In most countries, girls perform better than boys in reading but worse in mathematics. However, there is much variation between countries. Explanations for the gender gaps include the organisation of the school system, students' expectations and macro-societal factors. The purpose of this paper is to account for gender differences in both reading and mathematics among 15-year-old students using data from the OECD's 2000 Programme for International Student Assessment (PISA) project. In most countries, school system factors are associated with the gender gap in reading but not in mathematics. Generally, gender differences in students' occupational expectations do not account for the gender gaps, although expectations contribute to the gender gaps in reading in New Zealand and the United States. Although several macro-societal factors -- the proportion of women in the workforce, societal inequality and public sector spending -- are associated with the gender gap in reading, the correlations are only moderate, unstable and, importantly, are not associated with the gender gaps in mathematics. The much stronger association between the gender gaps in reading and mathematics across countries implies that they are both influenced by policy: the extent that countries have successfully implemented policies to promote the educational outcomes of girls and young women. In such countries the gender gap in mathematics is small or non-existent but the gender gap in reading is relatively large. Policies shift both gender gaps in tandem. (Contains 3 tables.)