Manipulation of food production data could lead to catastrophic social and economic consequences. The accuracy of official agricultural statistics has long been questioned in China. As a natural experiment, this paper studies the linkage between agricultural production data manipulation and the Granary Country Subsidy Policy (GCSP). Chinese government gave subsidies to the counties with annual grain production more than 200 thousand tons to encourage these local governments to give priority on grain production from 2005. In order to obtain the subsidies, the prospective counties with food production slightly below the threshold may have incentives to over-report their grain production. Based on the McCrary (2008) s density test, our empirical results confirm that the GCSP results in over-reporting of grain production in those countries. Furthermore, data manipulations are more likely to happen in major-grain-production, low-income and mid-western counties. The policy implication would be that the fiscal distribution rules of a central government should avoid data manipulation incentives in local governments, particularly should cut the linkage to the data which are self-reported by the local governments. Acknowledgement : Thanks