IES20010392A2 - Financial Data Processing - Google Patents

Abstract

A financial data processing system for processing data relating to monthly loan repayments by clients. Each client has a respective client memory 10 containing general client information and data relating to the progress of the payments. A risk curve memory (42) contains, for each month, a corresponding risk factor indicating the risk of the client going into arrears. Each month, the sum due that month is multiplied (41) by the risk factor for that month to obtain an advanced arrears write-off. The system may also include means (20-33) for determining, if the client goes into arrears, advanced default write-offs in advance of actual default. <Figure 1>

Description

The present invention is concerned with financial data processing, and more specifically with the management of installment loans by financial institutions.

Background Before describing the present system, it is desirable to outline the financial concepts which form the background to the system.

A loan consists in essence of the payment of a sum of money to the 15 borrower, who is expected to pay it back (with interest) over a set period by means of regular payments. Consider for convenience a loan of £ 48000 repayable in 48 monthly payments of £ 1000. (The £ 48000 is given by the product of the monthly payments and the number of months; the sum actually lent will be less than that, because the repayments include interest.) Ideally, the borrower will make all the monthly payments due, and will make them on time. However, there is the possibility that at some point the borrower will fail to make a payment due; if that happens, the debt is described as “delinquent”. It is convenient to divide delinquent debts into two classes. For a 25 fixed period, which will be taken to be 6 months, the borrower is regarded as “in arrears”. During this “in arrears” period, the borrower may manage to resume making the payments (additional interest and penalties being due, of course). If, however, the borrower fails to resume servicing the debt during the in arrears period, the debt is then described as “defaulting”. Once a debt goes into default, legal proceedings are started in an attempt atfrecovery. l ''' .....

OPEN TO PUBOC INSPECTION UNDER SECTION 28 AND RULE 23 JHlNo. J3£a_OF oC f- i 4-j iio \ The way that these possibilities are treated by the lender needs to be considered.

The simplest technique is for the lender to treat the loan as good until the 5 point of final default. At that point, an estimate can be made, based on previous experience, of the chance of recovery. The debt is then written down to estimated recovery value (ie the chance of recovery times the amount of the debt). This gives a bad debt value, which is amount by which the debt is written down, ie the value of the debt less the estimated recovery value. At that point, also, the bad debt (the amount written off) can generally be offset against profits for taxation purposes.

However, an alternative earlier-stage procedure and a system which implements it have been proposed. When a loan goes into arrears, the lender can assign a definite probability for the debt either being eventually paid off or going into final default. (This probability is determined by the lender from their experience of previous loans.) The effective value of the debt is then the outstanding loan value times the probability of the debt being eventually paid off. As a consequence, a provision for write-off value can also be calculated, as the outstanding loan value less that effective value; this is by definition equal to the value of the debt times the probability that it will go into default. Using this technique allows the provision for write-off value to be offset against profits for tax purposes at an earlier time. An example of this system is described in ZA 01/7749.

This system may be described as an advanced default loss provisioning system, because it deals with bad debts which arise as a result of default but provides against part of such debts before the point of default. The system provides a more satisfactory treatment of the risk associated with aging of arrears.

It is somewhat artificial when a single debt is considered, as a single debt either goes into default or is paid off; a single debt cannot split into a part which is paid off and a part which goes into arrears. But when the lender combines the figures for a large portfolio of debts, this approach provides a fairer view of the situation than the simple technique noted above.

Of course, some of the loans going into arrears will be paid off eventually, so the provisions against those loans will have to be cancelled eventually. But the rest of the loans going into arrears will default. For such a loan, the actual loss will be larger than the provision calculated when the loan initially went into arrears. So for those loans, there will be an additional provision for write-off, calculated at the point where the loan defaults, which will be the difference between the actual loss and the provision for write-off calculated when the loan went into arrears. But provided that the lender has determined the probabilities correctly, the write-offs to be cancelled and the additional write-offs should balance each other out. (Any discrepancies will of course have to be added to or subtracted from profits or losses.) The present invention provides a further improved approach.

According to the invention there is provided a financial data processing system for processing data relating to periodic payments by clients, comprising: for each client, a respective client memory containing general client information and data relating to the progress of the payments; a risk curve memory containing, for each period of payment, a respective risk factor; and means for multiplying, for each period, the sum due in that period and the risk factor for that period to determine an advanced arrears loss provision.

This system may be briefly described as an advanced arrears loss provisioning system.

SE Ο 1 0 3 9 2 A financial data processing system embodying the invention will now be described, by way of example, with reference to the drawings, in which: Fig. 1 is a risk curve; Fig. 2 is a block diagram of the system; and Fig. 3 shows a client record unit in more detail.

The Risk Curve The Applicants have carried out various studies into the pattern of arrears io emergence in the repayment of loans. Taking (as above) 4-year loans repayable monthly, it has been found that there is a typical pattern for the emergence of arrears, which can be summarized by a risk table and represented by a risk curve. Fig. 1 shows a typical risk curve, with the horizontal axis representing time (specifically, the 48 monthly intervals for repayment). The vertical axis shows, for each month of the loan, the percentage of loans which first fall into arrears in that month (equated to the probability of a loan falling into arrears in that month).

This curve is roughly in the shape of a single hump, with the peak of the hump occurring roughly a third of the way through the loan period. The curve starts off with a small secondary peak, reflecting situations where the loan never takes off - for example for return of merchandise. After this initial small secondary peak, the curve drops sharply to a low level. Thus few arrears emerge in the first several months of the loan. This, perhaps, reflects the positive equity recognized by the borrower in merchandise purchased when using installment finance. A relatively high level of arrears typically concentrates towards the end of the first half of the loan period. This is likely to reflect the borrower’s perception of declining equity, and, possibly, their business failure. Towards the end of the loan period, the incidence of arrears declines again. This is considered to reflect the borrower’s desire to clear the loan without trouble combined with their probable ability to do so without undue difficulty.

The significance of these observations is that the measurement of latent risk in non-delinquent debt must have due reference to the age of the portfolio. A lender, accordingly, should not deduce from an apparent low level of arrears on new lending that such loans are somehow of a superior quality to existing, relatively mature, loans. Conversely, a lender should not deduce from a rising level of emerging arrears that a mature portfolio is suddenly declining in quality the increasing maturity of the loans must be allowed for.

More generally, if a lender wishes to relate an existing level of arrears and write-off to the vast bulk of well performing loans (typically, over 97% of the total outstanding), they must have reference to the portfolio loan maturity time and to its current average age. This enables them to calculate how much of the risk has presented to date and how much remains to emerge.

The same principles apply, of course, to loans for longer and shorter maturities, with appropriate modifications of the risk curve.

The Applicants have conducted confidential studies of the loan portfolios of a number of banks, and from analysis of the results, a surprising degree of consistency in the rate of arrears emergence has been found, in conformity with the risk curve pattern just discussed. This consistency of arrears emergence noted across a number of banks is likely to partially reflect relatively homogenous lending policies, e.g. merchandise, financial products security, etc. With more diverse polices and diverse merchandise, etc, the risk emergence pattern may well vary between different lenders. Accordingly, wherever possible, the lender’s own risk curve should be identified. If a lender is unable to do this and deploys another institution’s risk curve, then this should be done with due caution.

The present system - concepts Returning to the treatment of loans in more detail, the risk curve means that provision should be made for default even on loans not in arrears. For such loans, there is a chance (reflected by the risk curve) that they will fall into arrears (the chance that they will then default is then determined as before). So provisions for write-offs on the loan should start straight away. These provisions for write-offs occur earlier than the provisions for write-offs, discussed above, which are calculated when a loan goes into arrears. The present provisions for write-offs are, like those discussed above, associated with specific loans, and can therefore also generally be offset again tax on profits. The present system may be described as an advanced arrears loss provisioning system, because it deals with bad debts which arise as a result of arrears but makes provision for writing off part of such debts before the point of going into arrears.

The provision for write-off on a loan will of course need to be adjusted each month. Since the risk curve gives the risk per month, the provision for write-off has to be calculated for each month in turn, as the outstanding amount of the loan multiplied by the risk for that month.

Once the loan reaches the point where the risk curve starts to fall, the risk curve generated provisions for write-offs will start to fall. (In fact, they will start to fall shortly before the peak of the risk curve is reached, because the outstanding amount also decreases month by month.) So there will be an initial sequence of rising provisions for write-offs, followed by a period in which the provisions for write-offs are reducing.

If the loan actually falls into arrears, then a provision for write-off value can be determined as discussed above. But the loan will already have generated a risk curve (advanced arrears) loss provision before it fell into arrears, in accordance with the use of the risk curve as just discussed. The arrears-generated (advanced default) loss provision therefore should either have the risk curve loss provision deducted from it, or experience an incremental provision requirement. -ί Considering the present system more generally, previous methods of estimating the likelihood of a customer defaulting on a loan concentrated on typing the customer to determine if they would default. However, none of these appear to assess in advance when a loan may go into arrears and subsequently result in a bad debt. The advanced default loss provisioning system, for example, is directed towards assessing the proportion of loans going into arrears that would result in a bad debt, but not when the loans would go into arrears.

An important advantage provided by the present system is that, from knowledge of the likelihood at a given time of a loan going into arrears and from that the likelihood of a bad debt provision needing to be made, specific accounts in a portfolio can be identified as giving rise to bad debts in a particular tax year, so enabling a tax provision to be made for those accounts in a current tax year rather than a later tax year when the accounts have actually been finally written off.

Thus the principal features of the present system relate to the calculation of risk on well performing debt, in contrast to the known systems which ignore risk either until final default or until actual arrears occur.

The discussion so far has been in terms of a loan of £ 48000 for 48 months. Obviously for loans of other periods, the risk curves will have to be adjusted appropriately. A simple time scaling of the risk curve will generally be reasonably satisfactory, but there may also be slight differences of shape for the risk curves for different lengths of loan. Similarly, for different sizes of loan; using a common risk curve will generally be reasonably satisfactory but there may also be slight differences of shape for the risk curves for different sizes of loan. Also, different loans may be of different types - eg whether the loan is for hire purchase or leasing, the type of merchandize for which the loan is made, the type of person or organization to which the loan is made, their geographical location of that person or organization, their credit rating, their age (if they are an individual rather than an organization), whether they are a new customer or not, and so on.

There are 2 main ways of dealing with such factors. One is to carry out a statistical analysis for the factors, and generate a risk curve accordingly. This can achieve high accuracy, but it requires a sufficiently large sample for the statistical analysis to be reasonably reliable. The other is to assign risk coefficients to the various factors, use these with suitable weightings to generate a composite risk, and use that composite risk to scale the risk curve. Of course, both methods can io be combined, by using statistical analysis for some factors and risk coefficients for others.

The present system - implementation Fig. 2 is a general block diagram of a system for calculating combined advanced arrears and advanced defaults. The system operates on a monthly cycle.

The system comprises a memory 10 which comprises a set of records of the clients (borrowers). As shown in Fig. 3, for each client, the record comprises a general section 11 containing client details, loan arrangements, etc, and a payment section 12 which consists of 48 locations (for the 48 months of the loan), each location containing a month number (column 13) and an amount (column 14). The payment section 12 includes a location 15 for the current month, a series of locations 16 for past months recording actual payments made for those months, and a series of locations 17 for future months recording the payments due in those months.

The client record 20 as just described is for a single-loan client. If a client has several loans, then a separate record will be needed for each loan, but the client details can of course be common to the different records.

The client records 10 are coupled to a set of three 1-month arrears accumulators 20-22 via a scan unit 23. Accumulator 20 is for the current month, and contains 6 registers, covering the 6-month arrears period described above. The scan unit scans each client record in turn, extracting the arrears (if any) for the last 6 months and passing them to the accumulator 20, which thus accumulates a set of 6 arrears values representing the total monthly arrears for the 6-month arrears period.

Considering first the components used to generate advanced defaults, io accumulator 20 feeds accumulator 21, which in turn feeds accumulator 22, so that the sets of arrears for the previous 2 months are available as well as the set of arrears for the current month. The accumulators 20-22 feed a quarter arrears register 23 through an adder 24. The register 23 contains 6 arrears values, each being the sum of the 3 arrears values for the corresponding locations in the accumulators 20-22. The use of register 23 smooths the values over a rolling quarter (3-month period), thereby reducing the effect of short-term statistical variations of client behaviour.

The contents of the quarter arrears register 23 are used to determine the provision for write-off values to be used for the various months of the arrears, as follows. The value in the current month (6th month) location of the accumulator 20 is the figure to be written off (subject to a reduction for expected legal recovery). Consider the figure in the 5th-month location of this register. Because the sums making up this figure are have not reached the end of the default period, we can expect some of them to be paid before the 6th month. The ratio of the 6th-month value to the 5th-month value is thus the fraction of the 5th-month value which we can expect not to be paid. This is the value which we can therefore make provision to write off in the 5th month. As noted above, instead of using the single-month values, the smoothed values from register 23 are used to calculate the fractions.

Specifically., a calculator unit 30 determines the ratio of the 6th-month location to the 5th-month location in the register 23. This fraction is fed to a calculator unit 33, which is also fed with the current month’s value (the 6th-month in arrears value) from register 20, and reduces that value by that fraction. The output from unit 33 is the provision for write-off value for the 5-months-in-arrears debts. A scan unit 31 then steps on in the register 23 to extract the values in the 5th-month and 4th-month locations, and the ratio of these is determined by the unit 30. This ratio is decreased by the previous ratio, which is fed back into unit 30, . so that the ratio fed by the unit to the unit 33 is the ratio of the 6th-month location to the 4th-month location. The 4th month arrears value is selected by unit 32 from the register 20 and reduced by the ratio from unit 30 to produce the provision for write-off value for the 4th month. The process is repeated for the arrears values for the remaining months.

Turning now to the components used to generate advanced arrears loss provisions, a risk curve memory 42 stores the risk curve in tabular form, as the series of monthly risks. The current month locations 15 in the client records 10 are selected in turn by a scan unit 40. For each client, the current month number is passed to the risk curve unit 42, this unit producing the corresponding risk value. The current value of the loan is also extracted from the client record, and passed to a multiplier 41, where it is multiplied by the risk value to generate the advanced arrears loss provision for the current month.

The advanced arrears loss provision is passed to a subtractor 43 whose output is circulated back to its other input. This will generate the incremental risk for the month, ie the risk for the current month less the risk for the previous month.

To combine the effects of the risk curve (advanced arrears) loss provisions with those of the advanced default loss provisions, the risk curve provision for write-off must be fed into the advanced default loss provision circuitry in the event of an actual default on the loan. This may be done, for example, by feeding the advanced arrears loss provision and the advanced default loss provision to a logic unit 44 which is controlled by a control signal CONT which is generated if the loan goes into arrears. The control unit normally passes the advanced arrears loss provisions, but if the loan goes into arrears, it passes the advanced default loss provisions, with the last advanced arrears loss provision being subtracted from the first advanced default loss provision.

In practice, the size of the client records will of course be dependent on the io length of the loan period. Similarly, there will normally be a variety of risk tables for different types and lengths of loan, means for adjusting the risk table figures in dependence on the risk assessment, etc, as discussed above.

It will of course be understood that the system may be implemented by 15 means of any convenient form of data processing apparatus suitably arranged to store and perform the necessary data and calculations.

Claims (5)

Claims

1 A financial data processing system for processing data relating to periodic payments by clients, comprising; for each client, a respective client memory containing general client information and data relating to the progress of the payments; a risk curve memory containing, for each period of payment, a respective risk factor; and ~~ means for multiplying, for each period, the sum due in that period and the risk factor for that period to determine an advanced arrears write-off.

2. A financial data processing system according to claim 1 further including: means for determining the proportion of non-recoverable debt at a specified final age; means for computing, for each age of debt, a first moving average of total debts outstanding at the end of each of a number of immediately preceding periods; means for computing a second, corresponding, moving average of total debts outstanding one period earlier and of one period lesser age; and means for multiplying the non-recoverable debt proportion for each age, starting with the final age, by the corresponding ratio of the second to the first moving average, to obtain an estimate of the proportion of non-recoverable debt for the next earlier age.

3. A financial data processing system according to either previous claim in which the payment period is 1 month.

4. A financial data processing system according to claim 3 in which the final age is 6 months.

5. A financial data processing system according to either of claims 3 and 4 in which the moving average is taken over 3 months.