This paper develops a two-step inference procedure to test for a local one-for-one relation of contemporaneous jumps in high-frequency financial data corrupted by market microstructure noise. The first step develops a new bivariate Lee-Mykland jump test for pre-averaged, intra-day returns. If a jump is detected in at least one of the two assets, then the second step tests for equal jump sizes. We apply the test procedure to pairs of nominal and inflationindexed government bond yields at monetary policy announcements in the U.S., U.K., and Euro Area. The analysis provides new high-frequency evidence about the anchoring of inflation expectations and central banks’ ability to push a measure of inflation expectations towards their inflation target.