Lyondell Chem. Company v. Occidental Chem., Company

United States Court of Appeals for the Fifth Circuit · 2010 · Evidence

608 F.3d 284 (5th Cir. 2010)

Updated August 22, 2026

EvidenceRule 408settlement communicationsCERCLAcontributionDaubertMonte Carlo analysisexpert testimony

Facts

French Limited hauled hazardous waste intended for its primary Highway 90 disposal site and sometimes dumped overflow waste at Turtle Bayou. In allocating CERCLA contribution costs, the district court treated the remediation areas as an indivisible whole, appointed Dr. Charles Newell to estimate waste volumes using Monte Carlo analysis, and used three inputs for Occidental's waste volume: a minimum based on driver testimony, an intermediate estimate based on the Smythe Reports, and a maximum based on the French Ledger. The Smythe Reports had been created by Occidental during earlier negotiations among French Limited customers to allocate potential liability for contamination at Highway 90 after the EPA threatened CERCLA litigation. The district court admitted those reports and ultimately allocated 15.96% of El Paso's response costs to Occidental.

Issue

Whether the district court properly admitted Dr. Newell's Monte Carlo-based expert testimony under Daubert, properly admitted the Smythe Reports despite Rule 408, and permissibly chose and applied its CERCLA allocation methodology and factual findings. The cross-appeal also asked whether the district court erred by placing Lyondell's uncollectible share on El Paso rather than distributing it among all liable parties.

Rule

Monte Carlo statistical analysis is not inherently unreliable under Daubert when used with transparent inputs and to estimate uncertain quantities within a known range. Rule 408 excludes conduct or statements made in compromise negotiations regarding a disputed claim, and the term "claim" is applied fact-specifically in light of Rule 408's purposes rather than by a rigid definition; settlement evidence arising from a shared factual nexus and bearing directly on present liability issues may be inadmissible even if generated in negotiations over a different site or dispute. In CERCLA contribution actions, a court may use such equitable factors as it determines are appropriate, and factual allocation findings are reviewed for clear error with flexibility because meticulous causal precision is often impossible.

See the holding & full analysis

Create a free KwikCourt account to unlock the rest of this brief — and practice the case.

The court's holding and reasoning
Doctrine tests, pitfalls & exam hypotheticals
10 practice questions + 4 AI-graded essays on this case

Free sample · practice this case

Test yourself

One of 10 multiple-choice questions for this case. Pick an answer to see why.

In a federal contribution bench trial over chemical cleanup near Baton Rouge, the judge appoints an environmental engineer to estimate how much solvent Ridgeway Coatings sent to a backup dumping area. Exact records are missing, so the judge supplies a minimum based on one driver's testimony, a maximum based on shipping invoices, and a midpoint based on admissible plant logs, and the expert runs a Monte Carlo simulation to generate likely disposal volumes.

Ridgeway objects that the testimony must be excluded because Monte Carlo analysis produces only a probability distribution rather than one exact number. How should the court rule?

Explanation. The majority held that Monte Carlo analysis is not inherently unreliable under Daubert. It is useful where an exact result is infeasible but the data provide a known range, and reliability can be assessed by examining the inputs and requiring transparency. The fact that the analysis yields a statistical range rather than a single exact number does not make it irrelevant or speculative; it can still help the factfinder choose among possible outcomes. (Derived from Lyondell Chem. Company v. Occidental Chem., Company (2010).)