Using two nationally representative datasets from household surveys conducted in India in 2005 and 2012, the present paper examines the causal effect of gender in private school choice. We argue that the gender of the child is potentially endogenous in India because parents continue to have children until they have a son. To redress this potential endogeneity, we exploit the variation in private school choice among 7-18 year olds born to the same parents within the same household in an attempt to minimize both child-invariant and child-varying household-level omitted variable bias. We then explore the nature of female (dis)advantage across different types of households, communities and years with a view to assess the role of parental preferences in this respect and also its change. Significant female disadvantage exists in both survey years, though the size of this disadvantage varies across sub-samples and years. Female disadvantage is significantly higher among younger (relative to eldest) girls and also in northern and northwestern (relative to western) regions, but it is lower among girls from poor (relative to rich) households, Christian (relative to Hindu high caste) households, and those with more educated mothers. We infer that the observed within-household variations in female disadvantage across sub-samples reflect variations in non-altruistic parental preferences linked to deeply held cultural norms (for example, sons acting as old-age security and the exogamy of girls), access to schools and other public goods, thus posing considerable challenges to securing 'education for all.'