2025年4月,美国启动新一轮全球税收重构计划,推动对多国贸易伙伴实施“对等征税”机制,迅速引发中国出台反制措施。全球税收规则的制度性重塑,构成新型外部冲击源,在短期内强化了国际资本的套利动机与趋利行为,触发跨境资本流动的周期错配与结构跃迁。相较传统基于利差与汇率波动的理论分析路径,此类制度性冲击更易引发政策预期变化、情绪放大效应及市场适应机制的非线性响应,传统方法难以有效识别其动态演化过程。本文聚焦全球税收博弈背景下中国短期跨境资本流动的非对称性调整机制,基于2008—2024年季度数据,构建多门槛回归模型,并引入生成式人工智能对政策文本进行语义建模,构造政策语义强度指数(TPSI),以系统识别资本流动的结构跃迁路径与调控政策的时变效应。研究发现:(1)全球税收冲击显著改变了资本对利差与汇率预期的响应结构,表现出区间跃迁与响应非对称特征;(2)由生成式人工智能生成的TPSI在识别资本情绪方向与流动趋势上具有显著领先性,弥补了传统变量对政策预期变化的响应滞后;(3)在高语义强度区间内,现行宏观审慎工具的调节效应边际减弱,短期资本流动存在“高频脱锚”与调控滞后风险。基于此,本文建议建立以语义预期识别为基础的跨境资本预警机制,前瞻性提升资本流动治理能力,并积极参与全球税收规则重构进程,增强国家在国际金融体系中的制度韧性与税收主权保障能力。
In April 2025, the United States launched a new round of global tax restructuring plan to promote the implementation of a "reciprocal taxation" mechanism for multinational trading partners, which quickly triggered the introduction of countermeasures by China. The institutional reshaping of global tax rules constitutes a new source of external shocks, which strengthens the arbitrage motive and profit-seeking behavior of international capital in the short term, and triggers the cyclical mismatch and structural transition of cross-border capital flows. Compared with the traditional theoretical analysis path based on interest rate spread and exchange rate fluctuations, such institutional shocks are more likely to cause changes in policy expectations, emotional amplification effects and non-linear responses of market adaptation mechanisms, and traditional methods are difficult to effectively identify their dynamic evolution process. This paper focuses on the asymmetric adjustment mechanism of China''s short-term cross-border capital flows in the context of the global tax game. Based on the quarterly data from 2008 to 2024, this paper constructs a multi-threshold regression model, introduces generative artificial intelligence to conduct semantic modeling of policy texts, and constructs the policy Semantic Strength index (TPSI). Systematically identify the structural transition path of capital flows and the time-varying effects of regulatory policies. The results show that: (1) the global tax shock significantly changes the structure of capital''s response to interest rate spread and exchange rate expectation, showing the characteristics of interval transition and asymmetric response; (2) TPSI generated by generative AI has a significant lead in identifying the direction and flow trend of capital sentiment, making up for the lagging response of traditional variables to changes in policy expectations; (3) In the high semantic intensity range, the regulatory effect of the current macro-prudential tools is weakened, and there is a risk of "high-frequency unanchoring" and lagging regulation of short-term capital flows. Based on this, this paper proposes to establish a cross-border capital early warning mechanism based on semantic expectation recognition, improve the governance capacity of capital flows prospectively, and actively participate in the restructuring process of global tax rules, so as to enhance the institutional resilience of countries in the international financial system and the ability to safeguard tax sovereignty.