November 06 2008
3:30pm - 5:00pm,
Cubberley Building, Conference Room 114
Assistant Professor, University of Pennsylvania
Topic: "Understanding Stereotypical Gender Differences in Test Scores"
Abstract:Understanding gender disparities in educational test scores is an important and often contentious issue. Much of the previous literature has emphasized the pervasive nature of gender differences in
test scores across various datasets and time periods. The strongest evidence suggests that, while average test scores are often similar for males and females, due to differences in the variance of test scores there are robust differences in the gender ratios scoring in the upper tails on standardized tests. The general findings match prevailing stereotypes -- with males overrepresented among high-achievers in math and science and girls overrepresented in high reading scores. This evidence has led to extensive discussion regarding the potential causes (genetic and/or environmental) of such differences. In this paper we approach this topic in a new way that sheds light on the consistency of gender differences and in so doing contributes new evidence to debates about the extent to which the sources of gender disparities in high performance are innate or environmentally driven. Using national test-score data from the U.S. , we analyze the variation that exists in the gender gap across geographic areas. We find that gender differences in test scores vary significantly across the U.S. and that this variation is systematic along several dimensions. First, states with a large gender gap in favor of males in math and science also tend to have a large gender gap in favor of females in reading. Hence, states with large math/science disparities are not simply failing their females, but rather appear to adhere more to prevailing stereotypes in all directions. The second systematic finding is that variation in stereotypical gender disparities is also geographically clustered. States with the most gender-unequal test scores are more likely to be located in the South and the Mountain West while states with the most gender-equal test scores tend to be located in U.S. New England and the Pacific Region. Finally, the levels of gender disparities that exist in each state are highly correlated with average state-level responses on questions to women’s issues on the General Social Survey.We argue that this variation is consistent with differences in environmental and cultural forces across states that affect adherence to gender stereotypes. Since there are many environmental factors that do not vary at the state level (e.g., national culture), the significant variation we find likely understates the effect of environment on gender ratios in the upper ranges of test scores. Our analysis provides clear evidence that test-score disparities are greatly affected by different environments within the , suggesting that attempts to address these disparities are by no means pre-destined to fail.