Climate Policy Stringency, Corporate Energy Efficiency and Sustainable Business Performance: A Synthetic Global Panel Simulation

Abstract

This study develops and demonstrates an integrated analytics framework to examine how climate‐policy stringency could shape corporate sustainability performance through corporate energy behaviour, while accounting for digitalisation as a contextual amplifier. Rather than drawing inferences from observed firm‐level data, the study uses a transparent policy‐calibrated synthetic panel simulation design covering 20 countries, 30 firms per country and 8 years (2017–2024), yielding 4800 firm‐year observations across five sectors. The simulation design is calibrated to plausible ranges and stylised relationships reported in recent climate–energy–ESG research and serves as a reproducible testbed for methodological integration. The simulation workflow combines descriptive diagnostics, clustered‐error panel regressions, a random‐forest prediction model with out‐of‐sample evaluation, a simulation‐based difference‐in‐differences policy‐shock exercise with event‐style diagnostics and mediation decomposition with bootstrap uncertainty. The results show that the framework recovers the directional relationships embedded in the simulation design: Higher policy stringency and renewable‐energy adoption are associated with higher ESG scores, whereas higher energy intensity is associated with lower ESG performance. Digitalisation is positively associated with the policy–energy–ESG linkage within the simulated environment. Feature‐importance rankings indicate that the policy index remains the dominant predictor, followed by financial valuation (Tobin's Q ) and energy intensity. In the simulated policy‐shock exercise, treated countries exhibit a modest decline in energy intensity and a small increase in ESG relative to controls after 2020. These estimated effects are directionally consistent with the programmed treatment structure but remain statistically imprecise, illustrating estimator behaviour within the synthetic simulation rather than providing evidence of real‐world policy impacts. Mediation estimates further suggest that approximately 15% of the simulated policy association operates through the energy‐intensity channel under the assumed structural ordering. The study contributes a transparent and reproducible workflow for integrating regression, machine learning, DiD diagnostics and mediation analysis in synthetic sustainability‐policy research.