Methods and system for assessing loss severity for commercial loans

• US 8,065,214 B2
• Filed: 09/06/2005
• Issued: 11/22/2011
• Est. Priority Date: 09/06/2005
• Status: Expired due to Fees

• Current Assignee: General Electric Company
• Original Assignee: GE Corporate Financial Services, Inc. (General Electric Company)
• Inventors: Cole, Mary Kennedy, Ruia, Sundeep, Keyes, Tim Kerry
• Primary Examiner(s): Kalinowski; Alexander
• Assistant Examiner(s): Vyas; Abhishek
• Application Number:
  US11/220,011
• Publication Number:
  US 20070055595A1
• Time in Patent Office: 2,268 Days
• Field of Search: 705 35- 45
• US Class Current: 705/36.R
• CPC Class Codes: G06Q 40/00 Finance; Insurance; Tax strategies; Processing of corporate or income taxes
• CPC Class Codes: G06Q 40/06 Asset management; Financial planning or analysis

Abstract

A method for predicting expected and unexpected loss outcomes for a portfolio of loans is provided. The loans are issued by a lender to a plurality of borrowers. The method includes recording key account and risk attributes for a historical portfolio of loans, recording actual default and loss information for each borrower included within the historical portfolio of loans, and comparing the key account and risk attributes with the actual default and loss information over a period of time. The method also includes selecting a sample of loans from the historical portfolio of loans to determine loss drivers based on the comparison of the key account and risk attributes with the actual default and loss information, building a regression tree based model representing relationships between the loss drivers, and expected and unexpected loss outcomes for the historical portfolio of loans, and predicting the expected and unexpected loss outcomes for a second portfolio of loans using the regression tree based model and the loss drivers for the second portfolio of loans.

Claims

1. A method for predicting expected and unexpected loss outcomes for a portfolio of loans using a computer system coupled to a database, the loans issued by a lender to a plurality of borrowers, said method comprising the steps of:

recording key account and risk attributes in the database for a historical portfolio of loans issued by the lender, the historical portfolio of loans includes loans issued to borrowers that have experienced a financial default with the lender and an associated economic loss;

recording actual default and loss information in the database for each borrower included within the historical portfolio of loans;

comparing the key account and risk attributes with the actual default and loss information over a predetermined period of time, the comparison is performed by the computer system;

dividing the historical portfolio of loans into a first sample of loans and a remaining hold-out sample of loans;

selecting the first sample of loans from the historical portfolio of loans, the first sample of loans including the key account and risk attributes;

determining loss drivers by using the computer system to electronically compare the key account and risk attributes of the first sample of loans with the actual default and loss information of the first sample of loans, the loss drivers being a set of the key account and risk attributes that are predictive of a loss;

building, via the computer system, a regression tree based model representing relationships between the loss drivers, and expected and unexpected loss outcomes using the first sample of loans, the building of the regression tree based model includes dividing the first sample of loans into buckets of expected and unexpected loss outcomes based on the loss drivers, each bucket showing corresponding expected and unexpected loss outcomes relative to a change in at least one of the loss drivers;

determining a final model by analyzing the remaining hold-out sample of loans with the regression tree based model and confirming the predictive accuracy of the final model, the final model representing a relationship between a set of loss drivers, and expected and unexpected loss outcomes; and

predicting the expected and unexpected loss outcomes for a second portfolio of loans using the final model and the set of loss drivers for the second portfolio of loans, the second portfolio of loans includes borrowers that have not experienced a default with the lender, the prediction is performed by the computer system.

2. A method according to claim 1 wherein recording key account and risk attributes in the database for a historical portfolio of loans further comprises:

recording variations in the key account and risk attributes in the database for the historical portfolio of loans over a predetermined period of time;

comparing the key account and risk attributes, and the variations in the key account and risk attributes with the actual default and loss information over a predetermined period of time; and

selecting the first sample of loans from the historical portfolio of loans to determine loss drivers based on the attributes and loss information comparison.

3. A method according to claim 1 wherein recording key account and risk attributes in the database for a historical portfolio of loans further comprises recording key account and risk attributes including at least one of the following financial measures for the corresponding borrower:

sales and EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortization), debt service coverage, liquidation coverage, senior debt multiple, capital structure, company leverage, default likelihood, industry sector, exposure and credit line, and macro-economic factors such as employment and interest rates.

4. A method according to claim 1 wherein predicting the expected and unexpected loss outcomes for a second portfolio of loans further comprises performing a Monte Carlo simulation analysis for the second portfolio of loans using model structures and assumptions made in building the regression tree based model from the determined loss drivers of the first sample of loans.

5. A method according to claim 1 wherein the expected loss outcome for each bucket is equal to a probability of default (PD) multiplied by a mean of a loss-given-default (mean LGD) for the bucket, and wherein the unexpected loss outcome for each bucket is a function of a mean of the PD, a standard deviation of the PD, the mean LGD, and a standard deviation of LGD for the bucket.

6. A method according to claim 1 wherein building a regression tree based model further comprises:

building the regression tree based model based on the determined loss drivers of the first sample of loans;

maintaining the hold-out sample of loans from the historical portfolio of loans, the hold-out sample includes a portion of loans from the historical portfolio of loans not used in building the regression tree based model; and

using the hold-out sample of loans to test predictive accuracy of the built regression tree based model.

7. A method according to claim 1 wherein predicting the expected and unexpected loss outcomes for a second portfolio of loans further comprises:

calculating an economic capital for the second portfolio of loans based on the predicted expected and unexpected loss outcomes;

recording the calculated economic capital on the balance sheet of the lender; and

using the calculated economic capital for at least one of pricing potential future loans issued by the lender, and for account and portfolio management optimization by the lender.

8. A method according to claim 7 wherein using the calculated economic capital for at least one of pricing potential future loans issued by the lender further comprises calculating expected returns for each potential future loan as a function of future transaction cash flows.

9. A method according to claim 1 wherein predicting the expected and unexpected loss outcomes for a second portfolio of loans further comprises:

generating default likelihood data for each borrower included within the second portfolio of loans;

combining the predicted expected and unexpected loss outcomes for each borrower included with the second portfolio of loans with the corresponding default likelihood data; and

generating a matrix having dimensions for each borrower that include a probability of default (PD) by a mean loss-given-default (mean LGD).

10. A network-based system for predicting expected and unexpected loss outcomes for a portfolio of loans issued by a lender to a plurality of borrowers, said system comprising:

a client system comprising a browser;

a centralized database for storing information; and

a server system configured to be coupled to said client system and said database, said server further configured to;

record key account and risk attributes in the database for a historical portfolio of loans issued by the lender, the historical portfolio of loans includes loans issued to borrowers that have experienced a financial default with the lender and an associated economic loss,

record actual default and loss information in the database for each borrower included within the historical portfolio of loans,

compare the key account and risk attributes with the actual default and loss information over a predetermined period of time,

divide the historical portfolio of loans into a first sample of loans and a remaining hold-out sample of loans,

select the first sample of loans from the historical portfolio of loans, the first sample of loans including the key account and risk attributes,

determining loss drivers by comparing the key account and risk attributes of the first sample of loans with the actual default and loss information of the first sample of loans, the loss drivers being a set of the key account and risk attributes that are predictive of a loss,

build a regression tree based model representing relationships between the loss drivers, and expected and unexpected loss outcomes using the first sample of loans, the building of the regressing tree based model includes dividing the first sample of loans into buckets of expected and unexpected loss outcomes based on the loss drivers, each bucket showing corresponding expected and unexpected loss outcomes relative to a change in at least one of the loss drivers,

determine a final model by analyzing the remaining hold-out sample of loans with the regression tree based model and confirm the predictive accuracy of the final model, the final model representing a relationship between a set of loss drivers, and expected and unexpected loss outcomes, and

predict the expected and unexpected loss outcomes for a second portfolio of loans using the final model and the set of loss drivers for the second portfolio of loans, the second portfolio of loans includes borrowers that have not experienced a default with the lender.

11. A system according to claim 10 wherein said server system further configured to:

record variations in the key account and risk attributes in the database for the historical portfolio of loans over a predetermined period of time,

compare the key account and risk attributes, and the variations in the key account and risk attributes with the actual default and loss information over a predetermined period of time, and

select the first sample of loans from the historical portfolio of loans to determine loss drivers based on the attributes and loss information comparison.

12. A system according to claim 10 wherein the key account and risk attributes comprise at least one of the following financial measures for the corresponding borrower:

sales and EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortization), debt service coverage, liquidation coverage, senior debt multiple, capital structure, company leverage, default likelihood, industry sector, exposure and credit line, and macro-economic factors such as employment and interest rates.

13. A system according to claim 10 wherein said server system further configured to perform a Monte Carlo simulation analysis for the second portfolio of loans using model structures and assumptions made in building the regression tree based model from the determined loss drivers of the first sample of loans, the simulation analysis is used for predicting the expected and unexpected loss outcomes for the second portfolio of loans.

14. A system according to claim 10 wherein the expected loss outcome for each bucket is equal to a probability of default (PD) multiplied by a mean of a loss-given-default (mean LGD) for the bucket, and wherein the unexpected loss outcome for each bucket is a function of a mean of the PD, a standard deviation of the PD, the mean LGD, and a standard deviation of LGD for the bucket.

15. A system according to claim 10 wherein said server system further configured to:

build the regression tree based model based on the determined loss drivers of the first sample of loans,

maintain the hold-out sample of loans from the historical portfolio of loans, the hold-out sample includes a portion of loans from the historical portfolio of loans not used in building the regression tree based model, and

using the hold-out sample of loans to test predictive accuracy of the built regression tree based model.

16. A system according to claim 10 wherein said server system further configured to:

calculate an economic capital for the second portfolio of loans based on the predicted expected and unexpected loss outcomes,

record the calculated economic capital on the balance sheet of the lender, and

use the calculated economic capital for at least one of pricing potential future loans issued by the lender, and for account and portfolio management optimization by the lender.

17. A system according to claim 16 wherein said server system further configured to calculate expected returns for each potential future loan as a function of future transaction cash flows.

18. A system according to claim 10 wherein said server system further configured to:

generate default likelihood data for each borrower included within the second portfolio of loans,

combine the predicted expected and unexpected loss outcomes for each borrower included with the second portfolio of loans with the corresponding default likelihood data, and

generate a matrix having dimensions for each borrower that include a probability of default (PD) by a mean loss-given-default (mean LGD).

19. A computer for predicting expected and unexpected loss outcomes for a portfolio of loans issued by a lender to a plurality of borrowers, said computer in communication with a database for storing information relating to each loan and each borrower, said computer programmed with instructions stored on a computer readable medium, the instructions, when executed, direct the computer to:

record key account and risk attributes in the database for a historical portfolio of loans issued by the lender, the historical portfolio of loans includes loans issued to borrowers that have experienced a financial default with the lender and an associated economic loss;

record actual default and loss information in the database for each borrower included within the historical portfolio of loans;

compare the key account and risk attributes with the actual default and loss information over a predetermined period of time;

divide the historical portfolio of loans into a first sample of loans and a remaining hold-out sample of loans;

select the first sample of loans from the historical portfolio of loans, the first sample of loans including the key account and risk attributes;

determining loss drivers by comparing the key account and risk attributes of the first sample of loans with the actual default and loss information of the first sample of loans, the loss drivers being a set of the key account and risk attributes that are predictive of a loss;

build a regression tree based model representing relationships between the loss drivers, and expected and unexpected loss outcomes using the first sample of loans, the building of the regression tree based model includes dividing the first sample of loans into buckets of expected and unexpected loss outcomes based on the loss drivers, each bucket showing corresponding expected and unexpected loss outcomes relative to a change in at least one of the loss drivers;

determine a final model by analyzing the remaining hold-out sample of loans with the regression tree based model and confirm the predictive accuracy of the final model, the final model representing a relationship between a set of loss drivers, and expected and unexpected loss outcomes; and

predict the expected and unexpected loss outcomes for a second portfolio of loans using the final model and the set of loss drivers for the second portfolio of loans, the second portfolio of loans includes borrowers that have not experienced a default with the lender.

20. A computer according to claim 19 wherein said computer is further programmed to:

record variations in the key account and risk attributes in the database for the historical portfolio of loans over a predetermined period of time;

compare the key account and risk attributes, and the variations in the key account and risk attributes with the actual default and loss information over a predetermined period of time; and

select the first sample of loans from the historical portfolio of loans to determine loss drivers based on the attributes and loss information comparison.

21. A computer according to claim 19 wherein said computer is further programmed to perform a Monte Carlo simulation analysis for the second portfolio of loans using model structures and assumptions made in building the regression tree based model from the determined loss drivers of the first sample of loans, the simulation analysis is used for predicting the expected and unexpected loss outcomes for the second portfolio of loans.

22. A computer according to claim 19 wherein the expected loss outcome for each bucket is equal to a probability of default (PD) multiplied by a mean of a loss-given-default (mean LGD) for the bucket, and wherein the unexpected loss outcome for each bucket is a function of a mean of the PD, a standard deviation of the PD, the mean LGD, and a standard deviation of LGD for the bucket.

23. A computer according to claim 19 wherein said computer is further programmed to:

calculate an economic capital for the second portfolio of loans based on the predicted expected and unexpected loss outcomes;

record the calculated economic capital on the balance sheet of the lender; and

use the calculated economic capital for at least one of pricing potential future loans issued by the lender, and for account and portfolio management optimization by the lender.

24. A computer according to claim 19 wherein said computer is further programmed to:

generate default likelihood data for each borrower included within the second portfolio of loans;

combine the predicted expected and unexpected loss outcomes for each borrower included with the second portfolio of loans with the corresponding default likelihood data; and

generate a matrix having dimensions for each borrower that include a probability of default (PD) by a mean loss-given-default (mean LGD).

25. A computer program embodied on a non-transitory computer readable medium for predicting expected and unexpected loss outcomes for a portfolio of loans issued by a lender to a plurality of borrowers, said program comprising at least one code segment that, when executed, prompts a user to input key account and risk attributes for a historical portfolio of loans issued by the lender and then:

records the key account and risk attributes in a database, the historical portfolio of loans includes loans issued to borrowers that have experienced a financial default with the lender and an associated economic loss;

records actual default and loss information in the database for each borrower included within the historical portfolio of loans;

compares the key account and risk attributes with the actual default and loss information over a predetermined period of time;

divides the historical portfolio of loans into a first sample of loans and a remaining hold-out sample of loans;

selects the first sample of loans from the historical portfolio of loans, the first sample of loans including the key account and risk attributes;

determines loss drivers by comparing the key account and risk attributes of the first sample of loans with the actual default and loss information of the first sample of loans, the loss drivers being a set of the key account and risk attributes that are predictive of a loss;

builds a regression tree based model representing relationships between the loss drivers, and expected and unexpected loss outcomes using the first sample of loans, the building of the regression tree based model includes dividing the first sample of loans into buckets of expected and unexpected loss outcomes based on the loss drivers, each bucket showing corresponding expected and unexpected loss outcomes relative to a change in at least one of the loss drivers;

determines a final model by analyzing the remaining hold-out sample of loans with the regression tree based model and confirms the predictive accuracy of the final model, the final model representing a relationship between a set of loss drivers, and expected and unexpected loss outcomes; and

predicts the expected and unexpected loss outcomes for a second portfolio of loans using the final model and the set of loss drivers for the second portfolio of loans, the second portfolio of loans includes borrowers that have not experienced a default with the lender.

26. A computer program in accordance with claim 25 further comprising at least one code segment that:

records variations in the key account and risk attributes in the database for the historical portfolio of loans over a predetermined period of time;

compares the key account and risk attributes, and the variations in the key account and risk attributes with the actual default and loss information over a predetermined period of time; and

selects the first sample of loans from the historical portfolio of loans to determine loss drivers based on the attributes and loss information comparison.

27. A computer program in accordance with claim 25 further comprising at least one code segment that performs a Monte Carlo simulation analysis for the second portfolio of loans using model structures and assumptions made in building the regression tree based model from the determined loss drivers of the first sample of loans, the simulation analysis is used for predicting the expected and unexpected loss outcomes for the second portfolio of loans.

28. A computer program in accordance with claim 25 wherein the expected loss outcome for each bucket is equal to a probability of default (PD) multiplied by a mean of a loss-given-default (mean LGD) for the bucket, and wherein the unexpected loss outcome for each bucket is a function of a mean of the PD, a standard deviation of the PD, the mean LGD, and a standard deviation of LGD for the bucket.

29. A computer program in accordance with claim 25 further comprising at least one code segment that:

calculates an economic capital for the second portfolio of loans based on the predicted expected and unexpected loss outcomes;

records the calculated economic capital on the balance sheet of the lender; and

uses the calculated economic capital for at least one of pricing potential future loans issued by the lender, and for account and portfolio management optimization by the lender.

30. A computer program in accordance with claim 25 further comprising at least one code segment that:

generates default likelihood data for each borrower included within the second portfolio of loans;

combines the predicted expected and unexpected loss outcomes for each borrower included with the second portfolio of loans with the corresponding default likelihood data; and

generates a matrix having dimensions for each borrower that include a probability of default (PD) by a mean loss-given-default (mean LGD).