David Hyman, Bernard Black, Kathryn Zeiler, Charles Silver, and William Sage have a new paper with this title on SSRN. The abstract:
Legal scholars, legislators, policy advocates, and the news media frequently use jury verdicts to draw conclusions about the performance of the tort system. However actual payouts can differ greatly from verdicts. We report evidence on post-verdict payouts from the most comprehensive longitudinal study of matched jury verdicts and payouts. Using data on all insured medical malpractice claims in Texas from 1988-2003 in which the plaintiff received at least $25,000 (in 1988 dollars) following a jury trial, we find that most jury awards received �haircuts.� 75% of plaintiffs received a payout less than the adjusted verdict (jury verdict plus pre-judgment and post-judgment interest), 20% received the adjusted verdict (within + 2%), and 5% received more than the adjusted verdict.
Overall, plaintiffs received a mean (median) per-case haircut of 29% (19%), and an aggregate haircut of 56%, relative to the adjusted verdict. The larger the verdict, the more likely and larger the haircut. For cases with a positive adjusted verdict under $100,000, 47% of plaintiffs received a haircut, with a mean (median) per-case haircut of 8% (2%). For cases with an adjusted verdict larger than $2.5 million, 98% of plaintiffs received a haircut with a mean (median) per-case haircut of 56% (61%). Insurance policy limits are the most important factor explaining haircuts. Caps on damages in death cases and caps on punitive damages are also important, but defendants often paid substantially less than the post-cap adjusted allowed verdict. Remittitur accounts for a small percentage of the haircuts. Punitive damage awards have only a small effect on payouts. Out-of-pocket payments by physicians are rare, never large, and usually unrelated to punitive damage awards.
Most cases settle, presumably in the shadow of the outcome if the case were to be tried. That outcome is not the jury award, but the actual post-verdict payout. Because defendants rarely pay what juries award, jury verdicts alone do not provide a sufficient basis for claims about the performance of the tort system.
An interesting follow-up would be for someone to take a qualitative look at the cases where the jury verdict was above policy limits that settled just below policy limits or higher. Does the threat of bad-faith insurance litigation force insurers to forgo meritorious appeals? Unfortunately, the underlying data isn't detailed enough to determine when settlements reflect ex ante high-low treatment.
The paper argues that damages caps are a redundant restraint on settlements, because it is rare for caps to be exceeded within policy limits. This suggests a strange irrationality amongst insurers for supporting caps; then again, medical lobbyists have been seen to argue that caps above $250,000 are largely pointless, which is consistent with this paper's findings, which, because of the time frame analyzed, is limited to the less restrictive caps on death and punitive damages.
I was surprised that the paper found no effect on settlement amounts over time; this would seem to contradict economic theory. The authors acknowledge bad-faith-insurance risk (insurers pay above policy limits about 40% of the time that there is a verdict above policy limits, and pay more than double the policy limits about 10% of the time), so policy limits are not an absolute constraint on settlement amounts; if I'm reading the regression correctly, the authors' model predicts that a $1 increase in award over policy limits results in an additional 16-cent payment over policy limits. Therefore, increasing variance in verdicts over time should have an effect on settlement amounts, at least in alleged-severe-injury cases. Alas, the authors don't provide a hypothesis explaining their results that is consistent with their acknowledgment of a Stowers effect in Texas. Without that explanation, I'm inclined to think the issue is one of a data artifact, but I'm wary to simply announce that given the possibility of confirmation bias. Anyone have ideas?
I do wonder if the time-value effect is muted by a tendency over time for those purchasing insurance to purchase at lower policy limits, which would certainly be a rational response by doctors to the paper's finding that policy limits do operate as a significant partial constraint. That data should be available, and a forthcoming paper by the same authors focusing on policy limits may address that question.
Too, as the authors argue, one possible effect of increasing variance in verdicts would be increasing trials rather than increasing settlement amounts; the paper found the proportion of cases going to trial did not change over time. But here there are confounding variables; one would also expect the number of trials to go down over time because of increased attorney expenses. The absence of defense verdicts with settlements under $25,000 and self-insured claims in the dataset also introduces noise that prevents a determination of the factors in the decision to go to trial as well as making it impossible to say absolutely what the direction is in the number of trials.
Finally, the dataset is so small that a change in the mix of cases of just one or two cases a year would have profound effects on the regression. What is the trend in birth injury cases, the cases most sensitive to both the effects of reform and the hypothetical settlement ratchet effect? This is not measured, and the dataset is probably too small to be cut that finely.
But if the results on this issue are repeated in future larger studies, reformers and economists will need to rethink the theory of the effect of large verdicts, and doing so has uncertain implications for optimal public policy prescriptions.