Stewart: Welcome to a special edition of the InsuranceAUM.com podcast. Today’s topic is very timely. It is the potential changes in capital charge treatment for CLOs, and we’re joined by two experts today that are going to help us understand the issues at hand. Amnon Levy, CEO of Bridgeway Analytics, led the revision of the C1 factors with both the NAIC and ACLI, and Kevin Croft, professor of practice at Drake University and director of the Kelley Institute for Insurance Innovation. Gentlemen, welcome. Thanks for coming on.
Kevin: Thanks, Stewart.
Amnon: Thanks, Stewart. Always a pleasure seeing you.
Stewart: I am way, way out of my element today. And, Kevin, just for our audience’s sake, your background, you were at American Equity prior to becoming a professor. So, can you talk just a little bit about your background and maybe start off introducing us to CLOs and the actual asset that we’re going to be talking about today?
Kevin: Sure. Happy to, Stewart. My background, I have an actuarial degree, coming out of school, spent a little bit of time pricing and doing financials for an insurance company and then shifted into risk management roles and ultimately into investment management roles. So, that’s where I wound up at American Equity and was leading their structured products portfolio where we grew it from about $2.9 billion up to $13 billion during the tenure while I was there. And that $13 billion included a $5 billion portfolio of CLOs ranging from the double A tranche near the high end down to the double B tranche. So, pretty full breadth across it.
You ask, what is a CLO? Well, it’s complex and simple all at the same time. It’s a pool of loans, leverage loans which are double B, single B-rated loans to lower-rated companies. And then, there’s a structure of securitization put on top of that. And so, it slices the risk into different pieces ranging from the lowest risk like triple A, all the way down to a double B risk, and ultimately an equity tranche. And the behavior of these will be driven by some documents that relate to this. So, it’s a managed portfolio of leveraged loans, they’ll trade, they’ll buy and sell the loans. You’ve got a structure that decide where cash flows go that’s driven by a document and you’ve got a whole lot of modeling that can be used to help evaluate the strength or weakness of one of those structures.
Stewart: So, when you talk about CLOs, it’s similar to a CMO in mortgage land where you’ve got underlying collateral and you’ve got a structure that’s been put in place and that prioritizes cash flows and that’s what creates the rating structure.
Kevin: Right. It’s part of the alphabet soup, securitized products, the MBS, the ABS, the CMBS, and then the CLOs.
Stewart: Can you talk a little bit about the size of the market, the players, the investors? It’s a big asset class for insurance investors, which is why we’re talking about it today. But can you give us just the broad brush strokes on the CLO market in general?
Kevin: Well, I’ll start with the leverage loan market. It’s about a $1.7 trillion market, CLOs, approximately $1 trillion. So, a little over half of all leverage loans are owned by CLOs. So, it’s an important tie-in there. Insurance companies hold around $200 billion of CLOs on their books. The bulk of it of insurance company holdings is in the high investment grade category. So, when I think of it, 78% of insurance company holdings are single A, double A, and triple A. Only 13% are in the triple B category across the whole industry. Some players will have higher percentages in different parts of that capital structure.
Stewart: That’s really helpful. So, Amnon, I want to bring you in. Just to recap from our latest podcast on the regulatory framework in the US, can you just give us some highlights there? Just give us a starting point when we start talking about the potential for changes in the capital treatment for CLOs?
Amnon: Absolutely, Stew. CLOs like other credit assets are treated homogenously within the current regulatory framework in the US with the exception of CMBS and RMBS. The C1 framework that is in place assigns capital based off of the rating that, or the designation that the NAIC assigns, which is generally based off of an agency rating or the second lowest agency rating. The framework really is limited in many regards and something that we touched on a number of times as we were working with regulators on the redesign of the C1 framework, you have limitations for example with the observed variation in risk across asset classes.
The last time I was on your show, I pointed to the fact that if you look at MIS A-rated corporate bonds, they have a cumulative default rate of about 2% over a 10-year window if you are looking at the average default rate over the last 50 years. Meanwhile, Munis are at roughly 10 basis points, a difference of 20 times. And yet these assets are treated identically within the C1 framework, similar in the context of CLOs.
Stewart: And so, when we talk about agency ratings, we’re talking about NRSRO ratings and being based on the second highest rating. And then, that is, maybe not always, but generally, the basis for the assigned category by the NAIC which drives capital charges, right?
Amnon: That’s correct. And the difference in capital charges could be quite broad. Triple A will get something just north of 10 basis points and equity gets something in the order of 30%. So, if you’re thinking about designing an investment strategy, that’s not only sensitive to the economic risks, but also sensitive to headlines, statistics at the market, and rating agencies care about from a capital ratio perspective, this is an issue that you’re going to be very sensitive to.
Stewart: And so, what are the proposed changes? There’s time sensitivity, because the comment period, as this podcast is going to be released, is still open. What are some of the big issues around CLOs that is being talked about right now?
Amnon: That’s a good and very complicated question. Yeah, let me get into it. The headline issue that the NAIC is dealing with is what’s referred to as a regulatory arbitrage that exists with CLOs. There’ve been observations that you could materially reduce your capital charges by simply taking a pool of credit assets, say, loans that would go into the underlying collateral of a CLO, and tranche it up and hold the tranches. Seems like you’d be exposed to the same level of economic risks. Kevin could explain why the nuance structure of a CLO has the risks be different, and something I’m hoping he’ll speak to, and I’m excited to hear his perspectives on.
But just taking the simple thought exercise of taking a pool of assets, tranching it up, holding the tranches, you would think the risks are the same, yet you could cut your capital down by a factor of four or possibly in the order of like 25% of what you would otherwise have to hold if you held the underlying assets. And that disconnect really flows from, at its core, I think, the simplicity of the C1 framework, the fact that it doesn’t take, say, maturity into consideration when assigning capital. The fact that loans that go into CLOs often have an expected life, a fraction of the tenure corporate bond data that was used to calibrate the C1 framework. And so, when you think about those disconnects, you’re going to have these arbitrage opportunities, and the NAIC is dealing with the fact that they have to address this limitation in a fairly short amount of time. They want to address something this year. Considering the complexities, often these issues require years to work through.
Stewart: Thank you. Kevin, do you want to talk a little bit about the tranching versus holding the underlying?
Kevin: Sure. I totally hear where Amnon is coming from, totally. 9.5% is the capital charge if you hold the loan portfolio and it’s under three if you hold the CLO. Why is that? Some of it is what’s in the documents and the way the cash flow waterfall, they call it, allocates activity during times of stress. So, what it does is when defaults rise or when downgrades rise, so the first warning of a default is a ratings downgrade. When the holdings of a CLO start getting downgraded, you start tripping these called junior OC tests.
And if you trip a test, if you’ll hold more than 7.5% of triple C assets in a CLO, then distributions no longer flow to the equity tranche. So, what you’re doing is you’re no longer paying the equity. Now, where do those cash flows go? They’ll go to one or two things. One could be they just pay down the senior-most tranche and de-lever the structure. Or the other one could be they buy more collateral, buy more loans. So, you’re increasing the asset base of the loans.
It seems a little counterintuitive, but that’s actually a really positive thing for the CLO structure. Because when ratings downgrades are happening, loans tend to trade at a discount to price. So, if I can buy loans at 0.80$ on the dollar, I’m creating a lot of upside into this structure that benefits both debt and equity investors. But importantly, so you can see the structure starts to de-lever during times of stress and you shut off cash flows to equity.
What you’re doing is you’re transitioning the risk, or transferring it, from the debt holders to the equity holders. They don’t get cash flow distributions. And then, during the financial crisis, they got shut off for easily four to six quarters for a lot of deals. So, they absorbed a lot of pain to preserve the performance of the debt holders. And, that’s one thing that is very different from holding the pool of loans all on their own, is this transmission mechanism. And that’s one of the real big value components of a CLO and the debt tranche of a CLO.
Amnon: Kevin, what I really appreciate when speaking to practitioners about nuanced assets like CLOs is bridging those nuances and the modeling frameworks that are ultimately used in the context of building these capital frameworks. I’ve shared, Stew, on the last podcast with you that I had the privilege of working on numerous regulatory capital frameworks, work on models that fed into numerous capital frameworks, including Basel and others. And one of the challenges is the complexity and acknowledging that, say, structured assets aren’t a homogeneous set of securities, very different if you look at an asset-backed security that’s not managed and one that is managed.
And often simplifications are needed in order to manage the dimensionality of the problem. And active management is one of the more complicated aspects associated with modeling that often wind up getting lost when trying to level-set risk measures across assets. And that was something that we observed as increasingly relevant in the context of insurers’ holdings. Since the financial crisis, as you point out with your history as well now, since the financial crisis with low yields increased holdings of more opaque and less transparent, and more complex credit.
And over the years when we were working with insurers to try to level-set that articulation of risk, these issues bubbled up as different groups in the investment community realized that important nuances were not getting captured in level-setting those risks. And that’s something that I think the NAIC is going to have to deal with. The observation, for example, that CMBS and RMBS, designations are entirely model-based that are core to which is economic scenarios effectively that try to differentiate the risks associated with different tranches. How do you level-set, say, the CMBS assets and the CLO assets? How do you level set those tranches, given that the scenarios require a fair amount of judgment in their design, let alone the reinvestment strategies that enter into a CLO? So, very nuanced complex issues that I think are going to take time for the NAIC in the community to figure out what’s right.
Stewart: Earlier in the show, we were talking about the reliance on the NRSRO ratings and relating that back to the capital charges that are assigned at the NAIC. And that is a central difference in the discussion that’s taking place with regard to the changes in the treatment of CLOs now, right? The possibility of those not being tied together. Can you talk a little bit about what the central discussion is around the change in that methodology and the potential impact?
Amnon: Right now, there are many open questions and to their credit, the NAIC and regulators have been allowing ample time for an open debate and have reassured the community that they will keep the process as transparent as they can and allow for comments to help them in designing the right final solution. Now, that being said, the exposures that are open for public comment this coming Monday suggest that CLOs will eventually go down the path of being modeled using what’s referred to as the RMBS, CMBS intrinsic price approach, which effectively takes lifetime expected loss and assigns your capital based off of some in spirit, some measure like that.
And that is a departure from the C1 framework in that when you’re talking about lifetime loss, it necessarily means that longer-dated assets are going to have a higher cumulative expected loss, all else equal. That departure is going to mean that shorter-dated CMBS are receiving lower capital charges. If the NAIC goes down that path for CLOs, it’s going to suggest that shorter-dated CLOs are going to receive lower capital charges, all else equal. There’s talk of breaking up the designations for the equity tranches that Kevin was referencing and having the riskiest equity being assigned a 100% capital charge. So, there are a number of changes that’ll likely make the characteristics of the CLO tranches really look either more or less attractive compared to similar rated, say, corporate credit or municipal credit.
Stewart: Kevin, your background includes modeling CLOs extensively while you were at American Equity. What did you learn from those efforts?
Kevin: Right. I’ll build on Amnon’s perspective, and I think the first thing to echo is being a regulator is a really tough job and that’s being played out in real-time right now for them. They are very thoughtful, very sharp people who have the unenviable task of trying to stay ahead of very sharp creative investment people and legal teams. And so, I do not wish to be sitting in their role. To Amnon’s point, they have been transparent, they have been patient, they have been listening. So, they’re doing their job well.
A few things that come to my mind though when I compare and contrast this and what I’ve learned from modeling. First, let me look, S&P put out a study in March of this year. And looking at the history of CLOs, only 15 tranches, investment grade tranches of CLOs since the mid-1990s have ever defaulted. Okay, 40 in total, including below-investment-grade launches from 1990 to 2009. And then, after the financial crisis, the CLO went through a refresh to a 2.0 type version, which is what it’s referred to. And in the 2.0 case, there’s 12,000 classes of CLOs across 1400 different deals. Okay? There have been only 10 defaults, three double B and seven single B tranches. So, the level of default activity in this area is quite low. So, I’m going to say this, history has shown that these are relatively low-risk, low-default type assets. So, high-performance assets.
Now, to your question, when we looked at, in trying to decide how will you invest money and what risks are you willing to take in your investment portfolio, we started investing in Triple Bs tranches to get a sense of, okay, what can it withstand? And I use the GFC as a critical component. So, as we all know, the economic cycle goes through a cycle, it’s called that for a reason. And we’ll go through periods of lower GDP called recessions.
And then, we’ll go through longer recovery stand times when we’ll have low defaults during recoveries, and then you have high defaults during those recessions and parts of the cycles. So, you wind up with this curve, peaks, and then lower down in tranches, and then it goes back up again later. So, that’s the modeling we did on our CLOs to mimic the environment they really experience.
And what we found is that defaults peaked in the GFC around 10%, and a triple B CLO tranche could handle 1.75 times the stress of the GFC without losing a dollar. It would take time. It transmissions a lot of that volatility to the equity tranche, like Amnon was referring to. And so, 1.75 times 16.8 is where the default rate can peak out for a year on a CLO before triple B tranches lose any money. For the double B tranche, that was about 1.25% or a 12% default peak. So, these are able to withstand significant levels of default for extended periods of time.
And then, they self-heal or cure as the cycle goes through its expansionary phase and you see ratings go up, you no longer are tripping those junior OC tests and so everything fits and then cash flows start back out to the equity tranches when the debt tranches are fully healed. So, that’s the transition we see. To Amnon’s point about modeling and the RMBS experience, I was deeply involved in that from 2009 to 2020.
And what we saw there is, I’ll say I’m a little concerned about the scenario choices, and there’s not an easy one-fit answer, but I’m a little concerned about the scenario choices the NAIC is using for CLOs. And the ones I’ve seen proposed, they were looking at lifetime losses, but they were using constant default rates. And that just is not reality, as we’ve seen through history and we’ve seen through our modeling. So, it’s normal to have about a 3% default rate on leverage loans and that’s baked into the modeling. And, that causes no difficulty to CLOs if you have 2% or 3% default rates year after year after year. If you start doing stress as a standard deviation, a two above that, and hold them constant, it will stress a CLO to a point. It needs to have that cycle to properly work. But also, in my opinion, to properly model.
Stewart: Whenever I hear terms that I don’t understand, I just try to step back and define them for our audience that maybe like me that’s not steeped so deeply in the CLO tea. So, when you said the junior OC test, OC, does that stand for over-collateralization?
Kevin: It does, yes.
Stewart: Okay. And so, what that means is that when you’ve got your collateral performing worse than expected, and you mentioned it has to cure; that’s the active management component of a CLO that helps get those collateralization tests back in line, if I got that right?
Kevin: It could be from active management decisions by selling low-rated loans and buying higher-rated loans. But actually, some of it just happens through time as well with the actions I mentioned earlier where cash flow gets diverted, that helps improve the ratio of assets to liabilities in the CLO structure. It also happens if, when you’re in the good part, the expansionary phase of the economic cycle, companies see their credit ratings get upgraded, and that will help improve the over-collateralization ratios as well. So, there’s several different levers that can help cure an OC ratio. And to give a general sense also, Stewart, CLOs have more assets than they have debt. And so, that’s where you get over-collateralization. There’s more collateral, there’s more loans than there are debt tranches issued.
Amnon: Another interesting dynamic that plays out with CLOs is a loan prepayment, which I think on average is maybe in the order of 30% a year or so if I’m not mistaken, which means that you are constantly having cash to reinvest. And the extent to which the CLO needs to shift and improve the credit quality of its collateral that could be done by reinvesting those funds in higher-quality credit.
Kevin: And that’s part of the credit cycle as well. We see prepayments decline during difficult environments, but they don’t go away. Let’s say a company gets acquired, they will prepay those loans. So, that’s one measure or one driver of prepayment that occurs. So, prepayments, 30% during good times, that may drop below 10% during difficult times, but it never goes away. And so, if I buy that loan at $0.95 on the dollar and a year later it gets prepaid at par, or maybe even there’s a prepayment penalty of 102, 102.5 on it, that is total support of both the debt and equity trenches of a CLO.
And as the NAIC is pulling together their scenarios, they have pulled prepayment out. So, they’re modeling zero prepayment, which again, no prepayment. And then, also the lack of a credit cycle or an economic cycle are the two big things I would say that I think they should consider re-adding.
Stewart: Kevin, is that to suggest that their approach is too onerous?
Kevin: I think it’s a little too aggressive and too onerous. The reality is debt and equity markets go through cycles and they’re not reflecting that. Another ancillary learning I’d say I have from the RMBS modeling that they’ve done, when they’d started that in 2009; that was a very beneficial change to the insurance industry, and I was very happy to see that come along. Then a couple of years later, the scenario definition is important because as we got further from the crisis, say, to 2011, 2012, the home price appreciation was improving, trends were changing, rates of things were changing.
And in one of those cycles, the NAIC made their scenario work so that the bottom got closer and deeper. And we were in recovery phase. And so, the industry held several calls saying, “I just don’t understand the perspective.” What are they trying to accomplish? And this is divorced from the reality that we’re currently experiencing. And ultimately, someone from the NAIC did explain their position and they were trying to set the level of losses on the structures. So, it’s not heavy-handed, per se, but they were targeting an outcome.
And so, it’s a little bit of, do you want your scenarios to be reflective of the environment, or are you trying to target an outcome? And that was a shift from 2009 to somewhere in the early teens. And I wonder if that’s a little bit of what they’re trying to accomplish with the modeling they’re doing now. I do want to add some support to what they’re doing with these difficult scenarios. What did they do? They modeled, I believe it was 946 insurers. And they found that five of them with their current modeling wound up to a point of having CLO losses that did impact solvency for the issuer or for the insurer. That is something they do want to care about. That is a big part of their role. And so, they’re in that tightrope of trying to really balance “How do we model, how are we proactive without being onerous?” And I don’t envy them. It is a difficult spot to be in.
Stewart: So, let me ask you this when we’re talking about… one of the central themes here is being able to assign, for the NAIC to assign capital charges not based on NRSRO ratings, but based on modeled scenarios, right? Whose model? Is that a proprietary model internal to the NAIC? Obviously, there’s more than one model in the world. Who decides whose model we’re going to use?
Kevin: The NAIC has said as part of their transparency promise, that the major models are Moody’s and Intex. And they said, you’ll be given the scenarios, you’ll be given the inputs, and you should be able to, if not identically, very closely mimic the results that they’re coming up with. So, they will be transparent to those that have those models. And Intex is extremely widely used in the industry. It’s on most every trader’s desk.
Amnon: One of the challenges that you’re bringing up, Stew, though, and Kevin, you’ve touched on it as well, is when you have different asset classes, different models that describe different risks, how do you level-set them? How do you level-set the RMBS, CMBS scenarios and how favorable or unfavorable they are to those that define the CLO scenarios? The variables that are being simulated are very different from one another. And the way that the Intex or Moody’s or the BlackRock for the mortgage securities, the way that those models are calibrated is different.
And one very important issue that I think the NAIC has to address is how they’re benchmarking performance across these different possible models. What exactly is, quite frankly, the definition of a designation? What’s a good model that will assign a designation? And what’s a bad one? What’s the performance criteria that is allowing you to differentiate credit and be able to assess just how much riskier one credit is from another? And that hasn’t been done yet.
One of the challenges with using NRSROs is that each agency has different performance criteria. Moody’s, for example state their corporate ratings are opinions of expected loss. S&P is an opinion on a default likelihood. Now, if that’s really the different measures, and they’re all going into the designation along with possibly a model from the NAIC, well, how do you level set all this? It’s an incredibly complicated problem.
Stewart: And one of the questions, I mean, just to pull this together, we are recording on the 7th of September and the comment period is through the 12th of September of 2022 because this podcast is going to go out, and I want to level-set the date. If I want to participate or have my voice heard in this discussion, how do I contribute, or how do I participate in this process of figuring out how CLOs are going to be treated going forward from a capital charge perspective?
Kevin: I would encourage any insurer to reach out to, number one, their insurance department. And so, they’ll have contacts there with the investment team that represents their insurance department. I happen to be in Iowa, so I would go to the Iowa regulator. Additionally, though, there are, and if you go to the NAIC website, just simply type in CLO, you’ll see a bunch of documents and letters and it names who’s involved in each of those scenarios. And so, it happens, Carrie Mears is one individual here in the state of Iowa who is leading the effort nationally. And so, you could write a letter to Carrie Mears, the Iowa Division of Insurance, or others that are listed on the memos. And access to those people’s names and email addresses is very open. So, I would reach out to your regulators in your local jurisdiction, in your domicile, and/or at the national level as well.
Stewart: And we’ll put some links up when we post this. We’ll put some links up so folks can find that. We’re getting as close to a public service announcement as we can get.
Amnon: I want to reinforce a message that Kevin has made a number of times, which is the hard work in front of regulators and how deliberate they’ve been in trying to make sure the process that they follow is the right one for the industry. And I also want to reinforce his message of reaching out to the relevant decision-makers and their openness. I’ve found regulators to be incredibly, incredibly open to industry comment. And the effort that they’ve taken in helping them understand the nuances associated with these different asset classes, these are complicated issues, and often the regulators have backgrounds in actuarial science and may not be as familiar with capital markets. And the learning curve for them is one that it’s important that the industry supports them through this process.
Amnon: So, something that is obviously very important not only for the insurance industry but also for capital markets. The role that life companies play in and capital markets is tremendous. We saw the yield curve shift after the C1 framework rolled out in the ’90s. You literally saw fewer triple-A-rated issues because they weren’t as attractive from a capital perspective. These things matter.
Stewart: Yeah, absolutely. I throw this quote out all the time, but insurance companies own 30% of the world’s invested assets, right, at some $35 trillion. And it has huge implications for the capital markets. And ultimately, as we all know, that’s how projects get funded in the real world. That’s how equity capital, that’s how innovation happens. That’s how companies are born and financed, and new ideas that can benefit us all. Full disclosure, we did this podcast, we got folks on the phone very quickly. And, Amnon, your suggestion to do this. And it’s a great one. I think it’s one that there’s a very important topic, a very important asset class for insurance companies, and I really appreciate you making yourself available and you too, Kevin, for bringing your expertise to bear and let people know. And I will commit to putting up links to resources that either of you feel are important for people to know when we put out the podcast as well.
Amnon: Great. Thank you, Stew.
Kevin: Yeah. Thank you, Stewart. If I could throw in one last thing, a desired outcome, and what I would hope for or campaign for is really I think the most equitable result is to increase the risk-based capital charges for the equity tranche. Because when I look at the history of CLO performance, they have been strong. NAIC admits they have been strong. And when I look at the modeling that we’ve done throughout my career, it supports these mezzanine tranches as more volatile than normal, but very default-resistant security.
Amnon: One thing, Kevin, that I think will need to be addressed in that statement, which I’ve heard others articulate and advocate for as well, is the observation that agency ratings, while for any single agency, they aspire to be equated across asset classes. The question of, well, why is a double B or triple B CLO tranche be less risky than an equally rated corporate? What is that? Is that really the case? And if so, why do the rating agencies use models that are as conservative as they are?
And the extent to which the NAIC is able to get something more accurate is really something that I think needs to be addressed. It’s not an easy problem. And when you’re talking about remote events over a 10-year, or for a CLO, it could be a 7-year horizon, it’s going to be a very complicated, challenging issue to overcome.
Stewart: Thank you very much, both of you, Kevin Croft, professor of practice, Drake University, and director of the Kelley Center for Insurance Innovation. Amnon Levy, CEO of Bridgeway Analytics. Guys, thanks for being on.
Kevin: Thank you.
Amnon: Thank you.
Stewart: This concludes a special edition of the InsuranceAUM.com podcast. My name is Stewart Foley. Thanks for listening.
(PDF) Staff Discussion of Responses to CLO Proposal
(PDF) Collateralized Loan Obligation (CLO) – Stress Testing U.S. Insurers’ Year-End 2020 Exposure
(PDF) U.S. Insurers’ Collateralized Loan Obligation (CLO) Exposure Jumps Almost 23% at Year-End 2020