Thursday, January 31, 2013
Class Composition and Career Choice
Massimo Anelli and Giovanni Peri have a paper on "The Long Run Effects of High-School Class Gender Composition"
The abstract reads:
"The long run earnings and career potential of individuals are strongly affected by their education. Among college educated individuals, the choice of college major is a very important determinant of labor market outcomes. In most countries men and women exhibit significant differences in this choice which is responsible for a large portion of the gender gap in earnings. In this paper we analyze whether the gender composition of peers (classmates) in high school affects the choice of major and hence long run earning potential. We use a newly collected and unique dataset covering 30,000 Italian students graduated from high school between 1985 and 2005. We exploit the fact that students are assigned to classes whose gender composition, within a school over time, varies exogenously. Moreover we are able to control for family, cohort, teacher and school effects in assessing the effect of peer-gender ratio on outcomes. We find that the gender ratio of peers in high school significantly affected the choice of major. A larger share of same-sex peers increases the probability of choosing majors associated to high earning jobs (Economics/Business, Medicine, Engineering). For women we also find that a large percentage of female high school classmates increases their long run performance in college and their earnings."
They say that "There are three reasons why we choose Italian high school students as ideal subjects to study the e ffect of peer gender on the choice of college major. First, in Italian high schools students are assigned to a class (usually made of 20 to 30 peers) when they enroll in high school and, except for small rates of attrition, they share with them the same teachers and in a pre-determined academic curriculum for 5 years (from 13 to 18 years old). These peers, therefore, interact with each other intensely during fi ve years of their school life, a period in which their personal and gender identity are fundamentally shaped."
"the class gender ratio was uncorrelated to any of the predetermined and observable characteristics of the students and of the class. This is the ideal context to separate school, cohort and teacher eff ects from e ffects of the class gender-ratio [..]. we now own data on 30,000 college preparatory high school students in the city of Milan (Italy) who graduated between 1985 and 2005. Those data include information on their school, cohort, class, identity of their peers, year of graduation, residence address when in high school, exit-test scores. Moreover we have merged this information with that on their college career (choice of major, graduation date, university attended) on labor market outcomes (earnings in year 2005, occupational choice) and on their family background."
Their main findings are
"First, we fi nd that, controlling for academic quality and family background, being assigned to a high school class with a larger share of people of the same gender increases the probability of choosing high earning college majors. This fi nding is true for both men and women. In particular women in high school classes with very high percentage of women have a 5-6% higher probability to choose high-earning majors than women in classes with high percentage of men. Similarly men in classes with high percentages of men have a 6-7% higher probability of choosing high earning majors. Second, those women who attended classes with large percentage of women and then choose high earning majors performed better than or as well as other women in those majors in terms of dropout rates and time to graduation. Hence the long run eff ect of attending high school classes with high percentage of women is, for a woman, a higher probability of enrolling and graduating from high earning majors and therefore higher expected wage. On the other hand men who attended high school classes with high percentage of men had higher probability of enrolling in high-earning majors but the final probability of graduating from high earning major was not diff erent from that of other men due to higher attrition. Finally, women who attended high school classes with high percentage of women and then enrolled in high earning majors had higher wages on the labor market relative to other women above and beyond (i.e. even controlling for) the e ffect driven by the major."
How do they rationalize their findings:
"These fi ndings can be rationalized with a simple human capital model in which the choice of major is a ffected by its costs and benefi ts. If the individual cost of attending a high earning major depends on abilities (negatively) and on the self-confi dence of a person (positively), the low enrollment of women in those majors may be due to a low degree of self-confi dence. The share of women in the same high school class may increase the self-con fidence of a woman, as also found in the experimental evidence presented in the behavioral economic literature (Gneezy, Niederle and Rustichini 2003, Niederle and Vesterlund 2007) and as hypothesized in the theoretical literature on gender identity introduced by Akerlof and Kranton (2000)."
For a paper on belief updating and gender differences in doing so, see also
Mobius, Markus M., Muriel Niederle, Paul Niehaus and Tanya Rosenblat, “Managing Self-Confidence: Theory and Experimental Evidence”, September 2012.
See my earlier post on gender differences and competitiveness for looking at psychological attributes that can account for gender differences in choices of areas of study. We find that competitiveness is correlated with such study choices. We do not find that overconfidence per se is correlated with choices, though feelings on absolute or relative mathematical ability are indeed very much correlated with study choices.