Does central clearing reduce counterparty risk in realistic financial networks?

Novating a single asset class to a central counterparty (CCP) in an over-the-counter derivatives trading network impacts both the mean and variance of total net exposures between counterparties. When a small number of dealers trade in a relatively large number of asset classes, central clearing increases the mean and variance of net exposures, which may lead to increased counterparty risk and higher margin needs. There are intermediate cases where there is a trade-off: The introduction of a CCP leads to an increase in expected net exposures but this increase is accompanied by a reduction in variance. We extend the work of Duffie and Zhu (2011) by considering general classes of network structures and focus on scale-free and core-periphery structures, which have been shown to be accurate models of real-world financial networks. We find that a CCP is unlikely to be beneficial when the link structure of the network relies on just a few key nodes. In particular, in large scale-free networks a CCP will always worsen expected netting efficiency. In such cases, CCPs can improve netting efficiency only if agents have some degree of risk aversion that allows them to trade off the reduced variance against the higher expected netted exposures. This may explain why, in the absence of regulation, traders in a derivatives network may not develop a CCP themselves.