An atlas of ordinary life

⚠ Decades before the 1960s published white-only male/female life tables with no all-races total — the placard carries the sexed figures and the chart renders the gap rather than a spliced concept.

⚠ Three trajectories on one axis: heart disease peaked early and fell steadily; cancer rose for decades before turning; stroke fell dramatically throughout. Cancer survival is charted separately (different unit and source family).

⚠ Different source family (SEER/NCI, not CDC NCHS) and different unit (% surviving, not deaths per 100,000) — not grouped with the mortality arcs.

⚠ The official poverty series begins in 1959 — earlier decades have no line.

⚠ The 1900s datum (5.31) is from the 1901 BLS wage-earner survey — 0.7 persons above the Census 1900 national average of 4.61; the survey and census describe different populations.

⚠ The 1940s point is 1947 (first postwar CPS tabulation), not 1940. These averages describe family size, not the composition of a particular household or the cause of changes over time.

⚠ CPS/F-8 counts all families (related persons living together). The 1940s point is 1947 — the first postwar CPS tabulation.

⚠ The 1980s source bridges 1978 central-AC and 1993 total-AC figures — no single-decade datum, so the decade renders as a gap.

⚠ 1970s–2020s: median square feet of new single-family houses completed (Census C-25/H-150 / Characteristics of New Housing). The 2020s point is 2024; the series is conceptually consistent across the whole span.

⚠ Nominal dollars — the CPI arc and affordability dashboard provide deflation context. No 2010s or 2020s point: Census Historical Housing Tables end at 2000; ACS rent data (Table B25064) would extend the arc in a future curation pass.

⚠ The 1900s source combines several technologies in one headline percentage, so its 5% is not plotted as a telephone datum.

⚠ 1910s–1930s sources counted telephones per 1,000 population, not households — those decades render as gaps rather than a unit splice.

⚠ The 2010s source reports subscriber counts amid cord-cutting, not a share.

⚠ 1990s–2010s sources report within-decade ranges; the placards carry them and the chart renders the gaps.

⚠ Price and anchor years differ by more than 3 years in: 1940s — the hours axis is not strictly inflation-free there.

⚠ Pre-1970 car prices are wholesale averages; 1970+ are BEA transaction prices — the jump at 1970 partly reflects the methodology change, not only real price change.

⚠ Share-of-income mixes income concepts across decades (family income (period cost-of-living survey), total money income); a step between points may reflect the concept change, not affordability.

⚠ Price and anchor years differ by more than 3 years in: 1900s, 1940s, 2020s — the hours axis is not strictly inflation-free there.

⚠ Survey populations differ across the century — 1901 wage-earner families vs modern consumer units; each bar links to its placard, which names who was measured.

⚠ From 2010s onward the source is ATUS (all adults 15+, hrs/day, narrower 'household activities'), not Ramey (prime-age women 18-64, hrs/week, broader 'home production') — a concept splice. ATUS facts remain available in their rooms, but render as linked gaps on this weekly axis rather than as a false decline.

⚠ Populations differ across the century — 1901 wage-earner families vs modern consumer units; every placard names who was measured.

⚠ The 1950s point is the nearest available survey (CEX 1960–61), stated plainly rather than back-cast.

⚠ Population changes mid-series: 1970s–1990s are 4-person CUs (CEX Table 4), 2000s–2020s are all consumer units (multi-year tables). Every placard names the measured population.

⚠ The aggregate ratio hides inequality — the bottom quintile spent 221.8% of income in 2012 (see us-2010s-cex-quintile-q1-ratio). The falling aggregate reflects high-income households pulling the ratio down, not broad-based affordability improvement.

⚠ Two complementary halves of one story: national healthcare spending rose from 5% to 18% of GDP while the out-of-pocket share fell from 47% to 10.5%. Insurance and public programs shifted costs from direct payment to third-party payment.

⚠ Different population from the NHE arcs (consumer units, not national health spending) and different concept (family budget share, not GDP share). The trend is the same direction.

⚠ Consumption (tax-derived), not prevalence. The chart's line ends 1994 where its table does: 1995-1999 have no table in the archived record, and 2000-2023 continue on the TTB-removals basis (series us-cigarettes-per-capita-ttb, a methodology splice) — those decades stay in the room placards rather than piling on this axis. 1993 provisional, 1994 projected.

⚠ NHIS survey years only; the survey begins 1965, so earlier decades render as gaps, not zeros. Each marker sits at its fact's own year. The decade placards come from the American Lung Association compilation of the same survey, which agrees with this table at every overlapping year except 2006 (20.8 here, 20.6 there); the 2000 marker is that compilation's value on a year this table leaves blank.

⚠ Apparent consumption: legal beverage sales and shipments divided by population. The source table prints '(Prohibition)' instead of values for 1920-1933. That gap is not zero consumption or proof that no alcohol was sold; it is not interpolated and no estimate is substituted for it.

⚠ Not falling, on purpose. The line is a discontinuity, not a trend: the 1911-1915 level is not reached again until the 1970s, and the modern peak (2.76 in 1981) sits above every year since.

⚠ The 1900s and 1910s cards carry five-year ranges rather than single years, because that is how the table publishes the pre-1934 record; the annual series therefore begins at 1934.

⚠ The denominator changes at 1970, from population aged 15 and over to aged 14 and over. It is disclosed, not corrected for.

⚠ The table's own note records that data before 1977 are from Hyman et al. 1980 — the early half of this official series is republished scholarly work.

⚠ A rate, not a count. Absolute road deaths peaked at 55,600 in 1972 and were still near 41,000 in 2023; what fell by a factor of thirty is the aggregate death rate per mile travelled. It does not measure an individual traveller's risk.

⚠ FHWA's own footnote records a definitional change: from 1976 the count includes only people who died within 30 days of the crash. Earlier years were counted on a wider window, so the two halves of this line are not quite the same measurement.

⚠ The 1900 figure rests on 36 recorded deaths against an estimated 100 million vehicle-miles. It is published, not reconstructed, but the exposure behind it is very thin and the travel figure is a back-cast.

⚠ The federal Census of Fatal Occupational Injuries began in 1992. Earlier estimates and other records exist, but this collection does not join them to the CFOI series.

⚠ Between 2006 and 2024 the displayed rate ranges from 3.3 to 4.2. It is not marked as a steadily falling arc; no statistical trend test is implied by that presentation choice.

⚠ The 1992-2005 rates CFOI did publish were computed against employment rather than hours worked. BLS replaced that basis in 2006, so those years measure something else and are not chained on; the 1990s slot is a gap, not a zero.

⚠ Not marked falling, because it does not only fall. The highest point in this selected-year table is 1960; that is not a claim about the annual historical peak. The selected points rise again between 1980 and 1990 before declining.

⚠ A rate over every woman aged 15 to 19, not over those who were sexually active. This age-specific rate does not establish marital status, sexual activity, or social approval.

⚠ The series is what NCHS prints: selected years, not annual, so the line connects published rows and the true peak between them is not shown. It begins in 1950 because the published table does; the museum's earlier rooms carry no card rather than an invented one.

⚠ Availability, not intake. These are disappearance estimates — production plus imports, less exports and non-food use, over population — so they say what the food supply made available, not what anyone ate. ERS publishes a separate loss-adjusted series for consumption-like figures and this is not it.

⚠ Farm weight, as ERS publishes it. Each sheet also prints a retail-weight column derived by a conversion factor; the factor is disclosed rather than applied.

⚠ The lines start where their sheets start, not where the commodity did: broccoli and bell peppers at 1960, avocados and grapes at 1970. Avocados end at 2017 and the others run later, because that is how far each published series reaches.

⚠ The shape is not a smooth rise. Avocados swing year to year through the 1980s — 2.37 pounds in 1986, 1.08 in 1989 — and broccoli is lower in 2022 than it was in 2000. An arc of steady progress would be the wrong picture even here.