Chapter 1: The Household Balance Sheet
A claim at a date
Wealth is often described as something a person has. A household balance sheet asks a stricter question: which claims can the household make, which claims can others make against it, and at what date are those claims valued? A home may provide shelter for decades, but a valuation in a survey is a statement about its estimated market value at an interview, not the number of years its occupants can remain. A pension account may be named as an asset, but an expected future public pension is a different kind of promise. A small business can give its owner both income and authority while being difficult to sell. The word *wealth* does useful work only if we say which of these things is being counted.
In the Federal Reserve's 2025 Survey of Consumer Finances report, family net worth is the difference between gross assets and liabilities. The associated one-page category map separates financial from nonfinancial assets, then sets total debt against their sum. It lists a primary residence, other residential property, businesses, vehicles, deposit accounts, securities, retirement accounts and several classes of borrowing. That map is a guide to the published categories, not a substitute for the survey codebook's exact questions and instructions.[1-1] The arithmetic is simple. The judgment about what belongs in each term is not.
Consider a constructed example. Two households each live in a property currently estimated at $400,000. Household A owns its home outright. Household B has a $310,000 mortgage secured on an otherwise similar home. If neither holds other assets or debts, the relevant housing claims are $400,000 and $90,000. The houses look much the same from the street. Their owners' residual claims do not. If a price estimate falls by $40,000 while the debt balance is unchanged, A's housing claim falls by ten percent; B's falls by more than forty percent. That comparison assumes the estimates, debts and dates are correct. It says nothing by itself about either family's income, mortgage payment, ability to move or security of tenure.
The example also warns against an easy mistake. The house is an asset in the gross column; the mortgage is a liability in another column. If a writer begins with *home equity*—already the house value less home-secured debt—and subtracts the mortgage again, that writer has made debt appear twice. If the writer uses only the house price, the debt disappears. The error can change the story of who is wealthy without any household having changed its position.
The denominator changes the question
A household's net worth can be described as a dollar amount, a share of aggregate wealth, a rank among households or a multiple of income. Those measures answer different questions. The same household may have a high dollar value in a costly housing market and a low liquid balance. A young owner with a large mortgage can sit high in gross assets and much lower in net worth. A retiree with a modest pension account and no debt may have a low asset total but little immediate repayment pressure. A claim about *the distribution* needs a chosen measure and a reason for choosing it.
The mean and median make the difference visible. The mean divides all measured net worth by the number of families. The median is the middle family in the ordered distribution. Very large holdings can move the mean substantially without moving the median by the same proportion. In its October 2026 release of the 2025 SCF, the Federal Reserve reported real median family net worth of $215,900, up two percent from the 2022 survey, while real mean net worth rose seven percent to about $1.24 million.[1-2] These are survey estimates for two cross-sections. The figures do not mean that a typical family gained seven percent, or that the same set of families was followed from one interview to the next.
A policy argument can go wrong by sliding between the measures. Suppose, hypothetically, that an asset-price rise benefits a small group holding most of a class of securities. Aggregate household wealth may increase. The mean may increase. The median may barely move, while some households that own no securities see no direct capital gain. Yet even that account is incomplete: a rise in share prices can affect pension funds, employer investment, tax receipts and future returns. The question is which channels matter, for whom and on what timetable. A distributional measure locates a change; it does not supply the entire causal explanation.
The rank can move without an asset moving. A family's measured wealth may stay constant while other families' asset values rise or fall. It may also change households: marriage, separation, adult children leaving home and older relatives joining a household alter the unit in which claims are counted. A chart showing the wealth of households in an age band at two dates is not a diary of the same people growing older. The 2025 SCF report notes that the share of families headed by someone at least 75 increased from 12.0 percent in 2022 to 13.4 percent in 2025.[1-3] That compositional change deserves attention before a difference between age groups is attributed wholly to saving or investment skill.
Value without a sale
A market value is useful because it offers a common unit for unlike claims. It is also provisional. A quoted security may have a recent trade. A home may have comparable sales but no buyer for that particular house at the estimated price. A private business may depend on its owner, a supplier relationship or a contract that cannot be transferred without consent. A household does not need to sell every asset for the balance sheet to matter; nevertheless, the estimate should not be read as cash available by Friday.
Return to the constructed households. B has $90,000 of estimated home equity. If it must move urgently, an agent's fee, repairs, legal costs and the need to accept a lower bid can reduce the amount left after repaying the mortgage. A can also incur those costs, but it has a larger cushion. Conversely, B may have stable income, a manageable fixed-rate payment and no intention of selling. The lower net claim does not prove a current hardship. The asset gives both households shelter; the debt determines a different exposure to a price decline. Wealth, liquidity and living standard meet but do not collapse into one number.
There is a countercase to the idea that the largest balance always gives the strongest position. Imagine A's debt-free property requires major repairs and sits in a place where finding a buyer may take a year. B's mortgaged property is close to reliable employment, and B holds a cash reserve outside the house. If an unexpected bill arrives next month, B may have more usable cash and a better income path despite lower recorded housing equity. This does not erase the net-worth difference. It establishes why a balance-sheet comparison should name the decision at issue: borrowing, staying housed, enduring a job loss, investing in a business or leaving an unsafe arrangement. Each decision demands a different part of the balance sheet.
Some assets are valuable precisely because they need not be sold. A dwelling provides a stream of housing services. A small firm may be a source of work, local relationships and discretion over time. Retirement savings may be protected from a decision that would consume them too early. Recording their estimated values allows comparison, but it can hide the rules that make those values usable. A deed, a loan covenant, a vesting rule and a tax provision do not add the same kind of dollar to a household. They determine what the household can do with the dollar it appears to own.
A transaction ledger behind the snapshot
A balance sheet is a stock, but decisions arrive as transactions. To understand how the stock changes, follow one household through a deliberately constructed month. At the opening date it has $8,000 in a transaction account, a vehicle valued at $12,000, and a $5,000 vehicle loan. Its net worth from these items is $15,000. The family earns $4,000 after tax during the month, spends $3,000 on current needs, repays $500 of loan principal and pays $50 of interest. If no other values change, the cash balance becomes $8,450, the vehicle debt $4,500, and net worth $15,950. The increase is $950: the $1,000 gap between income and consumption less the $50 financing cost. The $500 principal payment moved value from cash to reduced debt; it did not, by itself, create an extra $500 of net worth. This is an accounting illustration, not a transaction observed in the SCF.
Now change one assumption. If the vehicle's estimated sale value falls from $12,000 to $10,500 over that month, closing net worth becomes $14,450. The family saved from current income in the cash-flow sense but ended with less measured net worth than it began. If a writer says the family “did not save,” the writer has confused an asset-price change with the income decision. If a writer says the family is better off because its loan fell, the writer has ignored depreciation. These are separate entries in the bridge from one date to another. In real data, missing price observations and changing household composition make the bridge harder, not less necessary.
The constructed ledger also exposes a timing mismatch in published statistics. The SCF report's Appendix A explains that a family's income before taxes is generally collected for the calendar year before the survey year, while most balance-sheet positions refer to the interview date.[1-4] A 2025 interview can thus pair a wealth stock at the interview with income earned during 2024. The report converts dollar values to a common 2025-price basis for comparison, but conversion does not make the reference periods identical. A debt-to-income ratio built from those entries is useful under its stated convention. It is not a live monthly underwriting file. A family that changed jobs between the income period and the interview can have a ratio whose numerator and denominator come from different economic circumstances.
The dates matter even within a survey wave. Appendix B states that most 2025 interviews occurred between April and December 2025, while about one-seventh took place from January through May 2026.[1-5] A chart labeled “2025 survey” is therefore not a single closing balance sheet for December 31, 2025. The survey organizes interviews into one wave and reports standardized estimates. If a market moves sharply during fieldwork, families can answer under different prices. A comparison with a national-account quarter should align the timing concept before treating a gap as a disagreement about ownership. This does not discredit the survey. It tells us what event its value is closest to: an interview, not a fiscal-year end.
The ownership audit
The next question is whose claim enters the ledger. A family can live in a house without owning it. An adult can work in a business without holding equity. A person can be promised a future benefit without having an account that can be transferred today. Conversely, a family may own a share in an asset it cannot occupy or manage. Living in, using, controlling and owning are different verbs. A distributional account should state which one the data measure before treating a household's environment as its property.
Joint ownership complicates the question without making it impossible. Imagine two siblings who jointly own a rental property. One lives elsewhere and manages tenants; the other receives half the net income and has a legal right to half the sale proceeds. If the property is worth $300,000 and carries $100,000 of secured debt, total equity is $200,000. It would be wrong to assign $200,000 to each sibling's family. It may also be wrong to split the value exactly in half if their legal shares, guarantees or sale restrictions differ. The hypothetical requires a title record, debt agreement and household-unit rule. A survey respondent's answer is a measurement of an economic position, not a substitute for reviewing every legal instrument.
The SCF's own category map helps avoid a related double count. It lists “net equity in non-residential real estate” as a nonfinancial asset category and displays private business interests alongside it, while primary-residence value and secured home debt appear separately in the asset and liability branches.[1-1] The map is explicitly a map of Bulletin extract variables and directs readers to the documentation and programs for precise definitions. If an analyst copies a net business-equity number, adds the firm's gross assets and then subtracts its debt, the same enterprise appears in several places. The balance-sheet identity cannot protect against a mislabeled input. The analyst must ask, for each line, whether debt has already been netted and whether an asset is held through another asset category.
Consider a family business that owns its work premises. In a hypothetical set of accounts, the company has equipment worth $200,000, premises worth $300,000, and $250,000 of company debt. The family's equity in the company is $250,000 before other adjustments. A family-level survey may report the net equity as a business claim. Adding the premises once more to the family residence or real-estate column would create wealth from bookkeeping. The opposite mistake is to subtract the $250,000 business loan from a net $250,000 business-equity figure, leaving the family's claim at zero. Actual survey treatment depends on the legal ownership route and the question wording; the illustration only shows why the route must be checked. Chapter 4 returns to the valuation of the business itself.
Some claims sit at the edge of ownership because they depend on a future event. A life-insurance death benefit may be large, but it is not the same as the policy's current cash value. The SCF report's Appendix A counts a policy's nonzero current cash value as an asset and excludes term life insurance, which offers only a death benefit.[1-6] This is a coherent balance-sheet choice. It does not mean the term policy has no protective value to survivors. It means a current net-worth inventory and an insurance-needs analysis answer different questions. A family may be financially fragile today despite a future payout to beneficiaries after a death. Calling the death benefit current wealth would erase that distinction.
A pending lawsuit, an expected inheritance or an employee option presents a similar problem with different rights. The claim may be valuable, contingent, disputed, nontransferable or hard to price. Appendix A includes proceeds from lawsuits or estates in settlement in the SCF's “other financial assets” category, but generally excludes employment-related stock options before exercise because valuation under trading or exercise constraints requires strong assumptions.[1-6] A lawyer or beneficiary may reasonably consider a likely future payment in planning. A standard balance sheet needs a rule about when that expectation becomes a measured asset. If two datasets use different rules, their “wealth” levels will differ before any economic behavior does.
The ownership audit should therefore have four lines, not one. What legal claim exists? Who has the economic benefit? Can the claim be transferred or pledged? What valuation is used at the measurement date? A household may own a claim with poor liquidity; it may use an asset without owning it; it may control a business while holding only part of its economic equity; it may expect a benefit whose present value is not in the inventory. The map of claims is still worth making. Its power comes from naming these limits rather than pretending they disappear in a total.
What a valuation date can and cannot settle
A valuation date makes a comparison possible but not frictionless. A listed share can be marked to a closing price, though a large block might not sell at that price. A deposit usually has a stated balance, though access or insurance limits matter. A house is often valued through a respondent's estimate or comparable sales. A small business may have no recent arm's-length transaction at all. The SCF report's category descriptions are tied to survey questions and interviewer instructions, and Appendix A calls its summary a general exposition rather than the full codebook.[1-6] Treating every line as equally precise can create a false sense of certainty about the sum.
An exact estimate would also not make the household equally able to use each dollar. Suppose, hypothetically, a family has $250,000 of equity in a closely held firm and $250 in cash. Another has $50,000 in deposits and no firm. The first has higher measured wealth; the second can pay a surprise $5,000 bill without financing or sale. The firm owner may have a credit line, or may be able to take a distribution, but either action depends on contracts and business cash flow. An asset value is a claim at a date, while liquidity describes a path from the claim to a payment. The path may be quick, costly, conditional or closed.
This is why one should not call a balance sheet a complete welfare measure. It omits capacities that cannot be sold, such as education and relationships, while counting assets that may be essential to shelter or work. It may omit public pension streams and include account balances subject to withdrawal limits. It can identify material vulnerability, especially when debt overwhelms assets, but it does not automatically describe every family's risk. The right use is more demanding: retain the stock measure, then add the contract, cash-flow and institutional facts needed for the decision at hand. The arithmetic gives the first coordinate, not the destination.
The boundaries of the ledger
The SCF category map is a useful discipline because it forces the analyst to place each claim somewhere. It also exposes exclusions. The flowchart includes quasi-liquid retirement accounts, such as account-type plans and individual retirement arrangements. It does not mean every future retirement promise is the same kind of account balance. OECD's experimental national-accounts-aligned wealth exercise, for example, excludes employment-related pension entitlements from its core 2025 concept and explores their inclusion separately for selected data. The report cautions that adding them can change not only measured totals but the ranking of households.[1-7] A comparison between that concept and SCF Bulletin net worth needs a bridge. Calling both series “household wealth” is not enough.
The boundary matters when the public argument turns to inheritance. A parent may help pay rent, transfer a business share, guarantee a loan or leave a taxable estate. These are not identical events in the recipient's balance sheet, and not all appear in an estate-tax filing. The Internal Revenue Service's estate tables cover returns filed in a calendar year; the agency notes that those returns can concern people who died in several different years.[1-8] Filing-year totals cannot stand in for all deaths or all gifts. A book about wealth distribution must follow the legal and statistical route by which a transfer becomes visible before it claims to measure the route's scale.
This chapter begins with arithmetic because imprecision here spreads. Gross assets minus liabilities produces net worth under a specified definition. A rank or share locates that result among other measured households. Neither operation says why the claims arose, how securely they are held or who can convert them into an action. The next chapter asks how the people and claims enter the dataset at all. That is where a distribution becomes an estimate rather than a complete list of owners.
Notes and Sources
[1-1] Board of Governors of the Federal Reserve System, *Changes in U.S. Family Finances from 2022 to 2025: Evidence from the Survey of Consumer Finances* (October 2026), Table 2 note and Appendix A, https://www.federalreserve.gov/publications/files/scf26.pdf; Board, *Definition of SCF Bulletin Asset and Debt Categories in Calculation of Net Worth*, one-page category flowchart, https://www.federalreserve.gov/econres/files/Networth%20Flowchart.pdf. Read the report's relevant net-worth/table and category passages and the flowchart, not the complete report or variable codebook. The constructed household calculations are illustrations, not survey observations.
[1-2] Board of Governors, press release announcing the 2025 SCF, October 9, 2026, https://www.federalreserve.gov/newsevents/pressreleases/other20261009a.htm. The release provides the real median and mean changes and levels; the report's standard errors and sample design must be considered for inferential comparisons. Read the release's key findings and selected report methods, not all linked statistical tables.
[1-3] Board of Governors, *Changes in U.S. Family Finances from 2022 to 2025*, Appendix A, Table A.5 (age distribution of reference person), and net-worth discussion. The age-band comparison is cross-sectional and not individual follow-up.
[1-4] Board of Governors, *Changes in U.S. Family Finances from 2022 to 2025*, Appendix A, report pp. 39–41. Directly read the preceding-year income reference period, interview-date balance-sheet description and inflation convention. The monthly ledger here is explicitly hypothetical.
[1-5] Board of Governors, same report, Appendix B, report p. 53. Directly read the 2025 fieldwork-date description; no adjustment for within-wave asset-price movement is estimated.
[1-6] Board of Governors, same report, Appendix A, “Asset and Liability Categories,” report pp. 41–44. Directly read cash-value life insurance, other financial assets, stock-option and net business/real-estate category descriptions. The exact 2025 codebook and survey variables were not reprocessed.
[1-7] Organisation for Economic Co-operation and Development, *Distribution of Household Wealth in Line with National Accounts* (PDF displaying © OECD 2026), pp. 14–16 and 42–44, https://www.oecd.org/content/dam/oecd/en/publications/reports/2025/12/distribution-of-household-wealth-in-line-with-national-accounts_ec97b58e/65d3d98c-en.pdf. This is an experimental, abridged cross-country concept; full methodology and country tables remain to be read before broader use.
[1-8] Internal Revenue Service, *SOI estate tax data tables, filing years*, https://www.irs.gov/statistics/soi-tax-stats-estate-tax-filing-year-tables. Read the page's coverage note and table inventory, not the XLSX values. Filing returns do not represent all transfers.