Covid19: Measuring the Impact under IFRS9

A Black Swan

Background

The Reserve Bank of India (RBI), on March 27 2020, permitted lending institutions to offer a three-month moratorium to their borrowers with respect to all term loans. This is in the backdrop of the COVID-19 outbreak and the nationwide lock-down that began in late March, bringing most economic activities to a standstill. Most forecasters predict the COVID-19 to impact economic growth significantly in FY21. Companies have already seen their cash flows shrinking.

RBI also relaxed the 90-day non-performing asset (NPA) classification norms for accounts being granted the three-month loan moratorium. These also included loans which were already in overdue (but standard) buckets as on Feb 29, 2020, thus leading to an asset reclassification standstill during the moratorium period from March 1 to May 31, 2020.

These can be seen as substantial relief measures to all the cash constraint borrowers and step in direction of protecting small businesses from liquidity stresses in these testing times.

Impact on Lenders

However, for lenders specifically NBFCs (NBFCs are the only financial institutions under Ind-AS in India, with implementation deferred for both banks and insurance companies till further notice), this could lead to a Catch 22 situation. Not giving moratorium in these testing times might give rise to short term / long term liquidity issues in their borrowers’ balance sheets. Giving a moratorium may lead to higher intensity of defaults in the three months post moratorium if the recoveries in borrower’s collections are not sharp enough. These include already existing delinquent borrowers who would have anyways defaulted during the moratorium had the asset reclassification norms not been changed and the new set of borrowers who are hit hard by economic inactivity. Thus, there might be large scale asset reclassification post moratorium because of existing (or pre-existing) overdue, cash losses during the moratorium period, industry wide stresses, collapse of business models etc.

The implications of such a situation on lenders are eerily similar to many time travel shows made — take for instance the German masterpiece on Netflix — Dark (best sci-fi ever in my opinion). In the show, a fictional town Winden has tunnels which can take someone back and forth in time to year 1953, 1986, 2019 or 2052 — with a gap of 33 years.

Payment holidays are like such time travel but in a credit risk universe. Say, a borrower ‘X’ who has taken a loan of INR 1 mn from a lender ‘Y’. Now with a payment holiday (say from 2019 to 2052 for comparison), lender Y is transported in the future directly to 2052 through the time travel tunnel as there are no intermittent cashflows to it. However, borrower X has to live these 33 years in the regular world, where there can be two scenarios — one in which its business recovers and normalises, and another in which it faces income losses, net-worth erosion, liquidity issues etc. Y expects to recover INR 1 mn with accrued interest when X reaches 2052! Add to it, if there are 1–2 EMI overdues and the borrower was a special mention account already (say SMA 2 with dpd 60–90) when it all started in 2019.

Though the period may not be as long as 33 years, payment holidays may act in a similar fashion with a lot of information asymmetry between a lender and a borrower.

Brace for the Worst but include the mitigating effect of government relief

A good start to brace for such a bad landing post the payment holiday is to estimate or foresee the kind of credit losses NBFCs might face. This would also help in budgeting and financial planning for the year ahead. The remaining part of the article talks about the challenges and suggestions to compute the expected losses more accurately from an IFRS9 and ECL perspective.

(From BCBS document on ‘Measures to reflect the impact of Covid-19’)

There are high levels of uncertainty currently surrounding the forward-looking information relevant to estimating expected credit losses (ECLs) and to applying the IFRS 9 assessment of significant increases in credit risk (SICR). IFRS 9 is a principles based standard and requires the use of experienced judgement. At present, information available that is both reasonable and supportable on which to assess SICR and to measure ECL is limited. Regarding the SICR assessment, relief measures to respond to the adverse economic impact of Covid-19 such as public guarantees or payment moratoriums, granted either by public authorities, or by banks on a voluntary basis, should not automatically result in exposures moving from a 12-month ECL to a lifetime ECL measurement. Where banks are able to develop forecasts based on reasonable and supportable information, the Committee expects ECL estimates to reflect the mitigating effect of the significant economic support and payment relief measures put in place by public authorities and the banking sector. While estimating ECL, banks should not apply the standard mechanistically and should use the flexibility inherent in IFRS 9, for example to give due weight to long-term economic trends.

Challenges:

1. Staging: Days‑past‑due (DPD) metrics would reflect the impact of payment moratoriums where borrowers take advantage of a payment holiday and so amounts may no longer be past due.The DPDs being frozen will lead to stagnant Staging. The additional loans which could have moved to Stage 3 (non performing) had payment holiday not been in place will continue to remain in Stage 1 (performing) & stage 2 (under-performing). Similarly the loans which might have moved to Stage 2 without moratorium would remain in Stage 1.
2. Probabilities of Default (PD): The PDs themselves might be underestimated as the common PD models which use forward rates or roll rates to estimate PD (% of loans moving into DPD90+ bucket from lower buckets historically) will not show any change. Thus there will not be any change in the PDs just before and after the moratorium which may not be reflective of the actual situation on ground. Same is the case with the models which use no. of defaults as an input to PD modelling. These Through-the-Cycle (TTC) PDs may actually decline owing to a better ratio of no of defaults and no of entities, and resultant scaling factors — contrary to the economic environment.
3. Macroeconomic Trends: Difficult to find robust macroeconomic estimates to compute forward looking Point-in-Time (PIT) PDs. (ECL is a forward looking process where the TTC PDs arrived at using the current and historical data pertaining to existing portfolio of financial assets is typically log regressed to get future PIT PDs)
4. Lack of Information for SICR: Due to lack of information about the borrower during the payment holiday, SICR criteria for moving an asset from Stage 1 to Stage 2 owing to significant increase in credit risk (SICR) since initial recognition may not be very objective. This will lead to a few inherently Stage 1 assets being moved to Stage 2 while a few SICR assets remaining in Stage 1 due to misjudgement.
5. Lack of clarity on capital infusion: Many entities have their future objectives which are dependent on capital infusion, the lack of which might impact its business goals. Given the uncertainties, the tentative timelines for capital infusion might go for a toss. Some of the entities may cease to remain a going concern if the proposed infusion is cancelled.
6. Lack of projections of disease spread by geography: To account for impact in products where geographical risk is a factor, lack of sound projections can be a challenge to stress based on concentration to particular geographies.

To do some balancing, RBI has stipulated 10% as additional provisioning on overdue but standard accounts as on Feb 28, 2020. Such flat rates may remind one of the pre IND-AS era and are not typically reflective of provisioning under Ind-AS which is conceptually against rule-based adhoc provisioning.

(‘Entity’ refers to ‘Borrower’, ‘Payment Holiday’ and ‘Moratorium’ are used interchangeably in the article)

Measuring the Impact:

There are a few ways to estimate the credit losses more accurately in the post moratorium world for NBFCs (under IndAS):

1. Use updated macroeconomic trends or give higher shocks to downside scenarios: There are macroeconomic trends now available which factor in the slowdown due to COVID 19 pandemic (though accuracy of such forecasts cannot be ascertained right now). In case, the macroeconomic variable does not show weakening trends on account of the impending economic stress, it might be a good idea to give a higher downside shock to the variable in regression.
2. Increase the Downside Scenario weights: ECL is typically calculated using PD forecasting in three scenarios — Base case, Upside case and Downside case. If the process does not contain scenarios and relies only on the base case, it would be a good time to add these scenarios (at least the downside scenario) to stress the forecasts of macro-economy variable available. Weights of such scenarios which are typically much lower than the base case can be increased in line with the projections of the rating agencies or consultation firms outlook for the asset classes in the segment.
3. Projections of impacted geographies: For products, where geographical risk is a factor, the color coded zones might be helpful to incorporate higher stressing on red districts. To go into finer projections rather than just stressing based on the three zones, one may use the SIR model to estimate the final size of the epidemic for each district. If one wants to incorporate more factors like age dependency, gender, demographics, no. of ICU beds in a district, ventilators etc, following links describe how to build a model and fit the curve to real world data.

Infectious Disease Modelling: Beyond the Basic SIR Model

Infectious Disease Modelling: Fit Your Model to Coronavirus Data

SICR (Movement from Stage 1/12 month ECL to Stage 2/Lifetime ECL):

1. Accepting a moratorium or other relief does not (automatically) mean a loan is forborne: Note that there is a difference in treatment of SICR in case of payment holiday extended to a specific borrower and a blanket payment holiday. In the former case, the borrower will definitely move to Stage 2 as extension of payment holiday is an evidence of SICR. In cases of blanket payment holidays or moratoriums, which might be due to situations like the current pandemic, a specific set of borrowers can be de-staged based on analysis of their liquidity, operations, solvency.
2. Assume Stressed collections, Multiple Liquidity scenarios : Using scenario based approach for liquidity and stressing the collections may be one of the approaches. The scenarios need to also consider the positive impact of all forms of government relief directly and indirectly to the entities — like moratorium, delayed tax filings, govt guarantees, reduced taxes etc. A top-down approach for SICR can be adopted if specific entities could not be assessed due to lack of information. If it can be established that the liquidity issues in the entity are only transient, management may decide not to recognise SICR.
3. Discard proposed equity infusion / commitments: Any proposed equity infusions or debt raises which have not yet taken place should ideally be discarded and not be a factor in the projections. Assessment should be done on the current balance sheet, in order to recognise SICR. The entity can be moved back to Stage 1 (‘cured’) once equity infusion is made.
4. Status before Covid: Entities which were already struggling with their portfolio quality and liquidity management in the pre-covid world might be more vulnerable as these may not find any kind of support from investors/lenders.
5. Estimate Cure rate with a conservative U shaped recovery rather than an aggressive V shaped recovery of the overall economy: Due to the lower base effect of the already weak status of the pre-covid economy (analogy — going from 100 to 20, and then growing from 20 to 40 — it should not be seen as 100% growth!), the recovery might seem more robust for 1–2 quarters post pandemic than it actually is. Long term recovery may well be a timid variant of such a post covid short term spike due to lower base. The movement of entities back from Stage 2 to Stage 1 (cure rate) should reflect the long term recovery and not the short term spike.

Overlay vs Model change:

Things are changing very quickly so using credit provision overlays in the first instance may be more expedient than re-modelling. As better information becomes available, a combination of modelled outcomes and overlays may be required.

Conclusion:

IFRS 9 requires an unbiased probability weighted estimate, so not too conservative nor too optimistic approach should be taken. It will be important to ensure that the process is not overly prudent given the economic situation and implied tendency of higher risk aversion, if recovery is visible.