Assessing qualitative similarities between financial reporting frameworks using visualization and rules: COREP vs. pillar 3

• Date: 01 January 2019
• Author: Adriano S. Koshiyama,Wenmei Yang
• Published date: 01 January 2019
• DOI: http://doi.org/10.1002/isaf.1441

RESEARCH ARTICLE

Assessing qualitative similarities between financial reporting

frameworks using visualization and rules: COREP vs. pillar 3

Wenmei Yang

1

|Adriano S. Koshiyama

2

1

Large commercial bank supervision

department, China Banking Supervision

Commission, Beijing, China

2

Department of Computer Science, University

College London, London, UK

Correspondence

Adriano S. Koshiyama, Department of

Computer Science, University College London,

Gower Street, London WC1E 6BT, UK.

Email: as.koshiyama@gmail.com

Funding information

Chevening UK; CNPq Brazil

Summary

Financial institutions are struggling with larger volume, more specific and greater fre-

quency of regulatory reporting after the global financial crisis in 2008, especially those

that need to report to multiple jurisdictions. To help to improve reporting efficiency,

this paper aims to assess the existence of similarities between templates related to

credit and counter party credit risk of COREP and Pillar 3 regulatory reporting frame-

works by applying Correspondence Analysis and Association Rules Mining. Our results

suggest a high degree of overlap between these reporting frameworks, more promi-

nently the three business functions as Front office, Finance and Risk. These patterns

can be used as guidance for financial institutions to reshape their reporting architecture.

KEYWORDS

association rules mining, correspondence analysis, regulatory reporting framework, reporting

architecture

1|INTRODUCTION

After the 2008 Global Financial Crisis, financial regulators, interna-

tional institutions and governments around the world proposed

stricter reporting to more effectively monitor and mitigate risks.

New regulations, such as Basel III and Dodd Frank Act and Markets

in Financial Instruments Directive 2 (MiFID2) have, still and will drive

the major changes to regulatory reporting (Covi, 2017; Degryse,

2009; Hortin, 2016; Walker, 2011). Relevant amount of research has

been devoted on evaluating these regulatory frameworks (Acharya &

Ryan, 2016; Ryan, 2017), checking if they enhance financial stability

and the economic impacts of this disclosure. The overall picture is that

the volume, complexity and pace of regulatory reporting for financial

institutions are growing significantly.

For example, financial institutions under the supervision of U.S.

Federal Reserve must submit multiple files including call reports, stress

testing reports, in addition to increasing requirements on data granu-

larity and submission frequency (FSOC, 2014). This trend is indicated

by the document of Basel Committee on Banking Supervision (BCBS)

named Principles for Effective Risk Data Aggregation and Risk

Reporting (BCBS 239) (BCBS, 2016; EBA, 2016). Hence, financial insti-

tutions must fill and submit reports more frequently and with greater

level of detail. When the financial institution is operating globally,

the volume and complexity of reporting requirements increase dra-

matically (Covi, 2017; Leuz & Wysocki, 2016).

In this sense, financial institutions are forced to improve their data

quality and integration across business functions and product lines,

given the short implementation time frames as well as the uncertainty

in rule making (Acharya & Ryan, 2016; BCBS, 2013; BCBS, 2015; Ernst

and Young, 2012). Such challenges tend to increase especially if the

companies deal with each regulation separately (across different

departments, business lines and geographies, i.e. a fragmented

response), instead of addressing common challenges across different

reporting frameworks together (i.e. a harmonized response). It is easy

to perceive that there are similar requirements from various regulatory

frameworks, but very limited research has been devoted to building

tools to quantify and check eventual overlaps. With the knowledge

of these connections or affinities among the regulations, financial

institutions would be able to optimize their business processes, tech-

nology platform and data infrastructure.

In order to measure associations between a set of variables, the

financial literature is populated of research that applies the Principal

Component Analysis (Fontana & Scheicher, 2016; Litterman &

Scheinkman, 1991; Poynter, Winder, & Tai, 2015), mostly because

Received: 3 August 2017 Revised: 27 December 2018 Accepted: 28 December 2018

DOI: 10.1002/isaf.1441

16 © 2019 John Wiley & Sons, Ltd. Intell Sys Acc Fin Mgmt. 2019;26:16–31.wileyonlinelibrary.com/journal/isaf

the researchers are dealing with quantitative/continuous types of var-

iables (such as returns, prices and yields). However, in the case of reg-

ulatory frameworks, most of the data that can be extracted are textual

and categorical, being more conventionally displayed as contingency

tables rather than in a time series. In this case, similar approaches such

as Correspondence Analysis (Beh & Lombardo, 2014) and Association

Rules Mining (Agrawal, Imieliński, & Swami, 1993) are preferred tech-

niques, being both popularly used in other areas, such as psychology,

biology, marketing, etc. (Greenacre & Pardo, 2006; Higuera‐Mendieta,

Cortés‐Corrales, Quintero, & González‐Uribe, 2016; Költringer &

Dickinger, 2015).

Therefore, this work quantifies the similarity between elements of

two different regulatory reporting frameworks: the most up to date

Common Reporting (COREP) issued by the European Bank Authority

and Pillar 3 (BCBS, 2017; EBA, 2017). With this information at hand,

the financial institutions can improve regulatory reporting efficiency

and timeliness as well as reducing compliance costs. By applying the

Correspondence Analysis and Association Rules Mining techniques,

we aim to assess these relationships from different perspectives and

levels of aggregation as well as intend to show that our findings are

robust across methodologies. Our work also related to some previous

contributions in the realm of data visualization and parsing in finance

and business (Fisher, Garnsey, & Hughes, 2016; Goel & Gangolly,

2012; Jaffe, 2015; Kleinknecht & Ng, 2015).

Hence, in terms of contributions, the visualized patterns displayed

in this paper can be used as a guidance for financial institutions to

reshape their reporting architecture. For example, the templates which

are more closed regarding their data source were clustered by our

methods, i.e. three business functions as Front office, Finance and

Risk, were assembled. This result can be used to clear reporting

responsibilities and locate data in different IT systems. Secondly, the

correspondence among templates and a particular data item or group

of data items were uncovered. So, given a data item which is

interested in, the templates that are positively or negatively correlated

with particular data item could be figured out at a glance. All this can

be used to streamline reporting, optimizing the reporting architecture,

and assess points of failure along the process.

In this sense, this work is organized as follows: next section pre-

sents a background review, outlining some fundamentals on the

COREP and Pillar 3 frameworks and describing the two techniques

used in this work: Correspondence Analysis and Association Rules

Mining. The third section exhibits the modelling strategy, showing

how the data was fetched from the regulatory reporting frameworks,

how they were transformed and manipulated, and the procedure pur-

sued to apply both techniques appropriately. Then, we move to the

results and analysis section, starting from the Correspondence Analy-

sis and moving to Association Rules Mining. In both cases, we have

begun by reporting the results from a high‐level of aggregation and

then moving to analyzes at the template level. After suggesting the

main overlaps between COREP and Pillar 3, section 5 closes this work

with some final remarks.

2|BACKGROUND REVIEW

2.1 |Common reporting (COREP)

The European Banking Authority (EBA) published a standardized

reporting framework to address the Capital Requirements Regulation

and Directive (CRR/CRD) reporting. It applies to all credit institutions

and investment firms operating in European Economic Area, and

almost 30 European countries have adopted this reporting framework.

Since the first publication in 2006, the EBA has updated COREP sev-

eral times to the newest version of DPM 2.7 (Data Point Model, see

Table 1) in 2017, which contains 111 templates and covers capital

adequacy and group solvency, credit and counter party credit risk,

TABLE 1 COREP‐templates, timelines and frequencies (EBA, April 2017)

COREP return

category

No of

templates

Template no.

based on

COREP DPM

Reporting

frequency

Submission timing

from reference dates

Capital adequacy 6 C 01.00 to C 05.02 Quarterly 41 days

Group solvency 2 C 06.01 and C 06.02 Quarterly 41 days

Credit, counterparty credit, settlement

and securitization risk

20 C 07.00.a to C 15.00 Quarterly 41 days

Operational risk 5 C 16.00.a to C 17.02 Quarterly and

semi‐annual

41 days

Market risk 7 C 18.00 to C 24.00 Quarterly 41 days

Credit value adjustment risk 1 C 25.00 Quarterly 41 days

Large exposures 6 C 26.00 to C 31.00 Quarterly 41 days

Sovereign exposures 2 C 33.00.a to C 33.00.b Quarterly 41 days

Leverage 8 C 40.00 to C 47.00 Quarterly 41 days

Liquidity coverage ratio 30 C 51.00.a to C 54.00.w,C 72.00.a

to C 76.00.w

Monthly 30 days

Net stable funding ratio 8 C 60.00.a to C61.00.x Quarterly 41 days

Additional liquidity monitoring metrics 16 C 66.01.a to C 71.00.w Monthly 15 working days

YANG AND KOSHIYAMA 17