Insights

Challenges of IFRS 9 modelling in the UAE banking industry

ME PoV Spring 2020 issue

This year marks the 12th anniversary of the 2008 global financial crisis. It also marks the second year since the adoption of IFRS 9—a standard that has contributed to the improvement of the mechanisms of classification and measurement of financial instruments deemed as one of the main causes triggering the aforementioned financial crisis. Prior to the effective date of IFRS 9 (1 January 2018), the International Accounting Standards Board (IASB) undertook a substantial number of activities to support the implementation of the standard especially in relation to the Expected Credit Losses (ECL) model that requires a high degree of management estimates and judgements. The implementation was complemented with a pool of unexpected challenges to all adopters, and to the banks and financial institutions in particular though the initial impact of IFRS 9 on the banks’ financial results and regulatory capital resources has not been as severe as the market had  initially expected. In this article, we will cover the major challenges of IFRS 9 model developments, mainly around data availability and ECL computation for the UAE banking sector.

The struggles and challenges emanating from IFRS 9 reside not only in the lack of data availability, experience and available resources but also in the lack of clarity from regulatory expectations. The impact of IFRS 9 implementation has gone beyond a simple update of accounting policies.
It has impacted governance, the risk and finance functions, internal controls, information systems, regulatory reporting, disclosures and, in many cases, the underlying business models and strategies of banks themselves. Below are the major challenges that banks have come across in the first year of implementation and are still in the process of overcoming:

  Data: 

 ECL computation:
  • Unavailability of historical data:
    • Many banks still process customer information manually using Excel workbooks. Therefore, the extraction of original data is a tedious task where the original ratings of the customers (at the time of initiation) are captured. A comprehensive track of customer ratings is required to assess whether there is a ‘significant increase in credit risk (SICR)’ which moves the customer account from stage 1 to stage 2, and default from stage 2 to stage 3.
    • Due to the unavailability of a comprehensive record of the borrowers, banks have faced major challenges in generating recovery rates, cost of recovery, etc. which are essential components for the determination of ECLs.
  • Data quality: this issue is a result of the use of simple manual systems to capture client data.
  • Underlying models such as rating models, behavioral scorecards which have been used as inputs for ECL estimation have not been validated.
  • Unavailability of data/models to derive quantitative parameters like behavioral scorecards etc. to be used for SICR criteria and to be used for the curing of accounts criteria as well as the requirement for use of 5 to 10 years of historical data as per CBUAE requirements.
  • Vendor-based models relying on global data have the following issues:
    • Challenge in bridging global models to regional macroeconomics data for probability of default (PD) and  loss given default (LGD).
    • Limited control over model specifications, especially for LGD, where this percentage needs to be ‘unique’ to the bank and represent the recovery patterns of the bank itself rather than use of global data or regulatory estimates or existing vendor models.
  • Accuracy of inputs in models, especially which banks have relied on using Basel guidelines to develop their LGD models, whereas this has caused limitations in  the model to capture the expected recoveries from the account once the account has defaulted.
  • Justification of use of certain forward looking macroeconomic factors, and the determination of the use of the appropriate factors.
  • Availability of macro-economic forecasts, and use of the same/similar forecasts in setting the bank’s strategy as well as computing ECL.
  • Determination of appropriateness of overrides to certain accounts in the model—and the governance around the same
  • Estimation of exposure at default (EAD) for non-fund-based products and consideration of prepayment model.
  • Determination of lifetime of revolving products is a major challenge banks have faced and is still facing, in which limitation in data availability and quality has led to inconsistent behavior studies.

Whereas the above challenges do not cover areas outside of data and ECL computation, their extent varies from one bank to another, depending on the sophistication of the model and the tolerance and appetite of senior management to invest in more advanced technology-based models.

How can banks overcome these challenges going forward?

The banks were required to make many judgments in constructing models to comply with IFRS 9 impairment requirements. Differing approaches for certain key judgements may result in IFRS 9 impairment provisions behaving inconsistently, particularly during future periods of stress.

  Data: 

 ECL computation:

Banks need to:

  • Maintain historical data of customer ratings and move away from manual Excel data to information system-based data where data is more easily maintained and extracted.
  • Consider regulatory requirements (including curing and staging) when capturing customer-related information in order to have this data available for use in the model.
  • Develop an appropriate and robust mechanism for validation of rating models and behavioral score models, as well as quality of data used in the modelling.
  • Decrease reliance on third-party vendors and develop models specific to the bank’s portfolio.
  • Institute and implement a model risk management framework to manage the risks arising from use of internal models for ECL Computation.
  • Banks should develop a process to validate data used in the models using internal data, also to ensure appropriate transfer of knowledge from the vendor in order to refine the model and tailor the inputs and processes in the models to reflect the uniqueness of the banks’ data and to capture regional macroeconomic factors.
  • Banks may use new vintages to reach the minimum threshold required for the development of PD and LGD models.
  • Overrides in models need to be appropriately justified and documented as part of the ECL model and governance.
  • For non-fund-based products banks should collate historical patterns and customer behaviors to determine more accurately the EAD.
  • Banks should consider updating macroeconomic factors on period basis to reflect the point in time.
  • The Bank may consider developing LGD models using the Best Estimate of Expected Loss prescribed by Basel within IRB guidelines or develop a model using a workout approach for LGD analysis incorporating estimates on collateral recovery, cure and restructured nodes. 
Conclusion

ECL provisioning will have a direct, quantifiable impact on the financial performance of the banks and financial institutions and an indirect qualitative impact on a wide range of factors contributing to shareholder value. Going forward, any increase in the regulatory view of IFRS 9 impairment provisions is likely to have a detrimental effect on regulatory ratios and vice versa, any reduction is likely to benefit capital and solvency ratios. To align the expectations of the regulator being a key area of focus for banks in the UAE, overcoming the challenges faced by banks after the first year of adoption of IFRS 9 is a top priority in the coming years, with data availability and ECL computation being the most significant areas to tackle and improve.

by Firas Anabtawi, Partner and Marcelle Hazboun, Senior Manager, Audit & Assurance, Deloitte Middle East

Did you find this useful?