Back to main page

Case study background and problem formulations

Instructions for optimization with PSG Run-File, PSG MATLAB Toolbox, PSG MATLAB Subroutines and PSG R.

Problem 1a: problem_hull_meansquare
Minimize Meansquare(matrix_h) (minimize sum of meansquare error functions specified by matrix_meansquare1 and matrix_meansquare2)
Meansquare(matrix_meansquare1) +
0.0189*Meansquare(matrix_meansquare2)
subject to
Linearmulti(matrix_aeq) = 1 (constraint: system of linear equations specified by matrix_aeq)
Box constraints (constraints on individual decision variables specified by point_lb and point_ub )
——————————————————————–
Meansquare = Meansquare Error
Box constraints = constraints on individual decision variables
——————————————————————–

Data and solution in Run-File Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 2.66GHz (sec)
Dataset1 Problem Statement Data Solution 1,000 6 8.94 e-10 2.86
Data and solution in MATLAB Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 3.50GHz (sec)
Dataset1 Matlab code Data Solution 1,000 6 8.94 e-10 2.82
Data and solution in R Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 3.50GHz (sec)
Dataset1 R code Data 1,000 6 8.94 e-10 2.82


Problem 1b: problem_hull_variance
Minimize Variance(matrix_h) (minimize variance specified by matrix_h)
subject to
Linearmulti(matrix_aeq) = 1 (constraint: system of linear equations specified by matrix_aeq)
Box constraints (constraints on individual decision variables specified by point_lb and point_ub )
——————————————————————–
Variance = Variance
Box constraints = constraints on individual decision variables
——————————————————————–

Data and solution in Run-File Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 2.66GHz (sec)
Dataset1 Problem Statement Data Solution 1,000 1,000 6.45 e-08 0.64
Data and solution in MATLAB Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 3.50GHz (sec)
Dataset1 Matlab code Data Solution 1,000 1,000 6.45 e-08 0.62
Data and solution in R Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 3.50GHz (sec)
Dataset1 R code Data 1,000 1,000 6.45 e-08 0.62

Problem 2: Problem_CCC
Minimize Entropyr (matrix_h) (minimize entropyr specified by matrix_h)
subject to
Vector_bl ≤ Linearmulti(matrix_a) ≤ vector_b (constraint: system of linear inequalities specified by matrix_a)
Linearmulti(matrix_aeq) = 1 (constraint: system of linear equations specified by matrix_aeq)
Box constraints (constraints on individual decision variables specified by point_lb and point_ub )
——————————————————————–
Entropyr = Relative Entropy
Box constraints = probabilities should be non-negative
——————————————————————–

Data and solution in Run-File Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 2.66GHz (sec)
Dataset1 Problem Statement Data Solution 100 112 -3.47386 0.57

Data and solution in MATLAB Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 3.50GHz (sec)
Dataset1 Matlab code Data Solution 100 112 -3.47386 0.76

Data and solution in R Environment
Problem Datasets # of Variables # of Scenarios Objective Value Solving Time, PC 3.50GHz (sec)
Dataset1 R code Data 100 112 -3.47386 0.76

CASE STUDY SUMMARY
This case study estimates default distributions of a basket of credit default swaps using prices of CDO tranches. We calibrated probabilities with two approached in the so called “implied copula” model proposed by Hull and White (2006). The theory and the case study presented here is described in paper Veremyev, Tsyurmasto, Uryasev, Rockafellar (2012).
We solved two calibration problems:
– The fist calibration problem is a reproduction of the case study in Hull and White (2006). We solved the same optimization problem two times, using different PSG nonlinear functions: Problem 1a with the meansquare function and Problem 1b with the variance function.
– The second calibration problem (Problem 2) is based on “the maximum entropy principle.” We maximized the entropy of a distribution, subject to no-arbitrage constraints. The probabilistic distribution belongs to the class of Convex-Concave-Convex (CCC) distributions. The CCC distribution is specified by a system of linear inequalities for the discrete probability atoms.
This web page contains codes and data Problems 1a, 1b and 2 in PSG Run-File Text Environment.
Also, we posted MATLAB files to run this Case Study in the PSG MATLAB Environment. The link Matlab_codes_and_data_for_Case_Study Study refers to the zipped folder containing files and folders: file CS_Implied_Copula_Hull.m, file CS_Implied_Copula_CCC.m, folder “CDO_data”, folder “Data_Genearation”.
To optimize Hull model, run CS_Implied_Copula_Hull.m. To optimize CCC-distribution model, run CS_Implied_Copula_CCC.m . Data for solving optimization problems are already prepared in the files matrix_A_hull_mid.mat and matrix_A_100_5y.mat .
The MATLAB files for generating data, CDO_cashflow_generator.m, CDO_cashflow_generator_bin.m, CDO_matrix_generator_bid_ask.m are placed to the folder “Data_Genearation”. To generate data, run MATLAB file CDO_matrix_generator_bid_ask.m.
Folder “CDO_data” contains raw iTraxx data.

References
• Hull, J., White, A. (2006). Valuing Credit Derivatives Using an Implied Copula Approach. The Journal of Derivatives 14, 2, 8-28.
• Veremyev, A., Tsyurmasto, P., Uryasev, S. Rockafellar, R.T. (2012). Convex-Concave-Convex Distributions in Application to CDO Pricing. Research Report 2012-1, ISE Dept., University of Florida, June 2012.