Adult numeracy randomised controlled trials: Family numeracy

This report presents findings from a pilot randomised controlled trial (RCT) and Implementation and Process Evaluation (IPE) of a Family Numeracy (FN) intervention. The study was funded by the Department for Education (DfE) as part of a programme of Adult Numeracy Trials delivered through the Multiply programme. The trial compared the intervention to Business-As-Usual (BAU), for which individuals were put on a waitlist for the FN intervention. BAU may, or may not, have included activities by teachers, providers and local authorities seeking to encourage learners to enrol onto Further Education (FE) courses. The primary outcome for the pilot RCT was progression of parents/carers of Key Stage 1 Year 2 children to further maths courses. Secondary outcomes included parents’/carers’ confidence in maths, progression to other courses (including qualification-bearing maths courses at any level, Level 2 qualification-bearing maths courses, Level 2 qualification-bearing courses of any type, and any further learning) and child’s maths attainment. The IPE explored compliance, attrition, fidelity, engagement, feasibility, perceived outcomes, trial readiness and barriers and enablers to delivery. As a pilot, this study provided a small-scale investigation into the feasibility of implementing family learning trials and measuring relevant outcomes.

Key findings

A total of 22 treatment group parents/carers (10%) progressed to a further maths course, compared to none in the control group. This translates to an estimated 18 percentage point higher likelihood of progression attributable to the treatment, after controlling for demographic characteristics, provider effects and school-level factors. The pilot trial found that this effect was statistically significant at the 99% confidence level[1] on parent/carer progression to further maths courses (the primary outcome). The effect size was relatively large compared to other education trials, which may reflect the low probability of parent/carer progression to further maths courses without an intervention.

This effect was stronger among the 34% parents/carers who completed the minimum dosage of at least 3 sessions and 3 homework sessions which was defined as the level of attendance and homework that the product developer considered necessary to achieve parent/carer outcomes. The treatment effect was higher by 45 percentage points and statistically significant (p<0.001) for parents/carers who complied with the minimum dosage. This means that parents/carers who fully engaged with the intervention showed notably higher outcomes than those who did not.

Outcomes varied significantly across providers, highlighting their role in shaping delivery and learner experience. During interviews, FN tutors reported that they had signposted parents/carers to other courses offered by their organisations, with many parents/carers expressing interest in further learning opportunities. The correlation between impact on progression among parents/carers participating in FN in the same school was also notable, suggesting that school-level factors influenced outcomes to a certain extent.

There were 18 parents/carers who signed up to the trial post-randomisation. These individuals were potentially more highly motivated or had different characteristics from the randomised sample. However, adjusted estimates from sensitivity analysis showed that the impact on parent/carer progression to further maths courses (primary outcome) was robust and not driven by this group. There was no statistically significant impact on any of the secondary outcomes (parent/carer progression into any further course, parent/carer progression into any qualification bearing course, parent/carer progression into any qualification bearing maths course, parent/carer progression into any L2 qualification bearing maths course, parent/carer confidence in maths, child’s maths attainment).

The FN intervention was delivered in line with product developer guidance (i.e. with fidelity) and needed only minor adjustments to session length and content to meet the needs of parents/carers. It proved feasible for delivery within schools by adult education providers, and parents/carers found the sessions and support to be readily accessible.

Tutors attended all or most of the pre-delivery training (known as compliance) and found it to be accessible and efficient. They reported increased knowledge and confidence in delivering family learning courses and improved understanding of the Key Stage 1 maths curriculum. The product developers specified that parents/carers would need to attend at least 50% of the sessions and complete at least 50% of the home learning tasks for the intervention to be effective. Just over a third achieved this level of compliance, with others affected by obstacles in their personal circumstances, such as childcare or other caring responsibilities.

Parents/carers reported more positive attitudes towards maths and felt better equipped to support their children’s learning, both of which motivated them to consider enrolling in further maths courses. Treatment group engagement increased between baseline and endline. Additionally, school staff observed improvements in parent/carer engagement with both children’s learning and the school more broadly.

Overall, the intervention worked well within school settings and was delivered with high fidelity. Data collection proceeded as planned, demonstrating the feasibility of scaling up to a full trial. Recommendations for future delivery include using clearer, simpler language in recruitment materials to improve parent/carer take-up and allowing more time for recruitment.

[1] A statistically significant effect falls outside the margin of error and is therefore unlikely due to chance alone. Results at the 99% confidence level provide stronger evidence than the standard 95% level as they meet more stringent criteria. At 99% confidence, researchers can be 99% certain the findings reflect a true effect, with only a 1% chance the observed difference is due to random variation. At the 95% confidence level, this probability increases to 5%.