"I have argued that forecasts will be based on more information than is contained in econometric models and in general on information differing from agent to agent. I also want to argue that they will not necessarily use all the information contained in an econometric model. In fact, the two propositions are intimately linked though they seem to move in opposite directions. We have to assume that information-processing ability is scarce. As I have already said, this is one of the main justifications for and explanations of a decentralized economy. But then it follows that an individual concentrates on acquiring the information most useful to him and will have to crowd out the information which is less useful. In particular, information that is broadly pertinent to the economy as a whole may have very little predictive power for the future of an individual."