Classifying Counterparty Sector in EMIR Data

During the Pittsburgh Summit in 2009, G20 leaders agreed to reform the derivatives markets to increase transparency, mitigate systemic risk, and limit market abuse [14]. As a result of this internationally coordinated effort, counterparties trading derivatives in 21 jurisdictions are now required to daily report their transactions to trade repositories (TR) [16]. To accomplish the G20’s reform agenda, the EU introduced in 2012 the European Market Infrastructure Regulation (EMIR, hereafter).

However, the use of these data poses numerous challenges, especially when it comes to data aggregation [15, 16]. To enhance data quality and usability, over the past years public institutions and private entities have jointly worked to harmonize critical data fields [27]. The harmonization effort has focused on key variables, one of which is the legal entity identifier (LEI). The LEI uniquely identifies legally distinct entities that engage in financial transactions based on their domicile.¹ The LEI was introduced in 2012, and currently covers 1.4 million entities in 200 countries. It identifies entities reporting over-the-counter (OTC) derivatives with a coverage close to 100% of the gross notional outstanding, and debt and equity issuers for 78% of the outstanding amount, across all FSB jurisdiction [17]. LEIs are linked to reference data which provide basic information on the legal entity itself, such as the name and address, and its ownership (direct and ultimate parent entities). However, the counterparties’ sector is not included in the reference data. This information is crucial to derive the sectoral risk allocation in this global and diverse market, especially if the aim is to identify potential concentration of risk in specific sectors of the financial system. In EMIR data, even though counterparties are obliged to report their sector using a classification given in the regulation, the available information suffers from several conceptual and data quality limitations. In particular, the sector breakdown is not detailed enough to obtain a comprehensive view of the sectoral allocation of risk. For example, central clearing counterparties (CCPs), which play a key role in the market, are not readily identifiable, as they do not need to report to any sector. To fill this gap, we propose an algorithm to enrich the current classification and uniquely assign a sector to each counterparty trading derivatives, identified by its LEI. We employ a greedy algorithm [7] based on eight different data sources. Firstly we use lists of institutions available from relevant EU public authorities competent for various sectors. Even though comprehensive at EU level, these lists are not sufficient to gain the whole picture because of the global scale of the derivatives market, where many entities outside EU interact with EU investors. Therefore we complement the official lists with sector-specialized commercial data providers. Our work contributes to the existing body of published research dealing with the problem of assigning sectors to individual institutions. In [13] this is done by grouping firms according to their Standard Industrial Classification code in a way to have similar exposure to risk factors within the same group. Despite the popularity of this method in the academic literature, [5] showed that the Global Industry Classifications Standard (GICS) system, jointly developed by Standard & Poor’s and Morgan Stanley Capital International (MSCI), is significantly better at explaining stock return co-movements with respect to [13]. The GICS, however, is not very detailed for the financial sector, so not suitable to fairly describe the derivatives market. More recent works [32] have used deep learning to predict the sector of companies² from the database of business contacts.

The methodology presented in this chapter has a proven track record, as it has been used by several studies. It has been effectively employed to support analysis in the areas of financial stability [19, 12, 23] and monetary policy [6].

Our approach has three main advantages with respect to existing research: it is comprehensive and detailed, flexible, and helps reproducibility and comparability.

We use a multilayered taxonomy to allow a wide range of applications and granularity. The final classification allows classifying entities trading 96% of notional outstanding in the euro area at the end of 2018Q2 and is tailored for the derivatives market, recognizing entities having crucial roles (like market makers, large dealers, and CCPs).

The algorithm is flexible and can easily accommodate future changes in regulation regarding institutional sectors and can be used in other markets.

To this regard, our methodology contributes to the growing body of research using TR data [1, 29, 20, 15, 6, 10] by providing a stable building block to conduct a wide range of analyses. To show this potential, we present four cases studies where we use our classification on the sample of EMIR data available to the ECB.³ In the first case we describe, for the first time to our knowledge, the derivatives portfolios of euro area investment funds, with emphasis on their overall investment strategy. In the second, we disentangle the role of investment and commercial banks in the market. In the third, we measure how large dealers provide liquidity in the Credit Default Swaps (CDS) market. In the last, we show how relying only on the sector reported in EMIR data can lead to a very different picture of the euro area insurance companies activity in the market.

The rest of the chapter is structured as follows: Sect. 2 describes reporting under EMIR, Sect. 3 describes the methodology, Sect. 4 discusses the performance of the algorithm, Sect. 5 includes the four case studies.

EMIR enabled authorities in the EU to improve their oversight of the derivatives market by requiring European counterparties to report their derivatives transactions to TRs.⁴ The reporting obligation applies to both OTC and exchange-traded derivatives in all five main asset classes, i.e., commodity, equity, foreign exchange, credit and interest rate derivatives.

Since 2014 all EU-located entities that enter a derivatives contract must report the details of the contract, within one day from its execution to one of the TRs authorized by the European Securities and Markets Authority (ESMA).⁵ Each opening of a new contract should be reported by the counterparties to the trade repository as a new entry, and all life-cycle events must be reported as well (modification, early termination, compression, and valuation update of contracts). Intragroup transactions are not exempt from the obligation, and trades with nonfinancial counterparties must be reported alike.⁶

The EU implemented the reform with double reporting, i.e., counterparties of the trade have to be compliant in reporting the details of the transaction to one of the trade repositories active in the jurisdiction.

Daily transaction-by-transaction derivatives data are made available by the TRs to over one hundred authorities in the EU, depending on their mandate and jurisdiction. The ECB has access to trades where at least one of the counterparties is located in the euro area, the reference entity is resident in the euro area, to euro-denominated contracts or when the derivatives contract is written on sovereigns domiciled in the euro area: these trades constitute the sample for the implementation of the algorithm presented in this chapter.

With more than 2000 entities reporting every day roughly 30 million outstanding derivatives contracts, with an overall value of slightly less than €300 trillion, EMIR data can be classified as “big data.” On a daily basis, counterparties report roughly 250 fields, of which 85 are subject to mandatory reporting.⁷ These include information on entities involved in the transactions, the characteristics and terms of the contract, which are static and common across asset classes, and the value of the contract, which may change over the life cycle of a trade.

The regulation requires counterparties to report their own sector choosing from a specific list of codes as reported in EMIR.⁸ For nonfinancial corporations, a single letter distinguishes the sector each firm belongs to, while for others the relevant regulation assigns entities to a specific sector (as shown in Table 1).

The existing reporting requirements present five main drawbacks related either to data quality or to the level of granularity:

To overcome the limitations of the sectors available in EMIR, we define a greedy algorithm in order to uniquely identify to which sector each counterparty belongs to. As shown in Fig. 1 the algorithm comprises three parts:

Our algorithm is greedy as the “local” optimal is determined by looking at a single (ordered) source at the time, without considering whether the same LEI appears in another source later in the hierarchy.

In this section we test our algorithm on the ECB’s sample of EMIR data, including outstanding contracts as of 2018Q2, and we demonstrate its added value with respect to the EMIR sector classification, both as reported and processed to avoid ambiguous classification.²⁶

We first show in Table 7 how our sector classification (rows) compares to the sector reported in EMIR data (columns). To this aim, aggregation is based on the sector of the reporting counterparty.²⁷ By increasing the overall granularity from ten to seventeen categories (including subsectors), there is not only a reshuffling among existing categories but also a transition towards other sectors. As expected, the most significant transitions occur towards the sectors of CCP and investment bank, which are known to play a very important role in the market, but do not have a dedicated sector in EMIR classification. 88% of gross notional outstanding which was in the residual group (NULL) is now classified as traded by CCPs.²⁸ Furthermore, 69% and 73% of gross notional traded by credit institutions (CDTI) and investment firms (INVF), respectively, is allocated to investment banks according to our classification.

The sectors of insurance companies, pension funds, and nonfinancial corporations are also deeply affected. Forty-four percent (7%) of gross notional allocated to assurance companies (ASSU) are reclassified as investment funds (nonfinancial corporations) once we apply our classification.²⁹ Only 62% of gross notional outstanding reported by pension funds under EMIR remains as such, while the remaining 23% of gross notional is found to be traded by insurance companies, investment funds, other financial institutions, or nonfinancial corporations.

Our method shows its value also when compared to EMIR data as source for the sector of both counterparties. In this case, aggregation is based on the two sectors, and in order to assign a sector also to the other counterparty, EMIR data needs to be processed to avoid ambiguity.³⁰ Our algorithm reaches a coverage of 96% of notional amount outstanding, for which it successfully classifies both counterparties. For the remaining 4%, entities’ domicile is either located outside EU or not available.³¹ This compares with 80% when using only EMIR data as source, but this figure is inflated by the fact that one CCP is wrongly identified as a credit institution.³²

On top of the improved coverage, the detailed granularity of our classification enhances the understanding of the market structure (see Fig. 2). It allows to recognize that CCPs and investment banks play a key role in the market, being a counterparty in 76% of outstanding trades in terms of gross notional.

Specifically, trades between CCP and investment banks represent 32% notional (blue bubble CCP—Investment Bank in Fig. 2), while 14% is interdealer activity (yellow bubble Investment Bank—Investment Bank). Among CCPs, the volume of notional is concentrated in a few large players, with seven players clearing 98% of the market. The largest player covers 60% of the outstanding notional among cleared contracts, the second 15% and the third 14%, each specialized in some segments of the market: interest rate, equity, and credit derivatives, respectively. Some asset classes are characterized by a monopoly-oriented market in the provision of clearing services, where the first player clears more than 50% of cleared contracts in interest rate, commodity, and equity derivatives. While credit and currency derivatives show a sort of duopoly. Finally, two major European CCPs seem to benefit from economies of scope providing clearing services in the commodity and credit derivatives market, and currency and interest rate derivatives market, respectively. For further details on the CCPs’ business model, and their role in the derivatives market after the reforms, see, e.g., [28, 9, 25, 18].

Commercial banks trade mainly with CCPs and investment banks, with notional amounts of similar magnitude (9% each pair). On the other hand investment banks interact with all the other sectors in the market, owing to their market making and dealer activities. Notably, we find that 7% of notional outstanding is represented by trades between investment funds and investment banks (three red-labeled bubbles at the bottom).

When RIAD, and hence ESA classification, is employed instead of the official lists, results for some sectors change considerably. Most notably, 86% of notional allocated to CCPs according to our classification is allocated to OFIs (S125) with ESA classification. Furthermore, 14% of notional allocated to banks in our classification is allocated as OFI (S125), financial auxiliaries (S126), and captive financial institutions (S127), and 1% is not classified at all. Five percent of notional allocated to the insurance sector is not allocated in ESA while 8% is classified as nonfinancial corporations (S11) or pension funds (S129). Finally, using only ESA classification does not allow to classify 15%, 23%, and 22% of entities classified as nonfinancial corporations, OFI, and pension funds, respectively ,according to our classification.

Overall, the results show several advantages of our sector classification with respect to the reported EMIR sector classification. Firstly, it improves the coverage, allowing for a more comprehensive market description. Secondly, it introduces separate categories for key players in the market, CCPs and investment banks, providing a fairer representation of the market. Lastly, its detailed and multilayered granularity allows to better characterize the market structure.

This section presents four case studies that demonstrate our new classification effectiveness and robustness. At the same time, this section shows the potential of our method as a building block for economic and financial econometric research on the derivatives market. For example, it can be used to investigate market microstructure implications and price formation in these markets, to indicate whether a specific sector would bear more information than others or to study the pricing strategies of derivatives market participants aggregated at the sector level. The application of this algorithm could also be used to deepen the research on monetary economics, e.g., by studying trading strategies on underlyings subject to QE with a breakdown by counterparties’ sector. Finally, thanks to the level of automation the algorithm can support a time series setting and can be used to analyze the number of counterparties active in the euro area derivatives market, with a breakdown of the sector they belong to, or in econometric modeling and forecasting.

In some case studies the enhanced granularity provides further insight on the market or on investors’ behavior, in others, the extended coverage allows for more precise assessment of sectoral exposures. Case study I leverages on the dedicated taxonomy for investment funds, to show how their strategy significantly affects their portfolio allocation in the derivatives market; Case study II shows the role of investment and commercial banks in the euro area derivatives market; Case study III focuses on the euro area sovereign CDS market, showing the liquidity provisioning role of G16 dealers in one of the major intermediated OTC markets; Case study IV compares the derivatives portfolio of insurance companies as reported in EMIR to previous published reports.