集中清算与回购市场特殊性定价

Central clearing and the pricing ofspecialness in repo markets Piotr Danisewicz, Tobias Dieler,Loriano Mancini, Francesco Mazzari,Julian Metzler Abstract Repo markets clear either bilaterally over the counter (OTC) or through central counter-parties (CCPs), which differ in how counterparty risk is priced. In bilateral markets, reporates reflect borrower-specific risk, while CCP clearing pools counterparties and applies acommon pricing rule.We develop a model of security-driven repo in which repo rates arenon-linear in borrower risk. As a result, averaging borrower-specific OTC prices yields morenegative rates than pricing the pooled borrower in CCP markets. The model predicts thatthe CCP–OTC specialness gap compresses during periods of counterparty uncertainty andvaries with borrower and collateral characteristics.Using transaction-level data from theeuro-area interbank repo market around the March 2020 COVID-19 shock, we find evidenceconsistent with these predictions. Our results show that central clearing dampens specialnessin normal times but stabilizes repo pricing during stress. Keywords:Asymmetric information, Counterparty risk, Uncertainty, Collateral risk, Interbankmarkets JEL Codes:D47, D82, G14, G15, G21 Non-technical summary The paper studies how the structure of the euro area interbank repo market affects the cost ofborrowing collateral via over the counter (OTC) or centrally cleared (CCP) transaction. In repotransactions, one party provides cash and receives a security as collateral, agreeing to reverse thetransaction at maturity. When the primary motive is to obtain a specific security rather thancash, the transaction is referred to as security–driven (or reverse repo). In Europe, repos canbe traded either bilaterally in OTC markets, where counterparties know each other’s identityand directly bear counterparty risk, or via CCPs, where trades are anonymous and the CCPbecomes the buyer to every seller and the seller to every buyer. In CCP markets, counterpartyrisk is pooled and mutualised through a default fund. This institutional difference has importantpricing implications: in OTC markets, repo rates reflect borrower-specific risk, while in CCPmarkets rates reflect the average risk of the participant pool. The paper develops a theoretical model of security-driven repo tailored to the institutionalfeatures of the European market. A central insight of the model is that repo rates are a non-linear function of borrower default risk. In bilateral OTC markets, lenders can condition priceson borrower identity, setting a specific rate for each borrower type. The observed market rate istherefore an average of borrower-specific prices. In CCP markets, by contrast, pricing applies tothe pooled “average borrower”, since individual identities are not reflected in transaction-levelprices. Because the pricing function is non-linear, averaging borrower-specific prices is not thesame as pricing the average borrower. As a result, the average OTC rate is more negative thanthe CCP rate. Intuitively, in OTC markets lenders fully internalize borrower-specific risk whilein CCP markets, risk is diluted across the pool. This aggregation effect generates a systematicwedge between OTC and CCP rates. The model delivers three main predictions that align closely with the empirical findings.First, in normal times, special securities should trade at more negative rates in OTC marketsthan in CCP markets, reflecting the non-linear pricing of borrower-specific risk. Second, whencounterparty uncertainty increases – as during the COVID-19 shock – the CCP–OTC differentialshould compress. The reason is that as lenders become more concerned about the distributionof borrower types, pooling risk in CCP markets becomes relatively more costly, raising CCPrates towards OTC levels. Third, the magnitude of this compression should depend on borrowerand collateral characteristics. When borrower quality deteriorates, the non-linearity in pricing becomes stronger, so identity-based OTC pricing remains more distinct from pooled CCP pric-ing, dampening the compression. By contrast, when collateral is of particularly high quality andtherefore highly valued, the interaction between counterparty risk and collateral risk becomesmore pronounced in CCP markets, amplifying the compression for safe assets. Using a differences-in-differences setting applied to confidential transaction-level data fromthe Eurosystem’s Money Market Statistical Reporting (MMSR) dataset, we test the modelpredictions about the differential cost of borrowing on CCP and OTC repos around the Covid-19 pandemic. Following the COVID-19 uncertainty shock, the CCP–OTC differential compressessharply. Borrowing costs in CCP markets increase relative to OTC markets by roughly 6 basispoints. This compression is heterogeneous: it is weaker for riskier borrowers – measured usingnon-performing loan (NPL) ratios and credit default swap (CDS) spreads – and stronger forhigh-quality c