From The ‘Lectric Law Library’s Stacks
Modeling Potential Liability Costs in Large Class-Action Cases by Robert D. Shriner, Ph.D.
Large class-action product liability cases pose special analytical problems for defendant firms and their insurers. The firm needs to realistically assess the likely costs of resolving claims presented by class participants plus suits and other claims filed separately by opt-outs, often at an early stage at which the number of claimants and the nature of their claims is still uncertain. The firm’s chief finance officer may need to allow for potential claims payments in the firm’s cash flow planning and, in major liability situations, also determine whether the probable costs are great enough to exceed the firm’s insurance coverage or to be financially material to the firm and thus subject to mandatory disclosure under SEC regulations. The firm’s legal counsel and insurers must have the best available simulations of the effects of various possible settlement strategies, offers, and counter-offers in order to negotiate advantageously and effectively.
We have expertise in developing specialized simulation and forecasting models to estimate the probable extent of liabilities and the likely costs of settling claims arising from major product liability situations. These liability estimation models often combine claims data, marketing and product use data, medical and epidemiological data, medical care cost data, actuarial data, and other information, including expert opinions, which may be relevant to the situation. In simplified form, here’s how we do it.
The first task is trying to estimate how many customers are affected and the likely impact on them. How many claims have been filed? How many settled (if any) and at what cost? Are the problems attributable to a single batch, model, or configuration, or to several? How many units were produced? What kinds of problems are being reported? Does it appear that there is a single injury/illness mode, or several? What is the relative frequency among different modes if there appear to be several? What medical measures, hospital days, lost time, or other impacts appear to be prompted by each injury mode? Are the consequences different for different individuals, depending on their behavior or personal characteristics (age, sex, genetics, environment, etc.)?
If there are settled claims, use that as the starting point since the substance of the claims will have been checked by adjustors. Determine the average settlement amount and the ratio between settlement amount and medical costs or other losses cited by claimants. If there are multiple injury modes, or levels of severity, determine the average settlement amounts and ratios for each, along with their relative frequency. Since initial settlements are likely to be those which are easiest to achieve, these averages, ratios, and frequency distributions provide the foundation for estimating the low end of the likely cost range. They can be multiplied by actual losses reported in filed but unsettled claims to make a tentative estimate of the minimum likely cost of settling those pending claims. If there is enough time-series information on claims, it may be possible to assess whether there is a measurable pattern of change over time in either claims or settlements. However, this is rarely determinable in the early stages of a case.
If no claims have been settled at the time the analysis is begun, it becomes necessary to estimate the ratio between actual losses reported and expected settlement costs based on experience (your own, the adjustors’, or the lawyers’) with similar types of claims in other cases. The use of such “guesstimates” is admittedly crude but often close to the mark if the sources are highly experienced. Where expert opinions differ, they may provide a possible basis for estimating high and low ranges.
Next, estimate the likely number and cost of future claims that might be expected, based on internal company information about the number of units affected or produced, where they were sold and when, numbers recalled (if any), and similar production and marketing data. It may also be necessary to include expert medical information or estimates for example, in a food contamination situation, estimates on the threshold inoculum level for known symptoms and complications; likely variations in dosage response by age, sex, medical history, or other personal factors; and medical treatments normally prescribed under those circumstances as well as socioeconomic and demographic information. Because of the sensitive and highly confidential nature of this information, estimates of possible future claims are virtually always prepared under close supervision and control of the firm’s legal counsel and treated as “attorney’s work product”. Because of its confidential nature, this analysis is NOT prepared for use as part of expert testimony, nor should it be prepared by an expert who may be called upon to provide testimony related to the case.
Adding together the estimates of the three components — settled claims, filed but unsettled claims, and possible future claims — we now have a working estimate of the likely liability costs of the case. At the outset, this estimate is likely to be only a very rough ballpark figure. As the case proceeds, so that the weight of verified and/or settled claims rises and the weight of possible future claims declines, the range of uncertainty of the estimates improves with subsequent updates and revisions.
Some unsettled claims and some previously unfiled claims are likely to become law suits if the injuries are serious. As the number of individual suits related to a specific situation rises, it becomes increasingly likely that a class action suit may be filed to consolidate a substantial number of the cases. If the proposed class is certified by the court, prospective claimants and plaintiffs are normally offered the option of joining in the class action or of “opting out” and continuing their suits separately. Typically, each category of claimant will then require separate analysis and cost estimation, which substantially increases the complexity of the analysis.
The analysis becomes increasingly dependent on the quality and detail of information available from claims databases developed by adjustors, attorneys, or insurance carriers and of internal production and marketing information. In most instances, these data are at best poorly suited for analytical use; but they are usually all that’s available.
Claims data can be especially messy, often including coding errors, missing data, and entry errors. Sound analysis usually requires that extensive tests of logical consistency of claims information be devised to pinpoint problems or screen out “spoiled” records. From experience, we have found it essential to develop and use a wide array of range and consistency tests when using claims data. However, in spite of these limitations, claim-specific data generally provide a much better foundation for analysis than aggregate data tabulated by someone else, which may merely conceal the problems and prevent their being detected. Skill and creativity in scrubbing raw claims data often become critical to the quality of liability cost estimates.
Company data on production and distribution of a product or component can vary from excellent to ambiguous, depending on the sophistication of the firm’s enterprise information system, production and inventory control system, and marketing and shipping information system. In many firms, much of this information is not computerized and must be compiled from a combination of hand-written and automated records. As with claims data, close scrutiny of production and distribution data is needed to check for and avoid potentially significant errors.
Data and other information provided by technical experts are usually of excellent quality; but their use may require analytical techniques that may be unfamiliar to many economists, such as queuing models, transition matrix models, or other techniques more commonly used by industrial engineers, operations researchers, actuaries, biostatisticians, and other specialists. The broader the range of techniques with which the analyst has experience, the better able s/he will be to use the available technical information and expertise in developing and refining liability cost estimates.
As the case progresses and the number of filed and settled claims increases, estimates of liability costs are usually revised and refined repeatedly at regular intervals. This enables the firm’s top management and its defense counsel to track changes in the estimates reflecting the latest settlement and claims data and to assess the possible effects of various settlement proposals and strategies. It is not unusual for mid-point estimates of aggregate liability costs to change significantly from one time to the next, especially if the adjustors group batches of similar claims together for settlement. If the batches are large enough, or contain far-above-average claims, they may cause breath-taking rises in the estimate, unless adequate care is taken to realistically assess their relative frequency based on technical factors, demographics, and other identifiable features. On the other hand, if larger claims have been processed first on a priority basis by the adjustors, the estimated liability may actually decline as the remaining (and presumably smaller) claims begin to be processed.
As the case progresses, the assessment and development of projection models for liability costs can serve an important role in the formulation and evaluation of possible settlement and negotiation strategies. Its most valuable function is the identification of settlement strategies that may appear superficially similar or equivalent but which are likely to result in large increases or decreases in aggregate settlement cost to the defendant firm and/or its insurance carrier.
Anticipating possible settlement proposals based on other cases permits the analytical team to assess their likely cost in the present case and to evaluate the possible consequences of alternative counter-proposals before a negotiating session actually occurs. It is particularly useful to identify for the negotiating team the relative sensitivity of the results to changes in different components of a proposal. Depending on the specifics of the situation, there are likely to be instances in which changes that look small may have big consequences and changes that look big may have only minor consequences. By knowing where such instances may exist, the negotiating team may be able to devise negotiating strategies in advance that favor movement along desirable paths and minimize possible movement along undesirable paths. For example, a comparison of likely settlement costs of suits filed by “opt outs” with likely settlement costs if they were persuaded to remain in the class by “sweetening” the class settlement by Y amount might be very useful to the negotiating team.
Cases involving long-term disabilities or continuing injuries must, in addition, be evaluated with attention to the timing of events, the possible future stream of disability costs and benefits, appropriate discount and interest rates, and possible structures for long-term or deferred compensation settlements. In some instances, this analysis may be assigned to specialized actuarial advisors. However, because of its importance to the negotiation of settlements, any actuarial analysis should be incorporated into the overall settlement simulation model along with other expert judgments affecting the case.
* The above work by Robert D. Shriner, Ph.D. was provided by the Technical Assistance Bureau.