Data Analyst Task with Prof.
Anran Li
Overview
In this task, imagine that the United States Department of Education (ED) has contacted you to
provide data analysis that will inform policy decisions. As a representative of Cornell, your job is
to use real-world data to address the questions brought by ED and present your findings in a brief
memo.
Your memo will address the questions further below. It should be no longer than one page, single-
spaced, in 12-point font. Please include any tables or figures after your text; tables and figures do
not count toward the one page limit. Throughout your answers, please be brief. We value clarity
more than length. Additionally, please bear in mind that your audience includes policymakers who
may not have a technical background or training; it will be important to convey your results in a
way that can be understood in non-technical terms.
When you are finished, please combine the following in a single, zipped folder:
– your clean data
– any raw data files you used in their original format
– the script you wrote to produce your analysis. Please ONLY use Python to complete data
analysis.
– your policy memo for ED saved as a PDF. Include any tables or figures in your memo after
all text.
– a readme text file describing the contents of the folder
Your goal is to produce files that other members of your research team could use without much
further explanation. Please keep that in mind as you organize and document your work. If any
issues or questions arise while you’re working, please note them in an appendix section at the
end of your memo. This task will help us assess your current qualitative and quantitative research
skills, as well as your ability to convey results to policymakers with who may not have training in
quantitative research methods. A “perfect” score is by no means required for the job. With that in
mind, please do this work yourself without input from other people.
A. Assembling the Data
The first step is assembling a clean panel data set of colleges from the Integrated Postsecondary
Education Data System (IPEDS). These data are publicly available online through the US Depart-
ment of Education website (here). To download the files, visit the IPEDS homepage, click “Use
the Data” and then select “complete data files” from the dropdown menu in the “Survey Data”
section. You’ll draw on two sections of the IPEDS surveys:
– “Directory Information” files from the “Institutional Characteristics” survey
– “Student Financial Aid and Net Price” files
1 of 3
Your clean data set should contain the following variables:
– ID_IPEDS: a unique identifier for each institution
– stabbr: two-letter state abbreviation
– year: the four-digit academic year, where “2010-11” is coded as 2010, “2011-12” is 2011,
and so on
– degree_bach: a dummy variable that identifies bachelor’s degree-granting institutions
– public: a dummy variable that identifies public institutions
– enroll_ftug: the total number of first-time, full-time undergraduates
– grant_federal: total amount of federal grant aid awarded to first-time, full-time undergradu-
ates
Some details:
– Your panel should cover the academic years 2010-11 through 2015-16. We’ll use a balanced
panel, so include only the colleges for which data exist in all years.
– Exclude Washington, D.C. and U.S. territories. These have the following state abbreviations:
DC, FM, MH, MP, PR, PW, VI, GU, AS.
– Restrict the sample to undergraduate institutions, which may offer bachelor’s and associate
degrees, as well as diplomas and certificates. Institutions that only offer graduate degrees
should be excluded from your sample.
– Include both public and private institutions in your sample.
– Use total count variables where they exist rather than manually adding subcategories.
– Do not use the revised data files. IPEDS identifies these files with the suffix “_rv”. Likewise,
do not use any of the revised or imputed variables in the dataset. The original reported values
will suffice.
– Please download files listed under the “Data Files” category (as opposed to the “Stata Data
Files” category).
B. Questions: Enrollment
ED has heard anecdotally from a number of public, two-year colleges that their enrollment has been
declining from the 2010-11 to the 2015-16 academic years. Please address the following question.
For the sake of these questions, consider a “two-year college” as an undergraduate institution that
does not grant bachelor’s degrees.
2 of 3
1. ED would like to know whether the total number of students enrolled at all public, two-year
colleges in the U.S. has increased, decreased, or stayed the same from 2010-11 to 2015-16.
Use your dataset to address this question and create one figure to help illustrate. Describe
any possible shortcomings of using the dataset you compiled to answer this question.
C. Questions: Financial Aid
ED is interested in maintaining an equitable distribution of federal grant aid across states. Please
address the following questions. Use only data from the 2015-16 school year.
1. Vermont has complained that its neighbor New York receives substantially greater federal
grants on a per-student basis. Assess this claim.
2. ED is interested in maintaining a similar per-student federal grant amount across states.
Provide a few summary statistics detailing the spread of average per-student federal grant
aid across states. Describe what we can learn from these statistics.
3. Suppose that, rather than allocating federal grants directly to students, ED is consider-
ing a new system in which federal grants are allocated at the school level based on en-
rollment. Each school will be given federal grants according to the following formula:
grant_ f ederal = 1750 ∗ enroll_ f tug + .15 ∗ enroll_ f tug2 . Simulate this new allocation.
How does this new allocation formula affect the spread of average per-student federal grant
aid across states?
3 of 3