This paper comprehensively investigates the role of realized jumps detected from high frequency data in predicting future volatility from both statistical and economic perspectives. Using seven major jump tests, we show that separating jumps from diffusion improves volatility forecasting both...