Printer Friendly

Keeping an eye out for risk.

GNMA has successfully developed a sophisticated system for closely monitoring issuers' portfolios. The system now gives GNMA early warning signs of portfolio troubles that lie ahead.

Success stories in the realm of the federal government's financial and insurance regulatory activities have been difficult to find in the last several years. A notable exception, touted by both regulators and the regulated, is the Government National Mortgage Association's (GNMA) implementation of a concept called the Automated Risk Reduction Cycle (ARRC). This concept, developed by the Risk Analysis Group at Coopers & Lybrand, has brought about a fundamental change in the agency's managerial approach.

The GNMA mortgage-backed security (MBS) program is responsible for administering and monitoring more than $400 billion of mortgage-backed securities, covering more than 7 million federally insured and guaranteed loans. The issuance and servicing of these GNMA securities are performed by approximately 700 mortgage institutions, or issuers.

GNMA, like many of the other federal insurance programs, experienced financial pressures in the late 1980s, unlike that in previous years, with issuer defaults rising to an alarming rate. In 1989, GNMA's defaulted single-family portfolio rose from $0.4 billion to approximately $10.5 billion.

Important questions were raised at GNMA--questions without easy answers. How many more issuers were likely to default? Who were the potential defaulters? What, if anything, could be done to prevent future defaults? Was the insurance capital reserve fund sufficient to cover the cost of these future defaults?

GNMA contracted with Coopers & Lybrand in December 1988 to build the Mortgage-Backed Security Information System (MBSIS), which was implemented in June 1989. This mainframe system stores monthly pool and loan-level data received from the issuer community, showing the status and activity of the 270,000 pools and 7.4 million loans for which GNMA is responsible.

Even though MBSIS was updated monthly and contained valuable data, it was still difficult for GNMA to access the information necessary to answer the difficult questions at-hand. GNMA senior management realized they needed to convert the tremendous amount of data being collected into effective management information that would enable GNMA to monitor and evaluate its uncertain environment.

In January 1990, the C&L Risk Analysis Group began working with GNMA management to implement the ARRC concept. (Figure 1 shows MBSIS and the critical components that evolved from the implementation of ARRC and their interrelationship with GNMA and the issuer community.) The three major components that evolved from ARRC are: IPADS (Issuer Portfolio Analyses Database System), the expert system and the Issuer Feedback Report.

Issuer portfolio analysis data base system (IPADs)

Key to ARRC implementation was the development of IPADS. IPADS has become the cornerstone of a new managerial approach to not only monitor risk, but to reduce risk through industry self-regulation. IPADS provides GNMA senior management with ready access to summary level information on each of its major programs: single-family, manufactured housing and multifamily housing. In addition, IPADS maintains current information and statistics on each individual mortgage portfolio serviced by an issuer of GNMA mortgage-backed securities. IPADS is updated monthly and operates in a PC/LAN computer environment.

IPADS was designed to allow senior management and program analysts to:

* quickly access basic issuer portfolio information, such as remaining principal balance, number of loans and number of pools in the portfolio;

* evaluate a composite set of performance indicators to provide a broad perspective on the health of the issuer's portfolio;

* compare an issuer's portfolio statistics against established GNMA norms to provide a basis for analysis of the portfolio's condition;

* access 24 months of trend analysis on key variables that can be used to monitor improvement and/or deterioration in the portfolio; and

* perform easy ad hoc queries for more advanced analyses.

Figure 2 shows an example of what an analyst might typically see on what is called the "main screen" in IPADS for a hypothetical issuer called XYZ Mortgage Corp. This screen provides a snapshot of the key factors of XYZ's portfolio, including:

* size and yearly changes in size;

* delinquency-ratio analysis;

* twelve-month collection history;

* portfolio age and average mortgage rate; and

* FHA and VA proportional split.

Figure 3 shows another IPADS screen, the "trend screen," for XYZ Mortgage Corp. It displays the most recent 24-month history of the GNMA portfolio, highlighting remaining principal balance (RPB) and loan changes and delinquency ratio trends.

IPADS provides 13 screens on every issuer portfolio. Other IPADS screens include information on loan distribution by state, statistics on how efficiently an issuer is processing delinquent and foreclosed loans, standard income and balance sheet information and graphic displays showing delinquency rates and loan distribution by year of origination.

When deciding what information to include in IPADS, the Risk Analysis Group and GNMA determined key performance indicators associated with risk. Then, using the large computer files within MBSIS, the Risk Analysis Group aggregated and synthesized the data to develop risk indicators relating to issuer performance.

In developing the risk indicators, it was crucial to understand what edits needed to be run to filter data that might corrupt analysis and ways to eliminate analytical bias. For example, the analysis on a typical delinquency performance indicator at the pool level might include editing each pool for reasonable loan size and ensuring that delinquent counts were consistent with cumulative delinquent interest and principal dollars. To reduce size and age bias, an issuer's delinquency rate or foreclosure rate might be compared to a GNMA mean for issuers with portfolios of similar size and age.

Consistency--which was critical to this effort--was provided by multiple standardized performance risk factors and statistical analysis. Multiple indicators allowed for cross-reference checks between related risks. Standardized factors established a common language and basis for identifying and discussing risk between GNMA management, GNMA analysts and the mortgage community. Statistical analysis, using means and standard deviations, reduced the subjective interpretation and assessment of risk.

Once the performance indicators and other management information were compiled into IPADS, expert programs were created to analyze each issuer portfolio, and an Issuer Feedback Report was developed to send to each issuer. As a result of the expert programming analysis, GNMA can identify troubled portfolios and solve problems before they lead to default. The feedback report regularly shows each issuer how its portfolio compares to the industry as a whole, allowing for self-improvement without GNMA intervention.

Expert programming and analysis

When GNMA brought IPADS on line in June 1990, it was apparent that it contained valuable issuer information not readily accessible before then. However, it also became clear immediately that evaluating more than 700 portfolios on a monthly basis was still beyond GNMA's staffing capability. In addition, even the best analysts had difficulty maintaining a standardized evaluation among issuers. To address these issues, the Risk Analysis Group recommended that an expert system be created.

Before the development of the expert system began, the Risk Analysis Group set

down two principals to follow: determine a method for evaluating the results of the expert system output, and begin with simple program logic and add complexity to the logic only when necessary. The first principal, obviously, was necessary to validate the results of the expert system output and to counter skepticism about whether potential default risk could be evaluated. The second principal reduced the development cost. Simplicity in an expert system might seem paradoxical; however, if the program is too complex early in the development phase, problems will be difficult to locate and costs can soar.

To meet the first principal, two methods for evaluating the validity of the expert system output were used. First, since IPADS contained the same information on previously defaulted portfolios, as it did on nondefaulted portfolios, one test was to see if the expert system would identify the previously defaulted portfolios as problem portfolios. Second, 40 nondefaulted portfolios were chosen randomly to test. Three of the best analysts at GNMA and Coopers & Lybrand were then told to independently rank each portfolio on a scale from zero to ten, with ten being the worst case. Using correlation analysis, expert system results would be compared to those of the three analysts.

To meet the second principal of seeking to maintain system simplicity, the Risk Analysis Group tightly constrained the initial set of modular components used to program the expert system. Five major areas were chosen to evaluate portfolios and weights were assigned to them by GNMA in collaboration with the Risk Analysis Group. These five areas and their associated weights were:

* delinquencies (3.5);

* processing of problem loans (3.0);

* trend analysis (2.0);

* delinquent history of loans originated in the last two years (1.0); and

* proportion of VA loans, to account for no-bid costs (0.5).

After deciding on the various weighting factors, GNMA and C&L analysts discussed how each area should be evaluated and how the expert logic should be programmed. Initially, 10 programming submodules were designed. Areas that TABULAR DATA OMITTED were weighted higher had more submodules than those with lesser weights. For example, in the delinquency area, three separate methods for evaluation were programmed, whereas in the VA proportion area, only one method was used.

Because the size of a portfolio has an effect on statistical variance measurements, issuer portfolios were subdivided into four groups based on loan size. The raw scores of a submodule are sorted from worst case to best case for each of the four groups. The expert system scores the 10 percent of the issuers with the highest raw score in each group at the maximum rating for the submodule. The remaining issuers receive a prorated score from the maximum to zero, with the best 30 percent of issuers receiving a zero rating.

Initial results from the expert system exceeded expectations. Previously defaulted portfolios did show up as problems, and the correlation was high between the expert system results and those of the three analysts' test cases. In fact, the correlation between the expert system results and the three analysts combined was better than the correlation between any one analyst against the other two analysts and expert system combined. In effect, the results from the expert system were better than that of the best analysts.

The expert system's efficiency was as significant as its effectiveness. The three analysts took two days to adequately evaluate the forty test portfolios; the expert system evaluated all seven hundred portfolios in an hour. With the expert system, GNMA quickly could monitor the changing status of all portfolios without increasing staff. In fact, the expert system freed the staff to use their time more efficiently. Instead of using time to look for problems, the staff could solve problems that the expert system identified.

The initial expert system, which went on-line in October 1990, was a powerful tool, but it still had its faults. As GNMA began relying increasingly on the expert system, and as new data sources became available, improvements were proposed. For example, the initial expert system could not evaluate the effect of loan location on delinquencies. The expert system corrected this weakness when new loan-level data became available.

In addition, discussions with troubled issuers helped fine-tune analytical biases the expert system was not taking into account. As a result, the number of expert programming submodules increased from 10 to 18. Figure 4 shows how the current expert system is weighted, showing the submodular breakdown of the delinquency area.

Issuer feedback report

IPADS and the expert system are tools limited for internal use by GNMA. IPADS contains the data and provides the vehicle to view all the performance indicators. The expert system analyzes the performance indicators contained in IPADS and performs a composite risk evaluation on each issuer portfolio. An additional instrument, the Issuer Feedback Report, was developed to bring the issuers into the Automated Risk Reduction Cycle.

Launched in October 1990, the Issuer Feedback Report is mailed quarterly to each issuer servicing GNMA single-family and manufactured housing pools. Each report shows issuers, both numerically and graphically, how several performance indicators measuring aspects of their portfolios compare to industry standards. Even though most issuers generate their own performance indicators internally, few know how they are performing compared to industry-wide standards.

The industry standards used in the Issuer Feedback Report are accurate for two reasons. First, the entire population of pools and loans captured in MBSIS are used to develop these industry standards, instead of a sample set of data that could be biased. Second, GNMA has a long history of collecting data from the issuer community in a standardized format. Using statistical analysis and large data volumes makes consistent and fair analysis easy.

The Issuer Feedback Report is organized into several sections and includes the following:

* a letter from GNMA describing changes in the report and general information on the program;

* some "big picture" analyses, which change quarterly and offer exhibits of current interest;

* issuer-specific, pool-level statistics against GNMA norms;

* issuer-specific, loan-level statistics against GNMA norms; and

* a glossary of terms with definitions of the indicators.

The objective behind the Issuer Feedback Report is for issuers who rank consistently worse than industry norms to work internally on improving their performance. Improving the performance of the poorer portfolios moves the industry standard toward the optimum level of operation of the best managed portfolios. No one really expects such an industry optimum to be achieved; however, even small movements toward this goal can significantly reduce costs and financial risk, especially when the industry accounts for more than $400 billion in mortgage-backed securities and more than 7.4 million loans.

Even though the overall objective of the Issuer Feedback Report is to curtail risk to GNMA, it also helps to curtail issuer costs and reduce issuer risk. By stressing portfolio performance, the Issuer Feedback Report identifies and reduces risk by encouraging issuers to solve costly problems without government intervention. Through IPADS and the expert system, GNMA has vital information to help address problems with troubled issuers if it needs to; however, it is preferable for all concerned if such action by GNMA is not required. The Issuer Feedback Report's goal is to make this possible.

Following the "evolving product" philosophy of ARRC, GNMA has encouraged issuers to offer suggestions to improve the Issuer Feedback Report. These suggestions have permitted the original report to be fine-tuned. The first mailing of the report included an issuer-wide survey that allowed issuers to rate the different exhibits in terms of usefulness and to suggest improvements. This resulted in more effective exhibits. Overall, the issuer response has been extremely positive.

Through the use of statistical methods and quality control measures, the Issuer Feedback Report provides valuable information to GNMA issuers. An issuer's proactive portfolio management, in response to the report, completes the circle for the successful implementation of the Automated Risk Reduction Cycle.

Evaluation of ARRC implementation at GNMA

The Automated Risk Reduction Cycle concept has been in place now for almost two years and it has exceeded all expectations. When the implementation of ARRC first began at GNMA, the primary objective was to help reduce losses associated with future issuer defaults. Not only has ARRC reduced the amount of defaults beyond expectation, GNMA has found that many of the resultant ARRC products (for example, IPADS, the expert system and Issuer Feedback Reports) have enabled them to more cost effectively manage already defaulted portfolios.

And the ultimate result is that the anxiety at GNMA of "not knowing where the next shoe is going to drop," in terms of serious issuer problems, has been quelled. GNMA program analysts use their time proactively, rather than reactively, to solve problems. When GNMA senior management testifies to the Office of Management and Budget and Congress, they can answer many difficult questions that were impossible to answer two years ago.

Findings from the analyses performed in developing IPADS and the expert system also have resulted in GNMA correcting weaknesses in its policy standards. For example, the criteria used to evaluate issuer performance prior to the authorization of additional commitment authority has been tightened. The correlation between risk indicators and previously defaulted portfolios provided the knowledge to establish these tighter standards.

In the last two years significant economic and program events have put additional stress on the GNMA portfolio. Increased unemployment rates and stagnating home prices foster an increase in delinquencies and foreclosures.

Even so, GNMA has weathered the stress well. In the last two years, less than $100 million of single-family portfolios have defaulted to GNMA. This is less than 1 percent of the amount of such portfolios that defaulted to GNMA in 1989 alone. Considering the limited amount of new defaults, the normal paydown of GNMA-managed portfolios, and the sale of portions of previously defaulted portfolios, GNMA is currently managing about half the size portfolio as it was two years ago.

Based on an index in place since October 1990, when the expert system and the Issuer Feedback Report were first implemented, the delinquency rate for FHA/VA mortgages also is down significantly compared to that of conventional mortgages.

Much of the credit for this success is due to GNMA's commitment to and constant involvement with the Issuer Portfolio Analyses Database System, and with the expert system and Issuer Feedback Report developed in the ARRC concept. GNMA's commitment, combined with the C&L Risk Analysis Group's efforts, ensure continued improvements in risk reduction.

The Risk Analysis Group of Coopers & Lybrand in Washington D.C. jointly produced this article. The team of consultants who make up the Risk Analysis Group are: Jim Boswell, Jean Young, Xavier Gonzalez-Sanfeliu, John Kruszewski, Jim Moynihan, Teresa Luhn, Eunah Kim, Hakan Beygo, Dominique Gillerot, Derek Price, Tristan Bostone, Dawn Thomas and Lorrie Cataneda.
COPYRIGHT 1992 Mortgage Bankers Association of America
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Government National Mortgage Association's Autonomy Risk Reduction Cycle for monitoring issuers' portfolios
Publication:Mortgage Banking
Date:Oct 1, 1992
Previous Article:RESPA's new enforcement muscle.
Next Article:Managing the burnout factor.

Related Articles
The MBS markets: 1989 and beyond.
A promising outlook: the quiet progress now evident on the commercial mortgage securitization front mirrors the early growth stages of the...
Changing capital market reveals few vehicles.
Mortgage conduits emerge as new capital source.
From main street to Wall Street.
Riding herd on subprime.
CMBS market set '98 record despite fall disruption.
Finding a roadmap: a veteran of the commercial mortgage-backed securities business offers his unique and time-tested perspective on the current...

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