ARE214 Winter Quarter 2023
Final Exam
Instructions. As soon as you download this exam, send me an email so that I know when you are
starting. From the moment you download this exam, you have 48 hours to complete it. Once
you complete your exam, please upload a pdf version to Canvas. The latest you may submit is
5:00 pm on Friday, March 24. Please include both: 1) your write-up and; 2) your Stata (or other
software) code you use for your estimations. You may consult any notes or readings, but you
must work independently without consulting other people. If questions arise about the exam,
please email me.
The Impact of Drought Tolerant Maize
Convinced that drought risk was inhibiting agricultural investment and productivity growth in
Africa, the Bill Gates Foundations spent hundreds of millions of dollars to create drought
tolerant (DT) maize seed varieties for Africa. The result was a series of new DT seed varieties
that scientists claimed would withstand mid-season drought stress, protecting farmers’ income
and, by extension, protecting the well-being of their families.
With the goal of testing these claims, in 2016, UC-Davis researchers launched a randomized
controlled trial of DT seed varieties in East and Southern Africa (Tanzania and Mozambique,
respectively). Endline (post-treatment) data were collected from farmers in 2018. While the
results from this study have not yet been published, you have the opportunity to be one of the
first to analyze the data and see if Bill Gates spent his millions wisely.
In the two countries, researchers identified 56 “clusters,” each comprised of 3 villages which
were close to each other and shared similar characteristics. Of these 3 villages, one was
randomly assigned to the control group, and the other two were assigned to the DT treatment. In
treatment villages, farmers were provided with information on the DT seeds and the opportunity
to buy the seeds locally through a local village agent. For this exam, you will work with data
from 5,568 farmers (approximately 33 per-village) collected in 2018, after the DT program had
been implemented.
The Stata data file, “DTMaize.dta,” includes the following variables for this sample:
Yield: Kilograms of maize produced per-hectare planted;
Assign: This dummy variable takes the value of 1 for farmers living in villages assigned to the
DT treatment, and equals zero for farmers in control villages;
Use_DT: This dummy variable takes on the value 1 if the farmer planted DT seeds;
2
Drought: This dummy variable takes on the value 1 if the farmer suffered a mid-season drought
during the maize growing season;

Cluster: This is a categorical variable that identifies of the 56 three-village clusters the
household is located.
1. Estimate the Intent to Treat (ITT) effect on maize yield. To do so, use the following
specification:
= 0 + 1 + � 56
=1
� + (1)
where i indexes the farmer and c the triad where the farmer lives. is the treatment
assignment variable, and is a dummy or indicator variable that takes on the value of 1 for
any observation in cluster c. In Stata, it is easy to generate the 56 cluster dummy variables
by including “i.Cluster” as a regressor.
a. Report and write a short paragraph interpreting your results. Please be sure to explain
what the estimated ITT impact means, describe its statistical significance and comment
on its economic significance (to do this, you might want to compare the ITT to the mean
of yields for farmers in the control group).
b. Bill also wants to know about the impact of DT on farmer income. Although you don’t
have income data, you can do some “back of the envelope” calculations. The typical
farmer in the sample cultivates 2 hectares of maize and spends $100 per-hectare on seed,
labor and other inputs. Farmers can sell their maize for $0.50 per-kilogram. The typical
farm also earns $200 per year in the off-farm labor market. Using these numbers, what is
the annual income a typical control group farmer earns (you’ll need average control
group yields to answer this question)? Using your ITT impact estimate, please calculate
the post-treatment annual income of a farmer in a treatment village. Please again write a
short paragraph clearly interpreting your income impact results to Bill.
2. You proudly show the results of your analysis to your boss and tell her that you are ready to
send the results to Bill. She grumbles that the impacts do not seem “big enough,” says
something about checking for imperfect compliance and tells you to continue your analysis.
She reminds you that what Bill cares about is the whether or not DT seeds work for those
who adopt them.
a. Please calculate and report the fraction of the following three types of farmers in the
sample: “Always Participators”, “Never Participators” and “Participate if Assigned”.
Please describe what each type means in the context of the DT maize study.
b. Using the information from part 2a, calculate the Local Average Treatment Effect for the
impact of DT seeds on maize yields. Please write up a paragraph or two describing your
3
LATE estimates for Bill. Make sure you carefully describe the population for which you
can draw conclusions about the causal impact of the program using your LATE estimate
and clearly explain the difference between your ITT and LATE estimates.
c. Your boss likes your new impact numbers, but only wants to send one set of results to
Bill. Please write her a short paragraph explaining which estimate you should send on.
(Remember what Bill cares about!).
3. After completing the above tasks, you recalled several conversations with the scientists who
developed the DT maize varieties. Specifically, you recall asking if farmers should only
expect benefits from planting DT maize during drought years or if they would also benefit in
normal years. The scientists were very excited at your question and claimed that indeed the
benefits should be largest during drought years but that, because they used improved genetic
material, farmers should also see small positive benefits from planting DT maize even during
non-drought years. .
a. Please write down a modified version of the ITT regression model in equation (1) that
allows you to test the scientists’ claims that the impact of DT seeds is: i) positive in both
drought and non-drought years and ii) larger during drought years than non-drought
years. Carefully explain your model and formulate the hypotheses you want to test. Note
that your dataset includes the dummy variable, Drought.
b. Estimate your modified model and report your results. Write a short paragraph for Bill
explaining what your results tell you about whether DT functions as insurance for farmers.
Among other things, please use your results to report your estimate of the difference in
maize yields for a farmer in a control village who faces a drought versus a farmer in a
treatment village who faces a drought. Do the same when they do not face a drought.
c. Using the same assumptions that you used in problem 1b, please calculate the ITT impact
of DT seeds on famer income in drought years. Please explain your findings to Bill and
Warren.
d. Finally, to keep your boss happy, calculate the LATE estimate of the impact of DT seeds
on farmer income in drought versus non-drought years, and write an additional paragraph
explaining your estimates and their meaning.