Reengineering quality control.
A QUIET EVOLUTION IS CHANGING THE MORTGAGE QUALITY process. The new approach is being driven by a changed corporate mindset about the best way to implement quality control programs and by improved technology in the never-ending quest for total loan quality.
Lenders and investors are reevaluating their traditional quality control techniques, which relied primarily on a post-funding evaluation of funded loans. The new mindset focuses on implementing an array of prefunding controls, with an appropriate change in responsibilities for quality control personnel.
In addition to this shift in the timing of quality control efforts, technology now allows lenders to independently verify loan information in a real-time environment and underwrite loans and appraise property using artificial intelligence. How can management take advantage of these technological advances, while at the same time ensure the required levels of control? To get some answers to this question it's necessary to explore the differences between yesterday's, today's and tomorrow's quality control environment.
Quality control in the 1980s
Many quality control procedures prior to the 1990s consisted of testing a sampling of loans after the loans were funded. Verifications were formally reverified, loans were reunderwritten, appraisals were reviewed by an independent appraiser, occupancy was checked and tax returns verified with the Internal Revenue Service (IRS). The reason many of these procedures were performed on a post-funding basis was simple--time.
The funding requirements that prevailed at that time could not be supported by available technology in a way that could independently provide information prior to funding. As a result, fraud, more often than not, was discovered after the lender cut the check. The lender was then faced with another dilemma--a loan unmarketable on the secondary market.
Due to the sheer quantity of information that needed to be reverified in a mostly manual environment, most quality-control measures consisted of sampling only 10 percent of closed loans. For these loans, the quality control reviews involved completely reconfirming all verifications (regardless of their materiality), reunderwriting files and checking for owner occupancy. Unfortunately, many quality control staffs were understaffed and could not perform these reviews in a timely manner. As a result, the information provided in management reports was stale, more often than not.
Furthermore, there wasn't sufficient time for the quality control staff to evaluate controls in the various operational areas such as funding branches, corporate operations, post closing, shipping and the like. As a result, management was provided with information on lapses in internal controls after such an extended period of time that it precluded quick corrective action.
Notably, automated systems to support the quality control function did not exist. Targeted audits were almost nonexistent because quality control personnel had to rely on manual systems or memory to determine which loans, brokers and appraisers should be targeted for additional review.
Lenders recognized the need for specialized prefunding quality control procedures and they also understood that quality control/audit personnel should place a high priority on reviewing those controls. But the available technology did not allow for effective implementation of these controls or for efficient quality control systems allowing personnel to spend time in the field. However, today all that is changing rapidly.
Technology has advanced to the point currently where online controls are built into the lender's origination system. Automated appraiser and broker systems provide controls to ensure that loans only are processed by authorized individuals. Additionally, the lender has real-time access to data bases such as that of the Mortgage Asset Research Institute (MARI), a Reston, Virginia-based company. MARI maintains a data base of information derived from public records that contains names of individuals and firms with convictions as a result of infractions relating to the real estate industry.
Additionally, Foster Ousley Conley, a Walnut Creek, California-based company, offers Loan-Watch, a nationwide cooperative data base of loans in process that is updated daily by many of the country's major lenders. Using highly sophisticated matching technology, reports are produced to assist these lenders in the detection of occupancy fraud, undisclosed liabilities, simultaneous transactions and double-loan locks.
In selected markets, lenders have access to statistical valuation products that can be used as either the primary method of establishing value or the review method for verifying the original appraised value. One such system, Automated Real Estate Analysis System (AREAS), statistically predicts the home's current value using the latest neural network artificial intelligence technology. In an appraisal review mode, if the AREAS predicted value is within an acceptable range of the Uniform Residential Appraisal Report (URAR) value, then the URAR value is considered supported. Although the acceptable range of value is determined by the client, typically that range is within 5 percent to 8 percent.
By using artificial intelligence, a lender can eliminate costly appraisal review methods, as well as any potential bias on the part of an appraiser and provide consistency in the value prediction process.
Also new on the scene are automated underwriting systems using various forms of artificial intelligence. These underwriting systems swiftly screen loan applications to ensure that they meet the criteria for a particular loan program and comply with current requirements of the secondary market investor and lender.
These systems eliminate time-consuming and expensive paper processing, reduce production costs, increase speed without sacrificing quality and provide consistency in the underwriting process. Automated underwriting is a critical element in the prefunding quality control process because it allows for implementation of the internal modules that check for loan quality based on the lender's and the secondary market's qualification requirements.
There are several automated underwriting systems on the market, including one being used in a pilot program offered by Freddie Mac, another under development by Fannie Mae, proprietary systems run by large individual lenders and a system from Foster Ousley Conley called Automated Quality Control Artificial Intelligence Underwriting System (AQUARIUS).
AQUARIUS was designed to both automate the underwriting process and complete the quality circle. Automated underwriting systems make use of a powerful combination of advanced technologies to ensure quality, lower production costs and increase productivity. These technologies--rules-based logic, case-based logic, traditional statistics and neural networks--work together to control quality.
Rules-based systems use a series of rules or statements to screen loans against secondary investor requirements, criteria and guidelines. Rules are written so that each loan can be judged to determine if it meets, or fails to meet, each rule or requirement. Rules can be combined to conduct more complicated analysis. As a result, confidence is increased that funded loans meet the requirement of a particular investor and thus, are marketable on the secondary market.
Statistically based systems use historical data to predict or score loans, loan types or borrowers from past experience. This element also reduces risk as it gives the lender another measure by which it can evaluate a potential loan.
Case-based reasoning divides loans into hundreds of categories and teaches the system to recognize similar cases for similar decisions. Again, this provides lenders with an additional tool for measuring risk.
Finally, neural networks, the most advanced technology available today, uses very large historical data sets to model the complex, interdependent relationship involved in loan decisions and loan performance. For example, AQUARIUS can be trained to emulate expert underwriters and to evaluate the borrower's ability and willingness to repay the loan. Neural networks add speed and low cost to underwriting applications. Furthermore, because data is collected monthly and models are frequently rebuilt to track new market conditions and lender requirements, neural networks allow for easy adaptation to external and internal underwriting changes, as mortgage risks vary over time.
In a nutshell, an automated underwriting system using artificial intelligence learns how to judge good loans from bad loans by accessing vast amounts of data, then evaluates the loan application according to the lender's requirements. The system then makes a recommendation to the underwriter based on the judgment. Using such an approach, underwriting decisions are provided in a consistent, non-biased manner, derived from an extensive base of knowledge, in a fraction of the time and labor previously required.
Automated quality control
Another critical breakthrough in the budding field of mortgage lending quality control has been the development of automated quality control systems. ACES, developed by Engineered Business Systems of Coconut Creek, Florida and Second Look, developed by Tena Companies of Minneapolis, enable quality control departments to automate the processing functions of post-funding quality control--long a labor-intensive process.
These systems merge data from a lender's loan origination system then automatically select loans for quality control review according to the lender's risk analysis requirements. Reverifications are logged and printed with bar coding, responses are managed and second requests automatically generated. The systems also provide management with reports of reverifications not received that require alternate procedures to be performed. Underwriting modules document the analyst's review and provide for overall grading of the loan. Sophisticated reporting allows senior management to determine areas of weakness in the lender's operations and allows management to determine loans to be targeted for additional review, both from a pre-funding and post-funding standpoint.
Another major advance in the field of quality control is the advent of statistical sampling. During the past three years, major investors, including Fannie Mae and Freddie Mac, have supported the use of statistical sampling where appropriate, versus the 10 percent sample generally required. Proper use of statistical sampling may allow the lender to sample fewer loans with the same confidence level as a 10 percent sample.
The key advantage of an automated quality control system and the acceptability of reduced sample size is that they create an extremely efficient post-funding quality control process and allow management to completely reengineer priorities to focus on evaluating prefunding controls. Prior to the availability of automated systems and statistical sampling, quality control departments generally focused their energies as follows:
* Post-funding quality control reviews of closed loans (processing, underwriting, reporting): 85 percent;
* Targeted reviews: 10 percent;
* Quality control audits of functional areas: 5 percent.
With the advent of automation and statistical sampling, priorities were rearranged as follows:
* Quality control audits of functional areas: 70 percent;
* Targeted reviews: 20 percent;
* Post-funding quality control reviews of closed loans: 10 percent.
This change in priorities necessitates a reengineering of the set of skills that quality control department personnel bring to their jobs. In the past, quality control departments generally consisted of processors and underwriters; today's focus is more on individuals with audit and/or analyst skills. Additionally, a moderate level of information systems skills is required as well.
With all these advances, one might well conclude that the prefunding approach to quality control has all but locked out the chance of bad loans being made. Unfortunately, this is not the case. What weaknesses are still present in the prefunding area where technology is not available? The answer is many prefunding processes are still performed manually and paper-oriented. Lenders still rely on manually prepared verifications of employment and deposit, with verbally reverified selected portions of that information, prior to funding. Flood certification processing, with the exception of one service provider, is still labor intensive with lots of paper in each loan file. But technological help is around the corner.
Electronic Data Interchange (EDI) is a key component in tomorrow's quality control environment, especially in the process of loan data verification. Instead of the current, written and verbal confirmations of the borrower's employment, bank deposits and outstanding debts, this information will be automatically obtained through EDI. Various companies are already working on communications gateways that electronically access and process third-party information, among them are Freddie Mac's Goldworks network and Computer Power Inc.'s (CPI) network. Through these gateways, the borrower's employment and income can be electronically and independently verified directly at the source. Most banks already have technology in place for electronic release of deposit information. One key stumbling block that we perceive in this process is security related: Did the borrower authorize release of the information?
When such obstacles are removed, the current two-week verification process will be completed in a matter of minutes. Borrowers will no longer face the hassle of supplying pay stubs, bank statements, credit reports and other information to the lender. Furthermore, this electronic and independent method of information-gathering will remove the possibility of fraud currently prevalent in today's process.
While the transfer of information is an important function of EDI, its true value lies in the interconnectivity between the many participants in the loan approval process. When loan officers, brokers, appraisers, title companies and credit bureaus can electronically communicate with each other, information will flow smoothly, automatically and fast. This instantaneous means of providing data is certain to save time, money and provide more accurate information for all participants.
The way it will be done
With the new quality control technology and processes in place, let's follow the trail of a typical loan application as it runs through a prototype system. When the hard file hits, data is entered into the lender's front-end system and the verification process is automatically initiated. Because the system has access to numerous banks, credit agencies and the borrower's payroll provider or employer, it dials up the appropriate participants and requests data verifications. The borrower need not supply copies of tax returns as the system is linked with the IRS. Occupancy is automatically verified with an online city directory service and comparison of addresses provided by the other participants.
Upon confirmation of the data, the information then moves on to automated underwriting module for underwriting. In addition, the system conducts a flood-zone check, and, if flood insurance is required, refers to an insurance company for processing. The system also records Home Mortgage Disclosure Act (HMDA) information for Community Reinvestment Act (CRA) management reporting at a later date. The system verifies that the subject property isn't located in an environmentally hazardous area. Additionally, the loan is compared against other lenders' pipelines to detect the existence of possible occupancy fraud, undisclosed liabilities, simultaneous transactions and double locks. Direct links with title companies generate and transmit the required title searches and title policies.
The next stop is the appraisal. With the advent of artificial intelligence, it is anticipated that investors will allow circumstances where the computer can replace the traditional appraiser. The property will run through the lender's statistical valuation model and determine a value: all within a fraction of the time previously spent. Utilizing this scenario, loans can be approved and closing documents prepared within a few hours.
New age quality control
Based upon this scenario, what will the typical quality control department look like? The traditional post-funding emphasis of the department will be minimized to an immaterial percentage of an analyst's time. Because the lender will independently verify employment, deposits, occupancy and other items, prior to funding, the need to reverify these items on a post-funding basis will be reduced or eliminated.
Analysts will need to have advanced electronic data processing (EDP) audit skills in order to adequately evaluate the company's system development and program change controls over its artificial intelligence automated underwriting and appraisal systems. Additionally, audit-type skills will be emphasized as quality control personnel will spend an ever increasing amount of time in the field. Does that sound as though quality control is going to encroach on the internal audit function? As a result of technology, a natural merger between the internal audit function and quality control will occur, as the emphasis shifts from a micro-level review of loans to a macro-level review of the lender's internal controls. Perhaps a better name for this new merged function is quality assurance.
The industry has experienced an astonishing amount of technological growth during the past five years. Even so, the next three to five years will provide still more exciting opportunities and challenges for lenders and quality control personnel in their never-ending quest for loan quality. The main challenge ahead for management is to adapt their companies to this radical change in the industry by reengineering their scope, objectives and procedures. Benefits will be reaped in the form of better quality loans and the timely and accurate disclosure of crucial information to the lender--before the loan funds.
Stephen J. Weimer is vice president of the information services division for Foster Ousley Conley based in Walnut Creek, California. He has over 16 years of experience in the banking and mortgage lending industries, specializing in financial and EDP audit, quality assurance and quality control.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||mortgage banking|
|Author:||Weimer, Stephen J.|
|Date:||Aug 1, 1994|
|Previous Article:||Discrimination without intent.|
|Next Article:||Tallying up servicing performance in 1993.|