Supporting Information S1: The heterogeneous dynamics of economic complexity

Here we discuss a number of auxiliary results supporting the main findings of the principal paper and assess the robustness of the selective predictability scheme (SPS hereafter). We also provide a discussion of the results which would be obtained by means of standard regressive approaches. In the final part, we also expand the discussion of the validation of the forecasting power of the method through the backtesting procedure proposed. 1 Fitness-Income cloud from 1995 to 2010 The deviation of the metrics from the monetary information is the key point to uncover the hidden potential of the growth of countries. The natural candidate for this study is the fitness-income scatter plot (i.e. GDP per capita-fitness plane), as shown in Figs. S1 and S2 where we report the static plot for all the available 16 years of our dataset from 1995 to 2010. The red line represents an estimation of the expected income of a country given its level of complexity. This line is not a regression in the form GDP = αF + β but it is the result of the minimization of the Euclidean distance from the line weighted by the country GDP. At this stage, this line does not represent a statement of cause-effect relationship between fitness and income. However, we will show that by looking at the dynamics in this plane we are able to develop a predictive scheme in some specific regimes