Child undernutrition remains a major child health and developmental issue in low- and middle-income countries. The concentration (clustering) of underweight children among siblings at the family level is known to exist in India. This study examined the extent and covariates of clustering of underweight children at the sibling and family level in Uttar Pradesh, the largest state of northern India. Clustering of underweight (low weight-for-age) children was assessed using data on 7533 under-five children from the National Family Health Survey (NFHS) conducted in 2005–06, analysed using binary logistic and binomial regression models. Related bio-demographic, socioeconomic and health care variables were used as covariates in the models. The odds of being underweight for the index child were about two times higher (OR=2.34, p<0.001) if any of the siblings within the household was malnourished or underweight. A longer birth interval increased the odds of a child being underweight. The odds of underweight were significantly lower (OR=0.69, p<0.001) for children born to normal-weight mothers compared with those born to underweight mothers. Similarly, the odds of underweight were significantly lower (OR=0.49, p=0.01) for children born to educated mothers (high school and above) compared with those born to illiterate mothers. The results of the binomial regression model suggested that the deviations between observed and expected number of children were positive (3.09, 3.78 and 2.71) for 1, 2 and 2+ underweight children within the households of underweight women, indicating the concentration of underweight children among underweight/malnourished mothers. Underweight children were found to be clustered among underweight mothers with multiple underweight siblings. The findings suggest that policy interventions need to focus on underweight mothers with multiple underweight children.