When I talk to people about demographics, after a while we quickly run into *the* question: why are we having fewer children? A popular explanation is that more education for women means marrying later, having children later, and ultimately having fewer children. That narrative is so deeply ingrained in how we think about education and demographics that it seems almost self-evident. But is that really the case?

A recent, original study by Fu and colleagues from Japan attempts to dissect precisely that narrative. Not with the classic correlations – which we know – but with a clever quasi-experiment. The researchers make use of the so-called “Firehorse” year. We are talking about 1966, a year that, according to superstition, is supposedly unlucky for girls. The result: there were remarkably fewer births in Japan that year. Having fewer children also means less competition at school later on. And that is exactly where their entry point lies: some cohorts received, purely by chance, just slightly more educational opportunities.

What follows is methodologically truly impressive. Using a difference-in-differences approach and large administrative datasets, they attempt to isolate the effect of extra educational opportunities. Not perfect randomisation, but certainly much closer to causality than many previous studies.

And what is the result? More education is indeed associated with marrying later and having a first child later. But those effects are small. We are talking about weeks, not years. And even more importantly, in the longer term, the differences largely disappear. By middle age, women with more education are married and mothers about as often as others.

This, therefore, clashes with the dominant narrative. Education here seems to determine not so much whether women start a family, but rather when. It is thus more about timing than about ultimate choices. That does not mean we know for sure. On the contrary. This is one context (Japan), one cohort (late 1960s), and a specific institutional setting with fairly traditional gender roles. The authors themselves are cautious about this as well. Moreover, it remains a quasi-experiment: robust, but not foolproof.

But it does put a bit of perspective on what we tend to assume (too?) quickly. Many studies that find a strong negative effect of education on fertility are observational. They compare groups that differ in multiple ways. This type of design suggests that part of that effect might not be a pure effect of education, but is related to selection: those who study longer often differ from the outset.

What also stands out: the study explicitly points to other factors. Labour market structures, expectations regarding motherhood, and the division of care responsibilities. If that context does not change, more education can actually lead to tensions: you have more opportunities, but the institutional space to combine work and family remains limited.

The question is perhaps not so much whether education “keeps” women away from children. The question is how societies deal with what education makes possible. If you create more opportunities without adjusting the context, you mainly shift the timing. If you change that context as well, the story can look completely different.