Title of Mazzilli’s talk: “Modeling and forecasting countries’ growth: new methods and data“
Predicting countries’ growth is an extremely challenging and important part of macroeconomics. Standard models for GDP forecasting are complex and involve the combination of many different variables often gathered from different institution. The paradigm “more data more information” in the era of Big Data has to be carefully considered, in fact it is not always true. It is indeed unclear how pollution, transport, population etc, should be combined in a forecasting model and a high dimensional
space often leads to results hard to understand. We propose a different approach. Using only export data and simple statistical mechanics’ tools, we construct a forecasting model well grounded, easy to understand and able to compete with the International Monetary Fund forecasting. The advantages of using only export data are different. First, there is no need of complex combination involving regressions or other fits. Second, starting with raw data, taken from COMTRADE, we can enhance its quality and control noise. Third, our model is easy to understand and to reproduce. Using export and a Hidden Markov Model description of productive systems, we can estimate the state of competitiveness of each country in each commodity. Using such states we can evaluate the Fitness of countries, i.e. their potential of growth. The Fitness, combined with the GDP per capita, shows a non trivial dynamics in the corresponding bidimensional plane. Studying this dynamics, we can predict countries’ growth for a 5 years horizon
competing with IMF in terms of accuracy.