When Income Is Not the Whole Story
This chapter reads the main results of Dynastic Capital and the Inheritance of Social Advantage as evidence, not as a slogan. The focus is the study’s parent–child rank relationships: whether measured social position persists across generations, whether it predicts adult income beyond parental income and cognitive ability, and why its connection to the top of the distribution matters. The reader’s task is to understand what the coefficients establish, what they change about an income-only mobility statistic, and what they still cannot prove.
How to read the coefficients
The study compares parents and children using percentile ranks. A rank places a person within a distribution: someone at the 80th percentile has a higher measured position than roughly 80 percent of the comparison group. The relevant comparison groups are cohort-specific, because incomes, test scores, and access to elite positions change over time. Rank–rank analysis puts unlike parental characteristics on a common relative scale, allowing the researchers to ask whether a higher parental position is associated with a higher child position.
A rank–rank coefficient is a slope, not a destiny. If a parent–child coefficient is 0.20, a ten-percentile-point difference between two parents is associated, on average, with a two-percentile-point difference between their children on the measured outcome. It does not predict where either particular child will finish, and it does not say that moving a family up one percentile would cause the child to move by the coefficient. The estimate summarizes a pattern across many observations, holding specified variables constant when the model includes controls.
That distinction is especially important here because the paper compares three dimensions of parental background: social-capital rank, income rank, and cognitive-ability rank. They are not measurements of the same thing. Social-capital rank summarizes the institutionally recorded positions described earlier; income rank describes financial resources; cognitive ability is based on the study’s test measure for the relevant parental cohorts. A similar coefficient across the three dimensions means that each has comparable predictive content in that specification, not that money, ability, and institutional position are interchangeable in lived experience.
The first result concerns the transmission of social position itself. Without controls, the parent–child social-capital rank coefficient is 0.22. After parental income rank, cognitive-ability rank, and demographic controls are included, it remains 0.17. For context, the paper cites typical Norwegian parent–child income rank correlations in the rough range of 0.20 to 0.25. The controlled social-capital relationship is therefore not a tiny residue left after the familiar measures have explained everything. It remains a substantial association in a large administrative sample.
The comparison also works in the other direction. Parental income and cognitive ability are each positively associated with children’s social-capital rank. Once the dimensions are entered together, their coefficients fall, but the social-capital coefficient remains large and statistically precise. The correct conclusion is not that social position is the only inherited advantage. It is that the measured bundle of high-status institutional and social locations contains information about the child’s later position that parental income and ability do not fully absorb.
The word “information” is doing useful work. The result says that adding a variable changes what the data can distinguish. It does not tell us whether a parent’s position worked through a recommendation, a household expectation, a neighborhood, a school choice, a professional norm, or a combination of processes. Nor does it show that a child with a high-status parent received more help than every child with a lower-status parent. It shows that the observed intergenerational pattern is not adequately summarized by the conventional variables alone.
The diagram is deliberately labeled as a joint comparison and an observed association. It does not depict a proven causal pipeline from a parent’s rank to a child’s outcome. Its purpose is to show why the study estimates several outcomes at once: persistence in social position, long-run income, movement relative to parents, and entry into increasingly selective parts of the distribution.
What remains after income enters the model
The income outcome is measured at age 35. In the unadjusted specification, parental social-capital rank has a coefficient of 0.14 in predicting the child’s income rank. When parental income rank is added, the social-capital coefficient falls to 0.11; after cognitive ability and demographic controls are included, it is 0.08. The decline matters because it shows that social position is related to income and ability. The remaining relationship matters because it shows that those conventional dimensions do not explain it entirely.
The study’s comparison places the raw social-capital coefficient near the coefficients for parental income and cognitive ability, which are 0.16 and 0.14 in their corresponding unadjusted models. After the variables are entered together, all are attenuated, but social capital remains associated with child income. This is a “horse race” in the limited sense that the dimensions compete to explain the same outcome within a common model. It is not a contest with one winner. Each measure captures a different part of family background, and each loses some predictive power when correlated dimensions are acknowledged.
The result changes the interpretation of conventional income mobility estimates. In the baseline parent–child income model, the income rank coefficient is 0.16. When parental social-capital rank is added, it falls to 0.13, a reduction of slightly more than 20 percent. That reduction does not mean that one-fifth of every child’s income is caused by social connections, nor does it identify a precise share of income inheritance that policymakers could remove. It means that an income-only coefficient partly reflects the fact that high-income families are distributed differently across the measured institutional and social positions.
This is a measurement result before it is a policy result. A statistic is an interpretation of a relationship, and the interpretation changes when an omitted, relevant dimension is measured. If a simple income correlation includes both purchasing power and social position, calling the entire coefficient “the effect of parental income” is too strong. At the same time, the adjustment is not a complete decomposition. The index is a proxy, and unmeasured forms of advantage may remain in the income coefficient or in the social-capital coefficient.
It helps to imagine the logic of the comparison without turning it into a personal forecast. Start with families who occupy different social-capital ranks but are similar on the parental-income measure. The first model records a relationship between those ranks and the child’s adult income rank. Add parental income, then ability and demographics, and ask how much of the relationship remains. The fall from 0.14 to 0.08 says that part of the initial association overlaps with those other dimensions. The surviving coefficient says that the recorded social-position measure still distinguishes outcomes within that more comparable statistical group. It does not say that the remaining difference is a pure social-capital effect, because families may still differ in unmeasured resources, selection histories, expectations, or opportunities.
The same logic explains why the income coefficient should be interpreted carefully. When social position is added and the income coefficient falls from 0.16 to 0.13, the model has not discovered a transferable share of money that can be removed from a child’s life. It has shown that the original income relationship bundled together at least two correlated forms of background. A richer measure could change the estimate again; a weaker proxy could leave more hidden variation behind. The finding is therefore about the adequacy of the map. It warns against treating a familiar coefficient as if it named one mechanism simply because it uses one familiar variable.
The study also examines relative upward mobility: whether the child’s income rank at age 35 exceeds the parent’s rank. Here the dimensions point in different directions. Higher parental social-capital rank is associated with a greater likelihood of the child moving above the parent, conditional on the other background measures. Higher parental income rank is associated with a lower likelihood of outranking the parent. That negative relationship is compatible with mean reversion: children of very high-income parents have more room to fall relative to the parent’s exceptional position, even if their own absolute incomes remain high.
This distinction prevents an easy but misleading conclusion. High parental income can raise a child’s absolute position while reducing the probability that the child ranks above the parent. Social position can also raise the child’s absolute position, yet its association with relative upward mobility points in the other direction. “Doing better than one’s parents” is a comparison between ranks; “earning more” is an outcome level. They answer different questions about mobility.
The estimates are strongest when read as evidence of incremental predictive content. They show that the social-position measure helps locate which children are likely to have higher adult income and social status after conventional background measures are included. They do not establish that the index is a treatment that could be increased by a fixed amount, or that a specific institutional position is the active ingredient. A measured difference may reflect selection into positions, family processes that precede the position, or several channels operating together.
Why the upper tail changes the interpretation
Average rank relationships can hide where a background measure matters most. The study therefore examines children’s probability of entering the top 10 percent and top 5 percent of the age-35 income distribution. In specifications that include parental income and cognitive ability, a one-percentile increase in parental social-capital rank is associated with a 0.09 percentage-point increase in the probability of reaching the top 10 percent and a 0.06 percentage-point increase in the probability of reaching the top 5 percent.
Those are marginal associations across the parental rank distribution, not a guarantee that a child will enter an elite bracket. But they reveal why the distributional location of an effect matters. A social-position measure can be important for identifying who reaches scarce, high-reward opportunities even if its average contribution to explained variation looks modest. Once most children are concentrated in the broad middle of the distribution, the question “who reaches the top?” may expose differences that an average income comparison smooths over.
The paper’s results also show that parental social capital and parental income remain comparable predictors of upper-tail attainment after cognitive ability is included. That does not make them substitutes in every sense. A family may use money to pay for resources, while social position affects information, institutional familiarity, or access to people and organizations. The two dimensions can move together in some households, but the regression result is valuable precisely because it retains them as distinguishable predictors.
There is another reason to resist treating the upper-tail result as a simple story about inherited privilege. The measured social position is built from adult institutional locations, and the outcome is an adult income rank. Between the two lie education, occupational choice, employer sorting, residence, labor-market conditions, and personal decisions. The study’s data can show that the parent and child positions are linked; they cannot assign each part of the link a fixed causal percentage. The upper-tail association is a reason to investigate mechanisms, not permission to invent one.
The evidence also has scope conditions. The main estimates are ordinary least-squares or linear-probability associations, even when the paper conditions on several background measures. Large sample size makes the estimates precise, but precision is not the same as causal identification. The rank definitions and income thresholds determine what the coefficients mean. The comparison with Norwegian income mobility supplies context, not a new estimate of mobility in every country. And because the social-capital index captures only recorded dimensions of position, it may miss family advantages that are important but not visible in education, occupation, firm, or residential records.
A careful reader should therefore carry away three conclusions. First, social position is persistent across generations even after income and cognitive ability are considered. Second, it predicts adult income and changes the interpretation of the parent–child income relationship, so income alone is an incomplete summary of inherited advantage. Third, its importance is especially visible in upper-tail outcomes, where access to several complementary forms of advantage may determine who reaches scarce opportunities. None of those conclusions requires saying that income is irrelevant or that social position dictates an individual life.
The next question is how these patterns accumulate over time. A parent–child coefficient measured at age 35 tells us that a relationship exists, but not whether the gap was already present at school-leaving age, emerged during job entry, or widened as careers unfolded. Understanding that timing is necessary before deciding which mechanisms—and which kinds of intervention—could plausibly change the trajectory.