The emerging market crises of the past decade have led some observers to question the wisdom of a completely open capital account. But even if capital controls are desirable, empirical evidence is needed to demonstrate when and under what circumstances restrictions are effective. This paper investigates this issue for Brazil. Much analysis of policy effectiveness has been conducted with vector autoregressions. The validity of policy inference based on this technique has been cast into doubt, so this paper employs a more "narrative" method, modeling flows to Brazil as ARMAX processes. This approach yields the interesting result that controls can be effective when the reaction of financial market participants to restrictions is taken into account, and derivatives and debt market activities curtailed.