Comparison of rating scales of Russian credit rating agencies
The Bank of Russia publishes an updated table of the comparison of the national rating scales of Russian credit rating agencies (CRAs) included in the Bank of Russia register.
Table of the cluster-based1 comparison of CRA rating scales taking into account Russian CRAs’ data on one-year theoretical probabilities of default according to the national rating scale for the Russian Federation
| CQG | Average default rate | ACRA (JSC) | JSC Expert RA | NCR | NRA LLC |
|---|---|---|---|---|---|
| 1 | 0.15% | AAA(RU), AA+(RU), AA(RU) | ruAAA, ruAA+, ruAA | AAA.ru, AA+.ru, AA.ru | AAA|ru|, AA+|ru| |
| 2 | 0.37% | AA-(RU), A+(RU), A(RU) | ruAA-, ruA+, ruA | AA-.ru, A+.ru, A.ru | AA|ru|, AA-|ru|, A+|ru|, A|ru| |
| 3 | 1.16% | A-(RU), BBB+(RU), BBB(RU), BBB-(RU) | ruA-, ruBBB+, ruBBB, ruBBB- | A-.ru, BBB+.ru, BBB.ru, BBB-.ru | A-|ru|, BBB+|ru|, BBB|ru|, BBB-|ru| |
| 4 | 3.77% | BB+(RU), BB(RU), BB-(RU) | ruBB+, ruBB, ruBB- | BB+.ru, BB.ru, BB-.ru | BB+|ru|, BB|ru| |
| 5 | 11.78% | B+(RU), B(RU), B-(RU) | ruB+, ruB, ruB- | B+.ru, B.ru, B-.ru | BB-|ru|, B+|ru|, B|ru|, B-|ru| |
| 6 | 47.02% | CCC(RU), CC(RU) | ruCCC, ruCC, ruC | CCC.ru, CC.ru, C.ru | CCC|ru|, СС|ru| and С|ru| |
The approach employed has been developed as a follow-up to an earlier published comparison2 (based on the definitions of rating categories and the practice of assigning credit ratings to entities or bonds by Russian CRAs).
Specifically, in addition to the data on the practice of assigning credit ratings to the same objects by CRAs, the expected probabilities of default by credit rating level were used.
The approach chosen is a four-dimensional clustering (grouping of CRA credit rating levels) by six credit quality grades (CQG) which takes into account the statistics of matching and non-matching credit ratings assigned to one and the same entity by two or more CRAs.3
The table also contains information on the mean values of expected probabilities of default for CRA credit rating levels grouped into each CQG.
K-means is used as a clustering algorithm as it is widespread, provides sufficient interpretability and has a parameter to specify the required number of clusters.
The information on the comparison is intended for the general public.
The relevant statutory procedure4 stipulates that the results of the national CRA rating scales comparison be published at least every three years.
1 K-means clustering with a non-random initial distribution of centroids is used.
2 The previous comparison table was published on the Bank of Russia website on 30 December 2021.
3 The four characteristics of an object corresponding to a particular rating level of a particular CRA are as follows:
1) a natural logarithm of the probability of default (ln PD) for the rating level of the
2) ln PD for the rating level of the i+1-th CRA, weighted by the probability of transitioning from the rating level of the
3) ln PD for the rating level of the i+2-th CRA, weighted by the probability of transitioning from the rating level of the
4) ln PD for the rating level of the i+3-th CRA, weighted by the probability of transitioning from the rating level of the
4 Bank of Russia Ordinance No.
The reference to the Press Service is mandatory if you intend to use this material.