Learning loss due to school closures during the COVID-19 pandemic

approximately 350,000 Dutch primary-school pupils

Dutch primary-school children, national examinations before and after the spring 2020 lockdown, compared with the same window in the three previous years

The cleanest natural experiment available on instruction delivered at home by parents — and a deliberately unfavourable one. The Netherlands closed schools for only 8 weeks, funds schools equitably and has the world's highest broadband penetration, so this is close to a best case for emergency home learning. Robust to propensity balancing, maximum-entropy weights, and school and FAMILY fixed effects. The exposure is involuntary, unplanned, unchosen home learning with working parents and no curriculum — the opposite of the selection that defines home-schooling, which is exactly why it is informative and exactly why it must not be read as a test of home-schooling.

A learning loss of about 3 percentile points, or 0.08 SD — equivalent to one-fifth of a school year, the same period schools were closed. In other words students made little or no progress while learning from home. Losses were up to 60% larger among children from less-educated homes. The effect reflects cumulative knowledge not learned rather than test-day disruption.

Low. Assignment is a national closure imposed on every child simultaneously; family fixed-effects specifications compare siblings. Genes cannot differ between the exposed and the pre-pandemic comparison cohorts.

Replicated in direction and magnitude class by the Betthäuser et al. 2023 systematic review and by national analyses across many countries; the Dutch estimate is at the small end because the closure was short and the infrastructure unusually good.