LET'S TALK ABOUT YOU(TH)!
1Who answered, class by class
Change over time uses matched pairs only — students with both questionnaires and ≥80% of items answered.
The endline questionnaire was administered in the first 10 minutes of the speech class, immediately before students delivered their speeches (half the class that day, half the following week). Two consequences: (1) intervention students rated items like "I can present my ideas convincingly" while waiting to perform — pre-performance nerves plausibly depress exactly the self-efficacy items; (2) the endline was measured before the program's culminating experience — the speech, the audience, the feedback. Whatever those moments change is not in this data at all.
2How students experienced the program
Four endline items, intervention group only (n ≈ 120). This is the most direct evidence the questionnaire produced — the quality of the civic learning experience itself, the program's core promise. Compare sites, or drill to a single class.
Does the experience relate to change?
Within the intervention group, students who rated the experience higher also increased more on belonging (r = , p = , n = ).
Correlational, not causal — but it is what a working mechanism looks like: where the program landed as intended, the outcome moved.
3Who moved: students who started with low belonging
Split at the overall baseline median (3.25). Left: among students who started below it, how many increased, held, or decreased? Right: the same intervention low-starters, split by whether they had any previous community experience (school activities or outside volunteering).
4Where each group started
Baseline mean per item (1–4). The groups did not start from the same place — which is why raw change comparisons mislead.
5The ceiling problem
Share of students already at the maximum ("exactly like me") and at the minimum at baseline, per item. A student at the ceiling cannot show improvement no matter what the program does; a student at the floor cannot show decline.
6What changed — belonging and the individual items
Belonging is the only defensible composite; the agency and cognitive items are shown one by one. Left: raw pre→post means per group. Right: each student's change against their starting point — students who started low rose, students who started high fell, in all three groups.
7Item explorer
Full response distributions for any of the 10 pre/post items — Intervention Brașov, Intervention Voluntari, and Control side by side.
8What we can and cannot say (so far)
For donors and the public, the story this data supports: roughly 8 out of 10 students experienced the program as a space where they could be sincere, felt heard, and saw the connection to their real lives (§2). The students who arrived with the least — low belonging, no prior community experience — moved most consistently upward (§3). And where the program landed as intended, belonging rose with it (§2).
We cannot say: that the program raised self-reported belonging or the agency/cognitive items relative to control. After adjusting for unequal starting points, the group difference is zero (§6).
We especially cannot say: that control did better. That pattern is an artifact: control started lower (§4), at the floor of the instrument while intervention sat at its ceiling (§5), low scores drift up on their own (§6), and intervention students filled in the endline minutes before performing (§1).
The measurement lesson, stated plainly: the questionnaire could not have detected the program's effect on self-beliefs even if it occurred — too little response room, too few items, endline before the culminating experience. For 2027: harder items, wider response scales, behavior-frequency questions, a retrospective-pre block, endline after the speeches. The proximal layer — experience quality, speeches, observations, interviews — is where this evaluation's evidence lives.