Written for financial engineers, this book nevertheless can also be read profitably by anyone interested in mathematical modeling or mathematical finance. The authors discuss in fair detail the science of structured securities, which are financial products that are becoming more important as investors and financial firms continue to find more intricate ways of dealing with risk. For non-experts (such as this reviewer) in the field of structured finance, the book requires careful reading and attention to detail. Readers are expected to have an understanding of various mathematical topics such as Markov chains, linear algebra, Monte Carlo simulation, and probability and statistics. As an investment strategy, the authors describe structured securities as performing best in "controlled" environments. This involves the use of `transaction documents', which are used to keep their performance within an expected range, and also `macro-level' controls to assist in dealing with event shocks. The basic idea of a structured security is to assemble a credit or investment package from a variety of sources and allow them to be administered by third parties. This entails that the sources (the transferors) be completely decoupled from the transferee, the latter of which is called a `special purpose entity' (SPE), and which has an extremely low likelihood of becoming insolvent by its own activities. The SPE is an analogue of the obligor, and is also shielded from the consequences of the insolvency of a related party. Its assets are thus `perfected' against the claims of the transferor.
Early in the book the authors describe what they consider to be the two types of structured securities. The first, called the `long-term transaction model' applies to asset-backed, mortgage-backed, and collateralized debt issues with maturity at least one year. The second, called the `short-term transaction model' applies to asset-backed commercial paper markets.
If structured securities are to be used as an investment strategy, their value must be assessed in as fine a detail as possible. This assessment is of course the main goal behind the authors' book, and they therefore spend a fair amount of time in explaining why the usual credit rating strategies are inadequate for structured securities. One of those discussed is `benchmark pool analysis' which does not require a large volume of data and uses a microeconomic model of the obligors in a collateral pool to simulate the financial impact of economic shocks. Others discussed include the actuarial method, used for asset-backed and mortgage-backed transactions, and the default method, which is used for collateralized debt obligations.
The most interesting discussions take place when the authors attempt to formulate a more exact, analytical notion of rating for structured securities than what is available with the usual corporate rating model. Essentially the authors are advocating a "unification" of credit and market risk in structured finance in their attempt to replace the alphanumeric scale of the usual corporate credit rating by a numerical scale (they motivate this interestingly by discussion involving the `continuum hypothesis' from set theory). Most important in their approach is to view the pricing of structured securities as a nonlinear problem: rating and pricing are entangled with each other, in that to obtain the rating the promised yield must be known; but to find the yield, the rating must be known. There is of course a paucity of exact solutions to nonlinear problems, and so numerical techniques must be used. The authors spend a fair amount of time discussing these techniques in the book, and in formulating the problem of structured pools as one involving (Markovian and non-stationary) stochastic processes.
As a warm-up to the complications of asset behavior, the authors first discuss the modeling of liabilities. The collection and distribution of cash to various parties is contained in the `pooling and servicing agreement' (P&S), which is a legally binding document that contains a collection of payment instructions called a `waterfall' or `structure.' A waterfall codifies the payment prioritization taken from the funds that are available. Examples are given that illustrate their analysis.
For those not familiar with Markov chains, the authors give a short review, and argue that they are important to structured finance due to their ability to eliminate long-term static pool data requirements. The Markov chains used in structured finance are finite-state Markov chains, where the states correspond to recognized delinquency states of an issuer in some asset class. The transition matrices of the associated asset pools represent the credit dynamics of structured securities. The authors give three very detailed examples of their formalism, the first one of these, dealing with automobile receivable securitizations, should be familiar to most readers.
The last chapter of the book deals with `triggers', which generalizes the earlier discussion on liability modeling. The authors describe triggers as being the most `intricate' aspect of the analysis of structured securities. If one views them in terms of their physics analogy as control structures, they are fairly straightforward to understand. `Cash flow triggers' which allow a reallocation of cash but it does so without being too disruptive or expensive, are the only types considered in this chapter. The cash reallocation is obtained through the use of a `trigger index', which is usually dependent on transaction variables such as delinquencies or tranche principal balances. A trigger is `breached' if its trigger index is higher than a pre-selected threshold on any determination date.
The authors discuss four basic types of triggers, all of which are defined mathematically in terms of the proportion P(x(t)) of excess spread to be reallocated and some variable function x(t) of the trigger index: `binary', in which all excess cash is reallocated to the spread account when there is a breach at time t; `proportional', which allows a kind of "ramping up" of the triggering; `differential', where the excess spread is proportional to the first derivative of x(t); and `integral', where P(x(t)) is proportional to the integral of x(t) over a time interval with lower bound the breaching time and the upper bound the current time. Monte Carlo simulations are used to optimize trigger mechanisms.
Click Here to see more reviews about: The Analysis of Structured Securities: Precise Risk Measurement and Capital Allocation (Hardcover)
No comments:
Post a Comment