The impact of rural population shrinkage on cultivated land non-grain conversion—dual causal inference based on machine learning

This is the authors' abstract. We don't add key points for this paper.

Against the backdrop of rapid advances in new urbanization and continued rural population contraction, the expansion of cropland non-grain use poses a potential challenge to national food security. Based on 3,598 observations from 257 prefecture-level cities in China over the 2011–2024 period, this study introduces a double machine learning approach and combines two-way fixed-effects models, threshold models, and spatial Durbin models to systematically examine the mechanisms through which rural population contraction affects cropland non-grain use. The results show: (1) rural population contraction significantly raises the degree of cropland non-grain use: for every one-unit increase in rural population contraction, cropland non-grain use rises by 0.089 units. (2) The relationship exhibits nonlinear characteristics, with per-mu grain returns serving as the threshold variable. As per-mu grain returns increase, the effect follows a pattern of “strong promotion–weak promotion–significant inhibition.” (3) Heterogeneity results show that the promoting effect of rural population decline on cropland non-grainization is more pronounced in eastern regions and major grain-marketing areas. (4) Spatial econometric analysis confirms that rural population decline exhibits significant positive spatial spillover effects, with the magnitude of the indirect spillover effect substantially exceeding that of the local direct effect. These findings are expected to provide empirical evidence and decision-making references for advancing differentiated and tiered governance of cropland non-grainization and strengthening food security.