Firm-level correlates of low-carbon innovation in Chinese agribusiness firms

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Methodology for analyzing sparse innovation outcomes and firm-level correlates using Poisson pseudo-maximum likelihood and occurrence models.

🏢実務担当者:Insights into which firm characteristics support low-carbon patenting, useful for innovation strategy in agribusiness.

🏛政策担当者:Evidence that larger, more innovative firms drive low-carbon innovation, suggesting targeted support for such firms may accelerate sectoral decarbonization.

This study examines firm-level correlates of low-carbon innovation among Chinese agribusiness-related listed firms. Using a pooled firm-year panel of 316 firms and 3,920 observations from 1998 to 2021, it treats low-carbon patenting as a sparse innovation outcome rather than as a routinely observed continuous variable: 93.1% of firm-year observations record no low-carbon patent application. The empirical design separates two descriptive margins: whether a firm files at least one low-carbon patent in a given year and how the unconditional annual count varies across the full firm-year sample. The count outcome is estimated with pooled Poisson pseudo-maximum likelihood, while the occurrence margin is examined with pooled linear probability, logit, and complementary log-log specifications. The explanatory variables are organized into resource-capability conditions, adjacent innovation stocks, and supplementary linkage proxies. Across specifications, firm size and selected balance-sheet structure variables are the most stable correlates of low-carbon patenting. Pre-period green patent stock is the strongest adjacent innovation correlate. Because green and low-carbon patent categories may overlap, the study uses exact patent-level overlap exclusion as the main measurement check, and the positive green-stock association remains. Pre-period digital patent stock is more sensitive to model design and is more visible in occurrence and onset evidence than in full-sample count models. Branch-based geographic-reach proxies and static equity-link proxies add only limited incremental explanatory value after firm characteristics and lagged innovation stocks are controlled for. The findings are descriptive rather than causal and mainly reflect between-firm differences in the pooled panel. They suggest that low-carbon patenting in this sector is concentrated among larger and more innovation-capable firms rather than being a general outcome of network expansion.

• openalex https://doi.org/10.5281/zenodo.20970722first seen 2026-07-05 04:59:24