OBJECTIVE: To assess sample representativeness and the precision of estimates of immunization coverage obtained with the 30 by 7 cluster sampling method proposed by the World Health Organization, by applying the method to determine immunization coverage in two municipalities (Diadema and Sao Caetano do Sul) in the state of Sao Paulo, Brazil, in 2000. METHOD: The representativeness of the samples was determined by comparing the census sectors picked by lot for the surveyed sectors and for the nonsurveyed sectors in both municipalities, in terms of socioeconomic and demographic characteristics (age distribution of the population, schooling, proportion of households headed by a women, monthly income of household head, and sanitary conditions of the home (piped-in water, connected to the sewer system)). The precision of the coverage estimates for the vaccines in the basic immunization schedule-BCG; diphtheria, pertussis, and tetanus (DPT); poliomyelitis; hepatitis B; measles; and measles, mumps, and rubella (MMR)-was determined by calculating the design effect and the width of the confidence intervals. Precision was considered to be satisfactory if the design effect was below 2.0 and the confidence interval width was below 10%. RESULTS: In both municipalities the comparison between the surveyed and nonsurveyed sectors showed a similar distribution in terms of socioeconomic and demographic variables. Concerning the precision of the estimates, the design effect was below 2.0 for all the vaccines, both in Sao Caetano do Sul and Diadema. In Diadema, the confidence interval width was below 10% for all the vaccines, except for MMR (10.1%). In Sao Caetano do Sul, only 89% of the expected sample were included, so the width of the confidence interval was slightly above 10% for the poliomyelitis vaccine (10.3%), the hepatitis B vaccine (11.8%), the mumps vaccine (10.4%), the MMR (12.9%), and the complete schedule (11.2%). CONCLUSION: The cluster sampling method proposed by the World Health Organization produces representative data as long as the methodological procedures for selecting the sample are rigorously followed in the field. [ABSTRACT FROM AUTHOR]