Increasing the predictive power of convidence indicators using selection methods.

Every member state of the European Union carries out business and consumer surveys where individuals are asked about their expectations on future values of certain economic variables. The results of these surveys are then summerized as an indicator. Previous research did showed that the predictive power of these indicators is rather weak. In this project we investigate how the predictive power can be increased by selection methods. The indicators are composed on the bases of many survey questions, asked to an heterogeneous group of respondents. From this high-dimensional data set one wishes to keep the informative part, and filter out the noise. Statistical techniques will be developed to obtain a selection of the most pertinent components of the survey, in order to increase the predictive power.

Keywords:Business and consumer surveys, Prediction, Time series analysis, Econometrics

Disciplines:Applied economics, Economic history, Macroeconomics and monetary economics, Microeconomics, Tourism