This paper asks whether gender bias in education expenditure in rural India fell over the two-decade period from 1995 to 2014. We find that instead of falling over time, the channel through which gender bias is practiced changed dramatically over the 20 years. Secondly, the paper demonstrates the usefulness of distinguishing between the two potential channels of gender bias, namely bias in the school enrolment decision, and bias in the conditional educational expenditure decision, rather than in the single unconditional education expenditure decision; this distinction is shown to be important because gender bias in the enrolment decision has greatly fallen but bias in the conditional expenditure decision has significantly risen over time. Thirdly, we find that individual child level data has much greater power to detect gender bias in education spending, compared to household level data. Lastly, household fixed effects analysis shows that the observed gender biases in education spending are a within-household phenomenon in rural India.