Stirling Children’s Wellbeing Scale — Pilot Report
Same-school comparison · 24/25 Season · 188 primary pupils aged 8–11
Pilot Headlines (24/25 Season)
points higher average wellbeing among BFF participants (44.37 vs 41.63, p = .029)
statistically significant difference in Positive Outlook (22.6 vs 21.0, d = 0.39)
effect size (Cohen’s d) on overall wellbeing, a moderate difference
effect size on “I feel that I am good at some things”, the largest single difference
Why Wellbeing
The evidence linking physical activity to young people’s mental health has grown quickly. A 2026 meta-analysis of 375 randomised trials with 5 to 18-year-olds found that physical activity interventions reduced symptoms of depression (SMD −0.45) and anxiety (SMD −0.39) (Singh et al., 2026), and an earlier review of positive outcomes found improvements in wellbeing, positive mental health and resilience (Andermo et al., 2020). The evidence is thinner for primary-age children specifically (Rodriguez-Ayllon et al., 2019). In the UK, the Education Endowment Foundation rates social and emotional learning as moderate impact for very low cost, measured against attainment. Our lunchtime sessions combine these elements: structured football during the school day, led by youth workers trained in social and emotional learning.
How the Pilot Worked
The Setting
The pilot ran at our largest primary school partner, in Westminster. The school sits in an area ranked among the 10% most deprived in England on the Income Deprivation Affecting Children Index. We run lunchtime football sessions there, and children choose whether to take part.
The Instrument
We used the Stirling Children’s Wellbeing Scale, a 15-item questionnaire developed and validated with 1,849 children aged 8 to 15 (Liddle & Carter, 2015). Children rate statements such as “I’ve been in a good mood” and “I think good things will happen in my life” from Never to All of the time. Twelve items make up a total wellbeing score from 12 to 60, split into two subscales:
- Positive Emotional State: calm, cheerful, relaxed, in a good mood, getting on with people.
- Positive Outlook: feeling good at things, finding fun things to do, feeling proud, feeling cared for, making choices, optimism.
The remaining three items check for social desirability, meaning a tendency to give the answer a child thinks adults want to hear. The scale’s guidance gives a typical average of 44, with the middle half of children scoring between 39 and 48. We use that band as a reference point throughout.
The Design
Pupils completed the questionnaire at school over one week in the 24/25 Season and told us what they usually do at lunchtime. Of 195 responses, 188 pupils gave consent. 98 take part in BFF lunchtime sessions and 90 do something else, such as other games, free play or quiet activities. We compared the two groups’ wellbeing scores.
This is a same-school comparison at a single point in time. It is the pilot for a quasi-experimental study. It tested whether the instrument works with our age group, whether running it in school is practical, and whether a comparison within a school shows a difference worth investigating at scale.
Analytical Approach
We compared average scores between the groups and report the size of each difference as an effect size (Cohen’s d). Following the scale’s guidance, our headline results exclude 33 pupils whose answers to the social-desirability items suggested they may have been answering to please, leaving 155 pupils (82 BFF, 73 others). We checked every headline result against the full sample, against non-parametric tests, and in a model adjusting for age, gender and sport outside school. Details are in the technical section.
What We Found
Overall Wellbeing
BFF participants’ average sits at the scale’s typical level of 44. Other pupils at the school average just below it. The confidence intervals are wide, which is expected in a pilot of this size, and the true difference could be anywhere from very small (d = 0.04) to medium-to-large (d = 0.68).
Fewer Children With Low Wellbeing
Outlook and Emotional State
Looking at the individual statements, one stands out. “I feel that I am good at some things” averaged 4.1 among BFF participants and 3.3 among other pupils on the five-point scale, roughly the difference between quite a lot of the time and some of the time (d = 0.86). A sense of competence sits at the heart of our theory of change, and it matches what our coaches describe on the pitch: children growing in confidence as they find something they are good at.
Next Steps: The Full Study
The pilot gave us enough evidence to design a stronger test. The next phase extends our lunchtime model to eight inner-London primary schools, reaching around 800 children, and builds the evaluation into delivery from the start:
- Matched comparison schools. Four schools, matched on free school meals eligibility, Ofsted rating and size. Two begin sessions immediately and two start a term later, acting as the comparison group in the meantime.
- A baseline. 240 children complete the Stirling scale before sessions begin and again afterwards, so we can measure change in the same children rather than a single snapshot.
- Outcomes beyond self-report. School-recorded attendance and attainment sit alongside the wellbeing measure.
- Co-designed with NPC. We aim to co-design the evaluation in collaboration with New Philanthropy Capital (NPC), with independent analysis of the results.
The pilot also shaped the design. Measuring before sessions start removes the biggest weakness of this snapshot. Comparing whole schools means the comparison group no longer depends on which children choose to play. And the effect we observed tells us roughly how large a study needs to be. Detecting a difference of this size reliably in a simple comparison would need about 120 children per group, and a before-and-after design gets more precision from each child.
Technical Details
Five blocks. Click any to expand the underlying methodology and limitations.
Data Cleaning & Scoring
- 195 responses collected during the 24/25 Season, almost all within a single week; 7 pupils declined consent, leaving 188.
- Responses coded 1 (Never) to 5 (All of the time). Total = 12 scored items (range 12–60); Positive Emotional State = items 9, 10, 11, 12, 14, 15; Positive Outlook = items 1, 3, 4, 5, 6, 8.
- Group coding: pupils whose lunchtime answer named Bloomsbury Football were coded as BFF participants (98); all others as the comparison group (90).
- Sport outside school coded from a free-text question; answers describing informal play at home were coded as no organised sport.
- Social desirability screen: the scale’s three social-desirability items (2, 7, 13) were summed and pupils scoring 14–15 or 3 were flagged, per the scoring guidance (33 pupils: 16 BFF, 17 comparison). Headline results use the remaining 155.
- Four pupils gave the same answer to every item; excluding them changed no result (see Robustness).
- Internal consistency was good: Cronbach’s α = 0.88 for the total, 0.84 for Positive Emotional State, 0.77 for Positive Outlook.
- Timestamps and free-text answers were removed before analysis. Only group-level results are reported.
Statistical Methods
Group comparison: Welch’s t-test (unequal variances) on total and subscale scores, with Mann-Whitney U as a non-parametric check.
Effect sizes: Cohen’s d with 95% confidence intervals: (Mean_BFF − Mean_comparison) / SD_pooled. Conventional benchmarks: 0.2 = small, 0.5 = medium, 0.8 = large. We describe the headline d = 0.36 as a moderate difference.
Adjusted model: linear regression, Score ~ BFF + Age + Gender + Outside sport. Interaction terms tested whether the gap differed by gender or outside sport. It did not (p = .81 and p = .44).
Distribution: chi-square test on the share of pupils scoring below 39, the lower bound of the scale’s typical range.
Item-level: Mann-Whitney U on each of the 12 scored items, with Holm correction for multiple comparisons.
Resampling: 10,000-draw permutation test and 2,000-draw bootstrap confidence interval for d.
Analysis in R 4.5.3 (tidyverse, effsize, broom, boot). The script is held with our analysis files and reproduces every figure in this report.
Robustness Checks
The headline difference in total wellbeing, under different samples and tests:
| Check | n | Difference / effect | p |
|---|---|---|---|
| Headline: social-desirability screened | 155 | +2.74 pts · d = 0.36 [0.04, 0.68] | .029 |
| Full sample, no screening | 188 | +2.93 pts · d = 0.35 [0.06, 0.64] | .018 |
| Also excluding uniform responders | 154 | +2.66 pts · d = 0.35 [0.03, 0.67] | .035 |
| Mann-Whitney U (screened) | 155 | Medians 45.5 vs 41 | .023 |
| Adjusted for age, gender, outside sport (screened) | 155 | +2.88 pts [0.18, 5.57] | .037 |
| Adjusted for age, gender, outside sport (full) | 188 | +2.68 pts [−0.02, 5.37] | .051 |
| Permutation test (full) | 188 | 10,000 draws | .017 |
| Bootstrap d (full) | 188 | d 95% CI [0.07, 0.65] | – |
The direction and size of the difference are consistent across every check. The adjusted full-sample estimate sits just outside conventional significance, which is expected at this sample size.
Limitations
- Self-selection: Children choose to take part. Those who pick football may already feel more confident or capable. Longitudinal research shows that sport participation and wellbeing influence each other in both directions (Vella et al., 2017), so we cannot separate the two here.
- Group differences: The BFF group has more boys (67% vs 34% in the screened sample), is slightly younger (9.3 vs 9.6 years) and more often plays sport outside school (64% vs 39% of all respondents). Adjusting for these left the gap similar, but other unmeasured differences may remain.
- Make-up of the comparison group: When we split the comparison group by what children do at lunchtime, pupils in free play or non-physical activities scored similarly to BFF participants (44.0, n = 39), while pupils who described other physical activities scored lower (38.9, n = 34). These subgroups are small and self-described, so we cannot explain the pattern, and the full study compares whole schools for this reason.
- No baseline: Many BFF participants had attended sessions before the survey was run. This is a snapshot of children already taking part, so it cannot show change over time.
- One school, one point in time: Results may not generalise to other schools. The reference range comes from the scale’s original validation sample, not a London population.
- Self-report: Primary-age children’s answers can be influenced by mood on the day and by wanting to please. We screened for social desirability, which removed 18% of responses, and results held with and without the screen.
- Statistical power: At this sample size the pilot could reliably detect only differences of about d = 0.4–0.45 or larger. Confidence intervals are wide, and the subscale and item results should be read as signals to test, not settled findings.
Sources
- Andermo et al. (2020). Sports Medicine – Open, 6, 25. doi:10.1186/s40798-020-00254-x
- Eime et al. (2013). International Journal of Behavioral Nutrition and Physical Activity, 10, 98. doi:10.1186/1479-5868-10-98
- Education Endowment Foundation. Teaching and Learning Toolkit: Social and emotional learning.
- Liddle & Carter (2015). Educational Psychology in Practice, 31(2), 174–185. doi:10.1080/02667363.2015.1008409
- Rodriguez-Ayllon et al. (2019). Sports Medicine. doi:10.1007/s40279-019-01099-5
- Singh et al. (2026). Journal of the American Academy of Child & Adolescent Psychiatry, 65(2), 171–186. doi:10.1016/j.jaac.2025.04.007
- Vella et al. (2017). Medicine & Science in Sports & Exercise, 49(4), 687–694. doi:10.1249/MSS.0000000000001142
- Youth Endowment Fund. Toolkit: Sports programmes.
Further Evidence
Each topic pairs the wider research with what we saw in the pilot. Click any title to expand.
Activity & Wellbeing in Childhood
A review of physical activity interventions found effects on children’s positive mental health of around g = 0.4, with larger but more variable effects on wellbeing and resilience (Andermo et al., 2020). Sport England’s Active Lives survey (2024–25) finds that only 49% of children and young people meet the activity guideline, and that more active children report higher happiness.
How Bloomsbury Compares
The pilot difference (d = 0.36) is similar in size to trial-based effects in the literature. Our design is weaker than a trial, so we read this as a sign that the full study is worth running.
Trusted Adults & Competence
The Youth Endowment Fund identifies the relationship between a young person and their coach as a likely driver of impact in sports programmes. A systematic review of sport in childhood found participation associated with better self-esteem and social skills, with stronger associations for team sports than individual activities (Eime et al., 2013).
How Bloomsbury Compares
Our largest difference was on “I feel that I am good at some things” (d = 0.86), with smaller differences in emotional state. That pattern matches what the literature would predict for a coach-led team sport.