Upstreamness and downstreamness in input-output analysis from local and aggregate information

Abstract: Ranking sectors and countries within global value chains is of paramount importance to estimate risks and forecast growth in large economies. However, this task is often non-trivial due to the lack of complete and accurate information on the flows of money and goods between sectors and countries, which are encoded in Input-Output (I-O) tables. In this work, we show that an accurate estimation of the role played by sectors and countries in supply chain networks can be achieved without full knowledge of the I-O tables, but only relying on local and aggregate information, e.g., the total intermediate demand per sector. Our method, based on a rank-$1$ approximation to the I-O table, shows consistently good performance in reconstructing rankings (i.e., upstreamness and downstreamness measures for countries and sectors) when tested on empirical data from the World Input-Output Database. Moreover, we connect the accuracy of our approximate framework with the spectral properties of the I-O tables, which ordinarily exhibit relatively large spectral gaps. Our approach provides a fast and analytical tractable framework to rank constituents of a complex economy without the need of matrix inversions and the knowledge of finer intersectorial details.