Data Quality & Data Standards

Databases:

For details and samples of Global Credit Data databases please see the Analysis page here

Data Quality and Completeness

Data quality comes by having a logical database structure, well described definitions of data-fields, good instructions to the users, rigorous validations rules and active use and feed back from its contributors.

Global Credit Data works to continuously improve the quality of data in terms of consistency and comparability, completeness and quantity. The efforts toward constant data quality improvement are carried out in the following ways:

1. Consistency: all processes are governed by the Methodology Committee

2. Validation: multiple controls are run at portal level in the data submission process ensuring that any incorrect data is rejected and that fields are complete to a satisfactory level

3. In-cycle audit: all data submitted by each bank is scrutinised by Global Credit Data's expert data executives to ensure that it not only passes the mechanical validations but also makes sense, both for individual borrowers and loans and on a portfolio basis for that bank

4. Documentation: a suite of supporting documents belonging to Global Credit Data is available to all members (see standards).  The data dictionary and worked examples are constantly being reviewed and augmented to allow banks to consistently enter data of high comparability to other banks

5. Methodology Committee: A group of experts from member banks are in control of the data process, the documentation and the direction of research through working groups etc.

6. Scoring system: after each data submission, member banks' data sets as updated are scored, allowing the banks to compare their data completeness with their peers

7. Annual audit: audit report to each Member-bank on selected key criteria, including a dialogue around qualitative questions and increasing the completeness of data fields

8. Semi annual General Meeting: opportunity to share best practice in credit modelling and on how to use the database.  Global Credit Data data is analysed by executives and results presented to members for discussion

Data documentation

The LGD and EAD parameters are most demanding in terms of multiple and precise data on the obligor, the loan and surrounding circumstances. Defining the data-fields has required many interactions between the Member-banks, the Methodology Committee and the Data Agent. This work has been consigned in 3 documents described here below. At the current stage, Global Credit Data offers a mature structure for LGD/EAD pooling, which the Member-banks can and do adopt in their internal databases. Global Credit Data is actually setting up standards: these are extensively elaborated in a suite of supporting documents.

Of course, this implies that, going forward, changes in the template will be limited and announced well in advance in order to give Member-banks the possibility to adjust their systems accordingly.

Supporting Documents Suite:

Global Credit Data EAD & LGD Input Structure

This document is the elaboration of data fields necessary to model EAD and (workout) LGD. It is the result of the permanent co-operation organised by the MethCom to transform the dialogue into clear words and facts.

It is fundamental to enable a new Member-bank to organize its internal collection process in accordance with the specifications of Global Credit Data. The Data Input Structure document specifies and defines all data fields and shows all possible answers (the lookup tables). Each data-field is marked whether it is optional or mandatory. The document also includes short comments and references to the business rules (see DMUG).

The data model itself consists of 8 tables. For each reported defaulted loan, the tables can be repeated, when relevant, to reflect the evolution through time, allowing 5 different “snapshots” at “Origination”, “one year prior to Default”, “date of the Default”, “post Default” and, finally, at “Resolution”. The table Nr 8, “the Loan Transaction table”- is like a statement of account reflecting the entries in and out (recovery cash flow, write-off, etc) in-between the transfer of the loan in default and its resolution.

Global Credit Data EAD & LGD Data Model User Guide (DMUG)

Bank loans know many appearances which could be recorded in many ways in the data base. For sake of consistency and clearness, additional guidance is provided to the Member-banks how to interpret certain aspects of a defaulted loan and how to record them.

For instance it explains how to treat a loan being a contingent liability at the time of default (i.e. a bank guarantee) and where a claim is paid out during the workout process. In this example it is important to know what to label as a cash flow and what as a drawn position at the moment of default in order to have a proper calculation of the LGD.

The Business Rules reflect the consensus obtained in the Methodology Committee on various Data Issues, as and when they come. The DMUG records them all, thereby allowing each member to know when a new business rule is applied in the data template and what the status is of outstanding issues.

Global Credit Data EAD & LGD Data Validation Specifications

When data is submitted by the Member-bank to the Data Agent, each defaulted facility is subjected to a set of more than 300 validation rules. Based on these rules – set by the Methodology Committee - the Data Agent checks the data in terms of presence, consistency, integrity and alignment with the database structure. If the rules are not met, the files will be rejected with an explanation enabling the Member-bank to do the necessary adjustments. All these rules are listed in the Global Credit Data EAD & LGD Data Validation Specifications document.

Scoring System

In order to encourage the Member-banks to improve the data quality, Global Credit Data has created a scoring system. Based on 13 criteria – defined by the methodology criteria – the quality of each individual delivery is measured, i.e. expressed in percentages of fields filled in etc. Each Member-bank can see how its delivery compares to its peers’. The scoring is carried out semi-annually and reported to the Member-banks (on an anonymous basis). The content of the scoring is by nature evolutionary, reflecting how more or less essential each data is seen by the MethCom.

Annual Audit

Each Member-bank receives an annual visit (real or virtual) by a Global Credit Data executive to discuss:

· Improvement in data field quality

· Improvement of coverage of data fields (completeness)

· Improvement of the coverage of all realized defaults.

This visit results in recommendations by Global Credit Data how the Member-bank can improve its quality and an agreement between the Member-bank and Global Credit Data on what will be done and when. The executive involved will keep a list on where Global Credit Data stands on each Member-bank and for the Global Credit Data as a whole.

Semi Annual General Meeting

Every year, in December and in June, the updated data-base – after the Spring and Fall data submissions- is presented to the Member-banks at the General Meeting. On this occasion, the Methodology Committee will present the progress of the last 6 months and inform the Member-banks about new requirements and the progress of the various working groups under its supervision. In this meeting, the Member-banks are invited to show their own analyses on the database and what they have done with it. It makes this meeting an important date for the Member-banks, as they share knowledge, encouragement to improve data quality, and eventually their mutual trust in the dataset which is their common good.

Data Collection Process and Feed back

Two full data submissions and releases of the global data-set take place in May-June and in October-December, each year.

In 2008, the database was substantially enriched, whilst the quality of data was scrutinised in terms of completeness and consistency. A dialogue has been initiated between the Member-banks on “Best Practice Sharing”, i.e. how to best exploit the data output.