Do central counterparties reduce counterparty and liquidity risk? Empirical results

Abstract

A central counterparty (CCP) interposes itself between buyers and sellers of financial contracts to extinguish their bilateral exposures. Therefore, central clearing and settlement through a CCP should affect how financial institutions engage in financial markets. Though, financial institutions’ interactions are difficult to observe and analyze. Based on a unique transaction dataset corresponding to the Colombian peso non-delivery forward market, this article compares—for the first time—networks of transactions agreed to be cleared and settled by the CCP with those to be cleared and settled bilaterally. Networks to be centrally cleared and settled show significantly higher connectivity and lower distances among financial institutions. This suggests that agreeing on central clearing and settlement reduces liquidity risk. After CCP interposition, exposure networks show significantly lower connectivity and higher distances, consistent with a reduction of counterparty risk. Consequently, evidence shows CCPs induce a change of behavior in financial institutions that emerges as two distinctive economic structures for the same market, which corresponds to CCP’s intended reduction of liquidity and counterparty risks.

Keywords

Counterparty risk liquidity risk networks central clearing central counterparty D85 L14 G2

1 Introduction

A central counterparty (CCP) is a financial market infrastructure that interposes itself between counterparties in financial markets, becoming the counterparty to both the buy and sell-side of all trades and thereby ensuring the performance of open contracts. Central clearing and settlement through CCPs (i.e., hereafter central clearing) aims at mitigating counterparty risk 1 while increasing operational efficiency 2 and reducing opacity and complexity in financial markets (Ripatti, 2004; Bliss & Steigerwald, 2006; Manning, et al., 2009; Cecchetti, et al., 2009; Duffie & Zhu, 2011; Yellen, 2013; Acharya & Bisin, 2013; Wendt, 2015; Deng, 2017).

With central clearing, trading continues to take place on a bilateral basis, but once a trade agreement is reached between financial institutions, it is transferred to the CCP (Cecchetti, et al., 2009; Deng, 2017). Therefore, CCPs interposition should affect the network of relations (i.e., transactions and exposures) between financial institutions. Operating in a market that has agreed on central clearing should increase the availability of potential counterparties (Ripatti, 2004; Bliss & Steigerwald, 2006; Wendt, 2015) and thus should reduce liquidity risk 3 . Compared to bilateral clearing and settlement (i.e., hereafter bilateral clearing), central clearing should attain a more interconnected network of transactions in which financial institutions are closer to each other—counterparties are easier to find. After CCPs interposition, the network of exposures should be less interconnected and counterparties should be farther apart than in the network of transactions, therefore reducing counterparty risk.

Central clearing through CCPs has been studied from theoretical and modeling viewpoints (Jackson & Manning, 2007; Acharya & Bisin, 2013; Galbiati & Soramäki, 2013; Yellen, 2013; Garratt & Zimmerman, 2015; Deng, 2017). However, as reported by Loon and Zhong (2016) and Akari, et al. (2021), empirical studies of the effects caused by CCPs are scarce. Data corresponding to financial institutions interacting with the option to use central or bilateral clearing in the same market (i.e., the same jurisdiction, the same underlying asset, the same period) are elusive. This makes the empirical study of central clearing effects on financial markets challenging.

Further, financial markets are complex systems (Farmer, et al., 2012; Caldarelli, 2020). Studying how financial institutions interact and produce emergent connective structures requires suitable methods. Network analysis 4 has become a prevalent method for studying complex systems in economic contexts, including financial markets. 5 In this vein, following Caldarelli (2020), visualizing and describing the networks that result from financial institutions’ interactions is critical to understanding financial markets.

Accordingly, we implement network analysis methods to visualize and quantify the effects of the interposition of CCPs between financial institutions. We emphasize how transaction and exposure networks show alternate emerging economic structures, and how the differences between those structures correspond to the intended role of CCPs. We follow the approach that León and Sarmiento (2021) use to study the trade-off between counterparty risk and liquidity risk in money market networks. That is, we study the connectedness and distance of transaction and exposure networks corresponding to the over-the-counter (OTC) Colombian peso non-delivery forward market to empirically address a question: How does CCP clearing affect counterparty risk and liquidity risk in an OTC derivatives market?

Loon and Zhong (2014, 2016) address this question empirically. They study credit default swaps and bond trades in the United States and confirm that CCPs not only reduce counterparty risk but also liquidity risk. Mayordomo and Posch (2016) and Akari, et al. (2021) follow the research path of Loon and Zhong (2014, 2016). They all have in common that they address this question based on an econometric approach to the price of credit default swap trades in the United States market. Loon and Zhong (2014, 2016) and Mayordomo and Posch (2016) agree that CCPs reduce counterparty risk and liquidity risk. Akari, et al. (2021) find evidence of a reduction in liquidity risk only. 6 In our case, we address this question based on the analysis of the connective structure of transaction and exposure networks from the OTC Colombian peso non-delivery forward market.

We conclude that central clearing through CCPs effectively reduces liquidity risk in the transaction stage while reducing counterparty risk afterward. Financial institutions interacting under the central clearing option behave differently than those that interact under the bilateral clearing option, with significant differences in the emergent structure of the corresponding networks. This agrees with the expected role of CCPs and with results reported—from a different methodological approach—by Loon and Zhong (2014, 2016) and Mayordomo and Posch (2016) regarding the simultaneous mitigation of liquidity risk and counterparty risk.

The remainder of the paper is organized as follows. In the next section, we review how network structure is related to how financial markets balance counterparty and liquidity risk. In Section 3, we describe the network analysis methodology and the dataset. In Section 4, we present the results. In the last section, we conclude and discuss.

2 Network structure, counterparty, and liquidity risk

A network represents patterns of connections between the parts or elements of a system. The network structure is related to the outcome of a general optimization process that balances two opposing objectives: connectedness and distance between participants (Ferrer i Cancho & Solé, 2003; Gastner & Newman, 2006; Hojman & Szeidl, 2008; Newman, 2010; León & Sarmiento, 2021). Connections benefit network participants by providing access to other participants. However, connections entail costs and their benefits decrease with distance.

Minimizing distances and costs of connectedness generate different types of network structures depending on the weight assigned to each objective. Under this general framework, two limit cases of network structure are available. When connections entail no cost, a fully connected network (i.e., complete network) achieves the minimal distance between all participants (Fig. 1a). On the other hand, when connections entail a cost, but no participant is to be unconnected, a star network minimizes connectedness (Fig. 1b).

Fig. 1

Complete (a.) and star (b.) networks. In the complete network (panel a.) distance between participants is minimal. In the star network (panel b.) the cost of connectedness is minimal. Source: León and Sarmiento (2021).

This general framework accommodates financial networks too. As in Castiglionesi and Eboli (2015), absent any cost related to exchanging liquidity among financial institutions, a complete network is the most efficient network. However, counterparty risk, monitoring costs, capital requirements, finite liquidity, and heterogeneity among financial institutions deter them from connecting to all others and creates incentives for sparsely connected networks (Battiston, et al., 2012; Afonso, et al., 2013; Craig & von Peter, 2014; León & Berndsen, 2014; Castiglionesi & Eboli, 2018).

Accordingly, sparse structures are a stylized fact of financial networks (Boss, et al., 2004; Soramäki, et al., 2007; Battiston, et al., 2012; Martínez-Jaramillo et al. 2014, Craig & von Peter, 2014; León & Berndsen, 2014). This overlaps with evidence that financial institutions avoid excessive counterparty risk by establishing a few dedicated lending relationships (Cocco, et al., 2009; Afonso, et al., 2013; Temizsoy, et al., 2015) and by hoarding liquidity (Gale & Yorulmazer, 2013). In turn, sparse financial networks convey liquidity under-provision (Castiglionesi & Wagner, 2013). That is, there is a trade-off between the exposure to counterparty risk and the coverage of liquidity risk in financial networks (Castiglionesi & Eboli, 2018). From an empirical viewpoint, León & Sarmiento (2021) use the connectedness and distance trade-off of network optimization processes to study the exposure to counterparty risk and liquidity risk in the Colombian money market. Under this framework, they are able to characterize different segments of money markets (i.e., secured, unsecured) by examining how they balance counterparty risk and liquidity risk.

Although star networks are both constrained efficient and stable (Babus & Hu, 2017), financial networks do not converge to that structure by themselves. The literature agrees on the prevalence of core-periphery (Craig & von Peter, 2014; Martínez-Jaramillo et al., 2014), scale-free (Boss et al., 2004; Soramäki, et al., 2007) and modular scale-free (León & Berndsen, 2014) structures.

Financial networks’ convergence to star networks occurs sparingly. In the case of CCPs, they transform a bilateral network into a star-shaped network to reduce counterparty risk (and other related costs). The interposition of CCPs avoids excessive direct and indirect exposures among financial institutions, i.e., by concentrating counterparty risk in a robust dedicated financial market infrastructure. In this vein, CCPs are financial market infrastructures that are designed, regulated, and supervised to manage the single-point-of-failure risk this transformation entails. In the case of central banks, their liquidity facilities conform to a star-shaped network in which the central bank lends to financial institutions to reduce the distance (i.e., to reduce liquidity risk) between financial institutions during periods of uncertainty (see León & Sarmiento, 2021).

3 Methodology and data

A traditional representation of a network is the adjacency matrix. Let n represent the number of elements in the network, A is an adjacency matrix of dimensions n × n, with elements A_ij such that $A_{ij} = {\begin{matrix} 1 if there is a connection from i to j . \\ 0 otherwise \end{matrix}}$ (1)

In our case, financial institutions are the parts or elements of the system portrayed in the network. A connection from i to j corresponds to j holding a long US dollar position from i; in this case, A_ij = 1, and 0 otherwise. 7 Also, our adjacency matrix is directed, where a connection from i to j does not imply the existence of a connection from j to i.

Based on adjacency matrix A, network analysis (Börner, et al., 2007; Newman, 2010) provides measures of connectedness and distance. Density (d) measures connectedness as the ratio of observed to possible connections; it measures the cohesion of the network. Let n represent the number of participants and m the number of observed connections in a directed network containing no self-connections, $d = \frac{m}{n (n - 1)},$ (2)

where 0 < d ≤ 1

Mean geodesic distance (l) measures distance as the average shortest path between participants; unlike density, mean geodesic distance is determined by how connections are organized, and a closed-form solution is unavailable. Let g_ij be the shortest path (in number of connections) between i and j, the mean geodesic distance of a network (l) is calculated as the mean of l_i over all i-reachable vertexes, $l_{i} = \frac{1}{n - 1} \sum_{j (\neq i)} g_{ij}$ (3)

In the case of a complete network, density is maximal (d = 1) and distance is minimal (l = 1). In the case of a star network, as the number of participants increases, density tends to the minimum (d ∼ 0) whereas mean geodesic distance tends to 2 (l ∼ 2).

Network analysis encompasses many other measures. 8 We select reciprocity and transitivity as ancillary measures of connectedness. Reciprocity (r) measures the frequency with which a transfer from i to j is complemented by a transfer from j to i. As 0 ≤ r ≤ 1, r = 1 when the network is purely bidirectional (i.e., reciprocal) and r = 0 when the network is purely unidirectional. Transitivity (c), commonly referred to as clustering, is the ratio of the number of triangles (i.e., three nodes interlinked by three connections) to the number of connected triplets (i.e., three nodes linked by two or three connections). As 0 ≤ c ≤ 1, c = 1 when all triads are fully connected (i.e., complete connective transitivity) and c = 0 when no triad is fully connected (i.e., null connective transitivity). In the case of a complete network, reciprocity and transitivity are maximal (r = c = 1). In the case of a star network, transitivity is null (c = 0), whereas reciprocity is in the 0 ≤ r ≤ 1 range. $r = (\sum_{ij} A_{ij} A_{ji}) / - m$ (4)

$c = (number of triangles \times 3) / (number of connected triplets)$ (5)

Reciprocity and transitivity are relevant because they have been reported to convey information about trust between participants in financial networks (see Squartini, et al., 2013; Cimini, et al., 2019; León & Miguélez, 2021). In our case, akin to density, the higher the reciprocity and transitivity, the higher the exposure to counterparty risk. Moreover, reciprocity and transitivity are particularly interesting because they capture the main aim of the CCP as it becomes the counterparty to both buy and sell-side of all trades. That is, as the CCP interposes, transitivity drops to zero because no connection between financial institutions remains, whereas reciprocity increases because the CCP becomes the buyer and seller for all financial institutions.

Our networks are built from a unique dataset that contains transactions and exposures from the OTC Colombian peso non-delivery forward market, in US dollars, monthly from October 2011 to December 2018 (i.e., 87 observations). It is unique because the dataset is built by conciliating two different non-publicly available datasets from Banco de la República (Central Bank of Colombia, hereafter CBoC) and the sole local CCP (i.e., Cámara de Riesgo Central de Contraparte de Colombia S.A., hereafter CRCC). 9 We choose the Colombian peso non-delivery forward market because it contributes the most to the open interest of the CCP (42,0 percent as of 2018) and financial institutions may simultaneously operate under the options of bilateral or central clearing.

We build three time series of networks. The first corresponds to the networks of transactions to be cleared bilaterally; as no CCP interposes, it corresponds to exposures between financial institutions too. The second corresponds to the network of transactions to be cleared by the CCP. The third corresponds to the network of exposures after CCP interposition. Under the network optimization framework, density, mean geodesic distance, reciprocity, and transitivity enable us to evaluate counterparty and liquidity risk on the three time series of networks. We expect the first series (transactions to be cleared bilaterally) to display features consistent with higher counterparty risk aversion and lower market liquidity in the transaction stage, in the form of lower density, reciprocity, and transitivity, and higher mean geodesic distance than the second series (transactions to be cleared by the CCP). After CCP interposition, consistent with CCP’s purpose of mitigating counterparty risk, we expect the third series (exposures after CCP interposition) to display lower density and null transitivity, and higher mean geodesic distance than the second series—yet reciprocity should be high as the CCP acts as the buyer and seller to all financial institutions.

4 Results

Figure 2a corresponds to the network of transactions to be cleared bilaterally during December 2018. For the same period, 2b exhibits the network of transactions to be cleared by the CCP, whereas 2c displays the network of exposures after CCP interposition. Nodes represent financial institutions (circles) and the CCP (square). Arrows represent transactions/exposures, pointing to the counterparty holding a long US dollar position; width represents the contribution to the total value of transactions. Networks in Fig. 2a and Fig. 2b use a circular layout, whereas Fig. 2c uses a force (gravitational) layout.

Fig. 2

Networks of transactions for bilateral (a.) and central (b.) clearing, and of exposures after CCP interposition (c.), December 2018. Nodes represent financial institutions (circles) and the CCP (square). Arrows represent transactions/exposures, pointing to the counterparty holding a long US dollar position; width represents the contribution to the sum of all transactions. Networks in a. and b. use a circular layout, whereas c. uses a force (gravitational) layout. Source: authors’ calculations.

As expected, the network of transactions to be centrally cleared (b.) displays higher connectedness than that to be bilaterally cleared (a.). Higher connectedness (i.e., closer to a complete network) suggests that a lower distance exists between financial institutions. Likewise, after the interposition of the CCP (c.), the higher connectedness displayed in the transaction stage turns into a sparse star-shaped network with no connections between financial institutions.

Table 1 exhibits average density, mean geodesic distance, reciprocity, and transitivity for each series, from October 2011 to December 2018. Standard deviation is reported in brackets. A non-parametric test of distributional equality with respect to bilateral clearing is reported, i.e., Kolmogorov-Smirnov two-sample test (see Massey, 1951), at a 5 percent significance level. 10

Numerical results confirm the visual inspection of networks in Fig. 2. The series of transactions networks for central clearing display significantly higher density (d = 39.84), reciprocity (r = 81.20), and transitivity (c = 30.33), and significantly lower mean geodesic distance (l = 1.65) than those corresponding to the series for bilateral clearing (d = 8.41, r = 52.78, c = 9.70, l = 2.53). Therefore, with respect to transactions for bilateral clearing, we conclude that central clearing through the CCP reduces liquidity risk in the transaction stage: finding a counterparty is easier as the network is more interconnected and financial institutions are closer. 11

Exposure networks after CCP interposition display a density (d = 11.92) and transitivity (c = 0.00) that are visibly lower than those corresponding to the transaction stage (d = 39.84, c = 30.33). Further, the density after CCP interposition is not significantly different from that corresponding to networks for bilateral clearing (d = 8.41). However, reciprocity is higher after CCP interposition, which is expected as all positions (buy and sell) are novated by CCP to mitigate counterparty risk by neglecting any exposure between financial institutions—in the form of null transitivity. Consequently, with respect to networks of transactions for central clearing, we conclude that the interposition of CCP significantly reduces counterparty risk: the significantly lower connectedness (i.e., density and transitivity) and higher distance reveals that exposures between financial institutions decrease manifestly after the transaction stage.

Figure 3 exhibits the time series of density, mean geodesic distance, reciprocity, and transitivity for the three series of networks. Consistent with the analysis of Table 1, central clearing through the CCP attains lower liquidity risk in the transaction stage in the form of higher density, reciprocity, and transitivity, and lower mean geodesic distance. Similarly, after the interposition of the CCP, the counterparty risk is below that of the transaction stage, with lower connectedness (i.e., density and transitivity) and higher distance. All in all, Fig. 3 shows that conclusions drawn from Table 1 not only hold from a descriptive statistics viewpoint but also hold throughout the dataset. 12

Fig. 3

Time series of density (a.), mean geodesic distance (b.), reciprocity (c.), and transitivity (d.). Source: authors’ calculations.

Table 1

Mean and standard deviation (in brackets) of density, mean geodesic distance, reciprocity, and transitivity. Rejection of the null hypothesis of distributional equality with respect to bilateral clearing is marked with ^* (Kolmogorov-Smirnov non-parametric two-sample test at 5 percent significance level). The series of transaction networks for central clearing display significantly higher density, reciprocity, and transitivity, and significantly lower mean geodesic distance than those corresponding to the series for bilateral clearing. Therefore, central clearing through the CCP reduces liquidity risk in the transaction stage. Exposure networks after CCP interposition display a density and transitivity that are visibly lower than those corresponding to the transaction stage. The density after CCP interposition is not significantly different from that corresponding to networks for bilateral clearing. However, reciprocity is higher after CCP interposition, which is expected as all positions (buy and sell) are novated by CCP. Consequently, with respect to networks of transactions for central clearing, we conclude that the interposition of CCP significantly reduces counterparty risk. Source: authors’ calculations

	Transactions for bilateral clearing	Transactions for CCP clearing	Exposures after CCP interposition
Density (d) (× 100)	8.41 [1.31]	39.84^* [8.16]	11.92 [3.59]
Mean geodesic distance (l)	2.53 [0.36]	1.65^* [0.12]	1.87^* [0.04]
Reciprocity (r) (× 100)	52.78 [8.96]	81.20^* [6.29]	95.45 [3.66]
Transitivity (c) (× 100)	9.70 [7.17]	30.33^* [8.05]	0.00^* [0.00]

5 Discussion

Results show that the option to clear bilaterally or centrally creates two alternate emerging economic structures within the same market. Financial institutions interacting under the central clearing option behave differently from those that interact under the bilateral clearing option. Visualizing and analyzing the networks enable us to study the emerging economic structures that result from financial institutions’ behavior.

From a network optimization framework, the structure corresponding to central clearing shows that the interposition of the CCP reduces liquidity risk in the transaction stage while reducing counterparty risk afterward. This agrees with what is expected from CCPs from theoretical and modeling viewpoints prevailing in related literature. Moreover, this agrees with Loon and Zhong (2014, 2016) and Mayordomo and Posch (2016), who use different methodologies to find that credit default swaps and bond trades that are centrally cleared in the United States exhibit lower liquidity risk and counterparty risk. In this vein, our work adds to existing literature regarding the contribution of CCPs to the simultaneous mitigation of counterparty risk and liquidity risk but with a different methodology, different jurisdiction, and different asset.

Interestingly, our empirical results are related to a particular finding reported in Deng (2017). Although counterparty risk between financial institutions should be irrelevant as the actual exposure will be against the CCP, the transaction network corresponding to the central clearing option significantly diverges from a complete network. This overlaps with the model constructed by Deng (2017), who finds that aggregate (i.e., macro) risk causes full counterparty risk insurance to falter. In this vein, it is apparent that financial institutions in Colombia exert search effort in the transaction stage notwithstanding the subsequent interposition of the CCP; yet, as expected, that effort is significantly lower than when bilateral clearing is agreed. From a different perspective, results are related to Antinolfi et al. (2018), who find that mandatory central clearing cause a potential loss of information due to decreased incentives to monitor other financial institutions; in our case, as central clearing is discretionary, financial institutions retain incentives to monitor each other and this may explain why full counterparty risk insurance falters.

Results are noteworthy because it is the first time the effect of CCPs on the relations between financial institutions is visualized and quantified based on observed data. Also, it is the first time the emergent connective structures arising from bilateral and central clearing are studied based on observed data. This provides new elements for existing theoretical and modeling approaches to the study of CCPs. We acknowledge that the Colombian case displays certain idiosyncrasies, but we expect similar results for other markets; in this vein, although Loon and Zhong (2014, 2016) and Mayordomo and Posch (2016) use different methods to study the effects of CCPs in a different jurisdiction with different assets, their results point in the same direction. Also, our work may contribute to study other markets and other cases in which a trade-off between counterparty risk and liquidity risk exists, such as interbank lending markets.

Footnotes

Acknowledgments

Opinions and statements in this article are the sole responsibility of the authors and do not represent those of the institutions they are affiliated to. We thank Pamela Cardozo, Freddy Cepeda, Oscar Leyva, Clara Machado, and Hernando Vargas for their comments and suggestions to an early version of this manuscript. We thank the anonymous reviewers and the editor for their contributive work.

Appendix

Table A2

Mean and standard deviation (in brackets) of density, mean geodesic distance, and transitivity, calculated on net transactions and exposures; reciprocity is not reported as reciprocity is zero after netting transactions and exposures. Rejection of the null hypothesis of distributional equality with respect to bilateral clearing is marked with ^* (Kolmogorov-Smirnov non-parametric two-sample test at 5 percent significance level). The series of transaction networks for central clearing display significantly higher density and transitivity, and significantly lower mean geodesic distance than those corresponding to the series for bilateral clearing. Therefore, central clearing through the CCP reduces liquidity risk in the transaction stage. Exposure networks after CCP interposition display a density and transitivity that are visibly lower than those corresponding to the transaction stage for CCP clearing. The density after CCP interposition is not significantly different from that corresponding to networks for bilateral clearing but transitivity is null. Consequently, with respect to networks of transactions for central clearing, we conclude that the interposition of CCP significantly reduces counterparty risk. Source: authors’ calculations

	Transactions for bilateral clearing	Transactions for CCP clearing	Exposures after CCP interposition
Density (d) (× 100)	11.85 [1.71]	45.09^* [7.88]	12.59 [3.68]
Mean geodesic distance (l)	2.52 [0.36]	2.09^* [0.29]	1.78^* [0.05]
Transitivity (c) (× 100)	2.72 [2.57]	5.43^* [3.73]	0.00^* [0.00]

Counterparty risk is related to a counterparty not paying or delivering as agreed—due to intrinsic or systemic reasons (e.g., contagion).

In the form of multilateral netting of exposures and payments, and more efficient management of collaterals (see Cecchetti, et al., 2009).

Liquidity risk is related to a financial institution being unable to enter into a transaction to exchange its position into other.

Network analysis aims at describing and understanding an underlying system, focused on capturing the system’s structure (see Börner, et al., 2007).

As documented in , the Global Financial Crisis spurred research aimed at identifying financial networks’ observed connective structure.

Nevertheless, reports of increased market liquidity due to the availability of CCP clearing is common (see Ripatti, 2004, Da Silva, et al., 2017).

The dataset enables us to work with weighted adjacency networks, where connections are weighted according to the dollar value of the relation. However, in our case we focus on the structure (i.e., the topology) of the network. Further, as it has been reported in Squartini, et al. (2013), unweighted properties often convey more information about real-world economic networks than that corresponding to weighted properties.

Network analysis encompasses many other measures that describe the connective structure of the network or that measure the importance of the participants or the connections. However, we focus on connectedness and distance because they are the mainstay of the network optimization process framework introduced in Section 2.

Conciliating both datasets requires an especially designed algorithm for extracting the transaction data out of exposures data with great accuracy. The algorithm’s accuracy is about 96 percent.

The cumulative distribution of the series is reported in Fig. A1 (in the ). In those cases, in which the Kolmogorov-Smirnov two-sample non-parametric test of distributional equality is rejected (not rejected), the differences in the distribution are manifest (slight).

This conclusion holds when comparing with bilateral networks before central clearing became available in May 2010 (see Table A1 in the Appendix). This suggests that the introduction of the CCP increased the liquidity of the market—as reported by and Da Silva, et al. (2017) for the repo and credit default swaps markets, respectively.

Furthermore, as displayed in Table A2 (in the Appendix), conclusions drawn from are robust to using net transactions and exposures.

References

Acharya,

& Bisin,

, 2014. Counterparty risk externality: Centralized versus over-the-counter markets, Journal of Economic Theory149, 153–182.

Afonso,

, Kovner,

, & Schoar,

, 2013. Trading partners in the interbank lending market, Federal Reserve Bank of New York Staff Reports 620, Federal Reserve Bank of New York.

Akari,

M-A.

, Ben-Abdallaa,

, Breton,

, & Dionne,

, 2021. The impact of central clearing on the market for single-name credit default swaps. North American Journal of Economics and Finance, 56.

Antinolfi,

, Carapella,

, & Carli,

, 2018. Transparency and collateral: central versus bilateral clearing. Finance and Economics Discussion Series 2018-017, Board of Governors of the Federal Reserve System.

Babus,

& Hu,

, 2017. Endogenous intermediation in over-the-counter markets, Journal of Financial Economics125, 200–215.

Battiston,

, Delli,

, Gallegati,

, wald,

, & Stiglitz,

J.E.

, 2012. Liaisons dangereuses: Increasing connectivity, risk sharing, and systemic risk, Journal of Economic Dynamics & Control36, 1121–1141.

Bliss,

R.R.

& Steigerwald,

R.S.

, 2006. Derivatives clearing and settlement:Acomparison of central counterparties and alternative structures, Economic Perspectives30(4), 22–29.

Börner,

, Sanyal,

, & Vespignani,

, 2007. Network science, Annual Review of Information Science and Technology41 (1), 537–607.

Boss,

, Elsinger,

, Summer,

, & Thurner,

, 2004. The network topology of the interbank market, Quantitative Finance6 (4), 677–684.

10.

Caldarelli,

, 2020. A perspective on complexity and network science, Journal of Physics: Complexity1(2).

11.

Castiglionesi,

& Eboli,

, 2018. Liquidity flows in interbank networks, Review of Finance22(4), 1291–1334.

12.

Castiglionesi,

& Wagner,

, 2013. On the efficiency of bilateral interbank insurance, Journal of Financial Intermediation22, 177–200.

13.

Ceccheti,

S.G.

, Gyntelberg,

, & Hollanders,

, 2009. Central counterparties for over-the-counter derivatives, BIS Quarterly Review, September 45–58.

14.

Cimini,

, Squartini,

, Saracco,

, Garlaschelli,

, Gabrielli,

& Caldarelli,

, 2019. The statistical physics of real-world networks, Nature Reviews Physics1, 58–71.

15.

Cocco,

J.F.

, Gomes,

F.J.

, & Martins,

N.C.

, 2009. Lending relationships in the interbank market, Journal of Financial Intermediation1824–48.

16.

Craig,

& von Peter,

, 2014. Interbank tiering and money center banks, Journal of Financial Intermediation23322–347.

17.

Da Silva,

P.P

, Vieira,

, & Vieira,

, 2018. Central clearing and CDS market quality, Journal of Futures Markets38731–753.

18.

Deng,

, 2017. Counterparty risk, central counterparty clearing and aggregate risk, Annals of Finance13355–400.

19.

Duffie,

& Zhu,

, 2011. Does a central clearing counterparty reduce counterparty risk? , Review of Asset Pricing Studies1(1) 74–95.

20.

Farmer,

J.D.

, Gallegati,

, Hommes,

, Kirman,

, Ormerod,

, Cincotti,

, Sanchez,

, & Helbing,

, 2012. A complex systems approach to constructing better models for managing financial markets and the economy, The European Physical Journal214295–324.

21.

Ferrer i Cancho,

& Solé,

R.V.

, 2003. Optimization in complex networks. In Statistical Mechanics of Complex Networks (Eds. Pastor Satorras, R., Rubi, J., & Díaz-Aguilera, A), 114–125. Springer: Berlin-Heidelberg.

22.

Galbiati,

& Soramäki,

, 2013. Central counterparties and the topology of clearing networks. Working Paper Series480, Bank of England.

23.

Gale,

& Yorulmazer,

, 2013. Liquidity hoarding, Theoretical Economics8291–324.

24.

Garratt,

& Zimmerman,

, 2015. Does central clearing reduce counterparty risk in realistic financial networks?. Staff Reports, 717, Federal Reserve Bank of New York.

25.

Gastner,

M.T.

& Newman,

M.E.J.

, 2006. Optimal design of spatial distribution networks, Physical Review EE 74016117.

26.

Hojman,

D.A.

& Szeidl,

, 2008. Core and periphery in networks, Journal of Economic Theory139295–309.

27.

Jackson,

J.P.

& Manning,

M.J.

, 2007. Comparing the pre-settlement risk implications of alternative clearing arrangements, Working Paper Series321, Bank of England.

28.

León,

& Berndsen,

R.J.

, 2014. Rethinking financial stability: challenges arising from financial networks’ modular scale-free architecture, Journal of Financial Stability15241–256.

29.

León,

& Miguélez,

, 2021. Securities cross-holding in the Colombian financial system: A topological approach. Studies in Economics and Finance.

30.

León,

& Sarmiento,

, 2021. The trade-off between liquidity risk and counterparty risk in money market networks, Journal of Financial Market Infrastructures9(2) 1–27.

31.

Loon,

Y.C.

& Zhong,

, 2016. Does Dodd-Frank affect OTC transaction costs and liquidity? Evidence from real-time CDS trade reports, Journal of Financial Economics119672.

32.

Loon,

Y.C.

& Zhong,

, 2014. The impact of central clearing on counterparty risk, liquidity, and trading: Evidence from the credit default swap market, Journal of Financial Economics11291–115.

33.

Manning,

, Nier,

, & Schanz,

, 2009. The economics of large-value payments and settlement: theory and policy issues for central banks, Oxford University Press: Oxford.

34.

Martínez-Jaramillo,

, Alexandrova-Kabadjova,

, Bravo-Benítez,

, & Solórzano- Margain,

J.P.

, 2014. An empirical study of the Mexican banking system’s network and its implications for systemic risk, Journal of Economic Dynamics and Control40242–265.

35.

Massey,

F.J.

, 1951. The Kolmogorov-Smirnov test for goodness of fit, Journal of the American Statistical Association46(253) 68–78.

36.

Mayordomo,

& Posch,

P.N.

, 2016. Does central clearing benefit risky dealers? Journal of International Financial Markets, Institutions & Money4291–100.

37.

Newman

M.E.J.

, 2010. Networks. Oxford University Press: New York.

38.

Ripatti,

, 2004. Central counterparty clearing: constructing a framework for evaluation of risks and benefits. Discussion Papers, 30, Bank of Finland.

39.

Soramäki,

, Bech,

, Arnold,

, Glass,

, & Beyeler,

, 2007. The topology of interbank payments flow, Physica A379317–333.

40.

Squartini,

, van Lelyveld,

, & Garlaschelli,

, 2013. Early-warning signals of topological collapse in interbank networks, Scientific Reports33357.

41.

Temizsoy,

, Iori,

, & Montes-Rojas,

, 2015. The role of bank relation in the interbank market, Journal of Economic Dynamics & Control59118–141.

42.

Wendt,

, 2015. Central counterparties: addressing their too important to fail nature, Working Paper, WP/15/21, International Monetary Fund.

43.

Yellen,

J.L.

, 2013. Interconnectedness and systemic risk: Lessons from the financial crisis and policy implications. Board of Governors of the Federal Reserve System. American Economic Association/American Finance Association Joint Luncheon, San Diego, California, January 4^th.