Family Size and Children’s Cognitive Development

Children from larger families score, on average, slightly lower on cognitive tests and complete slightly less schooling than children from smaller families. The pattern has been observed for over a century. The harder question is whether having more siblings causes the difference, or whether it just correlates with other things that do — lower parental income, lower parental education, different cultural settings. Half a century of theorising and a series of clean natural experiments have given the answer roughly the right shape: family size has a real but modest effect on cognitive development, smaller than the raw correlation suggests, and significantly mediated by birth order and resource allocation within the family.

Two theories of how family size could matter

The literature has long centred on two competing accounts. The Resource Dilution Hypothesis, formalised by Blake (1981), starts from a budget constraint: parents have a fixed amount of time, attention, money, and emotional energy, and adding a child divides that pool more thinly. Each additional child gets a smaller share of parental investment in language exposure, conversation, supervised reading, educational materials, and the kinds of one-on-one engagement that support cognitive development. The hypothesis predicts a monotonic negative effect of family size: more siblings, less per-child investment, lower cognitive outcomes.

The Confluence Model, articulated by Zajonc (1976) in Science, takes the opposite tack on one specific point. Zajonc argued that the average intellectual environment of a household is the average maturity level of its members, and that older siblings act as tutors and intellectual stimuli for younger ones. In a family with three children spaced two years apart, the youngest grows up in a richer verbal environment than an only child would. The Confluence Model predicts non-monotonic effects of family size depending on spacing: closely spaced large families dilute, but widely spaced ones can enrich.

The two models are not strictly incompatible — the Confluence Model addresses sibling-as-stimulus while Resource Dilution addresses parent-as-investor — but they make different empirical predictions about what raw family-size correlations should look like across different family configurations. The modern literature has spent considerable effort separating which mechanism is doing what.

The cleanest causal test: Norwegian administrative data

Most family-size studies are confounded by selection: families that have more children differ on average from families that have fewer in ways that themselves correlate with children’s outcomes. Black, Devereux, and Salvanes (2005) used the entire Norwegian birth-cohort registry to address this directly. Their key innovation was an instrumental-variable design that exploits twin births: when a second pregnancy unexpectedly produces twins, family size jumps by one for reasons unrelated to parental preferences or characteristics. Children whose family size was inflated by a twin birth could then be compared against children from comparable single-birth families.

The headline result was striking. The raw negative correlation between family size and children’s educational attainment in the Norwegian data was substantial. Once birth order was controlled, the family-size effect shrank by roughly an order of magnitude; once both birth order and the twin-IV correction were applied, the family-size effect on years of education and on continuation to higher education was statistically indistinguishable from zero. The implication is that the apparent family-size effect in observational data is largely a birth-order effect in disguise — later-born children tend to do somewhat worse, and larger families have more later-born children, so larger family size correlates with worse outcomes without family size itself being the cause.

This is a clean econometric result on a high-quality dataset, and it pushed the field to take birth order more seriously as the primary causal channel. The implications for the Resource Dilution Hypothesis are mixed: it does not necessarily refute the dilution mechanism, but it suggests that whatever dilution happens shows up most clearly in birth-order differences within families rather than in raw differences between families of different sizes.

The longitudinal test: NLSY79 mother-child data

Juhn, Rubinstein, and Zuppann (2015) approached the same question from a different angle, using matched mother-child data from the U.S. National Longitudinal Survey of Youth (NLSY79). The NLSY79 follows mothers and their children over decades and includes repeated measures of cognitive ability and behaviour, plus mother-level controls including the AFQT (Armed Forces Qualification Test) score that approximates maternal cognitive ability.

Their results identified a quantity-quality tradeoff that survives a wide range of controls: increases in family size were associated with reduced parental investment per child, lower child cognitive abilities, and increased behavioural problems. Two findings were notable. First, the effects on cognitive abilities were larger for girls than for boys; the effects on behavioural problems were larger for boys than for girls — a gender-asymmetric pattern that any theoretical account of family-size effects has to explain. Second, the negative effects were concentrated in lower-AFQT-mother families. Children of mothers with stronger cognitive resources appeared partially buffered against the effects of large family size, presumably because higher-resource parents could maintain per-child investment despite increased family size.

The Juhn-Rubinstein-Zuppann findings differ from the Black-Devereux-Salvanes findings in part because the U.S. NLSY79 sample and the Norwegian registry differ in policy environment, fertility patterns, and selection on family size. The two together support a nuanced picture: clean causal evidence for a strong family-size effect on educational attainment in a high-resource European setting is weak, while observational-with-controls evidence for a moderate family-size effect on early-childhood cognitive outcomes in a U.S. setting (with its wider variation in family resources) is stronger.

The role of birth order and spacing

One of the field’s clearer messages is that birth order does most of the work that family size appears to do in raw correlations. Firstborn children tend to score modestly higher on cognitive tests than later-born children within the same family, and the firstborn advantage compounds across larger families: a fourth-born child in a family of four faces three older siblings competing for parental attention and a parental investment style that has shifted from the higher-intensity pattern that firstborns receive.

Spacing matters too. Closely spaced siblings — brothers and sisters born within two years of each other — appear to compete more directly for cognitive resources during the developmental years, while widely spaced siblings can experience something closer to the Confluence Model’s enrichment effect. The spacing literature is messier than the birth-order literature, but the pattern is consistent enough to inform parental decision-making at the margin.

Cultural context and family resources

None of the family-size effects are properly understood without context. Poverty and resource scarcity compound family-size effects: in households where one or two children would already strain parental investment, adding a third or fourth has larger marginal cost. Cultural norms around extended-family caregiving, godparent involvement, and grandparent residence can offset some of the resource dilution that the standard model assumes occurs only between nuclear-family parents and biological children. In settings with strong extended-family support, the per-child investment burden of additional siblings is meaningfully shared.

The contrast between Norwegian (high-welfare-state, low fertility variance) and U.S. (high inequality, high fertility variance) findings reflects this: family size matters more when parental resources are tighter and external support is thinner. This also intersects with the broader literature on environmental influences on intelligence, where heritability estimates rise as environmental variance shrinks — in a more equal society, residual cognitive variation reflects more genetic and less family-environmental variance.

What it means in practice

The honest reading of the evidence does not support strong claims in either direction. Family size has some effect on children’s cognitive outcomes, smaller than the raw correlation suggests, and concentrated in lower-resource households. The mechanism is partly resource dilution and partly birth order working through how parental investment shifts as family size grows. Schooling and other environmental inputs remain the more reliable levers for cognitive development; family size is one input among many, not the dominant one.

For parents weighing how many children to have, the empirical picture is reassuring rather than alarming. Adding a child to a well-resourced family produces small marginal effects on individual cognitive outcomes, while the within-family experience of having siblings carries social and developmental benefits the cognitive-outcome literature does not capture. The decision should be made on grounds the literature does not directly speak to; the cognitive-development data alone are not strong enough to recommend a particular family size.

Frequently asked questions

Does family size really lower children’s IQ?

The raw correlation is real but the causal effect is small once birth order and parental characteristics are controlled. Black, Devereux, and Salvanes (2005) found that the apparent family-size effect on Norwegian children’s educational outcomes shrank to near zero once birth order was accounted for. Juhn, Rubinstein, and Zuppann (2015) found a moderate effect on early-childhood cognitive outcomes in U.S. data, concentrated in lower-resource households.

What is the resource dilution hypothesis?

The idea, formalised by Blake (1981), that parents have a fixed amount of time, attention, and money to invest in children, so each additional child reduces the per-child share. The hypothesis predicts a monotonic negative effect of family size on outcomes that depend on parental investment.

What is the confluence model?

Zajonc’s (1976) account, in which older siblings act as intellectual stimuli for younger ones, and the average intellectual environment of a household depends on the maturity levels of its members. The Confluence Model predicts that widely spaced siblings can enrich rather than dilute, while closely spaced ones compete.

Are firstborn children smarter than later-born children?

The within-family birth-order effect is real but small — a few IQ points on average. The effect is what most of the apparent “family size lowers IQ” finding turns out to be in clean causal designs. See birth order and intelligence for the deep-dive.

Does spacing between siblings matter?

Yes, in the predicted direction. Closely spaced siblings (within roughly two years) compete more directly for parental attention during developmental years; widely spaced siblings can experience something closer to the Confluence Model’s enrichment effect. The spacing literature is messier than the birth-order literature, but the pattern is consistent enough to be relevant at the margin.