(b) The community leader believes that women and girls are neglected because there are few employment opportunities for women. What are three alternative explanations for gender discrimination? (c) The community plans to introduce an employment program where married women aged 30- 45 are given an opportunity to earn a salary outside the home. How would you measure the impact of this program on gender discrimination using a randomized evaluation? a. How would you select program participants? b. What outcomes would you measure? c. What comparisons would you make? (d) How could you measure the effect of this program using a "natural experiment" or "difference-in-differences" approach? (e) Describe a research design you could use to test whether one of the alternative explanations you mentioned in Part (6) was valid. What evidence or information would you collect? What types of comparisons would you make? What would your explanation predict you should find? 5. Barriers to Improving Education [10 points] Suppose that policymakers in your country are interested in learning about the relative importance of different barriers to educational attainment among the poor. One group of officials believes that the high costs of education are the major constraint while another group believes that what matters is a lack of information on the returns to education. Some economists at the World Bank offer theoretical arguments as to why both constraints matter and interact with each other. You are granted a budget to design a randomized experiment with 10,000 poor households. You can construct up to three treatment arms in this experiment (a) Describe up to three treatments that you would use to determine which of the three hypotheses is correct. How would you assign the 10,000 households across treatment and control groups? (b) Describe the research design in terms of (i) the key dependent variables of interest, and (ii) the procedure to estimate the treatment effects. (c) Do you need to have a baseline survey in order to identify the causal impact of the treatments? Or can you just use a single follow-up survey? Explain