Abstract: Cross-regional flow and operation of agricultural machinery is common and it makes the development of agricultural mechanization in each province no longer independent. Therefore, examining the spatial correlation is of great significance to scientifically evaluate the effect of provincial agricultural machinery purchase subsidy policy. This paper applied the Moran index to verify the spatial correlation of the mechanization level of each province, and constructed a spatial Dubin model to analyze the impacts of agricultural machinery purchase subsidy on the level of agricultural mechanization under the consideration of spatial effects. Results show that since the implementation of agricultural machinery purchase subsidy policy, the comprehensive mechanization level has been improved significantly. In 2017, the level of comprehensive agricultural mechanization doubled from the level of comprehensive agricultural mechanization in 2004, from 34.32% in 2004 to 66.26% in 2017. At the same time, due to the cross-regional operation of agricultural machinery across the country, there are spatial positive correlations among the levels of agricultural mechanization among various regions. Neglecting the spatial correlation could lead to the overestimation of the impacts of agricultural machinery purchase subsidy on the level of local agricultural mechanization, and the overestimation could be 1.5 times of the actual value. In summary, this paper provides the following suggestions: 1) continuing to implement and give a full play of the subsidy policy for agricultural machinery purchase; 2) improving the subsidy policy for agricultural machinery purchase in order to improve the efficiency of subsidy; 3) considering the spatial correlation when evaluating the impacts of subsidy for agricultural machinery purchase on the level of agricultural mechanization; and 4) improving the relevant policies for promoting cross-regional operation of agricultural machinery.