Digital Resource Disclosure Configuration and Innovation Output: Evidence from Chinese A-Share Manufacturing Firms

Abstract

Digital disclosure configuration provides little stable information about subsequent patent applications beyond a released multidimensional digital-transformation index. We study this measurement question in a public panel of 20,858 firm-years from 3,073 Chinese A-share manufacturers. Annual reports permit analysis at scale, although their disclosure frequencies do not necessarily describe resources combined in practice. The primary proxy multiplies active-category breadth by entropy normalized over four technology categories, thereby weighting richness twice. Joint fixed-effects estimation gives a standardized proxy coefficient of -0.0030 (95% confidence interval from -0.0242 to 0.0181; 20,495 observations), compared with an index coefficient of 0.0791 and an interval from 0.0377 to 0.1205. Alternative configuration definitions yield similarly inconclusive full-panel estimates. The treatment of low-richness observations matters: among firm-years disclosing at least two categories, conditional evenness has a coefficient of -0.0906, an interval from -0.1706 to -0.0107, and 2,726 observations. Because this unadjusted sensitivity changes the population, it remains exploratory and establishes neither a full-versus-restricted difference nor harmful resource orchestration. The chronological prediction comparison also requires caution. Breadth ranks descriptively above the composite, but no uncertainty comparison accompanies that ranking. Patent applications are the observed outcome throughout; strategic innovation supplies the motivation for asking whether disclosed configuration contributes information beyond the broad index. The full-panel evidence is limited, and the interpretation of balance depends on how observations with few active categories enter the analysis.