The Emperor's New Clothes

12/15/2012 SIFMA Prepay Speeds
December 18, 2012, 10:26 am
Filed under: Mortgage Servicing Rights, MSRs, Prepay Speeds

December 15th prepay speeds have been uploaded to the L1A valuation site. Speeds seem to have leveled off for the current coupon 30s. The 15 speeds have ameliorated somewhat (see graph below). Let’s hope this trend continues into year-end.

15 year conforming PSA

15 year conforming PSA

August Fannie Mae Prepay Speeds

Annualized one-month prepay speeds on Fannie Mae MBSs seem to support the prepay models high expectations.  The data below is based on the underlying mortgage coupon, not the pass-thru rate.  I also, excluded those tranches that had statistically insignificant volume. 

As you can see, everything above 4.5% is getting hit pretty hard.  The seasoned pools are not quite as bad as the moderately seasoned.  ARM pools (not shown) prepayed at an annualized rate (CPR) of 25%

There is a wide divergence of speeds by state.  The “bubble” states (AZ, CA, FL & NV), interestingly, were all among the highest prepay states.  Puerto Rico was the slowest.

We will be updating this monthly and, hopefully, start to be able to draw conclusions as we gather more data,

8/15/2012 SIFMA Prepay Speeds

Speeds for the 15th have been posted on our site. They appear to have diminished substantially from their 7/31 highs,

Refi’s Wreak Havoc on MSR Impairment
July 10, 2012, 3:21 pm
Filed under: Lessons Learned, Mortgage Servicing Rights, MSRs, Prepay Speeds

This has been a difficult year for servicers.  In addition to increasing costs and ever more regulatory burdens, we continue to see most of our new servicing coming from refinancings.  For the majority of servicers that do not utilize fair-value accounting, this can result in bloated temporary impairment reserves. This temporary impairment is really permanent, and should be written down as the loans pay-off.

From an economic perspective, refi’s are not beneficial to the servicing industry.  While the new loan added may be slightly more valuable than the pay-off due to higher balance, lower coupon and extended maturity, the increased value is oftentimes offset by the cost of paying off the old loan and setting up the new. 

Yet from an accounting perspective, many servicers add the value of the newly originated MSR to their capitalized servicing and continue to amortize the capitalized servicing on the old loan.  The theory behind this is that the amortization already reflects the expectation of premature pay-offs and thus, to write-off the loan, is redundant.  This, however, creates the opportunity for impairment when refi’s make up the preponderance of new originations.

To illustrate this, I took a hypothetical portfolio of 2,500 loans ($550MM) and valued it over six periods.  Each period has the identical prepay and other assumptions as the last.  Accordingly, the overall value hovered around 1.0% (5.5MM).  However, in each period I assumed that 100 loans refinanced.  The new loans have the identical characteristics as the old loan except for the origination and maturity dates.  Accordingly, the current principal balance of the servicing portfolio does not change over the time period reviewed.  The value remains relatively constant over time,  diminishing only slightly as the portfolio ages.  Basis equal to the fair market value of the new loans was added in every period, and amortization was calculated based on the portfolio’s decreasing economic life. 

The analysis was run twice: first with no write-off of the basis (“amortization only”) on the paid-in-fulls.  Amortization is thus relied on to keep the basis in line with market value; and second, with a “write-off” of the paid-in-fulls’ bases as the loans are refinanced.  The differing results are dramatic.

Amortization Only – The red line on the graph below shows the rapid escalation of basis as new loans, and their respective bases, are added.  This is, admittedly, an extreme scenario.  I have assumed a high volume of pay-offs, all of them are refinances, and 100% of the refis are recaptured by our hypothetical mortgage servicer.   It is clear why the impairment is growing so rapidly in this scenario.  The growth in basis is substantial and, while amortization is growing as well, the net of the two fails to adequately reflect the fact that the market value of the portfolio is essentially unchanged over this time period.  Since market value of the overall portfolio is essentially static, impairment grows exponentially. 

Write-off – Comparing that however, to a Write-off scenario, where the basis on paid-in-fulls are written down, the dynamics change dramatically.  The green line in the same graph shows the basis as adjusted for the write-off of the old refi’d loan.  As is evident, there is no impairment under this scenario.  The “write-offs” are simply recognizing as a permanent impairment that which would be considered temporary under the Amortization Only method.  Since the resultant basis is lower, this approach has the additional benefit of reducing amortization going forward.  Over time, there is no difference in the two methodologies.  100% of the basis will be written off eventually, it is just a matter of when and how. 

Writing-off the basis on loans that have refinanced may not be popular with the folks in originations.  After a particularly good month of originations, no one wants to hear that the net effect on the bottom line approaches zero.  However, we are deluding ourselves, in a high refinance environment, in thinking we are making progress.  The refinance business, from a servicer’s perspective is at best a zero sum game.

The mortgage industry’s growth has historically been driven by population growth and real estate appreciation.  In this growth scenario it may be defensible to add the basis on the purchase money loans added and rely on amortization to reflect pay-offs.  Neither is the case today.  Absent immigration, our population is expected to be stagnant through 2050, and real estate values continue their inexorable decline towards more sustainable levels.  Our total industry servicing rights accordingly, are not growing.  We are simply moving them around from one company to another.  In this environment, it is prudent to write-off loans that that have refi’d.

 As published in Mortgage Banking Magazine, May, 2012

May 15, 2012 Prepay Speeds

Mid-month SIFMA prepay speeds have been uploaded to the L1 Analytics valuation model.  As can be seen below, these PSA speeds have increased from last month-end.  This is not a surprise.

The Freddie Mac primary mortgage market survey shows the 30 year fixed mortgage at 3.83% as of May 10th. In concert with HARP2 and the mortgage servicer settlements, the MBA application index is up 1.7% for the week ended 5/4.

Servicing Discount Rates Need to be Revisited

The interest rate decline has not been kind to those who own mortgage servicing rights.  As interest rates dropped, prepay speeds soared, and values plummeted.  As can be seen from the graph below, all of the major servicers experienced significant declines in servicing values over this period (source: 10Qs).

It is most probable that we will eventually enter a period of increasing rates.  This is a time when the servicing asset really proves its worth.  It is also the time to prepare for the next rate downturn.  Discount rates (yield requirements) in our industry have always been “sticky” as rates rise.  As mortgage rates rise, discount rates increase, but nowhere near enough to take into consideration the rapidly growing risk that rates will eventually come down once again.  The yield that we require needs to increase appreciably as rates rise to compensate for this increasing payoff risk.

The two largest drivers of servicing values are prepayment speeds and discount rates.  There is always a great deal of attention focused on prepay speed projection methodology, but much less on the discount rate.  The discount rate should reflect the return we demand on this investment.  The rate currently used is often either subjective or based on rudimentary rubrics such as an x spread over a 10 year Treasury.  A more logical approach to deriving this rate will not only make it more defensible to auditors and regulators, but can also be constructed to prevent large write-downs in the next interest rate decline.

As with any other investment, the discount rate should reflect the returns we could receive on other alternative investments plus or minus increments for the various risk elements unique to the servicing asset.   In the author’s opinion, the basis for the discount rate should be the mortgage rate, not the 10 year Treasury.  Mortgage rates are more easily tailored to the characteristics of the servicing portfolio (e.g. 30 year, 15 year, 5/1 ARM, etc), are readily discoverable, and already reflect many of the risks inherent in the servicing asset.

The mortgage rate, however, does not exactly reflect all of the risk dynamics of the servicing asset.  For instance servicing has additional operational risks and is a less liquid asset.  It also could be argued that, in most cases, its credit risk is less than the underlying mortgage (absent repurchase risk). It has a similar maturity risk to the underlying mortgage but, because of its negative convexity, has a much greater volatility of return as expected maturity ebbs and flows.  That is, when rates drop and prepay speeds increase, the mortgage investors at least gets their investment back (more or less), while the servicing investors’ cash flows simply stop.

Given today’s somewhat limited market for servicing, it appears that this spread over mortgage rates for newly originated, 30 year, fixed-rate, agency servicing approximates 450 basis points.  This would put today’s base discount rate in the 8.5 – 9.0% range.  While this spread needs to be tested regularly, it should be a fairly constant function throughout an interest rate cycle.  It is the maturity risk that needs further focus.

If you look at the period between January 2000 and December 2011 (right), mortgage rates had a peak of 8.33% and a trough of 3.96%.  While this range of rates is nowhere near the range from its historical peak (mortgage rates hit 18.45% in October of 1981), it is still a wide enough rate swing to cause great consternation to holders of mortgage servicing rights.  (source Freddie Mac PMMS)

If you were to accept that a reasonable range of expected mortgage rates through an interest rate cycle is 4.5% to 8.0%, I would make the argument that the higher the current market rate when the loan is originated, the higher the probability that it will experience a lower rate environment at some point during its life.  Thus the riskiness of a higher coupon loan (i.e. its potential volatility of return), is greater than a mortgage originated at the low point of the interest rate range.

In fact, if I value a 5.1% thirty year, fixed rate, conforming mortgage as of April 2010 (current market rate at the time), I get a value approximating five times service fee. This is probably what the MSR would be booked at as of that date.  Unfortunately, subsequent to April, the market dropped to a low of 4.23% in October of that same year.  Accordingly, the valuation of these same loans dropped significantly, creating the potential for a significant impairment over only six months.

I looked at this phenomenon over all coupons (in 50 basis point increments) from  4.5 to 8.0%.  The results are below.  If the same loan as mentioned above were valued using discount rates and prepay speeds in common usage at the time the coupons were “market”, I get a range of values from a 5.0 multiple when rates are low, down to a 4.5 multiple when rates are high.

The problem comes in when rates migrate to the low end of their cycle.  As can be seen below, the higher coupon loans (8.00%) lose approximately a full multiple (or more) in value dropping to 3.44, while the lowest coupons lose nothing (as would be expected).

Unfortunately, the solution is to recognize this phenomenon at the time of booking the servicing asset.  At a current mortgage rate of 8%, there is substantial maturity risk.  It takes an additional 9.0% over and above the regular discount rate to adequately take this potential rate drop into consideration (see below).

I would not conclude, however, that in reality this full 9% needs to be added to the discount rate.  This is the extreme and assumes that rates drop to 4.5% overnight.  This has never happened.  Rates tend to migrate up and down rather than soaring or plummeting overnight.  Because servicing cash flows, and its economic benefits, are heavily skewed to the first three years of a portfolio life, you may feel comfortable with a much smaller “option spread”.

Option risk increases as rates rise and needs to be taken into consideration. Because we do not know when rates will rise or fall, it is a projected volatility that can be addressed in our yield requirement.  Now is the time to start reflecting this in capitalized value, while we are still in the trough of the interest rate cycle.

MSR 3rd Quarter 2011 values

As anticipated, MSR prices, as of September 2011 10Qs, are down once again.