This site has spent a lot of ink on a single study. Pfeffer and Killewald's Generations of Advantage (Social Forces, 2017), fifty years of American panel data, is the source behind our transmission channels model: the...
This site has spent a lot of ink on a single study. Pfeffer and Killewald's Generations of Advantage (Social Forces, 2017), fifty years of American panel data, is the source behind our transmission channels model: the finding that houses, degrees, and marriages carry family advantage, and bequests carry surprisingly little of it. We have quoted its correlations so often that a careful reader could walk away with exactly the wrong conclusion.
So let this piece say the quiet part. The study's headline number, a parent-child wealth rank slope of 0.39, is evidence that most of what happens to your family's wealth is not inherited.
That is not a concession the data reluctantly permits. It is what the number means.
Start with what a rank slope is, because the machinery matters here. Pfeffer and Killewald lined up 4,608 parent-child pairs from the PSID, the American panel study running since 1968, and ranked everyone's wealth against their own generation. Then they asked: if a parent stands ten percentile ranks higher than another parent, how much higher does their child stand, on average, than the other parent's child?
The answer is about four ranks. That is the 0.39.
Run the arithmetic forward and the number starts talking. A parent at the 90th percentile of their generation, comfortably rich by any local definition, should expect a child near the 66th percentile of the next one. Well above average. Nowhere near the parent's perch. A parent at the 10th percentile, poor by the same definition, should expect a child around the 34th. Still below average, and pulled a long way up from the parent's position. The statistical name for this is regression to the mean, and wealth regresses hard: on this slope, the expected family drifts roughly forty percent of the way back toward the middle every generation.
And that is only the average drift. Square the correlation to get the share of variation explained and you get roughly 15 percent. Read it plainly: knowing exactly where the parents rank tells you about 15 percent of what you would need to know to place the child. The other 85 percent of the variation in where children land comes from somewhere else. Choices, marriages, health, timing, business outcomes, macroeconomic luck, behavior. The channels our whole model is built on, homeownership at 28.4 percent of the correlation, education at 25.5, marriage at 14.2, bequests at 12.3, business at 8.0, are shares of the 15 percent, not shares of the child's fate.
Attach the critiques before anyone builds on this. The 0.39 is one country's number; transmission is tighter in some economies and looser in others, and nobody has measured most of Africa at all. Rank correlations also compress the extremes, so the very top of the distribution is stickier than the average slope suggests; dynasties with governance defend rank better than the middle does. And "unexplained variation" is not the same as "freely chosen variation"; a currency collapse is in the 85 percent too. Granted, all of it. The core reading survives every caveat: the modal outcome for a family, in the best transmission data we have, is substantial movement.
Movement is an abstraction until you watch it run in each direction.
Downward first, because the canon's most famous family is the proof. Cornelius Vanderbilt died in 1877 holding roughly one hundred million dollars, more than the United States Treasury at the time, the largest private fortune the country had produced. His descendants had everything the correlation is made of: the houses, the elite educations, the strategic marriages, the bequests. Within fifty years no Vanderbilt ranked among the wealthiest Americans. When 120 members of the family gathered for a reunion in 1973, as Arthur T. Vanderbilt II documents in Fortune's Children, there was not a millionaire among them. The richest starting position in America produced, within a century, a family clustered around the ordinary. If the correlation were destiny, that story would be impossible. It is instead the direction the arithmetic points: from the top, regression to the mean has only one way to go, and behavior decides how fast the trip is.
Now upward, at scale. Africa's private wealth is overwhelmingly first generation. PwC's Africa Family Business Survey finds the continent's family firms younger than their global peers and concentrated in the first and second generation; the founders are alive, many still at the desk. Every one of those balance sheets is a family that moved up faster than any inherited correlation can explain, because there was nothing to inherit. A trader in Kampala whose parents farmed a smallholding, a Lagos founder whose father was a schoolteacher, a Nairobi professional whose grandmother could not read: these are not exceptions the data tolerates. They are the 85 percent, walking around. A statistic built on American panel families would have assigned most of them a modest expected rank, and they ignored it, which is precisely what an expectation with 15 percent explanatory power invites.
The same openness that let them rise is the openness that took the Vanderbilts down. It is one number, read from both ends.
Here is why this arithmetic deserves its own article rather than a footnote. Transmission data, badly read, manufactures fatalism in two opposite flavors, and both are behavior-destroying.
The first fatalism belongs to families starting from little: we are cursed to lose, the game is rigged, the correlation is a wall. This reading treats 0.39 as if it were 0.9. It is common, it is understandable, and in the data it is simply false. A slope of 0.39 is a headwind, not a wall; the child of the 10th percentile is expected at the 34th, and the expectation itself explains a seventh of the outcome. The wall reading has a real cost, because a family that believes movement is impossible stops doing the things that produce it. Why finish the degree, title the land, formalize the shop, if the end is written? Fatalism downward is self-executing: believe the correlation is destiny and you will supply, through abandoned effort, the destiny the correlation never promised.
The second fatalism belongs to families that have arrived: our children are set. The house is bought, the schools are paid, the accounts are full; the machine will carry them. This reading makes the same statistical error in a more comfortable chair. The child of the 90th percentile is expected at the 66th, and that is the average case, before anyone's son meets a borrowed-money idea he likes. The Vanderbilts are what "our children are set" looks like when it runs for four generations unattended. Complacency upward is also self-executing: the families that hold rank in the data are, on the evidence of every dynasty this site has profiled, the ones behaving as if they could lose it, because they can.
Notice the symmetry. Both fatalisms take a probabilistic statement and read it as a verdict, and both then stop the exact behaviors that operate in the unexplained 85 percent. The honest statistical reading is also the motivating one, which is rare enough in this literature to be worth underlining.
If most of the variation is open, the practical question is what fills it. The study itself points at the answer from an unexpected angle: the channels that carry the correlation are themselves behaviors, and they remain behaviors for families with no correlation to carry.
Homeownership explains the largest share of transmission among families that have wealth to transmit. But buying and titling a first property is available, with effort and time, to families the data would place nowhere. Education carries a quarter of the correlation; enrolling, persisting, and finishing is the same act whether a grandparent funds it or a SACCO loan does. Marriage stability, business formation, the avoidance of catastrophic uninsured shocks: each is a measured channel of inherited advantage and each is, simultaneously, a lever that requires no inheritance to pull. The transmission channels are not a description of a closed club. They are a published list of what moved families in the one place anyone counted, and the list is mostly actions.
This is the sense in which behavior at the margins decides direction. The correlation sets the current. The current is real, and pretending otherwise is its own dishonesty; a child of the 90th percentile can coast for a generation and a child of the 10th cannot. But a 0.39 current is one a determined swimmer moves against, and the drift toward the mean means the current itself weakens toward the middle. Families rise by running the channels deliberately. Families fall by assuming the channels run themselves.
Name the one behavior you control that the data says matters most for your tier, and put a date on it.
If your family rents and holds no title, it is homeownership: the largest single channel, and the first rung is a titled plot, not a finished house. If the title exists but the next generation's schooling is unfunded or stalling, it is education completion, the quarter-share channel that pays in every economy ever measured. If house and degrees are handled, it is the stability of the marriages and the formality of the business, the channels that quietly decide whether the first two compound or leak. And if your family is already high on the ladder, your channel is the one this article exists to point at: governance against the drift, because from up there the mean is below you, and 0.39 is the speed of the fall for families that stop steering.
One behavior. Yours will be obvious within a minute of honest thought, which is rather the point: the data's other half says the obvious thing, done deliberately, is most of what was ever being inherited.
This piece did its job if you can say both sentences without flinching: the correlation is real, and the correlation explains a seventh of the story. Then go act like someone who believes the other six sevenths are being written now, by you, because they are.