School closures from COVID-19 have resulted in large learning losses, from 0.05 to 0.17 standard deviations in high income countries, equivalent to two to six months of lost learning. However, the extent of primary-level learning loss in low-income countries remains unclear, studies lack information on individual students' learning trajectories, and most do not include students who dropped out. This paper uses representative survey data from Malawi that includes unique longitudinal data on individual students (grade 4 at baseline), including those who dropped out, at three points in time: pre-COVID; 1-12 months before the seven-month school closures; and 14-20 months after schools reopened. Across math, English, and Chichewa, the local language, the average learning loss amounts to 18 months (78 points, 0.78 standard deviations), significantly higher than the loss documented in high income contexts. Decomposing this loss, the findings show that students lost 0.25 standard deviations of existing knowledge during the closure, and a further 0.23 standard deviations in foregone learning compared to the expected trajectory had schools remained open. Further loss comes from a slowdown in learning after schools reopened, with students gaining 7 points' less new knowledge in math per 100 days, the majority of which is not explained by increased dropout. Our findings are relevant for other low-income and lower-middle income contexts: remote learning during school closure was in general ineffectual, necessitating urgent action to remediate lost learning; and children who dropped out had the highest learning losses and now require out-of-school learning opportunities.