Monitoring supply networks from mobile phone data for estimating the systemic risk of an economy

Abstract

Remarkably little is known about the structure, formation, and dynamics of supply- and production networks that form one foundation of society. Neither the resilience of these networks is known, nor do we have ways to systematically monitor their ongoing change. Systemic risk contributions of individual companies were hitherto not quantifiable since data on supply networks on the firm-level do not exist with the exception of a very few countries. Here we use telecommunication meta data to reconstruct nationwide firm-level supply networks in almost real-time. We find the probability of observing a supply-link, given the existence of a strong communication-link between two companies, to be about 90%. The so reconstructed supply networks allow us to reliably quantify the systemic risk of individual companies and thus obtain an estimate for a country’s economic resilience. We identify about 65 companies, from a broad range of company sizes and from 22 different industry sectors, that could potentially cause massive damages. The method can be used for objectively monitoring change in production processes which might become essential during the green transition.

Revealing production networks from firm growth dynamics

“…If we consider each vector g (t) as a drawn from a joint probability distribution where the correlations are driven by the supply chain, Gaussian graphical models seem well equipped to reconstruct the production network if one ignore the fact that the growth rates do not have a Gaussian distribution. 13 We think nonetheless that, because the growth-rates show a Gaussian-like central region, as shown by [31], it is reasonable to use this model to attempt a reconstruction.…”

“…To tackle this problem, recent efforts have attempted to reconstruct the production network, inferring the topology of the network using only partial, aggregate or related data. For instance, [13] uses mobile phone data to reconstruct the supply chain network of an undisclosed European country, while [14] and [15] pioneered machine learning for link prediction in supply chains, leveraging topological features computed by hand or distilled automatically through Graph Neural Networks. A similar approach was used in [16] to predict links between firms using their financial, industrial, and geographical features.…”

Reconstructing supply networks

“…This approach was also used in [37] for an undisclosed European country. 18 A different application of the maximum-entropy principle, where constraints are imposed softly (see section 3.1), results in the solution used in [66] to reconstruct Ecuador's national production network and in [19] to reconstruct the transaction network between customers of two Dutch banks.…”

“…Zhang et al [34] and Wichmann et al [17] have provided a proof of concept that mining news and websites for supplier-buyer relations can be automated, and we have already mentioned that websites can be an important source of key metadata for link prediction (especially product-related information). While phone data is likely to be difficult to access, it is worth remembering the impressive result in [37] that firms with average daily communication of more than 30s/day have a 70% probability of being connected.…”

“…In [37], instead, the authors use phone calls between companies and survey data to track down supplier-customer relationships in an undisclosed European country. The survey asked companies to list their ten most important suppliers and customers.…”