The primary aim of this project is to measure the impact of the communications-bundle, pooling together the IRS and University letter groups, on AOTC take up.

The study also includes a number of secondary aims. These secondary aims include:
1. Exploring whether messenger effects from who mailed the letter—the IRS or the University—generate different effects on AOTC take up or other secondary outcomes of interest.
2. Exploring whether there are heterogeneous treatment effects of the communications bundle(s) (e.g., independent and dependent students for tax purposes).
3. Exploring the effects of the communications-bundle(s) on intermediary outcomes (e.g., filing taxes, when taxes are filed, college credits attempted) and downstream outcomes (e.g., refund amount, other aid distributed, progress towards degree, academic performance)
4. Exploring the treatment-on-treated effects of the communications-bundle.

The experimental design includes a business-as-usual group and two communications-bundle groups. The University’s Financial Services Office (the University) sent email notices when the Form 1098-T, which is the tax statement used to claim AOTC, became available to students in the business-as-usual group. The University sent the same Form 1098-T emails and 5 additional AOTC emails to students in the communication-bundles groups. Note that the University sends copies of e-mail messages—including the Form 1098-T and AOTC e-mails—to a student’s authorized payer(s), typically a parent or guardian, when one or more is registered.
The communication-bundle groups were also mailed one letter about claiming the AOTC to their permanent address. The letter came from the IRS for students in the IRS letter and communications-bundle group and the letter came from the University for students in the University letter and communications-bundle group.

Primary outcome
A dichotomous indicator for claimed AOTC during tax year (TY) 2019 is the primary outcome of interest, which we will report in the OES Abstract.
Secondary outcomes
In the OES abstract, we also will report two secondary outcomes of interest:
● AOTC credit amount (0 to $2,5000); and
● Credits earned at the University in January 2020 - December 2020 (i.e., Spring 2020, Summer 2020, and Fall 2020 academic periods).
Planned outcomes to be analyzed for other publication:
We list planned tax-related outcomes, mostly observed in IRS administrative data and planned education-related outcomes, mostly observed in administrative data from the University, in the analysis plan.

First and second research question: OLS regressions

Following questions: OLS regressions and sensitivity analysis

Total communications-bundle sample 9,541
Total sample 19,071

Sample randomly assigned
To better understand the external validity of our results, we note that the random assignment
sample excludes the following students:
● students who appear to always have been enrolled as graduate students (based on year in
school variables) or who appear to always have been enrolled in non-degree programs;
16 We use the academic year rather than the academic period to account for some students taking courses
during summer periods.
Modified adjusted gross
income (MAGI)
Following Guyton et al. (2018), we account for reported income
interacted with TY 2018 filing status using the following
mutually exclusive groups:
● Reported income less than $60,000;
● Reported income between $60,000-120,000; and
● Reported income greater than $120,000
13 of 21
● students with 120 or more transfer credit hours (the minimum hours required to graduate
with a bachelor’s degree);
● students enrolled less than half-time each term; and
● students missing Social Security Number and gender information.
Tax-related data
Tax credit, tax refund, Economic Impact Payment amount, and HEERF aid amounts claimed will be
top coded at the maximum amount eligible to claim or receive. Other standard IRS procedures will
be implemented for other potential data exclusions.
Education-related data
College credits attempted and earned will be top coded at 20 credits per academic period. GPA
will be top coded at 4.0. Courses attempted and credits attempted measures will exclude
withdrawn and dropped courses.

Tax-related data
Non-filers: Students who did not file their taxes and were not claimed as a dependent for tax
purposes will be considered not to have claimed higher education tax credits.
Students who did not claim a higher education tax credit: Students who did not claim a tax credit
(including non-filers), will be considered to have a zero dollar credit amount.
Education-related data
2019 Graduates: A known source of missing data for education-related outcomes is students who
graduated prior to randomization, in spring or summer 2019 (N = 1,379). These students will have
missing data for education-related outcomes and will be excluded from analysis of
education-related outcomes. In practice, we will exclude random assignment blocks for students
who graduated in 2019 (i.e., graduated in 2019 and registered authorized payer, and graduated in
2019 and did not register authorized payer).
Additionally, a subset of students who were seniors at the time of randomization will graduate
during the Fall 2019 academic period, which is prior to our main outcome period for academic
outcomes. These students will also be excluded from our primary specifications for
education-related outcomes; however, we will conduct additional analysis that includes these
students and the Fall 2019 academic period to better understand the robustness of our findings.
2020 graduates: Students who graduated during the outcome period (Spring 2020 - Fall 2020) will
be included in education-related measures in post-graduation terms as follows.
Continued enrollment or graduated: They will be assigned values in the same way as other
students.
14 of 21
Other education-related outcomes: We will run this analysis as if students did not earn or
attempt credits in academic periods after they graduated. Additionally, we will examine
the robustness of our findings by bounding our estimates and excluding 2020 graduates
from analysis.
Other non-enrolled students: Missing data from other non-enrolled students for education-related
measures will be created as follows.
Degree progress measures: Missing credits attempted in an academic term will be
interpreted as the student not being enrolled at the University in that academic period. As
a result, missing values will be imputed to zero for the degree progress measures.
17
Academic performance measures: We will run this analysis in three ways. In the first
approach, we will exclude other non-enrolled students in the analysis of academic
performance measures in academic periods where they are not enrolled. In the second and
third approaches, we will attempt to bound the estimates by imputing outcomes.