Prolonged school closure has been adopted worldwide to control COVID-19. Such preemptive implementation was predicated on the premise that school children are a core group for COVID-19 transmission. Using surveillance data from the Chinese cities of Shenzhen and Anqing, we inferred that children aged 18 or below are only around half as susceptible to COVID-19 infection as adults. Using transmission models parameterized with synthetic contact matrices for 152 jurisdictions around the world, we showed that the lower susceptibility of school children substantially limited the effectiveness of school closure in reducing COVID-19 transmissibility. Our results, together with recent findings that clinical severity of COVID-19 in children is lower, suggest that school closure may not be ideal as a sustained, primary intervention for controlling COVID-19.