A framework for predicting composite bond rating changes

It is clear from these results that rating momentum matters. 

2. Index methodology composite rating bias

The idea here is that the number of individual rating agencies contributing towards a given composite bond rating, and any alignment or otherwise disagreement between them, is helpful for predicting upcoming composite rating changes for that bond. Whilst noting whether a bond is already split-rated or not, we find that daily evolving single agency reporting events can give meaningful insights regarding its future composite bond rating changes. Figure 2 shows the summary hit rates for this indicator.

Judging by the magnitude of these hit rates, and comparing Figure 2 to Figure 1, we see that rating bias is potentially more important than rating momentum for predicting forthcoming composite bond rating downgrades and upgrades.

3. Credit watch flag

The final and most important concept to be aware of is that bonds with a credit watch flag have a greater likelihood of undergoing a near term rating transition (in the direction of the flag).⁶ Figure 3 shows the summary hit rates of this daily indicator versus future composite bond rating outcomes.

These results show that credit watch matters, even more so than rating momentum and rating bias particularly over the 1, 3 and 6-month horizons; 43% of US high yield names that were put on negative credit watch in our historical sample have had a composite bond rating downgrade over the subsequent 6 months.

Among all three indicators, it seems that the indicator was a more important driver in US high yield than in US investment grade.

Rating transition score

We combine our three daily indicators to produce a rating transition score for each bond and assess how accurately it has predicted changes in composite bond ratings. Figure 4 summarises the historical hit rates of the highest and lowest values of this score⁷ when used for forecasting composite bond rating upgrades and downgrades over future 3 and 12-month forecast horizons.

Focussing on the outcomes over the 12-month horizon, we see that the resulting upgrade and downgrade probabilities associated with extreme scores are reasonably strong, falling between 50% and 65%. To put these numbers into context, we compared these transition rates with those in US investment grade and high yield benchmarks.⁸

For example, whereas, empirically, there is a 7% chance that an investment grade index’s constituents will be downgraded over a 12-month horizon, you can actually improve portfolio outcomes by steering the allocation away from bonds with rating transition scores of -2 or lower (which each have a 51% chance of being downgraded) towards bonds with rating transition scores of +2 or higher (which each have a 62% chance of being upgraded). Analogously, in high yield credit, there is an 18% chance that high yield index constituents will be downgraded over a 12-month horizon. Likewise, you can improve portfolio outcomes by steering the allocation away from bonds with rating transition scores of -2 or lower (which each have a 56% chance of being downgraded) towards bonds with rating transition scores of +2 or higher (which each have a 65% chance of being upgraded).

Conclusion

We have developed a framework for constructing rating transition scores to predict composite bond rating events. The approach capitalises on momentum in bond ratings, published rating agency credit watch flags and the replication of index composite rating methodologies. The rating transition scores are produced at the bond level and, as such, can be used within portfolio construction and risk management contexts for avoiding bonds that are at increased risk being downgraded. Equivalently, the output can be used to identify and get ahead of future upgrade events. Other direct applications include the anticipation of capital charge evolution in insurance portfolios and the construction of forward-looking rating transition matrices, including multiple rating notch forecasts.

¹ Bond index providers commonly incorporate the bond ratings from the three main rating agencies into aggregate rating measures called composite credit ratings, which they use as part of their index membership criteria.
² The historical analysis was conducted using the US investment grade corporate and US high yield corporate bond universes. More generally, we have access to daily agency ratings, and their associated credit watch and outlook metrics, for all global corporate bonds and issuers. The historical analysis was conducted using over seven years of data.
³ We have also conducted other, more detailed, studies that assess not just the directional accuracy of these predictors but also their ability to forecast the magnitude of future composite bond rating changes. We will look to utilise these results in future blogs.
⁴ The Barclays Credit Research team has conducted a related study in a paper titled, “Ratings Recidivism: Downgrade Momentum in Corporate Credit”, 10^th March 2023. The paper focuses exclusively on rating momentum trends and credit watch lists.
⁵ We define hit rate as an empirical measure of how often, within our historical dataset, composite rating change predictions got materialised over given future time periods.
⁶ We tested both the credit watch and outlook indicators put out by the main rating agencies. Whilst our empirical tests have determined credit watch to be a useful predictor (of composite bond ratings), they have not found this to be the case for outlook. We also note here that less than half of all agency rating changes historically were preceded by a credit watch indicator.
⁷ i.e. +2 or better and -2 or worse.
⁸ Analysis conducted using index constituent composite rating changes during the period 2016-2022.

9a6cc630-eb59-11ed-9d22-eeee0af4d508