Finding needles in haystacks: a machine learning approach for the drivers of green innovation

Abstract

Motivated by theoretical insights from the resource-based view, dynamic capabilities, and social network theories, we examine how internal capabilities, prior innovation experience, and collaborative ties jointly shape firms’ green innovation. Hence, we study the drivers of green innovation using firm-level panel data from Spain (2003–2016), leveraging lasso and double machine learning (DML) methods. Our findings highlight that internal R&D expenditures, firm age, external R&D partnerships, and lagged product and process innovations are robust and important predictors of green innovation. We provide new causal evidence of the path-dependent nature of green innovation, with prior innovations exerting persistent treatment effects across multiple periods. The mediation analysis further reveals that collaborative R&D serves as a critical channel through which innovation capabilities are mobilized. These results underscore the complementarity between internal resources and external knowledge access, which enables firms to reconfigure their capabilities in response to environmental imperatives. This evidence has implications for innovation policy design and suggests that targeted support for R&D investment and collaboration can enhance firms’ adaptive capacities for green innovation.