Seth Walsh
Iconoclast
Contributor
- Joined
- Jan 12, 2020
- Posts
- 11,038
- Reputation
- 22,585
People talk about class as if it begins when you choose a degree, get your first job, or start earning money at 23.
That is almost backwards.
By the time you are born, researchers can already observe variables that contain a disturbing amount of information about your eventual life:
parents’ occupation + education + income + wealth trajectory + neighbourhood + family stability + expected lifetime resources.
You haven't made a single decision yet.
And your probability distribution is already different.
Eshaghnia, Heckman, Landersø and Qureshi construct measures of parents' expected lifetime resources rather than simply asking what Dad earned this year. Expected lifetime resources predict children's education and their own eventual economic resources substantially better than conventional snapshot income measures. Their newer causal estimates still find sizeable effects, with the effects of parental resources generally largest during early childhood.
Translation:
It buys the house accordingly.
Chooses the school accordingly.
Takes career risks accordingly.
Builds networks accordingly.
Invests in the child accordingly.
The researchers find expected lifetime wealth produces intergenerational elasticity estimates around 0.38–0.40 at the most predictive ages, considerably higher than estimates obtained using some ordinary annual-income measures.
This is the real class variable people constantly miss:
not income. Expected lifetime command over resources.
Their basic result:
Every 10-percentile increase in parental income rank was associated with about a 3.4-percentile increase in the child's eventual income rank.
That's an average, not destiny.
But think about what it means.
Before Junior has:
Society loves individual biographies.
The dataset loves conditional probabilities.
Britain's 1958 National Child Development Study followed roughly 17,000 people from birth.
Researchers explicitly recorded parental social class at birth.
Decades later, birth class remained connected through pathways involving childhood cognitive performance, educational attainment and eventual adult social class. In one analysis, social class at birth had a standardized association of 0.27 with cognitive performance at age seven, which then strongly predicted cognition later in childhood and subsequent qualifications and occupational class.
Read that again.
They didn't ask:
And the signal propagated through the life course.
A 2021 study followed 8,711 members of the 1958 cohort into middle age.
Compared with people born into the highest occupational class, people born into the lowest class had an estimated 42% higher frailty-index level at age 50. Adult socioeconomic position explained only part of the relationship.
Another British cohort of more than 11,000 people used the father's social class recorded at the actual time of birth.
Those born into manual rather than nonmanual social classes subsequently had a 52% higher hazard of coronary heart disease or stroke before adjustment; the association weakened substantially after accounting for childhood growth and education, which is exactly what you'd expect if later pathways mediate part of the original gradient.
The class system isn't merely:
rich kid gets nicer car.
It can propagate through:
nutrition → stress → cognition → education → behaviour → occupation → adult environment → health.
Lower parental income was associated with elevated risks across numerous chronic and pain-related disorders and hospitalization. Those from the lowest parental-income decile also subsequently had lower earnings and education on average.
And this is Norway.
Huge welfare state.
Universal healthcare.
Much lower inequality than America.
The gradient survives.
At age 30, people from the lowest tenth of parental-income backgrounds had, relative to the highest tenth:
Again:
Not deterministic.
But the distributions are nowhere near identical.
Rich parents differ from poor parents in thousands of ways.
Therefore merely correlating parental SES with child outcomes cannot tell you how much is actually caused by money or environment.
Correct.
Unfortunately, quasi-experiments still find causal effects.
Dahl and Lochner exploited changes in the US Earned Income Tax Credit. Their estimates implied that an exogenous $1,000 increase in family income raised children's combined reading and mathematics achievement by roughly 0.06 standard deviations, with larger effects among disadvantaged families.
Akee et al. studied families receiving effectively exogenous casino-income payments. Among initially poor households, an extra $4,000 annually was associated with roughly one additional year of schooling by age 21 and a 22% reduction in minor offending at ages 16–17.
So no, observational correlations are not purely a gigantic statistical hallucination.
Changing resources can actually change children.
When children moved to areas where existing children had better outcomes, their own eventual outcomes improved roughly in proportion to how many childhood years they spent there.
They estimated convergence toward the destination area's outcomes at approximately 4% per year of childhood exposure. They used sibling comparisons and other designs to distinguish exposure effects from simple family selection.
This is devastating for the pure “individual choices” model.
Two babies can have similar underlying potential.
Put them into different:
Then society meets them at 25 and explains their different outcomes through “mindset.”
Poverty between the prenatal year and age five predicted substantially worse adult attainment, including earnings and work hours, even after accounting for economic circumstances later in childhood.
The newer Heckman/Landersø work points in broadly the same direction: parental lifetime resources often appear especially consequential during early childhood, although the most relevant developmental period differs by outcome.
Which makes intuitive sense.
Human-capital formation is recursive.
Skill today → easier skill acquisition tomorrow.
Early advantage does not merely give you +1.
It can alter the return on every subsequent input.
A 2026 systematic review/meta-analysis covering 38 studies and 388,674 participants similarly found lower parental socioeconomic position associated with worse adult indicators including blood pressure, adiposity, CRP and allostatic load, although heterogeneity across studies was substantial.
Your parents' class is therefore associated not only with your LinkedIn profile.
It can correlate decades later with what is happening inside your arteries.
Brutal.
SES at birth
↓
parental human capital
expected lifetime income
wealth
housing
neighbourhood
family stability
nutrition
stress exposure
parental time
language environment
school quality
peer group
expectations
networks
extracurriculars
healthcare
ability to survive mistakes
↓
childhood skills + health + behaviour
↓
education
↓
first occupational placement
↓
partner market + geography + network
↓
income + ownership + wealth accumulation
↓
SES of the next child
That's why social class has such absurd persistence.
It's not one mechanism.
It is 30 correlated mechanisms firing in the same direction.
Chetty's own work shows enormous geographical variation. A child born into the bottom income quintile had about a 4.4% chance of reaching the top quintile in Charlotte versus 12.9% in San Jose in their data.
That does not disprove the birth lottery.
It proves the opposite:
the transition probabilities themselves depend on the environment.
An individual exception does not erase a distribution.
Someone winning €50 million on the lottery does not prove buying lottery tickets has an expected return of €50 million.
A single annual-income observation at birth can be noisy. Families move up and down. Parents receive shocks. Children experience different environments at different developmental stages. Recent research explicitly finds that expected lifetime family resources can outperform snapshot income, and the relevant timing varies across outcomes.
So the sophisticated version is much darker and much more accurate:
Not equality either.
Researchers can nevertheless look at:
your parents' occupation, education, income trajectory, wealth and location
and already make a meaningfully better prediction of:
your education, earnings, occupation, unemployment risk, health and eventual socioeconomic position.
Then the environment compounds for another eighteen years.
The biggest cope isn't saying social mobility exists.
It obviously does.
The biggest cope is pretending everyone arrives at adulthood at the same starting line and that the race begins when they submit their first job application.
That is almost backwards.
By the time you are born, researchers can already observe variables that contain a disturbing amount of information about your eventual life:
parents’ occupation + education + income + wealth trajectory + neighbourhood + family stability + expected lifetime resources.
You haven't made a single decision yet.
And your probability distribution is already different.
1. The brutal part: it isn't just “poor parents = less money”
The strongest modern work increasingly suggests that a parent's salary in one particular year is actually a mediocre measure of family advantage.Eshaghnia, Heckman, Landersø and Qureshi construct measures of parents' expected lifetime resources rather than simply asking what Dad earned this year. Expected lifetime resources predict children's education and their own eventual economic resources substantially better than conventional snapshot income measures. Their newer causal estimates still find sizeable effects, with the effects of parental resources generally largest during early childhood.
Translation:
The family knows its trajectory too.
It buys the house accordingly.
Chooses the school accordingly.
Takes career risks accordingly.
Builds networks accordingly.
Invests in the child accordingly.
The researchers find expected lifetime wealth produces intergenerational elasticity estimates around 0.38–0.40 at the most predictive ages, considerably higher than estimates obtained using some ordinary annual-income measures.
This is the real class variable people constantly miss:
not income. Expected lifetime command over resources.
2. Forty million Americans: parental rank predicts child rank
Chetty, Hendren, Kline and Saez linked tax records covering more than 40 million children and parents.Their basic result:
Every 10-percentile increase in parental income rank was associated with about a 3.4-percentile increase in the child's eventual income rank.
That's an average, not destiny.
But think about what it means.
Before Junior has:
- chosen his subjects
- developed a work ethic
- discovered LinkedIn
- “networked”
- manifested anything
- started grinding
Society loves individual biographies.
The dataset loves conditional probabilities.
3. Britain literally recorded social class AT BIRTH
This is where it gets particularly brutal.Britain's 1958 National Child Development Study followed roughly 17,000 people from birth.
Researchers explicitly recorded parental social class at birth.
Decades later, birth class remained connected through pathways involving childhood cognitive performance, educational attainment and eventual adult social class. In one analysis, social class at birth had a standardized association of 0.27 with cognitive performance at age seven, which then strongly predicted cognition later in childhood and subsequent qualifications and occupational class.
Read that again.
They didn't ask:
They knew the family's class when you entered the world.
And the signal propagated through the life course.
4. Fifty years later, the body remembers too
Same British birth-cohort tradition.A 2021 study followed 8,711 members of the 1958 cohort into middle age.
Compared with people born into the highest occupational class, people born into the lowest class had an estimated 42% higher frailty-index level at age 50. Adult socioeconomic position explained only part of the relationship.
Another British cohort of more than 11,000 people used the father's social class recorded at the actual time of birth.
Those born into manual rather than nonmanual social classes subsequently had a 52% higher hazard of coronary heart disease or stroke before adjustment; the association weakened substantially after accounting for childhood growth and education, which is exactly what you'd expect if later pathways mediate part of the original gradient.
The class system isn't merely:
rich kid gets nicer car.
It can propagate through:
nutrition → stress → cognition → education → behaviour → occupation → adult environment → health.
5. Norway tried the “but welfare states fix this” difficulty setting
Researchers linked tax and health registers for 429,886 Norwegians, observing parental income from birth through age 18 and health around ages 39–43.Lower parental income was associated with elevated risks across numerous chronic and pain-related disorders and hospitalization. Those from the lowest parental-income decile also subsequently had lower earnings and education on average.
And this is Norway.
Huge welfare state.
Universal healthcare.
Much lower inequality than America.
The gradient survives.
6. Finland: same movie, Nordic subtitles
A recent Finnish register study followed cohorts born from 1981–1989.At age 30, people from the lowest tenth of parental-income backgrounds had, relative to the highest tenth:
- about 3× the risk of low education
- about 2.5× the risk of long-term unemployment
- about 1.8× the mortality risk
Again:
Not deterministic.
But the distributions are nowhere near identical.
7. “It's all genetics/confounding bro” doesn't rescue the story
This is the obvious criticism.Rich parents differ from poor parents in thousands of ways.
Therefore merely correlating parental SES with child outcomes cannot tell you how much is actually caused by money or environment.
Correct.
Unfortunately, quasi-experiments still find causal effects.
Dahl and Lochner exploited changes in the US Earned Income Tax Credit. Their estimates implied that an exogenous $1,000 increase in family income raised children's combined reading and mathematics achievement by roughly 0.06 standard deviations, with larger effects among disadvantaged families.
Akee et al. studied families receiving effectively exogenous casino-income payments. Among initially poor households, an extra $4,000 annually was associated with roughly one additional year of schooling by age 21 and a 22% reduction in minor offending at ages 16–17.
So no, observational correlations are not purely a gigantic statistical hallucination.
Changing resources can actually change children.
8. Your postcode starts working on you before you can pronounce it
Chetty and Hendren studied more than seven million families who moved.When children moved to areas where existing children had better outcomes, their own eventual outcomes improved roughly in proportion to how many childhood years they spent there.
They estimated convergence toward the destination area's outcomes at approximately 4% per year of childhood exposure. They used sibling comparisons and other designs to distinguish exposure effects from simple family selection.
This is devastating for the pure “individual choices” model.
Two babies can have similar underlying potential.
Put them into different:
- neighbourhoods
- schools
- peer groups
- local labour markets
- social networks
Then society meets them at 25 and explains their different outcomes through “mindset.”
9. Early childhood appears unusually important
Duncan, Ziol-Guest and Kalil followed Americans from the prenatal period into their 30s.Poverty between the prenatal year and age five predicted substantially worse adult attainment, including earnings and work hours, even after accounting for economic circumstances later in childhood.
The newer Heckman/Landersø work points in broadly the same direction: parental lifetime resources often appear especially consequential during early childhood, although the most relevant developmental period differs by outcome.
Which makes intuitive sense.
Human-capital formation is recursive.
Skill today → easier skill acquisition tomorrow.
Early advantage does not merely give you +1.
It can alter the return on every subsequent input.
10. Even the biological markers get classpilled
A systematic review/meta-analysis found adults who had grown up in the least advantaged childhood SES groups had approximately 25% higher C-reactive protein levels than those from the most advantaged groups before further adjustment; elevated BMI appeared to mediate some of the association.A 2026 systematic review/meta-analysis covering 38 studies and 388,674 participants similarly found lower parental socioeconomic position associated with worse adult indicators including blood pressure, adiposity, CRP and allostatic load, although heterogeneity across studies was substantial.
Your parents' class is therefore associated not only with your LinkedIn profile.
It can correlate decades later with what is happening inside your arteries.
Brutal.
11. The actual model
The childish model is:The real model looks more like:
SES at birth
↓
parental human capital
expected lifetime income
wealth
housing
neighbourhood
family stability
nutrition
stress exposure
parental time
language environment
school quality
peer group
expectations
networks
extracurriculars
healthcare
ability to survive mistakes
↓
childhood skills + health + behaviour
↓
education
↓
first occupational placement
↓
partner market + geography + network
↓
income + ownership + wealth accumulation
↓
SES of the next child
That's why social class has such absurd persistence.
It's not one mechanism.
It is 30 correlated mechanisms firing in the same direction.
12. The nastiest misconception: “mobility proves class doesn't matter”
Of course people move.Chetty's own work shows enormous geographical variation. A child born into the bottom income quintile had about a 4.4% chance of reaching the top quintile in Charlotte versus 12.9% in San Jose in their data.
That does not disprove the birth lottery.
It proves the opposite:
the transition probabilities themselves depend on the environment.
An individual exception does not erase a distribution.
Someone winning €50 million on the lottery does not prove buying lottery tickets has an expected return of €50 million.
13. One correction to the pure blackpill
There is one thing you should NOT say:The literature does not show that.
A single annual-income observation at birth can be noisy. Families move up and down. Parents receive shocks. Children experience different environments at different developmental stages. Recent research explicitly finds that expected lifetime family resources can outperform snapshot income, and the relevant timing varies across outcomes.
So the sophisticated version is much darker and much more accurate:
Not destiny.
Not equality either.
TL;DR
You spawn with no CV, no degree, no money, no decisions and no “mindset.”Researchers can nevertheless look at:
your parents' occupation, education, income trajectory, wealth and location
and already make a meaningfully better prediction of:
your education, earnings, occupation, unemployment risk, health and eventual socioeconomic position.
Then the environment compounds for another eighteen years.
The biggest cope isn't saying social mobility exists.
It obviously does.
The biggest cope is pretending everyone arrives at adulthood at the same starting line and that the race begins when they submit their first job application.
