alan auerbach, a professor of economics and law at the University of California (Berkeley), was named a distinguished fellow by the American Economic Association in 2021. larry kotlikoff, a professor of economics at Boston University, is president of Economic Security Planning.
Illustrations by john dykes
Published April 21, 2023
Defining and measuring inequality is of keen interest to contemporary economists. And, of course, it’s much more than an academic game. Policymakers need to understand the distribution of well-being, and in particular how it is affected — and changed — by taxes and government transfer programs like Social Security and Medicare. Accordingly, researchers have been treading this ground heavily for decades. But we believe the methods they have used are ill-suited to answer a key question: who gets to spend what, not just this year, but over the remainder of a lifetime.
Here, we summarize a comprehensive approach yielding results that don’t always fit the conventional wisdom. Those seeking a deeper dive can read our technical article, “U.S. Inequality and Fiscal Progressivity: An Intragenerational Accounting” (with Darryl Koehler), which is forthcoming in the Journal of Political Economy. Spoiler alert: some of our conclusions should help to assuage the concerns of those who have previously concluded that government policy has done little to offset marketdriven inequality.
Now You See It…
To illustrate the problems of traditional methods, consider some seemingly straightforward questions that elude straightforward answers.
Who is better off financially? A person with a million-dollar inheritance but no marketable skills or their friend earning $100,000 a year on the job? How about a retiree, who has little income from investments? Is that pensioner better off than a grandchild, who works full-time at a modest salary?
Is the payroll tax that currently funds the lion’s share of Social Security and Medicare as regressive as it appears? After all, it does impose the same tax rate on minimum-wage workers as on far better paid salaried workers. And in 2023, the tax doesn’t apply at all to earnings above $160,200.
The retiree has less income but may have accumulated wealth during a lifetime. And even if that wealth doesn’t earn much income, the retiree can spend down some of it each year to sustain their living standard.
To what extent do differences in life expectancy among different income groups influence the distributional impact of government programs targeting the elderly?
We believe most analysis is insufficiently comprehensive to address the questions raised above. Why? Start with the first question, the one comparing the trust fund baby with the high earner. You can’t hope to learn much about the relative well-being of the two individuals by looking separately at distributions of wealth and income. But integrating the two requires a lot of data and some important assumptions: you can’t simply add wealth to income to draw relevant conclusions because this fails to account for size and uncertainty of the incomes one or both will be earning in the future.
Now for the comparison of the retiree and the grandchild. The retiree has less income but may have accumulated wealth during a lifetime. And even if that wealth doesn’t earn much income, the retiree can spend down some of it each year to sustain their living standard. Consider, too, that this comparison doesn’t account for taxes and transfers. With little income, the retiree’s taxes may be much lower than those of their grandchild. And this doesn’t even account for the fact that many government transfer programs, most notably Social Security and Medicare, provide substantial benefits (and tax breaks) to the elderly.
Or consider the progressivity of the Social Security payroll tax. The tax is only part of the story when considering the overall progressivity of the giant transfer system. One must also account for the impact of the programs that the payroll taxes help finance, since those who pay the taxes will eventually receive benefits in the form of disability insurance, pension checks, survivor benefits and medical insurance. Even though the payroll tax alone is regressive, taking a greater share of income from lower-income individuals than higherincome individuals, it doesn’t follow that the taxes and benefits combined are regressive.
Moreover, weighing the combined impact of taxes and program transfers isn’t enough to determine progressivity because it doesn’t account for differences in timing associated with age. For example, Social Security and Medicare are nearly universal programs, providing benefits to most U.S. residents when they become old and/or disabled. But only a fraction of the population receives those benefits in a given year, meaning that we ignore the value of these programs to the remainder of the population.
By the same token, it is impossible to evaluate the role of differences in mortality on fiscal progressivity by looking at taxes paid and transfer payments received by the current population. One must also consider the effects of no longer being a part of the population. Even if the system replaces more of the income of the working poor on retirement, on average they live fewer years to benefit.
To get to the heart of the matter of inequality and fiscal impact, a more comprehensive approach is needed. For starters, in estimating inequality it is important to look at all resources, not just tangible wealth or income. And it is necessary to account for future income as well in assessing the sustainability of an individual’s standard of living.
Second, it’s important to include as many taxes and government spending programs as possible in measuring fiscal progressivity/regressivity. That means taking account not only of what individuals pay in taxes or receive in transfers in the current year, but how these amounts will change across a statistical life. By the same token, one should avoid comparing individuals of very different ages, for which differences in income, wealth, taxes paid and transfer payments received may reflect differences in where they are in the life cycle rather than their access to material resources.
Our analysis relies on a couple of key concepts. The first is “lifetime spending power” (LSP), the amount that a household commands over the life of the household head (or the household head and spouse). That’s taking account of all the resources the household will have at its disposal: its current wealth and its “human wealth” (the latter equal to its current earned income and the present value of its future earned income), plus the transfer payments it will receive, all net of taxes (again, in discounted, present-value terms).
Note that by calculating LSP both with and without taxes and transfers, we can also compute the household’s “lifetime net tax rate” (LNTR), which is equal to the percentage reduction (or increase) in lifetime spending power when taxes and transfers are included. By looking at lifetime net tax rates across groups with different before-tax lifetime spending power, we are able to assess fiscal progressivity — that is, how fiscal burdens change as one moves from less affluent households to more affluent households of any particular age cohort.
An added benefit of this approach based on lifetime spending, taxes and transfer payments is that it lessens problems arising from the sometimes-ambiguous nature of fiscal labeling. For example, types of retirement savings accounts vary with respect to the timing of tax liability even though they share the general objective of encouraging retirement saving by lightening the tax burden. Traditional IRA and 401(k) accounts provide an income tax deduction for contributions, and then tax the account owners when funds are withdrawn. By contrast, Roth versions of these accounts offer no upfront tax break — taxes are owed in the year the account deposits are made — but thereafter allow tax-free withdrawals of principal and earnings. The punchline: the different timing of the tax collections for these two types of accounts can lead to significant differences in calculations of a household’s annual tax payments, but not in its lifetime tax payments.
While these concepts — LSP and LNTR — are straightforward to describe, estimating the numbers involves a monumental amount of data and economic modeling. Many readers may wish to take our word for it. If you don’t, check out the peer-reviewed technical article in the Journal of Political Economy on which our conclusions are based.
Among all Americans aged 20-79, the top 1 percent controls 37.2 percent of all wealth. Break this down by age group, though, and the numbers are a bit less daunting.
The distribution of wealth is generally more equal within ten-year age cohorts than for the overall adult population. Among all Americans aged 20 to 79, the top 1 percent controls 37.2 percent of all wealth. Break this down by age group, though, and the numbers are a bit less daunting. Except among the youngest adults, very few of whom have any significant accumulation, the share of wealth controlled by the top 1 percent is below the share for the population as a whole. For 40- to 49-year-olds, the top 1 percent (ranked by wealth) controls 32.6 percent of wealth; for 50- to 59-year-olds 31.3 percent; for 60- to 69-year-olds 31.7 percent; for 70- to 79-year-olds 33.5 percent.
Wealth is much more unequally distributed than before-tax lifetime spending power. The figure on the top of the next page compares wealth and remaining lifetime spending shares among 40- to 49-year-olds. Note that the richest 1 percent (ranked by lifetime resources) have 29.1 percent of this cohort’s wealth, but only 11.8 percent of its remaining spending power. The poorest 20 percent have only 0.4 percent of cohort wealth, but account for 6.6 percent of the cohort’s remaining lifetime spending power.
There are two factors at play here. First, current and discounted future wage and salary earnings — what we call human wealth — are far more equally distributed than tangible wealth. For example, the top 1 percent of the 40- to 49-year-old cohort commands 10 percent of this group’s human wealth, which is roughly a third of the wealth of the age cohort. Meanwhile, the bottom 20 percent has 4.3 percent of the total 40- to 49-year-old cohort’s human wealth — roughly ten times this group’s overall share of net wealth.
Second, as discussed below, government taxes and transfers are progressive, further reducing lifetime spending power at the top relative to the bottom.
The fiscal system is very progressive. The middle figure shows estimates of the remaining lifetime spending power for 40- to 49-year-olds, breaking out the five quintiles of lifetime resources (those in the bottom 20 percent, the next 20 percent, and so on up to the top 20 percent) and also showing separately those in the top 5 percent and the top 1 percent.
While the fiscal system hardly levels lifetime spending power, the proportional reduction in LSP (as measured by the orange bars in the figure) is considerably greater at the top of the resource distribution than at the middle or the bottom. Another way of illustrating this progressivity is through lifetime net tax rates (taxes minus transfers), shown by the blue bars in the figure. These tax rates rise steadily as one goes from the left to the right in the figure. Indeed, those in the bottom quintile of the resource distribution enjoy substantially negative lifetime net tax rates — that is, they receive far more in transfers like Social Security and food stamps than they pay in taxes. The tax rates for those in the top 1 percent are roughly three times the rates for the solidly middle class in the second quintile.
Calculations based on current income and net taxes understate the degree of progressivity in the fiscal system. The orange bars in the figure above show the tax rates estimated by dividing all taxes net of transfers for the current year by current-year income. This is the sort of calculation found in the more ambitious contributions to the existing literature that attempt to incorporate the same vast array of tax and transfer programs as we consider here. But comprehensiveness in terms of taxes and transfers included doesn’t solve the program of incorporating future resources. For all resource groups in this cohort, lifetime net tax rates are lower than current-year tax rates because long-term net tax rates account for the substantial old-age benefits that individuals will only receive far down the road.
The effect of including future resources is much larger for those at the bottom of the resource distribution for two main reasons. First, as we will discuss shortly, these transfer programs are progressive. Second, those at the top of the resource distribution will pay taxes on their investment income in the future — taxes that reduce their lifetime spending power but are not included in current-year calculations.
Transfer payments are very progressive when based on forward-looking calculations. The figure at left shows the distributions of wealth, human wealth (discounted lifetime earnings), taxes and transfer payments for those aged 40 to 49. As already noted, the tangible wealth distribution is much more unequal than human wealth distribution. The distribution of taxes is also more unequal, reflecting the progressivity of the tax system. But transfer payments have the most impact on the lowest quintile of the population — the poorest fifth receive more than a quarter of the total — as measured by before-tax lifetime resources.
Among the most important transfer programs contributing to progressivity are those related to health care. The figure at the lower left breaks down the transfer payments shown above for each group of those in the 40- to 49-year-old cohort. While all of the programs are progressive, the most progressive are the Affordable Care Act health insurance subsidies and Medicaid, which deliver the lion’s share of their benefits to those in the lowest quintile. Even Medicare, which provides slightly higher benefits to those at the top (one cause of which, discussed below, is the longer life expectancy of more affluent groups), delivers a substantially higher share of before-tax lifetime resources to those at the poor end of the distribution.
Current income is an imperfect proxy for assessing a household’s place in the lifetime resource distribution. For example, among 40- to 49-year-olds in the third quintile — those between the 40th and 60th percentiles of the lifetime resource distribution — onethird would not be in this quintile in a ranking based on current income. About half of those misclassified belong in the two neighboring quintiles — those between the 20th and 40th percentiles and those between the 60th and 80th percentiles. Note that this type of misclassification can lead to errors in judging the progressivity of fiscal policy, for it will treat some individuals as substantially more or less affluent than they actually are.
Differences in life expectancy reduce the progressivity of the fiscal system, although the system remains very progressive. Not surprisingly, affluent Americans live longer than the less affluent. Many factors contribute to this, including differences in education, diet, smoking, exercise and health care. Policies to reduce mortality in general — and in particular this difference in mortality — may thus make a valuable contribution to improving the well-being of the less affluent. Apart from that, the current difference in life expectancy has implications for the progressivity of the fiscal system. Most of the transfer payments received by the elderly, including Social Security retirement income, Medicare and the significant chunk of Medicaid that pays for the long-term care of the elderly, are of course provided only as long as the recipient is alive.
Policies to reduce mortality in general — and in particular the difference in mortality between the affluent and less affluent — may thus make a valuable contribution to improving the well-being of the less affluent.
For those who are predicted to die earlier, then, these programs on average pay out less. Thus Social Security is progressive in that it replaces a bigger portion of the earnings of lower-income beneficiaries. But since, on average, lower-income individuals don’t collect benefits for as many years, the progressivity created by the higher replacement rate is partially offset.
To assess the importance of this issue, we recalculated lifetime net tax rates assuming that everyone had the more favorable life expectancy enjoyed by the top quintile. For those in the lower quintiles, this counterfactual calculation resulted in lower long term net tax rates: -56.6 percent versus -44.4 percent for those in the bottom quintile; 7.8 percent versus 11.5 percent for those in the second quintile; 17.1 percent versus 18.9 percent for those in the third quintile, and 22.8 percent versus 23.4 percent for those in the fourth quintile. Thus, while the fiscal system would be even more progressive if affluence did not affect life expectancy, it is still quite progressive.
* * *
While our analysis offers some striking insights into distribution, we’ve only scratched the surface. For example, there is much to learn about how inequality has changed over time, and how much of the difference is due to changes in fiscal progressivity. Studies using more traditional approaches haven’t yielded consistent conclusions. Some of these disagreements are due to differences in methodology, but all of the studies face the limitations we discussed above. We think we can do better with our data-rich, computationally complex, but conceptually more meaningful, approach.