ECONOMICS 172: Issues in African Economic Development Problem Set 2
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
ECONOMICS 172: Issues in African Economic Development
Problem Set 2
1. African Economic Development and the COVID-19 Pandemic [4 points]
Please read the news article “‘An economic calamity’: Africa faces years of post-Covid instability” from The Guardian news (August 13 2021), which is posted in the bCourses Assignments folder. Based on the article, as well as the course readings, lecture slides and any additional materials you find relevant, briefly discuss how the COVID pandemic is likely to affect African economic growth and poverty reduction. Specifically:
What did economic and social impacts look like over the past two years?
What economic and political effects are likely over the next 5-10 years?
Make sure to discuss at least three potentially important channels of impact, at least one of which is likely to be positive.
Strictly enforced: your answer should be no more than 2 pages, with 12-point font and double-spacing; we will not read past page 2 and you will be penalized for using different font, spacing or margins.
2. Household Cash Transfers and Food Security in the COVID Pandemic [6 points]
The impact of income on nutrition is an important topic in international development policy. Starting in September 2020, thousands of households in Kenya received medium- sized cash transfers worth 600 Kenyan Shillings (valued at approximately US$12 in PPP terms) to assist them during the COVID- 19 pandemic. The amount is equivalent to a couple of days of work for typical unskilled laborers in Kenya. These unconditional cash transfers were targeted to relatively poor households (in both rural and urban areas) and delivered via mobile money. Approximately four weeks after the distribution of the cash transfers, the households were surveyed (by phone) and data was collected on economic and life outcomes, including some measures of food security, total spending, and household earnings, among many others. Importantly, before any cash grants were distributed, all households were also interviewed in a baseline survey round, and we can denote the baseline round time t=0 and the endline survey round t=1.
In reality, the program was designed as a randomized control trial (RCT), and the households receiving the COVID-19 cash assistance were randomly chosen. In this question, imagine that instead we only have data from the treatment households (cash=1), and not any of the control households, but we do have data on both survey rounds. Below you will estimate the impact ofthe cash transfer on nutrition using data on treatment households, from both the baseline and endline surveys.
a) [2 points] Discuss the econometric assumptions needed to make a pre-post-analysis (PPA) approach appropriate for understanding the impact of cash transfers on subsequent food security outcomes.
Given that we seek to understand the effect of household income on nutritional outcomes, why is it important that we have access to data from both the baseline and endline survey rounds?
How does the fact that the dataset cannot take advantage of randomization in the allocation of cash transfers across treatment households affect the estimation oftreatment effects, and how might it lead to omitted variable bias? Please illustrate your points with the discussion presented in class, as well as any other relevant material, and be sure to include the equations from lecture that illustrate potential bias.
In your view, are the required conditions likely to be met in the case of these COVID assistance cash transfers in Kenya?
Next please carry out some econometric analysis. As was the case for Problem Set 1, we will examine whether you have followed best practices in your R script. For instance, you should include detailed comments in your code for easier reproducibility. You may choose to produce an RMarkdown file that integrates your code (set echo=TRUE), your written responses, and tables displaying your regression results into a single PDF (please “knit to PDF” and turn in the resulting PDF file; do not simply turn in your .rmd file). Alternatively, you may choose to write up your answers using a word processor, include
copies/screenshots of your tables (note that you may find it helpful to use the type=”text” option of the stargazer command in this case) and attach a copy of your R script at the end. In either case, be sure that your submission includes, in some format, (1) your entire code/R script, (2) your written answers, and (3) your regression output. Please merge all documents into a single PDF before submitting.
b) [1 point] Please download the data to be used in the analysis. There are two ways to do this. You can click on the following link to copy the dataset for this exercise into your repository on U.C. Berkeley’s DataHub. RStudio should automatically open. You should see a new folder in your repository called ‘ECON-172-SP22’, and within that there will be a PS2 folder:
https://datahub.berkeley.edu/hub/user-redirect/git- pull?repo=https%3A%2F%2Fgithub.com%2Fds-modules%2FECON-172- SP22&branch=main&urlpath=rstudio%2F
Alternatively, you can download “Econ172_S22_PS2_data.csv” from the bCourses page. Once you have downloaded the data using either method, use the “read.csv” command to open it in R (or RStudio), either on your local computer, or on U.C. Berkeley’s DataHub, found onhttps://r.datahub.berkeley.edu/.
The dataset is a partial extract of data from the actual project, where each observation (row) in the dataset represents one household during one time period (“time”), and where the baseline survey round is denoted with time=0 and the endline round with time=1. Using the “lm” command, determine the average difference between households observed in the baseline round (time=0) and in the endline round (time=1) for each of the following two characteristics: an indicator variable for whether the household respondent is married (“married”), and an indicator variable for whether the household resides in an urban area (“urban”). (Consider these two characteristics one by one, that is, you should run two separate regressions here.)
Report the regression output for the two regressions and interpret the coefficients.
Please also discuss the standard errors and t-statistics. What differences (if any) do you observe, and why?
c) [2 points] Determine the average change in terms of households’ nutritional status (“nutrition”) between the baseline survey (time=0) and the endline survey round (time=1). Since all of the households in this dataset eventually received the cash transfer, the cash treatment indicator variable (“cash”) takes on a value of 1 for all households both at baseline and at endline, but recall that they received the cash between these two time points. The nutritional status measure is an index of food security and is created by combining survey questions on the respondent and their household members’ numbers of meals eaten, number of days that they had to cut back on or skip meals in the last week, and number of days they went to bed hungry. (The variables take on a sign such that positive values denote better outcomes, i.e., fewer days going to bed hungry is associated with a more positive value in the index.) Taken together, this index can be thought of as a summary measure ofthe household’s regular access to adequate food. The index is measured in standard deviations from the mean of zero, and can take on either positive (better) or negative (worse) values.
Estimate the difference in nutritional status again using the “lm” command in R. To do so, please carry out two regressions. In the first regression, estimate the effect ofthe cash transfer on nutritional status between the baseline round and the endline round without including any other explanatory variables. In the second regression, estimate the effect of the cash transfer on nutritional status while also including “urban” and “married” as explanatory variables.
Report the results of these two regressions and interpret all coefficients. Under what assumptions do these analyses have a causal interpretation and reliably estimate the impact of cash transfers? Do you feel these conditions are likely to hold here? Why or why not?
d) [1 point] Estimate the difference in total household spending between the baseline and endline survey rounds, while also including “urban” and “married” as explanatory variables.
Report the results of this regression and interpret all coefficients.
Do these patterns in any way help make sense of the results in part c. above? Discuss why or why not.
2022-05-10