The reliability of AO's data was tested by comparing the characteristics of cases as coded in the AO data with what researchers believed to be the more accurate information recorded by court clerks on individual case docket sheets, as obtained through the PACER system (Public Access to Court Electronic Records). In the large majority of districts, PACER allows public Internet-based access to docket sheets recorded since 1993; in some districts other case materials are also available. The study focused on two large categories of cases, torts and inmate civil rights, separating two aspects of case outcomes, i.e., which party obtained judgment and the amount of the judgment when plaintiffs prevailed. Subject to sample limitations, this study tentatively concluded that AO data can provide reasonably accurate estimates of the proportion of cases in which plaintiffs win damages judgements. A possible systematic understatement of plaintiff win rates is due to judgments recorded as for "both" plaintiffs and defendants that in fact tend to favor plaintiffs; however, this outcome classification accounts for a small percentage of trial outcomes. The analysis of awards distinguished between mean and median awards. The AO data systematically overestimated the mean award, such that studies relying on AO data to address issues about the level of awards probably overstated amounts paid to plaintiffs. For case categories with fairly large awards, substantially improved mean-award estimates can likely be obtained by substituting awards recorded on docket sheets for awards coded by the AO data as 9999. Estimates of median awards based on the AO data without further investigation appear to be of reasonable size and to provide useful upper bounds for true median awards. 10 tables and 104 notes