Does tutoring raise exam scores?

Al Waha Learning District (fictional) runs an after-school mathematics tutoring programme for grade-11 students at a cost of about SAR 1.8 million a year (an invented scenario figure). A board member circulated this year's results with a blunt comment: “Tutored students scored 5.0 points lower on the final exam than everyone else. The programme isn't working — cut it.”

The district's head of planning asks you to analyse the data properly before the board votes on next year's budget. The data (student_scores, synthetic) hold 400 students: school type, last year's mathematics score, self-reported weekly study hours, attendance, whether they attended tutoring, and this year's final exam score.

Important context from the programme office: tutoring places were offered mainly to students whose previous score was low. Participation was not random.

Stakeholders: the board (budget decision), the programme office (who wants the programme continued), and the parents of tutored students (who want to know if it helps).

The district, schools and students are synthetic, generated for teaching.