Litmus Analysis on agency performance

scales pension fund scheme trustee magazine investment training

Inevitably, any business that critiques others will not always be popular, and when they get something wrong they get a lot of criticism, write Litmus Analysis director Peter Hughes and partner Stuart Shipperlee. That’s certainly been the lot of three main capital markets agencies in recent years.

While some of this was undoubtedly deserved (as with sub-prime CDO ratings) it also reflects an issue we often see in the misunderstanding of what ratings really are (forecasts) and the consequential erroneous reliance on them as ‘facts’ rather than ‘inputs into decision making’.

Ratings are merely forecasts based on current and historic information. Rating agencies do not have access to any information that is not also available to a firm’s non-executive directors, auditors and regulators. Indeed, agencies have far less ability to require full objective information from the executive leadership than any of these other actors.

Moreover, with a degree of prudence, ratings reflect an agency’s view as to the most likely outcome for a (re)insurer’s financial strength, not the only outcome imaginable.  

So how should we assess agency performance?

Forecasts cannot always be correct, so a logical approach is to consider how they perform in aggregate. S&P and Moody’s publish data on bond defaults. From this we know, for example, that S&P’s ‘A’ rated firms have had a 0.07% history of defaulting within one year of having the ‘A’ rating ( a 1-in-1400 year event).

Many reinsurance people may well feel they know far more ‘A’ rated reinsurer failures than this would imply. That is because of both the challenge of defining what reinsurer failure is and when it takes place after a reinsurer first experiences problems. The path to definitive ‘failure’ can take a long time; simply going into run-off is not default.

A low observed percentage of default by companies at a given rating level increases dramatically when you combine lots of firms with that rating and observe this over a decade of one-year periods.

 A.M. Best publishes data that helps address the ‘defining reinsurer default’ problem.  

Rather than measure ‘default’ they measure ‘impairment’, meaning regulatory takeover following financial duress. Inevitably there is some subjectivity, but it’s a lot easier to define than claims payment ‘default’.

The general intention of regulatory intervention is to prevent a claims payment failure. Therefore it should be more frequent and happen sooner than defaults, and Best’s data shows this.

However, whichever approach to observed ratings performance we take, it does not address the key question of ‘how often should single A reinsurers fail’? The answer cannot be ‘never’, since they are risk-taking entities. Nor can it be ‘almost never’, otherwise they should be rated ‘AAA’.

Therefore how well the agencies perform is in the eye of the beholder. The point is that a rational user looks at the data, understands that ratings are forecasts and then uses them in decision-making accordingly.

How about comparing the agencies?

At Litmus we routinely analyse the ratings of two reinsurer cohorts; ‘Reinsurance Majors’ and the ‘L-Zebedees’ (firms operating in one of or more of the hubs of London, Zurich, Bermuda, Dublin or Singapore).

We calculate the average ratings of Best and S&P for each cohort where both have a rating on the same carrier (our methodology is available on our website).

For the Majors, both averages are at the lower end of ‘AA-‘, with S&P’s very slightly lower than Best. For the L-Zebedees, Best is towards the lower end of ‘a+’ whereas S&P is towards the upper end of ‘A’.

A common misconception is that a lower rating is somehow more rigorous, but a rating that is too low is as incorrect as a rating that is too high.

Nor would it be desirable for the agencies to always agree with each other, indeed, while they all look at the same issues they do not rate in exactly the same way. Generally we see Fitch and Best giving relatively more weight to capital and Moody’s/S&P to issues that drive operating performance.

However, we see one major difference between the four agencies - the degree of transparency in their criteria and rating reports.

All strive to be transparent and produce papers on criteria, but the degree of detail and need for case-specific judgments can make precisely interpreting these challenging.

Best, S&P (and now Fitch) have ‘factor -based’ capital models which help, however capital is only part of the process. Fitch and  Moody’s publish ‘scorecards’ and expected ratio ranges, but these too can only really define ‘measurable’ factors.

But last year S&P introduced a step-change in criteria transparency for insurance ratings. Each element of the rating process now features specifically defined factors and detailed guidelines for how each is evaluated, then a series of decision matrices combining the factors together to produce one outcome.

They also publish some of the intermediate steps (such as the ‘ratings anchor’), further adding to the degree of transparency.

We believe they can and should go further, but, for now at least, there is clear blue water on this between S&P and its three main competitors.