| This dissertation consists of two essays on the estimation of instrumental variable models. In the first essay, I propose the quasi empirical likelihood (QEL) estimator and apply the estimator to analyze household child labor supply in Côte d'Ivoire during 1993–1995—a period spanning a devaluation of the regional currency. My primary goal is to develop a method that reduces finite-sample bias in estimating instrumental variable models while remaining computationally feasible when the number of orthogonality conditions is large. The QEL estimator, which approximates the method of empirical likelihood via semiparametric efficient estimation, is shown to be nearly unbiased under Bekker's many instrument asymptotic approach—but at a fraction of EL's computational cost. Thus, QEL is robust to many instrumental variables, thereby eliminating the need to choose instruments. Monte Carlo experiments confirm the accuracy of my theoretical results. When applied to the household child labor market in Côte d'Ivoire, QEL produces intuitive estimates of the parameters for household child labor supply. That is, household child labor supply is increasing in the marginal value product of child labor and in the household's endowment of child labor, while decreasing in the price of rice and household net sales of rice. Further, land, hired labor, and chemicals are substitute inputs for child labor; adult familial labor is complementary to child labor.; In the second essay (joint with Xenia Matschke), we propose that labor market characteristics (e.g., unionization and immobility) are important determinants of trade protection. According to the Grossman-Helpman protection for sale model, labor is perfectly mobile and cannot organize politically, therefore having no role in the determination of trade protection. This essay generalizes these assumptions to allow for trade union lobbies, collective bargaining, and immobile labor. As a result, our model produces empirically verifiable implications that we test using data from United States manufacturing industries in 1983. Overall, we conclude that labor market characteristics do indeed play important roles in the determination of trade policy—we reject the restrictions imposed by the nested Grossman-Helpman model because the labor market variables enter our estimation to a statistically significant degree. |
