Method and apparatus for an integrative model of multiple asset classes

• US 7,024,388 B2
• Filed: 06/29/2001
• Issued: 04/04/2006
• Est. Priority Date: 06/29/2001
• Status: Expired due to Term

• Current Assignee: Barra Incorporated
• Original Assignee: Barra Incorporated
• Inventors: Stefek, Daniel, Goldberg, Lisa Robin, Hui, Ken Chorlam, Scheffler, Scott Steven, Torre, Nicolas Goodrich
• Primary Examiner(s): Knight, Anthony
• Assistant Examiner(s): Holmes, Michael B.
• Application Number:
  US09/895,605
• Publication Number:
  US 20030110016A1
• Time in Patent Office: 1,740 Days
• Field of Search: 705/35, 705/36, 705/37, 705/38, 705/7, 705/4, 706/13, 706/62, 706/925, 706/45, 706/46, 706/48, 706/932
• US Class Current: 705/38
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof
• CPC Class Codes: G06Q 40/08 Insurance
• CPC Class Codes: Y10S 706/925 Business
• CPC Class Codes: Y10S 706/932 Mathematics, science, or engineering

Abstract

The invention provides a method and apparatus for combining two or more risk models to create a risk model with wider scope than its constituent parts. The method insures that the newly formed risk model is consistent with the component models from which it is formed.

Claims

1. A computer implemented aggregate risk model for a plurality of asset classes, where construction of the aggregate risk model comprises the steps of:

integrating a plurality of factor models corresponding to the plurality of asset classes, where integration of the plurality of factor models is based at least in part on;

one or more global factors (g) that capture covariance among factors (f) associated with the plurality of factor models; and

exposures (Y) of the factors (f) associated with the plurality of factor models to the one or more global factors (g).

2. The aggregate risk model of claim 1, wherein the one or more global factors (g) are data representative of a a time series of global factor returns, and the factors (f) are data representative of a time series of factor returns for the plurality of factor models for the plurality of asset classes.

3. The aggregate risk model of claim 2, where integration is further based at least in part on:

data representative of a covariance matrix (G) of the one or more global factors (g); and

data representative of a covariance matrix (Φ

) of residuals (ø

).

4. The aggregate risk model of claim 3, wherein residuals (ø

) are data representative of a time series of the purely local part of the factor returns of (f).

5. The aggregate risk model of claim 4, wherein the values of f, Y, g and ø

conform to f═

Yg+ø

, and wherein substantially all sources of common covariance between the factors in distinct asset classes are captured.

6. The aggregate risk model of claim 5, where integration is further based at least in part on:

data representative of a factor covariance matrix (F).

7. The aggregate risk model of claim 6, wherein the data representative of the factor covariance matrix (F), comprises data representative of one or more blocks comprising the factor covariance matrix (F).

8. The aggregate risk model of claim 7, wherein the values of F, Y, G and Φ

conform to F═

YGY^t+Φ

.

9. The aggregate risk model of claim 8, wherein F is an initial estimate that is then rescaled to bring its diagonal blocks into agreement with the corresponding blocks of factor covariance matrices of the plurality of factor models, whereby a final estimate of F is achieved.

10. The aggregate risk model of claim 9, wherein the initial estimate of F is linearly rescaled with a matrix R, whereby the final estimate RFR^t is achieved.

11. The aggregate risk model of claim 10, wherein an asset—

by—

asset covariance matrix is formed based on data representative of the final estimate of F, exposures (X) of the factors (f) associated with each of the plurality of factor models, and specific covariance data (D) associated with each of the plurality of factor models, in conformance with XFX^t+D, wherein F is the final estimate of F.

12. A computer implemented method for constructing an aggregate risk model for a plurality of asset classes, comprising the steps of:

integrating a plurality of factor models is based at least in part on data derived from the steps comprising;

determining one or more global factors (g) that capture covariance among factors (f) associated with the plurality of factor models; and

determining exposures (Y) of the factors (f) associated with the plurality of factor models to the one or more global factors (g).

13. The method of claim 12, wherein the one or more global factors (g) are data representative of a time series of global factor returns, and the factors (f) are data representative of a time series of factor returns for the plurality of factor models for the plurality of asset classes.

14. The method of claim 13, where integration is further based at least in part on data derived from the steps comprising:

determining data representative of a covariance matrix (G) of the one or more global factors (g); and

determining data representative of a covariance matrix (Φ

) of residuals (ø

).

15. The method of claim 14, wherein residuals (ø

) are data representative of a time series of the purely local part of the factor returns of (f).

16. The method of claim 15, wherein the values of f, Y, g and ø

conform to f═

Yg+ø

, and wherein substantially all sources of common covariance among the factors (f) are captured.

17. The method of claim 16, where integration is further based at least in part on data derived from the steps comprising:

determining data representative of a factor covariance matrix (F).

18. The method of claim 17, wherein the data representative of the factor covariance matrix (F), comprises data representative of one or more blocks comprising the factor covariance matrix (F).

19. The method of claim 18, wherein the values of F, Y, G and Φ

conform to F═

YGY^t+Φ

.

20. The method of claim 19, wherein integrating further comprises the step of rescaling an initial estimate of F to bring its diagonal blocks into agreement with the corresponding blocks of factor covariance matrices of the plurality of factor models, whereby a final estimate of F is achieved.

21. The method of claim 20, wherein the step of rescaling further comprises linearly rescaling with a matrix R, whereby the final estimate RFR^t is achieved.

22. The method of claim 21, wherein integrating further comprises the step of forming an asset—

by —

asset covariance matrix based on data representative of the final estimate of F, exposures (X) of the factors (f) associated with each of the plurality of factor models, and specific covariance data (D) associated with each of the plurality of factor models, in conformance with XFX^t+D.

23. A computer implemented aggregate risk model for a plurality of asset classes, where construction of the aggregate risk model comprises the steps of:

integrating a plurality of factor models corresponding to the plurality of asset classes, where integrating of the plurality of factor models is based at least in part on;

data associated with the plurality of factor models corresponding to the plurality of asset classes; and

an estimated factor model with one or more global factors (g) that captures the covariance among factors (f) associated with the plurality of factor models.

24. The aggregate risk model of claim 23, wherein integration is further based at least in part on:

data representative of a covariance matrix (G) of the global factors;

data representative of a covariance matrix (Φ

) of residuals; and

data representative of exposures (Y) of the factors (f) associated with the plurality of factor models to the one or more global factors (g).

25. The aggregate risk model of claim 24, wherein integration is further based at least in part on:

data representative of a factor covariance matrix (F).

26. The aggregate risk model of claim 25, wherein F is an intial estimate that is linearly rescaled with a matrix R, whereby a final estimate RFR^t is achieved, and wherein the diagonal blocks of the initial estimate are brought into agreement with the corresponding blocks of factor covariance matrices of the plurality of factor models.

27. A computer implemented method for constructing an aggregate risk model for a plurality of asset classes, comprising the steps of:

integrating a plurality of factor models corresponding to the plurality of asset classes, where integration of the plurality of factor models is based at least in part on data derived from the steps comprising;

determining data associated with the plurality of factor models corresponding to the plurality of asset classes; and

determining an estimated factor model with one or more global factors (g) that captures the covariance among factors (f) associated with the plurality of factor models.

28. The method of claim 27, wherein integration is further based at least in part on data derived from the steps comprising:

determining data representative of a covariance matrix (G) of the global factors;

determining data representative of a covariance matrix (Φ

) of residuals; and

determining data representative of exposures (Y) of the factors (f) associated with the plurality of factor models to the one or more global factors (g).

29. The method of claim 28, wherein integration is further based at least in part on data derived from the step comprising:

determining data representative of a factor covariance matrix (F).

30. The method of claim 29, wherein F is an initial estimate, and further comprising the step of linearly rescaling the initial estimate with a matrix R, whereby a final estimate RFR^t is achieved, and wherein the diagonal blocks of the initial estimate are brought into agreement with the corresponding blocks of factor covariance matrices of the plurality of factor models.

31. A computer implemented method for combining two or more risk models to provide an aggregate risk model for a plurality of asset classes, for risk analysis, comprising the steps of:

determining a factor risk model, having one or more factors, for each of the asset classes;

combining the factor risk models to form and output an aggregate risk factor model based at least in part on;

data associated with each of the factor risk models comprising;

one or more time matrix

to form and output an aggregate risk model with wider scope than the factor risk models; and

wherein the aggregate risk model is consistent with each of the factor risk models.

32. The method of claim 31, wherein the step of combining the factor risk models to form and output an aggregate risk factor model is further based at least in part on data associated with one or more global factors.