Expert's Regression Model Struck as Unreliable: The Pitfalls of Unsupported Assumptions
Expert's Regression Model Struck as Unreliable: The Pitfalls of Unsupported Assumptions
As expert testimony becomes prevalent in litigation, expert witnesses must scrutinize the opinions being offered to the court in relation to various standards and rules of evidence including the Daubert standard and Rule 702 of the Federal Rules of Evidence. It is becoming more and more apparent that whenever an expert is hired to provide expert testimony, the opposing party’s attorney(s) will file a motion to exclude that testimony.
The Daubert standard is used by a trial judge to make a preliminary assessment of whether an expert’s testimony is based on reasoning or methodology that is scientifically valid and can properly be applied to the facts at issue. Under this standard, the factors that may be considered are:
1. whether the theory or technique in question can be and has been tested
2. whether the theory has been subjected to peer review and publication
3. the theory’s known or potential rate of error
4. the existence and maintenance of standards controlling the theory’s operation
5. whether the theory has attracted widespread acceptance within a relevant scientific community
Experts should also scrutinize the opinions being offered to the court in relation to the Federal Rules of Evidence, which allows “a witness who is qualified as an expert by knowledge, skill, experience, training, or education [to] testify in the form of an opinion or otherwise if:
1.the expert’s scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue
2. the testimony is based on sufficient facts or data
3. the testimony is the product of reliable principles and methods
4. the expert has reliably applied the principles and methods to the facts of the case”
One of the greatest pitfalls that an expert can succumb to is the reliance on assumptions without gaining an understanding or performing any independent analyses to support the assumption being made. The Court’s conclusion that Plaintiff’s expert’s testimony should have been excluded under Daubert in United States of America ex rel. Kurt Bunk and Daniel Heuser v. Birkart Globistics GmbH & Co., et al. (“Birkart Litigation”) illustrates this very pitfall.
The Birkart Litigation was remanded by the Fourth Circuit and tried before a jury in the United States District Court, E.D. Virginia, Alexandria Division. On August 1, 2014, the jury returned a verdict against various defendants. Those defendants filed a Renewed Motion for Judgment as a Matter of Law and Alternative Motion for New Trial.
In considering the defendants’ motion, the court also considered whether the evidence was sufficient to sustain the jury’s award of damages, which resulted in the assessment of the plaintiffs’ damages expert. The Court ultimately concluded that the damages expert’s testimony was not reliable and should have been excluded under Daubert. Since the plaintiffs did not produce any other evidence of damages at trial, the court granted the defendants’ motion and set aside the damage award.
The following section will lay a foundation of the Birkart Litigation by discussing the factual background of the case as it was memorialized by the court in the Memorandum Opinion dated December 24, 2014.
The U.S. enters into contracts with American freight carriers (“AFCs”) to transport the household goods of U.S. servicemen and women, in this particular instance, to Germany, which was known as the ITGBL program. The AFCs enter into contracts with local agents to take possession of the goods once they reach Europe and transport them to the final destination within Germany.
One of the local agents in Europe, Gosselin, provided services to the AFCs in relation to the ITGBL program. During November 2000, the cost of shipping began to drastically drop and, in response, Gosselin and other local agents entered into an agreement to adhere to a bundled rate for all services related to the transport of goods (the “Bundled Agreement”). Gosselin became the primary local agent who contracted with the AFCs to transport the goods to the final destination within Germany.
Realtors Kurt Bunk and Ray Ammons originally filed the consolidated actions in 2002 and the U.S. intervened as to all claims relating to the ITGBL program in 2008. The U.S. alleged that the Bundled Agreement constituted fraudulent conduct that inflated rates it paid to the AFCs under every ITGBL program contract that was awarded from 2001 to 2002 (the “Alleged Damage Period”). The U.S. claimed that the Bundled Agreement was a scheme to eliminate competition from the bidding process to inflate shipping prices.
The U.S. hired an economist, Dr. Robert Marshall, to present evidence to the jury as to the damages suffered by the U.S. as a result of the alleged actions of Gosselin and the other defendants. Dr. Marshall developed a multiple regression analysis model that purported to predict the prime rates3 during the Alleged Damage Period. His multiple regression analysis considered 18 “explanatory variables” which were reduced to only those variables that positively correlated with the prime rates during the period 1979 through 2000 and 2005 through 2007 (the “Benchmark Period”).
Once Dr. Marshall predicted the prime rates for the Alleged Damage Period using the multiple regression analysis, he employed a “but for” model to calculate the damages. He calculated the difference between the “but for” and actual prime rates and multiplied that difference by the amount of tonnage shipped during the Alleged Damage Period. Dr. Marshall opined that the U.S. paid approximately $41.5 million more than it should have during the Alleged Damage Period.
The following section will discuss the court’s analysis supporting its decision to exclude the expert’s testimony under Daubert as it was memorialized in the Memorandum Opinion dated December 24, 2014.
The court did not question Dr. Marshall’s level of expertise or the decision to develop a damages model based on the well-recognized and accepted underlying principles of multiple regression analysis. However, the court did question whether or not a multiple regression analysis can reliably predict the prime rates that would have been set under the opaque, unusual, and complex price setting mechanism for the ITGBL program.
Specifically, the court questioned whether Dr. Marshall’s multiple regression analysis accounted for all of the major factors that affected the setting of the prime rates during the Alleged Damage Period. Dr. Marshall’s model was built on the key assumption that AFCs acted rationally and that they would not submit a bid for transporting goods unless it was compensatory in the sense that it maximized their profits.
During trial, evidence was presented that established that AFCs would implement various strategies in submitting their bids related to the ITGBL program. Some AFCs would submit bids that reflected the true cost to transport the goods while others would submit bids that were artificially inflated for the purpose of being recognized as a bidder in the ITGBL program but not expecting or wanting to win the contract.
The evidence established that during Dr. Marshall’s Benchmark Period, AFCs submitted bids that set prime rates which were considered to be “non-compensatory” by other AFCs in the industry, meaning the prime rate did not cover all internal costs to transport the goods. As a result, other AFCs in the industry began decreasing their bids to stay competitive in the ITGBL program.
The court determined that a single carrier had the ability to set a prime rate that did not necessarily have any particular relationship to its internal costs or the cost and demand factors used by Dr. Marshall in his multiple regression analysis. Accordingly, the court found that Dr. Marshall’s multiple regression analysis was not reliable in predicting the “but for” prime rates during the Alleged Damage Period.
In addition, the court made the following observations:
1.It was not at all clear whether Dr. Marshall used the most appropriate data to find the explanatory variables that best correlated with prime rates during the Benchmark Period.
2. Dr. Marshall’s model did not adequately account for ocean shipping rates, which account for approximately 30% of the AFCs costs.
3. Dr. Marshall failed to explain the inconsistencies between the actual prime rates set during the Benchmark Period and the prime rates predicted using his multiple regression model for the same period.
The court ruled that it was left with a firm conviction that Dr. Marshall’s model was not reliably predictive of the prime rates within the alleged damage period and should have been excluded under Daubert.
Experts routinely use regression analysis, a statistical technique to develop an equation depicting the relation among variables, to estimate the incremental cost of “but for” sales and, in the litigation discussed above, to estimate the “but for” prime rates during the Alleged Damage Period. Regression analysis is a generally accepted methodology in damage calculations, but when used improperly it can produce “a biased and misleading estimate.”
As discussed previously, the court did not question Dr. Marshall’s decision to use a multiple regression analysis approach but did question Dr. Marshall’s application of the multiple regression analysis and, specifically, the factors that Dr. Marshall considered to be relevant to the prediction of the prime rates. The court determined that Dr. Marshall’s assumption that the AFCs would only submit “compensatory” bids that would “maximize their profits” was an unsupported, integral assumption that resulted in the multiple regression model being unreliable.
Not only was the assumption unsupported, but the defendants presented evidence at trial that prior to the Alleged Damage Period, AFCs were submitting bids that others in the industry considered to be “non-compensatory.” Thus, decreasing the prime rates below what the industry considered to be the norm.
In a commercial litigation engagement a financial expert using a regression analysis in his/her damages model should consider performing:
1. An analysis to determine all the major factors that affect the prediction.
2. Economic and industry research to gain an understanding of the general economy and industry performance before, during, and after the alleged breach.
3. Interviews of management to gain an understanding of the industry and any particular factors affecting the industry, such as the existence of “non-compensatory” bids.
4. Interviews of other knowledgeable individuals to gain an understanding of the industry’s trends and future outlook.
If an expert uses a regression analysis and decides to make an integral assumption and/or exclude a major factor affecting the prediction model, the expert should be able to clearly explain the reasons such decisions were made. In addition, an expert should consider performing a sensitivity analysis to determine the outcome of the analysis if evidence is presented to rebut the integral assumption.
Because Dr. Marshall was not able to explain the reasons behind his integral assumption and exclusion of major factors affecting the prediction model and did not perform any sensitivity analyses relating to his assumption and exclusion of major factors, the court found the multiple regression model to be unreliable in the prediction of the prime rates and excluded Dr. Marshall’s testimony under Daubert. Accordingly, the court granted the defendants’ motion for judgment as a matter of law as to damages and set aside the damage award because the U.S. did not produce any other evidence at trial relating to damages.
It is imperative to note that this court’s decision to exclude the expert’s testimony does not provide a clear-cut standard prohibiting experts from using a regression analysis in calculating damages. However, this court’s decision is a reminder that an expert should consider performing independent analyses to gain an understanding and support all assumptions that are integral to the damages model.
Generally, an expert’s responsibility is to assemble sufficient, competent evidential matter that supports and/or verifies the most significant assumptions that affect the expert’s damages model. Experts achieve this by being diligent in understanding the facts of the case, a company’s financial performance before, during, and after the alleged wrongdoing, the industry as a whole, and the impact of decisions made by company management.
Experts could do well to remember the court’s conclusion in the Daubert case: “reliability and relevance are the touchstones of the admissibility of expert testimony.”