A Model for Testing the Mediating Effect of Digital Capabilities on Corporate Innovation Performance

In the digital economy, many enterprises face a “high investment-low conversion” dilemma in digital transformation, where digital input does not automatically improve innovation performance. To clarify the transmission mechanism from digital capabilities to innovation performance, this study introduces knowledge reorganization efficiency as a mediating variable based on dynamic capability theory and the knowledge-based view. A hierarchical regression model is established, and a bias-corrected Bootstrap method with 5,000 repeated samples is used to infer indirect effects without relying on the normality assumption of the Sobel test. In addition, the industry-level mean of digital capabilities is used as an instrumental variable for two-stage least squares regression to correct endogeneity bias. Cross-situation robustness is tested by replacing the dependent variable and grouping samples by ownership type. Results show that knowledge reorganization efficiency has a positive partial mediating effect, with an indirect effect of 0.226 (p < 0.01), accounting for 34.93% of the total effect. The conclusions remain valid after endogeneity correction, providing a robust methodological framework for evaluating digital transformation effectiveness.