Redistributing Credit Transition Probabilities After Updating Default Risk

Summary

The document asks how to revise the non-default outcomes in a credit rating transition row after new information changes the estimated default probability. It presents three candidate relationships between the original and revised probabilities, each imposing a different proportional adjustment. The accepted response recommends preserving the relative ratios among the remaining outcomes. That choice keeps their relative likelihoods fixed while redistributing their total probability to accommodate the updated default estimate.

This is a practical modeling principle when the additional information is assumed to affect default risk but provide no new evidence about transitions among surviving rating states. It avoids implicitly changing the relative ranking of those non-default outcomes. The response points to a general probability-redistribution discussion as support, but supplies no empirical validation or worked numerical example. The recommendation therefore depends on the assumption that relative non-default probabilities should remain constant; if the new information also changes migration risk, the transition probabilities need a richer update.

Key ideas

• When only default risk is updated, preserve the relative proportions among non-default transition outcomes.
• Renormalize the non-default probabilities so the full transition row still sums to one.
• Changing the relative likelihoods of surviving rating states implies information about those transitions as well.
• The proposed adjustment is conditional on new information affecting default risk alone.

Tags

# Estimating credit transition probabilities from additional information


# Estimating credit transition probabilities from additional information












Let say $P_{i,j}, j = 1,2,3, DEF$ are the probabilities of transitions from an initial rating $i$ to rating $j$, where $P_{i, DEF}$ represents the default probability from that initial rating.

Now let say, based on some other information I require to modify above default probability to $R_{i, DEF}$ - which is considered to be more accurate estimate of probability.

Now, I need to adjust non-default probabilities, as I have now better information. I consider following 3 approached.

First approach is to use the formula $\frac{1-P_{i,j}}{1-R_{i,j}} = \frac{1-P_{i, DEF}}{1-R_{i, DEF}}, j = 1,2,3$

Second approach is to use the formula $\frac{P_{i,j}}{R_{i,j}} = \frac{P_{i, DEF}}{R_{i, DEF}}, j = 1,2,3$

And, third approach is to use the formula $\frac{P_{i,j}}{R_{i,j}} = \frac{1-P_{i, DEF}}{1-R_{i, DEF}}, j = 1,2,3$

Among above three approaches, which approach can be considered as best?

Your pointer will be very helpful

## Answer by Lsvob (score 1, accepted)

https://quant.stackexchange.com/a/71743

I found this thread on the math stack exchange: https://math.stackexchange.com/questions/3988333/how-to-redistribute-probabilities-when-one-outcomes-probabilities-changes

It suggests keeping the ratio of the remaining probabilities constant which in your case might be a good idea since you might want to avoid making the other states more or less probable in relation to one another as this would implicitly mean you have been given more information.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.