“Luck” here means the economy and policy treating you well at the moments in life when it counted: the job market when you started work, mortgage rates when you bought, the stock market while your savings were large, inflation once you were living on a fixed income. Each model scores how far those conditions sat above or below the 1950–2025 norm, life stage by life stage, then adds them up over a lifetime.
Scores are relative to the average cohort, in standard deviations of the underlying conditions. The full spread between luckiest and unluckiest is about 0.25–0.4. Years after 2026 are simulated 600 times from blocks of real 1950–2025 history, so cohorts with most of life ahead carry wide ranges and a pull toward zero.
Each row is a five-year birth cohort. The bar shows the 10th–90th percentile range across the simulated futures; a hollow dot means more than half of that cohort’s weighted life is still ahead.
This grid compares people at the same age across decades: a 30-year-old in 2026 lined up against a 30-year-old in 1972. Rows are birth cohorts, columns are ages, and each cell shows how favorable conditions were for that cohort at that age (blue better than the 1950–2025 norm, red worse). Read across a row to follow a cohort through life; read down a column to compare everyone at the same age. Diagonals mark calendar years. Model 3 counts bad years 1.3×, so its grid runs redder; childhood cells for 1955–79 are dominated by gasoline lead.
Average conditions from birth through the age you pick, for every cohort that has fully reached it, relative to the average of the cohorts shown. No projections in this chart. Hover a bar for the value against the 1950–2025 norm.
Each cohort ranked against every cohort that reached the same age. “Around that age” compares conditions during ages X−2 to X+2, so being 30 in 1972 lines up against being 30 in 2019. “Birth through that age” ranks the luck accumulated up to that birthday. A cohort appears once its youngest members have turned that age; when a window runs past 2025, only the years already lived count. Values are in SD relative to the average of the cohorts in that column. Hover a cell marked “why” for what made that cohort luckiest or unluckiest at that age. The lifetime column is the full model score, including projected years (“proj.” where more than half is still ahead).
Each column is a separate ranking of the cohorts that reached that age; height is a cohort’s place in that ranking, first at the top and last at the bottom, whatever the number of cohorts. It is not a lifetime average (that is the ranking chart at the top of the page). Click a line or a row in the table above to follow one cohort.
Contributions to each cohort’s score, grouped, relative to the average cohort. Bars right of the line helped; bars left of it hurt.
Model 3’s answer depends on how much each area of life counts. Move the sliders to reweight it; the ranking updates immediately (central estimate only).
Neither is part of the scores, but both test whether they make sense.
Save 10% of a median income from age 25 to 64 in a 60/40 stock/bond mix; wealth at 65 in years of final income (real). Hollow dots and bars are projections.
Early boomers win this one by a wide margin. They saved through the 1982–1999 boom and had low income growth, so early contributions dominated. Model 3 includes this; Models 1 and 2 do not compound returns.
Share of children whose household income at 30 beat their parents’ at the same age (Chetty et al. 2017, “The Fading American Dream”). Available for cohorts born 1940–1984.
This falls almost every five years, because it compares levels across generations and later parents were already richer. The luck scores compare each cohort to a fixed historical norm, so they line up only loosely (correlation 0.04–0.35 with luck banked by 30).
All three models run the same calculation. They differ in which conditions go in, and in when during a life each condition is allowed to count.
Years not yet lived are simulated 600 times. Each simulated future is stitched together from random 8-year blocks of real 1950–2025 history, so one future might contain a 1970s-style decade and another a 1990s-style one. 2026 is seeded with January–October 2026 actuals. The ranges on the ranking chart are the 10th to 90th percentile across these futures.
“Economic weather” means the general condition of the economy in a given year: is it easy to find a job, is the economy growing, are prices stable. It is what a news anchor reports, and it says nothing about your age. Grading a life this way is like averaging the weather over every day you lived, without asking whether it rained on your wedding or on a random Tuesday.
| Metric | Exact definition | Source | 1950–2025 average | Std. dev. | Better when |
|---|---|---|---|---|---|
| Jobs | Annual average unemployment rate | BLS via FRED UNRATE (1948+); BLS historical series for 1940–47 | 5.7% | 1.6 pts | lower |
| Growth | Real GDP growth minus population growth (growth in real GDP per person) | BEA via FRED GDPCA; population B230RC0A052NBEA | 2.1% | 2.2 pts | higher |
| Price stability | Distance of CPI inflation from 2%: |inflation − 2| (2%, not 0%, because deflation is also bad) | BLS CPI-U via FRED CPIAUCNS | 2.0 pts | 2.3 pts | smaller |
Each year’s weather is the average of the three z-scores. A cohort’s Model 1 score is the average weather over the years it was 18 to 70, so each of those 53 years counts 1/53. Five sample years:
| Year | Unemployment | GDP per person | Inflation | z jobs | z growth | z prices | Weather |
|---|---|---|---|---|---|---|---|
| 1969 | 3.5% | +2.1% | 5.5% | +1.35 | +0.02 | −0.61 | +0.25 |
| 1982 | 9.7% | −2.7% | 6.1% | −2.47 | −2.17 | −0.89 | −1.85 |
| 2009 | 9.3% | −3.4% | −0.4% | −2.21 | −2.49 | −0.13 | −1.61 |
| 2019 | 3.7% | +2.1% | 1.8% | +1.24 | 0.00 | +0.80 | +0.68 |
| 2022 | 3.6% | +1.9% | 8.0% | +1.25 | −0.09 | −1.70 | −0.18 |
What Model 1 can’t see: whether a bad year hit you at 22 or at 62, anything about asset prices or housing, and policy.
The same calculation, but each metric counts only at the ages where it affects a household, using one fixed life script for everyone: childhood to 17, start work at 22, buy a home around 30, retire at 65. Eleven metrics:
| Metric | Calculation | Ages that count (relative weight) | Points |
|---|---|---|---|
| Unemployment | Annual unemployment rate | 0–17: 0.25 (parents’ jobs) · 18–21: 0.6 · 22–25: 1.0 · 26–54: 0.8 · 55–64: 0.6 | 15 |
| Entry-recession scarring | Average unemployment z in the 3 years after starting work, applied again at 25–34. Graduating into a recession lowers pay for about a decade (Kahn 2010; Oreopoulos, von Wachter & Heisz 2012). | 25–34, fading from 1.0 to 0.3 | 5 |
| Real income growth | Annual % change in real median family income (Census, via FRED) | 0–17: 0.3 · 18–22: 0.6 · 23–54: 1.0 · 55–64: 0.7 · 65+: 0.1 | 20 |
| Stock returns | S&P 500 total return, deflated by Dec-to-Dec CPI (Damodaran, NYU) | Rises from 0.05 at 25 to 1.0 at 64 as savings grow · 1.0 at 65–74 · tapers to 0.6 by 90 | 15 |
| Bond returns | 10-year Treasury total return, after inflation (Damodaran) | Rises from 0.1 at 40 to 0.7 at 64 · 0.8 after 65 | 5 |
| Real savings rate | 3-month T-bill yield minus inflation (what cash savings earn) | 50–64: 0.3 · 65+: 1.0 | 3 |
| Inflation | |CPI inflation − 2%| | 0–17: 0.2 · 18–64: 0.5 · 65+: 1.0 (fixed incomes) | 10 |
| Home affordability at purchase | Mortgage payment as a share of income: (median home price ÷ median family income) × 80% loan × 30-year payment factor at that year’s mortgage rate. Prices: average of the new-home median (FRED MSPUS) and Shiller’s index. 2012: 16%; 1981: 42%. | Ages 27–35, peaking at 29–31 | 8 |
| Home price gains | Annual change in Shiller’s real home price index | 30–45: 1.0 · 46–64: 0.7 · 65+: 0.4 | 8 |
| College cost | Tuition + fees + room + board ÷ median family income (NCES Digest 330.10) | 18–21 | 4 |
| War and the draft | log(1 + US military deaths in thousands + 0.015 × draft inductions in thousands) | 18–26 | 7 |
Model 3 makes five changes to the calculation, then adds eight metrics.
1. Life stages move with the cohort. Later generations start work, have children, buy homes and retire later, so a tailwind at 30 means something different for each one.
| Born | Starts work | First child (avg. parent) | Buys a home | Retires | Parents’ estate arrives | Chance of reaching 85 |
|---|---|---|---|---|---|---|
| 1940 | 21 | 23 | 27 | 63 | 50 | 33% |
| 1960 | 22 | 25 | 30 | 65 | 54 | 38% |
| 1980 | 23 | 26 | 33 | 66 | 58 | 42% |
| 2000 | 23 | 29 | 35 | 67 | 62 | 46% |
Sources: mother’s mean age at first birth (NCHS) plus 1.25 years for fathers; the age a cohort reaches about 50% homeownership (Census); average age of leaving the labor force (Boston College Center for Retirement Research), projected for later cohorts. Survival follows a Gompertz curve whose modal age at death rises from 84 (1940 births) to 89.
2. Years with kids at home count 1.6×. From the first child to 18 years later (tapered over 3 years at each end), weights on jobs, income, inflation, home buying, mortgage cost, college and student debt are multiplied by 1.6. Money matters most when a household is raising children. This is the mechanism you described: a cohort that has kids at 30 gets more out of a good economy in its thirties than in its twenties.
3. Each age is weighted by the chance of being alive. Conditions at 85 count for less in a cohort where only a third reach 85.
4. Bad years count 1.3×. Any negative z is multiplied by 1.3, because a recession at 25 hurts more than a boom helps (diminishing marginal utility of money).
5. Exposure scaling. Stock-market luck is scaled by the share of households owning stock that year (about 20% before 1983, 60% now, ÷ 45%), so the 1970s crash counts for less because few families owned stock. Home-price gains are scaled by the cohort’s homeownership rate at 40 (72% for 1940 births, 60% for 2000).
The added metrics:
| Metric | Calculation | Ages that count | Points |
|---|---|---|---|
| Nest egg at retirement | Save 10% of income from 25 until retirement in a 60/40 stock/bond mix (real returns); income grows with real median family income. Wealth at retirement in years of final income, as a log ratio to what constant average returns would produce (9–11 years), divided by 0.51 (its spread across simulated histories), then scaled by stock ownership. This replaces most of Model 2’s year-by-year stock and bond scoring, which misses compounding; stock and bond returns keep 4 and 2 points, counted only after retirement. | Retirement to retirement + 20 | 9 |
| Real mortgage cost | For a home bought in year t: the mortgage rate, refinanced whenever rates drop 0.75 point below it, minus inflation, averaged over 10 years. Average for 1950–2015 purchases: 3.4% a year. Splits the home-buying budget with affordability (5 points). | Home-buying window | 3 |
| Student-debt overhang | College cost burden × share of students borrowing × share attending college, averaged over the cohort’s college years; ranked across cohorts | 15 years after leaving school | 3 |
| Health-care cost burden | National health spending as % of GDP (CMS NHE Table 1): 4.4% in 1950, 18% now | Working years 0.6 · retired 1.0 | 3 |
| Avoidable-death environment | Traffic + homicide + drug overdose + HIV deaths per 100,000 (approximate, from NCHS, NHTSA, CDC) | 15–54 | 3 |
| Childhood lead exposure | Thousand tonnes of lead burned in US gasoline (peak about 220 in 1972; zero after 1995) | 0–5 | 2 |
| COVID | Fixed shocks: −2.5 SD at ages 6–17 in 2020 and −2.0 in 2021 (school closures); −2.5 SD at 70+ in 2020 and 2021 and −1.0 in 2022 (mortality), half that at 65–69 | 6–17 and 65+ | 2 |
| Retirement security | Average of two cohort rankings. (a) Social Security’s real return for a medium earner, about 3.2% for 1940 births falling to 1.6% for 2005 births, minus 0.3 point × the share of retirement after 2034 (a 50% chance of the roughly 20% trust-fund cut). (b) Share of retirees with a defined-benefit pension, 45% falling to 15%. | Retirement to retirement + 25 | 6 |
| Inheritance per sibling | Household net worth ÷ disposable income in the year the parents’ estate arrives (Fed Z.1), divided by the completed fertility of the parents’ generation | 5 years either side of the arrival age | 4 |
Model 3 keeps Model 2’s metrics, reweighted to make room: unemployment 11, entry scarring 4, income growth 15, real savings rate 3, inflation 7, home affordability 5, home price gains 7, college cost 2, war 5. For 2027–2036 it lowers expected real stock returns from the 1950–2025 average of 7.8% a year to 2.5%, the earnings yield implied by a CAPE of 40.7 (October 2026).
Cohort-level metrics (nest egg, student debt, retirement security, inheritance) are single numbers per birth year, z-scored across birth years instead of calendar years, then spread over their ages like everything else.
Luck here is relative to the norm, not absolute living standards. A 30-year-old today has better medicine and technology than one in 1972. Counting that would hand the prize to whoever was born last, and it isn’t what “tailwinds” means.
These are national medians. The models don’t separate race, gender, region, or education, and those gaps were often larger than the gaps between cohorts. Inequality enters only through median income growth.
Weights are judgment calls. Model 3 picks 1940–44 in 87% of simulated futures, and in 75% when every factor’s budget is also randomly rescaled (log-normal, σ = 0.4, typically between ⅔× and 1.5×). It still picks 1940–44 with retirement security, the nest egg, health or inheritance dropped one at a time. Models 1 and 2 are closer calls: their leaders are #1 in 19% and 57% of futures.
Persistent conditions are assumed to continue: today’s tuition burden, health-spending share, low pension coverage and a 50% chance of the scheduled 2034 Social Security cut. That is most of why cohorts born after 2000 score where they do.
Data: S&P 500 and 10-year Treasury returns from Damodaran (NYU); CPI, unemployment, GDP, population, family income, T-bills, 10-year yields, mortgage rates, new-home prices and household net worth from FRED; real home prices and CAPE from Shiller; college cost from NCES Digest table 330.10; health spending from CMS NHE Table 1; absolute mobility from Opportunity Insights. Hand-entered approximations: pre-1971 mortgage rates, 1940–53 income growth, war deaths and draft inductions, the avoidable-death series, gasoline lead, stock-ownership rates, Social Security returns by cohort, pension coverage, parents’ fertility, and the life-stage ages.