Final Exam – Data Project ECON 6074: Causal Inference Module 2 – 2021/2022 Instructions Please submit your work in one zip file containing three files: a) write up of your answers (including tables), b) log file, and c) do-file that works with your matched data set. You will conduct empirical analysis using a dataset from the study by Duflo, Dupas, and Kremer (2011) “Peer Effects, Teacher Incentives, and the Impact of Tracking: Evidence from a Randomized Evaluation in Kenya” American Economic Review, 101(5): 1739-1774. Each student is assigned with a unique dataset. Look up your matched Final Exam Dataset ID number and download your unique data set “Final_M2_2022_ID[#].dta”. If your answers do not match with your assigned dataset you will receive 0 points. 1. [25 points] The Impact of Tracking on Students’ Achievements a. (5 points) What is the authors’ identification strategy to estimate the impact of tracking on students’ achievements Explain how this method can identify the treatment effect using the potential outcomes framework. What are the key assumptions associated with this identification strategy b. (5 points) Conduct a balance check to examine whether students in tracking and non-tracking schools are similar to each other. c. (10 points) Report the OLS estimate of the effect of tracking on standardized endline test score [stdR_totalscore] with and without individual controls (control variables: student’s gender [girl], age [agetest], being assigned to contract teacher [etpteacher], percentile in initial distribution [percentile]). d. (5 points) Provide a test of whether the effect of tracking on standardized endline test score differs
across students at different percentiles of initial distribution (e.g. students in the bottom half versus students in the top half). 2. [35 points] The Effect of Assignment to Upper Tracking Section on Achievement a. (5 points) Propose an identification strategy to estimate the impact of being in upper tracking (top half) section on achievement Use the potential outcomes framework to illustrate the identification strategy and key assumptions. b. (5 points) Create a graph which plots standardized endline test score [stdR_totalscore] on the assignment variable [percentile] and visually inspect whether there is discontinuity at the cutoff. c. (5 points) Create graphs to examine whether covariates [girl; agetest; etpteacher] are balanced around the cutoff point. Do students at the bottom of the upper section look similar to students at the top of the lower section d. (10 points) Use parametric approach to estimate the effect of being assigned to the upper section [tophalf] on test score. That is, regress test score on indicator for upper section controlling for the assignment variable, individual controls [girl; agetest; etpteacher], and school fixed effect [schoolid]. Use different polynomials of the assignment variable: 1. Linear, 2. Quadratic, 3. Cubic. And report the estimated coefficients on assignment to upper section. How do the estimates vary across different polynomials e. (10 points) Use local linear regression (nonparametric approach) to estimate the effect of being assigned to upper section. Report results from using at least three different bandwidths. How does the estimate from the nonparametric approach compare to the estimate from the parametric approach in part d