MBS & Real Time Information

September 21, 2009

Rob Wosnitzer conceived of the argument in this post and wrote it together with Martha Poon. It responds to the discussion of how real time information could contribute to MBS evaluation at the end of the previous post.

A.    MBSs have always been structured with an awareness of ‘real time’ shifts in the status of the individual loans/mortgages. Any loan pool – whether these are actual pools of loans or the derivative CMO/CDOs – is built around a set of assumptions about its likely prepayment rate, or even re-payment rate. A 30 year mortgage is scheduled to have a 30 year life.  However, homeowners can disrupt this schedule by paying more in principal at opportunistic moments, as well as by deploying refinancing strategies in various interest rate environments.

B.     Back in the day, the person whose job it was to interpret  ‘real-time’ information on mortgages was called a ‘tape-cracker’. This person would await the periodic release of a ‘data tape’ from the agencies – Fannie & Freddie – with reports on prepayment rates on pools of mortgages.  Tape crackers operated with a time-critical posture – the quicker the agency information was gleaned and input to re-price the MBS universe, the greater the advantage conferred to the trading desk which moved to reposition their book. (Rob: I always liked the name of that job: tape cracker — it was like possessing special, secret skills, like a safe-cracker or code-breaker spy.)

C.     The oscillating rate of prepayments has the potential to wreak havoc on portfolios.  For example, take a Portfolio Manager (PM) who purchases a CMO with an average life of 14 years. Let’s make the overall interest rate 8% (which translates into mortgage rates of roughly 8.75% or 9.00%). Now let’s say interest rates decrease and mortgage rates fall, by say, 1% to 7.75%.  In the new conditions, homeowners are more likely to refinance, paying off their mortgages early.  The PM, therefore, would get stuck with a pool of money that needed to be reinvested — but in a much lower interest rate environment overall.  So the only way the PM could ‘match’ his original 8% was to take on more risk.

D.    One of the consequences of this intrinsic instability in CMO’s was to create IO’s and PO’s – ‘interest only and ‘principal only’ streams – in which the mortgages would literally be stripped in half.  Here’s how it would work: if a PM thought interest rates would be stable or rise over time they would purchase the IO portion. Conversely, if the PM felt that interest rates were expected to drop, which would incite a wave of refinancing, they would buy the PO portion at a very deep discount to par (say, 80 cents on the dollar). If this second bet was correct the profit margin could be very handsome since prepayments pay 100 cents on the dollar.  (Rob: Over time, traders devised ways to create synthetic securities from these segregated streams. They combined IO’s with high-interest paying pools of mortgages and PO’s with low-interest mortgages.  Another strategy was to engineer a security from Treasury strips and IO’s from a CMO.)

E.     In the world of MBS, defaults are anticipated.  What ‘subprime’ means is that the security is structured to anticipate that a greater percentage of the pool may default.  This is why CMO deals are over-collateralized. The ‘residual’ holder of a CMO structure puts up pure equity in exchange for a percentage of the ‘residual’ mortgages which remain.  These residuals are rated below investment grade.  This means that if the pool pays off as expected – default rates remain under a certain threshold – the residual owner is compensated at an above-average spread and vice-versa.

F.      The debacle from sub-prime escalated because the residual holders were the banks themselves, or the loan originators (i.e. Countrywide).  As increasing defaults accrued, the protection provided by over-collateralization was exhausted.  So the problem was, indeed, sparked by declining credit quality, but what created systemic shock was the highly concentrated ownership of these residual tranches. Tett’s observation in ‘Fool’s Gold’ is that the problem of BISTRO structures was finding an outside party to assume the ‘residual’ risk. Enter AIG. As Tett showed, the issue became pronounced when an increasing number of banks started keeping the ‘residuals’ on their balance sheet (i.e. in SIV’s), and a few dominant players, such as AIG, stepped in to insure the associated risk.

G.    Given that defaults are already being anticipated in MBS’s – hence over-collateralization – how would real time information significantly alter the valuation process?  It was a known fact that (sub-prime) pools deteriorate.  The subprime problem wasn’t really related to time sensitive information.  It was the unusually sensitive structure of the loans (risk layering through exotic features in addition to weakly positioned consumers made subprime loans vulnerable to external shocks); and it was also a liquidity issue — buyers became unwilling to commit capital to these products.

H.    Consumer credit information in the form of FICO scores are only used at the time of MBS structuring.  As Martha’s work suggests, FICO scores helped MBS become liquid by enabling the credit quality of the borrower to be represented in a capturable and comparable fashion.  That FICOs change as credit behavior changes over time, might add real time element to MBS valuation. But it would also make the variability of FICO visible.  FICO becomes particularly volatile in economic conditions where unemployment and credit line retractions [see previous post] can dramatically affect consumer performance.  Visible variation would potentially erode FICO’s robustness as a secondary market facilitator.

I.       The assumption that real-time information would create some degree of stability anchors itself in the very idea of a self-regulating, self-policing market.  Given the history of CMO’s, it would be inevitable that traders and financial modelers would devise strategies to isolate those portions of the pools experiencing distortions, and create new vehicles to even out risk, just like IO’s and PO’s.  Rather than ‘regulate’ themselves, market participants would simply modify production to capitalize on the the newly traceable volatility implied by enhanced real-time information.