辅导ECON 383、data留学生讲解、讲解R语言、讲解R编程设计
辅导R语言编程|讲解SPSS
Assignment #1
This assignment uses data from a randomized experiment undertaken by Benhassine et al (2015) in
Morocco. The purpose of the experiment was to determine the effectiveness of cash transfers as an
incentive for families to make educational investments in their children. The experiment explicitly
targeted poor rural communities, where more than 40 percent of children drop out of school before
completing the full six years of primary education.
The experiment was undertaken in 320 rural school sectors (these are like school districts, or
communities). Each school sector was randomly assigned to a control group or one of four treatment
groups. All families with school-age children who resided in one of these 320 school sectors were part of
the experiment. Families in the control group received no cash transfers. Families in two of the treatment
groups received a small monthly cash transfer that was conditional: to receive the monthly transfer, their
children needed to attend school regularly. Within this group, families were further randomized into one
of two groups: one in which the cash transfer was paid to the mother, and one in which the cash transfer
was paid to the father. Families in the other two treatment groups received a small monthly cash transfer
that was unconditional: families were told that the purpose of the transfer was to support their children’s
education, but they received the transfer whether their children attended school regularly or not. Again,
families within this group were randomized into one of two groups: one where the transfer was paid to the
mother, and one whether the transfer was paid to the father.
So, to summarize, families were randomly assigned to one of five groups (group and benef are
variables in the data for this assignment, described further below):
group Transfer Type Transfer recipient (benef)
Control 0 None n/a
Treatment 1 Unconditional Mother
Treatment 1 Unconditional Father
Treatment 2 Conditional Mother
Treatment 2 Conditional Father
The data for this assignment are a subset of the data collected at the experiment’s “baseline”, i.e., at the
start of the experiment. At baseline, families had been randomly assigned to the treatment and control
groups, but did not know the details of the experiment and had not yet begun to receive treatment. You’ll
find the data in the excel file a1_baseline.xls. Each row corresponds to one household. Variable names are
in the first row, and are as follows:
Variable name Description
hhid ID number of the household
schoolid ID number of the school sector
group indicates whether family is in control group (0), unconditional transfer group (1),
or conditional transfer group (2); see table above
benef indicates whether transfer recipient is Mother or Father; see table above
province Province of residence
survey_status Indicates whether family completed baseline survey or not
hhsize Number of people residing in this home
dwellingsize Floor area of home in m2
electricity Indicates whether the home is connected to electrical network
watersuppply Indicates whether the home is connected to water network
child1 Indicates whether there is at least one child (child #1) age 6-15 residing in this
home
child1_inschool Indicates whether child #1 is currently enrolled in school
child2 Indicates whether there is a second child (child #2) age 6-15 residing in this home
child2_inschool Indicates whether child #2 is currently enrolled in school
child3 Indicates whether there is a third child (child #3) age 6-15 residing in this home
child3_inschool Indicates whether child #3 is currently enrolled in school
child4 Indicates whether there is a fourth child (child #4) age 6-15 residing in this home
child4_inschool Indicates whether child #4 is currently enrolled in school
When doing a randomized experiment, it’s usually a good idea to check whether there’s any evidence that
the randomization was done incorrectly. We can do that by looking for evidence that the baseline
characteristics of treatment and control groups are systematically different. That’s what you’ll do in this
assignment.
Answer each of the following questions using R. Assemble your R code, output, and a short written
answer to each question in a nicely formatted html document. To create that nicely formatted html
document, use the RMarkdown template provided to you (and don’t forget to knit it!). When
you’ve finished the assignment, upload the html file via Canvas.
1) Create a new variable called treatment that equals: zero if the household is in the control group;
one if the household received an unconditional transfer paid to the mother; two if the household
received an unconditional transfer paid to the father; three if the household received a conditional
transfer paid to the mother; and four if the household received a conditional transfer paid to the
father. Report the number of households in the control group and each of the four treatment groups.
2) For the control group and each of the treatment groups, report:
- the average number of people residing in the home
- the average floor area of the home
- the proportion of homes connected to the electrical network
- the proportion of homes connected to the water network
- the average number of children residing in this home
- the average number of children enrolled in school
Do these means and proportions provide any evidence that the treatment and control groups are
“unbalanced,” i.e., that there are systematic differences in treatment and control groups at baseline?
3) For each of the four treatment groups, test whether the average number of people residing in this
home is different than the control group mean at the 5% level of significance.
(No, we haven’t discussed how to do this test, but I have given you the tools you need to build a
reasonable test. There are many ways to do it. Be creative!)