The contributions of this paper are two‐fold, one methodological and the other substantive. First, we propose a novel way of estimating volatility impulse response functions (VIRF) employing the Markov chain Monte Carlo algorithm. As a useful byproduct, the associated confidence bands can be constructed. Second, we analyze volatility spillovers between the on‐ and offshore Renminbi exchange rates toward the US dollar. The VIRF show that the satellite CNH exchange rate promptly reflects the global market demand and supply, while the main CNY exchange rate reacts with a time lag.