Title:Forecasting countries' gross domestic product from patent data

Abstract:Recent strides in economic complexity have shown that the future economic development of nations can be predicted with a single "economic fitness" variable, which captures countries' competitiveness in international trade. The predictions by this low-dimensional approach could match or even outperform predictions based on much more sophisticated methods, such as those by the International Monetary Fund (IMF). However, all prior works in economic complexity aimed to quantify countries' fitness from World Trade export data, without considering the possibility to infer countries' potential for growth from alternative sources of data. Here, motivated by the long-standing relationship between technological development and economic growth, we aim to forecast countries' growth from patent data. Specifically, we construct a citation network between countries from the European Patent Office (EPO) dataset. Initial results suggest that the H-index centrality in this network is a potential candidate to gauge national economic performance. To validate this conjecture, we construct a two-dimensional plane defined by the H-index and GDP per capita, and use a forecasting method based on dynamical systems to test the predicting accuracy of the H-index. We find that the predictions based on the H-index-GDP plane outperform the predictions by IMF by approximately 35%, and they marginally outperform those by the economic fitness extracted from trade data. Our results could inspire further attempts to identify predictors of national growth from different sources of data related to scientific and technological innovation.