Bank substitutability and financial network resilience: insights from the first globalization
This work, co-authored with Olivier Accominotti (London School of Economics and CEPR) and Stefano Ugolini (Sciences Po Toulouse), is from Chapter 5 of Delio Lucena-Piquero’s PhD dissertation: “Beyond the Dyadic Approach in Social Network Analysis: Applications to Innovation Studies and Financial Economics” (2020). The London School of Economics has kindly provided the financial support for the collection of the data used in this work.
Résumé : The recent literature on financial network resilience has paid relatively less attention to the dimension of substitutability than to interconnectedness. In this paper, we apply a simple technique to simulate the upper-bound effects of bank defaults on firms’ access to credit at the global level during the first globalization (1880-1914). We find that, in stark contrast to today’s financial networks, in the early 20th century the global network displayed considerable resilience to shocks, as the level of substitutability of all banks was relatively high. This finding has implications for regulators, as it shows that a financial network not featuring highly-systemic banks can (and did) actually exist.
Since the 2008 crisis, academics and policymakers have been deeply concerned with the question of the identification of systemic actors in financial systems. Traditionally, regulators had been mostly identifying systemic intermediaries according to their size (the “too-big-to-fail approach”), but the catastrophic effects generated by the fall of Lehman Brothers (a relatively small bank) brought to light the multidimensional nature of systemicness. In 2013, the Basel Committee on Banking Supervision (BCBS) and the Financial Stability Board (FSB) jointly published new guidelines for the assessment of systemicness, based on five different dimensions of the concept: size, interconnectedness, substitutability, complexity, and cross-jurisdictional activity (BCBS, 2013)1. Among these five dimensions, interconnectedness is the one that has attracted larger academic attention in the last decade or so: many scholars have been applying network analysis and simulation techniques in order to identify “too-interconnected-to-fail” actors, thus reaching a number of important conclusions about the structural properties and resilience of modern financial networks. By contrast, substitutability has attracted relatively less attention to date. Assessing substitutability has however proved particularly problematic (Benoit et al., 2019), leading regulators to revise their guidelines under this respect (BCBS, 2018).
In this paper, we use simple network analysis and simulation techniques in order to assess more specifically the relationship between actors’ substitutability and financial network resilience. Our approach is straightforward: we provide an upper-bound evaluation of network disruption due to financial shocks by looking at how many actors remain isolated once intermediaries are removed. This is an upper-bound estimation as it rests on the (very strong) assumption that no other financial relationship can exist except those that are actually observed – meaning that an agent will lose market access if the intermediaries to which she is connected do default. Applying this methodology to contemporary financial networks yields very catastrophic results: networks break down as central nodes are removed, thus pointing to the high degree of unsubstitutability of a few actors (see e.g. Pröpper et al., 2008). This result might be taken as a confirmation of the conclusion (a general one for studies based on contemporary data) that financial networks inevitably feature a few highly systemic actors. However, in this paper we apply this methodology to an historical financial network and we reach very different conclusions: in our case study, node removal never generates sizable damage in the network, thus pointing to low levels of systemicness for all actors. Note that our case study does not consist of an idiosyncratic, peripheral financial market from an obscure historical epoch, but no less than the global money market at the heyday of the first globalization (1880-1914), when the international economy reached levels of interconnectedness comparable to those of the late 20th century (O’Rourke and Williamson, 2002). We are therefore able to provide, for the first time, evidence that a financial network featuring can exist and did actually exist globally at a time of high international integration. In our view, this finding yields potentially important implications for regulators.
The rest of the paper is organized as follows. Section 2 reviews the recent literature applying network analysis and simulation techniques, and highlights the originality of our paper. Section 3 describes our data and details our empirical strategy. Section 4 presents our results. Section 5 concludes and offers some speculations for regulators.
Pour accéder au texte intégral, cliquer ici.
- Size is defined as the total size of the bank’s liabilities. Interconnectedness consists of the network of contractual obligations that characterize the bank’s activities. Substitutability (sometimes referred to as “financial institution infrastructure”) is defined as the bank’s importance as a provider of client services. Complexity consists of the business, structural, and operational complexity of the bank (i.e., its involvement in sophisticated activities such as derivatives or other off-balance-sheet exposures). Cross-jurisdictional activity is the geographical dispersion of the bank’s activities (BCBS, 2013). [↩]