Abstract: To explore the ways to improve the production efficiency of hog breeding industry of different scales, based on a data of 15 major hog breeding provinces in China from 2005 to 2018, this paper examined the dynamic evolution characteristics of total factor productivity of different scales by the nuclear density curve, the fixed effect estimation, the two-stage least squares regression (2SLS), the comprehensive feasible generalized least square (FGLS), the spatial Dubin model and two district spatial Dubin model and analyzed the relationship between hog breeding industry agglomeration and the total factor productivity. Results show that: 1) the small-scale and medium-scale total factor productivity shows an obvious upward trend, while the large-scale total factor productivity is relatively stable; 2) the impact of industrial agglomeration on the total factor productivity of different scales shows the characteristics of an "inverted U" shape. Before the inflection point, industrial agglomeration has the greatest effect on the promotion of the small-scale total factor productivity; 3) industrial agglomeration has a significant spatial spillover effect on the total factor productivity of different scales; and 4) the spatial spillover effect of different provinces with different agglomeration levels has significant asymmetry. Therefore, this paper suggests to form a market structure system based on small scale and to promote the flow of production factors among regions. In addition, Hunan Province and Yunnan Province should extend the industrial chain and play their leading role, and the other 13 provinces should promote the process of industrial scale expansion.