Notes: *** and ** indicate significance at the 1% and 5% levels, respectively. The standard errors (in parentheses) are corrected for clustering at the village level. The base category for ‘parents’ bias’ is unbiased. The regressions are based on a full set of controls, viz., individual and sibling controls including age, age square, interaction of age and gender, number of brothers of the child, number of older siblings, and a vector of dummies indicating number of siblings; parental and household controls including a vector of dummies indicating the highest education level among parents and indicating the occupation of the household head, mother’s age, a dummy variable for whether the household head is Muslim and for whether the household has electricity, and the household’s landholding per capita; and a village dummy. The illness regression (Column 1) includes additional controls for access to safe water and sanitation; the regressions for formal treatment and treatment cost (Columns 2 and 3) also include additional controls for the duration of illness. The sample includes children aged 0–15 years in households involving joint decisions.

Source: Authors’ calculations.