Variability in Punitive Damages: An Empirical Assessment of the U.S. Supreme Court’s Decision in Exxon Shipping Co. v. Baker
24 Pages Posted: 26 Apr 2009 Last revised: 26 Apr 2014
Date Written: April 20, 2009
Exxon Shipping Co. v. Baker acknowledged what virtually all methodologically sound punitive damages research shows. The Supreme Court relied in part on an article by the present authors and others to state that empirical studies undercut the most audible criticism of punitive damages and that no mass of runaway punitive awards existed. Paradoxically, the Court simultaneously expressed concern about jury predictability based on a high mean and standard deviation in the punitive-compensatory ratio published in our article. The Court therefore reduced a $2.5 billion punitive award relating to the Exxon Valdez oil spill to $500 million to implement a 1:1 punitive-compensatory ratio and stated that “the constitutional outer limit may well be 1:1.” This article shows that our empirical findings relied on by the Court do not support the unpredictability concern or widely applying the limiting ratio. The high mean and standard deviation are artifacts of not accounting for the key variable that explains punitive awards - the compensatory award. Stratifying the mean and standard deviation of the punitive-compensatory ratio by the level of the compensatory award shows that the ratio is reasonably stable in high award cases and significantly and explicably more variable in low award cases. Basing doctrine on summary statistics that combine these heterogenous distributions is not statistically supportable. The award reduction in Exxon Shipping may have promoted consistency with other high compensatory award cases but the 1:1 principle the case hints at is not statistically supportable across the broad range of compensatory awards, and could contribute to an inability to tailor punitive awards to the facts and circumstances of particular cases.
Keywords: Punitive Damages, Tort, Empirical Studies
JEL Classification: K00, K13, K32, K41
Suggested Citation: Suggested Citation