UA 323
Problem Set 3
Development Economics
Due: April 3rd
This problem set if based on the paper “Missing Women and the Price of Tea in China: The Effect of Sex-Specific Earnings on Sex Imbalance”, by Nancy Qian (2008). Read the paper and answer the following questions:
1. What is the key question that the paper tries to answer? Not the practical thing they actually do, but the Big Picture question.
2. Could the author answer this question by comparing sex ratios and educational investments across households that have the same total income, but where the wife earns a higher share of the income? Why or why not?
3. To answer the Big Picture question, the paper uses a difference-in-differences approach.
3.1 Describe the approach.
3.2 What is the underlying assumption that must hold for this approach to deliver an unbiased estimate of the causal effect of increasing women’s income share within the household? Explain clearly what this assumption means in this context (it is not sufficient to mention the name of the assumption).
On Brightspace, you can find a file called TeaChina with some of the synthetic data that allow replicating the analysis in Qian et al. (2008), as well as an accompanying file with the variable labels.
4. Upload your dataset in R. Notice that the dataset is in .csv format, and not the standard .xlsx – this is avery common format for data. You can import data in this format using the read_csv command. So, if you wanted to upload a dataset called Dataset.csv in R you would use the following code:
data <- read_csv(“Dataset.csv”)
As usual, you should adjust the code accordingly.
5. Create a variable (call it tea) that indicates counties with above-median tea sown area, and another variable (call it orchard) that indicates counties with above-median orchard sown area.
5.1 What is the mean fraction of males in counties that have above-media tea sown area across all years? What is the mean fraction of males in counties that have above-media orchard sown area across all years?
5.2 Can we interpret the difference in the fraction of males across tea and orchard intensive counties as causal evidence that increasing female income improves survival rates for girls? Why or why not?
6. Create a variable (call it post) that is equal to one for years after the reform, namely years after 1979, and equal to zero for years up to (and including) 1979.
6.1 What is the mean fraction of males born up until (and including) 1979? What is the mean fraction of males born after 1979?
6.2 Can we interpret the difference in the fraction of males born before and after 1979 as causal evidence of the effect of increasing female income on the survival rates for girls? Why or why not?
7. Now create two additional variables. Create a variable (call it tea_post) equal to one for years after 1979 and counties that has above-median tea sown area. Create another variable (call it orchard_post) equal to one for years after 1979 in counties that has above-median orchard sown area. Now you have all the elements to calculate the difference-in-differences for the mean fraction of males outcome.
7.1 Calculate two separate difference-in-differences: one where the treatment is being born in a tea cultivation intense county, and one in which the treatment is being born in an orchard cultivation intense county. For each of these treatments, the output should be a table like the one we saw for Duflo (2001) on slide 46 of Lecture 9.
7.2 Are the results in line with households operating as described by the unitary model? Why or why not?
8. Derive the same difference-in-differences for (i) girls’ years of education, (ii) boys’ years of education.
8.1 Report the difference-in-differences in the same format table required in the previous question.
8.2 If you were a researcher helping the Chinese government to design an unconditional cash transfer program to increase female education in the country, what recommendations would you give to policy makers?
9. Refer to the education investment model we discussed in Lectures 9 and 10. Assume that the utility husbands and wives derive from educating their daughter are respectively:
Where: h(s1) = rs1 + 0.5ψs1(2). Notice that we are implicitly assuming that husband and wife share the cost of education equally. Assume households are not credit constrained.
9.1 Derive the optimal level of education the wife would choose if she was making all decisions within the household. Derive the optimal level of education the husband would choose if he was making all decisions within the household.
9.2 According to your answer to question 8, is mw higher, equal, or lower than mh?