This paper takes a step towards assessing the importance of the role of prices as signals of aggregate information, by developing an approach to estimating primitive parameters of a noisy rational expectations model under perfect competition, when both prices and terminal values are observable. We show that the equilibrium variance-covariance parameters in the Hellwig (1980) model can be inverted to obtain values of the primitive parameters. This lets us identify the MLE of the precision of private information, using only restrictions on parameter values from the underlying theory and provides, conditional on a level of risk aversion, the MLE of the variance of exogenous supply noise. Using data from a currency futures market we then present estimates of the primitive parameters, of the signal-to-noise ratio, of weights that agents place on different sources of information, and of coefficients in the linear price conjecture.

Market microstructure and the imperfect common knowledge literature in macroeconomics both analyze the effect of dispersed information on prices. This paper draws on both sources to understand exchange rate forecasting errors. A theoretical model is developed showing that forecasting errors depend on both forecast revisions as in the Woodford noisy information model and order flow as in the Evans-Lyons simultaneous trade model. This is applied to Brazilian data using a unique data set of daily consensus exchange rate forecasts managed by the Banco Central do Brasil along with order flow derived from the FX futures market. The results strongly support the theory.