Massachusetts Mutual Life Insurance v. Residential Funding Co.

Citations: 989 F. Supp. 2d 165; 2013 WL 6490125; 2013 U.S. Dist. LEXIS 172697Docket: Civil Action No. 11-30035-PBS

Court: District Court, D. Massachusetts; December 8, 2013; Federal District Court

SummaryOriginalCytator Case TreatmentBackward CitatorAnalysis

EnglishEspañolSimplified EnglishEspañol Fácil

Plaintiff Massachusetts Mutual Life Insurance Company (MassMutual) has initiated eleven lawsuits against various corporate and individual defendants, alleging violations of the Massachusetts Uniform Securities Act (MUSA) related to its investment in residential mortgage-backed securities (RMBSs). MassMutual purchased 121 securitized certificates totaling approximately $2 billion, linked to 95 securitizations and 99 Supporting Loan Groups (SLGs), which encompass 278,609 individual residential loans. The plaintiff claims these certificates contained material misrepresentations about the riskiness of the loans.

To assess these claims, MassMutual plans to "reunderwrite" the loans by reviewing original loan files to ensure they were originated according to proper standards. Given the extensive number of loans, MassMutual intends to analyze a sample of 100 loans from each SLG, resulting in the reunderwriting of 9,900 loan files. An expert report by Dr. Charles D. Cowan outlines the statistical sampling method to be used for this analysis, aiming to extrapolate a probable rate of misrepresentation across the entire SLG.

Defendants filed a motion to exclude Dr. Cowan’s testimony under Federal Rule of Evidence 702 and Daubert standards, which was denied following an evidentiary hearing. The amended complaints assert that the defendants made material misrepresentations in selling the securities, and under MUSA, liability can be established without demonstrating negligence, reliance, or loss causation. Additionally, in seven cases, MassMutual pursues "control person" liability against individual defendants, alleging they misrepresented the underwriting standards and appraisal processes used in the loans, which were purportedly riskier than disclosed.

Plaintiff alleges that defendants, including HSBC, reported false loan-to-value ratios and inaccurate owner-occupancy rates for properties. Defendants deny making any material misrepresentations during the marketing and sale of certificates. 

Expert witness Dr. Charles D. Cowan, who holds degrees in Economics and Mathematical Statistics, is the Managing Partner at Analytic Focus LLC and an adjunct professor of biostatistics. He has extensive experience in statistical analysis and has held prominent positions in various governmental and educational institutions. His credentials are undisputed by the defendants.

Dr. Cowan's expert report outlines his statistical sampling plan to analyze the loans underlying each securitization. He intends to select a sample of 100 loans from the pool, asserting that this sample size will yield valid conclusions for the entire population, with a margin of error of ±10 percentage points. Although a larger sample size could reduce the margin of error, he believes that 100 loans effectively balances cost and accuracy.

To enhance representativeness, Dr. Cowan will employ stratification, dividing the loan pool into subgroups based on FICO scores, which measure borrower creditworthiness. He will categorize loans into quartiles (high, somewhat high, somewhat low, and low FICO scores) and randomly select loans within these strata. The initial sample will consist of the first 25 loans from each stratum, with an additional 25 as a backup. This methodology aims to improve reliability while maintaining statistical rigor.

The sample from each SLG will consist of 25 loans across four strata, totaling 100 primary loans and 100 back-up loans. If a primary loan file is unavailable, a back-up loan from the same stratum will replace it. Dr. Cowan will evaluate both samples against the full loan population on eleven variables to confirm representativeness. The re-underwriting process, conducted by third-party servicers rather than Dr. Cowan, aims to identify material misrepresentations regarding loan origination compliance, appraisal standards, loan-to-value ratios, and, for HSBC, the percentage of loans tied to owner-occupied properties. After re-underwriting, Dr. Cowan will extrapolate the findings to the entire SLG population, utilizing a method that minimizes margin of error, though specific extrapolation techniques will not be predetermined. For example, if 50% of loans show misrepresentations, this could suggest with 95% confidence that between 40% and 60% of all loans similarly contain misrepresentations.

The section also introduces Dr. Arnold Barnett, a qualified expert in applied statistical analysis from MIT, who supports challenges against Dr. Cowan’s methodology through a declaration and testimony during the Daubert hearing. The discussion emphasizes the court's gatekeeping role in admitting expert evidence under Federal Rule of Evidence 702, requiring that expert testimony is based on sufficient data, reliable methods, and appropriate application of those methods to the case facts.

The trial court is tasked with assessing the reliability and relevance of an expert’s testimony, as well as the expert's qualifications. This assessment is guided by the Daubert standard, which emphasizes the expert's methodology as the primary focus, incorporating four key factors: (1) the ability to test the theory or technique, (2) peer review and publication status, (3) known or potential error rates, and (4) acceptance within the relevant discipline. While these factors are crucial, they are not exhaustive, allowing judges flexibility to consider additional factors. The court is reminded that vigorous cross-examination and the jury’s role in evaluating conflicting expert viewpoints are essential. 

In this case, the Plaintiff requested an early assessment of the reliability of Dr. Cowan's sampling methodology, which defendants argue is premature before application to reunderwriting results. A magistrate judge agreed to this early determination to streamline case management and avoid lengthy trials over data accuracy and methodology, recognizing the potential for significant expense and delay if the methodology fails a Daubert challenge later on.

Defendants do not contest Dr. Cowan's expertise but point out six methodological errors they believe undermine the reliability of his sampling protocol. A significant criticism is Dr. Cowan's lack of a specific extrapolation method, as he suggests determining this only after confirming sample design and testing results. Defendants argue this is essential for a valid scientific approach, though Dr. Cowan clarified that a specific extrapolation method is not necessary prior to reunderwriting to maintain valid sample design.

Dr. Barnett, an expert for the defendants, contends that Dr. Cowan is restricted in his choice of extrapolation methods due to his initial stratification of the population, which commits him to a specific formula related to proportional stratified sampling. Even if Dr. Cowan cannot reduce the margin of error below ±10 percentage points, his methodology remains valid. He plans to evaluate multiple extrapolation methods, potentially up to twenty, after obtaining reunderwriting results. The lack of a specified method in his report does not disqualify his sampling methodology in this preliminary assessment, provided he ultimately uses a reliable technique.

Defendants argue that Dr. Cowan’s methodology relies incorrectly on binary answers to complex misrepresentation inquiries, claiming that the inquiries related to compliance with underwriting and appraisal standards, as well as LTV and owner-occupancy rates, cannot be simplistically categorized as 'yes' or 'no.' They assert that underwriting guidelines allowed for flexibility and exceptions. While Dr. Barnett admitted that some questions may be binary, the defendants emphasize that LTV and CLTV ratios are continuous variables with multiple possible values. The plaintiff acknowledges the need for a materiality benchmark and indicates that the determination of average weighted LTV/CLTV will involve non-binary questions, yet maintains that once calculated, assessing whether these averages materially differ from figures in offering documents is a binary question.

The defendants' strongest argument focuses on the inability of Dr. Cowan's methodology to differentiate among various loan originators, particularly since 57 of the 99 SLGs involve loans from multiple lenders, each potentially adhering to distinct underwriting guidelines. They highlight that the prospectus for the ACE 2006-SL1 securitization only discloses criteria for two of the three originators, which could complicate the reliability of the analysis due to variations in practices among originators.

Deutsche Bank did not make any representations regarding the appraisal standards of Chapel Funding Corporation, thus cannot be held responsible for any misrepresentation related to it. The defendants contend that only loans from American Home and Residential Funding are relevant for sampling, leading the plaintiff to drop this securitization from liability claims. The core issue is not the liability of each originator for misrepresentations, but whether the defendants misrepresented underwriting standards and appraisal information in the certificates for all loans backing those certificates. The defendants assert that the proposed sampling methodology is inadequate, particularly since the offering documents differentiate among originators. Dr. Cowan acknowledged that sampling from a pool where representations apply to only a fraction of loans can increase the margin of error.

Defendants challenge Dr. Cowan’s ±10 percentage point margin of error as excessively wide compared to typical litigation standards, which usually observe a ±5 percentage point margin. They reference guidelines from HUD and Freddie Mac, which use a ±2 percentage point margin. They argue that a larger sample size would reduce the margin of error, as increasing sample size generally decreases random error. However, Dr. Barnett noted that a ±20 percentage point confidence interval does not inherently render a methodology unreliable. The margin of error impacts the sample's persuasive power rather than its admissibility.

Defendants dispute the use of FICO scores as a stratification variable, yet Dr. Cowan and defendants’ expert agree that stratification does not increase margin of error, but may enhance precision. Defendants fail to convincingly argue against using FICO scores, which indicate borrower creditworthiness. Additionally, Dr. Cowan’s method of assigning loans without a FICO score to the lowest quartile is based on his experience of such cases indicating underwriting breaches, though this assumption lacks clarity. Nonetheless, it is premature to challenge this methodology given the uncertainty regarding the number of missing FICO scores.

Defendants argue that the potential absence of certain loan files compromises the randomness and statistical reliability of the sample. Plaintiff counters that Dr. Cowan’s plan to implement a randomized 100-loan 'back-up' sample is a valid statistical approach to mitigate issues related to missing files. Previous court rulings have upheld similar methodologies, acknowledging that missing data is a common occurrence in statistical surveys. Dr. Cowan will consider the effects of absent files in his analysis, and defendants can contest his methodology later in the proceedings.

The court denied defendants' Motion to Exclude Dr. Cowan's opinions from his April 12, 2013 report without prejudice. MassMutual has ongoing claims against various individual and corporate defendants across multiple cases, with some claims in one case stayed due to bankruptcy proceedings. The text also references the role of credit scores in assessing borrower risk and notes that Dr. Cowan has experience with their use by lenders, citing the destruction of records by Hurricane Sandy as a specific instance of missing data.