The question of what drives corporate social performance (CSP) has become a vital concern for many managers and researchers of large corporations. This study addresses this question by adopting a multilevel, multistakeholder, and multimethod approach to theorize and estimate the relative influence of macro (national business system and country), meso (industry), and micro (firm-level) factors on CSP. Applying three different methods of variance decomposition analysis to an international sample of 2,060 large public companies over a time span of 5 years, our results show that firm-level factors explain the largest proportion of variance in aggregate CSP as well as CSP oriented toward communities, the natural environment, and employees. These results support our hypotheses according to which CSP is not primarily driven by macrolevel or mesolevel factors, except for shareholderoriented CSP, which is relatively more influenced by country-level factors. As a whole, our findings also point to the value of subdividing CSP into its stakeholder-specific components as this disaggregation allows for a more careful examination of distinct drivers of distinct aspects of CSP.Keywords: Corporate social performance; corporate social responsibility; decomposition of variance; hierarchical linear modeling; stakeholders; variance components analysis.Abbreviations: ANOVA = analysis of variance; CSP = corporate social performance; HLM = hierarchical linear modeling; MLE = maximum likelihood estimation; NBS = national business system; REMLE = restricted maximum likelihood estimation; VCA = variance components analysis.Unpacking 2