Cognitive competence is a fragile construct and a range of factors influence whether individuals perform above or below their ability.  Using rich longitudinal data collected by Young Lives from the state of Andhra Pradesh, India (2002-2009), this paper studies two cohorts of children at two points in time (at 5 and 8 years, and 12 and 15 years of age) and examines gender inequality in performance in language and math tests.  Newly developed quantile estimation techniques in conjunction with traditional Blinder-Oaxaca decompositions are applied to examine the boy-advantage (and changes over time) in these test scores.

The evidence suggests that indeed, looking at average differences can be misleading since substantial heterogeneity exists across quantiles. In addition, the production function (determinants) for language and math are different. Specifically, a widening of the boy-advantage takes place across quantiles and over time for the younger cohort in language, and for the older cohort in math.  This pattern is most pronounced for girls in math at 15 years, where the largest gap (and change in gap) is at the 95th percentile. Unexplained effects are found to contribute most to these gaps, which is often interpreted in the literature as evidence of gender discrimination. Additionally, child-level factors contribute most to these unexplained gaps. This highlights the possible influence of characteristics such as self-efficacy on the relatively poor test performance of (the brightest) girls in the current context.

This paper was presented at a conference on Inequalities in Children?s Outcomes in Developing Countries hosted by Young Lives at St Anne's College, Oxford on 8-9 July 2013.