Fig 1.
The relative time difference in hours between finishing the survey and the beginning of the exam.
Fig 2.
The time (in hours) relative to the exam when the treatment SMS was sent.
Fig 3.
The total number of treatment messages (e-mail and SMS) corresponding to an exam on a particular calendar date.
Note: E-mail messages were sent out at 8 pm the day before the exam. Text messages (SMS) were sent out at 7 am on the day of the exam. N of treatment e-mail = 14,974. N of treatment SMS = 14,277.
Table 1.
The difference in means between students allocated to Group A relative to Group B for each baseline covariate separately.
Fig 4.
Questions students answered before their first and second exam.
Fig 5.
The unconditional raw mean of primary and secondary outcome variables in treated and control groups with 95% confidence intervals.
Table 2.
Treatment effect on students’ endline exam grades, unstandardized regression coefficients.
Fig 6.
Conditional treatment effect of receiving the encouragement message on students’ endline exam grades, based on students’ baseline ability.
Table 3.
Treatment effect on students’ endline self-efficacy, unstandardized regression coefficients.
Table 4.
Treatment effect on students’ endline motivation, unstandardized regression coefficients.
Table 5.
Treatment effect on students’ endline test anxiety, unstandardized regression coefficients.
Table 6.
Hypothesized treatment heterogeneity.
Fig 7.
The explored conditional heterogeneous treatment effect (y-axis) on students’ endline test anxiety (left panel) and endline self-efficacy (right panel), based on students’ baseline self-confidence (x-axis). The left panel: corresponds to Model 10 in Table 5, N of observations = 2,014; N of respondents = 1,590. The right panel: corresponds to Model 10 in Table 6, N of observations = 2,016; N of respondents = 1,594.