On Economic Complexity and the Fitness of Nations

Abstract

Complex economic systems can often be described by a network, with nodes representing economic entities and edges their interdependencies, while network centrality is often a good indicator of importance. Recent publications have implemented a nonlinear iterative Fitness-Complexity (FC) algorithm to measure centrality in a bipartite trade network, which aims to represent the ‘Fitness’ of national economies as well as the ‘Complexity’ of the products being traded. In this paper, we discuss this methodological approach and conclude that further work is needed to identify stable and reliable measures of fitness and complexity. We provide theoretical and numerical evidence for the intrinsic instability in the nonlinear definition of the FC algorithm. We perform an in-depth evaluation of the algorithm’s rankings in two real world networks at the country level: the global trade network, and the patent network in different technological domains. In both networks, we find evidence of the instabilities predicted theoretically, and show that ‘complex’ products or patents tend often to be those that countries rarely produce, rather than those that are intrinsically more difficult to produce.

A New Interpretation of the Economic Complexity Index

“…We note that an alternative measure based on global trade data called Fitness [22] is much more strongly correlated with diversity [14]. Spearman rank correlations for Fitness and diversity tend to be between 0.94 and 0.97 [13].…”

“…The ECI has recently been described as 'standard eigenvalue centrality algorithm' [14]. Eigenvector centrality is defined as the eigenvector corresponding to the largest eigenvalue of an adjacency matrix and this definition is standard for symmetric matrices that represent undirected networks.…”

“…1) [8,4]. Second, the ECI has been compared to an eigenvector centrality measure [14]. This paper stresses that neither of these descriptions is accurate.…”

Better to Stay Apart: Asset Commonality, Bipartite Network Centrality, and Investment Strategies

“…Furthermore, recent publications have also proposed nonlinear versions of the algorithm to measure centrality in bipartite networks; the interested reader can refer e.g. to: Tacchella et al (2012Tacchella et al ( , 2013; Morrison et al (2017); Alshamsi et al (2018), among others. observables:…”

Maximum-Entropy Tools for Economic Fitness and Complexity

“…To conclude the comparison with the original results, we show that we have reproduced the global correlation that was found between the fitness of each country and its GDP per capita in Figure 2. We will not go into the supposed significance of this correlation, which in the first place was disputed in [5] and could be further questioned after a comparison with similar plots using higher z-score filters, as we will show in the following section. Figure 2.…”