The aim of this longitudinal ecological study was to analyze the trend in the proportion of low birth weight in Rio Grande do Sul State, Brazil, from 1994 to 2004 by panel data analysis and multilevel linear regression (two levels: by micro-region and time in years) to estimate risk factors associated with low birth weight. The proportion of low birth weight increased by 1.2% per year, and the multilevel model showed that the proportions differed between the micro-regions and increased over time, with the increase in the percentage of premature newborns, with the increase in the infant mortality rate, and with the increase in the cesarean rate. Among the micro-regions, the proportions of low birth weight varied positively with the urbanization rate and expenditures in the Unified National Health System and negatively with rate of participation in the workforce. According to the multilevel model, most of the variation in proportions of low birth weight was due to the effects of the micro-region of residence of the newborn's mother.