a single household, and measure inequality as between households? Does one look at individuals separately? If so, does one assume a particular split, such as 50-50, between spouses? This can actually have huge effects on how one measures wealth and income concentration at the top.

In the income tax literature, it’s common to note that spouses do not in fact generally share their income and wealth equally. For example, in a heterosexual marriage, it may be common for the man, especially if he is the higher (or only) earner, to have more sway. But this does not rebut the relevance of household-level assets and income to the material circumstances of all household members.

We can define a household as a group of individuals (often cohabiting) who, at least to some degree, pool their collective resources and then allocate them among the members based on internal rules or norms other than just “eat what you kill.” These rules may be hard for us to observe, but that does not reduce their potential conceptual relevance

There’s a well-known issue in the income tax debate, concerning whether it might matter distributionally if households enjoy economies of scale. (Even if the old saying that “two can live as cheaply as one” overstates it. There may also be non-rival consumption within households. For example, if I am extremely rich, and I newly find a partner (whether or not we are legally married), one might reasonably view the partner as now also being extremely rich, without my really now being any less rich. Even apart from mutually beneficial expenses (e.g., I’d rather vacation with a loved one than all alone), our enjoyment of the status and power that the wealth provides may be to a degree non-rival.

It’s hard to determine the “right” way to treat this set of issues. But it can have an anomalously large effect on how we end up measuring high-end inequality. Consider, for example, the relative power and status in multiple realms of the very rich versus the rest of us. How much of a difference does it make, to those of us in the bottom 99%, how equally or unequally spouses in the top, say, 1% or .1% or .01% “share” their income or wealth? Probably a lot less than it may seem to matter when one compares the bottom line results in studies that use different conventions with respect to households.