This paper proposes a Bayesian approach to incorporating specification and identification uncertainty into a VAR analysis of the dynamic effects of money supply shocks on the macroeconomy. The approach follows Poirier (1991) in averaging over discrete model specifications in forming posterior densities of the dynamic responses to such shocks. Two distinct means of identifying money supply shocks are considered here: one that imposes contemporaneous restrictions, and one that imposes long-run monetary neutrality. I estimate bounds on dynamic responses that account for specification and parameter uncertainty, and find strong evidence of short-run real effects of money on the economy, including a liquidity effect for both long-term and short-term interest rates. Furthermore, some results differ substantially across identifying restrictions if model uncertainty is ignored.