2012 Statistical Exercise on Matching and Settlement Efficiency

Transcription

1 18 September Statistical Exercise on Matching and Settlement Efficiency EXECUTIVE SUMMARY In 2009, ECSDA developed a common methodology to collect data on matching and settlement efficiency, thereby supporting more transparency and comparability across national markets for monitoring, among others, the level of settlement fails. The statistical exercise carried out in March 2012 shows that: Instructions directly input by the trading platform or CCP into the CSD s settlement system tend to represent a higher proportion of settled transactions than those that are bilaterally input by participants, although there are important differences from one CSD to another. Compared to 2009, the proportion of directly input instructions has increased, reflecting among others the greater importance of CCP clearing in securities markets. The matching rate, defined as the number of instructions already matched on the day before the intended settlement date (ISD), varies from 43% to 95% and even 100%, depending on local market practices. In March 2012, the settlement efficiency rate, measured by the number of transactions which settle by the end of ISD, was 98.9% in value terms and 97.4% in volume terms. This represents an increase of 1.5 to 2 percentage points (in value/volume) compared to 2009 figures. Among the 1.1% of transactions that fail to settle on time (value), most will settle the following day, with a settlement fail rate of less than 0.5% on ISD+1. When a given market wishes to reduce its level of settlement fails, it can use different tools such as regulation, a change of market practices, but also a number of mechanisms at the level of the CSD such as settlement optimisation techniques, incentives for early matching or penalties for late settlement. Going forward, ECSDA hopes that the present report will contribute to inform the debate on the optimal level of matching and settlement efficiency, notably in the context of the TARGET2-Securities project and the upcoming EU regulation on CSDs. ECSDA aisbl (European Central Securities Depositories Association), Rond-Point Robert Schuman 6, bte 5, B-1040 Brussels Phone: Fax: info@ecsda.eu Website:

2 1. Background and methodology In 2009, the European Central Securities Depositories Association (ECSDA) developed a common methodology to collect data on matching and settlement efficiency from its member CSDs. The objective of the exercise was to monitor matching and settlement performance at the European level and to allow for comparisons across national markets. The release of the first statistical exercise in February 2010 (based on November 2009 data) was welcomed by European regulators and market participants as a step towards more transparency and comparability, especially in view of monitoring the level of settlement fails. On 7 March 2012, the European Commission released a Proposal for a regulation on CSDs which includes provisions on the prevention, mitigation and reporting of settlement fails by CSDs. This proposal resulted in heightened interest for a harmonised methodology aimed at monitoring the evolution in the number of settlement fails on a European aggregate basis. Against this background, ECSDA decided to carry out an update of its 2009 survey, this time based on March 2012 data, so that the ongoing political debate on the optimal level of matching and settlement efficiency is informed by more recent figures. The 2012 exercise is based on the same methodology as in 2009 (see Annex 2). It is important to keep in mind that the outcome of the statistical exercise is not an assessment of the efficiency of a given CSD (and that of its settlement engine) but rather a way to look at the efficiency of a market as a whole. Indeed, timely matching and settlement depend in large part on the behaviour of market players, whether or not they are direct participants in the CSD. The March 2012 results cover 20 securities settlement systems (SSSs) in 19 markets 1. While Euroclear UK & Ireland provided data for both the UK and Irish securities markets 2, Iberclear in Spain provided separate figures for its SCLV and CADE platforms, for equities and fixed income listed on Spanish Stock Exchanges, and Public Debt Market and AIAF Fixed Income Market respectively. 2. Matching Early matching is encouraged by regulators and by market standards, such as the ESSF-ECSDA matching standards 3 adopted in 2006, as a way to reduce the likelihood of fails and to facilitate timely settlement. Statistics on matching are thus interesting in order to assess whether the pre-settlement process is efficient in a given market. CSDs however do not always collect matching information (especially if they primarily settle on-exchange transactions), and the data provided in the context of 1 The same 18 markets as in the 2009 statistical exercise plus Slovakia 2 Including Jersey, Guernsey and the Isle of Man 3 See 2

3 this exercise only covers 14 out of 20 SSSs (17 out of 20 for the proportion of directly input versus bilaterally input instructions). In most securities settlement systems (10 out of 17 respondents), the vast majority of instructions is directly input by the exchange, MTF or CCP, thus resulting in a relatively low proportion of transactions that require counterparties to send a matching instruction. Compared to 2009, the proportion of instructions that are directly input increased in almost all cases, reflecting among others the greater importance of CCP clearing in securities markets. In 7 settlement systems (the three ESES CSDs 4, Euroclear Finland, Interbolsa, VPS Norway and CADE, the fixed income platform of Iberclear) a majority of transactions are bilaterally input. One way to measure the level of matching efficiency in a given market is to look at the percentage of instructions matched on or shortly after the trade date (T and T+1), as well as one day prior to the intended settlement date (ISD-1). It should be noted that such matching rates are only calculated for bilaterally input instructions which are not pre-matched by the trading venue or CCP. The data collected by ECSDA reveals important differences in matching rates across national markets. Moreover, there does not seem to be a direct correlation between the matching rate on ISD-1 and the settlement efficiency rate on ISD. In other words, a high rate of matching efficiency does not necessarily translate into a proportionally high rate of settlement efficiency, although we intuitively know that early matching supports timely settlement. Possible explanations for this absence of a direct correlation include: The fact that the matching rate is limited to bilaterally input instructions and thus does not reflect the importance of pre-matched/directly input transactions coming from trading platforms and CCPs; The practice of instructing at the last minute in some markets, for example leading participants to send their matching instructions early on ISD once they are sure that settlement can take place (meaning that instructions are not sent when settlement is expected to fail 5 ). Markets with a small proportion of bilaterally input transactions tend to have a high matching rate, exceeding 89% on ISD-1 (and sometimes equal to 100%). On the opposite, in markets with a more important proportion of bilaterally input instructions, the matching rate is 43% in the ESES markets (on ISD-1), 46% in Portugal, 70% in Spain (CADE Platform), close to 88% in Finland and more than 95% in Norway. For bilaterally input instructions, the matching date is typically agreed by the counterparties and practices differ from one market to another. This explains why it is difficult to define and compare matching efficiency across Europe. 4 Euroclear Belgium, Euroclear France and Euroclear Nederland, the so-called ESES CSDs share a common platform for the settlement of Euronext-zone securities and thus have identical rules for the collection of matching and settlement data. 5 The latter behaviour can distort the reported settlement rate because such virtual fails will not be counted in the settlement fail rates. 3

4 3. Settlement efficiency In March 2012, the average settlement efficiency rate of European CSDs, measured by the percentage of transactions settled in the CSD on the intended settlement date, was 96.7% both in terms of value and in terms of volume. Compared to the last ECSDA statistical exercise in November 2009, this implies a slight decrease in the settlement efficiency rate by 0.7 percentage points in value terms and by around 0.4 percentage points in terms of volume. However, using a weighted average, which takes into account the share of each individual CSD in the total amount of transactions processed, is probably more meaningful than using a simple average across all CSDs (see Annex 2-D for the detailed calculations). According to this measure, the aggregate settlement efficiency rate is in fact significantly higher in terms of value (98.9% instead of 96.7%) and in terms of volume (97.4% instead of 96.7%). The weighted average moreover shows an improvement compared to 2009, with an increase of 1.5 percentage point on average in value terms and close to 2 percentage points in volume terms. Looking at the other side of the coin, this translates into an average settlement fail rate of 1.1% (in terms of value) and 2.6% (in terms of volume). It is also important to note that a large part of the transactions that fail to settle on ISD achieve settlement the following day. The settlement efficiency rate on ISD+1 reaches over 99.5% in value terms and 98.8% in volume terms. 6 At the level of individual markets, the settlement efficiency rate has generally increased or remained stable compared to In some cases, substantial improvements were recorded, such as in Austria and Switzerland, where the settlement efficiency rate in terms of value increased by more than 6 percentage points. Those CSDs that had already attained rates of over 99% in 2009 were in most cases able to maintain this very high level or could even achieve a further slight improvement. The very high settlement efficiency rate observed in some markets (100% in Greece, Romania and Slovenia) can be explained by the following factors: These CSDs mostly settle on-exchange transactions which are pre-matched and directly input in the settlement system in full straight-through-processing (STP) mode; In Slovenia, a buy-in is automatically triggered if an on-exchange trade fails to settle on ISD. OTC transactions entered into the CSD system are either settled on ISD (currently, only sameday settlement is allowed) or cancelled before the end of ISD, hence they are not included in the figures. The number of OTC trades that fail to settle on ISD is negligible (less than 10 trades per year), and the reason for those fails are usually technical and procedural difficulties. In Greece and Romania, buy-ins are also automatically enforced on ISD+1 for exchange trades which are not settled. 6 The settlement efficiency rate on ISD+1 was calculated for the sub-sample of CSDs having provided data on ISD+1 (13 SSSs in value terms, 14 SSSs in volume terms, see Annex 2). 4

5 More generally, CSDs use a number of different tools to promote settlement efficiency and prevent fails, such as credit facilities (whether provided by the CSD itself when it has a banking license or via an agent bank) and a securities lending and borrowing service. For example, it is interesting to note that in the ESES markets, clients using the auto-collateralisation mechanism with the Banque de France have a higher rate of early settlement, particularly in the overnight cycles, than those who do not use or are not able to use the facility, even though by the end of the day the rates are comparable. Unsurprisingly, the settlement efficiency rate tends to be lower in markets with a high proportion of bilaterally input instructions. In a few instances, the settlement efficiency rate declined compared to November 2009 data due to different reasons: In Finland and Portugal, settlement efficiency dropped by 6% and 17% respectively in value terms (the decreases are more limited in volume terms). In the case of Finland, a large part of the drop can be explained by the introduction of CCP clearing; In Hungary and Italy, the exceptional failure of some high-value transactions has affected the overall settlement efficiency rate in terms of value without impacting the rate in terms of volume. Especially in the case of Hungary, a few transactions cancelled on trade date represented a significant share of the total value of the transactions processed in March. The data collected so far does not show a correlation between the use of segregated accounts in some markets (Greece, Finland, Romania, Slovakia, among others) and the matching and settlement efficiency rates, although this aspect was not analysed in detail as part of the statistical exercise Conclusion Settlement efficiency is generally high across all European markets, with a settlement fail rate of around 1.1% in terms of value, and less than 0.5% on the day after the intended settlement date. The differences in settlement fail rates across markets and over time depend on a number of different factors, such as the matching process and whether or not a CCP is used. Settlement instructions are automatically generated by the trading platform or CCP into the settlement system, but in the case of OTC transactions the counterparties to the transaction have to send bilateral matching instructions to the CSD prior to settlement. The likelihood of fails is typically lower in markets where the proportion of directly input instructions outweighs the proportion of bilaterally input instructions. The level of settlement efficiency (and its corollary, the number of settlement fails) is primarily the responsibility of market participants, who are counterparties to the transactions and must ensure that they have the necessary cash or securities in their accounts to be able to deliver on time. Regulators 7 In Romania for example, Depozitarul Central recently extended the use of omnibus accounts for most securities traded on the Romanian capital market and the settlement efficiency rate was not influenced by this change. On the other hand, some CSDs have identified realignment issues with participants who are the greater users of segregation. 5

6 acknowledge that a certain level of fails can be considered as physiological and that safe and efficient financial markets can accommodate a settlement efficiency rate slightly below 100% 8. When a given market wishes to reduce its level of settlement fails, it can use different tools such as regulation, a change of market practices, but also a number of mechanisms at the level of the CSD such as settlement optimisation techniques, incentives for early matching or penalties for late settlement 9. Going forward, it will be interesting to monitor the impact of two major European public initiatives on the level of settlement efficiency, namely: The implementation of TARGET2-Securities, the Eurosystem project aimed at providing CSDs with a single IT platform for the settlement of (mostly euro-denominated) securities; The upcoming adoption of an EU regulation on CSDs and other aspects of securities settlement, which is expected to enforce a two-day settlement cycle for most securities as of January 2015, as well as to establish a harmonised framework for settlement discipline measures aimed at deterring settlement fails. In the context of the upcoming EU regulation on CSDs, it is interesting to note that the European Commission has tasked the European Securities Market Authority (ESMA) with collecting data from national regulators on settlement fails, including the publication of an annual report assessing the level of settlement efficiency for each EU market. We hope that the ECSDA methodology can form a useful basis for the future ESMA reports and we understand that there is an interest from regulators and market participants to gather more detailed information on further aspects of settlement efficiency, including by comparing the differences in fail rates among asset classes and between domestic and cross-border transactions (the difficulty, in the latter case, being to find a relevant definition of what constitutes domestic versus a cross-border transaction, especially in cases where the CSD has no information about the underlying investors). As regards settlement discipline, it should also be noted that penalties charged by CSDs for failed transactions (i.e. transactions not settled by the end of ISD), while already in place in many markets today, should not be designed in such a way that they discourage the practice of early matching. Indeed, instead of sending their matching instructions prior to ISD, in some cases market participants might decide to send instructions for matching only once they are certain that settlement can take place, i.e.as late as possible on ISD, in order to avoid a potential penalty fee. In view of these developments, ECSDA stands ready to pursue the dialogue with market participants and European authorities on the need to monitor and measure the level of settlement fails across the different European markets, as well as on possible ways to improve the settlement efficiency rate, when this is deemed necessary. 8 European Central Bank, Settlement fails - Report on securities settlement systems (SSS) measures to ensure timely settlement, 12 April 2011, available at 9 For further details on the whole array of existing settlement discipline measures, see the ECSDA Reports on market discipline regimes (1 September 2009) and on CSDs' discipline regimes for matching and settlement fails (7 July 2011), both available at 6

7 Annex 1: Results of the March 2012 statistical exercise Matching Settlement AT BE* CH DE DK ES- CADE ES- SLCV FI FR* GR HU IT % bilaterally input n.a n.a n.a % directly input n.a n.a n.a % bilaterally input matched by T n.a n.a n.a % bilaterally input matched by T+1 n.a n.a n.a % bilaterally input matched by ISD-1 n.a n.a n.a % settled on ISD (value, March 2012) % settled on ISD (value, Nov. 2009) % settled on ISD (volume, March 2012) % settled on ISD (volume, Nov. 2009) % settled by ISD+1 (value) n.a n.a n.a. % settled by ISD+1 (volume) n.a n.a Matching Settlement NL* NO PL PT RO SI SK UK&IE Average % Weighted average % % bilaterally input n/a % directly input n/a % bilaterally input matched by T n.a n/a n.a n/a % bilaterally input matched by T n.a. n.a n/a n.a n/a % bilaterally input matched by ISD n.a n/a n.a n/a % settled on ISD (value, March 2012) n.a % settled on ISD (value, Nov. 2009) n.a % settled on ISD (volume, March 2012) % settled on ISD (volume, Nov. 2009) n.a % settled by ISD+1 (value) n.a. n.a n.a ** 99.54** % settled by ISD+1 (volume) n.a. n.a n.a ** 98.81** * ESES CSDs ** calculated for the sub-sample of CSDs that have provided data for ISD+1 7

8 Annex 2: Methodology Most CSDs produce statistics in terms of settlement rates on Intended Settlement Date, but the means by which such statistics are produced are not always the same. Sometimes, there are features of a CSD platform that make such a comparison difficult. Matching rates are not generally reported. The ECSDA methodology thus sets out a framework for determining what transactions should be included in a statistical exercise both in terms of matching and in terms of settlement. The output of the exercise should be a set of statistics from all European CSDs that are produced on the same basis, which would allow regulators and market participants to monitor the level of matching and settlement efficiency on an aggregate level. Where market specific features have an impact on matching or settlement rates, those should be identified. For CSDs that operate different markets on different platforms, a questionnaire for each platform should be completed. For the 2012 exercise, data was collected for the month of March The data should be an aggregated look back over the month, not data collected on a day-by-day basis, to ensure that the widest possible dataset is considered. A. Matching methodology The purpose of the analysis on matching is to see whether non-directly instructed transactions (i.e. those where both sides of a transaction are not received directly from a stock exchange, multilateral trading platform (MTF), central clearing counterparty (CCP) or any other externally provided sources of transactions) are input in a timely fashion. The focus is therefore on bilaterally input instructions by CSD clients that do require matching. Note that such transactions may or may not relate to onexchange or MTF activity, and they could include CCP activity, but as a CSD is typically not aware of the business context of the transaction, it is not usually possible to split out non-exchange activity. As a working assumption, we assume that regulators and market participants are interested to know about the agreement between two parties on the terms of the bilaterally input instruction for greater than T+0 (same day settlement) transactions. We therefore include in the matched transactions figure pre-matched transactions and transactions that are matched but on hold (depending on the type of matching system employed by the CSD), but exclude transactions input for same day settlement. For the purpose of this exercise, an understanding of matching for same day settlement is not relevant this also excludes transactions which, for example, may represent significant volumes of activity in the collateral management services of some CSDs (same day substitutions being a major driver of such activity) which may distort the matching statistics were they to be included. As different markets will have different levels of bilateral versus directly input transactions, the first question to ask is: 8

9 What the proportion is of directly input versus bilaterally input instructions? For some markets, the vast majority of instructions will be directly input by an exchange, MTF or CCP. In other markets, the vast majority will be input bilaterally by participants. In some markets, instructions to settle against a CCP will also require matching, hence should be included in the bilaterally input figures. Having identified as far as possible the complete dataset that the questionnaire is covering, the next set of questions drill down into more detail on the matching rates for bilaterally input transactions. The 3 questions are: % of bilateral instructions pre-matched/matched by end ISD-1; % of bilateral instructions pre-matched/matched by end T; and % of bilateral instructions pre-matched/matched by end T+1; By end, we define this as being the end-of-day for the CSD in question i.e. when the settlement date rolls over. By S we refer to the Intended Settlement Date as agreed as part of the matching process, not the actual (or earliest) settlement date. By T we refer to the Trade Date as agreed as part of the matching process. Not all CSDs use trade date for matching purposes, although date of input could be used as a proxy. It is not essential to provide these two pieces of information, although it would be helpful in the work to determine whether a given market can move to a shorter settlement cycle. Clearly, for those markets already on a T+2 settlement cycle, the figure for T+1 and ISD-1 matching should be the same. By percentage in this instance, we are talking about the number of transactions that require matching we do not require percentages either by value of the transactions, or the number of shares in the transactions. To prevent double-counting of transactions in a cross-border context (where they can be identified) CSDs should only count inbound transactions in their assessment. Where splitting functionality forms part of a CSD s functionality, the original transaction is cancelled and (at least) two new transactions are created. It is ok to count each split (or sibling ) in the overall total. For transactions that are subject to partialling (i.e. the original is not cancelled but partially settled) we do not see any issues in counting these for matching purposes either. Finally, the data should, if possible, be collected on a retrospective basis, including the agreed periods for centrally cancelling unmatched transactions (20 business days as per the ESSF-ECSDA matching standards). Therefore, data for the month of March should be collated 20 business days after the end 9

10 of March, to ensure that all markets are reporting comparable data, and to ensure that late matching transactions are included when calculating the total dataset for comparing ISD-1 matching rates. A worked example of how to perform the retrospective calculation is included in section C below. B. Settlement methodology To have a comparable means of assessing settlement efficiency, we need to look at the fullest possible range of transactions. As this is not an assessment of the efficiency of a settlement engine but a look at the efficiency of the market as a whole, we need to ensure that transactions that have not been submitted for settlement by ISD are, where possible, also included in calculating the settlement rate. This means: Pre-matched or on-hold transactions need to be included, as well as those submitted for settlement; Transactions that either match late or are released for settlement after ISD need to be included in the ISD figure; All transactions, including those for same day settlement (T+0), need to be included this includes both directly input transactions and bilaterally input transactions. The dataset used to assess settlement rates is therefore wider than that used for matching rates, with the inclusion of same day settlement transactions and directly input transactions. Note that this may also include collateral management and certain CCP-related transactions. These transactions have high settlement efficiency and as such, may inflate the settlement efficiency above that which is achieved on bilaterally input and matched transactions. Settlement rates should be calculated both by the number of transactions settled and by value. As with matching rates, settlement performance should have a retrospective view. For ease, we suggest the period of March 2012 plus 20 days, as per the matching analysis, to ensure that late matched transactions are included in the ISD settlement rate. In terms of a distinction between transaction types, we do not expect to see a split between CCP and non-ccp settlement. If the information is available, it may be helpful to see a split between the type of asset (defined in the questionnaire as equities, bonds, funds, other) but again, this is not essential. If it is possible, corporate action related movements should be excluded from the figure (this includes the settlement of market claims, if they occur separately to the transaction). As with matching, to avoid double-counting the settlement of cross-border transactions CSDs should only include inbound links i.e. links where they are the issuer CSD. A worked example of how to perform the retrospective calculation is included in section C below. 10

11 C. Worked examples This section aims to provide examples to show how the data collection is envisaged to be handled. Clearly, CSDs have varying systems in place to manage transactions and different ways of retrieving and analysing data, but the end result (in order to be comparable) should be based on the output of similar procedures. There is an implicit assumption that CSDs will be able to access the required data from their production environments rather than having to utilise their archives. Matching example The data should be based on the following, looking back retrospectively from the agreed calculation date: To get the total data set, all transactions (direct input, bilaterally input, pre-matched, on hold or matched) with an ISD in the month of March 2012 should be counted; The proportion of directly input versus bilaterally input transactions should be calculated (note, this may be difficult for continuous net settlement (CNS) markets for such markets, an approximation should be fine); For the next steps, any same day settlement transactions (where Trade Date = ISD) should be excluded from the count this is the defined dataset for this exercise in matching comparison; For bilaterally input transactions a comparison between the date of matching (or pre-matching) and the Intended Settlement date (as instructed by the counterparties) should be performed; Out of the total number of pre-matched, on hold or matched transactions, the number where the match date is ISD-1 or earlier (to get the total matched by ISD-1, not just on) should be counted and expressed as a percentage of the defined dataset. This should also automatically exclude same day matched instructions. To get figures for matching on T or T+1, the same defined dataset should be used, but clearly the comparison should be between Trade Date and Matching Date, rather than ISD. Settlement example To calculate settlement rates over the period, the entire dataset of transactions should be used as the basis for the calculation direct and bilaterally input, pre-matched and on hold transactions as well as matched instructions. Cancelled instructions (if retained on the CSDs system) can be ignored, but unsettled transactions with an ISD within the period should be included. So, for the period in question, the CSDs system(s) should be queried to identify: All transactions where the actual settlement date (or value date) is the same as theisd; As above, but on ISD+1 (the resulting percentage is cumulative with that on ISD); These numbers should then be expressed as a percentage of the total defined dataset. 11

12 D. Calculation of average rates The 2009 statistical exercise included a simple average for the aggregate settlement efficiency rate, thus which giving the same weight to all markets, irrespective of the value and volume of transactions processed by a given CSD. In order to increase the informative value of these figures, the 2012 report adds a weighted average to the simple average. This weighted average takes into account the different sizes of national markets within Europe. The different weights attributed to each market are calculated based on data collected from ECSDA members in accordance with the ECB bluebook methodology 10. Two different figures serve as basis for calculating the weights. For the average settlement efficiency rate in volume terms, the weights reflect the share of each CSD in the total number of delivery instructions reported in Similarly, in terms of value, the weights represent each CSD s share in the value of the deliveries reported by each CSD in The following table shows the base figures and the resulting weights that were used to calculate the overall weighted averages displayed in Annex 1. In calculating, the figures the following assumption was made: The weight of each CSD in total deliveries is assumed to have remained stable from 2009 to 2012, hence the use of 2010 figures as a basis for the calculation of the weights of both 2009 and ISD ISD+1 (subgroup) Weighted average (volume) Weighted average (value) volume value Volume % of total Value (bn EUR) % of total (thousands) % of total % of total AT 1, % EUR % n.a. n.a. BE 1, % EUR % 0.65% 0.097% CH 57, % EUR 9, % 21.65% 1.940% DE 58, % EUR 68, % 22.02% % DK 11, % EUR 5, % n.a. n.a. ES 19, % EUR 88, % 7.43% % FI 5, % EUR % 1.91% 0.100% FR 26, % EUR 136, % 9.85% % GR 8, % EUR % 3.04% 0.009% HU 3, % EUR 1, % 1.30% 0.343% IT 25, % EUR 77, % 9.77% n.a. NL 4, % EUR 4, % 1.84% 0.946% NO 15, % EUR 2, % n.a. n.a. PL 14, % EUR 4, % n.a. n.a. PT % EUR % 0.36% 0.032% RO 1, % EUR % 0.45% 0.001% SI % EUR % 0.10% 0.005% SK % EUR % n.a. n.a. EUI 52, % EUR 168, % 19.64% % Total 309, % EUR 568, % subgroup 265,251 EUR 478, % % 10 Methodological note available at: 12