Can an Extra Year at School Prevent Dementia? Three Natural Experiments Give Different Answers

The question is not whether education is valuable. It is whether extending compulsory education changes the probability of developing dementia, and what the strongest available experiments can actually tell us.

In 1972, Britain raised its school-leaving age from 15 to 16. People born on or after 1 September 1957 faced the new requirement; those born immediately before the cutoff could leave under the old rule. More than fifty years later, that boundary has become a useful scientific experiment.

On 23 October 2026, a team publishing in the Proceedings of the National Academy of Sciences (PNAS) reported fewer recorded diagnoses of Alzheimer’s disease and related dementias among people whose education was affected by the reform. It was a striking finding, especially because the team also examined inherited dementia risk. [1]

But a separate team, publishing in Social Science & Medicine in February, investigated the same policy and found no statistically detectable reduction in dementia risk. [2] Neither result by itself settles what schooling does to the brain. Their differences reveal what has and has not been demonstrated.

What changed in 1972?

A conventional comparison between graduates and early school-leavers is hard to interpret. Family income, childhood health, parental education and cognitive ability all influence how long someone stays in school. They may also matter for health decades later.

A compulsory-schooling reform gives researchers a better comparison. People born very close to the legal cutoff grew up under broadly similar conditions but faced different minimum school-leaving requirements. In the PNAS data, the proportion staying until at least 16 rose by 14 percentage points at the cutoff. That means roughly 14 additional pupils per 100 stayed until that age because of the reform; it does not mean everybody received an extra year. The researchers estimated that the average school-leaving age rose by about 0.19 years in their main sample. [1]

This method, known as regression discontinuity, can provide unusually strong evidence about causes if other factors change smoothly across the cutoff. It cannot automatically tell us whether an outcome reflects less underlying disease, delayed symptoms or differences in medical diagnosis.

What the October paper really found

The PNAS investigators used UK Biobank data and diagnoses drawn from hospital records, death registers and participants’ reported medical histories. Their working samples varied by analysis, with more than 100,000 people in some comparisons and 38,647 in their preferred schooling-effect estimate. The primary dementia outcome was whether someone had been diagnosed by December 2022, when the first group affected by the reform was 65. [1]

In their preferred statistical model, the estimated effect of one extra year of education was a 0.88-percentage-point reduction in the cumulative probability of a recorded dementia diagnosis. The 95% confidence interval ran from a reduction of 1.763 percentage points to a reduction of 0.154 percentage points; the authors reported a p-value of 0.022. [1]

That distinction matters. A result can be statistically inconsistent with no reduction under a particular model while its central estimate remains a poor basis for predicting how many cases a policy would prevent. The study is evidence of a possible protective effect on recorded diagnosis, not a reliable promise that an extra year of schooling eliminates a specific fraction of dementia cases.

The genetic findings need similar care. The team reported that the decline was statistically detectable within its higher-genetic-risk group, but the statistical tests did not establish that the benefit differed reliably between the higher- and lower-risk groups. The paper itself acknowledges insufficient precision for that contrast. [1]

A second British study, a different test

Nathalie Monnet, Emma Nichols, Clémence Kieny and Mauricio Avendano examined the 1947 and 1972 school-leaving reforms using the English Longitudinal Study of Ageing (ELSA). Instead of focusing on hospital-record diagnoses, they used cognitive assessments and a model designed to estimate dementia or mild cognitive impairment risk. For cohorts around the 1972 cutoff, participants averaged about 56 years of age. [2]

The reform clearly raised educational attainment in their sample: about half a year, depending on the specification. But the study did not find a statistically detectable improvement in its cognitive or estimated dementia-risk outcomes for the 1972 reform. In one 24-month-bandwidth comparison, the modelled dementia-risk estimate was 0.001 with a standard error of 0.010; other specifications also failed to yield a clear effect. [2]

That is not proof that education had no effect. Some estimates were imprecise and the samples were considerably smaller. The PNAS paper and the ELSA study also measured different outcomes at different ages. A clinical diagnosis at 65 is not interchangeable with a predicted impairment risk or cognitive test result in a survey whose 1972-reform participants averaged 56.

The ELSA researchers did find more encouraging results for the 1947 reform, especially in certain analyses involving women and people from less-educated families. Those effects concerned cognitive outcomes and combined dementia/mild-impairment risk; the evidence for diagnosed dementia alone was weaker. The researchers themselves cautioned that effects depend on historical setting. [2]

The Swedish experiment changes the picture

These two British studies are not the only useful natural experiments. A 2021 investigation in the American Journal of Epidemiology followed 1,341,842 people exposed to a Swedish schooling reform. It studied dementia diagnoses at much older ages, up to roughly 79–96, rather than stopping around 65. [3]

The Swedish estimate was close to no effect: a hazard ratio of 1.01, with a 95% confidence interval from 0.98 to 1.04. That study also found limited effects of the reform on later socioeconomic circumstances, such as income. It raised a possibility that is highly relevant to the British comparison: a school-leaving law may affect dementia risk differently depending on whether it also changes qualifications, earnings, employment and subsequent health. [3]

That is a plausible mechanism, not a demonstrated resolution. The reforms occurred in different countries and periods, and the studies used different populations and designs. Yet the Swedish evidence is too substantial to leave out when assessing whether one additional compulsory school year is generally protective.

Does schooling prevent disease or change its appearance?

One important concept is cognitive reserve. Education could help a person maintain cognitive function despite some underlying age-related or neurological changes, delaying when symptoms become severe enough for diagnosis. That would matter enormously to patients and families. But it is not quite the same claim as preventing every biological process that causes dementia.

The October researchers also found changes consistent with other possible pathways, including higher household incomes and fewer strokes or heart attacks. Their design could not determine which mechanisms explained the dementia result. Nor could it prove that differences in diagnosis played no part. [1]

Separate evidence shows why ordinary education–dementia associations are hazardous to interpret. A 2024 Norwegian study of 207,814 men found that an apparent association between low education and dementia weakened markedly after controlling for cognitive-test performance around age 18: the reported hazard ratio fell from 1.37 to 1.08. This adjustment does not disprove a causal benefit from schooling, but it illustrates how strongly early-life differences can distort simple comparisons. [4]

UK Biobank adds another limitation. Its participants were volunteers and are not perfectly representative of the British population. Even a careful birth-cutoff design cannot automatically guarantee that an estimated effect generalises to everyone, especially groups not well represented in the study. [5]

The answer that has not arrived yet

The most defensible conclusion is narrower than an announcement that schooling prevents dementia, but more interesting than a declaration that scientists disagree.

One British natural experiment found evidence that raising the school-leaving age reduced recorded diagnoses by 65, with a statistically meaningful but imprecisely sized effect. A separate analysis of the same law found no detectable improvement on different measures among younger participants. A far larger Swedish experiment found little or no benefit from a different reform as participants aged much further into the years when dementia becomes common.

These results can coexist. They concern different measures, contexts and stages of life. The decisive next evidence is not another press release about an isolated effect size. It is follow-up of the British cohorts into older ages, together with direct comparisons of diagnosis, cognitive functioning and the pathways through which education may operate.

Britain’s 1972 law did more than change classrooms. It created a test of whether an intervention in adolescence can alter disease risk half a century later. That test has yielded an encouraging result, and serious reasons not to call the question settled.