Three problems pose severe challenges to identify the impact of corporate social responsibility (CSR) on firm value and profitability. These are construct validity, limited data, and endogeneity. To deal with them we use a broad composite measure of CSR and panel data with firm fixed and random effects, plus extensive covariates. We analyze a global sample drawn from 35 countries over 2003–12 and find an economically significant relationship between the overall CSR measure and firm value, but little impact of CSR and profitability. The results are driven by the social subscore. We show that both omitted firm characteristics and omitted aspects of CSR can lead to omitted variable bias and in studies that focus on a single aspect of CSR the omitted variables bias can be severe.