La Sapienza - Università di Roma

logo de La Sapienza per la stampa

Economic Forecasting and Complex Systems

An article published on-line on Nature Physics addresses a new method to forecast GDP growth that integrates and supports the forecasts made by the International Monetary Fund. The new method envisages economic growth as a physical system, whose dynamic processes can be forecast by analysing product export data via techniques derived from the Physics of Complex Systems. 
Models based on a country’s economy and developed in this way are extremely complicated and difficult. Moreover, although economists have access to a growing trove of data, producing reliable and replicable results is anything but simple. However, Physics of Complexity provides techniques to model such unfathomable systems in terms of individual components, such as the diffusion of epidemics or the movements of traffic.

Researchers at the Sapienza Department of Physics have developed, in collaboration with Andrea Tacchella from the National Research Council, a method for GDP forecasting based on the idea that a complex system with unknown microscopic dynamics may be forecast by observing the temporal evolution of historical data, a relevant and analogous system. These methods are reliable only if the system employed has a minimal number of dimensions, corresponding to the addressed variables. Indeed, the uncontrolled addition of further data will produce less reliable results.

The authors of the method have demonstrated that it is possible to make a competitive forecast of the GDP by using a two-dimensional model with the pro capita GDP of a country and its fitness, which characterises a country’s competitiveness and can be reconstructed by inserting exports data in relevant mathematical algorithms.

Comparing the forecasts obtained through this method with the forecasts made by using data published by the International Monetary Fund, the authors have demonstrated that, on average, their forecasts are 25% more accurate. Moreover, the fact that the errors in the different models are not correlated points to an increased potential for joint system forecasts. 


Dynamical System Approach to Gross Domestic Product Forecasting - A. Tacchella; D. Mazzilli; L. Pietronero - Nature Physics - Volume 14, pp. 861865 (2018). DOI: 10.1038/s41567-018-0204-y   

For further information

Luciano Pietronero
Department of Physics, Sapienza University

Andrea Tacchella
CNR, Rome