The Mechanisms of Transmission

The pattern is established. This lesson is about the machinery — the concrete, datable, largely legal steps by which a family's position when a child is born becomes that child's position three decades later.

Nearly all of it happens without anyone doing anything wrong, which is what makes it durable and what makes it hard to argue about.

Two children, one city, thirty-five years.

Return to 8.1.1's two seven-year-olds. Call them Aisha and Charlotte. Follow both.

Before school. Charlotte's household has savings, two adults, and a parent who reduced hours. She is read to daily, taken places, and spoken to in a stream of explanation and question. Aisha's household has one adult working two jobs, no savings, and a grandmother who does the childcare and adores her. By school entry there is a measurable gap in vocabulary and in familiarity with the school's way of talking. Neither child has done anything.

Primary. Charlotte's parents bought a house in a particular catchment. That was not a school choice; it was a housing transaction — and the price included the school. Aisha's school is a mile away and has a third more children with additional needs and a quarter of the staff turnover every year.

Eleven. Both are assessed. Charlotte, borderline on one paper, is tutored for four months. Aisha, borderline on the same paper, is not. They end up in different sets, which determines which examinations they will be entered for, which determines what they can do at sixteen.

Sixteen. They achieve similar grades. Here the divergence becomes a decision rather than an outcome. Charlotte's family assume she will stay on; the question is which subjects. Aisha's family are proud and worried; there is a job available; staying on costs money now for a benefit later, and nobody in the family has done it. Aisha stays on. Many in her position do not — and the ones who do not have the same grades as those who do.

Eighteen. Charlotte applies to universities her parents' colleagues attended, with a personal statement redrafted four times by an adult who has read hundreds. Aisha applies locally, so she can live at home. The two decisions are both rational. They are made with different information and different fallbacks.

Twenty-one. Charlotte takes an unpaid summer internship in a city where she stays with a family friend. This is a job that is only available to someone who can work for nothing and has somewhere to sleep. Aisha works in a call centre, which pays.

Twenty-four. Charlotte's application to a professional firm is one of thousands. The people shortlisting look for evidence of "polish" and "fit" — extracurricular activities, travel, the ability to talk easily about interests that overlap with theirs. The criteria are not about family income and they sort by it almost perfectly. Aisha, equally qualified, is not shortlisted.

Twenty-nine. Charlotte buys a flat with a deposit from her parents. Aisha rents, and pays more per month than Charlotte's mortgage. From this point the gap compounds automatically: one is acquiring an appreciating asset and the other is not.

Thirty-five. Charlotte earns more, owns more, and has a fallback. And when her employer restructures, she can wait three months for the right job. Aisha cannot, and takes the first one.

Now count. Nobody broke a rule. Nobody discriminated in any actionable sense. There were perhaps eleven points at which the gap widened, and at each one the mechanism was different.

The correlation is not in dispute. The question is which channels carry it, and how big each is.

Why this matters more than the correlation. A society that wants to change the pattern has to act on a mechanism, and the mechanisms have wildly different tractability. Some are cheap to change and some are close to impossible; some operate before school and some after hiring; and acting on the wrong one produces expensive policies that do nothing — which describes a great deal of mobility policy.

And there is a structural feature that makes it harder. The channels are not independent, and they substitute for each other. Block one and advantaged families use another — a finding with a name, compensatory advantage: when an advantaged child suffers an early setback, the family mobilises to correct it, so early disadvantages predict later outcomes much less strongly for them than for others.

Which means a policy that closes one channel and leaves the others open will find the advantage rerouted rather than removed. That is not an argument against acting. It is the reason single-channel policies disappoint.

The seven channels

One — direct financial transfer, which is the largest and requires no effort from anybody.

Deposits for property, help with rent, school fees, the cost of unpaid work experience, a guarantee on a lease, cars, the fees for a professional qualification, and the standing fact of a household that can absorb an emergency.

Two features make this channel distinctive.

It is invisible in earnings data (see 8.3.2), which is why studies of income mobility understate transmission.

And its timing matters. Inheritance proper arrives late — typically when the recipient is in their fifties or sixties — so it affects wealth rather than trajectory. The transfers that shape a career are the ones that arrive at twenty-two and twenty-nine, and they are gifts, not bequests.

Two — investment in the child, from before birth.

Nutrition, health, stability, housing quality, books, talk, activities, tutoring. The attainment gap by family background is present and measurable before school begins, and it grows.

And here a famous piece of evidence needs correcting, because this course does that. A widely reproduced chart appeared to show that low-ability children from rich families overtook high-ability children from poor families by around age ten — two lines dramatically crossing. It became one of the most cited images in education policy.

Subsequent methodological work showed the crossover was substantially an artefact of regression to the mean (see 7.6.4): children selected as extreme on an early noisy test will move towards the average on a later one regardless of anything else, and the apparent crossing largely reflects that.

What survives is still serious: the gap at school entry is real and substantial, and family background predicts later attainment strongly. What does not survive is the dramatic crossover, and it is worth knowing because the chart is still in circulation.

Three — the education system, which transmits advantage even when it is free.

This is the channel people expect to be equalising, and the mechanism by which it is not is precise.

The key distinction separates primary effects — differences in measured attainment — from secondary effects: differences in the choices made at branch points, conditional on identical attainment. Whether to stay on, which track to take, which subjects, whether to apply, where to apply.

Secondary effects are large, and they are where advantage is defended, because they operate on children who have already demonstrated the same ability. Aisha and Charlotte at sixteen is a secondary effect.

And two further mechanisms explain why educational expansion has done so little to relative mobility (see 8.3.1).

Inequality is maintained as long as there is room to maintain it. When a level of education expands, the advantaged take up the new places first; the association between origin and reaching that level only weakens once the advantaged are effectively saturated, and then the competition moves up to the next level.

And when quantity is equalised, advantage shifts to quality and type. Once everyone attends secondary school, it matters which school; once everyone has a degree, it matters which institution and which subject. Advantage relocates from whether to which — which is why formal equalisation of access has repeatedly failed to equalise outcomes, and why measuring participation rates as an equity indicator misses the point (see 7.2.2 on Goodhart).

Four — cultural capital and comportment.

Bourdieu's argument (see 5.8.1), and its most concrete demonstration is in the observed differences in how families deal with institutions. One pattern involves organised activities, extensive reasoning and negotiation with children, and an active, entitled stance towards teachers, doctors and officials — children are taught to question, request and advocate for themselves. Another involves more autonomous unstructured time, clearer directives, and a deferential stance towards professionals who are treated as authorities not to be challenged.

The second is not deficient parenting, and the research making the observation was explicit that its warmth, independence and kin relationships have real advantages. But institutions reward the first, because institutions are staffed by people from families that produced it. The advantage is not in the child; it is in the fit between the child's disposition and the institution's expectations — which is precisely 5.8.1's point, observed in living rooms.

Five — networks and information.

A large share of jobs are found through personal contacts, and — Granovetter's finding — disproportionately through weak ties rather than close friends, because distant acquaintances know about opportunities your immediate circle does not.

Which converts a family's social range directly into a child's opportunity set. Who mentions that a job exists, who explains what the role actually involves, who makes an introduction, who says "call my colleague". None of this is corrupt and all of it is unequally distributed.

Six — gatekeeper practice.

Audit evidence establishes that identical applications are treated differently by signalled characteristics (see 7.4.2). And the qualitative research on hiring at elite professional firms adds the mechanism: recruiters describe looking for "polish", "fit" and evidence of a well-rounded life — extracurricular achievements, particular leisure pursuits, travel, the ability to hold a certain kind of conversation. They report, sincerely, that they are selecting on excellence and cultural compatibility. The criteria are not about money and they sort by it with great efficiency.

Seven — place and housing.

Neighbourhood exposure has causal effects on children's outcomes, established by two designs with different weaknesses (see 8.3.2). And where school admission depends on address, the school is capitalised into the house price — a relationship demonstrated by comparing otherwise identical properties on either side of a catchment boundary, where prices jump.

Which means that in systems with catchment-based admission, a school place is purchased through the housing market. It is legal, invisible in any education statistic, unavailable to renters, and it is one of the largest single transfers in the whole list.

The genetic question, handled properly.

Some of the parent–child similarity in education and earnings is not environmental transmission, and a course that omitted this would be dishonest. Three things must be said together.

One — the evidence that there is a genetic component is real. Adoption studies, including one using a programme that assigned children to families in a manner close to random, find that adoptees' educational and economic outcomes correlate with their adoptive parents' position — substantially, and less strongly than for biological children. Both halves of that sentence matter: the environmental effect is clearly non-zero, and it is not the whole story.

Two — heritability is routinely misunderstood, in a way 7.6.4 already prepared you for. A heritability estimate decomposes variance between individuals within a particular population, in a particular environment. It says nothing about differences between populations, nothing about whether a trait is malleable, and nothing about any individual. A trait can be highly heritable and completely changed by an environmental intervention — height in populations that improved their nutrition is the standard example.

Three — and the finding that most complicates the simple picture. Studies using genetic scores have found that parental genetic variants that were not passed on to the child still predict the child's educational outcome. Since those variants are not in the child, the effect cannot be direct: it must operate through the environment the parents provide. This is called genetic nurture, and it means that a substantial part of what looks like genetic transmission is genes acting on parents, who then act on children's environments.

Where this leaves the argument. Both channels operate. The environmental channels are large, demonstrably manipulable, and the ones on which policy can act — the movers evidence, the adoption evidence and the secondary-effects evidence all establish that. And heritability is not a statement about groups, not a statement about destiny, and not an argument against intervention.

And the explanation that keeps being reached for and does not survive: aspirations.

The intuitive account — that disadvantaged families want less for their children — is not supported by the evidence. Survey and interview research across several countries consistently finds that parents and children across the income distribution report high aspirations, and that stated aspirations do not differ nearly enough to explain the gaps in outcomes.

What differs is not the wanting. It is the knowing how, and the having what it takes. Which examinations open which doors, when the deadlines are, what a personal statement is for, who to ask, what a given course leads to, whether the risk of trying and failing can be absorbed. Aisha at eighteen did not aspire less than Charlotte; she navigated with less information and a smaller floor.

This matters because interventions aimed at "raising aspirations" have a poor evaluation record, and it is exactly what you would predict if the constraint was never aspiration. It is also an example of what 8.2.3 warned about: an individualised explanation for a structured outcome, which locates the problem in the family's psychology and leaves every one of the seven channels untouched.

Because "which channel" is the whole of the policy question, and the honest answer about relative size is unsettled.

Decompositions disagree, they differ by country and by outcome, and the wealth channel is the most under-measured because earnings data cannot see it. Anyone who tells you confidently that one channel accounts for most of it is reporting a modelling choice.

Four things that are solid.

Secondary effects are large — the divergence at branch points among children with identical attainment.

Advantage relocates when a channel closes. Equalise access and it moves to type; equalise type and it moves to networks and gatekeeping. This is compensatory advantage operating at the level of a whole society, and it explains 8.3.1's constant flux.

Place has causal effects, established by converging designs.

And the largest transfers are legal, invisible in the statistics that get quoted, and require no effort: the deposit, the catchment, the unpaid internship, the guarantee.

Two questions to carry.

When you meet a mobility policy, ask which of the seven channels it acts on — and which of the other six the advantage will move to.

And when you meet an explanation of the gap, ask whether it locates the mechanism in the family's psychology or in the structure of opportunity. The first is easier to say, cheaper to act on, and has the worse evidence base.

Eleven divergence points between two children, none of them involving anyone breaking a rule.

Seven channels. Direct financial transfer — the largest, invisible in earnings data, and decisive in the form of gifts at twenty-two and twenty-nine rather than inheritance at sixty. Investment in the child, producing a real gap at school entry — though the famous chart showing rich low-ability children overtaking poor high-ability ones was substantially a regression-to-the-mean artefact. The education system, through secondary effects — divergence at branch points conditional on identical attainment — and through the pattern by which expansion only weakens the origin–destination link once the advantaged are saturated, after which advantage relocates from whether to which. Cultural capital, where the advantage lies not in the child but in the fit between disposition and institutional expectation. Networks, converting a family's social range into an opportunity set, disproportionately through weak ties. Gatekeeper practice, where "polish" and "fit" are sincere criteria that sort by background with great efficiency. And place and housing, where catchment-based admission means a school place is bought through the property market — legal, invisible in education statistics, and unavailable to renters.

The channels substitute for each other. Compensatory advantage: block one and advantaged families use another, which is why single-channel policies disappoint.

On genetics, three things together. Adoption evidence — including quasi-random placement — shows environmental effects that are clearly non-zero and not the whole story. Heritability is a within-population variance decomposition that says nothing about groups, malleability or individuals. And genetic nurture — parental variants not transmitted to a child still predicting that child's education — shows that much apparent genetic transmission runs through the environment parents create.

And the aspiration explanation does not survive: stated aspirations do not differ nearly enough to explain the gaps. What differs is navigation and the floor to fall on — and aspiration-raising interventions have a poor record, exactly as that predicts.

Inter vivos transfer — a gift during the giver's life; the transfer that shapes trajectories, as against inheritance, which arrives late.

Primary and secondary effects — differences in attainment; differences in choices at branch points given identical attainment.

Maximally maintained inequality — expansion reduces origin effects at a level only once the advantaged are saturated.

Effectively maintained inequality — when quantity is equalised, advantage shifts to quality and type.

Concerted cultivation / accomplishment of natural growth — the two observed patterns of child-rearing, and the institutional fit that rewards one.

Weak ties — distant acquaintances who supply information a close circle does not.

Cultural matching / "fit" — gatekeepers selecting on cultural similarity while sincerely selecting on merit.

Capitalisation — a school's quality reflected in local house prices, converting admission into a purchase.

Compensatory advantage — advantaged families mobilising to correct a child's early setbacks, weakening the link from early disadvantage to later outcome for them alone.

Heritability — a within-population variance decomposition; not a statement about groups, individuals, or malleability.

Genetic nurture — non-transmitted parental variants predicting a child's outcome, demonstrating an environmental pathway.

One — map your own eleven points. Write your own version of the timeline, from before school to the present, marking every point at which something a family had or did affected what happened next.

Two — find a secondary effect. Identify a branch point in your country's education system, and ask what proportion of similarly-attaining children from different backgrounds take each path. The data usually exists.

Three — price a catchment. Look up house prices on either side of a school catchment boundary near you. The difference is the price of the school, and it is paid to a previous owner rather than to the school.

Four — audit an unpaid opportunity. Find an internship, placement or volunteering role in a field you care about, and work out what someone would need in order to take it. That list is the eligibility criterion nobody wrote down.

Five — test the aspiration story. Ask three people from different backgrounds what they hoped for at sixteen, and then what they knew about how to get it. The difference will be in the second answer.