Portfolio analysis is undertaken to determine whether the present and expected future losses are consistent with the revenue expectations of the portfolio and enable it to generate a profit both now and in the future.
The phrase loss minimisation is not used since losses and revenue are closely linked in many products; nowhere more so that in credit cards and lines of credit.
The transition from a current, interest earning, account to a credit loss takes place via a series of different levels of delinquency. It is the analysis between these different levels of delinquency over time which form the core of the analysis.
Start with Objective Measures
The whole process falls apart if the measure of delinquency is not objective and consistent. Yet it is surprising how often this is the case amongst many of our leading consumer banks.
Issues include, amongst others:
- Erratic treatment of partial or multiple payments
- Loan rescheduling accompanied by re-setting of delinquency status and without separate monitoring
- Inconsistent treatment of fee payments
- Collector intervention in the ageing process.
Analysis that is performed on inaccurate data is like a house built without foundations. It is likely to fall down.
Monthly Delinquency Reports
It is common to report the status of a portfolio broken down into the different levels of delinquency as measured by contractual ageing. (That is the number of days delinquent is measured against the schedule established when the credit agreement was signed and the amount reported is the balance, shown as the “balance at risk” (net of any unearned income)).
| Delinquency | Jan - 21 |
| Current | 1,234,567 |
| 1 29 dpd | 49,383 |
| 30 59 dpd | 11,358 |
| 60 89 dpd | 4,998 |
| 90 119 dpd | 3,798 |
| 120 149 dpd | 3,342 |
| 150 179 dpd | 2,975 |
| 180 209 dpd | 2,796 |
| 210 239 dpd | 2,712 |
| 240-269 dpd | 2,658 |
| 270 299 dpd | 2,631 |
| 300 329 dpd | 2,605 |
| 330 359 dpd | 2,579 |
| .. | .. |
| 720 dpd | 2,566 |
A report of this kind will look like the one above. The numbers used are realistic for a mature portfolio. In the example we have drawn a line across the table at 180 days past due which is the point at which the accounts are written–off.
Write-off is often mistakenly understood to mean that there will be no future revenue from the account. In fact in most markets there is still significant value of an account that has been written-off. The debt for the customer still stands and there is still a legal obligation on behalf of the customer to settle the debt. The extent to which they are collected depends very much on either the local legal environment or the efforts of the collector. In countries, such as Germany, where it is usual for banks to apply to the courts for a portion of the customers wages to be assigned directly to the bank the rate of recovery can be very high. Where this legal support is not available the proportion collected will depend on the skill of the collections department and usually varies between 0% and 33%.
In most countries there is an active market for the sale of written-off debt with prices varying up to a maximum of around 8% of the face value of the debt. The logic for this maximum is that if the debt is recoverable at 33% then the cost of recovery will typically be less than 25%.
In our example the delinquency may not be important in terms of the reported results of a business but for the credit risk manager understanding the evolution of delinquency beyond write-off is critical.
Of the balances which are not written-off it is normal to consider the ratios of the percentage of balances which are greater than 30dpd and the percentage greater than 90dpd. In the case of our example the ratios are
30+ dpd delinquency - 2.2%
90+ dpd delinquency - 1.0%
The percentage of the receivable written-off in a given month which is called the gross credit write-off is 0.21%. If this is repeated in every month of the year the annual gross write-off would be 12 x 0.21% or 2.52%.
Static to Dynamic Delinquency
Although these numbers are commonly used they shed very little light on any possible cause of delinquency nor upon the impact of any changes in the portfolio.
If the delinquency figures for two consecutive months are written side by side then a direct comparison can be made between the two months. In the table underneath can be seen that not only are the receivables growing but so are the delinquent balances at each level of delinquency.
Unfortunately the table doesn’t say anything about the movement of accounts between the two months so accounts which for example, are 1-29dpd in Feb-21 may have been in any non-written-off bucket in January.
Of course the maximum an account can age between two months is 30 days but they can pay off as much of the overdue balance as they are able and can become less delinquent all the way back to current or anywhere in between.
Experience, however, tells us that the evolution of delinquency for accounts beyond 60 dpd is likely to be determined primarily by most customers making no payment and simply ageing to the next bucket.
| Delinquency | Jan-21 | Feb-21 |
| Current | 1,234,567 | 1,296,295 |
| 1 - 29 dpd | 49,383 | 61,728 |
| 30 59 dpd | 11,358 | 12,346 |
| 60 89 dpd | 4,998 | 5,111 |
| 90 119 dpd | 3,798 | 3,748 |
| 120 149 dpd | 3,342 | 3,228 |
| 150 179 dpd | 2,975 | 3,008 |
| 180 209 dpd | 2,796 | 2,826 |
| 210 239 dpd | 2,712 | 2,740 |
| 240 269 dpd | 2,658 | 2,658 |
| 270 299 dpd | 2,631 | 2,605 |
| 300 329 dpd | 2,605 | 2,579 |
| 330 359 dpd | 2,579 | 2,553 |
| .. | .. | .. |
| .. | .. | .. |
| 720 dpd | 2,566 | 2,566 |
It suits our purpose to make the simplifying assumption that all accounts arriving in a given delinquency bucket come from the previous (one lower) stage of delinquency. This assumed movement from month to month is called the “net flow” or “roll” of delinquent balances.
From the above table we will be saying that the flow from the 30-59dpd bucket in January to the 60-89dpd bucket in February is 11,358 -> 5,111 or a flow of 45%.With this simplifying assumption in place it becomes possible to analyse further what is happening to risk within a portfolio.
The Net Flow Matrix
On the following two tables, the first is a delinquency matrix which shows the balances delinquent for a series of 12 months, whilst the second table is a net flow matrix which shows the transition to a state of delinquency in a given month from a state of lower delinquency in the previous month under the assumption that accounts either only become one step more delinquent every month or cure entirely (a simplification but adequate especially for the later buckets).
| Delinquency | Jan-21 | Feb-21 | Mar-21 | Apr-21 | May-21 | Jun-21 | Jul-21 | Aug-21 | Sep-21 | Oct-21 | Nov-21 | Dec-21 |
| Current | 1,234,567 | 1,296, 295 | 1,361,110 | 1,429,166 | 1,500,624 | 1,575,655 | 1,654,438 | 1,737,160 | 1,824,018 | 1,915,219 | 2,010,980 | 2,111,529 |
| 1 - 29 dpd | 49,383 | 61,728 | 64,815 | 68,056 | 71,458 | 75,031 | 78,783 | 82,722 | 86,858 | 91,201 | 95,761 | 100,549 |
| 30 - 59 dpd | 11,358 | 12,346 | 15,432 | 16,204 | 17,014 | 17,865 | 18,758 | 19,696 | 20,680 | 21,714 | 22,800 | 23,940 |
| 60 - 89 dpd | 4,998 | 5,111 | 5,556 | 6,944 | 7,292 | 7,656 | 8,039 | 8,441 | 8,863 | 9,306 | 9,772 | 10,260 |
| 90 - 119 dpd | 3,798 | 3,748 | 3,833 | 4,167 | 5,208 | 5,469 | 5,742 | 6,029 | 6,331 | 6,647 | 6,980 | 7,329 |
| 120 - 149 dpd | 3,342 | 3,228 | 3,186 | 3,258 | 3,542 | 4,427 | 4,648 | 4,881 | 5,125 | 5,381 | 5,650 | 5,933 |
| 150 - 179 dpd | 2,975 | 3,008 | 2,906 | 2,867 | 2,932 | 3,187 | 3,984 | 4,184 | 4,393 | 4,612 | 4,843 | 5,085 |
| 180-209 dpd | 2,796 | 2,826 | 2,858 | 2,760 | 2,724 | 2,786 | 3,028 | 3,785 | 3,974 | 4,173 | 4,382 | 4,601 |
| 210 - 239 dpd | 2,712 | 2,740 | 2,769 | 2,801 | 2,705 | 2,669 | 2,730 | 2,968 | 3,709 | 3,895 | 4,090 | 4,294 |
| 240-269 dpd | 2,658 | 2,658 | 2,685 | 2,714 | 2,745 | 2,651 | 2,616 | 2,676 | 2,908 | 3,635 | 3,817 | 4,008 |
| 270 - 299 dpd | 2,631 | 2,605 | 2,605 | 2,632 | 2,660 | 2,690 | 2,598 | 2,564 | 2,622 | 2,850 | 3,563 | 3,741 |
| 300 - 329 dpd | 2,605 | 2,579 | 2,553 | 2,553 | 2,579 | 2,607 | 2,636 | 2,546 | 2,512 | 2,570 | 2,793 | 3,491 |
| 330 - 359 dpd | 2,579 | 2,553 | 2,527 | 2,502 | 2,502 | 2,528 | 2,554 | 2,583 | 2,495 | 2,462 | 2,518 | 2,737 |
| .. | .. | .. | .. | .. | .. | .. | .. | .. | .. | .. | .. | .. |
| 720 + dpd | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 |
| Jan-21 | Feb-21 | Mar-21 | Apr-21 | May-21 | Jun-21 | Jul-21 | Aug-21 | Sep-21 | Oct-21 | Nov-21 | Dec-21 | |
| 1 - 29 dpd | 5% | 5% | 5% | 5% | 5% | 5% | 5% | 5% | 5% | 5% | 5% | |
| 30 - 59 dpd | 25% | 25% | 25% | 25% | 25% | 25% | 25% | 25% | 25% | 25% | 25% | |
| 60 - 89 dpd | 45% | 45% | 45% | 45% | 45% | 45% | 45% | 45% | 45% | 45% | 45% | |
| 90 - 119 dpd | 75% | 75% | 75% | 75% | 75% | 75% | 75% | 75% | 75% | 75% | 75% | |
| 120 - 149 dpd | 85% | 85% | 85% | 85% | 85% | 85% | 85% | 85% | 85% | 85% | 85% | |
| 150 - 179 dpd | 90% | 90% | 90% | 90% | 90% | 90% | 90% | 90% | 90% | 90% | 90% | |
| 180 - 209 dpd | 95% | 95% | 95% | 95% | 95% | 95% | 95% | 95% | 95% | 95% | 95% | |
| 210 - 239 dpd | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | |
| 240-269 dpd | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | |
| 270 - 299 dpd | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | |
| 300 - 329 dpd | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | |
| 330 - 359 dpd | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | 98% | |
| .. | .. | .. | .. | .. | .. | .. | .. | .. | .. | .. | .. | .. |
| 720 + dpd | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 | 2,566 |
The net flow tells quite a detailed story about how the portfolio is being managed. First of all the evolution of an account through the different stages of delinquency gives the monthly gross loss (before recoveries). The simplest way of stating this is to say that the losses which arrive at 180 dpd in August 2021 were current in January of that year. So the percentage that becomes written-off from that month is 3,785 / 1,234,567 or 0.306% of the current balance. If this happens on each of 12 months of the year then the annualised loss rate is 3.68% of the current balance.
Another way of calculating the same number is to take the product of the monthly flows:
5% x 25% x 45% x 75% x 85% x 90% x 95% = 0.306%
The advantage of this way of looking at things is that it breaks the losses down into different stages of the evolution and if there is deterioration in any part of the process it can be quickly identified and remedial actions taken.
If the losses were to deteriorate by 20% from 0.306% per month to 0.366% per month then the reason may not be obvious immediately, however if it can be seen that the flow rates have changed to those underneath
5% x 25% x 54% x 75% x 85% x 90% x 95% = 0.366%
It can be seen that the flow from the 30pd bucket to the 60dpd bucket has increased from 45% to 54%, which accounts for the 20% increase in the gross losses. The problem lies in a specific part of the collections area. It could be due to staff shortages or some other such change in collections strategy but at least the problem has been located and can be treated.
Breaking the Flow
It has become common practice for lots of good reasons to break the transition from current good customer to recognised final loss into a number of stages. This is partially encapsulated in Basel II where the division is made between account prior to default (usually 90 dpd) and post default. This enables the Basel II formulation of expected losses as:
EL=PD x LGD x EAD
Many lenders would provide a more detailed breakdown of the process divining the loss given default transition into two separate parts; that prior to write-off and that post write-off.
This is because experience tells us that the analysis for predicting write-off from delinquent status works quite well using a net flow model but beyond that it is better to perform vintage analysis to forecast future losses.
The other major advantage of this kind of analysis is that it can be used to understand the true state of the portfolio without being misled by the impact of the effects of growth.
In the example that we are looking at on the previous page the accountants view of the percentage of losses would simply be the balances written-off divided by the sum of the previously non written-off balances. This is 0.203% of total balance (0.218% of current balance) or almost a third lower than the lagged value.
This is because the receivable that is used as the denominator refers to the month of the write-off. Since the portfolio has been growing since the time when the newly written-off account was current the denominator is higher and the ratio of losses is lower. Of course if the situation is reversed then the denominator shrinks and the loss ratio becomes higher.
A growing portfolio can conceal a multitude of problems and care needs to be taken to analyse its true quality.
Concepts on the Measurement of Loss
Another difference in the calculations is not related to the lagging factor but to the denominators used. The lagged flow gives a loss rate as a percent of the lagged current balance while the accountants’ loss rate is this month’s written-off balance as a percent of this month’s not previously written-off balances. In both approaches, this is a simplification with the lagged calculation assuming receivables only from current bucket and the standard approach total receivables. In reality losses should be measured as a percentage of performing, interest earning assets.
The lagged calculation could be seen as more prudent as it looks at only the major portion of interest earning balances and removes the effects of growth. The accountant’s simplified view of the world has an account as either performing or written-off and confuses numerators and denominators.
A more accurate calculation, however, would use a risk weighting/impairment on all buckets as delinquent balances, although still considered as assets, have a much diminished potential for earning interest.
Conclusions
The challenging part of the risk analysis is not the preparation of the data, although inaccurate data makes any analysis impossible, it is in the interpretation of that data.
The whole management process is geared to producing a profitable portfolio and the credit process must be designed to be consistent with this. This means to have an understanding of what are the acceptable level of losses and knowing what are the revenues and costs associated with a given product.
Once that is established the whole lending process from underwriting through limit setting, maintenance and collections must function in order to achieve the acceptable level of risk.