Dennis R Capozza, Ph.D.
University Financial Associates LLC

Mortgage lenders have become increasingly aware of the importance of location in the underwriting equation through bittersweet experience. During the 1980s, even prime mortgage loans acquired the default characteristics of the worst subprime loans. In Houston, for example, prime mortgages for some vintages defaulted at over 20%, measured in cumulative rates, while in New England loans of the same vintage exhibited almost no foreclosures.

But awareness that loan performance varies by location does not solve the problem for risk mangers. New tools are needed to deal with these local risks. Our approach at University Financial Associates LLC (UFA), which is implemented in our quarterly Nonprime Mortgage Report^TM (NMR), is to use NMR Multipliers™.

The NMR Multipliers are a second generation approach that evaluates the most likely scenarios for future local economic conditions. Expected loan performance is then linked to these future economic scenarios. This approach contrasts with the earlier methodologies that attempt to control for local economic conditions by estimating multiple scoring models by region. Implicitly, this earlier approach assumes that the economic conditions that prevailed in the recent past in an area will continue to prevail in the future.

By making the assumptions about future economic conditions explicit, lenders can enjoy many benefits, including:

• increased profitability from focusing on the best markets while avoiding the most risky ones,
• clearer benchmarking of loans by region,
• better aligned incentive compensation schemes,
• better control of macroeconomic risks,
• increases in lending volume and market share,
• and lower default costs.

Why Location Matters

Traditionally lenders have concentrated their underwriting efforts on credit and other borrower characteristics. The location of the loan, however, greatly affects loan performance by influencing both borrower ability to pay and the value of the underlying collateral.

Location affects loan performance through several channels. The first channel is through collateral. This is easily understood for mortgage loans where the security for the loan is a dwelling unit whose price can fluctuate widely depending on local economic performance. The past decade’s mortgage foreclosure experience in California illustrates the dependence of loan performance on collateral (see the figure below).  In California defaults rose dramatically as the local economy eroded in the mid-1990s and collateral prices fell.

A second channel is through the borrower’s financial condition. It is well known that borrower stress "triggers" many loan defaults. A period of unemployment will stress a borrower and increase the probability of delinquency and default. Divorce and health problems are other triggers that affect loan performance.

The third channel is through loan terms and competitive conditions. When lenders are highly profitable the natural forces of competition encourage entry of new lenders and a willingness to make loans on more attractive terms to weaker borrowers. Often aggressive lending will follow an economically favorable period, which produced the conditions of high profits that caused entry. Therefore, credit cycles are grounded in the underlying regional economy.

The New Approach

The Nonprime Mortgage Report, developed by UFA, is the first sophisticated analysis of mortgage risks by economic region. This research on mortgage default uses millions of loan observations on a quarterly basis.

Underlying the analysis and research is UFA’s proprietary ForeScore™ modeling framework. Four dimensions of mortgage loan risk are analyzed in detail and summarized in four component scores. The four dimensions are the customer, the product structure, the collateral and the regional economics. The borrower component includes commonly used credit scores as well as information on work experience, debt capacity and other financial characteristics. The new approach focuses on the effects of regional economics on mortgage defaults, prepayments and loan values, and provides a set of numerical values – the NMR Multipliers – which allow the user to mathematically account for these effects.

The NMR Multipliers address the effects of regional/local information through all these avenues to loan performance. Local information is drawn from economic, political and demographic measures.

All of this information is integrated into the analysis of the borrower, the loan terms and the collateral to obtain estimates of future loan performance by location. To standardize, the analysis is performed on a prototypical or constant-quality loan. The results for each location are then compared to the national average for a loan with the same characteristics. For example, if defaults in Ohio are projected to be 15% above the average estimate for loans in all states, then the NMR Multiplier for Ohio is 1.15.

How the New Tool Works

The essential output of the analysis of mortgage loan performance is a set of loan multipliers by state. The multipliers are typically between 0.7 and 1.45 (70% and 145%) and represent the level of expected defaults, repayments and prepayments over the life of a loan relative to the national average. These multipliers enable the user to risk-adjust any loan or pool of loans for the regional risks. For example, if a loan is located in Connecticut where the NMR Default Multiplier is 0.82, then expected defaults in Connecticut are 82% of expected defaults on the average loan in the U.S.  Similar multipliers are provided for voluntary repayments and total prepayments.

A Better Tool for Credit Risk Management and Underwriting

Most lenders use quantitative scoring systems of either generic (FICO® scores) or custom varieties. In either case underwriters will be using a "hurdle" score to reject loans viewed as too risky. Expected defaults are most critical to the underwriting decision.

To apply the analysis effectively, underwriters adjust the hurdle score upwards in unfavorable locations and downward in favorable ones. The exact adjustment needed depends on the relationship among the current scoring system, expected defaults and loan value. For example, if expected loan defaults are 20% lower when the FICO or custom score is 50 points higher, then in a state with a default multiplier of 0.90 (defaults are expected to be 10% below the national average), lowering the hurdle score by 25 points will result in the same average level of expected defaults in that state as in other locations. (See the Appendix below for an algebraic formulation.)

Advantages for Mortgage Servicers

Mortgage servicers are most interested in total prepayments since servicing fees are similar to an interest-only strip. Servicers can use the NMR total prepayment multipliers to adjust estimates of expected fee income as a function of loan location. For example, if a state has a NMR Total Prepayment Multiplier of 1.10 (10% above the national average), servicing rights will be less valuable since loans will not be outstanding as long, and will not generate as much fee income. A servicer can incorporate the higher level of expected prepayments into its valuations of servicing rights for that state.

Conclusion

The availability of high speed, low cost computing is enabling new technologies for lenders. One of the these new technologies, NMR Multipliers, have been developed by UFA to assist lenders with local market risks. In recent years defaults have been highly concentrated by state and region. The southwest, especially Texas in the 1980s, and California in the 1990s are examples where defaults occurred at more than twice the national rate.

The NMR Multiplier approach gives lenders a powerful new tool for their risk management arsenal.  With an understanding of local risks lenders can navigate aggressively through the nonprime lending space.

Appendix - An Algebraic Formulation

In the example in the text the sensitivity of defaults to the score, or score beta, is .4% given by:

The needed location adjustment is simply

Therefore, in this location, a lender can accept loans with a score that is 25 points lower and expect the same rate of defaults as in an average national location.