Wealth Inequality

Wealth inequality is the unevenness of the distribution of net worth across households. Net worth is the value of what a household owns less what it owes, measured on a given date, so wealth inequality is a statement about stocks held rather than about money received during a period. That distinction is the boundary with the neighboring term: income inequality describes how much people receive in a year, wealth inequality describes how much they own at a point in time, the two are measured by different agencies from different surveys, and they are not the same shape, because wealth accumulates over a lifetime while a single year's income does not.

In the United States the principal measurement comes from the Board of Governors of the Federal Reserve System, which runs the Survey of Consumer Finances. The Board describes it as "normally a triennial cross-sectional survey of U.S. families" whose data "include information on families' balance sheets, pensions, income, and demographic characteristics." Because the survey is triennial, the Federal Reserve also publishes the Distributional Financial Accounts, which it describes as "quarterly estimates of the distribution of a comprehensive measure of U.S. household wealth" built by using "distributional information from the SCF to allocate the Financial Accounts aggregate measures of assets and liabilities" across groups. The arithmetic of net worth itself, and how a household computes its own, belongs to the pages on net worth and the personal balance sheet.

The most consequential thing about measured wealth concentration is what the measurement leaves out, and the Federal Reserve says so in its own report on the survey. Its footnote on retirement assets states that "two common and often particularly important types of retirement plans are not included in the assets described in this section": Social Security, which the Board notes "covers the great majority of the population," and employer-sponsored defined-benefit plans. Neither appears anywhere in the survey's published net-worth definition.

The reason is a measurement problem rather than an oversight, and the Board states it: "future income streams from OASDI and defined-benefit plans cannot be translated directly into a current value because valuation depends critically on assumptions about future events and conditions—work decisions, earnings, inflation rates, discount rates, mortality, and so on—and no widely agreed-upon standards exist for making these assumptions." Putting a number on a promised lifetime income requires choosing a discount rate and a mortality assumption, and different choices give very different answers.

The consequence for a reader is precise and easy to state. Measured wealth concentration and lifetime-resource concentration are not the same quantity. A Social Security benefit and a defined-benefit pension are both resources a household will live on, and both are assets in every ordinary sense, yet neither appears in a survey net-worth figure. This does not make measured concentration wrong; net worth is a well-defined thing and the survey measures it consistently. It means the published statistic answers "how unevenly are balance-sheet assets held?" rather than "how unevenly are the resources people will actually live on distributed?"

What the survey does count is worth knowing, because the boundary is not intuitive. The Federal Reserve's own definition of net worth for its survey reports places account-type retirement holdings inside total assets, listing individual retirement accounts and Keoghs, account-type pensions on a current job, future pensions, and currently received account-type pensions among quasi-liquid retirement accounts, alongside transaction accounts, certificates of deposit, directly held stocks and bonds, mutual funds, the cash value of whole life insurance, annuities and trusts. On the nonfinancial side it counts vehicles, the primary residence, other residential and non-residential real estate, and business interests. Debt covers mortgages and home equity borrowing, other lines of credit, credit card balances after the last payment, installment loans including education and vehicle loans, and other debt. So a defined-contribution balance is in the measure and a defined-benefit promise is not, which means two workers with identical retirement security can record very different net worth depending only on the type of plan their employer chose.

One further caution about sourcing. Because the survey runs on a multi-year cycle, the phrase "the latest Survey of Consumer Finances" ages quietly, and a reader comparing a wealth statistic against a current income statistic is comparing a survey wave with a monthly or annual series. Any wealth-share figure should be quoted with the wave it came from.

Measurement runs from a household interview to a distribution. Interviewers collect the value of each asset and each debt a family holds, net worth is computed as assets less debts, families are ranked, and shares of total net worth are reported for groups such as percentile bands. Because the distribution is of a stock, one household can hold a large multiple of another's wealth while their annual incomes are similar.

A hypothetical shows why the excluded items change the picture. Consider two households, both aged 60. The first has $600,000 in a 401(k), no pension, and no other assets or debts, so its survey net worth is $600,000. The second has $100,000 in a 401(k) and a defined-benefit pension that will pay $36,000 a year for life, with no other assets or debts, so its survey net worth is $100,000. On the measured statistic the first household holds six times the wealth of the second: $600,000 ÷ $100,000 = 6.

Now price the pension, which the survey deliberately does not do. Valued as an income stream, $36,000 a year has some present value: at a 4 percent discount rate over 25 years, using the ordinary annuity factor, the figure is roughly $562,000, and at a 6 percent rate over 20 years it is roughly $413,000. Add either to the second household and its total resources become comparable to, or larger than, the first household's, and the six-times gap disappears. The point is not that one of these numbers is the truth. It is that the answer swings by more than $149,000 on the choice of discount rate and horizon alone, which is precisely the reason the Federal Reserve declines to put a single number on it.

What the measure is good for

• It captures the balance sheet, which determines a household's capacity to absorb a shock, to borrow, and to transfer resources to the next generation, none of which annual income shows.
• It rests on a long-running, documented survey with a published definition of every asset and debt category counted.
• The Federal Reserve supplements the survey with a quarterly distributional series, so the picture is not frozen between waves.

Where it misleads

• It excludes Social Security and employer defined-benefit pensions, which for many households are the largest retirement resources they have.
• Two workers with the same retirement security can record very different net worth depending only on whether their employer offered a defined-contribution or a defined-benefit plan.
• Survey wealth data arrive on a multi-year cycle, so a wealth statistic and an income statistic are rarely as of the same date.
• Net worth depends on asset valuations, so measured concentration moves with house prices and stock prices independently of anything households did.
• A distribution is not an explanation. The statistic describes a pattern and does not identify a cause or prescribe a response.