Rounding Error and Information Quality:
The case of Exxon Shipping Co. v. Baker
30 Jun 2008 in Information Quality
Last week the Supreme Court reversed an appellate court opinion that would have imposed $2.5 billion in punitive damages resulting from the 1989 Exxon Valdez oil spill. In the majority opinion written by Justice David Souter, the Court opined on a matter of maritime law for which there was neither a constitutional precedent nor operating law. The Court ruled that a 1:1 ratio of punitive to compensatory damages "is a fair upper limit in such maritime cases."
The Court obtained this ratio by conducting an ad hoc qualitative statistical analysis of trial court practice, which yielded a ratio of 0.65:1, then rounding up to 1:1. It is instructive to note the practical financial significance of the arcane information quality issue of significant figures (see here and here).
The Court's reasoning is summarized as follows:
There is better evidence of an accepted limit of reasonable civil penalty, however, in several studies mentioned before, showing the median ratio of punitive to compensatory verdicts, reflecting what juries and judges have considered reasonable across many hundreds of punitive awards. We think it is fair to assume that the greater share of the verdicts studied in these comprehensive collections reflect reasonable judgments about the economic penalties appropriate in their particular cases. These studies cover cases of the most as well as the least blameworthy conduct triggering punitive liability, from malice and avarice, down to recklessness, and even gross negligence in some jurisdictions. The data put the median ratio for the entire gamut of circumstances at less than 1:1, meaning that the compensatory award exceeds the punitive award in most cases. In a well-functioning system, we would expect that awards at the median or lower would roughly express jurors’ sense of reasonable penalties in cases with no earmarks of exceptional blameworthiness within the punishable spectrum (cases like this one, without intentional or malicious conduct, and without behavior driven primarily by desire for gain, for example) and cases (again like this one) without the modest economic harm or odds of detection that have opened the door to higher awards. It also seems fair to suppose that most of the unpredictable outlier cases that call the fairness of the system into question are above the median; in theory a factfinder’s deliberation could go awry to produce a very low ratio, but we have no basis to assume that such a case would be more than a sport, and the cases with serious constitutional issues coming to us have naturally been on the high side (ratio of 500:1). On these assumptions, a median ratio of punitive to compensatory damages of about 0.65:1 probably marks the line near which cases like this one largely should be grouped. Accordingly, given the need to protect against the possibility (and the disruptive cost to the legal system) of awards that are unpredictable and unnecessary, either for deterrence or for measured retribution, we consider that a 1:1 ratio, which is above the median award, is a fair upper limit in such maritime cases (Exxon Shipping Co. v. Baker at 40, internal references and citations omitted).
The Court did not reduce punitive damages to $507.5 million, as many news reports have incorrectly stated, but rather implicitly capped punitive damages at this amount assuming that the district court's calculation of compensatory damages is correct. The actual punitive damage award will be determined by lower courts, taking into account the Supreme Court's reasoning in this case and its description of the defendant's conduct as "without intentional or malicious conduct, and without behavior driven primarily by desire for gain." and the actions of these courts remains subject to review and challenge. (The jury's punitive damage award was $5 billion, which the Ninth Circuit reduced to $2.5 billion on the ground that $5 billion was excessive. It is this amount that Exxon Shipping challenged.)Nonetheless, the Court's reported ratios of punitive to compensatory damages, and the effects of different rules for rounding, are financially spectacular. The table below summarizes the amount of punitive damages using a variety of rules for rounding, all of which are arbitrary.
There is an interesting coda to the story. The Court determined that 0.65:1 was the median ratio of punitive damage awards -- a well-defined statistical term meaning that exactly half of all cases are above and half are below -- but it actually rounded up from the 0.62 reported in the study cited by the Court. Normally, 0.62 would be rounded down to 0.6 rather than rounded up to 0.7. Presumably, it is harder to justify rounding 0.6 up to 1.0 when hundreds of millions of dollars are in the balance.
Bill Henderson blogs on empirical legal studies alongside Theodore Eisenberg, who was the lead aithor of the study. Henderson notes approvingly concerns raised by Election Law blogger Rick Hasen, who highlights the perverse incentive against research implied by the Court's exclusion of data from mock jury studies. The Court excluded them because they were plaintiff-funded, not because they were methodologically defective. In short, the Supreme Court has committed a frank information quality error.
Of course, the Information Quality Act applies only to the Executive branch. Congress and the courts are exempt. One can only speculate what would happen if they were required to adhere to the law and its implementing guidelines.
| ALTERNATIVE ROUNDING RULES FOR ASSESSING PUNITIVE DAMAGES |
||
| Punitive:Compensatory Damage Ratio |
Description of Rounding Rule | Maximum Punitive Damages, $ Millions |
| 9.85:1 | Actual jury award in this case |
$5,000 |
| 5:1 | Missouri statute |
$2,538 |
| 4.93:1 | 9th Circuit in this case |
$2,500 |
| 3:1 |
|
$1,523 |
| 2:1 |
|
$1,015 |
| 1:1 |
|
$508 |
| 0.7:1 | 1 significant figure | $355 |
| 0.65:1 | 2 significant figures Reported median |
$330 |
| 0.62:1 | Median from cited study | $315 |
T. Eisenberg et al., Juries, Judges, and Punitive Damages: Empirical Analyses Using the Civil Justice Survey of State Courts 1992, 1996, and 2001 Data, 3 J. of Empirical Legal Studies 263, 278 (2006). A free working paper version is here.
