Essays on selection,treatment effects and gender.
详细信息
文摘
This dissertation consists of three essays on selection, treatment effects and gender. The first chapter is an essay on selection, with an application to the measurement of the gender wage gap. I explore the idea that selection into employment could be different in different parts of the female labor market to show that usual estimates of the gender wage gap could be biased. I propose a novel estimator which accounts for unobserved heterogeneity in selection and applies to comparable groups of men and women in terms of labor force attachment. Using US data I estimate a closing gender wage gap from 1976 to 2005. The second chapter, which is joint work with Lena Edlund, reassesses one of the causes of female empowerment, nowadays evident in all levels of society. An influential argument in the literature was that the contraceptive Pill allowed women to delay fertility, and therefore delay marriage and pursue higher education. However, in the early periods, Pill access was restricted to married women, and we show that marriage, combined with the Pill, acted as a facilitator, rather than a barrier, to womens educational and professional attainment. The third chapter, co-authored with Azeem Shaikh and Edward Vytlacil, establishes conditions under which we can use an instrumental variable to sign the average treatment effect. In a simple setting of binary outcome, treatment and instrument, we derive tight bounds on the average treatment effect using monotonicity restrictions on the outcome equation, the equation for the endogenous regression, and both restrictions together. We show that those restrictions imply testable conditions related to the strength of the intent-to-treat effect, and we use them to describe when the average treatment effect can be signed.