abstract bibtex

This paper investigates how the structure of re-hypothecation networks affects the dynamics of endogenous total liquidity and the emergence of systemic risk within the financial system. Re- hypothecation consists in the right of reusing the collateral of a transaction many times over. Re-hypothecation increases the liquidity of market players, as those players can use the collateral received to honor another obligation. At the same time re-hypothecation lowers parties’ actual coverage against counterparty risk, because the same collateral secures more than one transaction, and it can therefore be a source of systemic risk. To study the above issues, we build a model where banks are linked by chains of repo contracts and use or re-use a fixed amount of initial collateral. We develop few variants of the basic model, by introducing elements of increasing complexity and we performed several exercises aimed at evaluating the resilience and the liquidity of the system in presence of re-hypothecation. First, we analyse how the resilience and the liquidity depends on the network density. To this end, following previous results of some of the authors in the context game- theoretical analysis of economic networks evolution (Koenig ea. 2012), we generate a set of nested- split graphs of increasing density. Nested-split networks are known to be in the class of graphs that maximize the degree variance. At the other extreme, we also generate a set of graphs of increasing density, each of which is maximally homogenous in degree for that given density. We find that higher density generates more collateral, and thus more liquidity, but also it implies more losses given the same percentage shock to the value of initial proprietary collateral. Second, we find that also network sizes matters. Thus, having more actors in the financial networks increases the total liquidity created via the re-hypothecation networks. At the same time, it also implies less resilience, as total collateral losses (relative to shock of given size) also increase. Third, we find that – for given density - network architecture matters as well. Our results indeed show that moving from a degree-homogenous structure to a nested star structure implies an increase in total collateral created as well as in the size of after-shock total liquidity losses. Finally, we show that the above collateral loss estimates are just a lower bound of the potential negative impact of an adverse shock in presence of re-hypothecation. More precisely, we augment the model with liquidity hoarding effects (e.g. Gai et al., 2011) and we show that the impact on liquidity of a shock is in that case higher. The policy implications of the above results are also discussed.

@article{ title = {Collateral Unchained: Rehypothecation Networks, Complexity and Systemic Effects}, type = {article}, id = {6f322dab-6dc5-35b3-94b5-d6b90407e49f}, created = {2016-11-01T17:41:49.000Z}, file_attached = {false}, profile_id = {e7ad1df1-57da-3801-b924-2e51ef162a9d}, group_id = {bd0f1172-1660-3fe3-b7d0-9f2d0b4051ad}, last_modified = {2016-11-21T13:42:27.000Z}, tags = {DOLFINS_T1.1,DOLFINS_WP1}, read = {false}, starred = {false}, authored = {false}, confirmed = {true}, hidden = {false}, citation_key = {Luu}, abstract = {This paper investigates how the structure of re-hypothecation networks affects the dynamics of endogenous total liquidity and the emergence of systemic risk within the financial system. Re- hypothecation consists in the right of reusing the collateral of a transaction many times over. Re-hypothecation increases the liquidity of market players, as those players can use the collateral received to honor another obligation. At the same time re-hypothecation lowers parties’ actual coverage against counterparty risk, because the same collateral secures more than one transaction, and it can therefore be a source of systemic risk. To study the above issues, we build a model where banks are linked by chains of repo contracts and use or re-use a fixed amount of initial collateral. We develop few variants of the basic model, by introducing elements of increasing complexity and we performed several exercises aimed at evaluating the resilience and the liquidity of the system in presence of re-hypothecation. First, we analyse how the resilience and the liquidity depends on the network density. To this end, following previous results of some of the authors in the context game- theoretical analysis of economic networks evolution (Koenig ea. 2012), we generate a set of nested- split graphs of increasing density. Nested-split networks are known to be in the class of graphs that maximize the degree variance. At the other extreme, we also generate a set of graphs of increasing density, each of which is maximally homogenous in degree for that given density. We find that higher density generates more collateral, and thus more liquidity, but also it implies more losses given the same percentage shock to the value of initial proprietary collateral. Second, we find that also network sizes matters. Thus, having more actors in the financial networks increases the total liquidity created via the re-hypothecation networks. At the same time, it also implies less resilience, as total collateral losses (relative to shock of given size) also increase. Third, we find that – for given density - network architecture matters as well. Our results indeed show that moving from a degree-homogenous structure to a nested star structure implies an increase in total collateral created as well as in the size of after-shock total liquidity losses. Finally, we show that the above collateral loss estimates are just a lower bound of the potential negative impact of an adverse shock in presence of re-hypothecation. More precisely, we augment the model with liquidity hoarding effects (e.g. Gai et al., 2011) and we show that the impact on liquidity of a shock is in that case higher. The policy implications of the above results are also discussed.}, bibtype = {article}, author = {Luu, Duc and Napoletano, Mauro and Battiston, Stefano and Barucca, Paolo}, journal = {paper-progress} }

Downloads: 0