REQUIRED READING: Do Credit Characteristics Help To Identify Fraud?


It comes as no surprise that mortgage finance transactions are conducted very differently today than they were in the last few years. One could even say this is a ‘new world.’ Much of 2007 was spent re-evaluating business, operational and risk-management techniques and capabilities. In particular, the pendulum has swung back from production-oriented business processes to risk-managed business processes.

While mortgage fraud has been an industry consideration and consternation for several years, market condition changes have not caused the prevalence of fraud to wane. Based on millions of transactions being evaluated for fraud risk every year using automated fraud detection tools, it is possible to track fraud-propensity trends. Unfortunately, these trends are on the rise. The risk of fraud is 53% higher than it was three years ago.

As risk management practices are evaluated and rising fraud propensity is considered, can traditional risk characteristics be used to identify fraud? This article addresses that question by considering four of the most important and risk-predictive characteristics lenders traditionally use: loan-to-value (LTV) ratio, debt-to-income (DTI) ratio, FICO score and owner-occupancy status.

LTV is an important characteristic lenders use to evaluate credit risk because it measures a borrower's equity stake. LTV serves as an indicator of the borrower's seriousness and a buffer against foreclosure and loss severity in the event asset prices decline. A borrower with a low LTV is much less likely to experience a negative equity position over the life of the loan (even when asset prices decline) and would be able to refinance or sell the house and pay off the mortgage if he were unable to pay the monthly obligation.

But if the LTV is fraudulent, or more specifically, if the valuation in the LTV denominator is significantly inflated or unsustainable given local market conditions, the actual LTV is higher than reported. In many fraud schemes, particularly those used to extract sizable monetary gain from loan transactions, perpetrators inflate valuations to substantiate larger loan amounts at seemingly low-risk LTVs.

Therefore, the credit-risk characteristic is fraudulent and cannot be relied on to accurately assess risk. The lender must ascertain the true LTV – if collateral value is inflated or unsustainable, the LTV understates the transaction's true risk. It is therefore a circular argument to use the LTV as a variable to detect fraud when the LTV itself may be fraudulent.

The FICO score is certainly a strong predictor of risk, but when used in a credit-risk assessment, a lender must be certain the FICO score belongs to the borrower applying for the loan. If the borrower identity is stolen, a straw, or the FICO has been artificially inflated through the piggybacking of good credit on bad, the credit-risk assessment can be dramatically incorrect.

Misrepresenting owner occupancy status is one of the most common and historically prevalent fraud tactics. It is used because, all other characteristics being constant, lenders give owner occupants better interest rates and lower mortgage payments than they give investors. Lenders know from experience that an investor is more likely to default if a property is worth less than the mortgage; it is not the roof over the investor's head and is treated more like a financial asset than a home.

Furthermore, investor properties tend to have more condition issues than owner-occupied properties, increasing the lender's loss severity in the event of foreclosure. If an investor falsely indicates owner-occupancy status, credit risk is understated.

Identity-based fraud detection tools can validate borrowers' identities, determine their likelihood of being straw buyers, and identify if they already own multiple properties. Such identity checking helps to determine if the FICO score and owner-occupancy status used in the credit-risk assessment are correct. Income-validation tools quickly and efficiently rank the likelihood that an income is inflated and a DTI is therefore deflated.

DTI ratios, given less importance in recent years, are regaining popularity. They are an excellent barometer of a borrower's fundamental capacity to repay a recurring mortgage debt obligation.

As with LTVs, fraud is linked to the DTI because perpetrators often inflate income to create the illusion of a capacity to pay. It is important to note that fraudulent incomes, and therefore fraudulently low DTI ratios, are not only found in low- and no-documentation loan products. Income documentation fraud (falsified pay stubs, W-2s, and tax returns) existed long before stated-income loans.

Stated-income products simply removed falsifying documents from the to-do lists of those committing fraud. Income documentation fraud is expected to increase in full-documentation products as stated-income products become more difficult to obtain. Again, using DTI to detect fraud creates a circular argument.

To detect fraud, lenders must examine the loan transaction characteristics that are complementary to traditional credit-risk characteristics. Fraud detection tools do that by using alternative data sources and algorithms to validate the traditional characteristics that credit-risk assessment models use. Think of the fraud detection tool as the "fact checker" that identifies and assigns risk metrics to the likelihood of a "fact" being incorrect.

For example, collateral risk scores assess the risk of overvaluation and unsustainability of a proposed value (the value being considered for the denominator of the LTV). The higher the collateral risk, the more likely the given LTV is to understate the "true ratio."

Successful fraud identification lies in the ability to detect and recognize when the facts stated (value, income, FICO, occupancy status, etc.) are most likely to be untrue. Basing a loan credit-risk assessment on these "facts" without fraud detection fact screening yields a prediction of risk predicated on fraudulent or misrepresented information. Characteristics that can be manipulated and used to facilitate fraud cannot be used to assess fraud risk unless validated by other data sources and analytic algorithms.

Fraud detection tools provide a complementary view of risk, separating loans with a high risk of fraud into a group quite different than those assessed as high risk based on credit characteristics. An effective fraud detection tool does not use the credit characteristics to identify risk but validates the reasonableness of the credit-risk characteristics. Only when the fraud detection tool verifies those underlying facts will fraud-risk assessments and credit-risk assessments combine to yield effective fraud protection and informative risk predictions.

Mark Fleming is chief economist with First American CoreLogic. He can be reached at

Notify of
Inline Feedbacks
View all comments