The article by Izadi and Sarvimäki (2026) provides an enlightening and relevant description of the implementation of a nationwide two-year preschool experiment. It takes readers through the stages of developing the experiment, from how the policymakers agreed to a large and complicated experiment to the process of randomly allocating treatment, to discussing the data collected. Unfortunately, the last two phases were not completed by the deadline for submitting these written comments, making it impossible to comment on the experiment’s outcomes.
The experiment described looks at the effects of extending the existing one-year preschool to two years, meaning that children would enrol in preschool at age five rather than age six. Before the experiment, most five-year-olds were enrolled in daycare, meaning that the experiment, in some sense, tests the effects of enrolling five-year-olds in preschool rather than daycare.
It might be beyond the scope of the article to discuss in depth the differences between these two settings (daycare vs. preschool), but it is difficult for a non-Finn to grasp what is at stake here. One of the great strengths of the treatment description is, therefore, that the authors show how the quantifiable educational inputs have changed between the treatment and control groups.
I look forward to delving into the results when they are available. Researchers rarely have the opportunity to design experiments on this scale, and in such vital parts of the Nordic welfare model – childcare and schooling. Investing in early childhood education is considered highly cost-effective. At the peer-review conference, Sarvimäki made a strong case for greater knowledge of the effects of investments in early childhood education in settings where these investments are already relatively significant. I agree. To interpret the famous Heckman curve in the Nordic countries, we need to account for the baseline for additional investments. In this sense, this experiment is highly relevant.
On the other hand, I am slightly concerned – not having seen the results and the reactions to them – that the design, in a sense, puts all its eggs in one basket. By this, I mean that the article investigates the effects of a potential national policy to extend preschool to two years, but does not go into as much depth about the best way to implement such a policy change. I would expect the data to show local differences in effectiveness, differences that might be the result of how successfully the change has been implemented. The extent to which the researchers have collected detailed data on local differences in adaptation is unclear to me.
Conducting such a large-scale experiment places significant responsibility on the researchers coordinating it and the civil servants with whom they work. For the experiment to maintain its legitimacy, it is important that there is also broad support for the implementation. Data collection, in particular, can be perceived as disruptive and time-consuming, taking time away from the professional’s core task – education. I am therefore curious about how Izadi, Sarvimäki, and their collaborators have balanced the need to collect reliable assessment data on the one hand (carefully laid out in the paper) with the need not to disrupt more than necessary on the other. I would worry that if, for example, teachers find the data collection time-consuming and of no benefit to the children, the experiment might lose (some of) its legitimacy, and policy makers might be more reluctant to experiment in the future.