The total quota control and reserve of rare earths are important means for the sustainable development and utilization of rare-earth resources. Focusing on the government reserve of rare-earth products under stochastic demand, this paper analyses the interactive decisions of the government and the rare-earth firms from a game-theoretic perspective. The government determines the total quantity, reserve quantity and reserve–release quantity of the rare-earth products to maximize social warfare, while the firm decides the price of rare-earth products to maximize its own profit. The results show that the production cost and the expected net present value (NPV) of the reserve are important factors affecting the government’s decisions. When the expected NPV of the reserve is below a threshold, the government adopts the no-reserve strategy: it determines only a total quota index that maximizes the current-period social welfare but keeps no reserve. When the expected NPV of a reserve is higher than the above threshold but lower than the production cost, the government adopts the low-reserve strategy: it determines a total quota index and a low reserve that are both in increasing in the expected NPV of reserve, and will release the reserve as many as possible if there is a supply shortage. When the expected NPV of a reserve is higher than both the above threshold and the production cost, the government adopts the high-reserve strategy: it sets a total quota index which is sufficiently large to cover the entire market demand, reserves a large amount, and releases part of the reserve to completely fill the demand gap (if any).
Against the backdrop of China’s “dual carbon” strategy and regional integration, enhancing green innovation efficiency (GIE) has become a core issue for the Yangtze River Delta Urban Agglomeration (YRDUA) in achieving sustainable and high-quality development. This study employs the Super EBM model to measure the GIE of 41 cities in the YRDUA from 2012 to 2022 and further integrates a modified gravity model with social network analysis to uncover the structural complexity and spatial directionality of its spatial association network. In addition, the Exponential Random Graph Model (ERGM) is applied to explore the formation mechanisms of the green innovation efficiency network. Results show the following: (1) GIE presents a fluctuating upward trend, with the mean rising from 0.747 in 2012 to 0.906 in 2022 and disparities gradually narrowing, but provincial gradients persist, implying potential “Matthew effect” risks. (2) Network density continues to increase, with S-density rising from 0.0061 in 2012 to 0.0335 in 2022; supporting and basic connections serve as key drivers of network complexity, whereas the significant decline of edge connections may weaken the network’s extensibility. (3) Node connections display preference and attachment, causing polarization; transitivity and triadic cooperation rise markedly, increasing by 41.89% and 40.86%, respectively, reflecting strong self-organization. (4) Reciprocity and agglomeration drive network formation, and economic and technological differences promote it, while disparities in innovation input and government roles vary across periods. Geographic distance hinders formation, though its effect is weakening. These findings enhance the methodological approaches to sustainability research and provide insights for optimizing regional cooperation and advancing green integration in the YRDUA.