Technology Investment and Firm Performance in China’s Sporting Goods Industry: An Exploratory Panel Study Across the Industrial Chain

This study examines whether technology investment is efficiently transformed into firm performance in China’s sporting goods industry, a sector whose development cycle is short and whose disclosure practices are still maturing. Drawing on the resource-based view and dynamic capability theory, and following the common concern of the reviewers about sample coverage, we assemble an unbalanced panel of 44 Chinese listed sports firms covering the full industrial chain from 2016 to 2025 (389 firm–year observations), with sporting goods manufacturing as the principal component (31 of 44 firms); R&D intensity is observable for 35 firms (266 observations). The analysis is exploratory: our aim is to establish conditionally robust facts rather than to confirm a structural model. Using two-way fixed-effects estimators with firm-clustered inference, wild cluster bootstrap resampling, Romano–Wolf multiple-hypothesis correction, an Oster-type selection diagnostic, and a bootstrap mediation test, we obtain three main results. First, once operating revenue is controlled, current-period R&D intensity has no significant effect on ROA (β = −0.214, p = 0.554); the significantly negative estimate obtained without a revenue control reflects the mechanical scaling of R&D intensity by operating revenue, and adding back R&D expenditure expensed in ROA turns the coefficient positive and insignificant (β = 0.239, p = 0.507). Second, the intangible asset ratio does not mediate the investment–performance relationship (indirect effect a·b = 0.058, 95% CI [−0.081, 0.572]), consistent with a reporting gap between R&D input and observable capability accumulation. Third, the point estimates of the module contrasts run opposite to the hypothesized direction, and no contrast survives the Romano–Wolf correction (family-wise adjusted p = 0.846), so sub-module heterogeneity is reported as an exploratory observation rather than an established regularity. An Oster diagnostic (δ = 0.54) indicates that moderate selection on unobservables could account for the restricted estimate. The evidence points to transformation efficiency, rather than input scale, as the binding constraint, and we outline corresponding policy measures for building an intelligent sports industry.