精选博文

C++ implementation of a simple order book

Please refer to my github for the code:  https://github.com/DongliangLarryYi.  1.      Data Structure 1.1   A basic or...

Sunday, August 20, 2017

What information do we need to rate a CLO investment?

CLO is a kind of securitized products which build CLO tranches based on cash flows from a leveraged loans pool. In order to analyze the performance of CLO tranches, we need to know the information of the leveraged loans in the pool. These loans’ Probability of Default (PD), Recovery Rate (RR), Default Correlations and Prepayment rates, are needed for us to have a better understanding of a CLO deal.

Here I would like to introduce a Moody’s methodology [1] on CLO analysis. The methodology is aimed to give a rating to CLO tranches. Since prepayment do not affect the credit rating, so no prepayment is discussed here.  


Probability of Default

Moody’s calculates the expected average default rate with their rating factors weighted by par value. WARF is short for the Weighted Average Rating Factor. The rating factor number indicates the 10-year accumulative default probability of the asset. For example, Aaa asset has a rating factor of 1 which means a 10-year accumulative default probability of 0.01% and the Caa3 has a 10-year default rate of 80.7%.

Moody’s Default Probability Ratings vs. Moody’s Rating Factors
Moody's Default Probability Rating
Moody's Rating Factor
Moody's Default Probability Rating2
Moody's Rating Factor3
Aaa
1
Ba1
940
Aa1
10
Ba2
1350
Aa2
20
Ba3
1766
Aa3
40
B1
2220
A1
70
B2
2720
A2
120
B3
3490
A3
180
Caa1
4770
Baa1
260
Caa2
6500
Baa2
360
Caa3
8070
Baa3
610
Ca, C
10000














 Source: Moody’s Investors Service

Since the rating factor indicates the 10-year default rates, we need to do linear interpolation when the pool’s Weighted Average Life (WAL) is not 10 years. We may also adjust the rating when there are review on loans of upgrade or downgrade.

Default Correlation

A CLO deal normally have more than 150 loans in its pool, so explicit calculation of correlation would be explosive. Moody’s uses an implicit method to model the default correlation with the diversity score. This score considers the loan volumes of different industries, and how many industries are covered in the pool.

Moody’s uses a binomial distribution with diversity score as the total trial (round down to an integer) to derive the loss distribution. If the loan pool is heterogeneous along some measures, we may need to use more complex distribution like multi binomial approach or a full simulation.

Recovery Rate

The recovery rate is the percentage will be recovered when a default happens. According to the appendix 3 in the report [1], the higher rating loans have lower recovery rate. Moody’s examines the difference between the instrument rating and the Default probability rating. The higher the difference, the higher the recovery rate.

Recovery rates of First-Lien Senior Secured Loans (Instrument Rating - Moody’s Default Probability Rating)
Target Rating
-1 Notch
0 Notch
1 Notch
Aaa
40.00%
45.00%
50.00%
Aa1
40.00%
45.00%
50.00%
Aa2
40.00%
45.00%
50.00%
Aa3
41.70%
46.70%
51.70%
A1
43.30%
48.30%
53.30%
A2
45.00%
50.00%
55.00%
A3
46.70%
51.70%
56.70%
Baa1
48.30%
53.30%
58.30%
Baa2
50.00%
55.00%
60.00%
Baa3
51.70%
56.70%
61.70%
Ba1
53.30%
58.30%
63.30%
Ba2
55.00%
60.00%
65.00%
Ba3 to C
55.00%
60.00%
65.00%
Source: Moody’s Investors Service

Conclusion

The WARF is an important measure when we look at a CLO deal. Lower WARF indicates the pool has more highly rated loans, and it is normally safer than high WARF pool. The diversity score gives us an idea on the concentration risk, it also affects the portfolio level default rate and loss distribution. However, the recovery rate in the Moody’s reports looks not that intuitive for me, especially that the lower rated loans have higher recovery rates. I may need to do more research on it.


Reference

[1] ‘Moody’s Global Approach to Rating Collateralized Loan Obligations’, Moody’s Investor Service, Feb. 27, 2014

No comments:

Post a Comment