PIP: The Expanded Program on Immunization (EPI) is using a simplified cluster sampling method, based on the random selection of 210 children in 30 clusters of 7 children each, to estimate immunization coverage levels. This article analyzes the results of this method in actual and computer simulated surveys. Results from 60 actual surveys conducted in 25 countries were available for analysis, for a total of 446 sample estimations of immunization coverage. 83% of the sample results had 95% confidence limits within + or - 10%, and none of the surveys had 95% confidence limits exceeding + or - 13%. In addition, 12 hypothetical population strata with immunization coverage rates ranging from 10%-99% were established for the purposes of computer simulation, and 10 hypothetical communities were established by allocating to them various proportions of each of the strata. These simulated surveys also supported the validity of the EPI method: over 95% of the results were less than + or - 10% from the actual population mean. The precision of this method, as estimated from the results of both actual and simulated surveys, is considered satisfactory for the requirements of the EPI. Among the actual surveys, the proportion of results whose confidence limits exceeded + or - 10% was greatest (50%) when immunization coverage in the sample was 45%-54%.