Abstract: Land banking is an important means of urban land resource management. The strict land banking system has been carried out in China since the 1990s. Although it is an important public policy, local governments, as the main executors, often deviate from the objectives of public policy in the choice of local interests and public interests, resulting in inefficient implementation of land reserve policy. This paper takes the deviation of land banking policy implementation as the research object, by using the improved DEA method and Delphi method, a new method is constructed to calculate the deviation degree of land banking policy implementation based on the efficiency of land banking decision-making and analyzes the deviation degree of land banking policy implementation and temporal and spatial characteristics in 34 large and medium-sized cities. The empirical results show that, from the perspective of time dimension, the deviation of land banking policy implementation can be divided into three stages: relatively high status, U-shaped growth and volatility decline; from the spatial dimension, the cities with moderate deviation of land banking policy implementation from 2004 to 2015 are mainly located in the east and middle region, and the number of cities with moderate deviation and no deviation or approximately no deviation in the western region is equal number. Also, the degree of deviation varies greatly from city to city. On the evolution of spatial distribution, the number of cities with moderate and high deviation first increases and then decreases. The highly deviated urban distribution diffuses from the east to the middle. The research in this paper provides useful information for the evaluation and optimization of land banking policy implementation.

Key words: land banking, policy implementation deviation, decision-making efficiency