CSH researcher Christian Diem will present his work (together with András Borsos; Tobias Reisch; János Kertész; Stefan Thurner) at the Networks 2021 conference.
Abstract:
The current COVID-19 crisis has made the vulnerability of international and national production networks obvious. It highlighted the strong dependence of firms’ production on their suppliers and customer and showed that functional defaults can propagate along this firm level networks. Single companies can have substantial systemic risk for production networks, i.e. the default of a single firm can cause a large part of the system to shut down their production.
However, up to now it has been unclear how large the systemic risk is that single firms pose to large firm level production networks. We address this problem empirically and calculate the systemic risk for roughly 100,000 companies in the firm level production network of Hungary first described by Borsos [2020]. To quantify the systemic risk of a firm we simulate how much of the production network is affected in response to the firm’s failure. The network dynamics are based on an edge update scheme where a loss of inputs of a firm causes a reduction of its own production and in turn causes a loss of inputs for other firms.
Our simulations for the most realistic production shock propagation scenario – taking a combination of linear and Leontief production functions and the varying criticality of inputs into account – show that only less than 100 firms have the potential to destroy more than five percent of the national production network and hence pose a thread to the overall economy. We also simulate a lower bound of firm level systemic risk by assuming only linear production functions and upper bounds by assuming only Leontief production functions.
Our findings suggest that a few single companies can pose large threats to production networks by potentially affecting around 23% of production. The knowledge of the systemic riskiness of single firms is crucial for understanding and preventing potentially large failure cascades in these networks.