Towards a statistical mechanics of economies

Abstract

The relevance of a given good in the production network of an economy can be quantified by its degree. We find that, across a large dataset of IO matrices for different countries, degrees follow an exponential distribution when the level of aggregation of economic activity is coarse enough. We confirm this by studying the US economy, for which data at a finer classification are available, at different level of aggregation. We recover the same “universal” degree distribution in models of large random economies, upon aggregation. The convergence to the exponential distribution is faster when aggregation is performed disregarding the nature of economic activities. This suggests that the loss of information on microscopic details depends on the classification methods employed by government agencies.