Using the 2013 China Education Panel Survey (CEPS), we study the impact of a 2008 inclusive education policy, through which the central government mandated urban public schools to exempt migrant children from tuition and temporary schooling fees. Whereas the non-disclosure rule regarding geographical location of CEPS sampling units precludes the control of locational characteristics, we identify the causal effect of the policy through a novel identification strategy, which relies on the types of primary sampling units. Specifically, we only use non-migrant rural hukou children living in counties in the nationally representative sample as the control group (the never-takers), while, in the treatment group, we only include migrant children who are currently living in China's top 120 migrant-receiving counties or city districts, and Shanghai. We also distinguish migrant children who started urban schooling before and after 2008 as separate treatment groups of always-takers and compliers, respectively. Using the Inverse Probability Weighted Regression Adjustment (IPWRA) approach, we find that the average treatment effect of the policy on migrant children is around 0.18 SD, as measured by a standardised cognitive test score – a large effect. We also present complementary evidence that the average treatment effect tends to be larger for municipalities and provincial capitals, consistently with the notion that the (potential) value-added of attending urban schools is higher the larger the initial gap with rural schools.