Most empirical studies on Corporate Social Responsibility (CSR) use cross-sectional data or case studies, making causality hard to establish. We overcome this limitation by using panel data on Chinese firms. We find no effect of last year's profits on CSR ratings, although their negative contemporaneous relation suggests a trade-off. Managerial shareholdings reduce CSR ratings while rising wages and employment are the main drivers of increasing CSR ratings. This suggests the CSR agenda aligns with the interests of labour, but not capital. However, the positive effect of Tobin's Q may indicate CSR is associated with intangibles of value to a firm.