The Economic Complexity Index (ECI) introduced by Hidalgo and Hausmann (2009) has been successful in explaining differences in GDP/capita and economic growth across countries. There has been confusion, however, about what it means and why it works. The ECI was originally motivated as an enhancement of diversity which is defined as the number of products a country is competitive in. However, the ECI is orthogonal to diversity. Nor is the ECI an eigenvalue centrality measure - in fact, the standard eigenvalue centrality measure applied to the export similarity matrix is equivalent to diversity. Instead we show that the ECI can be understood in terms of spectral clustering. It corresponds to an approximate solution to the problem of partitioning a graph into two parts in order to minimize the connections between the parts. It can also be viewed as an optimal one-dimensional ordering that clusters countries with similar exports together and minimizes the distance between countries. We present two empirical examples that involve regional employment in occupations and industries rather than exports, in which diversity fails to be a distinguishing feature of the data. These particular regional settings illustrate how the ECI can be useful even when diversity is not.
↧