RESUMO
OBJECTIVE: To adapt and develop a method for finding out what proportion of the variation among small areas in the number of births to married women is excess (systematic) variation over and above the chance (random) variation. METHODS: We adopted a two-stage sampling procedure to select 20 sub-centres in south India. We contacted all households and collected information on recent births and socioeconomic variables from all ever married women aged 15-49 years. Systematic and random components of variance were estimated using Poisson regression, adjusting for socioeconomic factors. RESULTS: Of the observed variance in fertility, 65% is explained by the estimated systematic variation. Though many important explanatory variables are considered, the systematic variance is significant mainly among illiterate women, those aged under 18 years at marriage, the marginalized population, and those with labourer husbands. CONCLUSIONS: Poisson regression can be adapted to estimate the random and systematic components of variation in fertility among small areas. The systematic component of variance can further be adjusted for socioeconomic factors. Identification of the significant predictors will help health planners develop necessary interventions at the small area level.