© 2017 Elsevier Inc. We introduce a new market microstructure model to study a setting in which an authority (e.g. a firm manager or government policymaker) learns about the likelihood of a bad state by observing activity in the asset market, before deciding whether to undertake a costly intervention to improve the state. Intervention erodes the value of an investor's private information by weakening the link between the initial state and the asset payoff. Informed investors are reluctant to make large, informative trades in the bad state, undermining the market's informativeness and the welfare gains generated by the possibility of a corrective intervention. Fundamentally, the authority faces a tradeoff between eliciting information from the asset market and using the information so obtained. The authority can generate a Pareto improvement if she commits to intervene less often when the market suggests that intervention is most beneficial and more often when the market suggests that intervention is unwarranted. She thus may benefit from imperfections in the intervention process or from delegating the decision to intervene to a biased agent.