Using a unique longitudinal dataset collected from primary school students in Pakistan, we document four new facts about learning in low-income countries. First, children’s test scores increase by 1.19 SD between Grades 3 and 6. Second, going to school is associated with greater learning. Children who dropout have the same test score gains prior to dropping out as those who do not but experience no improvements after dropping out. Third, there is significant variation in test score gains across students, but test scores converge over the primary schooling years. Students with initially low test scores gain more than those with initially high scores, even after accounting for mean reversion. Fourth, conditional on past test scores, household characteristics explain little of the variation in learning. In order to reconcile our findings with the literature, we introduce the concept of “fragile learning,” where progression may be followed by stagnation or reversals. We discuss the implications of these results for several ongoing debates in the literature on education from Low- and Middle-Income Countries (LMICs).

使用从巴基斯坦小学生收集的独特纵向数据集，我们记录了有关低收入国家学习的四个新事实。首先，从 3 年级到 6 年级，儿童的考试成绩提高了 1.19 个标准差。其次，上学与更好的学习相关。辍学的孩子在辍学前的考试成绩与未辍学的孩子相同，但在辍学后没有任何进步。第三，学生之间的考试成绩增长存在显着差异，但小学阶段的考试成绩趋于一致。即使在考虑了均值回归之后，最初测试分数较低的学生也比最初分数高的学生获得更多的收益。第四，以过去的考试成绩为条件，家庭特征几乎无法解释学习的差异。为了使我们的研究结果与文献相一致，我们引入了“脆弱学习”的概念，即进步之后可能会出现停滞或逆转。我们讨论了这些结果对低收入和中等收入国家 (LMIC) 教育文献中正在进行的几项争论的影响。