Grandmother proximity and fertility

Why do couples living near a grandmother have more children than equally well-off couples living far from one?

Take two couples with the same income, education and health, one living twenty minutes from the wife's mother and one four hours away. Across many datasets and several countries, the near couple tends to have more children, and the next one sooner.

The obvious reading is that a grandmother nearby makes another child affordable. Before accepting it, notice what the sentence quietly assumes: that where a couple lives arrived independently of what they want. Did it?

Start with the mechanism the intuition proposes, because it is a real one and worth stating precisely. What makes an additional child expensive in a wealthy country is mostly not food or clothing — it is the mother's time, and specifically her attachment to paid work. The binding constraint is coverage: the hours between nursery closing and a shift ending, the week a child is too ill for school, the evening that cannot be moved.

A grandmother nearby is unusually well suited to that gap. She is unpaid, flexible at short notice, trusted enough to be left alone with an infant, and — decisively — available for the irregular, unbookable hours no purchased arrangement covers cheaply. So her proximity should not lower the total cost of a child by much; it should lower the cost of the marginal one, and lower it most where formal childcare is scarce, expensive or rigid. That is sharper than "help is nice", and testable.

It has a timing edge too. If what she supplies is coverage during the constrained early years, then proximity should compress the interval between births rather than simply raising the eventual total — a couple can start the second child before the first has aged out of the hardest phase.

Now the awkward part. Residence is chosen, and chosen by both generations. A couple who want a large family have a reason to stay near kin, or to move back. And the causation runs the other way too, on a timescale that is easy to miss: grandmothers move toward grandchildren. A researcher measuring proximity after the second birth may be measuring a consequence of the fertility they are trying to explain.

Nor is that the only route to the same correlation. People who stay in their hometown differ systematically from those who leave — in education, earnings trajectory, religiosity, housing costs, how much they value kin ties at all — and every one of those independently predicts fertility. And a nearby grandmother is also, on average, one who is alive, healthy and not herself in need of care, which tracks the resources of the whole family line.

So: one plausible causal channel, and at least three non-causal readings producing the same correlation. What would separate them?

Watching the same couple over time helps with part of it: if births follow a grandmother's arrival or retirement rather than preceding it, reverse causation is harder to sustain. Better is a shock to her availability the couple did not choose — a pension-age reform changing when she stops working, a workplace closure. Sibling comparisons hold the family line constant and ask why one adult child living near the mother differs from the one who moved. Each attacks a different confound; none removes all at once. Where no such design exists, the honest statement is that the association is robust in direction and the size of the causal part is not established.

Three refinements connect the steps above.

The first is that "grandmother" is not interchangeable with "grandparent". The help that relieves the coverage constraint is concentrated in the maternal line in most measured settings — a pattern with its own explanations, involving how certain each grandparent's link to the child is and where couples customarily live. Here it matters only as a reason not to average the four together.

The second is that the mechanism is conditional on surrounding institutions. If her contribution is coverage of the hours nothing else covers, her effect should be small where childcare is cheap, plentiful and flexible, and large where it is not; a finding that proximity mattered just as much under abundant subsidised childcare would be hard for this account to absorb.

The third is the falsification test proper. The account predicts that an exogenous change in a grandmother's availability — a retirement date set by a pension reform, a relocation driven by something other than the grandchild — is followed by a change in the couple's birth timing. The refuting observation is clean: if fertility did not respond to such shocks while remaining correlated with ordinary chosen proximity, then proximity is a marker of the couple's type rather than a cause of their choices, and the intuition we started with is wrong.

Two cautions. Everything here is a population average made of varied cases — a nearby grandmother can add work rather than remove it, and the aggregate tells no particular family what their own arrangement is doing. And I quote no numbers: published estimates vary by more than an order of magnitude across cohorts and countries, and depend heavily on how "nearby" is coded.

• What happens when the flow reverses and the grandmother becomes the one needing care during the same years.

Every term the collection defines is gathered in the glossary.