Can Agents with Causal Misperceptions be Systematically Fooled?

The conventional rational-expectations postulate rules out the possibility that agents will form systematically biased forecasts of economic variables. I revisit this question under the assumption that agents' expectations are based on a misperceived causal model. Specifically, I analyze a model in which an agent forms forecasts of economic variables after observing a signal. His forecasts are based on fitting a subjective causal model - formalized as a direct acyclic graph, following the "Bayesian networks" literature - to objective long-run data. I show that the agent's forecasts are never systematically biased if and only if his graph is perfect - equivalently, if the direction of the causal links he postulates has no empirical content. I demonstrate the relevance of this result for economic applications - mainly a stylized "monetary policy" example in which the inflation-output relation obeys an expectations-augmented Phillips curve.