13   Attribution Analysis

Attribution Analysis enables institutions and regulators to monitor the quality and health of financial institutions through identifying reasons for higher provisions, which result from a varied range of factors including but not limited to strategic decisions, quality of customers, and economic circumstances.

The Attribution Analysis feature offered in the OFS Loan Loss Forecasting and Provisioning application enables you to compare the results of two Runs (Processes) and understand the contribution of each factor that is involved in the computation process to the change in the result value.

The attribution analysis component is integrated into the OFS AAI application's Process Modelling Framework (PMF). PMF is a design and execution framework that enables Process Pipeline developers to implement various pipelines modeled by business analysts. PMF consists of process modeling components for modeling pipelines and process monitor components to monitor instantiated pipelines of OFSAA applications. For detailed information about the usage of PMF, see the OFS Analytical Applications Infrastructure Process Modeling Framework Orchestration Guide.

Attribution analysis requires access to OFSAA and defining the attribution pipeline in PMF. Preparation of an attribution analysis definition and execution of the attribution pipeline are the final steps in attribution analysis. For detailed information about the usage of Attribution Analysis, see the OFS Attribution Analysis User Guide.

Specific to the Expected Credit Loss (ECL) computation, the Attribution Analysis process compares the account level ECL of two dates or Runs and computes the contribution of any or all factors to the actual change in ECL either in the same date or different dates. Later the reasons for the change in ECL are also calculated. The sum of all such contributions by each factor sums up the actual change in the ECL.

You can derive insights about changes in ECL potentially caused by any or all the following causes:

·        Change in IFRS 9 Stage

·        Change due to Carrying Amount

·        Change due to Undrawn Amount

·        Change in Credit Conversion Factor

·        Change in the Probability of Default (Marginal or Cumulative)

·        Change in the Loss Given Default

·        Change in the Exchange Rate

·        Change in the Provision Rate

·        Change due to Cash Flows or Forward Exposures

·        Change due to Effective Interest Rate

·        Change in the Time Factor (Bucket)

·        Historical Cash Flows changes (Cash Flows paid off)

·        Other changes are not covered in the preceding causes.

If an account is moved from one Classification category to another (reclassified), the account is treated as derecognized under the first category and newly recognized under the second category.

Detailed Attribution Analysis is performed for CVB and Collateral dependent methods, based on Mitigant Effect on Specific Provision, Forward Exposure, and CVB. Specific Provision and Forward Exposure methods use collateral. New factors such as Mitigant Value, Amortized Cost, and Deferred Balance are introduced and are used for Attribution Analysis performed on Specific Provision, Forward Exposure, and CVB.

The variable to the column mapping must be additionally configured in the FSI_LLFP_ATTR_COMP_VAR_MAPPING table that allows multiple columns to a single variable. This configuration is required if you are adding additional variables apart from the out-of-the-box provided variables.

Execute Attribution Analysis

Execution of attribution analysis is a UI-driven approach. Oracle Financial Services Attribution Analysis is a feature that is integrated into the OFS AAI application's PMF. This component specifically addresses the requirement of financial institutions to compare the execution runs in models by varying one factor and studying the impact on other factors that are linked to it.

The comparison of the two executions, a base run instance and an inspect run instance, provides an insight into the impact of the outcome of the run when the variables are altered for every attribution run anchoring all but one variable, keeping all variables constant except one. You can create a group of tasks, runs, and models, and execute the processes configured in a parallel or a serial sequence as required. The results can be explored for analysis and reporting. For detailed information about the usage of Attribution Analysis, see the OFS Attribution Analysis User Guide.

Process Flow

The following figure depicts the process flow of the Attribution Analysis process.

Figure 18: The Reconciliation Process Flow

Attribution Analysis Process

The following is the attribution analysis process on execution:

1.     The Application obtains the values of all the relevant factors to compute the ECL from both dates.

2.     For each factor, at an account or cash flow granularity, the application compares the values on either date or runs and computes the change in ECL due to the change in the value of that specific factor.

3.     This is repeated for all factors that were selected by you in the application.

4.     After all the selected factors are attributed, the remaining difference in ECL values is attributed to Other Changes and Multiplier Effect.

Primarily, the factors are grouped under the following headers:

·        New Assets

·        Derecognized Assets

·        Remeasurement

·        Model or Risk Parameters

·        Stage Changes

·        Foreign Exchange and

·        Change in Exposure

·        Other Changes and Multiplier Effect

Reconciliation Reports

Reconciliation disclosures are intended to explain factors that are attributable to change in asset values over a period for attribution analysis. Because the ECL directly impacts profitability and therefore the financial health of a bank, management and regulators alike are interested in understanding what important factors caused a change in the ECL amount from one period to the next.

OFS LLFP application has a dedicated UI-driven run with a range of factors for you to choose and determine important factors to help explain or attribute changes in ECL.

ECL Reconciliation Analysis for Accounts

This report is a waterfall chart providing a visual description of the movements between two reporting dates. The tabular report is located following the waterfall chart that details various contents of the chart. Drills are accessible also on the tabular report apart from on the waterfall chart, see the following explanation:

·        The first and the last columns represent the run1 and run2 ECLs respectively.

·        Across the x-axis, factors are grouped based on similarity and provide a high view of reasons for changes between the runs. A group can further have individual contributing factors, the summation of their combined net effect per group is shown as a column on the waterfall chart.

·        A red column indicates the ECL increased between the two runs and green indicates a reduction in ECL.

·        If there are no columns visible for a group, say in as Derecognized Assets in the sample image as follows, it means that either constituent factors movements have canceled each other out or as a group, they did not impact in increase or decrease in ECL.

Figure 19: The ECL Reconciliation Analysis Report

This report can be drilled down from the column that opens up a separate report:

·        Remeasurement

·        Change in Exposure

·        Stage Changes

·        Model or Risk Parameters

Each drill-down opens a combination of a horizontal bar chart and a tabular report depicting ECL contribution per factor and its percentage contribution to the factor group.

Figure 20: The ECL Reconciliation Analysis Report

35 H and 35 I

In addition to the preceding report, the reconciliation feature also caters to the quantitative disclosure requirements of 35 H and 35 I report as per IFRS 7, but with more details or granularity.

The report shows the roll forward of Expected Credit Loss and Carrying amount values, from the beginning period balance, through the various parameters or values affecting the ECL and carrying amounts to the ending balances of the current period. The report is designed in a manner that allows you to view it in either a simple format or a detailed one. The reporting lines are created with hierarchies that allow you to expand a reporting line to view the subsections, in a rolldown or rollup format. The rolldown format shows the roll forwards for every factor affecting the ECL or carrying amount, while the rollup view clubs many of these factors under major headers such as Model or Risk Parameters, Remeasurement, and so on.

The following table explains the list of reporting lines that are part of this report and the corresponding factors associated with it.

The following is a screenshot of the 35 H and 35 I report with sample data (for representational purposes only).

Figure 21: The detailed view of the 35 H and I Reports

Figure 22: The simple view of the 35 H and I Reports

Sample Computation Logic

The following is an example of a Loan account spread over two time periods with changes in various parameters. Assuming a PD or LGD approach for ECL computation, the following table shows various parameters related to the said account, including the ECL, computed against both dates.

Note that in the second period the account has faced a $200 Write-off in addition to a repayment of $100.

Based on the computations, the Expected Credit Loss (which needs to be provisioned for, in the Balance sheet) in period 1 is $35.00 which increases to $220.50 in period 2.

The reconciliation process parses through the given information for both the dates and identifies the change in ECL due to each parameter.

In this case, the actual change in ECL between the two dates is $185.50

The changes due to the various parameters are arrived at by computing the ECL with the date 1 values except for the given parameter, which is replaced with the Date 2 value.

Taking PD as an example:

With an account in Stage 2, ECL computed on Date 1 would have been $70.00

Now, replacing only the PD with Date 2 value, compute the ECL as follows:

·        Outstanding of Date 1 = $1000.00

·        LGD of Date 1 = 70%

·        PD of Date 2 = 45%

·        ECL computed = $315.00

·        Change in ECL due to PD = $315 - $70 = $245.00

The same process is repeated for other parameters.

Summing up the changes in ECL attributed to each of these factors, you can notice that the cumulative change is different from that of the actual change. This difference is due to the multiplier effect of individual changes and thereby posted under the header Cumulative or Multiplier Effect.

The waterfall report based on the given example is as follows.

Figure 23: The values in the Waterfall Report based on the Example

From the disclosure�s perspective, the 35 H and 35 I report values are as follows.:

Note that Other Changes include changes due to factors not selected in the reconciliation run and the multiplier effect. In this example, Other Changes includes the Multiplier Effect of - $73.50 and an additional impairment loss of $200 that equals the write-off amount. 

ECL Reconciliation Analysis for Credit Line or Facilities

This report is a waterfall chart providing a visual description of the movements between two reporting dates. The tabular report is located following the waterfall chart that details various contents of the chart.

·        The first and the last columns represent the run1 and run2 ECLs respectively.

·        Across the x-axis, factors are grouped based on similarity and provide a high view of reasons for changes between the runs. A group can further have individual contributing factors, the summation of their combined net effect per group is shown as a column on the waterfall chart.

·        A red column indicates the ECL increased between the two runs and green indicates a reduction in ECL.

·        If there are no columns visible for a group, say in as Derecognized Assets in the sample image as follows, it means that either constituent factors movements have canceled each other out or as a group, they did not impact in increase or decrease in ECL.

·        The below report displays only the lifetime ECL on Undrawn Amount data. 

Figure 24: The ECL Reconciliation Analysis Report

Credit Line Reconciliation Report

In addition to the preceding report, the reconciliation feature also caters to the quantitative disclosure requirements of the undrawn amount but with more details or granularity.

The report shows the roll forward of Expected Credit Loss and Undrawn amount values, from the beginning period balance, through the various parameters or values affecting the ECL and Undrawn amounts to the ending balances of the current period. The report is designed in a manner that allows you to view it in a detailed format.

The following table explains the list of reporting lines that are part of this report and the corresponding factors associated with it.

Use the Reconciliation ID drop-down list filter and select a value.

For every Credit Line Reconciliation report, a unique Reconciliation ID is generated for the selected Base Run or Run1 Date and run_skey and Inspect Run or Run2 Date and run_skey.

To  know the Run 1 and Run 2 details for a Reconciliation ID, execute the following script in the atomic schema:

Select distinct N_RECONCILIATION_ID,BR.FIC_MIS_DATE BASE_RUN_MIS_DATE, N_BASE_RUN_SKEY, IR.FIC_MIS_DATE INSPECT_RUN_MIS_DATE, N_INSPECT_RUN_SKEY from FCT_LLFP_CREDIT_LINE_RECON CLR INNER JOIN DIM_RUN BR ON BR.N_RUN_SKEY = CLR.N_BASE_RUN_SKEY INNER JOIN DIM_RUN IR ON IR.N_RUN_SKEY = CLR.N_INSPECT_RUN_SKEY;

Figure 25: The Reconciliation ID Filter

The following is a screenshot of the Credit Line Reconciliation Report:

Figure 26: The Credit Line Reconciliation Report