Are there any possible situations in which the state of the economy tomorrow depends on that of the economy today revealed by the government? If so, does the government have any ~incentives" to manipulate statistics? Using a simulation approach based on a model of evolutionary cellular automata, this paper tackles the issue by taking explicitly into account self- fulfilling expectations and the existence of multiple equilibria. We find that the government will not always lie, es- pecially when agents use the Bayesian learning algorithm to adjust their reliance on government statistics. Nevertheless, there is an incentive for the government to lie under certain circumstances, that is, when the economy, in terms of our model, is in a cloudy zone or the scale of the pessimistic shock is moderate.
Proceedings of the Fisrt Asia-Pacific Conference on Simulated Evolution and Learning (SEAL'96)