• Blog
  • PFRS9 Implementation and Its Impact on Business

Aug 20, 2020

PFRS9 Implementation and Its Impact on Business

PFRS9 (Philippine Financial Reporting Standard 9) 'Financial Instruments' requires a business to categorise the customers to three-stage risk-based classification. PFRS9 implementation requires a quantitative model-based approach for assessing the customer risk and making provisions accordingly. 

However, moving to a completely data-based approach may pose several challenges to the business entities. Some of these challenges include little incentive to identify low-risk segments, inadequacy and reliability of data going into the calculations, etc. At the same time, there are several possibilities of risk assessment going wrong due to underestimation or overestimation of specific data points. 

Introduction to PFRS9

PFRS9 requires a quantitative framework for estimating the Expected Credit Loss on the receivables. The requirements are aligned with IFRS 9 (International Financial Reporting Standard 9), which requires multiple data points viz. 12-month PD (Probability of Default), Lifetime PD, staging of assets, EAD (Exposure at Default), CCF (Credit Conversion Factor), LGD (Loss Given Default), ECL (Expected Credit Loss), etc. 

Basic terms and definitions

Once we know the problem, it is always easier to have a quantitative and qualitative process to measure such a problem, benchmark it and eventually solve it. PFRS9 helps the business entities assess the risk for their clients which can also serve as primary input in pricing the risk to the interest rates being charged from the customers. 

Here are the basic terms and definitions you must know before we talk about risk assessment and pricing our customers rationally:

  1. Probability of Default (PD) – It refers to the probability of the borrower defaulting on the payment obligations. In simple words, it reflects the likelihood of the borrower not being able to repay the due amounts in time. This can be benchmarked with the borrower's credit rating/ credit score, its historical credit profile, etc. 

  2. 12-Month PD – It is the likelihood of the borrower defaulting in the next 12 months. Since the lender has the recent credit profile and financial data, the 12-month PD is easier to be assessed. 

  3. Lifetime PD – It refers to the likelihood of the borrower defaulting across the lifetime of the loan. Since the loan may run through several years with PD being the aggregation of Probability of Default across different years, Lifetime PD will always be higher than 12-month PD. 

  4. Staging of Assets – PFRS9 requires the receivables to be classified into three stages – Stage I, wherein loans do not reflect any inherent credit risk; stage II, where the credit risk has increased compared to the initial risk assessment Stage III, which are credit-impaired assets. The staging is predominantly based upon the DPD (days past due) status of the borrower, which denotes the number of days for which the amounts have remained overdue with the borrower. However, DPD may not be the only criteria. There may be several other indications of deterioration in credit quality, e.g., default with other lenders, decrease in the credit score, consistent high utilization of the credit limits, etc. 

  5. Exposure at Default (EAD) – It reflects the total exposure from the party which may be at the risk of default and includes the current outstanding balance, including interest thereon and any unconditional commitments.

  6. Credit Conversion Factor (CCF) – PFRS9 applies to both on-balance sheet exposures and off-balance sheet exposures. However, for off-balance sheet exposures, a CCF is required to be applied to calculate the exposure amount. The Regulator or the Statutory Bodies may provide specific guidance on such CCF, and in the absence of the same, maybe determined by the entity based on the past experience. 

  7. Loss Given Default (LGD) – It refers to the loss an entity can incur if the borrower defaults. In simple words, it denotes the haircut expected out of the borrower in case of the default.

  8. Expected Credit Loss (ECL) – ECL is the statistical product of PD and LGD and reflecting the provisioning to be made on the borrowers. 

How are the markets reacting to PFRS9?

PFRS9 has brought out a forward-looking approach for estimating credit risk and related provisioning. Here is how the industry is reacting and adapting to PFRS9:

  1. Banks seem to have understood the benefits of using data-driven statistics, whereby they are also trying out different approaches to eliminate subjectivity in the process.

  2. Data-driven decisions into areas like staging, Lifetime PD, etc., are evolving with time. Being early stages of PFRS9 implementation, the availability of more data points over time will help polish the existing approaches towards improving the overall credit model.

  3. Most of the implementations are very manual, with processes run monthly. Hence, the operational efforts in creating and presenting the numbers are very high. This should gradually move towards automation. 

Implementing a data-driven provisioning through ECL model is considered to be better when compared to a simple regulatory ratio approach. This is because entities need to make provisions based on estimated losses in future, thus being more realistic and accurate. In contrast, regulatory ratio approach tends to be more prudent, which may require higher provisions with conservative approach. A higher provisioning also means higher capital requirement, which may be much more expensive for the financial entities in general. 

Challenges in implementing PFRS9

Entities are experiencing several challenges in implementing PFRS9; some of the major ones are as below:

  1. While an entity may be inclined towards low-risk segments, there is very little incentive in terms of credit risk provisioning due to the minimum ECL requirement of 1%. 

  2. The use of individual data points requires regular backtesting of the credit model to ensure its reliability. 

  3. Multiple approaches are currently in vogue for calculating Lifetime PD, viz. Markov and vintage rates, etc. There may not be a single approach that fits all situations appropriately.

  4. Staging criteria must evolve beyond the 'dpd' criteria and capture the increase in risk assessment even when there are no overdue amounts. 

Special considerations for Covid-19 risk assessment

Covid-19 pandemic has brought out some specific risk areas, which may be required to be built-in within the model, but at the same time, are also required to be made exclusive to the existing model. This will be helpful to get back to the core credit modeling once the Covid-19 impact fades out. Here is a brief snapshot of how one should approach PFRS9 for Covid-19 risk assessment:

  1. No change in core logic should be undertaken. Build the strategy as one would have, assuming Covid does not exist.

  2. Create a Covid strategy on the top of the core logic – and keep them modular – so that from an implementation standpoint – you can quickly go back to your primary strategy.

  3. Potential modular Covid strategy may include evaluating moving the moratorium customers to Stage 2, considering the impact on amortization rates of loans due to moratorium and restructuring, LGD to be stressed to worst levels, calculation of scenario weighted ECL, etc.

What may go wrong in data-driven ECL implementation?

Like any other statistical model, the outcome of the credit risk model may differ significantly if the inputs to the data decisions are not correct. This may eventually lead to practical issues in underwriting and customer management, as under-estimation or over-estimation of the risks due to different data points may eventually lead to a variable risk perception towards several borrowers. 

Here are the key risk areas in the process:

  1. Use of wrong application score may lead to estimating higher/ lower risk of default, potentially leading to low-risk customers moving to Stage 2/3 or high-risk customers continued in Stage 1.

  2. A high cut-off towards credit risk assessment may lead to incorrect decisions regarding the portfolio mix.

  3. Pricing and assigning limits to the borrowers may be required to be risk-based instead of setting fixed rates/ limits, leading to an imbalance in the risk-reward trade-offs.

  4. Incorrect estimation of the behavioural score/early warning signals may lead to incorrect impact on provisioning and staging, estimating macroeconomic overlays and associated factors.  

Considering the importance of precise data points in the credit risk assessment framework, an entity must have a robust mechanism for data validation and unbiased and independent process implementation. 

How can CRIF help you in bridging the gap?

Owing to its industry experience and availability of large customer data with several variables, CRIF can help you bridge the gap between the existing credit model and PFRS9 requirements and build a robust credit risk assessment framework. Our key services range from assisting in scorecard development, validation of data sampling, monitoring data processes and implementing the process to automate credit risk grading. 

Here are some of the specific areas wherein CRIF can assist you:

  1. Getting an independent assessment of underwriting, collection, or PFRS9 process/ results and identifying major/minor gaps against industry best practice/benchmark.

  2. Leveraging CRIF experience with similar projects in many financial institutions and banks in Europe and other countries, including detailed analysis of several different angles of your processes.

  3. Providing clear recommendations on the initiatives to be implemented to close gaps against the leading practice and best planning to optimize ROI.

  4. Implementation of the best-in-industry practices to assess and benchmark the credit risk assessment, leading to better asset quality and consequently better returns on investment. 

    To learn more about how we can help you bridge the gap to build a robust credit risk assessment framework, contact us to start a conversation

    Get In Touch