Abstract:    The spatial agglomeration of digital-skilled talent is becoming increasingly evident, forming a substantial digital "labor pool" that provides critical support for enterprise innovation. This study utilized the extensive online recruitment data to construct indicators of digital-skilled talent agglomeration and applied the causal forest method to address traditional econometric endogeneity issues. It investigated the impact of digital-skilled talent agglomeration on enterprise innovation levels and structures, as well as its transmission mechanisms. The findings revealed that the spatial agglomeration of digital-skilled talent significantly enhances the overall level of green innovation among local enterprises but does not lead to substantial changes in innovation structures. In cities with digital development levels above the median, digital-skilled talent agglomeration has a more pronounced effect on promoting non-green innovation activities while "crowding out" some high-risk, low-return green innovation projects. This agglomeration drives enterprise digital transformation, amplifies scale effects, and consequently promotes the overall development of green innovation. These results have suggested that governments should be cautious of the potential "crowding-out" effect of advanced digital technology development on green innovation projects, thus providing important theoretical insights for achieving the goals of "carbon peaking and carbon neutrality".

Key words:  labor agglomeration, green innovation, machine learning, crowding-out effect, digital transformation