To identify groups within Dhaka slums that report similar patterns of livelihood, and to explore nutritional and health status.

A random sample of households participated in a longitudinal study in 1995–1997. Socio-economic and morbidity data were collected monthly by questionnaire and nutritional status was assessed. Cluster analysis was used to aggregate households into livelihood groups.

Dhaka slums, Bangladesh.

Five-hundred and fifty-nine households.

Main outcome measures: Socio-economic and demographic variables, nutritional status, morbidity.

Four livelihood groups were identified. Cluster 1 (n = 178) was the richest cluster with land, animals, business assets and savings. Loans as well as income were higher, which shows that this group was credit-worthy. The group was mainly selfemployed and worked more days per month than the other clusters. The cluster had the second highest body mass index (BMI) score, and the highest children's nutrition status. Cluster 2 (n = 190) was a poor cluster and was mainly dependent selfemployed. Savings and loans were lower. Cluster 3 (n = 124) was the most vulnerable cluster. Members of this group were mainly casual unskilled, and 40% were femaleheaded households. Total income and expenditure were lowest amongst the clusters. BMI and children's nutritional status were lowest in the slum. Cluster 4 (n = 67) was the second richest cluster. This group comprised skilled workers. BMI was the highest in this cluster and children's nutritional status was second highest.

Cluster analysis has identified four groups that differed in terms of socioeconomic, demographic and nutritional status and morbidity. The technique could be a practically useful tool of relevance to the development, monitoring and targeting of vulnerable households by public policy in Bangladesh.