We hypothesize and present strong evidence that dividend increases (decreases) result in a general decrease (increase) in the opportunity cost of equity capital (Ke), measured by the discount rate implicit in analysts' forecasts. Estimates of Ke obtained from analyst forecast data likely capture priced information risk that is not reflected in cost of equity capital estimates customarily obtained from empirical excess returns data. In the presence of a full menu of control variables, our measured changes in the cost of equity capital are shown to provide high explanatory power for the market reaction to dividend change announcements. We also hypothesize and demonstrate that the impact of dividend changes on the cost of equity is conditional on how preannouncement Ke relates to preannouncement return on equity (ROE). Specifically, dividend increases result in a reduction in the cost of equity capital only when currently experienced ROE < Ke. This is consistent with shareholders preferring earnings to be reinvested by managers to earn a higher rate than their opportunity rate. When ROE > Ke, on the other hand, the cost of equity capital actually increases. For dividend decreases, the cost of equity capital increases only when ROE > Ke, consistent with firms currently experiencing positive economic income using dividend cuts to signal anticipated permanent earnings declines. Together with extensive robustness tests, our results indicate that dividend changes significantly affect shareholder value, contrary to the longstanding dividend irrelevance argument.