In this article (part 1), I covered the basics of non-agency mortgage-backed securities. Now in part 2, I look into how hybrid mortgage REITs manage their non-agency MBS investments, and into the risks associated with these securities. As a reminder, I covered how to analyze agency REITs in this article.
Some hybrid REITs I am using as examples are AG Mortgage Investment Trust (MITT), Invesco Mortgage Capital (IVR), Two Harbors Investment Corp (TWO), American Capital Mortgage Investment Corp (MTGE), and Chimera Investment Corp (CIM). I also include Ellington Financial (EFC) as a useful comparison benchmark.
As we will see, it is not only "the housing market" that drives risks and value; it is also servicing behavior, and particular differences in bond structures can have a huge impact. One important consequence (to be more developed in a future article) is that certain servicer stocks can act as natural hedges for hybrid mREITs.
As with the above-referenced articles, I make here some gross over-simplifications for the sake of clarity.
REITs exposure in the non-agency market
While agency securities are generally levered 5 to 10 times in mortgage REITs, the leverage on non-agencies is usually between 0 and 2, and financing costs are higher. This is a logical consequence of the lower liquidity in this market, combined with more intense (historical) volatility. The "fair" value of agency MBS probably trades in the -50 to +150bps option-adjusted spread range (see here) at a maximum, which in price terms translates approximately into 5% relative price moves. In the subprime market, historical prices have moved by over 50% in a matter of weeks or months, as fundamental credit anticipations and/or market risk premiums changed.
Large broker/dealers, who are the counterparties to REITs' MBS trades, do recognize however that there are big differences between bonds in the whole non-agency market. Prime paper, or Re-REMIC senior bonds, can command lower haircuts and cheaper financing than risky back-ended subprime bonds. So IVR, which holds a good amount of senior Re-REMICs, can get more leverage than TWO, for example, which holds mostly subprime paper.
So a first important question is how much exposure a given REIT actually has to the non-agency market. Simply looking at the dollar amount in non-agencies relative to equity does not properly distinguish between safer or riskier non-agency bonds, so I prefer to look at this question of exposure another way: by measuring how much they contribute, net of hedge costs. Before going into this, we need to quickly review some agency mREIT-related concepts.
A REIT's agency holdings typically generate significant amounts of yield. But as I explained in the introductory article on how to analyze agency mREITs, large portions of that are only the fair retribution for duration and convexity risk. As a matter of fact, the true "net" value added of agency MBS net of all these more or less hidden costs is measurable, and called the Option Adjusted Spread. As I explained in that article, I'd expect the OAS on typical mREIT holdings to be in the 1% range. This is like calculating an mREIT's net spread, factoring in the cost of all risks, even those they might chose not to hedge. Or one can see it as the pure "alpha" they generate. For example, on MITT, out of $2.3bb in agencies holdings, this would imply $23mm of agency-related annual "alpha" return.
The non-agency holdings, in contrast, will normally cost less in terms of duration and convexity hedges, for some of the reasons covered in part 1: barring the case of clean prime paper, the borrowers are less likely to refinance, and hence do not hold so much of a prepayment option; the bonds are most often structured as floating rate, which reduces duration. So if we turn back to MITT's holding for example (reported as roughly prime paper), the equivalent of the OAS we calculated for agencies holdings would be equal to the reported yield (6.3%), minus the reported funding cost of 170bps, and minus 20bps of option cost (I arbitrarily took about two fifths of a typical agency option cost given that these are not fully refinanceable loans). This amounts to 4.50%. Relative to $440m in non-agency holdings, we end up with an annual "alpha" of $20mm.
Adding up the agency and non-agency parts, I get $43mm, out of which the non-agencies account for 46%. This is, in my opinion, the least bad metric that we can use to capture the true exposure to non-agencies in an mREIT -- how much alpha they bring in. This factors in:
- Relative leverage (through the holdings amounts)
- Main rates risks (since we assume them hedged and pay the cost for it)
- True contribution in terms of "alpha".
By this measure, I find that your typical hybrid mREIT, while only carrying about 15% of its portfolio in non-agencies, has a true exposure of about 50/50 in agencies/non-agencies.
My estimate of non-agency exposure
*More guessing on these numbers due to the very limited amount of available information. I estimated their exposure by reading between the lines of their latest report (now 1 yr old...)
Now we can turn to the risks that are more specifically related to non-agencies in mREITs:
- Liquidations (defaults)
- Home prices (severities)
- Loan modifications
- Servicer behavior
As explained in part 1, the liquidation rate is the annualized rate at which the loans default and are liquidated in a non-agency pool. This rate is quite variable from deal to deal, as well as across time. As servicers have been working to modify loans, and in some states there were foreclosure moratoriums, the liquidation rates went down form the highs seen in 2008-2010.
It might seem logical that being long non-agency bonds through REITs, we want liquidations as low as possible. This is actually not true: quite a few bonds benefit from increases in liquidations. It depends (as is often the case with these products). More precisely, it mainly depends on structure, price and collateral.
To see this, let us consider a simplified Alt-A bond:
- Price of $60
- Expected principal recovery of 80 cents on the $
- Over 4 years of maturity / average life
- Coupon is negligible (very low)
The yield is about 7.5%: 60 * (1 + 7.5%) ^ 4 = 80.
Now let us say that liquidations come in faster/higher than expected. As a result, we will not recover as much as 80 cents on the dollar, but less, only 65 cents on the dollar (close to 20% reduction in recovery, in relative terms).
But since the liquidations take place faster, the return of principal happens much earlier, let us say within 6 months. Hence keeping a constant yield of 7.5%, the bond should be now worth about $62.7 instead of $60 (since 62.7 * (1 + 7.5%) ^ 0.5 = 65).
The key aspect here is that the faster return of cash flows, combined with a relatively high discount rate, more than compensates for the fact that the total cash flows are smaller.
In general, one finds that generic distressed subprime bonds are not largely exposed to changes in liquidation rates. So news about an increase or a decline in the total number of foreclosures are not that relevant for non-agency investments, in and by themselves.
Impact of home prices (and loss severities)
Home prices matter for non-agency valuations, but not any home price. As a matter of fact, the current delinquency pipelines, combined with observed liquidation rates, usually lead to the conclusion that a majority of subprime, Alt-A and option-ARM borrowers will eventually default (here by majority I mean in the 50-80% range).
As a result, it's not really the home prices measured by standard home price indices that matter, but rather distressed home prices, that is, the price at which liquidated properties are sold. These will have a direct impact on loss severities, which in turn will normally directly affect bond cash flows by increasing losses and reducing the return of principal. An increase in loss severity can only affect the bonds in one way: negatively.
Distressed home prices do not necessarily evolve in parallel with those based on transactions between willing participants, and that is why it's not really sufficient to just look at broad home price indices to form an opinion on how things are evolving for non-agency bonds. Zillow, for example, publishes a wealth of analysis and details on home prices, including distressed prices.
But it is important to also realize that the non-agency market prices in some fairly stressed assumptions regarding distressed house liquidation values. For example, when market participants assume future subprime loss severities over 80%, this is based on massive liquidation discount expectation, on the order of 75% (i.e. it assumes that houses are sold at a quarter of their initial appraised value).
So if you read somewhere that distressed properties are trading at basement levels, it is fairly irrelevant as far as non-agencies are concerned. What's relevant is (1) how these prices evolve and (2) to what extent that evolution is priced in the market.
As an alternative to liquidations, loan mods have been offered and tried with delinquent borrowers. Their impact on non-agency bonds is very sensitive to the type of structure.
Loan mods essentially turn a small short-term cash flow (a liquidation, hence maybe 30% of the loan's balance in 6 months from now) into a larger long-term cash flow (the loan's balance and coupon are reduced so that over time the borrower will pay maybe 60% of the current balance). So generically speaking, shorter bonds that "want" a quick return of principal will suffer when there are modifications, while longer bonds will be neutral or slightly up.
Naturally, the recent elections would point towards a continued effort by the government to help with loan modifications. Unfortunately, unless one has a more detailed view into a REIT's book than what typically transpires from quarterly reports, it is not possible to know which way the book would be impacted.
If I had to make a wild guess I would say that as a proxy, non-prime books with an average price over 70 would suffer with more mods, while they would moderately benefit under 70.
Servicer behavior (advances)
Servicers naturally drive liquidations and modifications, which we have addressed already -- but here we look into a very particular effect: servicer advances.
When loans are delinquent in a securitized pool, the servicer is supposed to advance their payments to the structure. The servicer gets their money back upon the loans' liquidation. Over the past few years, servicers have realized that finding the money to make these payments on subprime or Alt-A pools (something like 50% of these loans are delinquent, over extended periods of time) can be extremely expensive and essentially destroy their business model. So some servicers have aggressively reduced the amounts they were advancing to the trusts. Larger, bank-owned servicers (like Bank of America servicing for example) have been less aggressive as their holding companies were able to fund themselves more easily than smaller independent operations.
The impact of servicer advances on bond cash flows can be extremely significant, particularly for shorter bonds. On longer bonds, the fact that there is less cash flow initially (less advances) is compensated by the fact that in a few years, losses are smaller (since the servicer does not take money out of the recovered moneys from liquidations). Hence we have a comparable situation to the exposure to loan modifications: it depends on a lot, but by and large, high dollar price short bonds (above 70) suffer more from a cut in advances than low dollar price longer bonds.
When aggressive servicers like Ocwen (OCN) buy mortgage servicing rights, it is expected they will manage them as efficiently as possible (and presumably more efficiently than their previous owners). This will likely lead to a decline in servicing rate, and hence affect certain bonds -- by a larger amount, and in less time, than changes in home prices could. Here again, not knowing what a given REIT holds, it is not possible to infer precisely their exposure to this kind of risk.
Headline home prices, foreclosure or loan liquidation information might seem relevant for the non-agency holdings of mREITs, but all things considered, we've seen that they're not.
The value of distressed properties is more important, but typically not easily available, and takes a bit of digging.
Finally, the intensity of loan modifications, and of consolidation in the mortgage servicing sector both can have a significant impact on non-agency MBS, but this impact is difficult to measure without detailed information on positions. Nevertheless, lower price (and seemingly worse quality) bonds are generally in a better position to withstand these servicer-related risks.
Bottom line: in non-agencies, things are usually not what they seem, and you must dig deep to understand, otherwise your "analysis" is just noise.