The two sources

Everything on the wealth page is computed from two Federal Reserve releases, both US-government works in the public domain. The Distributional Financial Accounts (DFA) give quarterly shares and dollar levels of household net worth by wealth percentile, generation, race, education, age and income, from 1989 Q3 to 2026 Q1. The Survey of Consumer Finances (SCF) gives median and mean family net worth by characteristic for each triennial wave, 1989 to 2022. We build a parquet from each and query it at build time; no number on the page is typed by hand.

How the DFA is built

The DFA does not measure households directly. It takes the aggregate household balance sheet from the Financial Accounts of the United States (the Z.1 release) and distributes each aggregate across households using the distribution observed in the SCF, interpolating between the triennial SCF waves and extrapolating forward from the latest wave. The consequence that matters for reading it: the DFA net-worth level, summed across the five wealth groups, equals the Z.1 “Households; Net Worth” total by construction. We verify that identity at build time against the Z.1 aggregate, so the headline total and the shares beneath it cannot drift apart.

Two caveats follow directly. First, the most recent quarters are the most heavily model-based, because they extrapolate past the latest SCF wave; the Fed revises the DFA with each Z.1 release and each new SCF wave, so recent points move. Second, the intra-wave quarters are imputations, not observations: the DFA interpolates the distribution between surveys. Both are properties of the method, not errors, and both mean the quarterly wiggles carry less information than the multi-year trends.

Households, families, and why there is no DFA median

The DFA counts households; the SCF counts families (a primary economic unit that can be a subset of a household). The two units differ, and we state which one an exhibit uses rather than pretending they are interchangeable.

More importantly, the DFA has no median and cannot have one. It is a distribution of an aggregate: it answers “what share of the total does group X hold,” not “what does a typical member of group X have.” A median is a property of a distribution of families, which only the SCF provides. Any figure presented as a “DFA median” is a category error. On the wealth page, every median is SCF, every share is DFA, and the race exhibit shows the two in separate panels precisely because they answer different questions.

Why methods disagree about top wealth

Measuring the very top of the wealth distribution is genuinely contested, and the DFA's 31.6% top-1% share (2026 Q1) is one method's answer, not the answer. Three approaches, three biases:

Surveys (SCF). Even with the SCF’s special high-wealth oversample, surveys under-cover the extreme top through non-response and top-coding, and the DFA inherits the SCF distribution it is built on.
Administrative capitalization (Saez and Zucman). Capitalizing tax-reported capital income into wealth stocks reaches the top through tax records, but the result is sensitive to the assumed rate of return: divide a given flow of capital income by a lower assumed return and you infer a larger stock of wealth.
Heterogeneous returns (Smith, Zidar and Zwick). Smith, Zidar and Zwick (“Top Wealth in America: New Estimates under Heterogeneous Returns,” Quarterly Journal of Economics 138(1), 2023, 515–573) show the wealthy earn systematically higher returns, so a given capital-income flow implies less underlying wealth than a homogeneous-return capitalization assumes. Their top-1% wealth share reaches 33.7% by 2016, below the Saez–Zucman capitalization estimates though still high and rising.

The DFA sits closest to the survey tradition, benchmarked to Z.1 totals. We display it as the maintained quarterly series and flag, here, that its top share is a lower bound in spirit relative to the capitalization estimates.

Wealth is not income

This page is about the distribution of wealth (a stock). The distribution of income (a flow) is a different object on a different clock, and its top shares are contested in their own way: the Piketty–Saez–Zucman series (maintained in the World Inequality Database) puts the US top-1% pre-tax income share around 20% and rising, while Auten and Splinter (“Income Inequality in the United States: Using Tax Data to Measure Long-Term Trends,” Journal of Political Economy 132(7), 2024) allocate untaxed and retained income differently and find roughly 14% with little rise. Neither is refuted; the gap is the current state of knowledge. Income concentration, with that divergence shown in full, belongs to the cross-country inequality module rather than here.

Where wealth comes from: firm formation and asset ownership

The "where wealth comes from" section adds two public-domain series. The Census Business Dynamics Statistics (BDS) count firms and establishments by year since 1978. The startup rate we plot is the count of firms at age 0 (a year's births) over all firms, the Decker-Haltiwanger-Jarmin-Miranda definition, computed from the two BDS counts at build time; it ran from 13.7% in 1978 to 9.1% in 2023. The Census establishment entry rate is a separate, directly published measure and is quoted as a cross-check, not merged with the firm startup rate.

Asset ownership comes from the SCF's Tables 6 and 9, "percentage of families holding asset", by wave and by wealth and income percentile group. We read the percentage-holding block only, not the value-of-holdings block beneath it. "Directly held stocks" is the SCF's Stocks column: it counts families holding individual shares directly and is far below the direct-plus-indirect stock participation the Bulletin reports separately (which also counts stock held through funds and retirement accounts), so it is labelled precisely on the page. Ownership is a share of families, on the SCF's triennial clock; in 2022, 66.1% of families owned their home, 21.0% held stocks directly and 14.6% held business equity. These rates carry no causal claim in either direction.

Sources and licence

Federal Reserve Board, Distributional Financial Accounts (federalreserve.gov/releases/z1/dataviz/dfa) and Survey of Consumer Finances (federalreserve.gov/econres/scfindex.htm). Both are US-government works in the public domain. We display these figures with attribution and republish no bulk data: no download of the DFA or SCF files is offered here. The SCF median we headline, $192,900 for the 2022 wave, matches the SCF Bulletin's own all-families figure.

U.S. Census Bureau, Business Dynamics Statistics (census.gov/programs-surveys/bds), also a US-government work in the public domain, accessed through the Census data API. As with the Fed sources, these figures are displayed with attribution and no bulk data is republished.

Methodology