We evaluate the effects of participation in the Swedish version of the Science and Technology for Children Program on content and process skills in sciences, in grade 9. The Swedish version, called Natural Sciences and Technology for All (NTA), is predominantly employed in grades 1–6. Our outcome measures are scores and grades on nationwide standardised tests, and course grades, in biology, chemistry and physics, for the years 2009 and 2010. A nationally representative random sample of almost 16,000 test‐taking students is coupled with multi‐level information about the NTA, and background factors. Non‐random selection into the programme is addressed by propensity score analysis. The matched sample has almost maximum common support and is well behaved in terms of propensity scores. Accounting for selection is shown to be very important. We find significantly positive effects on national test scores (effect size 0.24) and national test grades for physics, but not for biology and chemistry. With respect to course grades, we find no significant effects at all. We consider explanations for the differences in the estimated effects across science subjects and between types of outcome variables, i.e. national standardised tests versus course grades.