A meta-analysis of students’ academic learning losses over the course of the COVID-19 pandemic.

Background: In spring 2020, the COVID-19 pandemic led to a disruption of students’ education around the world. School closures led to a shift from in-person learning at school to remote learning, with changes in education persisting over the following years. Aims: We conducted a meta-analysis to determine if the COVID-19 pandemic was associated with learning losses for school students and to identify potential moderators. We examined learning losses during the pandemic (April 2020–June 2022) and focused on learning losses at different timepoints. Sample(s): 763 effect sizes from 103 studies conducted in 45 different countries were included in our meta-analysis. Methods: After an extensive, AI-supported literature search, relevant information from the primary studies was coded. Effect sizes were transformed into Cohen's d. An average effect size was estimated and moderator analyses were conducted for school level, learning domain, country's Human Development Index level, publication type, and study quality. Additionally, overall changes in learning losses were examined in more detail, including a moderator analysis for timepoint of data collection during the pandemic. Results: Our results showed average learning losses of Cohen's d = −0.20, SE = 0.04, p <.001. Learning domain was the only statistically significant moderator with learning losses being largest in mathematics. The most pronounced learning losses occurred shortly after the onset of the pandemic, but significant learning losses were observed at the majority of the included timepoints. Conclusions: Learning losses were observed more than two years into the pandemic, highlighting the need to ensure long-term recovery.