Containing the COVID-19 pandemic by non-pharmacological interventions is costly. Using high-frequency, city-to-city truck flow data, this paper estimates the economic cost of lockdown in China, a stringent but effective policy. By comparing the truck flow change in the cities with and without lockdown, we find that a one-month full-scale lockdown causally reduces the truck flows connected to the locked down city in the month by 54%, implying a decline of city’s real income with the same proportion in a gravity model of city-to-city trade. We also structurally estimate the cost of lockdown in the gravity model, where the effects of lockdown can spill over to other cities through trade linkages. Imposing full-scale lockdown on four largest cities for one month would reduce the national real GDP by 8.6%, of which 11% is contributed by the spillover effects.