Ideally, to evaluate the empirical support for assertions about the benefits and costs of Rule 23 (b) (3) class actions, we would select a statistically representative sample of class action lawsuits and measure key characteristics and outcomes of those lawsuits. To select such a sample we first would have to construct a complete list of class action lawsuits, either for the nation or for selected jurisdictions. At the present time, developing such a list would be very difficult. Prior to 1995, reports of class action filings in the federal courts contained many inaccuracies; since then, the Administrative Office of the U.S. Courts believes that its effort to improve district reporting of class action status has improved the reliability of its database. However, the change has been so recent that many of those cases would still be pending and we would not be able to measure outcomes. Moreover, many class action lawsuits are filed in state courts that do not identify class action lawsuits in their reports of civil case filings. The only way to identify class actions in state courts would be to systematically review individual case files, a gargantuan task to undertake in multiple jurisdictions. If we could construct a sampling list, we then would need to collect information about each case we chose during our sampling. Because much information about class action litigation processes and outcomes is not contained in court records, we would need both to collect court record data and to survey key participants in each lawsuit to obtain data for analysis. To draw statistical inferences using this approach, we would need to sample a large number of cases and survey the participants in them.
We did not have the resources available to adopt such a large-scale approach, nor were we confident that we could surmount the considerable sampling and data collection problems it would entail. Moreover, many of the data we were interested in pertained to litigation processes; survey research techniques are not as well suited to studying processes as they are to measuring characteristics of individuals and organizations. Instead, we adopted a case study approach.
Case study research calls for selecting a few examples of the phenomenon to be studied and then intensively investigating the characteristics of those examples