Short Shrift for Surveillance
There seems to be a scramble on at the rating agencies, but we
think we’ve seen this scramble before.
Beginning late in 2006 the subprime and alt-A residential
mortgage-backed securities (RMBS) market began its long downward slide into
unimaginable losses.
The credit rating agencies – of which S&P,
Moody’s, and Fitch have the leading market shares – appeared to be slow to
respond. Perhaps there was good reason in the early months of
2007. Or perhaps not. An argument against promptly
downgrading the RMBS was to wait for a few monthly reporting periods to
validate the trend of rising mortgage delinquencies. The risk in
waiting, of course, is that it suddenly becomes “too late” and you are required
to implement larger, more frantic downgrades than in the alternative in which
you had been downgrading incrementally if and when appropriate. But
while ratings were yet to be downgraded there was no doubt, however, that the
values and prospects of both the RMBS and their underlying mortgages were
tumbling violently.
We’re reminded of the iconic J.P. Morgan himself who once said
(paraphrase): “Every man has two reasons for doing
something. There’s the reason he gives that sounds good and then
there’s the real reason.”
The rating agencies had at least one “real” reason for not
downgrading RMBS bonds promptly: for RMBS and other structured products, they
were ill-equipped to provide the services! From our perspectives
having been in the industry, reviewing Congressional testimony and materials
from the Financial Crisis Inquiry Commission, and having conferred with others
playing similar roles at other agencies, we can tell you that the rating
agencies did not have the capacity to methodically or efficiently oversee their
models and ratings on the RMBS bonds they purported to be
monitoring. There was very little that was programmatic to it. The
process for the supposedly continuous, ongoing surveillance of thousands of
RMBS bonds was (shockingly) manual. Or it was not done at all.
The rating agencies were much more concerned with clipping high
revenue-generating new deal rating tickets, and much less concerned about
providing the required surveillance on existing deals. Monies for
surveillance, after all, get paid whether or not the service is provided.
So the RMBS downgrades came in waves spaced out over many months
in 2007. The rating agencies placed many bonds on downgrade watch prior
to the requisite modeling and analysis based on mortgage characteristics,
transaction vintage, and broad market performance. This “watch
period” supposedly bought the rating agencies time to perform the actual analysis
before specifying actual downgrades.
Of course, while the existing ratings were intermittently “wrong”
or awaiting correction, there was much ado about how to rate new deals or other
existing deals that depended on those outstanding – but yet to be addressed –
ratings.
Now it’s 2020 and we wonder if the ratings surveillance process
has improved at all. The tremendous economic stress of the
coronavirus (“COVID-19”) raises doubts of the sudden inability of homeowners to
continue making mortgage payments. Many of these debtors are now
unemployed with much uncertainty for the future. The house prices
themselves may be greatly depressed – we won’t know until the end of this
suspension period for real estate transactions. Falling home prices
in 2007-2008 themselves prompted many mortgage defaults. Thus, it’s
fair to say that RMBS bonds have much greater risk of default than they did at
the beginning of the year.
Moody’s acknowledges the impact of the coronavirus, and the
substantial uncertainty it brings with it.
“Our analysis has considered the increased uncertainty relating to the effect of the coronavirus outbreak on the US economy as well as the effects of the announced government measures which were put in place to contain the virus. We regard the coronavirus outbreak as a social risk under our ESG framework, given the substantial implications for public health and safety.”[1]
Importantly, Moody’s accepts
that “It is a global health shock, which makes it extremely difficult to
provide an economic assessment.” However, Moody’s is in the business
of making these assessments, meaning it needs a robust and methodical way of
tackling the new variable, and accounting for the uncertainty that comes with
it.
The rating agencies have two very
interesting vocations at present.
First, they’re actively downgrading
RMBS (and most other bonds too).
Next, they’re actively rating new
deals, even though some of the new deals are backed by assets that they’re
simultaneously downgrading. For example, rating agencies rated five
new CLO deals, backed by loans in April; meanwhile, as of early April JPMorgan
was showing that loan downgrades were outpacing upgrades at a rate of 3.5-to-1.
Part 1: Ratings Downgrades
Moody’s announced, on April 15, its
placement of 356 RMBS bonds on review for downgrade and the actual downgrade of
48 bonds to the Baa3 level.[2] The downgraded certificates remain on watch for further downgrade.
Placing bonds on watch for
potential downgrade is not, alone, all that interesting. But the
downgrades were interesting. Moody’s rationale:
“The rating action reflects the heightened risk of interest loss in light of slowing US economic activity and increased unemployment due to the coronavirus outbreak. In its analysis, Moody's considered the sensitivity of the bonds' ratings to the magnitude of projected interest shortfalls under a baseline and stressed scenarios. In addition, today's downgrade of certain bond ratings to Baa3 (sf) is due to the sensitivity of the ratings to even a single period of missed interest payment […]”
There are a few things that
are interesting here.
First, the lack of specificity.
Moody’s does not specify any quantitative elements of its analysis – Moody’s
does not cite to changes in default rates, severity rates, recovery rates or
prepayment speeds – that were adjusted to accommodate COVID-19 or that justify
the downgrade action or explain the specific rating as Baa3, which is (curiously) the
lowest level of investment-grade. It is all rather vague.
Second, you might have noticed some
unusual language there. The “downgrade of … ratings … is due to the
sensitivity of the ratings…” That is all meaningless. One does not
downgrade a rating because of the sensitivity of the rating. All
ratings, aside perhaps from some Aaa ratings, are “sensitive.” One
downgrades because of a higher potential for default or loss, as ascertained by
an analysis performed – not because of the sensitivity.
That brings us to the next issue:
what analysis was performed? The short answer is, well, that there
was no proper analysis performed! Aside from Moody’s failure to
describe the specific analysis performed, we have at least two reasons to
believe that no analysis was performed.
- Moody’s does not have a methodology for incorporating the coronavirus stress: Moody’s explicitly acknowledges that the principal methodology used in these ratings was the "US RMBS Surveillance Methodology" published in February 2019, which of course was pre-COVID-19
- Moody’s acknowledges that it did no modeling: “Moody's did not use any models, or loss or cash flow analysis, in its analysis. Moody's did not use any stress scenario simulations in its analysis.”
Altogether,
we see 48 bonds downgraded to Baa3, absent:
- an updated methodology that reflects the key new coronavirus risk, despite acknowledgment of the key importance of this risk to the deals;
- a proper application of the existing methodology from Feb 2019 or any explanation as to the deviations from the methodology; and
- the application of any models at all.
Strictly
speaking and adhering to the meaning of Moody’s ratings, it is extraordinarily
difficult to imagine how Moody’s could determine that 48 bonds should
simultaneously earn the new, lower rating of Baa3 without
analysis. With no quantitative estimate for increased loss or diminished
and re-directed cash flow, how can Moody’s determine the precise new rating
level?
One
thought, however, is that Moody’s guessed at the rating based on the
methodology, and perhaps because there were no models to apply: in the rush of
rating new deals, maybe Moody’s hadn’t gotten around to building
well-functioning models to replace the old ones? Or, equally
plausible is the possibility that Moody’s models do not exist in the cloud,
meaning that the Moody’s analysts and supervisors working from remote (home)
locations could not access the necessary models, data or tools to perform their
reviews. What an astounding disclosure this would be, if true, for a global
data-centric organization!
Whatever
the reason, the Baa3 ratings are flawed. No analysis by an outsider,
or perhaps even a Moody’s insider, when applying Moody’s methodology, can
credibly achieve an outcome of Baa3. It is just a guess.
Part
2: New Ratings
The
rating agencies are actively downgrading bonds, loans, and credit instruments
across most sectors and industries.[3] Many of these downgrades may
be well-intended, due to newly introduced COVID-19-related
risks. But while they are downgrading these (newly unstable) credits
they are rating new structured finance credits that are supported by these very
same unstable credits. Their structured finance ratings
problematically look to their own ratings of the credits, which are being
downgraded daily. Thus, new ratings cannot be said to be robust to
the degree the rating agencies lack confidence in the sturdiness of their own
underlying ratings. Moreover, the new structures are not being rated
according to any new, post-coronavirus methodology. So we might
expect those newly-created bonds to be similarly downgraded, too, in short
order.
When
S&P rated Harriman Park CLO, a new deal, on April 20th, S&P said
nothing at all about coronavirus in its press release. In explaining
how it came about its ratings, S&P cited only to related criteria and
research from 2019 and earlier. There is no evidence that any
scenario was run at all differently by S&P. No mention is made
of any newly-imposed stress scenario being run.
When
Fitch rated this same deal, Fitch explained its thought
process and how it is deviating from its pre-existing
methodologies. That’s a whole lot better than S&P in this case,
but even Fitch failed to explain the “why.” Fitch simply explained
what it is doing, but not why what is doing makes sense. How have they calibrated their models (if at all)? How do we know the new assumptions are adequate or comprehensive?
“Coronavirus Causing Economic Shock: Fitch has made assumptions about the spread of the coronavirus and the economic impact of related containment measures. As a base-case scenario, Fitch assumes a global recession in 1H20 driven by sharp economic contractions in major economies with a rapid spike in unemployment, followed by a recovery that begins in 3Q20 as the health crisis subsidies. As a downside (sensitivity) scenario provided in the Rating Sensitivities section, Fitch considers a more severe and prolonged period of stress with a halting recovery beginning in 2Q21.
Fitch has identified the following sectors that are most exposed to negative performance as a result of business disruptions from the coronavirus: aerospace and defense; automobiles; energy, oil and gas; gaming and leisure and entertainment; lodging and restaurants; metals and mining; retail; and transportation and distribution. The total portfolio exposure to these sectors is 9.9%. Fitch applied a common base scenario to the indicative portfolio that envisages negative rating migration by one notch (with a 'CCC-' floor), along with a 0.85 multiplier to recovery rates for all assets in these sectors. Outside these sectors, Fitch also applied a one notch downgrade for all assets with a negative outlook (with a 'CCC-' floor). Under this stress, the class A notes can withstand default rates of up to 61.8%, relative to a PCM hurdle rate of 53.9% and assuming recoveries of 40.6%.”
While it continues to rate new deals,
S&P (for example) has placed roughly 9% of all its CLO ratings on watch
negative since March 20th.[4]
The rating agencies suffered
tremendous reputational damage when they were found to be rating new CDO deals
in 2007 while they already knew they could no longer rely on the RMBS ratings
supporting those deals, as they would imminently be downgraded. Once
the CDO rating analysts knew the RMBS ratings were unstable and about to be
downgraded, the CDO analysts were taking a legal risk in producing
ratings they knew would not be robust. It can be difficult to turn
away the sizable revenues that come with rating new deals – even when you do not have a
sustainable ratings methodology or any conviction in the credibility of the data
(including ratings) you are relying on.
The essence of formulating credit ratings for all entities (structured, corporate, municipal, etc.) requires the ability to estimate future revenues, expenses, and liabilities for a multi-year time period. Such estimates are critical to the rating process. As we write, forecasting for the broad economy and for most specific entities is highly uncertain. We do not see how it is possible to perform meaningful rating analysis amidst this uncertainty. Hence, the credit rating agencies should arguably pause all new ratings until uncertainty declines or until they develop rating methodologies that fully incorporate the wide uncertainty.
Part 3: Summary/Conclusions
Stale
ratings, inflated ratings and other faulty ratings can sometimes go unnoticed
during an upswing. But shortcomings become most pronounced during an economic
downturn or crisis. We are concerned that we are (again) seeing the
result of years of prior, weak, ratings mismanagement.
When
crises occur, rating agencies should be able to swiftly update their
methodologies, models and ratings. And rating
agencies should stave off rating new deals until and unless they have strong
and current methodologies in place to explain their new ratings and how their
ratings accommodate the upheaval, whatever it may be.
Rating
agencies seem quickly to have forgotten (or simply to have ignored) the lessons
they should have learned from the last crisis. Regulatory
settlements with the DOJ for subprime crisis-era misconduct (S&P for
$1.375 billion[5], and Moody’s, in the amount of
$864 million[6]) concerned issues in which they
deviated from their code, which required them to provide objective, independent
ratings. Those were not about the rating agencies being wrong, or
failing to predict the downturn, but closer to argument that the rating agencies failed to
believe their own ratings.[7]
In the above case, we would
have to ask how Moody’s can be sure that Baa3 is the right rating – for all 48
bonds, across different deals, structure types and vintages – consistent with
its methodology, given it failed to apply any modeling?
