Essays on Treatment Effects from Multiple Unordered Choices
Galindo Pardo, Camila Andrea
Urzúa, Sergio S
MetadataShow full item record
I study some of the methodological and empirical challenges associated with estimating treatment effects of one option versus another, in contexts where agents can choose from many alternatives with no clear rank (i.e., one option is no better than the other, for everyone). In particular, I focus on educational decisions throughout the life-cycle, such as parental choice of childcare, students' choice of high school, and college enrollment. First, I present a strategy to overcome a limitation of instrumental variables in these settings, where there are many endogenous choices. I use this strategy to provide empirical evidence from an early childhood development intervention in Colombia, where parents can choose among different childcare options (e.g., small centers, large centers, or home care). In the third chapter, I focus on the Chilean high school context where students can choose from three types of schools: academic, vocational, or hybrid. I find that, while academic schools seem to improve the student's academic achievement, the effects of hybrid and vocational schools depend on the student's fallback option (i.e., what they would have chosen if their preferred option was not available). Last, in the Colombian context, jointly with Maria Marta Ferreyra and Sergio Urzúa, I examine the labor market returns to short-cycle degrees versus bachelor’s degrees and versus obtaining a high school diploma. Chapter 2 presents a strategy to estimate causal effects in settings where agents can choose from many options along with empirical evidence from an early childhood development intervention in Colombia. I exploit the joint effect of discrete and continuous instruments on the probability of choosing an option. These combined effects of different instruments have been recognized and studied in contexts where there are only two alternatives. In turn, current methods for multiple unordered choices implicitly assume that the potential response to one instrument is the same across the distribution of other instruments. Instead, I allow for the response to the variation in one instrument (for example, an offer of a slot at a childcare center) to differ depending on other instruments (for example, proximity to the center). To do so, I employ a latent utility framework and model agent's responses to the instruments through their effect on each option's costs. With assumptions motivated by economic theory (i.e., convexity of cost functions), I define conditional vectors consisting of combinations of potential choices that differ along the distribution ofa second instrument. I use conditional vectors and recent advances in the instrumental variables literature to estimate local average treatment effects. With this strategy, I empirically assess the effect of different types of childcare (e.g., small centers, large centers, or home care) on the cognitive, nutritional, and socio-emotional development of children from 0-5 years of age in Colombia. My results suggest that childcare centers with better infrastructure and services could improve some children's cognitive development. In contrast, existing estimation methods would find overall negativeeffects of these centers on cognitive development. In Chapter 3, I estimate the effects of different high school types on educational achievement, such as high school completion and higher education enrollment. I find evidence that suggests that attending a vocational high school does not have a differential effect on the probability of enrolling in a vocational college. Moreover, while hybrid schools seem to foster student enrollment in bachelor’s programs, this effect largely depends on the student's fallback option. In particular, there is no evidence of improvements in educational achievement among students who would have chosen academic schools instead of hybrid schools. In Chapter 4, with Maria Marta Ferreyra and Sergio Urzúa, we provide evidence of diversion and expansion effects of changes in the local supply of short-cycle degrees, in the context of higher education for Colombia. Our results suggest that most students would divert from bachelor's- and into short-cycle- degrees as the local supply of short-cycle degrees changes. For these students we find significant gains, particularly among women, in terms of participation in the formal labor market and years of experience.