Short-Term Forecasting of GDP via Evolutionary Algorithms Using Survey-Based Expectations

Business and consumer surveys are the main source of agents' expectations. In this study we use survey expectations about a wide range of economic variables to forecast GDP growth. We propose an empirical approach to derive mathematical functional forms that link survey-based expectations to present and future economic growth. Combining symbolic regression with genetic programming we generate two indicators: a leading one, using the expectations about the future, and a coincident one with the perceptions about the present. Our examination of the forecast accuracy of both indicators to track the evolution of economic activity in fourteen European countries indicates that the coincident indicator always outperforms the leading indicator. In order to find the optimal combination of both indicators that best replicates the evolution of GDP growth in each country, we use a portfolio management procedure known as index tracking. By means of a generalized reduced gradient algorithm we derive the relative weights of both indicators. We find significant differences across countries, which suggests that survey-based indicators should not equally weight the expectations about the future and the perceptions about the present. In most economies, the survey-based predictions generated with the composite indicator outperform a baseline model for one-quarter ahead forecasts. These results highlight the usefulness of survey data for the estimation of the evolution of economic activity