Investors in mortgage-backed securities were unaware of the widespread financial fraud involving securitized nonagency loans during the recent financial crisis. Misrepresentations by homebuyers, appraisers, loan originators, and underwriters all contributed to the losses experienced by mortgage-backed security investors.
The authors review home loans originated during the period leading up to the recent financial crisis to detect unreported second liens, misreported owner occupancy, and overstated appraisal values. They examine three misrepresentations to determine who was likely responsible for initiating fraud involving securitized nonagency loans, as well as the implications of misrepresentation for loan performance.
How Is This Research Useful to Practitioners?
Misrepresentation is associated with 48% of loans examined by the authors, a rate that is consistent across loans that required low/no documentation and those that required high documentation. Unreported second liens occurred on 13.4% of loans reporting no second lien, and these loans were 97% more likely to default. Two-thirds of unreported second liens were issued by the same originator as the first-lien loan, and the interest rate on unreported second liens is similar to the rates on loans correctly reporting second liens, which indicates that originators were aware of the misrepresentation.
Occurrences of this misrepresentation are clustered at securitization thresholds, suggesting that misreporting was done with the intention to securitize the loans.
Misrepresentation of owner occupancy may have occurred on 7.7% of loans marked as owner occupied, and these loans were 8% more likely to default. Interest rates on these loans were lower than that on correctly reported second homes, suggesting the misrepresentation was initiated by buyers and undetected by originators.
Appraisal values were overstated on 44.9% of loans, and these loans were 34% more likely to default. Interest rates were slightly higher on these loans, suggesting that originators were aware of the misrepresentation. Refinances were 74% more likely to contain an overstated appraisal; originators may have been leading appraisers to target certain values.
An investigation by the Financial Crisis Inquiry Commission revealed that firms performing loan monitoring and due diligence on behalf of underwriters typically review only 3%–10% of pooled loans and that banks replace only a portion of detected trouble loans. A sample reviewed during a congressional inquiry showed that 28% of loans failed to meet credit and compliance guidelines. This research is useful for mortgage-backed security (MBS) investors, underwriters, and regulators and highlights the need for thorough due diligence and transparency in MBS pools.
How Did the Authors Conduct This Research?
Loan-level and home valuation information is from Lewtan’s ABSNet Loan and HomeVal datasets. House characteristic and transaction information is from DataQuick’s Assessor and History files. A matching algorithm links the data from these sources to provide a detailed sample of 3,143,755 first-lien loans originated between 2002 and 2007.
Unreported second liens are detected by comparing MBS loan-level data showing only a first lien and county-level data showing a second lien originated on the same day as the first-lien loan. Misreported owner occupancy is detected by matching MBS loan-level data indicating that the property is owner occupied with county-level data showing that the tax records are sent to a different, nonbusiness address. Appraisal overstatements are detected by comparing the stated appraisal value on loans with Lewtan’s proprietary automated valuation model (AVM). Appraisals 5% higher than the AVM value are generally considered overstated, but the detailed regressions use a threshold of 20% above the AVM value to conservatively allow for estimation error of the model.
A logit regression is estimated using a loan delinquency dummy as the dependent variable and the misreporting indicators as the independent variables. This regression controls for determinants of loan performance identified by previous researchers, as well as mortgage complexity, original interest rates, reported second liens, and occupancy status. All continuous variables are standardized. Core-based statistical area is used to cluster standard errors as well as to ensure that results are not explained by geographical home price movements.
By focusing on only three types of misreporting indicators, the authors may conservatively reflect the lower bound of misrepresentations that actually took place leading up to the financial crisis. Falsification of borrower financial information seems to be an obvious additional misrepresentation that may have been initiated by either the buyer or the originator.
The most prevalent misrepresentation is unreported second liens. It could be that originators attempted to rationalize this misrepresentation through protocol that the second-lien loan not be issued until immediately following issuance of the first-lien loan, thus making the claim of the first-lien loan technically accurate at issuance, even if only for a brief while.