This dissertation consists of two chapters on topics in economics of education. In the first chapter, I develop and estimate an equilibrium model of endogenous instruction and student effort in order to empirically investigate the relationship between instructional choices, classroom composition, and student achievement in elementary school. The model allows teachers to vary in instructional ability and to value differently the achievement of students with different levels of prior knowledge. Using a unique dataset that combines school administrative data with rich information on instructional practices from five US school districts, I find that teachers attach a higher value to the achievement of students at lower quantiles of the distribution. I further explore the model’s implications by simulating a counterfactual scenario where I track students into classrooms based on prior test score performance. Results show that tracking has heterogeneous effects on students with different levels of prior knowledge. Moreover, the distribution of these effects depends on the mechanism used to assign teachers to classrooms. In particular, the combination of tracking with the assignment of high-ability teachers to lower tracks would benefit students at the bottom tercile of the distribution despite the lower level of peer quality. The second chapter (co-authored with Flavio Cunha and Kenneth Wolpin) reports the results of the evaluation of a parenting intervention developed and implemented by the Alief Independent School District in Texas. The goal of the intervention is to encourage and train parents to teach their children foundational skills for Pre-K. The results of a randomized controlled trial based on three yearly cohorts show that the program impacted parental investments and child development as measured by two different tests of school readiness. I go beyond reporting program impacts by building and estimating a model of parental choice of input levels. The model allows for a production function of knowledge that features individual-specific coefficients that capture the marginal productivity of parental inputs. I find that the mechanism posited for the program’s impact is validated by the model estimates.