Economics - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemMoment Condition Verification and Treatment Effects Estimation in the BETEL Framework( 2022)In this thesis, we expand upon the literature to develop new advancements to the BETEL method, and we demonstrate them in the context of well-studied empirical problems. The first two chapters address the issue of moment condition validity. In Chapter 1, we develop a moment condition verification procedure by expanding on the Chib et al. (2018) framework with a hierarchical spike-and-slab prior similar to that by Geweke (1996). This allows us to test joint and elaborate hypotheses based on moment conditions in the Bayesian framework, which is shown to exhibit ideal properties under simulation. Additionally, the contributions in this chapter allows for a broader application of BETEL in complex empirical problems, including in structural VAR analysis, as considered in the following chapter. In Chapter 2, we apply the methodology developed in Chapter 1 to macroeconomic structural VAR models, verifying multiple sources of identification presented as moment conditions for the fiscal policy model of Blanchard and Perotti (2002) and Mertens and Ravn (2014). Verifying and comparing different sources of identification sheds new light on the validity of different identification strategies proposed in the literature. In Chapter 3, we depart from moment condition validity and introduce a BETEL-based estimator for causal treatment effects in the presence of skewness due to an unobserved state variable. This is motivated by previous studies, e.g. by Oreopoulos (2006) and Devereux and Hart (2010), showing returns to schooling for females to be statistically insignificant under standard approaches. This is driven by standard instrumental variables analysis not considering the differences in the female and male earnings distributions. In particular, the considerable negative skew in the female earnings distribution can be attributed to much higher, but unobserved part-time status, which is not accounted for by standard IV methods. Individuals in the left tail of the earnings distribution tend to have a lower return to schooling than others, for example, due to part time work (Ermisch and Wright, 1993; Connolly and Gregory, 2008; Hirsch, 2005). We show that this skewness is informative to identifying the treatment effect in the proposed BETEL-IV framework for the “subpopulation” of females from the non-skewed part of the earnings distribution, or that with comparable skew to males. Extending the method of Liu et al. (2017) to the IV setting, we are able to exploit this skewness, yielding a positive and significant return to schooling for females. This method is robust when applied to the male population, yielding similar results to the current literature, and when compared to other existing methods. Additionally, extensive simulation studies illustrate the performance of the estimator in identifying treatment effects for the relevant subpopulation, down-weighting the contribution of individuals in the tail due to the unobserved state variable. The methodological contributions in this thesis exploit and demonstrate the flexibility of the BETEL method and show how the BETEL framework can be expanded to address different empirical problems. Furthermore, this thesis presents empirical contributions arising from the BETEL framework, moving beyond illustrative examples to answering important empirical questions currently addressed in the literature, showing how developments in the BETEL framework can shed new light on existing empirical problems.