We present fresh evidence on the validity of the dividend signaling hypothesis (DSH), by using a new testing approach. Using a simple dividend signaling model, we derive three empirically identifiable drivers of the marginal net benefit of signaling: cash flow predictability, market-to-book ratio, and past equity returns. Our empirical tests support the DSH. There is a significantly greater association between current dividend changes and future earnings performance for firms with low cash flow predictability, low market-to-book ratio, and low past equity returns. We also present evidence that the marginal signaling benefits at the firm-level are time-varying, increasing (decreasing) in booms (recessions) and in periods of high (low) aggregate stock market performance.