REIT Analyst

Reits, research analyst, bonds, long/short equity
REIT Analyst
REITs, research analyst, bonds, long/short equity
Contributor since: 2012
"They don't delta hedge and they don't and can't hedge the basis risk between treasuries and mortgages"
That's wrong I think. Buying IOs (or MSRs, or any comparable structure) will naturally hedge a large amount of duration and basis. Many mREITs do it.
Way I see it, these MSRs they have are not much worse or better than an oversized strip IO position, and as such are a useful/efficient instrument to hedge both the basis and duration. I'd rather they own them, because if they did not their overall spread exposure and duration would be worse.
Jerkim, I did not mean that they would run anything or do anything at 0 CPR. I just meant that as an extreme example to illustrate how and why the runoff and time decay are two fundamentally different things.
Hello Jerkim,
Regarding the double counting, I was wondering about that, and then did a little thought experiment. Consider MSR or some IO off of a bullet. You would have 0 prepays, 0 prin payments. Hence no runoff (until the very end). However you would still have a reduction in value over time: in the last few periods left with just a couple of monthly payments your strip would be worth very little.
An IO being like a premium on steroids, you have that premium amortization taking place notwithstanding what is going on in prepays, I think.
With prepays/runoff you kill off an entire stream of future cash flows. With time decay you take out a one-month cash flow out of every loan.
Hi Robert,
This is an interesting story you have here, thank you for the detailed analysis.
I looked into this following a slightly different angle, and I get something that is not as clear cut as you did, but definitely along the same lines.
Back of the envelope:
94mm income over MSR value is about 26bps annualized, as one would expect. Then if I subtract the costs (including the ones you logically allocated to the line of business) I get 16.6bps annualized. Versus a 95bps value, that means a multiple of 5.7, for a 13 CPR type of asset.
Then looking at the cheaper range of IOs or strips, that pay around 13 CPR (and a bit of interpolation) I get that such an IO paying 13c and at a wide OAS (above 400) would be at a multiple of ~5.2. So that is in the ballpark of TWO's MSR valuation relative to their net income, with the MSRs on the rich side.
As you pointed out strip IO yields are crap, but they can however be fundamentally cheap thanks to the OASs, which are in a sense reflecting the benefit of the very negative duration. For these IOs or MSRs, I'd say duration is somewhere around -50.
To me this is really not to be dismissed, and is extremely significant. From a hedging perspective the negative basis exposure (values go up when spreads widen), and negative duration is worth a lot. For example, and oversimplifying a bit, -50 duration means you could go long 6x 10-yr Trsys, so yielding 9% (assuming ~50bps in financing costs).
So if you reflect the gain from a hedging perspective it does really change the picture. Still, you are right in that using MSRs to gain this negative duration might not be the most efficient way, and they could improve it. Do realize however that you can't easily go and buy 500mm in market value of strip IOs just like that. So in that type of size there are not many alternatives to MSRs.
(btw, I am long TWO, but I am ok with this MSR issue - in a sense the glass is half full for me rather than your half empty view)
The book value in most cases is a true reflection of liquidation value, since these are all liquid and identifiable products. Of course if they do not all liquidate the same day, in which case the MBS markets would be affected.
There is ~2-400bb a day in agency passthroughs trading. Liquidating a 10bb portfolio would be noticeable, but not that much. Besides, if done over a few days, it would have an even smaller impact.
FYI, an inverted yield curve is only a marginal issue. And if it plays a role at all it is not because of what you are describing but rather indirectly through the potential cheapening of MBS in that situation.
Overall, REITs run~6yrs long rate durations, and ~ -4 yrs slope duration, at the equity level.
So curve goes flat 100bps, they lose 4% in BV (reflecting all the future earnings stuff etc).
Curve goes up in parallel 100bps, they lose 6% in BV.
Nothing that dramatic, especially in relation with their carry. As DResearch said, they hedge most of their rates/curve risks.
I would be much more worried and focused on mortgage value versus swaps, which has declined substantially over the past couple of months. Buying mortgages when they're cheap, that's what matters for REITs. And incidentally I think that is why they are not buying back their stock that much. It's a relative value call: makes more sense to stay long mortgages right now. They'll buy back stocks when/if they have deep discounts and MBS are fair to rich.
I reckon mREITs across the board run ~-4 yrs in slope duration (a few more than that, a few less). This means if the curve flattens 100bps they lose 4% in BV. Since July the 2s-10s Trsy curve has flattened maybe 50bps. And in swaps, which matter more, the curve has moved even less.
So who cares?
Interesting, and it makes no sense. All these REITs now are running negative net durations.
Many REITs run negative durations, and most of them have substantially reduced their rates exposure, so BV will not benefit that much from this rally.
And 1/3rd of AI's book value comes from a tax deferred asset which is effectively worthless if you consider it on an apples to apples vs a typical mREIT (which pays 0 taxes).
And how old are the basic "models" to value stocks? Dating back to the mid 30s for discounted CFs? Mid 60s for CAPM? How is the age of these research pieces relevant?
I would not expect to find that of info in a 10Q! And the articles from analysts that address these points are certainly not either from stock analysts. Here are a couple of pointers however:
On MSR valuation:
On IO durations vs mortgage durations:
Where they show that basically IO-like paper is 10x "riskier" (within a properly developed notion of multidimensional risk) than passthroughs (unstructured / basic MBS).
This statement "Its Mortgage Servicing Rights are not leveraged at all [...] This significantly reduces risk" is misleading.
Thank God they are not leveraging their MSRs. MSRs are already massively leveraged to mortgages. They are like IOs, and concentrate a very large amount of prepayment and credit risk.
Depending on how you measure it, MSRs concentrate about 8-15x the risk of mortgages. So in that respect I'd argue that 1x on MSRs is like 12x on typical agency or non-agency bonds.
A general comment: does it not make more sense that REITs strive for 0 duration, and you have the opportunity to buy 30-yr Trsys on leverage if you want to bet that rates go down? Why would you want to trust an mREIT's management to forecast rates? They might be good at managing a mortgage portfolio, but that does not involve forecasting rates. Besides, they have tended to be wrong in their views anyway.
The lower rates on these ARMs is reflected in the price. In fact from a yield or spread standpoint, ARMs are more attractive than fixed-rate paper. Also, it's wrong to say that the ARMs have the same "life expectancy" as FRMs. ARMs have a shorter duration, because of their faster prepayment speeds.
Technically, prepayment risk is the risk that prepayments are different from what is initially anticipated given the rate environment.
What you described (fairly clearly, thank you) is the fact that prepayments will change as rates move, in other words the directionality of prepayments to rates, but that is well understood / modeled / hedged (for those who decide to hedge that directionality). That is called negative convexity, it is very much akin to the risk you would bear selling interest rates options.
Prepayment risk would be rather related to the risk that your hedging model is wrong, for example because home prices are stronger than expected and the refinancing response is steeper than expected on some borrowers.
I do not think that the rate at which the swaps are entered is really relevant to the hedging ability of these swaps, because for a fixed amount paid or received from a counterparty you can change the fixed payment to whatever you want. I had written a piece on swap hedging by mREITs some time ago, which addresses that aspect to some extent among other things.
So there is something to be learned here, at least for me, and I think it is rather important: Wall Street research reads SA, and they position themselves relative to articles published in SA.
Given all the various (and I think, valid) points that have been raised, it would seem to me it would be the least of things that the authors address some of them at least. For an article published under the Pro category, I would have expected more responsive authors on these points.
Aren't they financing it through securitization? If they don't they would always have that option which would delink financing cost from their credit rating.
This is an interesting story, however it feels like you might not have fully analyzed the mortgage servicing business. In your discussion of "rights to MSR", you might want to google "excess servicing". It is fairly standard and common that servicing be cut up in excess vs "real" cost of servicing. Also, excess servicing rights are qualifying REIT assets. So if you set up a REIT holding them you don't pay taxes anyway.
Hence, I have a hard time believing there is much of a tax advantage in holding what amounts to excess servicing through a Caiman entity, vs a plain old REIT.
Hello Michael,
Thank for your interesting comment.
As you point out I did not discount the current assets, but I viewed that as an investment portfolio, which is essentially fairly liquid securities and could be turned into cash very quickly. But you're right that applying a discount is a useful metric to look at. You can easily adjust the numbers: A 10% discount on the investment book would be about $1, and taking $1 off will reduce the "net" valuation. The relative effect is not very large however, about 6% in relative terms.
Feel free to message me with your questions.
If I follow you correctly, your analysis is this:
"Clearly [...] will be worth $9 or $10 from the charts".
And your state that there is no point in putting together any other analysis (particularly fact-based or knowledge-based) since you already know the answers.
On all accounts, your comment is entirely unsubstantiated, so I would appreciate if you could actually provide elements that justify your points.
Why is this momentum stuff more valid than analysis?
Why is it "clearly" so?
Why do all the numbers or analysis I discussed in the article not answer the question?
Are you saying that there is no difference between's something's potential value and its market price?
Why expect a 15-25% gain? Why not something else?
We know rather well how home prices have affected agency loan performance, by looking at how high default rates were through the crisis, as a function of local home price appreciation. On rather bad paper (2007 vintage), defaults were about 18% in places with -45% HPI locally, and about 6% in places where HPI was about flat. In a case where home prices would be down 10% (which would already be a very serious stress relative to what home prices have been doing recently), you would expect defaults to increase, and if on top of that you assumed that MGIC's portfolio was all as bad as the 2007 vintage, then you would expect about 9% of defaults. The first stress I discussed in the article is even stronger than that.
So you are right that bad home prices are a negative, but (1) reasonable scenarios would not be very bad and (2) even unreasonable scenarios do not have an overly negative impact on the analysis.
Reperforming loans are to a large extent modified loans, it's only recently that one has started seeing loans curing for real. The historical performance of these modified loans was very ugly a few years back, but at that time, loans that had been never delinquent so far were also defaulting at very high rates. The important aspect here is how much worse these modified/reperforming loans are versus squeaky clean loans. And the data seems to indicate they are about twice as bad.
Hi Rob,
Thank you for your comment.
First about the dividend, I did not factor any payment (which of course on the plus side would send a good signal, but on the minus side takes cash out of the company) in order to simplify the analysis of fair value. In a sense whether the cash is in your pocket or in the company does not make a huge difference to the financial aspect of the calculation.
Separately, there's the question of business growth, and there I also wanted to keep things simple and try to err on the conservative side.
The large discrepancy you point out in yields (and which I should have thought of addressing in the article -- sorry) is simply due to the fact that the net income is evolving over time. Right now there are large amounts of claims coming through, which for the moment implies a fairly small net income. But as I think I have explained in detail, these claims will be logically declining over time, at a fairly fast rate, as (a) bad legacy paper is flushed out and (b) claims on new paper ramp slowly and to a much lower level.
I think the right way to frame this question would be in terms of spread volatility on the particular assets that CMO has been holding. Period.
Then this would raise the question of ARM vs FRM production, future home price growth (only driver of net mortgage production), and refis from ARM into FRMS or reciprocally, which will depend on the slope of the curve.
Also as a side note, the book's duration, given that it's presumably hedged, has essentially nothing to do with the required spread.
Still, if you were correctly hedged (ie reflecting CYS's effective duration of over 20 yrs) and had the right amount of 10-yr trsy shorts, you would have been about flat today.
Investing in mREITs without a proper hedge for residual interest rate exposure just means taking an unknown amount of interest rate risk.
CYS might go to 1x book, but what if book drops by 20%?
Ever heard of hedging interest rate risk?