This paper considers the relationship between child labour and a child's academic achievement in rural China. Using a unique longitudinal, multi-level survey, the Gansu Survey of Children and Families (GSCF) which was enrolled in the Gansu province, I use a quasi-maximum likelihood estimation (QMLE) and find that more than 1 h of child labour in the previous time period has a negative effect on a child's academic achievement in the subsequent period after controlling for child talent. I also show that previous academic achievement has no strong significant effect on current child labour by applying a logistic model. Based on the data, the fact that those effects are not very big or not significant suggests that child labour in China is not a big problem when compared with other developing countries (Bacolod & Ranjan 2008).