Compensation–Skills Dual-Dimensional Configuration and Sustainable Upgrading of Export Technological Sophistication: A Re-Explanation of the “Value Paradox” of Skill Mismatch

Amid the restructuring of global value chains (GVCs) and the artificial intelligence (AI) revolution, how firms’ reallocation of human capital relates to escaping value chain lock-in has become a question of growing theoretical and practical importance for China. Traditional static skill-matching theory treats skill mismatch purely as a distortion in resource allocation. In the course of China’s industrial transformation and upgrading, the co-existence of configuration deviation and industrial upgrading is widely observed. Using panel data on Chinese listed firms from 2015 to 2023, this study measures firms’ relative configuration deviations along the compensation and skill dimensions against industry-year benchmarks. Different types of configuration deviation are then identified according to the sign combinations of deviations in these two dimensions. The results show that the overall magnitude of configuration deviation is not significantly associated with firms’ export technological sophistication (ETS). Instead, the direction and combination of deviations matter: firms with a High-high configuration are associated with significantly higher ETS than those in the reverse configuration. Decomposition by dimension shows that the Compensation-premium is significantly and positively associated with ETS but negatively associated with innovation output, whereas Skill-underutilization is positively associated with innovation yet negatively associated with contemporaneous ETS. Channel tests indicate that the High-high configuration is positively associated with two interconnected prospective channels: innovative-talent accumulation and AI orientation. No robust regional heterogeneity is identified in the association between the High-high and ETS. All coefficients of interest are interpreted as conditional associations estimated within a two-way fixed-effects framework, rather than as identified causal effects. Moving beyond the static job–skill matching paradigm, we interpret these findings through a dynamic capability-building lens and discuss sustainability-oriented policy implications for developing countries seeking to design differentiated talent and digital technology policies.