Summary: Predictions on the same economic scale can be performed by different actors and through different methods, depending on the election predicting the final degree of accuracy that it would lead partner. Given that each method used and every official involved can capture different aspects of the information, it seems appropriate to undertake a combination of predictions. As a result of these considerations, this report explores the possibilities offered by information measures in the context of the economic forecast, distinguishing assessment and the combination of predictions. In regard to the evaluation, the initial proposals based on information measures are Theil (
1966), who develops measures based on the uncertainty of Shannon (1948). In line with this approach, this report examines new alternatives that include both indicators based on the uncertainty (a measure of vagueness quadratic) and others built from the concerns (quadratic information associated with the predictions based on quadratic and imprecise relative errors) . The empirical analysis of the behavior of these indicators and their comparison with the usual measures allows draw some conclusions of interest. As for the combination of predictions, the pioneering work relate to Bates and Granger (1969), who proposed a prediction techniques for synthesis from linear combinations of individual predictions, whose weights are derived from the corresponding variance. In this study we investigated the potential of Information Theory to develop techniques combination whose weights calibrated unevenly predictions singles. Specifically we employ the principle of Maximizing Entropy for measures of uncertainty Hannon, quadratic and double quadratic, and use the empirical evidence available to compare the weights ladies and predictions combined with the average artimética. In addition, there is a combination through regression based on quadratic errors, where the function is a concern to minimize double quadratic, and compares the forecast obtained under this approach combined with the ordinary least squares. Since it is also possible to develop predictions based on the combination of subjective information, the report analyzes different methods of quantifying expectations, comparing their ability predictive assessment measures in both traditional and based on information measures. It also discusses some indicators that summarize the expectations on various economic variables, and examine its ability to anticipate the evolution of the economic cycle.
