Survey methods

In late 2014 we completed a household census in village hamlets in the Udaipur district in Rajasthan, India to confirm the eligibility of the hamlet, enumerate the population, and identify potential respondents for inclusion. We conducted a baseline survey in early 2015 among 3177 mothers with a child between one and six years of age living in 160 village hamlets from five blocks. After describing the study objectives, procedures, potential risks, potential benefits, voluntary nature, confidentiality and privacy protections, and compensation, each eligible respondent was asked if they would consent to participate. This was done in written form for respondents who could read and write and orally for those who could not. Each respondent who agreed to participate received a blanket, valued at 100 rupees (Rs.), as compensation at the completion of the survey interview.
The survey questionnaire was translated from English to the local Hindi language by a professional translator, and the software for recording responses via tablet was by the Institute for Financial Management and Research—Leveraging Evidence for Access and Development (IFMR-LEAD) software team. To minimize error a survey supervisor observed each surveyor for one survey each day, monitoring for correctness in asking survey questions, providing necessary clarifications, and responding to respondent questions. A random set of 10 % of completed surveys was selected for back-checks–a list of questions asked in the survey that are unlikely to change over a short time period, like the number of children, respondent’s marital and educational status and employment history. The back-check team independently re-interviewed the respondents and entered their responses into a software program designed to test these questions for identical responses. A third party clarified any discrepancies between the original survey response and the back-check response. More detail on specific outcome measures and covariates are provided below.