Printer Friendly

Calculating risks.

To risk means to expose to a chance of loss. Known or accurately predicted losses are not risks; they are considered part of the price of doing business. Some risks stem from the choice made by mortgage lenders to retain the servicing on a loan rather than opting for the risk-free choice of selling it.

For instance, prepayment risk stems from the chance that prepayment forecasts will not turn out fully on-the-mark. Likewise, default risk arises from the possibility that foreclosure probability or default severity will be worse than anticipated.

Risk, in and of itself, is neither good nor bad. Lenders must assume some risks, however, because a risk-free rate of return will not compensate a mortgage business for the value of its capital. Lenders who survive and prosper will be those who get the best assistance from sophisticated management tools designed to help price loans and manage risks most effectively.

One such portfolio management system that has been created to help mortgage bankers is the RF/Spectrum system developed by Reserve Financial Systems Corporation, the technology affiliate of Miami-based Reserve Financial Group.

For four years, clients have been assisted in making many types of management decisions with the help of this system. Here are some examples of the types of decisions and information that companies have used the system for:

* Decisions on selling servicing in order to meet accounting needs for income and to achieve reduction of prepayment risk.

* Evaluations of condition and composition of potential purchases.

* Evaluation of servicing department performance for such items as advances, uncollected late charges and delinquency progression that are not otherwise immediately apparent.

* Pricing of new production.

* Executive compensation decisions. (Can be used to implement a top management bonus plan based on real value gains and losses on servicing values measured quarterly.)

* Servicing hedging decisions. The system can evaluate whether a hedging proposal is worth as much, or more, than it costs.

* Dynamic portfolio optimization. This helps lenders set goals for servicing in the areas of REO, interest rate sensitivity and geographic risk distribution. The portfolio is reviewed quarterly to examine the gap between actual servicing position and goals that have been set. The system sets action plans to close those gaps.

* Management of default risk is handled by a subsystem of RF/Spectrum called the Residential Mortgage Default Risk System (RMDRS). Used by banks making acquisitions, this program tests a target bank's portfolio for default risk.

In mid-1991, this portfolio management system will become accessible directly to end-users through a mainframe system available through a non-exclusive licensing arrangement with Computer Power, Inc. (CPI).

The role of a decision-support system is to provide the insights, analysis and guidance to take faster and better actions. Although the mortgage industry has many talented risk-management specialists in secondary marketing, pipeline management and hedging, these tasks are highly challenging, and most of these specialists can only handle a set number of loans at one time.

The size and complexity of mortgage servicing portfolios, however, is beyond small systems' abilities. The sheer volume of data, lack of a central source and the complexity of the decisions to be made require large-scale technology with the speed to satisfy executives who require immediate action.

RF/Spectrum is designed to analyze portfolios by incorporating hundreds of bits of information about each loan to consider dozens of key elements affecting demographic, behavioral and real estate trends. These trends, in turn, help the lender evaluate risk. This article shows in detail how default, prepayment and production-pricing probabilities can be forecast using a large-scale portfolio-management system.

Types of risks

All risks can be divided into two categories. Systematic risks are common to all mortgage assets. They cannot be stripped away by making a portfolio more diversified, because they are associated with broad economic trends and financial market conditions such as the health of the national real estate economy and the movement of interest rates. For example, falling mortgage rates hurt servicing values of existing loans by increasing prepayment expectations and shortening the expected life of servicing assets.

Nonsystematic risks, by contrast, are specific to individual assets and can be reduced by diversification, or they can be eliminated surgically, one loan at a time. An example are those produced by damaging regional market conditions, such as the high prepayment behavior of 1987 California loans, or the losses produced by 1981 Texas originations. Another example of nonsystematic risk would be loan-type risks, such as high loan-to-value ratios or investor loans.

An obvious example of a nonsystematic risk-control measure would be the underwriting of a new borrower. Although this measure occurs just once in a loan's lifetime, the default risk is ongoing.

Mortgage risk management involves several components. First, it requires an understanding of each of the risks in servicing or asset lending and precisely how these factors interact to produce losses. In mortgage lending, those risks center on the behavior of the borrowers and the performance of their real estate.

Second, it requires discovering the exact degree to which those factors exist in a specific portfolio. This requires making a detailed analysis of portfolio composition and condition, sometimes integrating data from origination records.

Third, it involves pricing risks and quantifying their true effect under various scenarios of the future. Risk-pricing is the heart of risk management, and it requires the highest skills and the best analytic tools.

Fourth, the process of mortgage risk management requires acting to avoid, reduce or transfer assets that do not provide sound risk-adjusted returns, while retaining only those that do. Risk-reducing actions include servicing-released and bulk sales, regional pooling of servicing, nonrecourse securitization, pool insurance and hedging. The costs of those actions must be less than the value of the risk.

During the 1970s, mortgage products were relatively simple, and ongoing housing appreciation was an unquestioned assumption. Mortgage performance decisions were based on large, nationwide "benchmark" groups of loans, like the FHA Sec. 203 (b) or the private mortgage insurance companies' databases, which resembled actual portfolios in many ways.

The 1980s brought three developments that demolished much of the simplicity of the earlier business environment. First, there was an amazing proliferation of loan products, with adjustable-rate mortgages (ARMs), balloons, convertibles, 15-year products and many more. Portfolios today no longer resemble any benchmark group of loans against which mortgage bankers can project performance. Second, loan-quality risks increased--underwriting standards were relaxed, while more borrowers used bankruptcy and other methods to defend against complying with their mortgage debt obligations. Also, the "nationwide" real estate market vanished, along with the myth of uninterrupted growth in property values. Differences appeared betweeb regions, as first one housing market, then another, took punishing drops in real estate values.

The result is that now mortgage bankers are dealing with a market that harbors greater risk--but more opportunity. However, it also means that risk management is more complex, and the consequences of poor decisions are exponentially greater. Knowledgeable players, however, can profit in the relatively inefficient servicing markets where many participants still routinely misprice risks or fail to value them at all.

Risk management decision support

The diversity of loans and real estate market conditions, plus the materiality of risks, have led risk managers to demand loan-level decision-support systems that can evaluate the unique pattern of predictive traits of every loan. This "idiosyncratic" analysis is the new standard for risk management, replacing obsolete methods of aggregated, or "peer-group," analysis.

Decision-support systems are integrated structures that analyze, report and distribute information to support the most advanced problem solving. To be practical, they must be "action oriented," to provide fast, clear solutions to problems in servicing, production and marketing. These systems use comprehensive databases--not merely data derived from servicing--and make reports directly to managers, without intermediaries or complex reporting languages that require on-staff programmers to translate.

Additionally, a good decision-support system should provide ease of training for managers and other staff who do not have computer backgrounds. Last, the cost of a decision-support system should be measured and should be substantially offset by the benefit it delivers.

The RF/Spectrum system reflects these characteristics and can be used to evaluate prepayment, default and pricing risk, and to track delinquency progressions.

Evaluating and pricing prepayment risk

The value of servicing is so integral to mortgage banking that proper valuation is crucial. All servicers face this question: "How do I value servicing rights and correctly price prepayment risk in the face of interest rate volatility?" The answer is three-pronged: use the concept of "expected value," use the weighted average of multiple interest rate scenarios and avoid using static-case valuations.

Prepayments stem from three things: refinancings; demographic prepayments resulting principally from home sales; and defaults, or involuntary terminations.

Borrowers have many complicated refinance choices--they can have a new adjustable-rate loan, a 15-year fixed-rate, a 7-year balloon or even a loan from a stockbroker at prime rate. Refinances, therefore, are not related just to long-term interest rates, but to the slope of the yield curve, the attractiveness of competing loan products and interest rate expectations. Demographic prepayments are characterized by the vigor of housing resales and are partly nonsystematic, because they can be highly localized to a single market. The more active the market, the higher its prepayment rates.

Our system's sophisticated prepayment model links prepayments to loan type and interest rates and distinguishes between regions and local housing markets to allow for demographic differences. The slope of the curve of expected prepayments is illustrated by Chart 1.

Multi- vs. static-case scenarios

A static-case evaluation is one in which the valuation is computed under the assumption that the key variables such as market interest rates, prepayments and default patterns will not change over the life of the loan. This would imply that interest rates and prepayments will never change--an outlook that is wholly unrealistic.

Static-case values predicting prepayment behavior are misleading because prepayment risk is extremely nonlinear. That is, declining mortgage rates hurt servicing values far more than higher rates help them. This "negative convexity" is one reason it is crucial to use multiple scenarios of rates, which assume that rates will go unchanged, higher and lower. Chart 2 clearly shows a nonlinear response: a drop in mortgages rates of 2 percent causes a current rate portfolio to lose 80 basis points in servicing value, while a corresponding increase in rates only adds 12 basis points.

Yet, many market participants still use static-case methods to evaluate their portfolios. The reason for this is that many people are not aware of the pitfalls of static-case methods, or if they are aware of the problems, they don't know of any feasible alternative, since multi-scenario valuations require sophisticated programs that are relatively new.

RF/Spectrum calculates the value of servicing (and prepayment risk) in a series of steps. It begins with a valuation under a well-chosen base-case scenario, using a prepayment model sensitive to loan types, interest rates and demographic prepayments.

Next, the system performs valuations under several alternative rate scenarios, (for example, 100 basis points higher, 100 basis points lower) to test value under changes of conditions. Then, it calculates the expected value by first computing the probabilities of the alternative rate scenarios under differing levels of yield volatility. To get expected value, multiply the probability of each scenario times its value, and add the results together. Servicing buyers should use the expected value as the basis for servicing valuation and bids.

The difference between static-case value and expected value is the price of prepayment risk, unique to each segment of each portfolio. In Table 1, for example, the results of valuations are shown using five different static-case environments. The expected value of 162 is constructed by weighting the probability results of all the static-case scenarios. As can be seen, the difference between 1.00 and 1.92 in servicing value is nearly 10 percent between the two methods.

Each owner, buyer or seller of servicing should learn the risk-adjusted value of each segment of the portfolio. Portfolios with high risk-price relative to value should be sold. When hedging, the value of prepayment risk can be used to assess whether the cost of a hedge proposal is in line with its benefits.

Evaluating and pricing default risk

The various regional downturns that have affected real estate markets in the past decade have made default risk the pre-eminent concern for lenders. Whether setting loss reserves or pricing new loan products, the most pressing need for lenders may be in learning how to measure and price default risks in residential loans. RF/Spectrum's residential mortgage default modeling system allows servicers to run a stress-test on a portfolio, assuming various changes in housing market conditions to see the default results produced by different scenarios.

Pricing default risk requires three preliminary calculations: finding the probability of the default, determining the likely severity of default loss and arriving at the net present value of the expected loss.

Estimating default probability

Default probability has two major components: real estate values and loan-level risk--the most important of which is loan-to-value ratio (LTV). Default probability is calculated by the RF/Spectrum system by using default experience from a nationwide benchmark pool of loans, stratified by the most important factor--the original LTV. The system then adjusts the benchmark to reflect differences between the previous 8 percent home price escalation and the 5 percent escalation expected in the future, which translates into higher estimates of defaults. The default modeling program then uses this benchmark with each loan's original LTV to project base foreclosure rates.

Housing prices and default levels have not been uniform across the country. Even with nationwide appreciation, many individual regions and homes will suffer declines in price. To reflect that risk, the default modeling system stratifies the nationwide data into five housing market classifications--ranging from very strong markets with high historical appreciation to depressed markets duplicating the Texas market conditions in 1981 to 1982. Chart 3 shows how these five housing market classes stratify nationwide default probabilities, adding precision to forecasts. These default models can also be used to model any level of appreciation or depreciation.

The default modeling system applies housing market classes to 12 interstate regions, 50 states, 53 metropolitan housing markets and the origination year of the loan, for example, Houston, 80 percent, 1983; Houston, 75 percent, 1990; etc. With this capability, the system can differentiate not only between Riverside, California, and the stronger Sacramento, California, market, but also between Los Angeles loans made in 1986, which saw huge appreciation, and those made in 1990, which could perform far differently.

In addition to housing market class/LTV and origination year, the default modeling system applies up to 11 risk-weighted factors to every loan to distinguish it from the benchmark. These factors are called "default probability indices" and include loan type, loan subtype, amortization method, program, purpose and property types, occupancy, documentation, loan size and status. The annual expected default likelihood of a mortgage is calculated as the probability from the housing market/LTV and origination year model, multiplied by the product of all the default probability indices of the loan.

Calculating the price of default risk

As in the case of prepayment risks, default risks should not be priced under a single scenario, but under varying scenarios that show stable, improving and deteriorating real estate prices. Unlike interest rates, the probabilities of these scenarios will be subjective because the calculation of default risk lacks statistical methodology.

An accurate price of default risk, however, allows several extremely important objectives to be realized.

* Loan or servicing acquisitions can be priced more accurately, because subtracting the price of the default risk gives risk-adjusted value.

* Existing portfolios can be evaluated for setting foreclosure loss reserves.

* Servicing or securitization structures that involve recourse can be priced accurately, so that the risk-price of default is considered.

* New loan plans can be designed to minimize risk, and the price of that risk can be included in the loan pricing.

* Existing portfolios can be scanned for individual loans the carry a risk-value that is too high--leading to better sale targeting.

* Stress tests can design contingency plans to be executed when and if regional conditions deteriorate.

Establishing value-based production


Every mortgage banker shares a common concern for production profitability, particularly with volatile servicing prices. It's not uncommon to accept losses on originations and treat them as the cost of acquiring the servicing. A key question is: How can value-based production pricing ensure that the wholesale and retail operations are producing servicing that is worth more than it costs?

RF/Spectrum includes a focused component called a value-based originations (VBO) system, which operates to ensure that servicing values are in line with costs. This feature of the portfolio-management system uses a servicing production pricing matrix, (SPPM) which is produced for "generic" street pricing and boils down the complexities of servicing valuations into easy-to-use price options. The SPPM is given to branch managers or wholesale account officers who use its servicing values to create price quotes for new loans.

Another component called the "VBO Manager" captures loan information by direct download from an origination system on a daily, weekly or monthly basis and analyzes the actual loans, comparing them to the generic prices originally quoted by the SPPM. Actual loans will vary from the generic loans in the SPPM. Their true values must be captured early in the origination process, so that if their risk-adjusted value is unacceptable, they can be sold to someone who values them more highly.

The SPPM gives 24 simple pricing options for servicing production values in each state. The SPPM reflects the differences in value among states with interest on impounds, higher servicing costs, greater prepayments or poor default risk-pricing.

The VBO's third component creates "hindsight" management reports that compile the volume and value of loans produced by every originator or seller. This value is then contrasted to the price paid for acquiring them. When combined with RF/Spectrum's other analytic capabilities, a mortgage lender can have continued surveillance of producer quality, condition and prepayments.

A production pricing program continuously compares value received to price paid, making several important actions possible:

* The implied servicing prices used in pricing loans can be set to reflect true value received.

* Originators' and sellers' incentives can be better aligned with values, not just volume, produced.

* The quality and composition of loans from producers can be carefully tracked to meet agency requirements.

* Money-losing correspondents, producers or products can be quickly repriced or eliminated--increasing overall profitability.

* Better choices can be made between retaining each servicing right or its possible sale.

Delinquency progression behavior

The most important function of servicing may be delinquency management. Servicing systems are rich with delinquency data, but not with intelligence for decision support.

One of the challenges of servicing is to provide effective collections, while controlling costs. The costs are pervasive--a borrower not making principal payments isn't paying taxes or insurance, adding costs in several areas. Some estimate that the total costs of delinquency in every department add to more than 50 percent of all servicing expenditures.

A particular desire of servicing managers is to gain greater knowledge about the behavior of delinquent borrowers, so that scarce collection resources get optimal allocation. Obviously, multi-delinquents will always get attention, but what about loans in earlier stages? The residential mortgage default risk system will identify which loans present the greatest immediate financial risk to a servicer.

Mortgage lenders can find help with these questions by using the delinquency progression behavior program, another function of RF/Spectrum that analyzes which loans are progressing from current status into multi-delinquency. The delinquency progression behavior function offers users a powerful feature of RF/Spectrum--multiple-period reporting. The program also offers lenders an explanation of why certain loans progress into multi-delinquency.

This information also permits a new view of collection performance-- better collectors return greater percentages of early delinquencies to current status, allowing fewer delinquencies to progress, which cause losses. If, for example, it was found that 20 percent of 60-day accounts became current, while 30 percent of the same accounts became 90 days past due, a servicing manager would ask: "Is that acceptable performance? Did we cure the right loans, and what can we do to improve that ratio?"

Negative progressions are not in and of themselves harmful. Losses are harmful to mortgage lenders, and negative progressions lead to losses. Thus a servicing manager needs to make sure that loans are moving toward the less delinquent from the more delinquent--not the other way around. Systems such as RF/Spectrum are designed to help the collections department spend its valuable time curing the right loans.

By using a sophisticated decision-support system, several advances can be achieved:

* The loan factors that correlate to negative progression can be identified, and similar loans can be tagged for better, earlier contact.

* More meaningful performance standards for collectors' performance can be created.

* Loan and servicing acquisitions can be analyzed for progression, to provide better insights into delinquency dynamics.

* Automated intervention methods can be used to improve collections, such as five-day letters, a 30-day statements, etc., thus improving early detection in high-risk cases.


RF/Spectrum provides information and solutions for many decisions in servicing, production and marketing. The example above are just four of the serious issues mortgage lenders face daily. Some didn't exist 10 years ago. As the mortgage market continues its evaluation, and as real estate becomes even more challenging, there is little doubt that other issues will come to light that are, as yet, unknown.

Hunter W. Wolcott is founder, president and CEO of Reserve Financial Group, Miami, Florida, an investment banking, technology and information services group serving the mortgage industry for more than 11 years.
COPYRIGHT 1991 Mortgage Bankers Association of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:management tools for effective loans and risk management
Author:Wolcott, Hunter W.
Publication:Mortgage Banking
Date:May 1, 1991
Previous Article:An inside job.
Next Article:Appraising the profession.

Related Articles
Guarding against risk.
Managing mortgage credit risk.
Establishing a value-at-risk framework.
Secondary systems: managing risk.
A life preserver for loans.
THE COMPLETE Debt Risk Puzzle.
FNIS unveils risk-management tool and technology. (Tech Newz).
Frictionless compliance: compliance risks can be better managed by making use of automated tools before a loan is closed.
Getting help with fair lending: as regulators continue to turn up the heat over fair-lending compliance, new technology is coming to the rescue of...

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters