ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

How Should Banks Estimate Their Expected Loan Loss Provisions to Survive in Difficult Times?

This article explains how banks can use their forward-looking internal credit risk estimates and apply on loan cash fl ows over different time horizons and assess the impact on loss provisions. Such an estimate based on longer historical data will enable the banks to better foresee the uncertainty pertaining to repayment status of their loans and make loss provisions in a more proactive manner.

As part of its continuous efforts to improve financial stability as well as to ensure uniform practices of accounting globally, the International Accounting Standards Board (IASB) and Financial Accounting Standards Board (FASB) issued a guidance on how to recognise and measure financial instruments. Required in more than 100 countries, all financial entities must incorporate a new expected credit loss model and adhere to new accounting standards within the next few years. The new acc­ounting standards aim to simplify and strengthen risk measurement and the reporting of financial instruments in an efficient and forward-looking manner. To address the “too little, too late” problem arising from the incurred loss model, the new accounting standards necessitate a “forward-looking” impairment model for the estimation of loss provision by commercial banks.

This forward-looking approach requires banks to update and recognise expected credit loss (ECL) for financial assets from the initial acquisition or origination date. In the view of rising stress on banks due to poor-quality loans and falling profitability, the Reserve Bank of India (RBI) has postponed the adoption of the standards by Indian banks. However, sooner or later the banks have to mandatorily adopt this standard to better deal with the uncertainty. Banks that have better data management systems and a prudent risk internal risk culture have been using these three risk drivers as per internal estimation models to measure loss provisions as well as capital requirement for credit risk.

The purpose of this article is to explain how ECL-based provisioning would benefit the banks to manage the impairment of assets in this difficult time posed by the COVID-19 pandemic situation. A longer historical data on delinquency rates can be used for this purpose. Since the loan assets on initial recognition have to be classified into three stages depending on their credit-quality maintaining data as per days past due buckets would facilitate better estimations and forecasting of probability of default (PD) of a loan for each stage. This will enable the banks to accurately estimate their future credit provisioning requirements while conducting credit planning.

Impairment Model

The International Financial Reporting Standards (IFRS) 9 introduces a new imp­airment model that requires the recognition of expected credit losses on all fin­ancial assets at an amortised cost or at a fair value through other comprehensive inc­ome (other than equity instruments), lease receivables, certain loan commitments, and financial guarantee contracts. The expected credit loss must also consider forward-looking information to recognise impairment allowan­ces earlier in the lifecycle of a product.

The IFRS 9 recommends a three-stage app­roach to impairment as follows:

Stage 1—the recognition of a 12-month ECL, that is, the portion of lifetime ECLs from default events that are expected within 12 months of the reporting date, if credit risk has not increased significantly since initial recognition;

Stage 2—lifetime ECLs for financial instruments for which cre­dit risk has increased significantly since initial recognition; and

Stage 3—lifetime ECLs for financial instruments that are credit impaired.



Twelve-month ECLs are the portion of the lifetime ECLs that results from default events that are possible within the next 12 months weighted by the probability of that default occurring. At initial recognition, a 12-month ECL is provided for. At the next reporting date, ECL will be classified into the lifetime category, that is, the losses that might ­occur in the whole life of the asset if there is a significant increase in the credit risk of the account. Finally, if the account becomes impaired or non-performing, it is categorised as stage 3 and lifetime ECL is calculated for it.

The IFRS 9 defines lifetime ECLs as those that result from all possible default events over the expected life of a financial instrument (that is, an entity needs to estimate the risk of a default occurring on the financial instrument during its expected life). They would be estimated based on the present value of all cash shortfalls over the remaining expe­cted life of the financial asset, that is, the difference between the contractual cash flows that are due to an entity ­under the contract and the cash flows that the holder expects to receive.

Significant increase in credit risk (SICR) is another concept introduced in the standards. This is the basis on which accounts are classified into either of the three stages. SICR is not defined by the standard; banks will have to use their own knowledge and experience in judging what will be a significant increase in credit risk. Otherwise, they can apply the RBI’s backstop measure that uses 60 days past due to non-retail exposures and 30 days past due for retail exposure on the reporting date for SICR.

It can be estimated in many ways: inc­rease in cumulative probability of default above a threshold, movement of rating from super grade low investment grade, worsening of rating or behavioural score changes can be considered as triggers. For example, statistically significant cha­nge in PD, benchmark relative change in PD (say a 20% ­increase), etc.

Expected Credit Loss Estimation

The ECL calculation method for both stages 1 and 2 acco­unts is based on derived PD, loss given default (LGD), and exposure at default (EAD).

Expected credit loss at the borrower level is calculated by using the following ­formula:

ELi = PDi × LGDi × EADi … (1)

where EL is expected loss. The term PD is the probability of default over the lifetime of the asset (cumulative probability of default) if lifetime ECL is calculated. For the 12-month ECL, PD will the proba­bility that the account defaults within next 12 months. EAD is the adjusted exp­osure at default. It shows the total amount that the bank has lent to the borrower adjusted for some haircut. LGD is the loss given default of the asset, which is the amount the bank might not recover out of the entire exposure.

Each entity should define their own definition of default (for example, 90 days past due), which should be consistent with the definition used for internal credit risk management purpose.

Table 1 presents an exa­mple of the repayment schedule of a term loan with the corresponding conditional PDs, the 12-month and lifetime ECLs.

Effective interest rate (EIR) is the rate that exactly discounts estimated future cash payments or receipts through the expected life, or when appropriate, a shorter period of the financial asset or financial liability to the gross carrying amount of a financial asset or to the amor­tised cost of a financial liability. The calculation of EIR includes all fees, transaction costs, and all other premiums or discounts that are directly rela­ted to the acquisition of financial assets on the book of an entity.

The cumulative PDs are obtained from marginal probability of default (MPDs) that are derived from a yearly rating migration study (transition matrix). In this ­exercise, this has been estimated using historical data through a static cohort analysis, which is followed by leading credit rating agencies (CRAs).

Table 2 provides estimates for conditional PD, which has been used in estimation of ECL using EAD cash flows reported in Table 1 Panel C.

The conditional PD for each period is conditioned on the fact that the borrower has not defaulted in the previous period and defaults in the current period. Thus, in the second year, conditional marginal PD is estimated as:

=1.59% × (1-1.27%) = 1.57%.

Similarly, in the third year, conditional marginal PD for the same BBB borrower would be: = 2.76% × (1-1.59%) = 2.72%.

The survival benefit is passed to the borrower to derive credit risk for each yearly cash flow.

The cumulative PD is estimated from historically obtained marginal PDs (MPDs) using the following expression:

CPD(n)=d (1)+d(2)×(1-d(1))+d(3)
×(1-d(1))×(1-d(2))+….+d(n)(1-d(n)) … (2)

For example, cumulative probability of default in the second year (CPD2) is estimated as:

= 1.27%+(1-1.27%)*1.59%=2.84%

In the same manner, CPD in the third year (CPD3) can be obtained by using ­the expression:


Similarly, CPD in the fourth year (CPD4) is estimated as:

1.27% =6.72%

In this case, we have used equation 2 which is used by CRAs.

It is interesting to note that in many default study reports, CRAs publish their cumulative PDs for different time horizons. From these CPDs, one can obtain forward-looking conditional marginal PDs directly using the following expression.

Conditional marginal PDn
= 1-{(1-CPDn)/(1-CPDn-1)} … (3)

where n is the number of years.

This forward-looking PD is a pure historical view. However, one can adjust macroeconomic factors through regressions or by using the Z-index approach to link PD with macroeconomic scenarios. It is important to note that, alternatively, one can use the matrix multiplication method to derive cumulative PDs and conditional PDs for the estimation of ECL.

In the Ind AS 109 prescribed methodology for ECL computation, the 12-month ECL for stage 1 accounts multiplies the estimated EAD with the one-year PD and LGD. We have assumed 40% LGD for the loan pool. The estimated EAD is the sum of the principal balance outstanding and interest accrued if the customer defaults on that loan account at the end of the year. This has been shown in Table 1 Panel C.

The lifetime ECL for stage 2 accounts first estimates the expected loss on the account for each future year of potential default during the loan lifetime by multiplying each future year’s EAD with the one-year marginal conditional PD and LGD. Each future year’s EAD is estimated as the sum of principal balance outstanding and interest accrued if the customer defaults on the loan in the middle of that year. Each future year’s expected loss is then discounted using the EIR.

Ultimately, we obtain lifetime ECL (for stage 2) using conditional PDs estimated as $24.15. For stage 1, one-year ECL is estimated as $3.56. This has been further summarised in Table 3.

In Table 3, it is important to note that the discounting for ECL based on amortising EAD is crucial in the analysis of loan loss provisioning, which should also be applied for the exposure of each borrower, especially those in stage 2. This will lead to a redu­ction in the lifetime ECL provisioning amounts and will actually benefit the banks.

The difference between the 12-month and lifetime ECLs depends on many factors, including the loan’s contractual maturity and “lifetime” as well as how default risks (PDs), recovery values (LGDs) and exposures (EADs) are expected to evolve over the life of a loan.

Annual cash flow projections can be constructed from the principal and inte­rest repayment schedules provided by the bank (since PD is annual). It has been observed that bank can save ECL provision mainly for stage 2 accounts if they are able to track year-wise cash flows and apply conditional PDs instead of simply using cumulative PDs on current outstanding for estimation of their ECL-based provision. Portfolio segmentation into accounts which will be assessed at individual level and accounts that will be assessed as part of a homogeneous credit risk group also provides more risk-sensitive and granular ECL estimates.

ECL Estimation for Marketable Instruments—Corporate Bonds

In Table 4, we have demonstrated the ECL estimation based on cash flow obt­ained from a portfolio of bond instruments. A CDS spread method has been used to estimate the 12-month as well as lifetime PDs and based on that, stage 1 and stage 2 ECLs in dollar amount have been estimated.

The risk-neutral default probability is implied from the CDS spread using the following expression:

CPD= 1-e^(-S × t/(1-R)) … (4)

where S is the flat credit default swap (CDS) spread and R is the recovery rate. Similarly, 1-R is the LGD. The expression represents time. In India, instead of a CDS spread, a bond market spread can be used to estimate PDs for corporate ­borrowers.

Accordingly, using the above expre­ssion, one can estimate cumulative PD for the first year, which is also the 12-month PD = 1-EXP ((-0.50%×1)/(60%)).

Similarly, for the second year also, CPD2 can be estimated using figures reported in Table 4.

One can use forward rates by tracking bond spreads from the yield curve to­estimate the 12-month PD as well as lifetime PD for different maturities.

One can clearly check the difference between the lifetime ECL and 12-month ECL as reported in Table 4. The diffe­rence is quite significant. However, if we do not consider five-year cash flow and directly estimate ECL from by multiplying one-year EAD with five-year cumulative PD, the lifetime expected loss amo­unt will be greater (without discounting). Hence, cash flow-based ECL is a better method.

Concluding Discussions

For the purpose of Ind AS 109 ECL computation for loan portfolios, banks should actually follow the discounted proba­bility weighted cash flow-based EAD method and multiply with the deri­ved 12-month PD for stage 1 and lifetime PD for stage 2 accounts to more precisely estimate their loan provisions. For this, loan-level cash flow data associated with the bank’s term loans need to be culled out. For non-fund-based exposures, app­ropriate credit conversion factors need to be applied. For stage 2 classification loans, banks can choose the criteria that if days past due is more than 60 days, then the term loan is in stage 2, else in stage 1. Banks can also choose other internally set criteria.

Early detection of a significant increase in credit risk may incentivise the bank to bring back those accounts to standard category and shred their further non-performing asset burdens. Banks may consider classifying restructured accounts as stage 2 for a certain number of reporting periods (say for one or two quarters) post restructuring and hold lifetime ECL provisions for such accounts. Banks will have to take an int­ernal policy decision with regard to when they will revert a restr­uctured exposure in stage 2 to stage 1 classification. This can be based on monitoring the repayments on restructured exposure, and if there have been no overdue in the last two to three quarters, the account can be upgraded to stage 1. Such risk-based analysis will enable the banks to maintain their risk appetite and sustain their lending business even in difficult times.


Bandyopadhyay, A (2019): “The Accuracy of Agency Ratings,” Economic & Political Weekly, Vol LIV, No 36, pp 15–17.

Hamilton, D H (2002): “Historical Corporate Rating Migration, Default and Recovery Rates,” Credit Ratings: Methodologies, Rationale and Default Risk, Michael K Ong (ed), Chapter 2, London: Risk Publisher.

Mukherjee, S and S Maji (2017): “Credit Risk Modelling Challenges in IFRS 9,” Prajnan, Vol XLV, No 4, pp 383–404.

Moodys (2006): “Measuring Corporate Default Rates,” November.

Roy, S (2018): “Expected Credit Loss Estimation: Embedding the Forecasts of Future Economic Conditions as per IFRS 9 Guideline,” Prajnan, Vol 46, No 2, pp 185–94.

Saunders, A and M M Cornett (2008): Financial Institutions Management: A Risk Management Approach, New York: McGraw-Hill/Irwin.


Dear Reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Updated On : 31st May, 2022
Back to Top