Child fostering

Why do households with food to spare still send their own children away to be raised by relatives?

In a good many societies, a substantial share of children spend part of childhood raised by someone other than their parents — an aunt, a grandmother, a mother's brother — while the parents are alive, together, and able to feed them.

The reflex explanation is hardship: you send a child away because you cannot keep them. But that reading predicts fostering should run only up the wealth gradient, from poor households to richer ones, rising in bad years and vanishing in good ones. Where the practice has been studied carefully, flows run in both directions, and prosperous households send children too. So what is accomplished by a movement that hardship does not explain?

Ask first what a household is, functionally. It is a unit that has to match a set of tasks to a set of people over decades — and the match is chronically bad, because children arrive in an order nobody controls. One household needs help with an infant; another has a trade to teach; a third is near a school and a fourth a day's walk from one. None of this is poverty. It is mismatch, and it is unavoidable.

Now notice that a child is the one household member who can be relocated without dissolving anything. Move a child and both households' matches improve. That reading recurs in the ethnographic record: children placed to learn a craft, placed with a relative near a school, placed where an extra pair of hands is needed and there is capacity to feed them.

But that is not sufficient, because it predicts placements chosen by fit alone — send the child wherever the need is greatest. Observed placements instead follow kin lines with striking regularity, and often specific ones, such as a mother's brother rather than an equally convenient neighbour. Why would the relationship matter if the point were matching?

Here is the step that makes the practice legible. A child in another household is not only a child being raised. It is a standing tie between two households that outlives any particular exchange. The sender has an ongoing interest in the receiver; the receiver has a claim on the sender's future help; the child, grown, has two sets of adults with obligations to them. Fostering is a durable connection made out of a person rather than a promise — which is the point, because a promise between households with no court to enforce it is worth very little.

That suggests a second function: redundancy. A household that keeps all its children at home and then suffers a shock — a death, a bad harvest, a displacement — has everything exposed to one event. A household whose children are distributed across several others is connected to several independent sets of resources, and so is each child. Its value is highest where formal insurance does not exist.

Does the account make a prediction sharp enough to be wrong? Yes. If fostering were distress relief, it should be concentrated in bad years and among the poorest senders, and run one way. If it is training, matching and alliance, it should persist in good years, run in both directions across the wealth gradient, follow kin lines rather than proximity, and be most common where schooling and craft training sit away from home. Those are different worlds, and distinguishable in data.

What would refute the reading offered here? A well-measured setting in which fostering flows ran only from poorer to richer households, rose sharply with hardship, and disappeared in good years, would leave nothing for training or alliance to explain. Equally, if placements were unrelated to kinship once distance and household resources were accounted for, the alliance leg would collapse.

Several things must be held together here.

First, the mechanisms are not alternatives. Training, task-matching, schooling access, alliance and risk-spreading coexist, and any given placement typically serves more than one. Their relative weight varies by society, and I will not claim a ranking.

Second, distress fostering is real. Placements after a parental death, a separation, or severe hardship are well documented, and nothing above denies them. The argument is narrower: distress cannot be the general explanation, because it fails to account for the substantial share of placements from households under no such pressure. Lumping the two together is what makes the practice look mysterious.

Third — the hardest point — the large literature comparing outcomes for fostered and non-fostered children should be read with real caution. Placement is chosen, not assigned, and which child goes where is decided by the people who know most about that child and household. Any difference in later schooling, health or earnings therefore confounds the effect of fostering with the reasons for the placement. The designs that get closer are comparisons within families, asking why one sibling was sent and another kept, and settings where an external shock moved placements for reasons unrelated to the child. Where neither exists, the effect of being fostered is not identified, and I decline to state figures.

Fourth, an interpretive caution. "Alliance" and "risk-spreading" are functional accounts of why an institution is stable, not claims about what anyone is thinking. A grandmother taking a grandchild is not running a portfolio calculation, and reading these mechanisms as motives misdescribes her.

Finally: this describes institutions in the societies where they have been studied, not a template, a recommendation, or a comment on how anyone should raise a child.

• Why some societies foster to the mother's kin and others to the father's, and what descent rules have to do with it.
• What happens to these arrangements when schooling becomes universal and local.

Every term the collection defines is gathered in the glossary.