' tot al_blind_score the total score obtained by an exam when it was graded blindly 0 tot al_score the total score obtained by an exam when it was not graded blindly 0 brahmin_org a variable that takes value 1 if the exam was actually solved by a child in the Brahmin caste and 0 otherwise ' brahmin_ob a variable that takes value 1 if the exam was randomly assigned a Brahmin identity when it was given to the second set of of teachers to grade 0 age_ob the age of the student randomly assigned to an exam when it was given to the second set of teachers to grade 0 f emalegb a variable that takes value 1 if the exam was randomly assigned a female identity when it was given to the second set of teachers to grade (a) (5 points) Explain why simply observing that children belonging to the Brahmin caste obtain higher scores on exams might not necessarily indicate discrimination in grading. Use a Stata command to support your answer. Show the Stata command and its output. Interpret the output. (b) (5 points) We want to test whether there is discrimination in grading in our experiment. Think about the exams assigned randomly to Brahmin caste as the Treatment group and the exams assigned to other castes as the Control group. We want to rst test that the two groups have balance when it comes to the underlying quality of answers. Use a Stata command to run the balance test. Show the Stata command and its output. Interpret the output. (c) (6 points) Do teachers assign larger grades when they think an exam was written by a child in the Brahmin caste? In other words, do teachers assign larger grades to exams in the Treatment group? Use a ttest to answer this question. Show the Stata command and its output. Give an answer to the original question using the relevant part of Stata output and a 10% signicance level. (d) (6 points) The papers are graded by two sets of graders within this experiment. How would you apply the difference-in-differences approach in this context to account for that difference? Please provide a numerical value for the result of this approach, but do not engage in statistical inference. Does this approach provide a better answer to the question of grading discrimination than the previous approach? Explain in 1-2 sentences