MANHATTAN (LN) — A federal judge on Thursday cleared all six expert witnesses to testify trial of S. Kenneth Leech, the former chief investment officer of Western Asset Management Company accused of running a Treasury futures cherry-picking scheme that routed profitable trades to favored accounts while allocating losers to disfavored clients.
U.S. District Judge Gregory H. Woods denied both the government's and defense's Daubert motions in full, ruling that none of the proposed expert testimony — spanning competing statistical analyses, a blind allocation exercise, and dueling regression models — warranted exclusion under Federal Rules of Evidence 702, 403, or 704.
The government's November 2024 indictment charges Leech with investment adviser fraud, securities fraud, commodity trading advisor fraud, commodities fraud, and making false statements to the Securities and Exchange Commission. The government alleges that from January 2021 through October 2023, Leech waited to see how Treasury futures and options trades performed before allocating winners to his Macro Opportunities strategy and losers to the Core Strategies — and then lied to the SEC about it.
The government's experts, Harvard Business School professor Lauren Cohen and NYU Courant Institute professor Petter Kolm, propose to testify that trades Leech allocated to Macro Opps had an average first-day return of $0.24 million, while trades allocated to the Core Strategies had an average first-day return of negative $0.31 million, and that permutation testing ruled out random chance as an explanation.
The defense countered with a blind allocation exercise designed by Charles River Associates vice president Aaron Dolgoff and conducted by NYU Stern professor Bruce Tuckman, who allocated 935 trades across 50 randomly selected trading days without knowing Leech's actual allocation decisions or first-day performance results. Dolgoff proposes to testify that Tuckman's allocations matched Leech's at a statistically significant rate of 56.8 percent — and that Tuckman's Macro Opps allocations nonetheless generated statistically higher first-day performance than the overall trade pool, suggesting portfolio management needs rather than cherry-picking could explain the pattern.
Woods rejected the government's argument that the CRA Exercise should be excluded because it failed the four Daubert reliability factors, finding that those factors were designed for scientific testimony and that the exercise reflected sufficient intellectual rigor given the professional expertise of its designers. He also rejected the defense's bid to exclude Cohen and Kolm on qualification grounds, finding that Cohen's academic research on asset manager performance and Kolm's private-sector experience designing fixed-income trading strategies placed their proposed testimony within their areas of expertise.
On the most contested legal question — whether the government's experts would impermissibly opine on Leech's mental state in violation of Federal Rule of Evidence 704(b) — Woods drew a line between testimony that evidence is consistent with a defendant using information to make decisions, which he found permissible, and testimony that expressly states the inference of criminal intent, which is not. He distinguished the case from United States v. Hwang, where Judge Hellerstein sustained objections to testimony that trading was consistent with the intent to minimize price impact, noting that Cohen and Kolm do not propose to use the word intent or its derivatives.
Woods nonetheless flagged language in both sides' disclosures as potential trial hazards. He ordered the parties to meet and confer within two weeks on mutually agreeable phrasing — specifically targeting the government experts' use of using in connection with allocation decisions, and the defense's use of proxy to describe Tuckman's duration and convexity targets, which the judge said risked implying Tuckman was attempting to stand in Leech's shoes.
The trial trade blotter covers more than 33,000 Treasury futures and options trades spanning 595,236 rows and 48 columns — nearly 28.6 million cells of data — that the jury will be asked to evaluate with the help of all six experts.