The Search for a Universal ESG–FDI Relationship: Why Development and Investment Context Matter

Abstract

ABSTRACT The empirical literature has not established a consistent relationship between environmental, social, and governance (ESG) performance and foreign direct investment (FDI). One reason may be that the association varies with countries' level of development and the structure of their investment flows. This study examines this possibility using a balanced panel of 10 upper‐middle‐income and 10 high‐income economies over the period 2002–2023. The baseline estimates are obtained from country fixed‐effects models with Driscoll–Kraay standard errors and are supplemented by lagged and dynamic specifications, pillar‐level analyses, nonlinear models, income–group interactions, and country‐level sensitivity checks. The results point to a negative and statistically significant association between ESG performance and FDI inflows in upper‐middle‐income economies, and this finding is relatively stable across alternative specifications. For high‐income economies, however, the estimates are considerably more sensitive to the composition of the sample. The ESG dimensions also exhibit distinct patterns across the two income groups, suggesting that the aggregate ESG measure masks important underlying heterogeneity. The interaction estimates provide further evidence of differences in the ESG–FDI slope across income groups, although both the magnitude and direction of this difference are sensitive to the inclusion of the Netherlands. Once the Netherlands is excluded, the estimated difference in slopes reverses, and the implied ESG coefficient for the high‐income group is no longer statistically significant. A similar sensitivity is observed in the linear and nonlinear specifications, indicating that the Netherlands plays an unusually influential role in the high‐income results, plausibly reflecting its position in international investment and financial‐routing networks. Taken as a whole, the findings suggest that the ESG–FDI relationship is better understood as context‐dependent rather than universal. They also underscore the importance of distinguishing robustness across model specifications from robustness to sample composition, particularly when aggregate FDI data include economies with substantial conduit investment activity.