October 3, 2026: car-payment scenario scope and counts
Scenario Lab previously counted positive SCF vehicle-payment groups as individual loans and included other-vehicle financing, such as aircraft, in an ordinary car-loan counterfactual. Some source fields aggregate payments on several vehicles, so those counts were not reliable exact-loan counts. In one sparse renter match, aircraft financing supplied the only donor and produced an implausibly large car-payment scenario. These were official public source values, including imputed payment amounts and frequencies, not proof of those exact real-world bills.
The revised scenario uses personal cars, trucks, vans and SUVs and labels its counts as positive payments, not all outstanding loans. It retains a residual-car aggregate as one payment only when the source’s owned-car count identifies exactly one remaining car; otherwise the positive aggregate has an unknown payment count and is excluded. Nonempty correctly scoped samples remain visible with their actual counts and warnings. Historical Norms and the full SCF debt breakdown retain the broader observed/imputed all-vehicle totals, explicitly including other vehicles. No payments are capped or rewritten to make the result look plausible. Federal Reserve construction definitions.
October 3, 2026: SCF weight-concentration diagnostics
Wealth and debt comparisons previously showed raw family counts but omitted effective sample size and the largest family’s survey-weight share, so the interface could miss low-precision warnings when at least 50 families were selected. New core and supplement releases add those diagnostics, counting each family once across its imputed records. Existing medians, quartiles, confidence intervals, population estimates and participation context are unchanged. Large neighboring-age median changes can reflect a different selected distribution, not an abrupt life-course change.
October 3, 2026: visible samples, consistent parent groups and valid denominators
The previous CEX sample cutoff hid nonempty, ordinary reader scenarios. Size cutoffs also hid calculable housing and loan-donor points. The new release retains their actual household counts and point estimates, warns below 50 independent households, and flags low effective samples or concentrated survey weights. SCF balance comparisons and observed principal tables follow the same nonempty-point policy. Established preferences for larger donor pools remain in place when a candidate meets them; otherwise the closest nonempty declared pool is retained and its scope is disclosed. A confidence interval can remain unavailable when a replicate is empty; that no longer removes an otherwise calculable point estimate. Genuinely empty or missing required component sources remain unavailable, not zero.
Singleton donor uncertainty is unavailable rather than measured zero. A two-borrower scenario now discloses both source pools, including a small-sample warning for an additional-borrower pool even when the first pool is larger.
Parent selections previously used the SCF’s count of children of all ages, unlike the minor-child boundaries in the spending data. Child ages recovered from the saved full public files now distinguish minors from adults in all three waves. This changes parent-group wealth, debt and modeled cash-flow results. The surveys still differ in economic units and marital status; the methodology describes those differences.
Housing spending share previously replaced very low or nonpositive income with an invented $100 monthly denominator. It now uses positive income only, with eligible counts and excluded shares. Ratios above 100% remain possible and are not capped. Survey-weight concentration diagnostics now sum a household’s interview weights rather than averaging them. Personal-budget income-dependent percentile grids now use the same five-imputation ordering as their headline medians. Spending means dominated by a few unusually large gifts or expenses carry concentration notes; the source values are not winsorized to look plausible.
Prior immutable releases remain in the research archive. These are prelaunch data corrections, not a claim that 2022 observations describe conditions in 2026.
October 3, 2026: education beneficiaries are not legal borrowers
The student-loan scenario previously claimed the SCF identified legal adult borrowers. Its actual question asks whose education was financed. The controls now describe modeled debt financing one or two adults’ education, with an explicit education-beneficiary proxy limitation. The source cannot establish Parent PLUS or cosigned borrower responsibility. Debts financing others’ education are outside the proxy but remain in Historical Norms’ full household payments; the custom combined-payment option can represent them. The payment calculations are unchanged.
September 22, 2026: SCF population counts
The population labels for the SCF wealth and debt measures divided the public summary-extract weights by five a second time. The Federal Reserve already makes that adjustment in WGT, so the displayed population counts were one-fifth of their correct size. The corrected calculation sums WGT across the five implicates once. For example, the 2022 age-35 all-household cohort represents approximately 10.6 million families, not 2.1 million.
Releases scf-2026-09-22-r7 and scf-2026-09-22-supplement-r2 correct the population metadata. Medians, percentiles, confidence intervals, independent sample counts, and suppressed cells are unchanged. The earlier immutable releases remain in the research archive. Source weight definition.
September 22, 2026: income types, taxpayer ages and retirement contributions
The previous personal-tax input treated all CEX income as wages and omitted taxpayer ages. The corrected specification separates wages, self-employment, Social Security, pensions, transfers and other income, preserves reported business/property losses, and supplies taxpayer and spouse ages. The component sums reconcile with every retained source income total across all five imputations. Food-stamp benefits remain nontaxable resources, not unrestricted cash.
The earlier retirement calculation smoothed and calibrated combined pension and Social Security deductions before subtracting unadjusted payroll deductions. That could produce a contribution for a household reporting none. Direct retirement-contribution fields are now smoothed and calibrated separately, leaving payroll deductions unchanged.
These changes affect modeled after-tax income and Breathing Room, released as cex-2026-09-22-r20 and bri-2026-09-22-r30. The latter also tests alternative income-taxability assumptions and retained source spending shapes. The revisions do not turn the residual after spending into a measured saving rate. Accounting and tax method.
September 22, 2026: SCF participation context and debt-payment definition
Core release scf-2026-09-22-r8 and supplement scf-2026-09-22-supplement-r3 add participation shares, conditional medians and housing-tenure composition with sampling and imputation uncertainty. Original medians, percentiles, intervals and suppression remain unchanged. Participation shares and conditional medians show how a zero all-household debt median can coexist with large balances among borrowers.
Core release scf-2026-09-22-r9 then corrects the debt-payment label and definition. The source ratio includes reported and imputed payments and uses a positive floor for very low or nonpositive monthly income. It is a share of prior-calendar-year before-tax income, not legally required bills divided by disposable income. Numerical values are unchanged. SCF source definitions.
Correction policy
Each entry will include the date, affected pages or estimates, the original issue, the corrected treatment, and whether historical values changed. Metric and variable-mapping versions will remain available so readers can identify which definition produced a result.
Correction record
There are no material corrections in the registry.