Friday, October 27, 2006

Some REAL bond trader shit...

On Wednesday I made an off-handed comment that I'd probably be moving to a more neutral position on MBS from an overweight. Steve Feiss made a nice comment on my MBS idea, so I thought I'd flesh out my philosophy on MBS.

Now, I warn you, I spend a lot of time talking about interest rates on this blog. That's something that most people involved in the investment business follow and enjoy reading. Today's post is some real bond trader shit, that probably Steve Feiss and 1-2 other readers will care about. I'm probably looking at the lowest number of hits in months. What the hell, its my blog.

Scott Simon, head of MBS trading at PIMCO, once said that investment managers could add or subtract more alpha in MBS than any other sector, including corporates. I'd say anecdotally that very few investment managers would agree with that. But I know I have added far more value in my career through MBS trading than other sectors. And it hasn't been through aggressive trading. In fact, I'd say that maybe 20% of the MBS positions I've taken over the years were sold before simply paying off.

MBS are extremely complicated on one hand. If you buy a 30-year pass-through with 100 loans in it, then you've really bought 100 30-year non-call 0 loans. That's bond lingo for saying that you really own 100 different loans with 30-years to maturity that can be called at any time (ergo, non-call for zero years). But unlike, say an agency bond structured as a 30-year non-call 0, each of the 100 borrowers may or may not prepay their mortgage for non-economic reasons. Plus a portion of the loan's principal is repaid each month, so while there is technically 30-years to maturity, even with no prepayments most of the principal in the MBS will be repaid well before 30-years.

Many investors attack MBS from an OAS framework. I am dead-set against this. OAS vastly over simplifies a mortgage holders decision making. I've never seen an OAS model that didn't either A) assume a mortgage holders prepayment decision was entirely economic, B) make a blanket assumption about the impact of various characteristics of the underlying loans, or C) assume that prepayments due to "life events," such as moving to another city, occur "normally." So the result is the foundation of the OAS model is built upon assumptions that everyone knows are invalid! By this I mean that A) we know the decision isn't entirely economic, B) we know that the impact of loan characteristics are not consistent from one economic cycle to the next, and C) home mobility is greatly impacted by economic cycles and home price appreciation, so its not likely to occur in a smooth pattern. On top of this shaky foundation of assumptions, OAS models layer on more assumptions about how prepayments will shift due to movements in rates. Again, prepayment models are full of assumptions that everyone knows are invalid.

But if we separate what we know about MBS from what we don't know, the sector becomes rather simple. We know what the coupon rate is on a MBS. We also know that in the course of time, rates are likely to fall such that the underlying loans have a refinancing opportunity. That assumption might be questionable for a small segment of MBS with extremely low coupons, but it holds for most. We also know that in the course of time, most people move out of their current residence.

So what can we do with what we know here? If we figure that eventually a given MBS will come into the money (i.e., refinancable), we can assume that a large percentage of the MBS holders will prepay their loan at that time. We don't know when this might happen or what percentage of the holders will prepay, but we can vary both those factors enough to see the range of likely returns for the security. All we do is assume a small percentage of the loan will prepay in each period up until the time that it becomes refinancable, at which point prepayments will spike.

We then compare that to alternatives, bearing in mind that if the MBS has moved in the money, that alternative bonds would also enjoy capital gains. Typically, the MBS position pays a lot more in yield than other options, but in a rally is likely to substantially underperform. If using this methodology, you find a particular coupon/term structure is way overvalued, you can bet that its a structure that happens to look good in a lot of OAS models, but would break down in a refi spike.

The real skill then becomes seeing if you can find particular MBS that are likely to prepay more favorably than a generic. For example, let's say that you buy a premium priced MBS, so one that would perform terribly in a near-term refi spike. You might be able to find a particular pool where the underlying loan rate is below average for that coupon. For example, a typical 6.5% coupon MBS would have an underlying loan rate of 7%. But its possible to find bonds with an underlying loan rate as low as 6.75%. While we don't know what interest rates are going to do, we know that at any interest rate level, the 6.75% bond will have a lower refi incentive than a typical bond with the same coupon. The funny thing about it is, you rarely if ever have to pay a premium for what is an obviously superior security.

A similar story exists for other loan characteristics, such as average credit score, loan size, geographic distribution, etc. Granted, its a little more difficult to assume how these factors will impact prepayments, but since you can usually buy MBS with these characteristics at the same price as generic bonds, you are in essence making a bet without putting any money in. It almost has to work in your favor over the long run.

For example, I recently added some 30-year 5.5% coupon bonds. These are trading in the $98 range, so what I want to avoid is these bonds paying slower than average. In this market, its widely believed that the "life event" prepayments will occur slower because home price appreciation is down. The logic is that if a borrower doesn't have the cash to put down on a new house because his current house hasn't appreciated, he won't be able to move.

So if I buy a discount MBS, I want to find bonds that won't have the home price appreciation problem. That would include older loans, because those borrowers have already enjoyed some home price appreciation, and loans where the borrower is on strong financial footing. You can tell by looking for high credit scores, larger loans, and a loan-to-value below 80%. (The opposite of loan-to-value is equity, so a LTV of 65% means the borrower has 35% equity). I have no problem finding 30-year 5.5% MBS which are 6-12 months old, and appear to be full of loans given to borrowers with strong finances. And I paid nothing over generic MBS. If it turns out that these bonds don't pay a cent faster than generic, no harm no foul. But if they do, that's adding alpha.