Newsletter Sign-up

By NEAL DIHORA, financial consultant; KYLE MROTEK, principal and consulting actuary; and MIKE SCHMITZ, principal and consulting actuary in the Milwaukee office of Milliman

Recently, securities such as mortgage-backed securities (MBSs) and collateralized debt obligations (CDOs), which are essentially bonds that include tranches of MBSs, have declined precipitously in value due to soaring defaults of subprime and other risky home loans, as well as prime loans. That makes getting a handle on the true worth of these securities all the more important.

BROAD ASSUMPTIONS RAISE RISK

One might be tempted to use a normal distribution of outcomes to quantify the risk, yet this will likely generate a false sense of security at best. Unexpected external events also disrupt normal distribution patterns, making the approach an ineffective way to measure risk. Long Term Capital Management, a hedge fund that made large bets on many different types of credits around the world, had to be bailed out in 1998 after Russia defaulted on its bonds, leading to an unexpected worldwide flight to safety and into U.S. Treasury bonds. Few could have predicted the savings and loan crisis in the late 1980s, the Mexican peso crisis in 1994 or the fall of Enron in 2001.

Normal distribution patterns do not account for such events, and relying on them tends to underprice risk.

For example, if the daily-price changes on the Dow Jones Industrial Average followed a normal distribution, they would have moved more than 4.5 percent only six times from 1916 to 2003. But price changes exceeded this level 366 times. This simple example underscores the need for more sophisticated measurements of the range of outcomes for mortgage credit losses which, like stock market returns, do not follow normal distributions.

USING DATA AT HAND

In order to best position themselves against losses, insurers need to measure the true risk they are holding among all the asset-backed securities on the balance sheet. Available data in the loan pool, along with economic considerations, can go a long way to estimating the frequency of default and severity related to the underlying collateral pool.

Holders of MBSs can collect loan-level data on the collateral of the underlying assets, such as interest rate, loan type, loan-to-value (LTV) ratio, loan size, occupancy, loan amortization type (fixed or adjustable rate), and credit scores. These data are readily available from various sources including servicer of the underlying loans. Risk officers can also gauge macroeconomic conditions.

COLLATERAL CREDIT LOSS

All of this data provides a looking glass into the ultimate performance of the collateral underlying the security. Common actuarial methods can be employed to forecast pool-level collateral performance. The first major component is collateral performance to date. Pool-level collateral loss is tracked and can be an input into forecasting ultimate collateral loss.

The second major considerations are the loan-level underwriting characteristics of the loans. Predictive modeling techniques can be used to relate credit and persistency performance to borrower, property and loan characteristics. For example, a loan that has very little documentation is more likely to default than one that has full documentation. A typical characteristic of Alt-A loans is limited documentation with respect to borrower income and assets. Subprime loans generally have low credit scores.

Economic- and macro-level items should also be considered. The equity a borrower has in the property is a major driver of collateral performance both with respect to credit and persistency. The borrower's equity is a function of the initial LTV and subsequent home price appreciation. Keep in mind that in recent months, independent financial institutions and federal, state and local governments have been introducing innovative programs to assist homeowners who are unable to meet their monthly payments.

As data and information on these programs and their effects develop, adjustments can be made to measure the impact of these and other programs to overall loss estimates.

USING CASH FLOW MODELS

Many MBSs and all CDOs have multiple tranches that have different payment triggers for scheduled principal, interest and losses. A cash-flow model delineates the overall collateral performance from original borrowers to the different security holders based on the deal prospectuses.

By feeding assumptions of ultimate collateral performance into a security cash-flow waterfall program, the asset holder can determine future cash flows based on interest, principal, and security losses. The cash flows can be then discounted to obtain a net present value.

GETTING CLOSER TO INTRINSIC VALUE

Holders of mortgage debt securities can use this method to arrive at an intrinsic value at any given time. Changing assumptions and repeating the process can allow for a general understanding of the sensitivity of the value. All of this provides the asset holder with a better understanding of the true worth of the security. This puts the owner in a better position to assess not only its own financial position but its competitive position on the marketplace, while providing critical information needed for strategic business decisions going forward.

August 1, 2009

Copyright 2009© LRP Publications