This is a socio-economic household survey dataset which was collected in urban centres in the state of Maharashtra, India. The dataset covers information about social capital and cooperation, household assets, the household’s experience of urban riots and various other shocks, perceptions of safety, coping strategies, and specific demographic information for each household member. The main purpose of this study is to fill this theoretical, empirical and policy gap by analysing how the relationship between populations living in contexts of violence and armed non-state actors controlling or contesting those areas results in forms of local governance and order, and how these in turn affect the access to and effectiveness of livelihoods adopted by individuals and communities in contexts of violence. The study is based on comparative qualitative and quantitative empirical work in Colombia, India, Lebanon, Niger and South Africa.

Data collection method: Districts in Maharashtra were categorised (based on Maharashtra police data) into three categories: high rioting districts (5 or more riots per district per year), medium rioting districts (more than 1.5 and less than 5 riots per district per year), and low rioting district (less than 1.5 riots per district per year). We took into account the geographical spread of the state by choosing districts that represented all administrative regions and socio-cultural divisions in the sample. Our final selection included three districts in each of the medium- and low-rioting clusters, and four in the high-rioting cluster. Within these 10 districts, 45 neighbourhoods were then randomly selected from the list of voting-booth zones obtained from the Maharashtra Election Commission corresponding to urban sites with a history of violence. The field team began household interviews simultaneously from a set of starting points in each of the neighbourhoods agreed a-priori, working their way inwards in each neighbourhood making sure that no alley, no matter how small, was missed by following a right-turn pattern at all junctions. Households were randomly selected through a skip pattern, which for larger neighbourhoods was 7 or 8 households, while for smaller neighbourhoods was 4 to 5 households. This multi-staged sampling framework resulted in a final sample of 1089 households, spread across forty-five neighbourhoods, in ten districts in Maharashtra. Households were interviewed face-to-face with a structured questionnaire.