The Emperor's New Clothes


MSA Limits in Basel III
August 23, 2010, 5:00 pm
Filed under: Mortgage Markets, Valuation Tools | Tags: , , , , ,

The proposed capital rules under Basel III would limit capitalized mortgage servicing assets (MSAs) to essentially 10% of Tier 1 capital.  This could adversely impact servicing market values.  Accordingly, I took a quick look at whether or not there may be a problem.  My conclusion is that there may be some dislocations at the bank level, but there is not a substantial systemic risk to servicing values.

There were 7,941 banks and thrifts in the United States as of 3/31/2010.  Of these, 1,137 had capitalized servicing (MSAs) on their books.  67 of these institutions had MSAs that exceeded 10% of Tier 1 capital, the remaining 1,070 were under 10%.  The bad news is that in order to reduce their MSAs to 10%, the 67 institutions would need to reduce their holdings by $24.8B.  This equates to approximately $2.8T of mortgage servicing principal balance at an assumed value of 90 basis points.  The good news is that the remaining 1,137 institutions, that are under 10% concentration, have adequate capital to absorb essentially all of this $2.8T ($2.6T anyway) if so desired.  This assumes that the banks that currently have no servicing wish to remain that way (a good bet for the most part).  It also assumes that non-bank mortgage servicers will not absorb some of this product.  This is probably not the case.

There are several ways an institution can address their overage:

  • Sell part of the portfolio – Only 28 of the 67 “over limit” banks are over by greater than 10% of Tier One capital and may need to sell. Their overage aggregates only $350B of servicing principal balance. 
  • Accelerate amortization and sell more loans servicing-released – It is conceivable that the other 39 institutions will manage their concentrations down through a more accelerated amortization combined with more servicing-released sales.  Additionally, normal prepayments and curtailments will also reduce their exposure materially before the proposed regulations take effect in 2012.  

Implementation of these capital limits, while non-sensical, should not create a large supply/demand imbalance and, therefore, should have little impact on servicing value. 

NB … please let me know if you would like to see the bank level data that went into this analysis.  Also, I would appreciate your thoughts on this subject.



OAS Analysis is not a Valuation Tool
August 11, 2010, 12:44 pm
Filed under: Valuation Tools | Tags: , , , , , , ,

Option-adjusted spread (“OAS”) analysis is not well understood by our industry.  OAS should not be used to generate an assessment of value but, rather, as a very useful output of the valuation process.  It is an important financial tool that adds tremendous insight into the risk dynamics of mortgage related assets, but it does not produce a market or economic value.

The process of developing an OAS is very telling:

  • Generate a large number of randomly generated future interest rate paths;
  • Produce cash flows along each path at a risk-free rate plus a spread;
  • Obtain a simple average of the net present value of these cash flows;

This produces a simulated price.

  • If this simulated price does not equal the market price, choose another spread and rerun;
  • Repeat this process until the simulated price equals the market price

The resultant spread is the OAS.  This OAS is calibrated to market value; not the other way around.

The current usage of OAS analysis as a tool to determine price implies that we know the risk spread that the market demands and, thus, can use this spread to determine price.  This is intellectual hubris.  The spread is predicated on a large number of very complex assumptions and models, few of which are directly observable in the marketplace.  These assumptions include, but are not limited to:

  • Rate volatility (the speed at which future short rates change from their current implied values
  • Rate constraints (high/low)
  • Mean reversion properties
  • Distribution (normal or log normal)
  • Yield curve model (instantaneous forward rate or short rate)
  • Interest rate path generation (single or multi-factor)

Additionally the OAS, while ostensibly a stochastic measure, is heavily influenced by a deterministic prepay model.  While most analysts would test their OAS with sensitivities based on 90% or 110% of these prepay models, this simply “measures the sensitivity of OAS to consistent misestimation of the prepayments … not to random fluctuations around the model’s predictions.”[1]

And, even if these assumptions were observable and defensible, the option adjusted value is still meaningless from a valuation perspective.  It is an average of hundreds of different interest rate paths.  Possibly one of them is the correct path (i.e. what actually occurs), but there is no assurance that this is true, nor is there even a meaningful probability that this path is the “average” path that the OAS concludes.  “It is extremely unlikely that a security will actually earn its calculated OAS.”[2]

If you look at the distribution of values that an OAS analysis produces, you will see my point.  A greatly simplified example is shown below. 

This simplified example shows only five possible outcomes: a yield on the investments of: 0%, 7.5, 15, 22.5 or 30%.  The model generates a frequency distribution of these returns as indicated from a low of 3% to a high of 40% (naturally they sum to 100%).  As is typical with mortgage related investments, the distribution is not normal; it is negatively convex (skewed to the left).  Because an OAS value is a simple average of values over the entire distribution, the OAS value in this case would result in an average return of 15% (the “mean”).  Not only is this value far from a certain outcome, but it is not even the most probable outcome.  The most probable outcome (the “mode”) returns only 7.5%.

I would not pay a price that equates to a yield of 15% when the most probable outcome is a yield of 7.5%.  Additionally, I would look closely at the dispersion of the expected returns emanating from these hundreds of paths.  If the deviation around the mode is small, I may be willing to pay the price related to the modal price (never the mean price).  If the dispersion, however, is large, I would discount the price substantially.

OAS provides very useful information about the expected cost arising from the mortgagors’ ability to prepay at will.  It does not, however, measure credit risk nor does it provide the answer to the question of “what is the value of this asset”?

OAS has become a fad; one number that ostensibly summarizes the entire range of financial dynamic of the mortgage asset.  Yet this is not what the architects of OAS intended.  They never “intended the OAS to be viewed as a ‘yield takeout’ over Treasuries.  Because it’s the result of an averaging process.”[3] Financial analysts love to talk about OAS while disparaging scenario and other analyses.    Yet, from a market value perspective, OAS is a very elegant and expensive way of being wrong.

————————————————————————————————————

N.B.  Blogs do not allow for a lengthy discussion of any subject.  I will be publishing a more comprehensive discussion of this topic in Mortgage Banking Magazine and will advise those who are interested as to its publication date.

My next blog will address level 1 (under FAS 157) alternatives to valuing mortgage related assets.  This will not be available until the first week of September.  For the remainder of this month I will be focused on producing an article for Mortgage Banking Magazine (October 2010 edition) on identifying paradigm shifts in our industry before they wreak their havoc on us.


[1] Robert W. Kopprasch, Financial Analysis Journal, May-June 1994; page 45

[2] Robert W. Kopprasch, Financial Analysis Journal, May-June 1994; page 43

[3] Robert W. Kopprasch, Financial Analysis Journal, May-June 1994; page 43



The Impact of Credit on Residential Mortgage Pricing
August 3, 2010, 1:17 pm
Filed under: Credit | Tags: , , , ,

Everyone knows there is a relationship between the cost of a mortgage loan and the borrower’s credit strength.  The precise relationship, however, is less clear.  Lenders expend a great deal of resources to determine the true cost of defaults so that their pricing is accurate on a risk adjusted basis.  Level 1 Loans has examined seventeen of the largest lenders’ rate sheets to quantify the effect that credit scores and loan-to-values have on their pricing; all other product characteristics were held constant.  We examined the pricing of a hypothetical loan with the following characteristics:

  • $300,000 principal balance; 1st lien;
  • 4.50% fixed rate coupon; 30 year term;
  • Collateralized by property located in Ohio;
  • Full documentation; conventional

We looked at pricing from our database of the major aggregators’ product offerings, underwriting guidelines, stipulations and rate sheets.  This data is updated daily.   The results (below) show a spread of 310 basis points between the highest price offered for this hypothetical loan (102.17 @ 60% LTV & 780 FICO) and the lowest (99.07 @ 95% LTV & 620 FICO).  As can be seen, this is not a linear function.  In fact, pricing is almost flat for FICOs >= 720 with LTVs >= 65%.  Below this 720 threshold, however, the expectation of losses climbs and prices plummet.

Loss expectations are driven by the probability of default (PD%) and the severity of a projected loss, i.e. the loss given default (LGD%).  While mortgage lending is a behavioral science, and many demographic factors relate to mortgagor defaults, we have limited this review to loan-to-value and credit scores.  The PD% is primarily driven by credit score, while the LGD% is more a function of LTV; although the two are inextricably intertwined.  Credit scores have minimal impact on pricing at the lower LTVs (62 bp swing @ 60% LTV) while they influence pricing by 285 bps at 95% LTV.  Likewise, LTV variations have de minimus impact on pricing (26 bps) at credit scores over 720 but a 248 bp swing in pricing for the lower credit score mortgagors.

On a $300,000 loan, a swing in pricing of 310 basis points implies expected losses of approximately $9,300.  This expected cost is passed onto the mortgagor in terms of either upfront points and/or a higher coupon.

———————————————————————–

For  additional information please feel free to access our daily database of loan pricing indices (LPI) at http://www.L1Loans.com