The aim of this ecological study was to determine the association of the frequency of dental health activities conducted as public health service and sociodemographic characteristics with caries prevalence in Japanese 3-year-old children using data for each municipality as one unit. Data on caries prevalence in 3-year-old children in 2000 were obtained from 3,251 municipalities (almost all municipalities in Japan). Caries prevalence for each municipality was recalculated using the empirical Bayes estimation model to make adjustments for variations in municipalities with small numbers of children. Explanatory factors included annual frequency of dental health activities such as dental health education and topical fluoride application programs and various sociodemographic characteristics. Multiple linear regression analysis was applied to examine the associations between caries prevalence and annual frequency of dental health activities with adjustment of other sociodemographic variables. Variations in caries prevalence in municipalities with small populations became smaller after being adjusted by the empirical Bayes estimation model. Multiple linear regression analysis showed significant associations of caries prevalence with sociodemographic characteristics such as the proportion of residents with university degrees, total fertility rate, and the proportion of workers in the manufacturing industry (p < 0.001). Annual frequency of topical fluoride application showed a significant but relatively weak association with caries prevalence (p < 0.001), explaining only 0.8% of the total variation in caries prevalence, whereas no correlation was found between caries prevalence and annual frequency of dental health education. It was concluded that the effect of the fluoride application service on caries prevalence in 3-year-old children, although statistically significant, was of far lower clinical significance than sociodemographic characteristics.