As of 1 January 2005, Nedbank has to follow with the IFRS ( International Financial Reporting Standard ) attack to loan loss provisioning. In footings of IAS 39, Fiscal Instruments: Recognition and Measurement, an entity shall measure at each balance sheet day of the month whether there is any nonsubjective grounds that a fiscal plus or group of fiscal assets are impaired.

Damage is the hazard, or uncertainness, that some of the involvement or capital of a fiscal instrument may non be paid back in full. Therefore, an plus is considered impaired if and merely if, there is nonsubjective grounds of an impairment event happening after initial acknowledgment of the plus and that this loss event has an impact on expected future hard currency flows.

The proviso, which is set up, is the difference between the plus ‘s carrying sum and the present value of expected future hard currency flows discounted at the fiscal instruments original effectual involvement rate.

An income statement charge is so calculated, by taking into consideration the gap and shutting balances of the proviso demands every bit good as write-off sums.

IAS 39 farther stipulates that a differentiation should be made between specific and portfolio proviso sums. The bank must first assess whether nonsubjective grounds of impairment issues for single fiscal assets and so jointly for fiscal assets that are non separately important.

If there does non go out nonsubjective grounds of damage on an single footing for a fiscal plus, it is so pooled together for damage losingss, because some of the fiscal assets in the pool might go impaired subsequently on.

A differentiation is made between the undermentioned proviso sums:

Specific Commissariats: A proviso sum that is established against a loss in an single loan.

General Provision: A proviso sum, that ‘s established against concealed losingss that are known to be, but can non as of yet be recognised to single loans.

In Baura ( 2004:2 ) , it is stipulated that the aim of Basel II is to guarantee that a loaner has sufficient commissariats or capital to back up its expected losingss over the following 12 months and back up any unexpected losingss. The aim of IFRS is to guarantee that the fiscal statements adequately reflect the losingss that are incurred at the balance sheet day of the month.

Therefore, it is apparent that Basel II works on statistical modeling of expected losingss while IFRS, although leting statistical theoretical accounts requires a trigger event to hold occurred before they can be used. IAS 39 specifically provinces that losingss that are expected as a consequence of future events, no affair how probably, are non recognised ( Baura, 2004:4 ) .

The Provisions Department of Nedbank has two theoretical accounts, an existent monthly commissariats theoretical account and a forecast/budget theoretical account.

Each history is allocated to put on the line sections depending on the arrear count position of that history. Each hazard section relies on a specific expression in order to cipher the needed proviso. There are five different hazard sections:

IBNR ( Incurred, but non realised ) ( General )

Impaired ( Specific )

Default ( Specific )

Write-off ( 100 % Provided ) ( Specific )

Securities ( Specific )

The existent theoretical account runs on existent month terminal informations and uses passage rates, chances of default ( PD ‘s ) and loss given defaults ( LGD ‘s ) to cipher the proviso figure on the balance sheet per portfolio, per section.

The prognosis theoretical account, which is used for budgeting, efforts to foretell the entire outstanding balances at each month terminal for future periods. The chief inputs for this theoretical account are:

Base month balances

Forecasted growing Numberss

Full set of passage rates.

Provision sums are so calculated on the predicted outstanding balances.

## .

## 2. PROBLEM DESCRIPTION

In order to provide the reader with a comprehensive apprehension of the sum job faced by the client, a elaborate description of the working procedure followed by the Provisions Department is given:

The prognosis theoretical account is used for budget intents, and prognosiss proviso Numberss. These Numberss are captured in the General Ledger ( GL ) .

Actual proviso computations are conducted monthly on the existent motions in the retail book.

Each month the existent proviso Numberss are compared to the Numberss in the GL.

The difference between these two Numberss is therefore the IAS 39 diary accommodation.

The purpose of this undertaking is the decrease of the discrepancies between the forecasted and existent sums by bettering the current prediction theoretical account.

Three chief inputs are used for the prediction theoretical account. Each merchandise provides growing Numberss, the base is existent balances of the month chosen, and therefore the lone input go forthing infinite for accommodation is the set of passage rates.

Presently these rates are forecasted by taking the norm of the historical rates.[ 1 ]

An probe will therefore be performed by agencies of clip series analyses in order to heighten the prediction theoretical account.

## 3. Undertaking OBJECTIVES

In order to turn to the range of this undertaking the undermentioned aims must be achieved to add value to the Provisions Department of Nedbank.

Acquire an in-depth apprehension of IAS commissariats, both for the existent IAS computations and for the prediction of proviso Numberss.

Develop a Microsoft Excel theoretical account ciphering IAS proviso Numberss. This theoretical account needs to be highly user-friendly, as it will be used during preparation Sessionss, explicating the IAS proviso methodological analysis followed by Nedbank.

Research the feasibleness of heightening the current commissariats calculating theoretical account utilizing seasonal and cyclical effects.

If the research indicates that the theoretical account ‘s truth can be improved the sweetenings should be implemented into the current SAS theoretical account.

## 4. Undertaking METHODOLOGY

## 4.1. Excel Model: Actual Provision Calculations

## 4.2. Forecasting With Seasonal Effectss

## Nedbank ‘s current prediction theoretical account operates in SAS where axial rotation rates are forecasted by taking norms from past values. Averages taken scope from 12/18/24 months.

The undermentioned methodological analysis illustrates the probe in order to see if important sweetenings can be made in the prediction of bad debt, if the axial rotation rate theoretical account is adjusted with seasonal effects.

Datasets contain many axial rotation rates, therefore to look into each one separately would be an tremendous and time-consuming undertaking. For the intent of this undertaking there was decided to work merely on selected forward rates and to use calculating techniques to these rates.

In order to find what the seasonal accommodations for associated months and quarters need to be, a silent person variable theoretical account is applied where dummy variables are allocated for associated months/quarters and mensurating the consequence of the associated month/quarter.

With the associated seasonal accommodations in topographic point the prediction of the forward rates can continue. The average prognosis is adjusted with the seasonal factors for the associated months/quarter.

## Correlation Structure

In order to recover the prognosiss of the remedy and remain rates we make usage of a correlativity construction. This entails the computation of correlativities between all the other rates in a hazard section with the forward rate. Through this construction all rates are forecasted.

In order to measure whether the forecasted values are an sweetening with regard to the current prediction system back proving is performed over a period of 1 twelvemonth, where comparings can be made against existent values.

## 5. LITERATURE STUDY

Minimal sample size demands for seasonal prediction theoretical accounts. ( Hyndman & A ; Kostenko, 2007 )

## 6. Result

6.1 Vehicle Asset Finance ( VAF )

/*Description new wave VAF Boek*/

## The VAF Dataset

The undermentioned merchandise splits are present in the VAF dataset:

Client Group 1 ( CG1 ) : Private Banking

Client Group 2 ( CG2 ) : Personal Banking

Client Group 3 ( CG2 ) : Markets

Client Group 4 ( CG2 ) : Enterprises

Client Group 5 ( CG2 ) : Nedbank Staff

NedEnterprise ( NED )

Old Mutual Bank ( OMB )

South African Scottish ( SAS )

The undermentioned hazard pails are used for VAF.

0: IBNR30

1: IBNR60

2: IMPAIRED60

3: Default

4: 100 % PROVIDED

Therefore in the dataset a axial rotation from the Client Group 3 IBNR60 hazard pail to the Client Group 3 IMPAIRED60 hazard pail is indicated by CG3CG3 $ 001002.

During the analysis of these axial rotation rates there will be mention to bring around, remain and send on rates.

A forward rate implies a axial rotation rate where the balance moves to a worse pail.

E.g. CG2CG2 $ 002003. This forward rate implies a axial rotation from the IMPAIRED60 hazard pail into the DEFAULT hazard pail.

A stay rate implies that the balance remains in the same pail. E.g. CG1CG1 $ 000000. This stay rate implies a axial rotation from the IBNR30 hazard pail to the IBNR30 hazard pail.

A remedy rate implies that the balance moves from a worse hazard pail to a better hazard pail. E.g. NEDNED $ 002001. This remedy rate implies a axial rotation from the IMPAIRED60 hazard pail to the IBNR60 hazard pail.

The dataset consists of monthly axial rotation rates with 77 informations points runing from January 2001 to June 2007. The notation used in the dataset is ROLL02040205.

ROLL & lt ; Year & gt ; & lt ; Month the balance is turn overing out of & gt ; & lt ; Year & gt ; & lt ; Month the balance is turn overing into & gt ;

ROLL02040205: The twelvemonth 2002.

ROLL02040205: Calendar month: the balance is turn overing out of – April.

ROLL02040205: The twelvemonth 2002.

ROLL02040205: Calendar month: the balance is turn overing into – May.

## Dimension Reduction

The original dataset consisted out of 176 axial rotation rates. In order to cut down the dimension we decided to analyze merely the forward rates for different merchandise groups. All other rates were so deduced from the forecasted forward rate through a correlativity construction. With the debut of this measure the figure of axial rotation rates which required intense analysis and prediction was reduced to 28.

Each associated forward rate will be treated as a clip series since each axial rotation rate is recorded at a specific clip t. These clip series are considered as a distinct clip series.

## VAF Client Group 1 ( CG1 ) : Private Clients

The client group 1 – Private Clients,

## Measure 1: Designation and Graphical Analysis

CG1CG1 $ 002003

## VAF Client Group 2 ( CG2 ) : Personal Banking

## Measure 1: Designation and Graphical Analysis

CG2CG2 $ 000001

CG2CG2 $ 001002

CG2CG2 $ 002003

## Measure 2: Appraisal of Seasonal Adjustments

After application of the silent person variable theoretical account, the undermentioned seasonal accommodations were obtained:

Table 1: Seasonal Adjustments for the forward rates ( CG2 )

## VAF Client Group 3 ( CG3 ) : Markets

## Measure 1: Designation and Graphical Analysis

## CG3CG3 $ 000001

Figure 1: Axial rotation Rates – IBNR30 to IBNR60

From Figure 1 it can be depicted that a displacement in the average degree has occurred from January 2005. In order to set the prediction theoretical account with seasonality, one would wish to see similar increases/decreases in specific months.

Figure 2: Axial rotation Rate into December

From Figure 2 it is apparent that the specific accommodation for December can non be deterministic. From this graph an statement might be that accommodations should be made in two twelvemonth intervals.

Figure 3: ACF and PACF Plots

From the initial graphical analysis, no evident forms can be detected with the oculus. Let ‘s turn to a more scientific attack in order to see if seasonality is present in the information.

## CG3CG3 $ 001002

Figure 4: Axial rotation Rates – IBNR60 to IMPAIRED60

From Figure 4 it can be seen that, apart from the spikes happening in the beginning of 2005, the axial rotation rate from IBNR60 to Impair 60 is comparatively stable. A important addition can be seen from October 2006 to February 2007. From this graph a possible probe arises, since additions in the forward rate occurs rather a few times over the December months.

## CG3CG3 $ 002003

Figure 5: Axial rotation Rates – IMPAIRED60 to Default

From Figure 5 the spikes at Roll02060207 and Roll02120301 are noticeable since they are the lowest points in the clip series, whereas the remainder of the clip series operates in the same assurance interval. With the oculus, seasonal forms are non that easy noticeable, therefore we will hold to turn to statistical processs to observe whether seasonality is present.

## Measure 2: Appraisal of Seasonal Adjustments

After application of the silent person variable theoretical account, the undermentioned seasonal accommodations were obtained:

Table 2: Seasonal Adjustments for the Forward Rates ( CG3 ) .

The accommodations in table 2 are quarterly accommodations. The quarters are defined as follows:

One-fourth 1: December, January and February

One-fourth 2: March, April and May

One-fourth 3: June, July and August

One-fourth 4: September, October and November

To contrast these estimations performed on 77 informations points, with the estimations which was obtained with merely 24 informations points the decision can be made that much better consequences were obtained with more informations points.

Table 3: Quarterly Seasonal Adjustments performed on 2 old ages informations

From the tabular array above it can be depicted that the seasonal accommodations do non do sense from a concern position. When the 0 – 1 ( IBNR30 – IBNR60 ) pail is investigated the estimations indicate a 10.32 % addition in the first one-fourth, but in the 2nd one-fourth a comparative lessening of -8.44 % is suggested. However, from a concern position it is unrealistic, since the first half of the twelvemonth should be seen as an addition in the mean forward rates.

Figure 6: Percentage Seasonal Adjustment – CG3CG3 $ 000001

However, better consequences are seen in the estimations calculated from 77 informations points. For CG3CG3 $ 000001 ( plotted in the top figure ) a lessening is expected in the declining rate from December through to May. And so a lessening is expected in the months from June till November.

Figure 7: Percentage Seasonal Adjustment – CG3CG3 $ 001002

For the axial rotation rate from IBNR60 to IMPAIRED60 a positive seasonal accommodation is verified in the first and 4th quarters ( i.e. this implies that through the rhythm from September to February the mean theoretical account should be adjusted positively with an upward seasonal consequence ) . The lone badgering factor is the seasonal addition that starts in September. This is unexpected from a concern position.

During the months from March to August ( i.e. one-fourth 3 and 4 ) a comparative lessening of -5.32 % and -8.11 % will be applied to the norm.

Figure 8: Percentage Seasonal Adjustment – CG3CG3 $ 002003

The seasonal accommodations for the axial rotation from the IMPAIRED60 hazard pail into the DEFAULT pail can originate some concern but this is problematic. The estimations presented suggest a lessening of -3.24 % in the first one-fourth and merely an addition in the 2nd one-fourth ( i.e. from February to May ) of 9.08 % . It is expected that if a important upward seasonal accommodation exists

## Measure 3: Prediction

The current SAS theoretical account predicts future axial rotation rates by taking the norm of the old 12 months and bring forthing mean prognosiss. In the probe to heighten the current prediction theoretical account, the seasonal accommodations calculated in Step 2 will be applied to set the mean with the associated per centums

Figure 9: CG3CG3 $ 000001 – Seasonal Forecast V. Mean Forecast

Figure 10: CG3CG3 $ 001002 – Seasonal Forecast V. Mean Forecast

Figure 11: CG3CG3 $ 001002 – Seasonal Forecast V. Mean Forecast

## Measure 4: Establish the correlativity construction

In order to infer all the axial rotation rates in the associated hazard buckets the correlativities the specific rates have with the forward rate is calculated.

## CG3CG3 $ 000001

Table 4: Correlations with CG3CG3 $ 000001

From the above tabular array it can be seen that a strong correlativity exists between the forward rate ( CG3CG3 $ 000001 ) and the stay rate ( CG3CG3 $ 000000 ) which is equal to -50.123 % . The negative mark indicates that an opposite relationship exists between the stay and forward rate. This makes sense from a concern position, i.e. these two axial rotation rates will travel in opposite waies.

The reader might be interested how a axial rotation rate e.g. CG3CG3 $ 000002 might be possible ( i.e. a balance turn overing from the IBNR30 hazard pail to the IMPAIRE60 hazard pail in 1 month ) . These uneven minutess happen seldom but do occur, therefore they need to be incorporated in the construction. When measuring the existent values of these axial rotation rates it can be seen that they are really little, i.e. merely a little per centum of the balances follow this behavior.

Figure 12: Remedy Rate ( CG3CG3 $ 00100 ) – Seasonal Forecast V. Mean Forecast

Figure 13: Stay Rate ( CG3CG3 $ 001001 ) – Seasonal Forecast v. Mean Forecast

From the above two figures it can be seen that through the correlativity construction, seasonality effects are still catered for in the remedy and remain rate, by infering it from the forward rate multiplied with the associated correlativity.

## CG3CG3 $ 002003

Figure 14: Remedy Rate ( CG3CG3 $ 002001 ) – Seasonal Forecast V. Mean Forecast

From the above Figure, it can be seen that the remedy rate is adjusted with seasonality through the correlativity construction. It corresponds to the mean overall, but a negative accommodation is made between March and May.

## Measure 5: Estimating the Income Statement Charge

Figure 15: Balance Sheet Provision Requirement – CG3 Arrear 1 Bucket

Figure 16: Balance Sheet Provision Requirement – CG3 Arrear 2 Bucket

Figure 17: Balance Sheet Provision Requirement – CG3 Arrear 3 Bucket

Figure 18: Entire Provision Amount Required for VAF

Figure 19: Predicted I/S Charge for VAF CG3

From the above figure it can be seen that an addition in the predicted I/S charge is anticipated in the first two quarters of the twelvemonth ( i.e. from December through to May ) .

Table 5: Predicting the I/S Charges of the Seasonal Model V. Mean Model

## VAF Client Group 4 ( CG4 ) : SBS ( Small Business Services )

## Measure 1: Designation and Graphical Analysis

CG4CG4 $ 000001

CG4CG4 $ 001002

CG4CG4 $ 002003

## VAF Client Group 5 ( CG5 ) : Nedbank Staff

## Measure 1: Designation and Graphical Analysis

CG5CG5 $ 000001

CG5CG5 $ 001002

CG5CG5 $ 002003

## VAF NedEnterprise ( NED )

## Measure 1: Designation and Graphical Analysis

NEDNED $ 000001

NEDNED $ 001002

NEDNED $ 002003