Following the 2008 financial crisis, loan loss provisioning was changed to reflect ‘expected’ losses rather than ‘incurred’ losses. This made the impairment reserves of banks more responsive to changes in credit quality, but it also introduced a distorting day 2 effect.
Under US GAAP most expected loan losses are charged to profit up front. This ‘prudent’ approach may be liked by banking regulators, but it can produce performance metrics that are confusing for investors. The distortion is greatest for growing loan portfolios, particularly following acquisitions, as illustrated by the Citizens Bank purchase of Silicon Valley Bank.
The demise of Silicon Valley Bank (SVB), and the acquisition of parts of its business by First Citizens BankShares (Citizens Bank), has already provided us with two topics for discussion in The Footnotes Analyst.
- In ‘Fair values and interest rate risk’, we examined the use of amortised cost accounting for loan receivables, how this failed to adequately reflect losses incurred by SVB due to interest rate changes, and why investors need to read the fair value footnotes.
- In ‘Negative goodwill may not mean a bargain purchase’ we explain how business combination accounting resulted in a negative goodwill bargain purchase gain for Citizens Bank, and why we do not believe this accounting profit represents a true economic gain for shareholders.
In this article we focus on the accounting for loan impairments1Various terms are used to describe the balance sheet provision and associated expense related to loans that are not expected to be recoverable, including loan impairments, provisions for credit losses and loan loss reserves. and how the ‘prudent’ lifetime expected loss allowance approach used in US GAAP can produce a misleading profit and loss effect. The effect is most pronounced, and potentially investors will be most confused, where there is significant growth in a loan portfolio – such as following a business combination. This is exactly what happened to Citizens Bank in Q1 2023 due to its acquisition of SVB.
Citizens Bank provision for credit losses
Citizens Bank acquired the lending business of Silicon Valley Bank on 27th March 2023, just before the Q1 2023 reporting date. The quarterly results included a large provision for credit losses equal to 65% of gross interest income, which contrasts with an average 7% credit loss rate in the previous 3 quarters. This large increase in credit losses resulted in a 90% reduction in net interest income, which had a significant impact on earnings in that quarter.
An earlier acquisition in Q1 2022 resulted in a similarly high credit loss charge for Citizens Bank in that quarter. This followed a period of very low and even negative credit losses (the write back of previous provisions that turned out not to be needed) in the preceding quarters.
Citizens Bank quarterly provision for credit losses
First Citizens Bankshares 10Q filings
The question we ask is whether the large credit losses reported in Q1 2023 are a realistic reflection of events in the period? Has profitability really declined that dramatically, and if so, is this perhaps an indication of a poor acquisition decision?
Confusingly for investors the increase in reported credit losses and reduction in profit is largely due to the accounting methodology that is required – it is not a true reflection of the underlying economics. In our view, the expected credit loss approach used in US GAAP (and IFRS, albeit less so) is more focused on ‘prudent’ reporting than on providing investors with a realistic measure of performance.
To appreciate the challenge for investors we need to understand the accounting for credit losses (loan impairments).
Loan impairments under US GAAP and IFRS
Both US GAAP and IFRS apply an ‘expected loss’ approach to loan impairments. The recognition of credit losses was changed after the 2008 financial crisis, during which the then ‘incurred loss’ approach seemed to result in a delay in the recognition of impairment provisions by banks. The phrase “too little, too late” was often used to characterise the old incurred loss approach during the financial crisis.
Under incurred loss accounting, a bank2Credit losses apply to any company with a receivable but, of course, it is of most relevance to banks. could only record a provision, and a loss, when it had “objective evidence” that a loss had been incurred. Essentially this meant waiting for an actual default, or for the borrower to have such difficulties that default became highly likely. In the financial crisis it became obvious that the incurred loss based provisions (at least for some banks) were too low and that some losses that could be anticipated were not being recognised in financial statements.
One of the accounting problems was that the incurred loss approach was applied differently in different jurisdictions. Even though the requirements of IFRS and US GAAP were essentially the same, the application of the incurred loss approach by US banks was much more conservative (higher provisions) than in most IFRS jurisdictions. There were also differences in application between IFRS reporters, with some of them being closer to US practice.
In part, these differences were due to the varied interpretations of what ‘incurred loss’ actually meant in the then accounting standards. However, a major factor was the actions of local banking regulators who, to varying degrees, encouraged higher provisions than arguably justified by the accounting standards. In fact, in the case of the US, the prior application of the incurred loss approach was, in practice, similar to recognising expected credit losses.
Expected credit loss (ECL) based accounting has now been in place for several years. While the FASB and IASB initially intended to develop a common approach to expected credit losses, the final standards are not the same. Unfortunately for investors the differences are significant.
Both US GAAP and IFRS use essentially the same measure of expected credit losses. An ECL is the estimate of how much of a loan balance will not be recovered, after allowing for collateral and other credit enhancements. It is based on current economic conditions and reasonable (albeit subjective) assumptions about the future.
The ECL measure is an expected value that considers the probability of default, even if that probability is low. Therefore, even for loans that are likely to be repaid, a loss allowance is still required that reflects the probability-weighted expected loss. Importantly, unlike the incurred loss model, there is no need for a ‘loss event’, or for a default, for a provision to be recognised.
The problem is that banks can implement an expected credit loss approach in different ways – there are 3 approaches used in practice:
Lifetime expected credit loss approach – US GAAP
The US GAAP accounting applied to the vast majority of loan receivables is the ‘current expected credit loss’ (CECL) approach. It is relatively simple – a balance sheet provision for credit losses must be made for the full amount of future losses expected to be incurred. The net balance sheet value for a loan portfolio is therefore the loan amount outstanding (after allowing for the effects of deferred issue costs and any premium or discount) less the lifetime impairment allowance.3The provision also includes expected losses for lending that is committed but not yet actually made.
The provision is updated each period as new loans are made, loans are repaid and, importantly, as credit quality and credit conditions change. The amount reported in profit and loss is the change in balance sheet allowance in the period, after allowing for any amounts actually written-off.
For an individual loan, or a portfolio of loans originating in the same period, the lifetime loss approach results in the full extent of expected losses being recognised in the first reporting period – often referred to as day 2 (or sometimes day 1) losses. The losses are therefore ‘front-loaded’ and in subsequent periods only catch-up adjustments are required in profit and loss to allow for actual defaults differing from previous expectations, and for changes in future expected defaults.
Build-up of expected credit loss approach – IFRS
Under IFRS not all expected lifetime credit losses are recognised when a loan is first originated. For loans where there has been no default and no “significant increase in credit risk”, only a 12-month ECL provision is recognised. The 12-month allowance is the expected loss considering the probability of default in the next 12 months rather than the full lifetime probability of default. Where the credit quality of a loan has significantly deteriorated, or the loan has defaulted, the credit loss provision is increased to the same lifetime loss allowance that is applied to all loans under US GAAP.4You will often hear the IFRS approach being described as a 3-stage model. Stage 1 is the 12-month allowance for performing loans for which there is no significant increase in credit risk. Stage 2 applies to loans with a significant increase in credit risk where a lifetime allowance is recognised. Stage 3 essentially represents defaulted loans where a lifetime allowance also applies, but for which there is, additionally, no accrual of interest income. The effective lifetime loan loss allowance for stages 2 and 3 are the same, with just a difference in presentation in profit and loss.
The IFRS approach also produces a day 2 loss, but this is only the 12-month allowance and therefore significantly less than under US GAAP, particularly for longer term lending. At the end of the first period after loan origination the reported credit loss allowance is the lifetime allowance for any loan that has already defaulted (or for which the increase in credit risk since origination is significant) plus the 12-month allowance that is required for all other loans. The remaining lifetime losses are built-up gradually in the following periods as and when loans deteriorate or default. This applies even if the future defaults are expected.
You may think that the IFRS approach understates the allowance by not including all future expected losses; however, this is not the case. Expected future losses do not necessarily result in a current economic loss.
Day 2 accounting losses are not economic losses
The problem with both approaches above is the day 2 loss recognition. If a loan is priced correctly, with the interest rate charged sufficient to cover expected credit losses and also provide an adequate return on the investment, then there is no economic loss at the time a loan is originated (or indeed purchased if the purchase price is fair value). A day 2 loss approach may be more ‘prudent’ and therefore to the liking of prudential regulators5The prudential regulation of banks is based on the reported financials but with an overlay specified by the regulators. Since that overlay is flexible one would think that the regulators should not be overly concerned by how ‘prudent’ the accounting is. However, regulators seem to prefer that the prudence is embedded in the financial statements, which explains their close interest in the development of the expected credit loss approaches., but it does not faithfully reflect the economics of a lending business.
In effect, US GAAP initially double counts (and IFRS partly so) the effect of credit losses when a loan is first recognised. Expected losses are already included in the origination (or purchase) price, considering this price allows for the contractual interest charged which, in turn, allows for credit losses. Recognising an immediate additional allowance is double counting.
The day 2 loss effect was so contentious when the US GAAP expected credit loss model was being developed that the amendment to US GAAP was only approved by 5 out of the 7 FASB members. The full dissenting view of Jim Kroeker and Larry Smith is quite long, but we have picked out a few sentences to give you a flavour of their concerns.
Messrs. Kroeker and Smith dissent from the issuance of this Accounting Standards Update because they disagree with the requirement to recognize a credit loss at origination or purchase, at an amount equal to the “lifetime expected credit loss” for financial assets. ….
this conceptual shortcoming of this Update thereby results in financial reporting that does not faithfully reflect the economics of lending activities. …
Recording a credit loss at initial recognition (along with a related allowance for loan losses) results in a balance sheet presentation that reflects the credit risk twice; it is reflected in the price paid (which is based on the terms of the instrument, including the stated interest rate) and it is reflected in the allowance for loan losses. In an arm’s-length transaction …. an entity would not be expected to incur an economic loss on the day a loan is made or the security is purchased.
Click here to see the full text of the dissenting view on pages 235 to 240 of the US GAAP Credit Losses standard (topic 326) – we think this should be required reading for any investor analysing US banks.
The immediate write-down of a loan to an amount below the transaction price may not present too much of a problem in practice if high-quality loans are originated and therefore expected credit losses are small. Furthermore, the front-loading effect will have a smaller profit and loss impact where banks originate new loans each period. This because the overall profit and loss effect is a combination of day 2 losses for new loans and only the smaller catch ups for others. In fact, if overall lending is stable, the day 2 loss effect completely disappears in profit and loss (but not the balance sheet).
However, where loans of lower credit quality are originated, and particularly where ‘credit deteriorated’ loans are purchased, the day 2 loss effect could become so large that the misrepresentation of the economics becomes intolerable. This is where both US GAAP and IFRS turn to a third method of allowing for credit losses – the purchased credit impaired approach.
Purchased credit deteriorated approach – both US GAAP and IFRS
Both US GAAP and IFRS apply an alternative impairment approach to a subset of assets. For US GAAP it applies to assets that are purchased (i.e not originated loans) for which credit quality has deteriorated by a “more-than-insignificant amount” since origination. For IFRS it is those assets that are “credit-impaired” at the time of either purchase or origination (the method is called ‘purchased credit impaired’ under IFRS). Although the requirements differ, the most common situation where this approach applies is the same – loans acquired in a business combination for which credit quality has deteriorated since the loans were first originated.
The purchased credit deteriorated approach is, in effect, the same as the US GAAP lifetime allowance approach, but with one modification – the lifetime expected credit loss estimated at the time of purchase (or origination under IFRS) is amortised over the period of the loan and is not recognised as an initial (day 2) loss. Subsequent changes to the ECL estimate are immediately recognised in profit and loss.
Under US GAAP, this approach is applied by using a ‘gross-up’ methodology, where the initial ECL is added to the loan balance sheet carrying value. This creates an issue premium that is amortised to profit and loss as an expense over the loan term. The full lifetime ECL is still recognised, but as a deduction from the higher loan balance. Therefore, on initial recognition, the loan purchase price is the same as the net carrying value in the balance sheet and no day 2 loss is reported.
The equivalent purchased credit impaired (PCI) approach under IFRS is described in a different way, but the outcome is essentially the same. IFRS explains the PCI method as a revised present value calculation rather than as a simple gross-up of the loan balance. However, the gross-up method is a way of implementing the IFRS approach in practice.6The reason for the difference between how the PCD (PCI) approach is described by US GAAP and IFRS is that the two accounting systems define amortised cost differently. Under US GAAP amortised cost represents the transaction price, with adjustments made for issue costs, fees, and redemption premiums and discounts. However, under IFRS, amortised cost is a present value calculation with the contractual cash flows discounted at the original effective interest rate (EIR). Issue costs, fees, and premiums and discounts are included in the EIR.
Because amortised cost is a present value under IFRS, impairment allowances are also described in a similar way. In the IFRS purchased credit impaired approach, the expected cash flows (after forecast credit losses) are discounted at the credit adjusted effective interest rate. The credit adjusted EIR is the internal rate of return of cash flows expected when the loan originates or is purchased.
The IFRS approach is numerically more complicated than that described by US GAAP. Nevertheless, the result is a loan balance that is the same as the grossed up cost less lifetime impairment allowance presented under US GAAP (assuming consistent inputs and consistent application of the time value of money in the grossing up).
The purchased credit deteriorated approach gives a much better representation of the economics of lending. The initial expected losses are ‘matched’ against the interest income that compensates for those losses. Furthermore, the net balance sheet carrying value of the loan is much closer to the economic value (present value of expected cash flows), albeit within an amortised cost system and therefore not updating for the effect of changes in interest rates.
However, this approach does back-end the recognition of losses relative to the two main approaches above (and compared with the previous incurred loss approach if incurred losses are higher in the early part of a loan), which was of particular concern to regulators at the time of the last financial crisis.
Perhaps the best way to explain the different approaches is with a simple example. Here is an embedded an interactive model to show the calculations.
Interactive model: Illustration of alternative loan impairment approaches
— iPhone and iPad users: This model formats best if viewed in Google Chrome —
To see the full workings behind this model, please access the downloadable version.
Please enter your email address to receive an excel version of this model
The input data when the model is first loaded (you can change any of the blue figures) shows that, when this portfolio of loans is originated, 5.0% of the loan principal is not expected to be recovered. In subsequent periods the actual defaults may differ from prior expectations and future expected losses are subject to amendment. At the end of the 5 years aggregate losses end up at 5.8% which means that in total 58 must be charged to profit and loss.
The difference between the approaches is simply how this 58 is allocated to each period:
- US GAAP lifetime loss approach: The lifetime loss expected at initial recognition is 50 (5% of 1,000), which is fully expensed in the first period. Year 1 also includes the effect of any change in estimates between origination and the end of that first period. Subsequent profit and loss effects are merely catch up the adjustments which add up to a further 8.
- IFRS 3-stage build up approach: The allowance in year 1 is the loss incurred from defaults in that period (the stage 3 loans) plus a 12-month ECL allowance for the remaining ‘performing’ (stage 1) loans. This equals 29, with the remaining losses built up in a similar manner over the remaining loan term.
- Purchased credit deteriorated approach (both IFRS and US GAAP): The basis of the PCD approach is that the initial forecast ECL of 50 is covered by a portion of the interest charged to all borrowers – in this case about 1% p.a. Therefore, the initial ECL is recognised over the 5 years to produce a loss of about 10 each year, which is in effect matched against the compensating interest income. Under US GAAP (IFRS describes the approach differently but the outcome is essentially the same) this is achieved by grossing up the loan balance and amortising the resulting premium. By grossing up in this way, the full lifetime ECL can be included in the balance sheet without creating the problematic day 2 loss. This lifetime allowance is then updated each period, with changes immediately included in profit and loss.
Our model is highly simplified and obviously the precise numbers are determined by the assumptions regarding the amount and timing of losses and how estimates change over time for this particular loan portfolio. However, we think the general pattern of losses, and differences between the approaches, we highlight is representative of what you might expect in practice.
Citizens Bank disclosures about day 2 losses
Returning to the Q1 2023 report of Citizens Bank: the extract below shows the loans it acquired from SVB analysed between those which are purchased credit deteriorated and those which are not. Notice that they identify a $200m impairment allowance for the PCD loans, which is about 10% of the acquisition date fair value. This is the amount that is used to ‘gross-up the loan amount’ which means it is, in effect, recognised as a loss over the life of the loans. The $200m would have increased the day 2 loss had PCD accounting not been applied.
For the non-PCD loans, the lifetime loss recognised on day 2 is $462m, which is about 0.7% of the loan fair value. It is this loss (together with a similar lifetime allowance for loan commitments) that has such a significant impact on the quarterly profit in Q1 2023, as we summarised in our earlier table.
Citizens Bank accounting for non-PCD and PCD loans
Pages 16 and 21 of First Citizens BankShares Q1 2023 10Q. (UPB stands for unpaid principal balance.)
In evaluating the performance of Citizens Bank, we think you should exclude the artificial effect of the day 2 loss recognition for the SVB acquisition (and the similar effect one year earlier for a different acquisition). However, the immediate recognition of a lifetime loss on day 2 affects more than just the period of an acquisition. In subsequent periods, credit losses for these loans are likely to be very low because the only profit and loss effect is to update the lifetime loss estimate. The interest earned on these loans reflects the compensation for credit losses, but the losses themselves have already been recognised in an earlier period. Therefore, in comparison with the application of the IFRS build-up effect, or the purchased credit impaired approach, the reported profits of Citizens Bank in subsequent quarters will be overstated.
The problems caused by the day 2 effect are particularly apparent immediately following a significant acquisition. However, they can also cause a distorting effect for a bank that is experiencing significant and persistent organic growth relative to its peers, where the day 2 losses create a profit ‘headwind’.
In our view, if the method used for purchased credit deteriorated loans were applied to all lending, performance metrics would be more understandable and more representative of the economics. In the FASB members’ dissenting view, they also highlight the seemingly odd outcome of it being only the less risky assets which are immediately written down to an amount that is below the loan amount (or fair value in the case of assets recognised in a business combination).
The amendments in this Update require that no loss be recorded for expected credit losses for the subset of financial assets for which credit has deteriorated by a more-than-insignificant amount (since origination) when the assets are acquired in a purchase transaction. Messrs. Kroeker and Smith agree with the approach for those assets, which is, in effect, the same as the method they suggest should be applied to other financial assets. However, they question why it is deemed necessary to require a Day-1 loss at an amount equal to expected credit losses for the vast majority of less risky assets when no such loss is recorded for the riskier credit deteriorated assets. This seems to be an acknowledgement that recognition of Day-1 losses is not a fundamentally sound basis of accounting. Messrs. Kroeker and Smith fully support the result of the model required for assets that have experienced a more-than-insignificant credit deterioration, and they see no reason (conceptually or otherwise) that the model should be applied only to purchased financial assets with credit deterioration.
An impairment methodology that was very similar to the above purchased credit deteriorated approach was initially proposed by the IASB immediately following the 2008 crisis. This would have applied to all loans, much as suggested by the FASB members in their dissenting view. However, the proposal was not liked by banks (too complex and the result too volatile) or prudential regulators (not prudent enough) and the idea was regrettably dropped. As is often the case in accounting standard setting, the first ideas are often the best ideas.
What the IASB ended up with – the 3-stage build up approach – does result in a day 2 effect but this is smaller than under US GAAP. Nevertheless, it is still something that investors will need to consider in their analysis and, in some cases, may make performance metrics unreliable.
We have focused on the day 2 impairment headwind that banks experience when they grow either by acquisition or organically. However, there is an equally problematic effect in cases where lending declines. If there are fewer new loans originating (or purchased) than are maturing, the impairment headwind turns into a tailwind and bank profitability is artificially improved. This is because there are fewer day 2 losses, and subsequent impairments for existing loans are automatically low (particularly under US GAAP).
The FASB is currently reviewing their credit losses standard and have recently issued a proposal to extend the application of the purchased credit deteriorated gross-up approach to all purchased assets. This would eliminate all day 2 losses arising on the acquisition of a loan book. However, it would not affect the similarly problematic day 2 losses for originated loans. We see no economic difference between a purchased and an originated loan, and suggest FASB extends the gross-up method to all loans, as recommended in the FASB members dissenting view that we highlight above.
Insights for investors
- Credit induced impairments under US GAAP involve making a lifetime expected loss provision immediately following the origination or purchase of a loan. IFRS has a similar day 2 loss but restricted to a ’12-month allowance’.
- Day 2 losses may be prudent, and to the liking of prudential regulators, but they fail to reflect the economics of lending. Interest rates charged include a credit spread to cover expected losses.
- For banks in steady state, the impact of day 2 losses on profit and loss is mitigated. But for banks that are growing a loan book, the impairment allowance creates an artificial profit headwind. This is most apparent following significant acquisitions.
- The expected loss approach does at least ensure that allowances are responsive to changes in current and forecast economic conditions, although, of course, impairments require judgement.
- A different impairment approach applies to purchased credit deteriorated loans (purchased and originated credit impaired under IFRS); this does not produce a day 2 loss. In our view, this approach better reflects the economics of lending.