Event studies show that the Federal Reserve’s announcements of forward guidance and large-scale asset purchases had large and desired effects on asset prices but these studies do not tell us how long such effects last. Wright (2012) used a structural vector autoregression (SVAR) to argue that unconventional policies have very transient effects on bond yields, with half-lives of 3 to 6 months. The present paper shows, however, that this inference is unsupported for several reasons. First, accounting for model uncertainty greatly lengthens the estimated persistence. Second, and more seriously, the inference is unreliable because the SVAR is structurally unstable and forecasts very poorly. Finally, the implied in-sample return predictability from the SVAR greatly exceeds a level consistent with rational asset pricing and reasonable risk aversion. Restricted models that respect more plausible asset return predictability are more stable and imply that unconventional monetary policy shocks were fairly persistent. Estimates of the dynamic effects of shocks should respect the limited predictability in asset prices.