In the midst of a mortgage lending meltdown that continues to spread pain in many directions, what are the lessons to learn about score use?

While many lenders use the same tools—such as the FICO® score—to measure risk, their risk management practices that usually make the biggest difference in terms of performance. That’s an important point that can get lost in the noise around what contributed to the mortgage mess.

As lenders look forward, many want to go beyond simply tightening the reins, and instead want to control risk while also managing growth. Here is what we have learned as “best practices” from those lenders who manage risk well.

1. Understand the difference between credit scores and underwriting.

FICO® scores are an important part of lending decisions, but not the only part. The score provides an excellent overall assessment of consumer credit risk, but good underwriting considers a much broader view of risk.

Consider a pool of 100 mortgage applicants with identical FICO® scores but with different loan products and underwriting criteria:

50% booked with a 3/1 ARM with high LTV (loan-to-value), high DTI (debt-to-income) and no documentation requirements
50% booked with 30-year fixed rate loans with moderate LTV/DTI and full documentation.

Despite having the same FICO® scores, these loan pools would not be expected to perform the same. What drives the performance difference between these two loan pools is underwriting, not credit scoring. This is an example of how additional risk factors, not measured by FICO® scores, can heavily influence the loan performance for mortgage-backed securities, whether in the prime or subprime space.

Credit scores provide a risk assessment that informs lenders of the general rank-ordered risk associated with a given consumer at that point in time, based on past behavior. Numerous studies have proven that FICO® scores effectively rank-order risk across various consumer segments and economic conditions. What any given score means in terms of future borrower behavior will vary by lender portfolio and loan pool, especially when different underwriting factors are applied.

Underwriting, especially for mortgage, is a complex process that involves many factors, including a score-based risk assessment of the borrower. Many factors drive differences in the risk of a portfolio, including:

Score cutoffs
Underwriting requirements (DTI, LTV, documentation, owner occupied or investment, etc.)
Product terms and features (fixed, ARM & reset, conforming, jumbo, HELOC ‘piggy’, etc.)
Market forces (economy, competitive landscape)
Lender/broker market and sourcing strategy (Internet, branch, direct mail, telemarketing, brand recognition, etc.)

For example, product features—such as interest-only or other “exotic” loan features—can attract higher-risk consumers and result in a generally higher-risk portfolio. In recent years, some lenders have relaxed underwriting requirements such as DTI, LTV and income documentation. While this may enable more consumers to qualify for a loan, the performance of those loans will differ from the performance of loans booked with stricter underwriting criteria.

Understanding the consumer’s total debt load, not only as it stands today but also how it may change as conditions and introductory rates adjust, is equally important. To compensate, some lenders have recently adopted policies to evaluate consumer DTI based on reset rates rather than strictly introductory rates.

The point is that FICO® scores only give you part of the risk picture. Best practices suggest that lenders evaluate as many factors as possible — and understand the risk associated with all the factors when determining lending strategies and evaluating the performance of a portfolio of loans.

2. Be more cautious with risky consumer segments and ‘exotic’ loan products

The need for caution in dealing with risky consumer segments may sound obvious. After all, consumers in the subprime space are already projected as high-risk by the FICO® score — in fact subprime is defined in part by a low FICO® score. A high proportion of low-scoring consumers are expected to exhibit serious delinquency and default.

While lenders understand that subprime and non-traditional loans are riskier, it may not be as clearly understood that these segments/loan types may also be more sensitive. It takes less to change the performance expected — from bad to worse.

Even a little extra pressure—whether from increased debt or a change in economic conditions that dampens ability to pay—is likely to exacerbate the delinquency and default rate. This can dramatically impact the odds-to-score performance. Take a subprime segment where the normal odds-to-score relationship at 600 is one bad for every five goods—one more bad in this group makes a bigger difference that at a high score where the odds are, say, one bad for every 20 goods.

FICO® scores rank-order risk even among consumers with low scores such that the higher the score, the lower the risk. But default rates among low-scoring consumers may rise faster than historical results indicate, especially when market conditions exert additional repayment pressures.

This same phenomenon can also happen in near-prime or prime score ranges. For example, the odds-to-score relationship could change for a portfolio segment of customers whose monthly mortgage payment drastically increase when their introductory adjustable rate more than doubles. These lower-risk people suddenly find they have exceeded their capacity to repay not only their mortgage, but potentially other debts as well.

It is not surprising that when lending standards are loosened, loan pools showed higher default rates than would otherwise be expected within a given FICO® score range. As is happening today, market conditions can result in worse consumer performance at any given score band. In these circumstances, best practices indicate that lenders should anticipate higher rates of default than would otherwise occur.

3. When it comes to new strategies—test, test, test

In general, it is more difficult to gauge the impact of newer product and underwriting features on the risk of portfolios than the impact of changing a score cutoff. That is why it’s so important to test.

Controlled strategy testing has been proven one of the most systematic and reliable ways to determine the validity of a new product in a new market—without the risk of rolling it out on the entire portfolio.

With champion/challenger testing, a lender can test a strategy with known performance against one with unknown performance. If the challenger outperforms the champion, it becomes the new champion—until a later challenger overtakes it. This method can be used to improve anything from product features to market segment to decision criteria, such as income documentation.

While the cycle for results is longer for a mortgage loan than for other consumer loan products such as credit cards, many of the loans booked using high-risk lending strategies have defaulted early in the loan cycle. This outcome might have been better managed with judicious testing.

Today’s lenders must make increasingly complex decisions in decreasing timeframes, often in unfamiliar markets. Waiting for test results can be frustrating, but mass rollouts of untested strategies during speculative market conditions can be disastrous. Controlled strategy testing continues to be a component of best practices. This is true even in highly competitive environments, where capturing market share early is important.

4. Track score distributions and portfolio performance over time

In recessionary times — as you might expect — score distributions change, such that consumers tend to score lower on a macro level. But the rate and amount of score changes varies, and lenders need to understand their portfolios’ performance as it relates to scores, underwriting and market conditions, in order to forecast changes..

Trends among mortgage acquisitions show score distributions have been shifting downward as lenders have sought growth via underserved markets. Economic impacts such as regional unemployment, interest rate changes and property value shifts result in more consumers falling behind in all debt payments.

Lenders should also understand there may be a lag between the “triggering” event that may cause a consumer to behave differently and the time when that new behavior is captured in the credit information at the bureau and can be considered by the score.

The type and degree of impact to distributions and performance expectations can vary for each product type within a given lender portfolio. Changes in score distributions and performance can also be associated with a wide variety of other factors—including lender target marketing strategies, brand recognition, competitive environment, underwriting criteria, and offer and treatment strategies.

The point is that consumer and lender behavior are inter-related and change with the times—and so should lender practices. By tracking score distributions of your portfolios over time, you can employ strategies—such as cutoff adjustments or changes in loss reserves—that better manage the overall risk and quality of each portfolio.

Even among an overall group of consumers with the same score, odds-to-score may shift by portfolio or product depending on product features.

Unique lender practices will drive differences in the observed odds-to-score relationships, depending on, for example, which consumers are attracted by the offer extended and how these customers are managed. Any portfolio or loan pool that substantially represents a new or unique feature set will attract different consumers who may perform worse than the overall group of consumers with the same scores. (See sidebar.)

In today’s market, it appears that multiple lender actions and products are compounding this effect. It is especially important to evaluate changes in approval volume and default risk during times of uncertainty. Best practices also suggest that the more sudden and serious the impact, the more frequently those reviews should occur.

Thus, while the FICO® score continues to rank-order risk, it would not be surprising if a “flatter” odds relationship (less differentiation between goods and bads) occurred for a pool of high LTV/DTI, no-documentation loans than would be observed with a pool of moderate LTV/DTI, full-documentation loans that scored the same as the first pool.

Lenders should evaluate the distributions and performance of each product and underwriting variant separately to better inform future product terms and underwriting guidelines. Lenders who continually assess their loan performance by FICO® score and underwriting levels in this manner can adjust strategies in response to market shifts or to meet changing business objectives.

5. Fully leverage risk assessment tools

Lenders who better leverage risk assessment tools can gain competitive advantage. Fair Isaac redevelops FICO® scores to keep pace with changes in credit reporting data, lender practices, new data reporters, consumer behavior and market conditions.

Lenders who adopt the most recent FICO® score models can gain the benefit of improved predictions. Fair Isaac’s redeveloped FICO® 08 score is designed to provide lenders with improved risk prediction across the entire spectrum of credit risk.

The FICO® 08 score can increase predictive strength by 5%-15% without lenders needing to make dramatic changes to systems and operations. The largest benefits will address the high-risk segments that lenders are targeting for additional growth:

Consumers seeking new credit
Consumers with limited credit experience
Consumers in the subprime sectors that have blemished credit histories at the time of evaluation

The new models also remove authorized user account information from consideration, so that authorized user account abuse will not affect the FICO® score.

In addition, lenders can improve their risk assessment when extending credit to both underserved and traditional markets by using the FICO® Expansion™ score, which is based on non-traditional data such as demand deposit accounts, utility payment, and public records. When the FICO® Expansion™ score is used as a standalone score for populations with no traditional credit history or in conjunction with the FICO® score, lenders can further distinguish risk and augment lending strategies for subprime as well as prime markets for a wide range of credit products.

As consumer debt loads continue to rise and lenders seek additional growth opportunities, the ability to assess consumers’ capacity for additional debt is particularly important. Fair Isaac’s research into consumer capacity shows we can help lenders identify which consumers are best equipped to responsibly manage incremental revolving debt. Research is also underway to explore mortgage debt capacity.

Drive at your own risk

At Fair Isaac, we make the ultimate scoring machine—but you do the driving. Speeding in new territory is exciting and rewarding, but also comes with risk. Lenders who incorporate best practices into their lending policies should find themselves with better control on balancing risk and growth to steer profitability in any economy.

Careen Foster is the FICO® Score Product Manager at Fair Isaac. To find out more download the archive webinar from our Credit Management Series on the Scoring Quarterly Update from October 18, 2007, Scoring for Subprime—Expectations in an Uncertain Market.