Three Years, Two Economies: What the Survey Measures

On 9 October 2026 the Federal Reserve published the results of its 2025 Survey of Consumer Finances, the triennial study that remains the most detailed reading of what American families own, owe, earn and spend, and the first since the survey of 2022 [1]. Two pairs of numbers from its opening page describe the three years in between better than any commentary could. Real median family income rose 7 percent, to $82,200, while real mean family income fell 6 percent, to $145,200; real median net worth rose 2 percent, to $215,900, while real mean net worth rose 7 percent, to $1.24 million [1][2]. Read together, the pairs say that incomes narrowed and wealth widened over the same period: the typical family earned more relative to the average family, and the average family grew richer relative to the typical one. An economy in which the income distribution compresses while the wealth distribution stretches is not one economy moving at two speeds. Two economies occupy the same country, one that lives on wages and one that lives on assets, and the survey measures the distance between them with a precision no monthly release can match.

This catalogue argued the shape before the data existed. The K-Shaped Fragility, published in February, described an equilibrium whose headline indices signalled resilience while fragility compounded beneath them, and The Silent Fracture measured the gap between what Americans believe about the distribution of wealth and what the Federal Reserve's own accounts record. The survey closes the argument with the ledger itself. The Census Bureau's income report for 2025, issued in September, confirms the wage leg of the pattern from a different sample: median household income reached $87,460, the highest in a series that begins in 1967, up 2.6 percent on the year, while the Gini index of income inequality stood at 0.490 and did not change significantly, income at the 90th percentile rose 1.7 percent and income at the 10th percentile did not change significantly at all [3]. What follows reads the household balance sheet section by section: who spends, who owns, who is late, whose pay has kept pace with what they produce, and what the record of consent says about a ledger that looks like this one.

The Decile That Spends: One Tenth of Families, More Than a Quarter of Spending

The 2025 survey added questions on spending for the first time, and the answers explain why the aggregate consumption statistics have stayed so calm while the underlying households have not. Families spent $48,700 on average over the twelve months before the survey, on everything other than housing, and the median family spent $33,700 [2]. Families in the bottom quintile of usual income spent $17,600 and accounted for 7.2 percent of all spending; those between the 80th and 90th percentiles spent $68,700; and families in the top decile of income spent $138,800 each, nearly double the group immediately beneath them, and accounted for 28.5 percent of everything the survey's families spent. Sorted by wealth rather than income the picture is the same: the poorest fifth spent $23,400 and the richest tenth $136,000, or 27.9 percent of the total [2].

Mark Zandi, chief economist of Moody's Analytics, described the dynamic behind those shares in a conversation recorded in July: over the past three years, he said, overall real consumer spending has grown about 2 percent a year, spending by the top 20 percent of the income distribution has grown "close to 4%" and spending by the remaining 80 percent "close to zero", a pattern he called, without hesitation, K-shaped [4]. His explanation was the stock market. The Fed's Distributional Financial Accounts put figures under the explanation: in the second quarter of 2026 the top 1 percent of households held 32.5 percent of all household net worth, the highest share in a series that begins in 1989 and that opened at 22.8 percent; the top 10 percent held 68.9 percent; and the bottom half of the country held 2.3 percent, down from 3.5 percent at the start of the series [5]. The top 1 percent's share rose 0.8 of a point in a single quarter, from 31.7 percent, which is what a rising equity market does to a ledger in which the equities belong to one group.

The structure matters more than the level. An economy whose consumption growth depends on one decile, whose spending depends on the paper value of assets, whose value depends in turn on a narrow set of equity prices, has delegated its stability to that decile and to that market. The aggregate figures will read as resilient for exactly as long as the market cooperates, and the eight households in ten whose spending, on Zandi's measure, has not grown will not appear in them until it fails.

The Quartile in Negative Territory: Seventeen Hundred Dollars at the Median

The survey's wealth tables divide the country into groups whose experiences since 2022 do not resemble each other. Among families in the bottom quartile of net worth, median net worth fell from $3,800 to $1,700 and mean net worth went from negative $5,800 to negative $4,300, which is to say that the average family in the bottom quarter of the United States owes more than it owns and has done so in both surveys [2]. Among families in the top decile of income, median net worth rose 31 percent to $3,669,900; among families in the top decile of wealth, the median rose from $4,152,700 to $4,935,200 and the mean reached $8,975,200 [2]. Families in the lowest fifth of income saw their median and mean net worth fall 6 and 4 percent while families between the 40th and 80th percentiles gained between 5 and 15 percent, a pattern the Fed contrasts with the broad-based gains of 2019 to 2022 [2]. Families younger than 35 saw median net worth fall 23 percent, to $33,000 [2].

Ownership of the instruments that compound explains the divergence. Eight percent of families in the bottom half of the income distribution owned a private business in 2025, against 40 percent of families in the top decile [2]. The homeownership rate held at 66 percent and the median owner's equity rose from $218,900 to $230,000, but the survey's own affordability measure, the median home at more than 4.5 times median family income, sat near its historical low, so the gain accrued to those already inside the door [1][2]. A balance sheet that holds a house, a business and a portfolio rode three years of asset inflation upward. A balance sheet that holds a car loan, a card balance and a deposit rode the same three years sideways or down, and the median family in the bottom quartile ended them with seventeen hundred dollars.

Late: One Family in Five Behind on Its Payments

The most consequential pages of the survey concern not what families own but whether they can pay. In 2025, 19.6 percent of families reported being late on a loan payment in the preceding year, up from 12.2 percent in 2022, and the Fed's own summary states that families "were more likely to be behind on their financial obligations than at any point since the 2010 survey" [1][2]. The series across the last five surveys reads 14.9, 13.5, 12.3, 12.2 and 19.6 percent; the share late by sixty days or more rose from 4.9 to 8.2 percent; and the share of families whose debt payments exceed 40 percent of their income, the Fed's threshold for a particularly heavy burden, rose from 6.5 to 8.6 percent, "a level last seen in the 2013 survey" and the largest since 2010 [1][2]. The median debt payment-to-income ratio rose two points to 15.4 percent and the aggregate debt-to-income ratio from 89.4 to 94.9 percent, while the leverage ratio, debt measured against assets, fell to 26.9 percent because asset prices rose; the Fed's authors note that families "may have money tied in illiquid assets that may not easily be drawn down to finance debts" and conclude that "some families may be increasingly less likely to be able to service their loans" [2].

The cost of carrying a balance explains part of the arithmetic. The bulletin records that the average credit card interest rate rose from 14.6 percent to 21.4 percent between the two surveys; 44.7 percent of families carried a card balance after their last payment, the median balance rose to $3,100 and the mean rose 19 percent to $8,000 [2]. The Fed's G.19 release for August 2026 puts the rate on accounts assessed interest at 22.36 percent [6]. Twelve percent of families carried a balance on a buy-now-pay-later plan, up from 7 percent, with a median balance of $400 [2]; the Fed's separate Survey of Household Economics and Decisionmaking found that 16 percent of adults used such plans in 2025 and that 11 percent of those users had a payment trigger an overdraft or insufficient-funds fee at their bank [7].

The credit-bureau data confirm the survey. The Federal Reserve Bank of New York's report for the second quarter of 2026 records household debt of $18.8 trillion, card balances of $1.26 trillion, 4.7 percent of all balances in some stage of delinquency, and annualised flows into serious delinquency of 6.97 percent on cards and 3.00 percent on auto loans, which the bank's economic policy advisor described as "elevated levels" that it would "continue to monitor" [8]; a companion Liberty Street Economics note reconciles the bureau and lender measures of card distress [9]. The bank's Survey of Consumer Expectations adds the forward view: in September 2026 households put the mean probability of missing a minimum debt payment in the next three months at 12.2 percent, at 17.4 percent for households earning under $50,000 against 8.3 percent above $100,000, and the December 2025 reading of 15.3 percent was the highest since April 2020; 42 percent of respondents said they were worse off than a year earlier [10]. The household survey's broader measures read the same way: 73 percent of adults said they were doing at least okay financially, 63 percent could cover a $400 emergency expense with cash or its equivalent, down from 68 percent in 2021, and 16 percent had not paid all of their bills in the previous month [7].

One line of income in the ledger is protected by formula. In October the Social Security Administration set the 2027 cost-of-living adjustment at 3.6 percent for 76 million beneficiaries [11], on a September consumer price report that showed prices 3.7 percent higher than a year earlier, with energy up 20.4 percent [12]. Indexed benefits keep pace by statute. Wages keep pace only by bargaining, and the sections that follow measure how that bargaining has gone.

Who Owns the Market: Nine Tenths of Equity in One Tenth of Households

The recovery since 2022 was an asset recovery, and the survey records who held the assets. Participation in the stock market, direct and indirect, fell from 58 to 56 percent of families, a decline the Fed attributes to "families outside the top 10 percent of the income distribution", while the median holding among those who still owned stock rose 36 percent, from $56,900 to $77,400 [1][2]. Retirement-plan participation held near 65 percent [1]. The Distributional Financial Accounts give the aggregate: in the second quarter of 2026 the top 1 percent of households owned 50.9 percent of corporate equities and mutual fund shares, the top 10 percent owned 88.1 percent, and the bottom half of the country owned 0.6 percent; the same bottom half owed 51.6 percent of consumer credit, against 13.5 percent for the top decile [5]. Zandi's rounded figure, that the top 10 percent "account for 90% of those equity holdings", is the Fed's own number to within two points [4].

The arithmetic of the last three years follows from those shares. The doubling in the value of listed equity that Zandi describes accrues almost entirely to the tenth of households that owns nine-tenths of it; a rise in the cost of consumer credit falls almost entirely on the half of households that owes half of it. The two speeds are not an accident of the cycle. They are the direct consequence of who holds the asset that appreciated and who holds the liability that repriced, and a policy that lifts asset prices while holding borrowing costs high will reproduce them in every subsequent survey until the ownership of the assets changes.

Ninety-Four Versus Thirty-Three: Productivity and Pay Since 1979

The wage leg of the ledger is the oldest story in the series. The Economic Policy Institute's productivity-pay tracker measures net productivity and the hourly compensation of the typical worker from the same base: from the fourth quarter of 1979 to the second quarter of 2026 productivity rose 93.7 percent and typical compensation 32.7 percent, an annual average of 1.4 percent against 0.6 percent, where the three decades after 1948 had delivered 2.5 percent against 2.1 percent [13]. The difference between those two lines is the share of the output of American labour that has gone somewhere other than American labour for forty-seven years, and the wealth tables above show where it went.

The current reading is no kinder. The Federal Reserve Bank of Atlanta's Wage Growth Tracker put median nominal wage growth at 3.9 percent in September 2026; by quartile of the wage distribution, on a twelve-month average, the lowest-paid quarter of workers saw 3.4 percent, the second 4.0, the third 3.9 and the highest 3.8, and workers who stayed in their jobs saw 3.4 percent against 4.7 for those who changed them [14]. Against consumer prices 3.7 percent higher than a year earlier, the lowest-paid quartile's wages and the typical job-stayer's are losing ground in real terms [12][14]. The Census Bureau's income report records the same tilt at the extremes, a 1.7 percent gain at the 90th percentile and no significant change at the 10th [3]. The official poverty rate fell to 10.2 percent, 34.5 million people, and the Bureau's supplemental measure stood at 13.1 percent, with Social Security alone lifting 28.8 million people above that line [15], which is to say that the floor under the bottom of the distribution is a transfer, not a wage.

The Fading Dream: From Nine in Ten to One in Two

The consequence of a half-century of that divergence has a measure, and it was published in Science in 2017 by Raj Chetty and his colleagues under the title The Fading American Dream. Comparing children's household income at thirty with their parents' at the same age, they found that the share of children who out-earned their parents fell from about 90 percent for the cohort born in 1940 to 50 percent for those born in the 1980s, and from 95 to 41 percent when sons are compared with fathers [16]. Their decomposition is the part that matters for anyone designing a model. Restoring the growth rates of the 1940s and 1950s while keeping today's distribution would raise absolute mobility only to 62 percent; keeping today's growth and restoring the distribution of the 1940s would raise it to 80 percent, reversing more than two-thirds of the decline; under the current distribution the economy would need real growth above 6 percent a year to return to mid-century mobility [16]. Growth alone cannot repair the dream because a large fraction of growth now goes to a small fraction of households. The survey's youngest families, whose median net worth fell 23 percent in three years, are the cohort whose odds that study measured [2].

A ledger that looks like this one has a history, and the history is not reassuring. The white paper that underlies this catalogue, Reform Before Rupture, builds its case on Walter Scheidel's survey of inequality from the Stone Age to the present, whose finding is exact: "Four different kinds of violent ruptures have flattened inequality: mass mobilization warfare, transformative revolution, state failure, and lethal pandemics" [17]. No peaceful reform of comparable size appears in that record. The paper's argument is that concentration on the scale now recorded erodes the growth, the credit and the political consent on which the returns to capital themselves rest, and the survey supplies the measurements that argument needs: a bottom half with 2.3 percent of net worth and half of consumer credit, a top percentile with a third of net worth and half of listed equity, and one family in five behind on its payments.

The precedents this catalogue has already examined describe the end-state. The Silent Fracture showed that Americans systematically underestimate how concentrated wealth is and would choose a far flatter distribution if asked; the gap between the ledger and the belief is where consent erodes first. The Liquefaction of Forty Trillion traced what a sovereign does when its arithmetic no longer closes, in Buenos Aires in 1975, 1982, 1989 and 2002 and in Washington in 1933 and 1942, and showed that the inflation which pays down a public debt is itself a transfer from the households that hold currency and wages to the institutions that hold assets and issue claims. A two-speed ledger is the condition under which that transfer stops being tolerated. None of the precedents required a majority to understand the mechanism. They required only that enough households notice that the terms had changed without their consent, and the survey suggests that a fifth of American families have had that notice delivered by their lender.

The SAVI Reading: The Mechanism That Moves the Median

The model this firm commits to write into the legal terms of every fund it raises is built for the ledger the survey describes, and its first tenet is the one mechanism in this catalogue that moves the median rather than the mean. Under Tenet 1 of The SAVI Capital Model, fifty percent of a portfolio company's net profit is distributed in cash to all of its employees, in equal shares, separate from salary, encoded in the fund's governance documents rather than left to a board's discretion each cycle. The distinction between that instrument and every other remedy in circulation is the line on which it lands. A rising stock market reaches the median family, if at all, as a retirement balance it cannot spend today, and reaches the bottom half of the country, which owns 0.6 percent of equities, not at all. A profit pool paid to the people who work in the company reaches the family in the bottom wealth quartile as a deposit, and a deposit is the one asset that retires a card balance carried at 22 percent.

The white paper's worked example states the cost and the scale without rounding in the model's favour. An illustrative $100 million fund whose companies earn five times their cost returns its investors 3.84 times their contributions under conventional terms and 2.38 times under the model, and in the same example $200 million reaches the Human Capital Pool, or $400,000 per full-time employee over five years at the paper's illustrative headcount of five hundred [17]. The paper claims no outperformance for the model and says plainly that no study yet tests sharing at half of profit. Set the illustration against the survey: a median family net worth of $215,900, a bottom-quartile median of $1,700, a median card balance of $3,100. A distribution of that order does not nudge the mean. It moves the household from the second economy into the first, and it does so through the ledger rather than through the tax code, by a term the investor signed rather than a vote the legislature took. The remaining tenets complete the structure: a maximum ratio between the chief executive's pay and the lowest-paid worker's, written into each company's charter; stewardship as a governance accountability standard; and, under Tenet 4, the direction of every dollar above five times investors' total contributions, measured on net distributions, to The SAVI Ministries Endowment, a term as binding as the preferred return.

The survey of 2025 will be followed by the survey of 2028, and the shape of that ledger is being decided now, in the terms on which capital is raised. Consent in a market economy is not a sentiment but a function of the ledger, of who owns the asset that appreciates and who owes the liability that reprices, and the ledger is a design choice. The two speeds are one design. The model is the other.