“[T]he ratio of the wealth of the wealthy to the wealth of the less wealthy …may reflect something about the adequacy or inadequacy of social insurance arrangements…” Lawrence Summers, comments at Peterson Institute panel, October 17, 2019.
Wealth inequality in the United States is obvious to everyone. The Federal Reserve’s triennial Survey of Consumer Finance (SCF) documents the glaring and persistent divide between rich and poor, confirming that ownership of financial and real assets in the United States has been highly concentrated for decades (see our earlier post). The most recent 2016 estimates suggest that the top 10% of the wealth distribution own nearly three-quarters of all marketable assets, with the top 1% owning more than half of that. And, Saez and Zucman (SZ) estimate that the U.S. distribution has been getting worse, with the top 1% share of marketable wealth rising by more than 10 full percentage points since 1989.
But, as Catherine, Miller and Sarin (CMS) recently highlight, adding in the present discounted value of Social Security benefits (net of taxes) to construct a more comprehensive measure of wealth alters these patterns. First, according to CMS’s estimates, the share of marketable wealth in total wealth has plunged by more than 18 percentage points since 1989. Second, over the past three decades, the top 1% share of total wealth has risen only modestly, while the share owned by the top 10% has declined somewhat.
In this post, we highlight the CMS results, and decompose their changes in total wealth shares into two parts: the changes in marketable and Social Security wealth shares accruing to each group, and the aggregate decline over time of marketable wealth as a share of total wealth. We show that the latter dominates the overall trend in this more comprehensive measure of inequality.
To help understand the importance of including social insurance programs, start by considering the three primary reasons that individuals accumulate wealth: saving for retirement (smoothing consumption over one’s own lifetime); bequeathing wealth to one’s heirs (smoothing consumption over generations); and having a rainy-day fund for unanticipated expenses or declines in income (self-insurance against temporary shocks).
The more generous social programs are, the weaker these motives to save (see the opening quote from Lawrence Summers). Here is a simple, if extreme, example. If the government promises a retirement annuity that will cover 100% of a person’s pre-retirement consumption, aside from the bequest motive, there is little need to accumulate wealth during one’s working years.
So, how large is Social Security wealth as a share of total U.S. wealth? The answer: substantial and rising. The following chart plots inflation-adjusted (real) wealth divided into two parts. The gray area is the SZ estimate of marketable wealth. This includes housing, the value of sole proprietorships and partnerships, and private pensions, in addition to financial assets. Measured in 2018 dollars, from 1989 to 2016, marketable wealth increased by roughly $50 trillion, rising by a factor of 2.6. However, the present discounted value of Social Security wealth computed by CMS (shown in red) rose from $4 trillion to $34 trillion—that’s an increase of 8.3 times.
Total Value of Marketable and Social Security Wealth (2018 dollars), 1989-2016
Sources: Saez and Zucman; Catherine, Miller and Sarin; and authors’ calculations. CMS construct estimates every three years—when SCF data are available.
Given that, since 1989, equity and property prices rose substantially in real terms, it may seem surprising that that the present discounted value of Social Security benefits rose more than twice as fast. There are three reasons for this: changes in institutional arrangements, increases in life expectancy, and a fall in interest rates. CMS describe how expansions of Social Security coverage and benefits account for some of the increase in Social Security wealth. In addition, over the past 30 years, life expectancy has risen by over 2 years, adding to the present value of the government life annuity.
While these two effects may be important, the fall in interest rates swamps them both. To see why, consider the forward rates (computed from a zero-coupon Treasury yield curve) that CMS use to compute the present value of Social Security benefits. These interest rates have fallen dramatically over the past 30 years. To give some sense of how much, note that the 10-year constant maturity Treasury rate less inflation fell from an average of 3.7% in 1989 to 0.6% in 2016. (This plunge accounts for the fact that CMS’s estimate of the rise in Social Security wealth exceeds that reported by the Social Security Administration.) In the first half of this year, that real yield has averaged less than zero.
We now turn to the main CMS result. The following chart presents the impact of including Social Security wealth in total wealth. The black lines, solid and dashed, show the SZ estimates of marketable wealth shares. Focusing on the period starting in 1989, there is an increase of 10.3 percentage for the top 1% share and of 8.9 percentage points for the top 10% share. The CMS estimates in red differ markedly. First, for the top 10%, the CMS share averages 10 percentage points lower. (The 1% share is 7½ percentage points lower.) More importantly, the dramatic increases in the SZ measures disappear. While the CMS estimate of the 1% share rises by 2½ percentage points—one quarter of the SZ rise. And, rather than rising, the top 10% share falls by 3 percentage points.
Top 10% and Top 1% Shares of Marketable and Total Wealth (Percent), 1950-2016
Source: Saez and Zucman; Catherine, Miller and Sarin; and authors calculations.
What accounts for the difference between the SZ and CMS estimates? To see, start by noting that we can decompose the share of total wealth of the top 1% or 10% into two parts. The first part is the product of their share of marketable wealth times the fraction of marketable wealth in total wealth. The second part is the product of their share of Social Security wealth times the fraction of Social Security wealth in total wealth. Put differently, as a matter of accounting, the difference in the levels of the SZ and CMS estimates can arise from two sources: the aggregate fraction of marketable wealth in total wealth or a group’s claim on Social Security wealth. And, the divergent paths of the two estimates can arise from a combination of the changes in these same two quantities.
To highlight why the evolution of the two paths looks so different, in the following table we report a decomposition of the changes in the CMS estimates into these two parts. Start by noting that the share of marketable wealth in the total declined by 18.2 percentage points. This plunge sharply offset the SZ estimates of increased shares of marketable wealth held by the top 1% and 10% of 10.3 and 8.9 percentage points, respectively. As a result, the contributions to the changes in their total wealth shares were +2.4 percentage points and –5.2 percentage points.
Similarly, while the Social Security wealth shares of both groups declined (by 1.4 percentage points for the top 1% and 8.3 percentage points for the top 10%), the increase in the aggregate share of Social Security wealth in the total was so large that it led to modest positive contributions from Social Security wealth to the top 1% and 10% shares in total wealth (of +0.2 and +2.2 percentage points, respectively).
So, the conclusion is that the overall changes in total wealth shares—the 2.6 percentage-point increase for the top 1% and the 3.0 percentage-point decline for the top 10%—are largely a consequence of the increase in the share of social security wealth in total wealth (or, equivalently, of the large decline in the share of marketable wealth).
Decomposing the Changes in the Top 1% and Top 10% Shares of Total Wealth
Source: First row is from Saez and Zucman. We thank Catherine, Miller and Sarin for providing the information in the remainder of the table.
CMS are extremely diligent in demonstrating the robustness of these patterns in the distribution of total wealth. Among other things, they examine the implications of using different interest rate assumptions, of adjusting for risk, of allowing the wealthy to have a longer life expectancy, and of ways in which the funding gap might influence future Social Security benefits. The results remain. In addition, our back-of-the-envelope estimates suggest that if we use information from the Social Security Administration’s Office of the Chief Actuary in place of the CMS estimates for Social Security wealth, the CMS patterns are largely unchanged.
Now, on top of changes in Social Security, U.S. government health care coverage also has changed dramatically since 1989. Over the past 30 years, the combined real annual expenditure of these programs has grown faster than that of Social Security and is now of similar magnitude. A very rough calculation suggests that including these health care benefits in total wealth modestly reinforces the CMS conclusions. That is, the top 1% and 10% shares of total wealth are a few percentage points smaller, and the changes over time are slightly lower. But we caution that our health care results could easily be mistaken. We hope that someone will take up the painstaking and complex challenge of adding Medicare and Medicaid to wealth to see how much they matter for the distribution.
So, what should we conclude? First, the CMS results confirm what we have long known: even after adding in Social Security wealth, the distribution of total wealth remains glaringly unequal. The top 1% lay claim to more than one-quarter of total wealth. In addition, despite greater Social Security benefits, many U.S. households still doubt that their savings are adequate for retirement (see, for example, here). By itself, however, the widening inequality over the past 30 years in marketable wealth does not imply a reduced ability to meet retirement needs for those below the top 10% of the distribution. Put differently, while we need to do more, changes in the safety net over the past 30 years have offset somewhat the welfare implications of rising inequality in the ownership of marketable wealth.
It is exactly for this reason that research to improve measures of wealth inequality, broadening them to include the value of various social insurance programs, is so important. And, as we see in the important new work of Catherine, Miller and Sarin, doing this carefully can affect policy-relevant conclusions.
Acknowledgements: We thank Sylvain Catherine, Max Miller and Natasha Sarin for sharing their data, and Greg Mankiw for sharing his discussion of their paper.