Data Analysis for Impact Evaluations – Quiz 3

1. Suppose that you designed a clustered impact evaluation, but then you forget to cluster your standard errors during analysis. Assume that intracluster correlation is positive. Which of the following will be true about your results? (select all that apply)

Regress total Year 3 endline scores on treatment and child_female, estimating robust standard errors. Now run this regression again with standard errors clustered at the village-level.

2. What is the standard error on treatment in the second regression? Round to the nearest 0.01.

3. Which of the following statements is true about the output from these two regressions? (select all that apply)

4. Which of the following statements is true about how clustering affects the precision of impact estimates, compared to non-clustered designs with the same sample size? (select all that apply)