What type of historical data should an entity collect to develop a provision matrix?
The first step is for an entity to gather information on past history of uncollectable accounts, and generally the profile of payment within its accounts receivable balances. This could be a period of one year, three years or even longer, dependent on the typical business cycle of the entity. Linked to the section above in assessing portfolios, we would expect entities to take this data and split accounts receivable balances into different populations before applying the provision matrix. This could be based on geographical regions, product type, customer ratings, collateral (letters of credit or trade credit insurance), and the nature of the customer (for example, wholesale versus retail). In all cases, the objective is to try to understand the drivers of credit risk for the underlying receivables. For example, one population could be product A in region B being sold to customer type C. The level of segmentation required is a matter of judgement and, in developing segments, the entity should consider whether further segmentation would be expected to lead to only immaterial changes.
Can entities make a specific provision against a particular customer?
Within the HP Upstream division, the business is split into two individual businesses: exploration; and development & production. In both businesses, it has looked to consider how to most appropriately group its accounts receivable to take into account the drivers of credit risk. It has concluded that the size of the customer and its ultimate geographical area are the best determinants of credit risk.
HP has split the size of customer into two broad categories: large established businesses (that is, around 1,000 employees or more); and small / medium-sized businesses. In terms of geography, it has again settled on two categories: entities with ultimate headquarters in OECD countries; and other countries. This has created four differing provision matrices within each business. In addition, where there are specific instances of problem customers, individual provisions will also be created.
HP did consider whether further subdividing the data was required; however, based on historical experience, it concluded that further groupings would not arrive at a materially different answer. The audit team of HP reviewed management’s analysis made to support this conclusion and agreed with the course of action. This was then clearly documented within their audit file, and discussed with management that this judgement would need to be refreshed at each reporting period.
In some cases, the population might be as specific as individual customers. For example, where a particular customer is known to be in financial difficulty, it might require an increased or specific provision compared to historical averages. In such a scenario, it is important to consider and avoid any double counting of losses as a result of the balance being provided for specifically and also being included within the wider general provision default rate for that customer type.
If an entity’s credit control policy requires it to obtain letters of credit or credit insurance, does this mean that it does not need to record a provision under IFRS 9?
IFRS 9 is clear that ‘credit enhancements’, the term it uses to refer to collateral posted or the effect of insurance taken out, cannot be used to justify an assumption that there is no probability of default. However, provided that the credit enhancement is integral to the receivable, which typically means that it has to be taken out at the same time, it can be considered when looking at how big any loss might be on the receivable.
Letters of credit and credit insurance might help to reduce the probability of default (PD) to that of the letter of credit/insurance provider, or reduce the loss given default (LGD), dependent on whether the letter of credit/insurance reduces the likelihood of default, or mitigates the loss after a default has occurred.
Therefore, so long as the entity’s insurance company or the bank supplying the letter of credit has a high credit rating itself, the risk of a significant loss on that customer should be mitigated. A provision should still be calculated, but it might be much smaller than without the ‘credit enhancement’ (see chapter 45 of PwC’s Manual of Accounting at FAQ 45.73.1).
What if an entity does not collect this detailed level of data, or cannot access the information in a cost-effective way?
It would be very unusual for an entity to have no historical data on collection of accounts receivable. The information required might not be readily available within the accounting team but, in the wider finance function or business, such information is likely to exist. As well as the core accounting system, there might be information in a separate billing system, customer relationship management or equivalent sales system, or within a credit control system.
Even if the information does exist, it might be in a format that is challenging for an entity to aggregate and summarise, to provide a full and complete payment history on which to base its historical collection experience. Whilst it might be complicated and difficult, that does not relieve the entity from estimating an expected credit loss on accounts receivable.
Like any accounting estimate, the level of information and documentation that the entity provides to support its position and judgement affects the level of work that an auditor would have to independently undertake in order to obtain the required audit evidence and challenge the decisions that the entity has made.
Therefore, where there are challenges with respect to data availability or ease of collection of such data, an entity could undertake a sampling approach and track the payment profiles and uncollected debts of a subset of its total population. It could do this by grouping customers with similar characteristics, as described above, and then picking a sample of individual customers within each group to consider their specific payment profile. Assuming their groupings do share similar credit risks, these individual payment profiles could then be used as an estimate for each group as a whole. Care would need to be taken to ensure that groups selected are reasonable, based on materiality, and that a sufficient sample is inspected to allow for it to be concluded that the sample is representative of the whole group.
In such a scenario, the auditor would need to ensure that appropriate consideration is given to the approach taken; in particular, it would need to consider and test whether any of the remaining data not considered in the analysis contradicts the assessment and conclusions. In such a scenario, the entity would need to ensure that it gathers and retains the relevant data required for the provision matrix going forward, with this simplification only to support its historical data challenges.
What if an entity does not have the necessary data because it is a start-up or moving into a new market?
In HP Downstream, within the aviation fuel business unit there has been an implementation of a new enterprise resource planning (ERP) system, called ‘TAP’, in the period. The old system was decommissioned and, while some payment data was retained, it is complex to access and not available in a total summarised report. Management has therefore grouped its accounts receivable based on its assessment and understanding of what it deems to be the drivers of credit risk, namely size and any customer-specific factors. These are:
- Group 1 – large international airlines groups (that is, Air Britain, French Airways, Epsilon);
- Group 2 – regional/national airlines (that is, Southeast, Branson Australia);
- Group 3 – regional/national airlines with trade credit insurance (that is, Markair, simpleJet);
- Group 4 – private charter/small business; and
- Group 5 – specific customers (that is, Italian Airlines, Crown).
For each of the above groups, the following information and approach have been taken by management:
- Group 1 – 20 customers – AR @ 31/12/16: £45m – sampled 5 customers
- Group 2 – 86 customers – AR @ 31/12/16: £12m – sampled 7 customers
- Group 3 – 54 customers – AR @ 31/12/16: £9.5m – sampled 12 customers
- Group 4 – 450 customers – AR @ 31/12/16: £2.5m – sampled 12 customers
- Group 5 – 5 customers – AR @ 31/12/16: £7.25m – sampled all customers
Following the sampling approach, management has arrived at payment patterns and bad debt profiles for each of the above groups based on their historical experience. Where trade credit insurance has been taken out in relation to specific receivables, those receivables have been assessed separately, and the extent to which the trade credit insurance mitigates credit risk will be taken into account in the expected credit loss. Management’s next step will then be to consider whether this is likely to be reflective of their future experience.
NB The above sample sizes are not necessarily reflective of PwC Audit sampling methodology; rather, they have been determined by the entity prior to consultation with their auditors.
Where the entity is entering into a new line or area of business, or engaging with new customers, there are different approaches that could be undertaken to estimate the initial impairment provision. This could be in making a judgement that other experience that it has is also relevant for this new product line, because the customer base and its likely credit risk are similar to an existing product line. Alternatively, the entity could look to consider external information, either generally or industry-specific, to inform its initial thinking on an assessment of credit risk. A good place to have these discussions could be with the credit department, which is likely to need similar information in extending initial credit to new customers.
Are there any other methods that an entity might use to gather enough data to calculate lifetime expected credit losses on its accounts receivable?
The use of a provision matrix is not mandatory, but it is a common method. Other potential methods would include customer-by-customer analysis where the entity has a small number of large contracts, application of expected loss rate linked to credit default spreads (‘CDS’) – that is, market-based derivative instruments that provide insurance in the event of default of a country or company – of the country of residence of customers for online retailers, application of dedicated analysis for the main customers, and application of a provision matrix to the residual batch of smaller customers.
How might an entity calculate its expected credit loss provision for a contract asset recognised under IFRS 15?
For contract assets such as unbilled receivables, under IAS 39 a bad debt provision was, in many cases, not recognised. Since the bill had not yet been raised, there was no objective evidence that the customer could not pay their bill, and so the receivable was not impaired. Because IFRS 9 requires an expected credit loss to be recognised on contract assets, there might be a lack of historical credit risk information.
On transition, one approach might be to consider that the collection (and therefore the bad debt) profile will mirror the overall trade receivables portfolio. This approach might overstate the provision, because the credit risk factors associated with the trade receivables might not all be applicable to the contract assets. Alternatively, since the contract assets have not yet been invoiced, the entity could take the view that these all have the same likelihood of experiencing credit losses as the current receivables. For example, all contract assets have a comparable likelihood of not being recovered as accounts receivable which have just been invoiced or are not yet due. This approach might understate the provision, because it does not take into account the ageing or population of the contract assets.
The most appropriate approach to take on transition will be dependent on the facts and circumstances of the entity and its contract asset portfolios. For both approaches, the contract assets portfolio should be actively monitored, following transition, to develop a more accurate expectation of future credit losses.
In some jurisdictions, there might also be regulations in place designed to protect consumers from long-dated bills yet to be invoiced; for example, any bill not raised within 90 days might be prohibited by law from being raised at all. Such a regulation would drive a write-off policy of unbilled receivables over 90 days. (See chapter 11 of PwC’s Manual of Accounting at FAQ 11.295.1 for further considerations on contract assets.)
What steps should be taken to audit this information?
The steps and approach to auditing the historical information used in a provision matrix under IFRS 9 are no different from the way in which any other input to an estimate or judgement would be audited. An understanding is needed of the entity’s process, the methodology that it has adopted, the source of its data, and how it has arrived at its position. Areas to consider include:
- Data: What is the source of the data used as part of the provisioning calculation? Is this from a system that we already have as part of the audit scope for Information Technology General Controls (‘ITGCs’), and is there a new report that we need to consider and test? Does it create a new IT dependency, and does this change the nature and requirements of the current ITGC testing? Should a walkthrough be performed of the process to generate the data, to understand the information flows and to document any relevant controls? If this is coming from a new system, do we need to consider whether we would want to undertake ITGC work, or will we plan to test the data within the report substantively back to source information (that is, bank statements / invoices), or systems where we already have controls evidence?
- Methodology: How has the entity arrived at its methodology for assessing historical data? Has it used a complete set of data for all customers? From what period of time has it considered data? Is this appropriate, based on the payment profile and standard business practice in the industry? If a sampling approach has been undertaken, how has the entity split its accounts receivable into different groups? Are the groupings reasonable and based on an assessment of the credit characteristics of customers? Are there sufficient groupings based on the total population and materiality? Have we looked for customers not used in the sample to assess and challenge the approach?
- Execution: Have we reviewed and checked the entity’s calculations, to ensure that the input data is complete and accurate? Are there any errors in the entity’s calculations or workings? If the entity has grouped customers, have we checked to ensure that they have been included and classified correctly in the right group? Are we aware of any specific accounts receivable balances that might need to be considered separately, due to current increases in credit risk?
In order to answer some of the above questions, we will need to undertake substantive testing, likely on a sample basis; and, in doing so, we use our core standard approaches of target testing, non-statistical sampling and accept-reject as appropriate.
What if an entity has never experienced defaults or instances of non-collection?
The engagement team of HP Downstream have considered the approach undertaken by the aviation fuel business unit. They have reviewed the paper prepared by management that details their basis of undertaking a sample-based approach for the review of their historical data. In the current year, they expect their component materiality within this business unit to remain broadly consistent with the prior year, and therefore would be £2m.
Management has performed a dry run of its analysis on data as at 30/09/2017, and it has provided the results to the engagement team during the interim audit. It has done so to allow the audit team to agree an approach and to feed back any observations in advance of its use for the balances on transition. It has used two years’ worth of data in its analysis.
The engagement team have concluded that they agree with the groupings proposed by management. They did initially challenge management on its approach for Group 3, because the grouping seemed rather large (450 customers) and it had only sampled 12 (lower than our accept-reject methodology would consider appropriate). However, it was concluded that, because the outstanding debtor balance for this group remains broadly at £2.5m, and materiality is £2m, it is highly unlikely that there could be a risk of material misstatement, since almost all of the debtors would need to be impaired.
They have therefore performed the following audit procedures:
- Obtained a listing of the accounts receivable balances and customers as at 30/09/2017.
- From the total listing of customers, performed accept-reject testing on a sample of 16 to ensure that they have been included within the appropriate category (for example, if they are in Group 2, they do represent regional/national airlines).
- For each of the customers included within management’s sample, obtained the relevant reporting from the decommissioned system showing their payment history and any uncollected debts.
- In addition, performed accept-reject testing on an independent sample of accounts receivable balances by agreeing the payment history and any uncollected debts to reporting from the decommissioned system and to cash.
- For each of the receivables for which there is trade credit insurance, obtained the trade credit insurance agreements. Reviewed key clauses in the agreement to ensure that the trade credit insurance is specific to the related accounts receivable and to understand the extent to which credit risk is mitigated for those receivables by the trade credit insurance.
- Agreed the payment data back to underlying reports from the banking system over which the team has controls evidence.
- For those balances listed as uncollected, agreed their write-off balance to the general ledger.
- Obtained management’s calculation and recalculated it to ensure accuracy.
Based on the procedures above, the team have concluded that the historical data used as the starting point for the provision matrix is complete, accurate and reasonable.
Even if an entity has never experienced historical defaults, a provision based on expected losses will still be required. In such circumstances, entities might look to derive probabilities of default and loss rates from external credit ratings, industry-specific data, credit bureaux, or other credit data sources. In addition, it is likely that all entities will have some form of credit risk policy and controls in operation, setting limits on the level of sales / accounts receivable for differing counterparties.
The credit risk policy and controls might be a good place to find further information that an entity has already prepared regarding its assessment of the creditworthiness of counterparties. In such circumstances, based on the limits imposed we might be able to understand and challenge an entity’s own assessment (for example, an entity might be less willing, or might have a lower internal credit limit against some counterparties compared to others that might share certain factors, such as size or credit rating) and use this to draw conclusions between counterparties where actual losses have not been previously incurred.
Although low historical defaults do not mean that no provision will be required, entities should still consider the level of actual profit or loss write-offs over the last few years, assuming a constant level and type of sales activity, because this does provide good historical evidence of losses. That historical data should then be supplemented by forward-looking information.