4.1. The Macro Network
Following Castrén and Kavonius (2013) and Castrén and Rancan (2014), we model the EAA data, introduced in
Section 2, as a macro-network. The macro-network consists of a set of bilateral links between the main
institutional sectors which constitute the nodes of the network. The links are the EAA who-to-whom statistics
for the different financial instruments. Separate macro-networks are drawn for the different financial
instruments. The macro-network allows us to model the financial system as an intertwined set of agents that is
particularly suitable to account for shock propagation and feedback effects.13
Figure 1 shows the status quo macro-networks of two separate instrument categories,
deposits (Panel A) and
debt securities (Panel B). The directions of the links between the nodes (the sectors) show the direction of a
claim (from liabilities to assets). In the case of deposits, the households (HH), the non-financial corporates (NFC)
and the rest of the world (RoW) sectors hold deposit claims that are issued mostly by commercial banks
(monetary financing institutions, MFIs).
The network is incomplete and dominated by strong links between a
small number of sectors. By contrast, the network of debt securities is much more complete, as the issuance
and holdings of these instruments are more evenly distributed across the sectors.14
13 There is now an extensive body of literature on financial networks. In their study of bank runs, Allen and Gale (2000)
demonstrated the different contagion effects implied by complete versus incomplete network structures. Several papers
study contagion effects across financial institutions, using interbank loans as financial links (e.g., Upper and Worms, 2004;
Gai and Kapadia, 2010; Mistrulli, 2011; Glasserman and Young, 2015). Some authors have considered a broader set of
interlinkages between banks, both on the asset and the liability side, with the aim of better characterizing the way in which
financial institutions are connected to each other (Aldasoro and Alves, 2018; Poledna et al., 2015;
Bargigli et al., 2015; Caccioli
et al., 2014). Papers that investigate network structures and their properties include Craig and von Peter (2014) and Peltonen
et al. (2014). Departing from the micro-level analysis, some authors treat the network nodes as more aggregate entities,
such as countries (see, e.g., Kali and Reyes, 2010) or industries (see, e.g., Acemoglu et al., 2016).
14 Note that because the data from Euro Area Accounts are non-consolidated, they include intra-sector exposures. For the
clarity of the presentation, the intra-sector links are not shown in the graphs, but they are always accounted for in the
calculations.