How we built “What a typical household actually has saved for retirement — and why the “average” is so misleading” — the metric, the sources, the limitations, and how to reproduce every number.
The metric
Retirement-account balances by age of household reference person, from SCF 2022 RETQLIQ (IRA/Keogh + account-type pensions): participation rate, UNCONDITIONAL median (all households, zeros included), CONDITIONAL median (account-holders), and mean (account-holders). The finding is the gap between a commonly-cited 'average' and a typical household's holdings, decomposed into mean-vs-median skew and population coverage.
Breakdown dimension: age of household reference person (SCF AGECL brackets). Data vintage: Federal Reserve Board, Survey of Consumer Finances (SCF) 2022, Summary Extract Public Data (rscfp2022.dta) (2022); Vanguard, How America Saves 2025 (2024).
Sources
- Federal Reserve Board, Survey of Consumer Finances (SCF) 2022, Summary Extract Public Data (rscfp2022.dta)
Canopy Press analysis of Federal Reserve Survey of Consumer Finances 2022 microdata; Vanguard How America Saves 2025 cited for recordkeeper context.
How it’s computed
All figures are deterministic arithmetic on the official series above — no estimates, no modeling. Raw API responses are snapshotted at the stated vintage; every published figure must reproduce from those snapshots before publication (an automated audit gate recomputes the table, the chart values, and the CSV byte-for-byte, and blocks publication on any mismatch).
Limitations
- Two populations, labeled distinctly: the UNCONDITIONAL median (and participation rate) cover ALL households including the ~46% with no retirement account; the CONDITIONAL median and the mean cover account-holders only.
- The gap between the 'average' and a typical balance has TWO causes — mean-vs-median skew (a few very large accounts) AND population coverage (who is counted) — and is not attributed to either one alone.
- SCF figures are 2022 (latest SCF wave, nominal 2022 dollars); the Vanguard recordkeeper figures are year-end 2024 — different years, so the comparison is directional, not a like-for-like timing claim.
- SCF measures all households' retirement accounts (IRAs + account-type pensions); Vanguard measures balances of participants who hold an account on one recordkeeping platform. Different populations, presented as parallel same-direction evidence.
- SCF medians/means use the 2022 Summary Extract (RETQLIQ), all five imputation implicates pooled and weighted by WGT.
Reproduce it
Pull the series listed above from the official APIs at the stated vintage and recompute. Verification values (recomputed before publish):
| Check | Value |
|---|---|
| scf_participation_overall_pct | 54.3 |
| scf_median_uncond_overall | 4000.0 |
| scf_median_cond_overall | 87000.0 |
| scf_mean_holders_overall | 334097.0 |
| scf_skew_overall_x | 3.84 |
| scf_median_uncond_under35 | 0.0 |
| scf_median_uncond_75plus | 0.0 |
| scf_median_uncond_45_54 | 20000.0 |
| scf_median_cond_45_54 | 115000.0 |
| scf_participation_45_54_pct | 62.2 |
Accountability
Canopy Press research is produced by a methodology-driven, human-overseen pipeline. John Fox, Founder & Head of Research, is responsible for the data methodology and editorial standards behind this asset. Corrections: [email protected]. See our Editorial Standards and the research changelog.