Production Efficiency of Raw Milk and Its Determinants: Application of Combining Data Envelopment Analysis and Stochastic Frontier Analysis

China’s raw milk production is confronted with problems such as high production costs, stringent environmental constraints, weak industrial interest linkages, etc. The extensive and inefficient operation mode seriously restricts the further development of China’s dairy industry. How to increase the production efficiency of raw milk and realize the transition from “a country with high milk production” to “a country with high-efficiency milk production” has become the key to further developing China’s dairy industry. In order to explore the upgrading path of the raw milk industry in China, this study used the DEA-Malmquist model to estimate the production efficiency of raw milk in China and analyze its spatial and temporal distribution characteristics based on raw milk production input and output data at four scales (i.e., free-range, small-scale, medium-scale, and large-scale) from 2004 to 2020. It then adopted the SFA model to explore the relationship between raw milk input and output factors and the driving factors of production efficiency. Finally, robustness was discussed according to the existing research differences. The study draws several valuable conclusions. First, the production efficiency of raw milk in China from 2004 to 2020 fluctuated upward and showed specific regular regional distribution characteristics, but the spatial–temporal differences were minor. Second, raw milk production efficiency is significantly influenced by various factors, including the proportion of concentrate to roughage consumption, medical and epidemic prevention investment, the price of raw milk, the wage level, and fixed assets. Third, the spatial–temporal distribution and driving factors of raw milk production efficiency are less robust, so efficiency analysis and improvement measures should fully consider analytical methods, scale heterogeneity, indicator systems, and temporal heterogeneity.