Introduction to Causal Inference — DU B.A. (Hons.) Economics Previous Year Question Papers

There are two Sections, A & B with a total of 6 questions. Answer any 4 out of the 5 questions from Section B. Marks are given at the end of each question. 1. (a) State the Stable Unit Treatment Value Assumption (SUTVA) and explain its two components with examples. (b) Explain the difference between the Average Treatment Effect (ATE) and the Local Average Treatment Effect (LATE). Under what conditions are they equal? (c) When treatment effects are heterogeneous, an additional bias term appears in the SDO decomposition beyond the selection bias. State this decomposition (with all three terms) and explain the role of the heterogeneous treatment effect bias. (d) Consider the following hypothetical data where both potential outcomes are observed for 8 Individual Y(0) Y(1) D (Treatment) Compute the ATE, ATT, and ATU for this population. Compute the Simple Difference in Outcomes (SDO). Using the SDO decomposition, compute the selection bias. Interpret your findings. 2. Under the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA), unskilled agricultural workers are entitled to a statutory minimum daily wage. In a particular state, the daily wage cutoff for a wage top-up subsidy is ₹250 per day — workers with a pre-program baseline daily wage below ₹250 are eligible for the top-up, while those earning ₹250 or more are not. A researcher uses a sharp RDD to estimate the causal effect of the wage top-up on the number of days worked per month. The running variable is baseline daily wage, and the cutoff is ₹250. Data from workers with wages between ₹200 and ₹300 are used: Days_Workedᵢ = α + τ·Dᵢ + f(Wageᵢ − 250) + εᵢ where Dᵢ = 1 if the worker's baseline wage < ₹250 (eligible for the top-up). Variable Coefficient Std. Error p-value (a) What is the estimated causal effect of the MGNREGA wage top-up on days worked per month? Interpret the RDD estimate. (b) What is the predicted number of days worked per month at the wage threshold for a worker who just qualifies for the subsidy? (c) Is the estimate statistically significant at the 5% level? Justify your answer. (d) What is the key identifying assumption required for this RDD to yield a valid causal estimate? (e) How might workers or employers manipulating reported baseline wages around the ₹250 cutoff affect the validity of the RDD? What type of bias could this introduce? (f) Describe a diagnostic test or visual analysis that could help assess whether the RDD assumptions hold in this context. (3 + 2 + 2 + 4 + 4 + 3) 3. In 2023, a state government launched a rural road construction program targeting villages with a connectivity index score below 50 (on a 0–100 scale). Villages scoring 49 or below received all-weather road connections; villages at 50 or above did not. A researcher estimates the impact of road connectivity on annual agricultural productivity (₹ per hectare) using two methods: Regression Discontinuity Design (RDD): Villages with connectivity index scores between 40