Trial 1
H1: There will be a higher proportion of re-engagement among people who received a message in their home language as compared with those who received a message in English (A1 > A0).
H2: There will be differences in re-engagement between the Department of Home Affairs messenger group and the service provider messenger group (B1 ≠ B0)

Trial 2
H1: There will be a higher proportion of enrolments among people who receive the letter/email in their home language as compared with those who receive the letter/email in English (A1 > A0).

Trial 1: A 2x2 factorial design, students will be randomised at the address level so people at the same addresses are in the same treatment group (clustered). We will stratify by service provider region to ensure balance in treatment group assignment across the four regions.

Trial 2: Potential students will be randomised at the address level so people at the same addresses are in the same treatment group (clustered). We will stratify by service provider and language region to ensure balance in treatment group assignment across the four regions.

Primary outcome
Trial 1: The primary outcome measure is re-engagement in the AMEP, indicated by a scheduled activity date (when the participant contacts the service provider to schedule returning to class).
Trial 2: The primary outcome measure is enrolments in the AMEP, as measured by registration date.

Secondary outcome
Trial 1: The secondary outcome will be the proportion of participants who click on the link embedded in the text message.
Trial 2: The secondary outcome will be the proportion of participants who visit the link embedded in the email and letter.

Trial 1
For both primary outcome measures, we will estimate treatment effects using Ordinary Least Squares (OLS) regression with standard errors (CR2) clustered using the following specification:
Y= α+ β_1 A+ β_2 B+ γ_1 X+ γ_2 X*A+γ_3 X*B+v+w
Where Y is either a binary variable indicating whether the student re-engaged. A indicates whether the cluster is allocated to the home language or English language group, B indicates whether the cluster was allocated to the Home Affairs or provider messenger group, X indicates the set of de-meaned baseline covariates (years of education) and vector of block indicators, X * A , X * B are the interaction of these covariates with the treatment indicators, v is a cluster-level error term and ω is the individual-level error term.
We will not adjust p-values for multiple comparisons. We will conduct logistic regression as a robustness check for our primary OLS specification and report average marginal effects.

Trial 2
For the primary outcome measure, we will estimate treatment effects using Ordinary Least Squares (OLS) regression with standard errors (CR2) clustered using the following specification:
Y= α+ β_1 A+ γ_1 X+ γ_2 X*A+v+w
Where Y is a binary variable indicating whether the student enrolled. A indicates whether the cluster is allocated to the home language or English language group, X indicates a vector of block indicators, X * A is the interaction of this vector with the treatment indicator, v is a cluster-level error term and ω is the individual-level error term.
We will not adjust p-values for multiple comparisons. We will conduct logistic regression as a robustness check for our primary OLS specification and report average marginal effects.

Expected sample size:
Trial 1: 26,000
Trial 2: 16,990