Central Counterparty Resilience in Financial Networks: An Analysis of Default Waterfall Mechanics

Sammanfattning

Central counterparties (CCPs) have become systemically important institutions inmodern financial markets following the regulatory reforms introduced after the 2007–2009financial crisis. By inserting themselves as the new buyer to every seller and new sellerto every buyer in a derivatives transaction, CCPs eliminate counterparty credit riskbetween clearing members. When a clearing member cannot meet its obligations, lossesare absorbed through the so called default waterfall, an ordered sequence of prefundedfinancial resources drawn upon layer by layer before losses can propagate to the broaderfinancial system. Despite their importance, existing research tends to examine the design of these resources and the structure of the financial network in which a CCPoperates in isolation, leaving open whether resilience benefits are robust across structurally different clearing environments. This thesis addresses that gap by integratingthe financial network clearing model of Eisenberg and Noe (2001) with the CCP defaultwaterfall framework of Ghamami et al. (2022) across three synthetic network topologiesof increasing structural complexity. The framework allows for the controlled variationof two key design dimensions simultaneously, namely the severity of an exogenous market shock, and the allocation between initial margin (IM), member-specific collateralposted to cover individual position risk, and the default fund (DF), the shared poolthrough which losses exceeding IM are mutualised across surviving clearing members.Our numerical studies render three main findings. First, heterogeneous, hub-drivennetworks exhibit lower shock thresholds for CCP default and sharper loss escalationthan dispersed structures. Second, the Cover-2 calibration standard is more sensitive to hub concentration than a uniform regulatory framework. This implies holdingnetwork size and connectivity constant, greater exposure concentration among hubmembers raises the required default fund. Consequently, identical regulatory rulesproduce meaningfully different levels of effective protection across markets. Third, reallocating IM toward the DF reduces CCP default probability more effectively thanincreasing total collateral by an equivalent amount, with the benefit materialising atlower shock thresholds in concentrated networks. This gain, however, comes at a directcost to surviving members, whose resources are drawn upon to cover defaulting members’ shortfalls, creating rational incentives to resist the system-optimal level of lossmutualisation. Collectively, these results suggest that regulatory frameworks cannotrely on a uniform approach. Accounting for both network topology and the incentivetensions inherent in loss mutualisation is imperative for effective CCP oversight.