With the aim of meeting the abovementioned research objectives, the researcher made a review of the literature on financial contagion, suggesting various models to estimate system wide losses due to the direct and indirect balance sheet interlinks of banks. Ms Saroyan started a comparative review of existing models organized in two groups 1) “default contagion models” are models in which the default of a financial firm can induce the default of other financial firms through default on obligations (e.g. Eisenberg and Noe 2001; Rogers and Veraart 2013; Greenwood, Landier, and Thesmar 2012, etc.); 2) “distress contagion models” are models in which not only default but also distress matter, including in particular the DebtRank model introduced by the hosting team and its further extensions (e.g. Battiston et al. 2012; Bardoscia et al. 2015; Battiston et al. 2016a,b; Bardoscia et al. 2016). One of the goals of this review was to adapt existing distress contagion and stress-test frameworks to the FINREALNETS’ research goal, i.e. the identification of Systemically Important Real Sectors for financial stability purposes. The stress-test framework developed by the hosting team mainly focuses on the contagion through direct, interbank interlinkages. During this phase it was explored the idea to develop an extension of the DebtRank-based stress-test model where distress would propagate via common exposures of banks to common assets or common asset classes in terms of sector and/or country. Moreover, it emerged that developing an appropriate definition of portfolio overlap is key and that this should build on the notion of leverage matrix (Battiston 2016a). Therefore, the researcher simultaneously explored the existing literature seeking to estimate the overlapping portfolio-based systemic importance of banks (Greenwood, Landier, and Thesmar 2012; Caccioli et al. 2012, 2015; Banwo et al. 2016; Di Gangi, Lillo, and Pirino 2015; Cont and Wagalath 2014, etc.).
There is both empirical and theoretical evidence that active risk management of banks, aiming to keep a constant probability of default, makes banks’ book value leverage procyclical and may induce the system into overvaluation or devaluation loops, depending on the sign of the initial shock (Adrian and Shin 2013, 2009). These shock amplifying effects of leverage procyclicality are in line with FINREALNETS’ macro-prudential policy objectives, because they are more or less pronounced depending on bank's promptness to comply with capital adequacy rules (Tasca and Battiston 2012). Moreover, those effects are subject to the market liquidity of revalued assets. Although these fire-sales-related feedback effects were present in the internal multi-round models (Battiston et al. 2016b), little attention has been paid to the post-shock credit-portfolio management and optimisation strategies of banks conditional to the existing risk-based regulatory framework (Koehn and Santomero 1980; Nielsen 2015; Crouhy, Galai, and Mark 2000). A better understanding of the assumptions on individual banks’ portfolio diversification strategies are key for estimating market procyclicality effects in network-based stress-test models(Tasca and Battiston 2016). Therefore, in the FINREALNETS framework the fellow started to investigate the issue of portfolio diversification vs overlap, which had not been fully explored yet in hosting team.
One of the researcher’s objectives under FINREALNETS was to acquire sufficient skills in stress-test modelling and algorithm writing. Therefore, she studied the available Matlab code written for the paper Battiston et al. (2016b). The existing Matlab code focuses on the contagion via interbank leverage network with different distress propagation mechanisms (DebtRank, Eisenberg&Noe and Default Cascades) providing different outputs in terms of systemic vulnerability.
