Frequency analysis in settlement

Many settlement conferences take a long time to get off the ground. They start with unrealistic posturing on both sides. Very low and very high numbers are thrown out with little hard analysis. The point is often precisely to posture, to send signals of toughness, to set expectations. Perhaps it’s done because counsel has heard about anchoring, the cognitive fallacy that causes people to make gut estimates based on irrelevant exposure to high and low numbers. (When people are first asked to recall the last few digits of their social security number before bidding, those with high numbers (897) will bid higher than those with low numbers (003).)

Getting past these posturing numbers and into the merits of the case takes time. Estimating the worth of a case is hard even when parties are trying to be analytical, because the analysis usually stops after just a couple of steps - the plaintiff claims $100, maybe she has a 50% chance of winning, so $50 seems right, discounted by something for settlement purposes.

But when there are multiple causes of action each with its own calculation of odds, or a spectrum of reasonable possible damages such as with pain and suffering, punitive damages, statutory penalties, or a series of items (like future medical bills) which the jury may or may not endorse, calculations rapidly disintegrate to gut reactions. We just estimate the estimates. But those won’t persuade the other side.

Too many settlement talks go off the rails when a party, confronted with a weakness on one issue, retorts that it has strength on some other issue. It’s a time-tested tactic – “my client might be a felon but you don’t even have an expert” – but I call it legerdemain.

Problems with estimating settlement value arise in other contexts, such as when lawyers try to convince a judge to approve a PAGA or class action settlement; that a settlement is within the ballpark of reasonableness in a good faith settlement motion (CCP § 877.6); and sometimes when a judge is thinking about the validity of a CCP § 998 offer. I have often seen in e.g. class action settlements vague discussion of the merits, apparently baseless estimates of liability, and conclusory suggestions of the fairness of a proposed dollar amount.

So I offer a schema designed to enable principled discussions of proposed settlements in all these situations: frequency analysis.

Most of us already do something like this. The simplest form is e.g., a 20% chance of a $100 settlement, generating a settlement value of $20. But most cases are more complex. There is e.g. a 20% chance of a defense verdict, a 40% chance of a $100 verdict, perhaps 30% chance of a $80 verdict, and so on. These sums are then added up to generate a settlement suggestion. For example, in a case where there is no substantial chance of a defense verdict and a small chance of a complete plaintiff victory (all punitives, maximum reasonable pain and suffering etc.), we might have this estimate:

5% odds of $100 = 5 (all punitives, all pain and suffering)

10% odds of $80 = 8 (no punitives)

50% odds of $50 = 25 (no punitives, moderate pain and suffering)

10% odds of $40 = 4 (low pain and suffering)

15% odds of $10 = 1.5 (compromise verdict with token award)

10% odds of $0 = 0 (defense verdict)

Total estimate settlement value = $43.5

(Of course this estimate doesn’t account for other important factors, such as the consequences of further litigation including delay in payment (or, for defendant, the benefits of delay – assuming no prejudgment interest), costs of litigation, the toll on the parties, insurance coverage issues, business interruption, publicity, precedent setting, and so on.)

Anyone can make up such a chart. I did. The point, really, is the next step: the analysis organizes discussion of what went into the numbers. In class action settlements of multiple wage and hour claims, we’d start with an estimate of the odds of certification, which might comprise a series of estimates that the court would find individual issues, manageability problems, and so on. Then, there might be a separate calculation for each claim. Each claim would be associated with a maximum reasonable recovery, and odds of that; as well as odds of smaller recoveries. In a personal injury case there might be odds associated with liability, odds that the causation expert will be allowed to testify and whether the jury will believe her, and various major elements of damages. For each line item, the parties and mediator can focus their discussion on the likelihood of a variety of outcomes: if plaintiff says there’s a 90% chance the jury will find future brain surgery (at $500,000) will be needed, and defendant disagrees, the parties can point to the evidence on their estimates of those odds.

Important factors can be subject of further detail, e.g., a 50% chance of a $500,000 surgery, 40% chance of a $400,000 surgery, 10% chance no surgery will be found to be needed; thus valued at $410,000.

This approach encourages the identification of decisive issues such as the admissibility of expert opinion or the assumption the jury will believe a certain witness. And it sums the odds of the many factors that feed an accurate estimate of settlement value.

Frequency analysis also gives the mediator a variety of ways to tweak the numbers to suggest to the parties that their gut estimates of the value of the case – or their apparently random starting offers and demands – are unsupported. Undertaking frequency analysis before a settlement conference will focus the parties on the evidence and the decisive issues, instead of their instinctual desires, before they have staked out their public claim. For some, it may be a revelation.

There’s another advantage. Many mediators (including me) try to get the very high and very low estimates off the table as soon as we can, in order to focus on what we think of as a reasonable intermediate range. But parties may feel shortchanged, upset that we are not accounting for their estimates which (while optimistic) are nevertheless actually possible. Frequency analysis literally does account for the high and low estimates.

I do have one caution. Frequency analysis assumes each factor is an independent variable; it does not handle reinforcement among factors. For example, some lawyers believe that if the jury likes the plaintiff, that will carry over to believing the plaintiff’s expert and will also increase damages.

Settlement is always about estimates – and guesstimates. Intangibles can make a difference at trial. So there is an art to settlement, not just number counting. Settlement of essentially the same case between different parties will generate different numbers. Many cases settle for reasons only distantly related to the merits. In these instances, frequency analysis won’t be effective.

But in other cases, and when explaining to a judge why a settlement reflects the merits, frequency analysis may provide a useful framework. The parties will be starting with a thoughtful analysis of numbers, not an untethered instinct.