No more tears without tiers? The impact of indirect settlement on liquidity use in TARGET2

Paulick, Jan; Berndsen, Ron; Diehl, Martin; Heijmans, Ronald Article — Published Version No more tears without tiers? The impact of indirect settlement on liquidity use in TARGET2 Empirica Provided in Cooperation with: Springer Nature Suggested Citation: Paulick, Jan; Berndsen, Ron; Diehl, Martin; Heijmans, Ronald (2023) : No more tears without tiers? The impact of indirect settlement on liquidity use in TARGET2, Empirica, ISSN 1573-6911, Springer US, New York, NY, Vol. 51, Iss. 2, pp. 425-458, https://doi.org/10.1007/s10663-023-09597-6 This Version is available at: https://hdl.handle.net/10419/318029 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Empirica (2024) 51:425–458 https://doi.org/10.1007/s10663-023-09597-6 1 3 ORIGINAL PAPER No more tears withouttiers? The impact ofindirect settlement onliquidity use inTARGET2 JanPaulick1,2 · RonBerndsen3,4,6,7,8· MartinDiehl1· RonaldHeijmans5 Accepted: 20 October 2023 / Published online: 13 December 2023 © The Author(s) 2023 Abstract We study the impact of tiered payments originating from client banks on the liquidity consumption (relative intraday liquidity use) of settlement banks. Estimates of a panel data model, employing wholesale payments in euro, show that a higher share of tiered payments reduces liquidity consumption by settlement banks. Metrics on timing, delay, and payment priorities suggest that settlement banks use more leeway in settling tiered payments from client banks compared to in-house payments. Payment timing as a proxy for external delay suggests that tiered payments help smooth liquidity positions. Payment delay within the system does not follow a clear dynamic over time, whereas banks consistently de-prioritize tiered payments. Thereby, settlement banks employ tiered arrangements to manage intraday liquidity more efficiently. To a certain extent, this hints at “free riding” or higher recycling of liquidity from client banks’ payments. However, the results are also consistent with settlement banks’ monitoring role or tiered payments potentially exhibiting different characteristics which may be attributable to contractual arrangements. Keywords RTGS systems· Banks· Payments· Tiering· Liquidity· TARGET2 JEL Classification E42· E58· G21 Responsible Editor Julia Wörz. The authors thank participants of the 55th Annual Conference of the Canadian Economics Association and the 13th Payment and Settlement System Simulation Seminar at the Bank of Finland. We thank James Chapman, Anneke Kosse and Segun Bewaji for their very helpful comments and suggestions, Constanza Martínez for an excellent discussion of preliminary results and two anonymous referees for their excellent comments. Outstanding research assistance was provided by Josefine Quast during the early stages of the work. Diehl, Heijmans and Paulick are or were members of one of the user groups with access to TARGET2 data in accordance with Article 1(2) of Decision ECB/2017/2080 of 22 September 2017 on access to and use of certain TARGET2 data. The Deutsche Bundesbank, De Nederlandsche Bank and the MIB have checked the paper against the rules for guaranteeing the confidentiality of transaction-level data imposed by the MIB pursuant to Article 1(4) of the above mentioned issue. The views expressed in the paper are solely those of the authors and do not necessarily represent the views of their affiliated organizations. Extended author information available on the last page of the article 426 Empirica (2024) 51:425–458 1 3 1 Introduction Payment systems form the basis for the settlement of debt obligations in an economy. A universal feature of payment systems is the settlement of payments on behalf of clients by direct system participants, called tiering. Instead of directly sending payments to a payment system, some banks1 choose to delegate settlement, akin to correspondent banking arrangements. The arrangement of an indirect participant (client bank) processing its payments through a direct participant (settlement bank) forms a tiered arrangement. The underlying economic reasons that influence banks’ decision on how to access a payment system are manifold. For smaller banks, it might be more cost-efficient to choose tiered settlement arrangements, avoiding costs related to operational setup and liquidity management. In addition, many jurisdictions restrict direct access to a payment system for foreign banks. The Principles for Financial Market Infrastructures (PFMIs), developed by CPSSIOSCO (2012), specify high international standards for financial market infrastructures (FMIs) such as payment systems and offer guidance on potential sources of risk and risk mitigation. Principle 19.4 of the PFMIs states that “an FMI should regularly review risks arising from tiered participation”. Such risks include credit, liquidity and operational risk. According to the PFMIs, these risks may be especially large for highly tiered systems. From a regulator’s view, tiering comes at the potential cost of concentration and hence operational bulk risks. A large share of tiered payments may increase exposures of settlement banks to client banks and vice versa. Liability issues may arise for such exposures in the event of a default. However, tiered participation not only entails risks but can increase the efficiency of payment systems. Costs of payment settlement decrease with tiered participation, as direct participants can profit from economies of scale [see for example Adams etal. (2010); Chapman etal. (2013)]. Tiering generally decreases liquidity consumption, as payments offset each other when concentrated among fewer direct participants. Pooling liquidity leads to lower cost of capital, as higher traffic volumes offset payments within the system or because banks settle payments internally on their own books without drawing on liquidity in their payment system account.2 However, little is known about the effect of tiering at the participant level and how tiering factors into banks’ active liquidity management. From a settlement bank’s perspective, tiered payments feed into the overall liquidity disposition of payments that are settled in a way to minimize liquidity use. Monitoring intraday liquidity is part of the Basel framework to ensure banks are able to meet payment obligations. Banks monitor intraday liquidity metrics in accordance 1 With regard to terminology, we use the term credit institution interchangeably with the term bank throughout the paper. Direct participants in a payment system are referred to as settlement banks, while indirect participants are referred to as client banks. For brevity, we at times refer to settlement banks as banks and direct participants as participants. 2 Pooling here does not refer to strategically timing payments but rather to liquidity savings due to offsetting when payments are concentrated among fewer participants. 427 1 3 Empirica (2024) 51:425–458 with Basel Committee on Banking Supervision (2019), also factoring in secured and unsecured credit lines and unencumbered assets. Central bank reserves and participation in wholesale payment systems are only one part of banks’ liquidity management, but arguably one of the most important parts. In this study, we empirically investigate the effect of tiering on the relative intraday liquidity use—referred to as liquidity consumption here—of settlement banks in TARGET2, the Eurosystem’s wholesale payment system.3 Liquidity consumption is defined as the maximum amount of liquidity needed in the course of a day to settle the payments of a settlement bank relative to all payments sent by that bank. The measure indicates how efficiently a bank uses liquidity to settle its payments. Using transaction-level data from TARGET2 spanning 10 years and a total of more than 1200 direct participants, we find that higher shares of tiered payments reduce the liquidity consumption of settlement banks. This finding sheds light on why banks have an incentive to provide settlement services for other banks. The results are robust controlling for pooling effects via bank fixed effects, general payment activity and other factors.4 The main driver appears to be that banks have more discretion in settling tiered payments. Payment timing suggests that tiered payments help settlement banks to more effectively manage intraday liquidity, for example by assigning lower priorities to them compared to banks’ own or in-house payments. As a result, settlement banks employ tiered arrangements to more effectively manage intraday liquidity. To some degree, this indicates discriminatory practices, as settlement banks treat their own payments with higher urgency, thus using more liquidity for settling own payments relative to tiered payments. Settlement banks reducing their cost of liquidity is consistent with their role in monitoring client banks and offering cost-efficient settlement services based on private information relevant to creditworthiness [see for example Chapman etal. (2013)]. From this perspective, settlement banks mitigate risk from tiered arrangements by smoothing their liquidity positions. At the same time, the findings are also consistent with tiered payments exhibiting different characteristics. Tiered payments may arrive later in the day and by nature be less urgent, so settlement banks can use these payments to optimize their liquidity positions. As acquiring liquidity for payment purposes is usually costly, settlement banks exhibiting lower values of liquidity consumption settle payments in a more costefficient manner. Liquidity in the form of central bank reserves can be assumed to be costly as it is acquired from the central bank or the interbank market. There is an opportunity cost to dedicating liquidity for the purposes of settling payments and from pledging collateral. When liquidity is scarce and interest rates are high, the cost of acquiring liquidity is thus expected to be larger. Consistent with this reasoning, liquidity consumption is lower when liquidity is scarce and interest rates are high. 3 TARGET2 refers to the second generation Trans-European Automated Real-time Gross Settlement Express Transfer System operated by the Eurosystem. 4 Pooling effects stem from banks settling more payments overall, which reduces their relative liquidity consumption compared to a situation in which multiple banks settle their own payments. 428 Empirica (2024) 51:425–458 1 3 The paper estimates the effect of tiered payments on liquidity consumption and sheds light on whether a predominantly risk-based view of tiering is warranted. The results indicate that tiered payments give banks more leeway in liquidity management. Benefits and risks should be weighed more carefully by system designers and overseers. 2 Tiering inlarge‑value payment systems Large-Value Payment Systems (LVPSs) typically settle transfers that are high value or high priority. Many LVPSs, such as TARGET2, settle transactions immediately on a gross basis and are also referred to as Real-Time Gross Settlement (RTGS) systems. This stands in contrast to net settlement systems in which net payment positions are settled at a specified time. Hence, RTGS systems require higher amounts of liquidity for settling payments. Aside from central banks and government entities, direct access to an LVPS is mostly restricted to banks. Banks can choose to access payment systems directly or through a direct participant (correspondent bank), though there are regulatory restrictions and access criteria that apply.5 Access criteria often restrict direct access for foreign banks. Participation in monetary policy operations may require direct access to an LVPS. At the same time, banks with direct access may still settle payments via other banks due to considerations concerning risk management or operational efficiency. Payment transactions include a sender and receiver bank and, for tiered transactions, an originator bank for sent payments or a beneficiary bank for received payments, as illustrated in Fig.1. The relationship between indirect participants and direct participants is subject to bilateral agreements. The level of tiering differs widely across systems. While there are over 1000 direct participants in TARGET2, there are only around 30 direct participants in the UK’s payment system CHAPS.6 As indirect participants, almost 700 credit institutions from the European Economic Area (EEA) and more than 4000 correspondents worldwide can settle payments via TARGET2.7 The number is quite similar for CHAPS, with roughly 5000 financial institutions being able to settle payments via CHAPS. The ratio of direct to indirect participants is roughly 1:5 for TARGET2 and 1:160 for CHAPS, meaning CHAPS is a much more highly tiered system than TARGET2. The ratio of direct and indirect participants gives an indication of how broadly banks access a system. In addition, the number of direct participants hints at the number of options potentially available to client banks. However, not all direct 5 For an overview of RTGS system features and institutional design see CPSS (2005). 6 See ecb.europa.eu/paym/target/target2 and bankofengland.co.uk/payment-and-settlement/chaps for information and recent numbers. 7 In TARGET2, there are so-called indirect participants and addressable BICs (Bank Identifier Codes). In both cases, banks use a direct participant to connect to TARGET2, but only supervised credit institutions established within the EEA can become indirect participants. In the context of this study, the difference is not relevant, and we refer broadly to indirect participants. 429 1 3 Empirica (2024) 51:425–458 participants offer settlement on behalf of client banks. In TARGET2, out of 1209 participants in the sample, 438 do not send any tiered payments, only 266 settle tiered payments making up more than 1 percent of their traffic, and it is only in the case of 128 participants that tiered payments make up more than 5 percent of their traffic. Smaller participants are much less likely to engage in settlement on behalf of client banks. Therefore, we restrict the sample to larger participants in some specifications for robustness. The concept of tiering employed here refers to volumes and values of payments rather than the number of participants. For a detailed overview of tiered arrangements in TARGET2 against the background of the regulation of systemically important payment systems (SIPS), see Glowka etal. (2022). Tiered payments may be settled internally in the accounts of a settlement bank. These payments do not provide a source of intraday liquidity for the settlement bank or act as a drain on its intraday liquidity as they are not linked to the payment system. However, these internalized payments do have implications for exposures and liquidity positions between settlement and customer banks and thus for potential risks. For banks, outgoing payments settled internally save liquidity compared to payments that are settled via a payment system. Surveys of correspondent banks in the UK have shown that internalized payments make up around one third of interbank payment values [see Adams etal. (2010)]. In the case of TARGET2, it might be assumed the share is lower as the system is less tiered. Levels of tiered participation depend on institutional design and the system’s pricing policy. Depending on what outcome a regulator desires, legal requirements and rules of access may be designed in a way to encourage direct participation. Policy makers and regulators often emphasize the risks of tiered settlement. As described by Finan etal. (2013), the Bank of England persuaded large indirect participants to become direct participants in the UK’s highly tiered CHAPS system on account of financial stability considerations. CHAPS can be considered as an extreme example, with historically few direct participants. However, even in this setting, Benos etal. (2017) find that the effects of the largest indirect participants becoming direct participants (de-tiering) have a low impact on risk measures. For other systems, such as the RTGS system Fedwire in the US, information on tiered payments is not available from transaction data. Thus, the analysis of risk relies on information gathered from other sources. Overall, risks for Fedwire from tiered arrangements are believed to be small and manageable through regular Originator bank Senderbank RTGS system Receiverbank Beneficiary bank Fig. 1 Tiered settlement 430 Empirica (2024) 51:425–458 1 3 Table 1 Benefits and risks of tiered participation from a client bank’s perspective Positive effects of tiering Negative effects of tiering Credit risk Risk exposures against other banks may be more efficiently managed by settlement bank Credit risk against settlement bank. Risk exposures against other banks may accumulate if inefficiently managed by settlement bank Liquidity risk Liquidity may be more efficiently managed by settlement bank. Client bank is charged cost of liquidity and profits Settlement bank may draw on liquidity provided by client banks. Costs of liquidity and operations may be lower than settlement bank fees Operational risk No operational setup costs and fewer own resources dedicated to operations. High operational proficiency via outsourcing to larger players Operational proficiency dependent on settlement bank. Dependency on settlement banks may lead to lock-in effects 431 1 3 Empirica (2024) 51:425–458 reviews and by the mitigation of risks posed by direct participants [see Fedwire Funds Service (2019)]. Tiered participation also reflects the banking system structure and historical developments. For example, the Australian RTGS system previously imposed restrictions on tiered arrangements. These were lifted in 2003, allowing participants whose RTGS payments are less than 0.25 percent of the total value of RTGS payments to settle payments via direct participants. There has been inertia in banks adjusting their access, potentially due to setup costs [see Arculus etal. (2012)]. A variety of factors influence the decision on how to access a payment system. Table1 summarizes the benefits and risks of tiering from a client bank’s risk perspective. Banks balance cost-effectiveness and exposure to risks. Direct participation may entail operational setup costs and investments in liquidity management. Indirect participation may give rise to credit risk, as exposures accumulate during the day against settlement banks. In addition, payment services to client banks may be bundled together with other services, thus making direct participation less attractive. Typically, smaller domestic banks and foreign banks are more likely to become indirect participants. Direct participants offer tiered settlement when profits outweigh the cost of providing settlement services. Direct participants may profit from economies of scale and tiering may help recoup some of the investment cost for operational setup. Banking structures also affect the degree of tiering. For example, head institutions of savings banks and credit cooperatives often provide services including payment settlement to member banks. This not only includes settlement in RTGS systems but also payments settled in internal giro systems. Tiering often leads to uncollateralized credit positions between banks. Rochet and Tirole (1996) study tiered arrangements in the context of interbank monitoring and systemic risk. Kahn and Roberds (2009) discuss the trade-off between widespread access to an LVPS versus the efficiency gains achieved by private monitoring in tiered relationships. Chapman etal. (2013) show that tiered arrangements can arise via two channels. The first is through settlement banks monitoring client banks. Settlement banks leverage private information on creditworthiness by offering different settlement modes. The modes of settlement are similar to system-level differences between deferred net settlement systems and RTGS systems. Tiering represents a balance between deferred settlement, with lower liquidity costs but higher credit risk, and immediate settlement, with high liquidity costs but low or absent credit risk. The second channel is through settlement banks benefiting from economies of scale that reduce overall costs in the system. Given their roles, failures of settlement banks would lead to substantial welfare losses in terms of operational risks and loss of information. From a central bank perspective, monitoring payment system activity is crucial for risk mitigation. A variety of approaches are available to identify different risks. Berndsen and Heijmans (2020) develop a traffic light approach based on different indicators to identify credit, liquidity and operational risk in TARGET2. Triepels etal. (2018) apply an unsupervised learning method to detect anomalies in RTGS systems. Sabetti and Heijmans (2021) apply a similar approach to Canadian LVPS data and discuss how deep-learning methods could be implemented by operators. 432 Empirica (2024) 51:425–458 1 3 Rubio etal. (2020) built on their work to assess deep networks to detect anomalies in the largest systemically important payment system in Ecuador. Aside from anomalies that can relate to different sources of risk, liquidity risk is of particular interest for the smooth functioning of payment systems and financial stability. Heuver and Triepels (2019) apply supervised machine learning in an experimental setting to identify banks encountering liquidity stress. From a settlement bank’s perspective, liquidity needed to fund payments in RTGS systems needs to be obtained from the central bank or the interbank market at a cost. The central bank may also offer overdraft facilities for banks to fund payments. Additionally, payments received allow banks to recycle liquidity from other participants to fund outgoing payments. McAndrews and Rajan (2000) develop a measure to decompose different sources of payment funding and find incoming payments accounting for 25–40 % of liquidity sources during the day in Fedwire. With increases in reserves, the funding of payments shifts to banks using available balances (Garratt etal. 2014). Intraday behavior in RTGS systems is also studied by Bech and Garratt (2003) using a game theory approach. Typically, banks have an incentive to postpone payments when liquidity is costly and they thus delay payments and recycle incoming payments. To account for banks changing behavior during disruption events, rather than assuming a given behavior, Arciero etal. (2009) employ agent-based modeling to simulate payment activity. Liquidity saving mechanisms in RTGS systems can affect banks’ behavior, illustrated by Martin and McAndrews (2008). One example is the use of limits in TARGET2 that allow maximum bilateral or multilateral exposures to be set [see Diehl and Müller (2014)]. More broadly, Alexandrova-Kabadjova etal. (2023) study the determinants of intraday liquidity usage in LVPSs across different countries. Banks relying heavily on incoming payments as a liquidity source can be labeled free-riders. Diehl (2013) provides an overview of different measures and interpretations in the context of free-riding in TARGET2. Heijmans and Heuver (2014) show that banks react dynamically to stress events and some banks delay payment. They find that timing indicators can help in detecting liquidity problems. Abbink etal. (2017) study the effect of disruptions on banks’ reactions in an experimental setting. The path dependency of disruptions may lead to inefficient coordination outcomes at the system level. Concerning market structure, a homogeneous market could relate to a highly tiered system with few active banks. The study finds that a heterogeneous market structure achieves efficient coordination more easily due to a leadership effect. Depending on banks’ use of liquidity, costs incurred by direct participants are passed on to indirect participants. Adams etal. (2010) simulate the emergence of tiered arrangements in a network structure where banks balance the liquidity costs incurred through direct participation and the service fees they pay as client banks. The service fee consists of direct participants’ liquidity costs and profits. The cost of liquidity is found to influence choices regarding system participation. Liquidity pricing is modeled proportionally to liquidity usage or up to a certain amount as free when banks have to post collateral to the central bank for prudential reasons. In such 439 1 3 Empirica (2024) 51:425–458 The respective average receiving time of payments TR is given by: The difference between payments sent and received TD indicates whether payments are recycled or whether individual banks, on average, send out payments before incoming payments arrive. The measure can therefore be interpreted as a proxy for the external delay of payments: TD calculated on a system level would result in a value of zero. However, this does not hold for TD across different payment categories as payments are, for example, tiered on the sending side but not on the receiving side. Timing indicators serve as a proxy for bank behavior, as banks can actively decide on when to send payments of client banks to the system. Assuming there are no structural reasons for timing differences between direct and indirect participants, differences in TD for non-tiered and tiered payments would result from direct participants treating tiered payments differently in terms of timing, for example via internal queue management. Contractual arrangements between direct and indirect participants are unknown. Therefore, postponing settlement of tiered payments may be in line with contractual provisions. A negative value of TD indicates that banks send payments later than they receive them, while a positive value shows that banks send payments earlier than they receive them. Abstracting from potential structural differences, a negative value implies that banks recycle liquidity rather than providing it. If it is assumed that all payment instructions arrive at banks independently, meaning without structural differences in the timing of payments sent and received across categories, the difference in timing would measure external delay. Differences in timing would occur if banks rearranged payments and thus delayed payments outside (external to) the system.16 The actual transmission and obligation to pay is unobserved, as payments show up in the data only upon entering the system. Assuming that payments do not differ structurally in terms of when direct participants receive payment instructions, payment timing can be regarded as a proxy for how participants manage their payments outside the system. Payment timing across different categories of payments can serve as an approximation for the treatment of payments in internal queues. (10) TR b i=∑ n i=1(r b i(t)∗t ) ∑ n i=1 (rb i ) (11) TDb i =TR b i −TS b i 16 It could be the case that tiered payments are sent to settlement banks later in the day. Note that on a system level, the timing of all payments sent and received is equal if all participants are observed. This is not the case for different categories of payments, such as tiered payments. The sending leg and receiving leg of payments may fall into different categories. 440 Empirica (2024) 51:425–458 1 3 4.3 Delay indicator Through delaying payments, direct participants may hold back liquidity and rely on incoming funds for making payments. Delays occur in two ways. First, as described above, participants can externally delay sending payments for settlement in the payment system. Second, within the system, a vdelay can occur between when payments are sent to the system and when they are actually settled in the system. Delays between when direct participants become aware of payment instructions and when payments are sent to the system are only observed indirectly. By contrast, delays within the system can be observed directly. Delays within the system occur when liquidity is not sufficient for settlement and payments are queued. Banks may also use different liquidity saving mechanisms available in TARGET2. One channel is assigning settlement priorities for processing in the system. Banks choose payment priorities ranging from normal to highly urgent, according to which payments are queued in TARGET2.17 In addition, banks can reserve liquidity for highly urgent and urgent payments which is then not available for lower priority payments. Participants may also set bilateral and multilateral limits, thus limiting their net positions vis-à-vis other participants. Following Kaliontzoglou and Müller (2015), we measure the delay in payments by comparing the introduction18 and settlement time in the system relative to the latest possible settlement time. The latest possible settlement time considered here is the close of business. The indicator of delay is stated as: where t1,i is the time during the business day when the payment is available to be settled, t2,i is the actual settlement time of the payment and T is the end of day, i.e. the latest possible settlement time.19 5 Results The results are organized starting with the overall levels of tiering and liquidity consumption. To formally test the effect of tiering on liquidity consumption, we then estimate a panel data model on the settlement bank level. Timing (12) DI b i=∑ n i=1(s b i(t)∗(t2,i−t1,i ) ∑ n i=1 (sb i (t)∗(T−t 1,i ) 17 Given some payment types such as CLS payments have higher priorities but exhibit lower levels of tiering, this could influence results to some degree. 18 Participants can specify the date and time when a payment should be executed. The first attempt for settlement by the system will be made at that point in time. In those cases, we use the time for payment execution rather than when the instruction for later settlement reached the system. 19 Cut-off times differ for different types of payment. For simplicity, we assume the latest cut-off for all payments to be the end of the day. 441 1 3 Empirica (2024) 51:425–458 and delay indicators then identify the channels via which tiering reduces liquidity consumption. For the interpretation of results, the following is implicitly or explicitly assumed: • Tiered and non-tiered payments do not differ structurally in terms of when payment obligations arise and when incoming payments are received by other participants. Without active liquidity management, similar arrival and sending times are expected. This assumption holds if the payment categories do not differ structurally due to their underlying business cases, emergence from activity in different time zones or other considerations by banks. Testing the assumption would require banks’ internal data and business logic. • Banks actively manage liquidity to limit intraday peaks. They can shuffle payments to some degree in order to limit their overall liquidity position across payments from different client banks as well as intragroup and their own payments. • Direct participants have some leeway in when they settle payments. Given internal queuing mechanisms for payment settlement, this assumption holds. However, contractual arrangements may limit leeway. • Resulting from the previous points, payment timing in the system differs largely due to liquidity management rather than different average instruction times across tiered and non-tiered payments. 5.1 Tiering andliquidity consumption The share of tiered payments in total payments lies roughly at between 15 and 25 percent over the observation period (see Fig.2). The number of tiered payments is higher on the sending side. However, in terms of values, the share of tiered payments is similar on the sending and receiving side. This means the average size of payments on the receiving side is larger for tiered payments. At the same time, indirect participants send higher volumes of payments than they receive, which can either indicate that client banks have a greater number of lower denominated payment obligations or that they break up payment obligations into smaller tranches compared to payments received. Overall, the level of tiering in TARGET2 is relatively low compared to other jurisdictions.20 Figure3 shows liquidity consumption based on Eq.(7) calculated separately for tiered and non-tiered payment legs. Directly comparing outcomes in terms of liquidity consumption shows that participants use less liquidity for tiered transactions. However, isolating different categories of payments here does not take into account the overall liquidity position of participants. There might be a bias, as liquidity management may change during the day, depending on a participant’s net overall position. It cannot be ruled out that banks’ own payments are by nature (and not by choice) of higher priority and need to be settled earlier in the day, thereby increasing banks’ liquidity use for their own payments. Settlement banks also have no influence 20 For our subset of the data, tiering levels are higher compared to values on the overall system level. For details on yearly levels of tiering in TARGET2, see the respective Annual Reports on TARGET2. 442 Empirica (2024) 51:425–458 1 3 0.1 .2 .3 .4 Share of tiered payments 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Volume sent Value sent Volume received Value received Fig. 2 Share of tiered payments on system level. Note The share of tiering is calculated using the number and value of tiered payments divided by all payments included in the sample. Tiered arrangements are identified on the sending and receiving side, meaning the same payment can fall into different categories .15 .2 .25 .3 .35 LC 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Overall Tiered payments Fig. 3 Liquidity consumption on system level. Note Moving averages over 30 calendar days. Liquidity consumption is calculated at the system level versus outgoing and incoming tiered payments only. For the indicator of tiered payments, non-tiered payments on the sending or receiving side respectively are ignored. Intraday balances do not reflect actual liquidity positions, but rather the hypothetical scenario in which only tiered payments would be processed 443 1 3 Empirica (2024) 51:425–458 on when they receive payments. Aside from such caution, the consistently lower levels of liquidity consumption for tiered payments indicate that tiered payments leave settlement banks more discretion, enabling them to use less liquidity. 5.2 Model ofliquidity consumption To derive the effect of tiering on settlement banks’ daily liquidity consumption, we estimate a panel data model using bank and time fixed effects. We prefer fixed effects over random effects, as the latter assume the unobserved bank-level effects are uncorrelated with the independent variables. As the level of tiering and size of settlement banks probably factor into the unobserved effects, fixed effects seem more appropriate here. However, the results are robust to employing random effects. Liquidity consumption is calculated daily across direct participants. As the independent variable of interest, the share of tiered payments is included. The share of tiering regards the sending side, as settlement banks can manage outgoing payments but not at what time they receive tiered payments. Table2 reports summary statistics for the variables in the model. The share of tiering lies at 3 percent. This is the average across business days and banks, whereas many banks do not engage in tiering and relatively few large banks settle the majority of tiered payments. The share of tiered payments on a system level in Fig.2 is therefore much higher. We use the log of overall payments sent by direct participants as controls to account for size. Direct participants with more payments should be better able to manage liquidity, as they can smooth their liquidity usage by pooling payments [see Adams etal. (2010)]. Accounting for size makes it possible to abstract from such pooling effects. The average priority of the direct participant’s sent payments controls for the urgency of payments.21 TARGET2 payments have the classifications normal, urgent or highly urgent. The difference in the average timing of payments sent and received proxies the degree of active liquidity management. In addition, we include the concentration of payments sent and received respectively, calculated as the Gini coefficient of payment values. The concentration of payments determines to some extent how granularly participants can manage liquidity. A higher concentration of payments inhibits participants from shuffling payments, as only a few large payments can be rearranged compared to a situation with smaller payments that can allow for more granular liquidity management. As controls for the cost of liquidity and the overall levels of liquidity, the overnight interbank money market rate and overall liquidity22 are included. The money market rate is calculated using an algorithm proposed by Furfine (1999), applied to TARGET2 data following Arciero etal. (2016) and Frutos etal. (2016). We use a modified version of the latter to 21 Generally, priority setting is either an internal queue-management process within banks or observable when banks assign priorities for settlement using the functionality within the payment system, the concept employed here. 22 Calculated as the sum of current account holdings and use of the deposit facility, minus use of the marginal lending facility. 444 Empirica (2024) 51:425–458 1 3 calculate the euro money market rate.23 The algorithm identifies interbank loans by matching payments with plausible repayments the next business day. We estimate the model with data from 2010 to 2019 using fixed effects for direct participants and time effects on a yearly basis to account for changes over time. Changes over time occur as a result of shifts in banking structures or payment processing. Events such as Brexit may trigger changes in how banks access TARGET2, for example by consolidating liquidity management or client banks using a different direct participant to route payments.24 The effect of tiered arrangements may partly be picked up in bank fixed effects. Specifications without fixed effects exhibit higher coefficients and significance levels for tiering and other control variables.25 The estimated model is therefore a conservative estimate of the effects of tiered arrangements. The model for liquidity consumption is stated with the share of tiered payments by settlement bank i on business day b as the independent variable of interest and different control variables in vector X′ ib . Bank-level effects are denoted as 𝛼 and yearly time effects as 𝜋 . (13) LCb i =𝛼 i +𝛽 1 tiering ib +𝛽 2 X ib +𝜋 y +𝜖 it Table 2 Summary statistics The share of tiered payments is calculated as the value of tiered payments sent relative to all payments sent by a participant, the log value sent is the log-transformed value of overall payments sent, the time difference is the difference in average timing between all payments sent and received, the concentration is measured by the Gini coefficient for outgoing and incoming payments, the priority of payments is the average priority of payments (values between 1 and 3 for normal, urgent or highly urgent), the money market rate is expressed as a percentage (calculated via loans identified from TARGET2 data), and logtransformed overall liquidity is measured in millions of euro (ECB data) Variables N Mean SD Min Max Tiering share 1,726,472 0.03 0.11 0.00 1.00 Liquidity consumption 1,726,472 0.43 0.33 0.00 1.00 Cost-based liquidity use 1,726,472 0.00 0.01 − 0.12 0.13 Concentration out 1,726,472 0.79 0.22 0.00 1.00 Concentration in 1,726,472 0.83 0.20 0.00 1.00 Priority of payments 1,726,472 1.63 0.69 1.00 3.00 Log value sent 1,726,472 17.69 3.05 6.91 25.90 Time difference 1,726,472 − 0.15 3.14 − 10.85 10.96 Money market rate 2555 − 0.02 0.42 − 0.54 1.63 Log liquidity 2555 13.43 0.81 11.64 14.54 24 The model is robust to employing time fixed effects on a monthly basis. However, including monthly fixed effects leads to multicollinearity with the prevailing money market rate and overall liquidity. We therefore prefer the yearly fixed effects to allow for the interpretation of the effects of the money market rate and liquidity conditions. 25 Results are available upon request. 23 For a discussion on the measurement of money market rates, see Müller and Paulick (2020). 445 1 3 Empirica (2024) 51:425–458 We estimate the model for the full sample between 2010 and 2019 using fixed effects for direct participants and yearly fixed effects. One issue in the case of TARGET2 is that direct participants with very low payment activity may distort results using relative measures. Small participants may only access TARGET2 for certain types of payments or are simply very small and do not actively engage with the system or play any significant role within the system. We estimate the model for all direct participants, and for sub-samples of direct participants with at least 0.1 percent (128 direct participants) of overall traffic value and a threshold of 0.5 percent (50 direct participants). Results are presented in Table3. In terms of significance and magnitude, the effect of tiering is quite stable within different sub-samples. The results for the sub-samples of participants are more meaningful, as larger settlement banks are more relevant in the context of tiered arrangements and of higher interest due to their importance in the payment system. Including only the largest 50 settlement banks seems most useful to investigate differences for those participants that are most critical to the system and most active in offering tiered arrangements. The share of tiered payments has a negative impact on liquidity consumption in all specifications, meaning a higher share of tiered payments leads to participants using less liquidity relative to their payment obligations. The effect is statistically significant at least on the 10 percent level, and significance increases when only including larger participants. In terms of economic significance, the effect increases as smaller participants are dropped. While the change in one unit of tiering has an effect of roughly 0.05 on liquidity consumption, the effect increases to around 0.21 for large participants. The effect of tiering does not constitute a mere pooling effect, given the control variables and estimation using fixed effects. With regard to liquidity risk, the findings suggest that settlement banks’ liquidity risk decreases as the share of tiering increases. This result holds controlling for other factors relevant to liquidity management and to settlement banks’ business models. Therefore, tiering allows settlement banks to save on liquidity input beyond mere pooling effects. The size of direct participants measured by log value sent leads to increases in liquidity consumption. Larger participants thus appear to provide more liquidity relative to payments to the system, but the effect is not significant when smaller direct participants are dropped. This is counter-intuitive to the hypothesized direction. The fixed effects specification of the model may partly capture the effect of the size of participants, as larger participants take advantage of pooling effects, which could explain this result. Unsurprisingly, the average difference in the timing of payments leads to increases in liquidity consumption and is significant at the 1 percent level in all specifications. Participants sending payments earlier than they receive them, on average, use more liquidity. As expected, a higher concentration of outgoing payments increases liquidity usage, while the opposite is true for the concentration of incoming payments. Highly concentrated sent payments give direct participants less leeway for liquidity management, as few large payments affect intraday balances. For incoming payments, the same reasoning applies, as receiving banks have less leeway in adjusting liquidity management when payments arrive in larger bulks. The 446 Empirica (2024) 51:425–458 1 3 Table 3 Liquidity consumption model (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Liquidity consumption Tiering share − 0.054** − 0.046** − 0.054** − 0.046** − 0.165*** − 0.154*** − 0.165*** − 0.154*** − 0.213*** − 0.209*** − 0.213*** − 0.209*** (0.024) (0.023) (0.024) (0.023) (0.048) (0.050) (0.048) (0.050) (0.057) (0.057) (0.058) (0.057) Log value sent 0.008*** 0.018*** 0.008*** 0.018*** 0.014 0.011 0.015* 0.012 0.003 0.002 0.003 0.002 (0.003) (0.003) (0.003) (0.003) (0.009) (0.008) (0.009) (0.008) (0.012) (0.011) (0.012) (0.011) Time difference 0.053*** 0.055*** 0.053*** 0.055*** 0.057*** 0.058*** 0.057*** 0.058*** 0.051*** 0.051*** 0.051*** 0.051*** (0.001) (0.001) (0.001) (0.001) (0.004) (0.004) (0.004) (0.004) (0.006) (0.005) (0.006) (0.005) Concentration out 0.041*** 0.041*** 0.225*** 0.224*** 0.179 0.180 (0.013) (0.013) (0.083) (0.083) (0.112) (0.112) Concentration in − 0.464*** − 0.464*** − 0.192** − 0.192** − 0.192 − 0.191 (0.018) (0.018) (0.090) (0.090) (0.124) (0.125) Priority 0.014** 0.014** 0.036** 0.036** 0.031 0.031 (0.005) (0.005) (0.014) (0.014) (0.031) (0.031) Money market rate − 0.024*** − 0.023*** − 0.030*** − 0.028*** − 0.023** − 0.021** (0.003) (0.003) (0.006) (0.006) (0.009) (0.009) Log liquidity 0.009*** 0.010*** 0.014*** 0.015*** 0.015** 0.016** (0.002) (0.002) (0.005) (0.004) (0.007) (0.007) Constant 0.278*** 0.424*** 0.176*** 0.310*** − 0.047 − 0.066 − 0.216 − 0.243 0.181 0.156 − 0.003 − 0.038 (0.047) (0.045) (0.056) (0.053) (0.195) (0.195) (0.204) (0.203) (0.269) (0.286) (0.285) (0.309) 447 1 3 Empirica (2024) 51:425–458 *** p< 0.01, ** p< 0.05, * p< 0.1 All the specifications include settlement bank and year fixed effects. Heteroskedasticity robust clustered standard errors (for serial correlation) in parentheses. The main variable of interest is the share of tiered payments, calculated as the value of tiered payments sent relative to all payments sent by a participant. Additional controls include log-transformed overall payments sent, the difference in average timing between all payments sent and received in hours, concentration measured by the Gini coefficient for outgoing and incoming payments, the average priority of payments (values between 1 and 3 for normal, urgent or highly urgent), the money market rate in percent (calculated via loans identified from TARGET2 data), and log-transformed overall liquidity is measured in millions of euro (ECB data) Table 3 (continued) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Observations 1,726,472 1,726,472 1,726,472 1,726,472 268,496 268,496 268,496 268,496 110,830 110,830 110,830 110,830 R-squared 0.231 0.265 0.232 0.265 0.339 0.348 0.340 0.349 0.298 0.304 0.299 0.305 Direct participants 1216 1216 1216 1216 128 128 128 128 50 50 50 50 448 Empirica (2024) 51:425–458 1 3 coefficients are not significant for larger participants. Larger participants here refer to those settling higher payment values. Even when large participants’ payments are highly concentrated, there might still be leeway to rearrange payments. In contrast, a higher concentration for participants with few payments makes granular liquidity management difficult. A higher average priority increases liquidity consumption, although the effects are not significant for larger participants. While higher average priorities lead to payments being settled in a timelier manner and thus act as a drain on liquidity, larger participants may predominately use internal queuing mechanisms rather than priorities within the system. It is expected that overall liquidity will have a positive impact, while the price of liquidity, as measured by the overnight money market rate, will have a negative impact. When money market rates are high, liquidity becomes more expensive for banks, prompting them to exercise greater caution in managing their liquidity. High levels of overall liquidity arguably loosen the liquidity constraints on banks and provide less incentive for active liquidity management. The effects are substantial and significant in all specifications for the money market rate. The effect of overall liquidity is positive and significant in most specifications. The R-squared is lower for specifications including smaller direct participants. A likely explanation is the heterogeneity of direct participants in those specifications. Direct participants with little payment activity and probably little liquidity management likely lead to the lower levels of explained variance. For robustness, we estimate the model with an alternative outcome variable, the cost-based measure of liquidity need cLN: Results in Table4 show a similar picture. The effect of tiering is slightly less consistent and the significance of some control variables changes. The effect of payment concentration becomes less significant and changes direction for payments sent. Meanwhile, the effect of timing differences stays highly significant. The effect of liquidity cost mostly remains negative and that of overall liquidity is positive. However, for the cost-based liquidity need they are not statistically significant. Liquidity conditions and cost may be picked up to some degree by the yearly fixed effects. Notably, the explained variance is lower for the cost-based measure compared to liquidity consumption. 5.3 Timing As one route of explanation for the results on liquidity consumption, timing differences are observed for tiered and non-tiered payments. Payment timing serves as a proxy for external delay, as banks queue payments internally before sending them to the system. The timing indicators from Eq.(11) are calculated for larger participants (0.1 percent threshold) and all other participants. For simplicity, these are called large and small participants, respectively. Types of payments are all payments sent and received, payments on banks’ own behalf, intragroup payments and tiered payments. Figure4a shows timing differences are positive for large participants for their (14) cLNb i =𝛼 i +𝛽1tiering ib +𝛽2X ib +𝜋 y +𝜖 it 455 1 3 Empirica (2024) 51:425–458 As system overseers and operators typically have no access to bank internal contracts and data, our analysis relies on inference and system internal dynamics. Policy makers need to balance efficiency gains and potentially emerging risks. Future research could build on findings here and in the literature to derive welfare effects 0.005 .01 .015 .02 DI 2010 2012 2014 2016 2018 2020 All payments Own payments Intragroup payments Tiered payments Fig. 5 Delay indicator for different payment types. Note Moving averages over 30 calendar days 11.2 1.4 1.6 1.8 Weighted priorities 2010 2012 2014 2016 2018 2020 All payments Own payments Intragroup payments Tiered payments Fig. 6 Value-weighted priorities for different payment types. Note Moving averages over 30 calendar days. Priority categories constructed with range from 1 (normal) to 3 (highly urgent) 456 Empirica (2024) 51:425–458 1 3 of tiered settlement. Arguably, internal processes of banks would need to be better understood to holistically evaluate the risks posed by tiered arrangements. Funding Open Access funding enabled and organized by Projekt DEAL. Declarations Conflict of interest The authors have no conflicting and competing interests to declare that are relevant to this article. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. References Abbink K, Bosman R, Heijmans R, van Winden F (2017) Disruptions in large-value payment systems: an experimental approach. Int J Centr Bank 13(4):63–95 Adams M, Galbiati M, Giansante S (2010) Liquidity costs and tiering in large-value payment systems. Bank of England working papers 399, Bank of England Alexandrova-Kabadjova B, Badev A, Bastos SB, Benos E, Cepeda-Lopéz F, Chapman J, Diehl M, DucaRadu I, Garratt R, Heijmans R, Kosse A, Martin A, Nellen T, Nilsson T, Paulick J, Pustelnikov A, Rivadeneyra F, doCouttoBastos MR, Testi S (2023) Intraday liquidity around the world. BIS Working Papers 1089, Bank for International Settlements Arango CA, Cepeda FH (2017) Non-monotonic tradeoffs of tiering in a large value payment system. J Financ Market Infrastr 6(1):45–78 Arciero L, Biancotti C, D’Aurizio L, Impenna C (2009) Exploring agent-based methods for the analysis of payment systems: a crisis model for StarLogo TNG. J Artific Soc Soc Simul 12(1):686 Arciero L, Heijmans R, Heuver R, Massarenti M, Picillo C, Vacirca F (2016) How to measure the unsecured money market: the Eurosystem’s implementation and validation using TARGET2 data. Int J Centr Bank 12(1):247–280 Arculus R, Hancock J, Moran G (2012) The impact of payment system design on tiering incentives. RBA Research Discussion Papers 2012-06, Reserve Bank of Australia Arjani N, Heijmans R (2020) Is there anybody out there? Detecting operational outages from Large Value Transfer System transaction data. J Financ Market Infrastr 8(4):23–41 Basel Committee on Banking Supervision (2019) Supervisory review process, Liquidity monitoring metrics. Technical Report SRP50, Bank for International Settlements Bech M, Garratt R (2003) The intraday liquidity management game. J Econ Theory 109:198–219 Benos E, Ferrara G, Gurrola Perez P (2017) The impact of de-tiering in the United Kingdom’s largevalue payment system. J Financ Market Infrastr 6(2/3):31–62 Berndsen R, Heijmans R (2020) Near-real-time monitoring in real-time gross settlement systems: a traffic light approach. J Risk 22(3):39–64 Chapman J, Chiu J, Molico M (2013) A model of tiered settlement networks. J Money Credit Bank 45(2–3):327–347 CPSS (2005) New developments in large-value payment systems. CPMI Papers67, Bank for International Settlements, Committee on Payment and Settlement Systems 457 1 3 Empirica (2024) 51:425–458 CPSS-IOSCO (2012) Principles for Financial Market Infrastructures (PFMI). CPMI papers 101, Bank for International Settlements, Committee on Payment and Settlement Systems and International Organization of Securities Commissions Denbee E, Garratt R, Zimmerman P (2014) Variations in liquidity provision in real-time payment systems. Bank of England Working Papers 513, Bank of England Diehl M (2013) Measuring free riding in large-value payment systems: the case of TARGET2. J Financ Market Infrastr 1(3):31–53 Diehl M, Müller A (2014) Analysis of the use and impact of limits. J Financ Market Infrastr 3(1):33–60 Fedwire Funds Service (2019) Disclosure. 23 December, Fedwire Finan K, Lasaosa A, Sunderland J (2013) Tiering in CHAPS. Bank Engl Quarter Bullet 53(4):371–378 Frutos JC, Garcia-de Andoain C, Heider F, Papsdorf P (2016) Stressed interbank markets: evidence from the European financial and sovereign debt crisis. Working Paper Series 1925, European Central Bank Furfine CH (1999) The microstructure of the federal funds market. Financ Mark Instit Instr 8:24–44 Garratt R, Martin A, McAndrews JJ (2014) Turnover in fedwire funds has dropped considerably since the crisis, but It’s Okay. Liberty street economics, Federal Reserve Bank of New York Glowka M, Müller A, Friz LP, Testi S, Valentini M, Vespucci S (2022) TARGET2 analytical tools for regulatory compliance. ECB Occasional Paper 300, European Central Bank Glowka M, Paulick J, Schultze I (2018) The absence of evidence and the evidence of absence: an algorithmic approach for identifying operational outages in TARGET2. J Financ Market Infrastr 6(2/3):63–91 Heijmans R, Heuver R (2014) Is this bank ill? The diagnosis of doctor TARGET2. J Financ Market Infrastr 2(3):3–36 Heuver R, Triepels R (2019) Liquidity stress detection in the European banking sector. DNB Working Papers 642, De Nederlandsche Bank Kahn CM, Roberds W (2009) Payments settlement: tiering in private and public systems. J Money Credit Bank 41(5):855–884 Kaliontzoglou A, Müller A (2015) Implementation aspects of indicators related to payments timing. In: Diehl M, Alexandrova-Kabadjova B, Heuver R, Martínez-Jaramillo S (eds) Anal Econ Financ Market Infrastr. IGI-Global, pp 169–190 Klee E (2010) Operational outages and aggregate uncertainty in the federal funds market. J Bank Financ 34:2386–2402 Lasaosa A, Tudela M (2008) Risks and efficiency gains of a tiered structure in large-value payments: a simulation approach. Bank of England working papers 337, Bank of England Leinonen H, Soramäki K (1999) Optimizing liquidity usage and settlement speed in payment systems. Research Discussion Papers 16/1999, Bank of Finland Martin A, McAndrews J (2008) Liquidity-saving mechanisms. J Monetar Econ 55(3):554–567 Massarenti M, Petriconi S, Lindner J (2012) Intraday patterns and timing of TARGET2 interbank payments. J Financ Market Infrastr 1(2):3–24 McAndrews JJ, Rajan S (2000) The timing and funding of Fedwire funds transfers. Econ Policy Rev July:17–32 Müller A, Paulick J (2020) The devil is in the details, but so is salvation - Different approaches in money market measurement. Discussion Paper 66/2020, Deutsche Bundesbank Rochet J, Tirole J (1996) Interbank lending and systemic risk. J Money Credit Bank 28(4):733–62 Rubio J, Barucca P, Gage G, Arroyoa J, Morales-Resendiz R (2020) Classifying payment patterns with artificial neural networks: an autoencoder approach. Latin Am J Centr Bank 1:1–4 Sabetti L, Heijmans R (2021) Shallow or deep? Detecting anomalous flows in the Canadian Automated Clearing and Settlement System using an autoencoder. Latin Am J Centr Bank 2(2) Triepels R, Daniels H, Heijmans R (2018) Detection and explanation of anomalous payment behaviour in real-time gross settlement systems. In: Hammoudi S, Śmiałek M, Camp O, Filipe J (eds) Enterprise information systems, vol 321. Springer, pp 145–161 Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 458 Empirica (2024) 51:425–458 1 3 Authors and Affiliations JanPaulick1,2 · RonBerndsen3,4,6,7,8· MartinDiehl1· RonaldHeijmans5 * Jan Paulick jan.paulic[email protected] Ron Berndsen [email protected] Martin Diehl mar[email protected] Ronald Heijmans ronald.hei[email protected] 1 Deutsche Bundesbank, Frankfurt, Germany 2 Bank forInternational Settlements, Basel, Switzerland 3 Tilburg School ofEconomics andManagement, Tilburg University, Tilburg, TheNetherlands 4 LCH, 10 Paternoster Square, EC4M7LSLondon, UK 5 De Nederlandsche Bank, Amsterdam, TheNetherlands 6 LCH, 18 Rue du Quatre Septembre, 75002Paris, France 7 Fnality UK Ltd, 41 Luke Street, EC2A4DPLondon, UK 8 Quantoz Payments BV, Europalaan 100, 3526KSUtrecht, TheNetherlands