After covering the analysis of agency mortgage REITs, I now turn to their close cousins, hybrid mortgage REITs. Readers interested in agency mortgage REITs should also read this other article by Portfolio Management 101 which also looks at REITs with a good and sound rational approach.
Hybrid mortgage REITs hold non-agency residential mortgage-backed securities (MBS) in addition to agency MBS. REITs must hold a significant fraction of agency MBS or whole-loans in order to qualify as REITs, so there are no "pure" non-agency REITs.
Given the extent of the subject, this article is broken down into two parts, the first one here focusing on the raw material purchased by these hybrid REITs, i.e. non-agency MBS, and how to analyze them. Part 2 will focus more specifically on how the REITs manage their non-agency portfolios and how to measure and quantify their potential returns and risks. Both articles are applicable to REITs such as American Capital Mortgage Investment Corp. (MTGE), AG Mortgage Investment Trust Inc. (MITT), Invesco Mortgage Capital Inc. (IVR), Two Harbors Investment Corp. (TWO), Apollo Residential Mortgage (AMTG), or Chimera Investment Corp. (CIM). All these REITs hold significant portfolios of non-agency MBS.
I make significant over simplifications in this article, but I believe the core points remain. Here in Part 1, I explain the basics of non-agency MBS and argue that in spite of a strong rally over the past year they remain fundamentally attractive.
Non-agency residential mortgage-backed securities are bonds backed by pools of loans that were not securitized by Fannie Mae, Freddie Mac or Ginnie Mae, and therefore do not carry a government or quasi-government guarantee. Issuance in the non-agency market was as large as agency issuance back in 2005 or 2006, but this market ground to a halt with the credit crisis. So when we talk about non-agencies, it means the secondary trading of roughly $1.5tr of paper outstanding today that was issued up to 2007. We call a "deal" a given securitization where the collateral, a bunch of loans (one or several "pools") back several bonds together (called "tranches").
There are 4 broad sectors in non-agencies, each with several hundred billion dollars outstanding:
- Jumbo prime: normally clean credit loans that were not securitized through the agencies because the loans' sizes were too large.
- Alt-A: loans with worse credit than those going to the agencies (low documentation, investor properties etc). For these, the cost of getting the agencies' guarantee was more than the cost of securitizing them as non-agencies
- Option-ARMs: adjustable-rate mortgages with very low teaser rates and negative amortization (you don't pay the full required payment, and the difference is added up to the loan's balance).
- Subprime: loans made to borrowers who had a problematic credit history. In addition the loan-to-value ratios (LTV) were substantially higher than on other sectors.
There are essentially three metrics to track the performance of these loans over time, and to express future performance:
- Prepayment speeds, CPRs (see this article for some explanations). In the order of 15 to 25% on prime loans, versus less than 1% on subprime in many cases
- Liquidation rates, CDRs (conditional default rates), that is the proportion of loans that are liquidated on an annualized basis. These are in the order of 5% on prime to about 12% on Alt-A/Subprime, but peaked in 2009 closer to 25%.
- Loss severity, which is the loss amount relative to the unpaid balance at the time of liquidation. So if a $100k loan is liquidated, and at the end of the day the bond holders have $40k left that is distributed, and $60k loss, that would be a 60% loss severity. Loss severities range from about 40% on prime paper, to 80% on subprime. It is largely driven by the massive discount suffered by foreclosed properties relative to where non-distressed homes would trade.
Credit enhancement and tranches
When these bonds (tranches) were issued, several mechanisms were put into place so that some of them would protect some others. The simplest is called subordination: if the loans generate losses, the subordinated tranches will take losses in priority in order for the other tranches to be protected, up to a point.
For example let us say there are $100mm of loans, backing 3 tranches: "A1" of size $75mm, "M" $20mm and "B" $5mm. B is the most subordinated, followed by M, and then A1. If the loans generate less than $5mm losses, they will all go to B only. So that tranche is leveraged to losses 20 times. Losses of 1% of the loans would translate into a loss of 20% of the principal on B. For losses under 5% of the loans' balance, A1 and M are hence unscathed.
If losses were $10mm, or 10% of the loan's balance, then B is wiped out: it never pays principal, and will probably just pay a little interest. But after B is toast, M is next in line and will take the losses in excess of what has been absorbed by B: M therefore takes $5mm of losses in this case, amounting to 25% of its principal balance.
If losses went further up, A1 would be OK as far as they stay below $25mm. Beyond that point, A1 will have to recognize some loss. For example if losses on the loans amount to $30mm then A1 would take $5mm of losses, amounting to 6.67% of its principal balance.
The amount of losses that the loans can take, until some bond begins to take a loss, is called the amount of "credit enhancement", or "credit support" of the bond. In the example above, A1 has 25% of credit enhancement, M has 5%, and B has 0%.
Note that it is not because a bond has some credit enhancement that it is particularly safe. M is a good example, because while it does benefit from some protection, it is inherently levered to the pool's losses: if losses went above 25%, then M would be entirely wiped out.
The subordinated bonds are called "junior", and the protected ones "senior". The senior bonds would typically get a AAA rating, while the junior bonds would be AA, A, etc. Today the majority of bonds in this market have been downgraded to junk.
With complex payment rules between all the bonds, one would usually find in a securitization deal:
- A pack of senior bonds, so that while they were all senior, some would get their principal repaid faster than others ("front sequentials" vs. "back sequentials")
- A stack of junior bonds, normally structured so that they would get their principal only after the seniors had been paid down, and so that the most juniors of these subordinated bonds protect the more senior of the subordinated bonds (so BBBs protect As, and As protect AAs etc).
A popular type of non-agency investment with some REITs has been so-called "reremics", meaning re-securitizations of non-agency bonds. To illustrate the principle, let us take a non-agency bond, just like the "A1" example I developed above. If collateral losses so far have been 30%, then the bond has already taken some losses, $5mm actually. So it is rated as a defaulted security. But maybe that given the collateral characteristics one does not expect to see many more losses, maybe only an extra couple of %. So one could cut A1 which now has a face value of $70mm into two new bonds, call them A for $50mm and R for $20mm (for residual), so that R is subordinated and protects A. Then, one can buy either A, with a lower yield but less risk, or R with a higher yield and more risk. Typically, bond A is re-rated as a AAA, hence fitting some bank portfolios and trading at an even lower yield because of that, which makes the subordinated piece R all the cheaper and more attractive.
For example CIM has been a heavy user of this technique, buying the subordinated pieces (like R in the example above) at attrative yields. This is clearly apparent on their books: they have to report the entire reremic structure on the assets side and the senior reremic on the liabilities side. Using the example above that means that instead of just being long R, they must report it as long A1, short A.
IVR on the other hand, has been buying senior Reremics.
The role of servicers
The servicer is the entity collecting payments from the borrowers so that they can be distributed to the bond holders. The servicer also manages delinquent borrowers, and has full authority to foreclose or not, or modify the loans or not. When loans are delinquent, the servicer is supposed to advance the payments to the deal (to the bond holders). These advances are to be recouped when the loan is finally liquidated. The servicer can decide not to advance the payments on particular loans if it looks like there will not be enough left after liquidation (depending on estimates for the house value, costs etc). We'll see in more detail in part 2 how servicers can significantly affect the value of non-agency MBS.
A distressed market
It's very important to realize this market is distressed: it does not trade the same way as agency MBS. Agency MBS trade in generic form by the hundreds of billions every day, while non-agencies trade through so-called bid-lists (granular and opaque auction mechanisms), to the tune of $2-5bb each day.
Until issuance stopped in this market in 2007, there were something like 6,000 deals done, having issued about 75,000 bonds. Each one of these deals is different from the next, and all these bonds are different. Before the whole market blew up, large chunks traded in a generic fashion, a "new issue 1-year subprime sequential" was treated as the same as some other "1-year subprime sequential". Nowadays, this is no more the case and the market has recognized that each bond requires a thorough analysis. The consequence is that there are as many prices as there are market participants, and the only way to know the value of some bond for sure, is to offer it for sale, or to try to buy it.
The difficulty in analyzing and trading non-agencies, combined with their tainted reputation after causing the credit meltdown, are the main reasons for the significant yield pickups they offer, net of credit risk. There is a fundamental selling pressure on non-agencies, as large commercial banks still hold significant amounts of them, and the carrying cost of these bonds is expensive, and getting more expensive over time as international prudential regulations come into play (Basel 2, 2.5, 3 etc). In addition, in many cases institutional investors will stay away from non-agency MBS because of the reputation risk attached to them.
This is a good thing, because it means that these bonds offering attractive yields tend to stay cheap.
Finding the value in non-agencies
In terms of value the most important thing is that statements such as "the bonds have such credit support", or "the bonds are backed by such and such type of loans", without factoring in all the other moving parts, are entirely irrelevant.
The 3 aspects that one must consider, at a minimum, when looking at non-agency paper are: collateral, structure, and price.
It should now be clear that subprime might sound worse than prime to some (collateral), but a subprime front-pay bond with 90% enhancement is orders of magnitude safer than a deeply subordinated prime bond (structure), while at the end of the day maybe I'd rather own the deeply subordinated prime bond at $15 rather than the subprime senior bond at $99 (price). Note this is really a minimum, in most cases it is impossible to say anything intelligent about a bond unless one has access to an algorithmic representation of its structure (on Bloomberg for example, or on other advanced and expensive tools).
The only valuation data that REIT investors will generally find on non-agency MBS is from the REITs themselves. As far as I know this presentation (from TWO) contains the most precise example of non-agency MBS investment analysis on page 17 published by a REIT. Not all REIT investors are able to measure whether this type of analysis is sound or not, but for those who can this is invaluable. In this particular case, I agree with the analysis and that's part of the reasons why I own TWO. It would be great that other REITs published the same type of detailed information, so that one could put more, or less, credence in their reported portfolio yields. In TWO's case, the reported yield on the non-agency portfolio is 9.6% (page 7 of the presentation), which is consistent with discounted subprime.
The non-agency yields reported by various REITs closely follow the type of non-agency exposure they have. See the table below for some summarized information.
|REIT ticker||Type of MBS||Reported yield||Source|
|TWO||Mostly subprime, some option-ARMs and Alt-A||9.6%||Pages 7 and 8|
|IVR||Prime bonds and Reremics||5.4%||Pages 6|
|MITT||Senior prime bonds mostly||6.3%||Pages 10 and 13|
|MTGE||Alt-A, option-ARMs, subprime||7.4%||Pages 5 and 16|
Now it's not because TWO reports a much higher yield than IVR that their portfolio is better (as a matter of fact I own IVR too). They just have different types of risks.
Prime MBS do exhibit some of the negative aspects of agency MBS: the prepayments on jumbo prime are directional relative to mortgage rates, and since the bonds trade at high dollar prices they are subject to negative convexity (see the above-mentioned article on the analysis of agency mortgage REITs for more details on this subject). They also exhibit some duration exposure due to their fixed-rate coupons, unlike many non-prime bonds that pay floating coupons.
However, different types of non-agency MBS do not get the same level of haircuts in repos -- in other words, prime or less credit-exposed paper can typically be levered more than straight subprime.
The impact of non-agency prices
In discussions and comments, some have mentioned the rally on non-agency MBS as a good thing for hybrid mortgage REITs.
First, one has to define what we mean by non-agency prices. Since they are not public, or even appear on "screens", the only continuous source of non-agency pricing is through some particular credit derivatives indices. For subprime, there are the ABX indices (historical prices are available here by clicking on particular indices), and for prime there are the PrimeX indices (available here). The correlation between these credit derivative indices and the cash bonds a REIT might own is naturally not 100%, but they are the best publicly available approximation.
If we look at the PrimeX index with the highest price, PRIMEX.FRM.1, we can see that over the past 6 months it has gone from about 106 to 109.5. This is not a very significant move. This is representative of prime paper with a strong credit quality. So we should conclude that the portfolios with mostly senior prime exposure should not have gained significantly in value.
In subprime, ABX.HE.AAA.06-1 is the highest dollar price indices. It's composed of more seasoned subprime bonds mostly from 2005, with a good level of original credit support (for subprime). It's gone from 89 to 94 over the past 6 months, a 5% increase. More credit-exposed subprime paper, such as those backing ABX.HE.AAA.07-2, have gone from 35 6 months ago to 45 today, a 28% increase.
We would therefore expect REIT non-agency portfolios exposed to subprime to have improved in value over the past several months, which is obviously a positive.
Higher prices could mean tighter yields on these non-agency bonds, but the principal repayment on most of the lower dollar price securities is not large relative to the bond's market value, unlike on agency MBS. In other words, a REIT holding typical subprime bonds will not have large amounts of principal to reinvest at a potentially lower yield. Hence the negative impact of lower yields is not likely to act as a strong negative relative to the benefit of higher prices.
Non-agencies are complicated and there are no shortcuts to analyzing them or forming an opinion on their value. There is however one simple truth, which is that it's precisely because many people think they're toxic, that they can offer attractive yields.
There are however particular risks attached to these securities, which we will go through in part 2, as well how they fit in various REITs' portfolios.