Income, debt payments and the money left over

Reference people around age 35. Three historical periods, not the same families followed over time.

Cash-flow estimates pool CEX interviews from 1988–1990, 2006–2008 and 2021–2023, selecting the birth cohort centered on each comparison year. SCF balances and payment breakdowns use the individual 1989, 2007 and 2022 surveys. Nonempty point estimates remain visible with their independent-household counts; below 50 they are low precision. Missing required sources remain unavailable, not zero. An empty replicate leaves the confidence interval unavailable without removing the point estimate. An available SCF result does not make a small CEX result precise.

The 1989 cash-flow pool is restricted to complete-income respondents; later periods use five income imputations. Its weighted population describes that eligible sample, not all U.S. households. The displayed nonpositive-income exclusion is a separate ratio rule, not the percentage of missing income responses. This historical response difference remains a model limitation.

Ages describe the household reference person, not every person of that age. Young adults living in a parent's consumer unit generally appear under the parent's age. Negative remainders can reflect borrowing, asset drawdowns, transfers, income/spending timing or model assumptions; they are not measured delinquency rates. Results at retirement ages do not measure whether accumulated wealth can support spending.

Uncertainty: Intervals include CEX replicate-weight and five-income-imputation uncertainty, conditional on assigned donor values. They exclude donor/model uncertainty; dependence sensitivity is separate. Singleton recipients or required donor sources retain points but have unavailable sampling intervals. No generational ranking is supported by this extension alone. The original breathing-room model's wider sensitivity analysis still applies. CEX income includes restricted food benefits; it is not all unrestricted cash.

What is counted as a debt payment?

These are weighted mean payments across all selected SCF families, including nonpayers, not median borrower bills. The categories add to total payments before rounding. Their 95% intervals use five implicates and 999 replicate weights. These are a different sample from the CEX cash-flow households above.

SCF CCPAY assumes 2.5% of the outstanding revolving balance; not an observed monthly card bill or universal contractual minimum. Other revolving payments include applicable buy-now-pay-later amounts; the remaining consumer category includes pension, insurance, margin and miscellaneous source-classified loans. The 1989 education payment is reconstructed from education-purpose installment loans and removed from other installments to avoid double-counting. Some source amounts are modeled; these are not all observed contractual minimums. Normal-approximation intervals can extend below zero; that represents estimation uncertainty, not a negative payment.

2022 student-payment caution: This survey overlaps the federal payment pause. A few large positive reports can substantially affect the mean; it is not a typical required federal-loan bill. The source does not establish whether each payment is private, voluntary or otherwise outside the pause. The historical-cost details in the Explorer show payer participation, payer medians and outlier influence.

How much does matching matter?

The broader-debt result assigns additional SCF payments to CEX household records. The default preserves the payment distribution with independent assignment. Matching larger payments to higher or lower incomes tests dependence. This is a sensitivity range, not a confidence interval and not the full model-risk envelope.

1989: effective sample n=201.93; largest family weight 1.34% . 2007: effective sample n=285.69; largest family weight 0.61% . 2022: effective sample n=201.04; largest family weight 1.15% .

1989: 313 additional-payment source families (exact_age_household_tenure). 2007: 331 additional-payment source families (exact_age_household_tenure). 2022: 305 additional-payment source families (exact_age_household_tenure).

These ratios do not preserve each SCF family's observed debt-to-income relationship. Use the within-SCF before-tax debt-payment ratio for that relationship. Neither the synthetic ratio nor the national aggregate below is interchangeable with it.

The remaining debt-balance categories

These three categories complete the balance breakdown alongside education, vehicle and credit-card debt. Their household-level amounts add to total debt, but their medians do not. Nonempty SCF samples remain visible with counts. Small samples below 50 are unstable. Five implicates and 999 replicate weights inform intervals; an empty replicate leaves the interval unavailable without removing the point estimate. Normal intervals can extend below zero; that is approximation uncertainty, not negative debt.

Weight diagnostics describe survey-weight concentration, not a replacement for the replicate-weight intervals. Small effective samples and a family weight above 2.5% are flagged even when the household count exceeds 50.

Why the consumption comparison differs

The consumption-basis measure counts purchases when acquired, including cars and other durable goods, instead of counting the later principal payments again. It includes flexible spending that the headline breathing-room measure leaves for a second layer. It uses the original CEX expenditure total instead of replacing housing with another survey. Those differences, pension saving and cash gifts all contribute to the gap between these two medians, so the gap is not the amount of principal repaid.

A principal payment reduces a liability; it still requires cash. The homebuying scenarios separately show principal, interest and upfront equity. These historical surveys do not provide a consistently observed principal/interest split for every outstanding loan; the consumption view uses the expenditure accounting basis instead of inventing that split.

Observed mortgage-principal outlays where recoverable

Reported principal outlays can include lump sums and home-equity loans, not only scheduled amortization. These are not contractual minimums and must not be subtracted from or used to split the separate AHS/ACS housing donor estimates. They reduce liabilities; interest does not. The complete 1989 pool lacks comparable respondent-level fields and remains unavailable. CEX FMLI source records, exact interview joins, consumer-unit-state means, 44 replicate weights. Summary-field definitions begin in 2000. No income-imputation uncertainty is added to observed expenditure fields. These are a separate CEX source view, not an additional deduction from breathing room. Zero overall medians include families making no such outlay; the mean captures the population-level amount. Amounts and interval endpoints are rounded to the nearest $10, so a displayed zero can also mean a small positive estimate. Negative normal-interval endpoints indicate approximation uncertainty, not negative principal repayment.

Download observed principal outlays and source-field mappings. The source fields sum previous- and current-calendar-quarter portions of the interview's three-month reference period, then divide by three; they are not six months of spending.

Independent national debt-service check

Annual average of four seasonally adjusted quarterly ratios of required aggregate debt payments to disposable personal income, old Fed methodology. It is a ratio of national aggregates, not a median household or an age-specific estimate.

The national historical debt-service ratio rises from 1989 to 2007 and falls by 2022. This supports checking a lower 2022 payment burden for plausibility, not equating cohort levels or proving a generational advantage. Use the archived historical methodology consistently in 1989, 2007 and 2022. The newer credit-bureau-based methodology begins in 2005 and is not spliced into this comparison. Federal Reserve source.

Sources and downloads

BLS expenditure definitions, Federal Reserve payment formulas, and the full methodology. Release cash-flow-2026-10-03-r2, extending bri-2026-10-03-r31.

Download cash-flow cohorts, definitions, intervals and source checksums. Download the additional debt-balance categories. Download national debt-service context.