Fast Computation of the Expected Tranche Loss of CDO Credit Portfolio

An algorithm for fast computation of the expected tranche loss of CDO credit portfolio.

The code is explained in the article P. Okunev, "A Fast Algorithm for Computing Expected Loan Portfolio Tranche Loss in the Gaussian Factor Model", LBNL-57676, 2005.

http://www-library.lbl.gov/docs/LBNL/576/76/PDF/LBNL-57676.pdf

Futher refinments of this algorithm are descibed in Okunev, Pavel, "Using Hermite Expansions for Fast and Arbitrarily Accurate Computation of the Expected Loss of a Loan Portfolio Tranche in the Gaussian Factor Model" .

http://ssrn.com/abstract=748625

This is a MATLAB code. It's relatively easy to adapt it for VBA.

ATTENTION: This code was tested and works well for portfolios of size 125. The accuracy will decrease for smaller portfolios. Higher accuracy can be achieved using the methos described in Okunev, Pavel, "Using Hermite Expansions for Fast and Arbitrarily Accurate Computation of the Expected Loss of a Loan Portfolio Tranche in the Gaussian Factor Model" .

http://ssrn.com/abstract=748625

This implements one factor Gaussian model.

[loss]=gsloss(L,w,p,a,d,N)
L = exposures,as fraction of total
portfolio, taking into account the recovery rate
Example: loan 1 is 0.01 fraction of the total portfolio, recovery rate is
40% then L(1)=0.01*(1-0.4)
w = loading factors
p = default probabilities
a = attachement point
d = detachment point
N = number of names in the portfolio
loss = expected tranche loss as percentage of the portfolio nominal
expressed in basis points

Copyright by Pavel Okunev 2005
E-mail: [e-mail]

This code is provided as is. The author provides no warranty and assumes no responsibility for any losses due to the use of this code.

You are granted permission to use this code for personal use and for academic research.

This code may not be used for commercial purposes without explicit permission by the author.

Permission for commercial use can be obtained by writing to [e-mail]

You may make and distribute a small number of copies of this code if you include this copyright notice with the code.

If you distribute a modified version of this code you must include the copyright notice and conspicuously indicate that the code was modified.

We are thankful to John Weare, LBNL and UC Berkeley, for his assistance with debuging the code.

Please report bugs to: [e-mail]