Incorporating uncertainty when predecting the future of the economy
UC3M/DICYT A research project the Universidad Carlos III of Madrid (UC3M) formulates new ways of anticipating changes in the economy that incorporate uncertainty. Using these methods it is possible to be more realistic with regard to macroeconomic predictions.
Governments take into account the macroeconomic predictions made by economists when they make important decisions that affect all of us. For this reason, experts strive to find new ways of fine tuning their previsions in an attempt to produce the minimum possible number of deviations in them.
There are different methods that aid in predicting what will happen in the future. Nevertheless, these predictions do not usually take into account uncertainty, that is, “the lack of knowledge we have about what is going to happen, and which we cannot predict with the information that is available when we are making the prediction,” explains one of the authors of the study, Esther Ruiz, a tenured professor in UC3M’s Statistics Department. In practice, predictions to not have to be exact, and “intoducing uncertainty makes it possible to make more realistic forecasts,” she adds.
The new methods proposed by the researchers are based on models in which, as often occurs in practice, the observation of the variable that is subjected to prediction includes some error. Such methods take into account not only the error that is attributable to the non-observation of the variable, but also to the uncertainty that comes from the estimation of the model. “When we take into account the fact that the parameters have been estimated, so we do not know their values for certain, the prediction intervals for future values of the variable are wider, so they can change our perception regarding what we can expect in the future,” states Esther Ruiz.
In the article the authors have published together with researchers from the Universidad de Concepción (Chile), in the journal Computational Statistics and Data Analysis, it is also notable that the proposed prediction methods have the further advantage of being simple from a computational point of view, while their results are an improvement over those that can be obtained by means of alternative techniques. The researchers, who point out the need to bear in mind the concept of uncertainty when making predictions about the economy using statistical methods, point out that “there are important differences between predicting macroeconomic variables and financial variables since, with the latter, uncertainty plays a central role due to the changes that can be observed in the volatility of financial returns.”
The new techniques that have been proposed allow the margins for error associated with some key variables, such as the rate of unemployment, inflation and production, to be changed over time, so that different intervals for future values can be obtained in function of the present uncertainty at any given moment.
Alejandro Rodríguez, Esther Ruiz. Bootstrap prediction mean squared errors of unobserved states based on the Kalman filter with estimated parameters. Computational Statistics and Data Analysis (2012). doi:10.1016/j.csda.2011.07.010