We decompose currency returns into permanent changes in intrinsic value and transitory movements, called respectively, intrinsic-value and expected-return shocks. We then explore these components and their interactions with institutional investor currency flows. We find that: expected-return shocks are much larger than intrinsic-value shocks; returns overreact to intrinsic-value shocks; expected-return shocks are reliably related to flows whereas intrinsic-value shocks are not; and that intrinsic-return shocks are, as theory would predict, positively related to forecasts of cumulated innovations of interest differentials.