The role of primary school composition in affective decision-making: a prospective cohort study

Table 1 summarises the sample characteristics and proportion of missing data in the sample. The analytic sample comprised 3,141 children, 1,545 of whom were males (49%) and 1,596 females (51%). The majority of children were white (N = 2,563; 82%), lived with both biological parents (N = 2,442; 78%), resided in urban areas (N = 2,573; 82%) and did not have SEN (N = 3,325; 74%). Moreover, 19% (N = 559) had mothers with a university degree and 12% (N = 346) were eligible for FSM. By age 11 years, 70% were showing signs of puberty. In terms of sex differences, males scored significantly higher in delay aversion and risk-taking, had more externalising problems and were more likely to have SEN (all p values < 0.01). Females scored significantly higher in deliberation time, had higher KS1 scores, and were more likely to be showing signs of puberty by age 11 years (all p values < 0.01). No other comparisons between sexes yielded statistically significant results.

Table 2 summarises the bivariate associations between the primary school characteristics in the analytic sample. Almost all correlations were statistically significant and reflected an ecology of disadvantage. The strongest (negative) associations were found between school’s average KS1 scores and proportion of children with FSM eligibility (rho = –0.65, p < 0.01), and between school’s average KS1 scores and proportion of children with SEN (rho = –0.52, p < 0.01). There was a strong positive association of the proportion of children with FSM eligibility with the proportion of children with SEN (rho = 0.57, p < 0.01).

Tables 3 and 4 summarise the results of the multiple linear regression models performed to assess the role of school characteristics for decision-making in males and females, respectively. In males (Table 3), after adjustments for confounding, school’s proportion of native English speakers was significantly and positively associated with longer deliberation time (b = 0.09, p < 0.05); school’s higher proportion of children with FSM eligibility was significantly associated with worse quality of decision-making (b = –0.01, p < 0.05); school’s proportion of children with SEN was positively associated with better quality of decision-making (b = 0.01, p < 0.05); and school’s average academic performance was negatively associated with delay aversion (b = –0.01, p < 0.05). Of the covariates considered, increased levels of externalising problems were associated with increased delay aversion (b = 0.01, p < 0.01); own FSM eligibility with poorer quality of decision-making (b = –0.05, p < 0.05); and own academic performance with decreased delay aversion (b = –0.01, p < 0.01), shorter deliberation time (b = –0.06, p < 0.01), better quality of decision-making (b = 0.01, p < 0.01) and greater risk adjustment (b = 0.06, p < 0.01).

In females (Table 4), after adjustment for confounding, the school’s proportion of children with SEN was associated negatively with risk-taking (b = –0.01, p < 0.05), while the school’s average KS1 score was positively associated with deliberation time (b = 0.06, p = 0.02) and negatively with risk-taking (b = –0.01, p < 0.05). Of the covariates considered, higher maternal education was associated with improved quality of decision-making (b = 0.04, p < 0.05) and greater risk adjustment (b = 0.16, p < 0.05). Own academic performance was associated negatively with deliberation time (b = –0.07, p < 0.01) and positively with quality of decision-making (b = 0.01, p < 0.01) and risk adjustment (b = 0.04, p < 0.01). Finally, own FSM eligibility was associated with less risk-taking (b = –0.03, p < 0.05). Mean VIF values in Models B in males and females were 1.43 and 1.40, respectively, indicating that none of the predictors were highly collinear (highest VIF value was observed for school’s proportion of FSM eligible among females, VIF = 2.37).¹

The findings of this study suggest that certain primary school characteristics are associated with individual children’s decision-making. In partial agreement with our hypotheses we found that after adjustment for individual and family-level confounding, those schools with a higher average academic performance showed more delay aversion among individual males, and higher deliberation time and lower risk-taking among individual females. In schools with a higher proportion of native English speakers individual males had higher deliberation time. In schools with a higher proportion of pupils eligible for FSMs individual males showed lower scores on quality of decision-making. Finally, in schools with higher proportions of children with special educational needs individual males showed better quality of decision-making and individual females showed lower risk-taking. Prior to examining these results more closely we note that they are associative and cannot be used to infer causal links between a school’s compositional characteristics and a pupil’s affective decision-making. Future studies aiming to test causal links should attempt to control for pre-existing differences in decision-making and to use mixed-methods approaches to identify additional individual and school-level factors which might be driving the significant associations that we found in this study.

Turning to the results of our study, the fact that the proportion of native English speakers in the school is positively associated with individual pupils’ deliberation time was rather unexpected given the positive impact that interacting with peers of other ethnicities can have on children’s cognitive maturity [18]. Speed of information processing, largely captured by deliberation time on the CGT, is associated with activations of brain areas similar to the ones observed for inhibitory control, a cognitive mechanism that serves to block affective impulses and, therefore, enables deliberative decision-making even in affect-charged situations [50‐52]. Howbeit, a school’s proportion of pupils with English as an additional language (EAL) cannot necessarily be considered a good proxy of the ethnic diversity of the student body. Moreover, as discussed in greater detail below, we could not control for native English speaker status at the individual level since the MCS recruited children born in the UK and such analyses would be severely underpowered. This finding should be interpreted with this caveat in mind. Our results also suggest that male pupils in schools with higher average SES show better quality of decision-making, independently of their own SES. Importantly these associations survived adjustments for individual and family-level confounding known to be associated with school selection and decision-making. If these associations were causal, they suggest that interventions aiming to support lower-achieving schools and those with less affluent intakes could help to improve all boys’ decision-making. For females, attending a higher performing school was related to less delay aversion and less risk-taking but also longer deliberation time, independently of their own academic performance. Previous evidence also suggests advantageous effects of such schools on behavioural adjustment and academic performance [4, 53], both of which are closely linked with reduced delay aversion but also quality of decision-making [7, 24]. Regarding the positive association found between attending a higher performing school and one’s deliberation time, we believe that this finding might be an artefact of higher rates of females in schools with a higher average academic performance. Girls have been shown to have longer deliberation times than boys at age 11 years [7] and a disadvantage in processing speed (reaction time) tasks including motor responses, e.g. finger tapping [54]. At the same time, they outperform boys in school [55]. Taken together, this line of evidence suggests that it is likely that better performing schools have a higher proportion of female pupils which might explain the positive association between average school performance and deliberation time among individual female pupils. In the absence of sex ratio data for our schools, we could not test this hypothesis explicitly.

An additional surprising finding of the study was the association between the proportion of pupils in the school with SEN and individual females’ reduced risk-taking and individual males’ better quality of decision-making. An explanation may be that inclusive schools with higher proportions of pupils with SEN tend to be more innovative in their teaching approaches and thus benefit all children. For example, schools with more pupils with SEN which are unable to offer individual support might adopt more creative ways of engaging all children in learning, thus benefitting all pupils’ quality of decision-making either directly or indirectly by enhancing the school climate and cultivating an inclusive school culture [56]. Although we could not determine how inclusive the MCS schools with a higher proportions of children with SEN were, we think that these findings provide support for the benefits of inclusive learning. This is particularly relevant in light of evidence suggesting that pupils with SEN in English schools are disproportionately excluded [57]. At the same time, there is evidence suggesting that all children who attend highly inclusive schools perform better academically and interact more positively with peers compared with pupils educated in low inclusive settings [58]. Our study supports this notion that diversion should be at the heart of inclusion since inclusive learning environments might also benefit aspects of learners’ decision-making.

Our study, however, also has several limitations, as already outlined, and the results should be interpreted with these caveats in mind. First, in the absence of measurements for pre-existing differences in decision-making our results are associative and cannot be used to infer causal links between the school’s compositional characteristics and an individual pupil’s affective decision-making. Nonetheless, by controlling for emotional and behavioural problems and academic performance (at age 7) which are correlated with decision-making [3, 7] it is likely that prior differences were, at least partially, taken into account. Second, not all characteristics measured at the school level were also captured at the individual level in our analyses. For example, EAL at the individual level was excluded due to the small number of non-native English speakers in the MCS given that the cohort had all been living in the UK since infancy. Third, school characteristics were measured in NPD in 2008 and were considered in our study to be invariant until 2011/12 when the MCS children took the CGT. Since we excluded from the analytic sample those children who changed school between the two assessments this assumption is likely to hold to a large extent. Fourth, the analytic sample (N = 3,141) consisted of 77% of the total sample of children eligible for our study with available data from the January 2008 census and the outcomes considered (N = 4,076). However, we attempted to eliminate biases resulting from attrition and non-response using study-specific weights which can account for the disproportionate attrition of participants in MCS. Fifth, the adequacy of FSM eligibility as a proxy of SES has been questioned in the available literature [59], however, there is also evidence that FSM eligibility comes very close to identifying a disadvantaged group of children [60]. Sixth, the variable used to capture SEN at the school-level excluded children with formal statements but that was not the case for the variable for SEN at the individual-level. Since only 46 children in the analytic sample (1%) had a statement according to the teacher survey this discrepancy is highly unlikely to have influenced the results. Seventh, due to small degree of clustering of pupils within school in our analytic sample we could not take the hierarchical nature of the data into account by modelling children as being nested within schools by means of multilevel models. However, by considering simultaneously characteristics at both the individual and school levels, we avoided committing the ecological fallacy, whereby inference occurs at the group level (school in this case), but is actually attributable to confounding by individual factors [61]. Nonetheless, it is strongly recommended that future studies utilise datasets specifically designed to model school effects by providing an adequate degree for clustering of pupils within schools. Finally, we only found few significant associations between school characteristics and the decision-making outcomes considered, and effect sizes were rather small. Nevertheless, such associations had not been investigated before, and one would not expect them to dominate individual/family-level effects. They may be weak, but they do exist.

The findings of this study suggest that primary school composition is associated with important aspects of individual children’s decision-making. Pupils in less socioeconomically disadvantaged schools and those in better performing schools were making, on average, better and less impulsive choices, even after controlling for their own level of socioeconomic disadvantage and academic performance. The findings of the study suggest that interventions aiming to support lower-achieving schools and those with less affluent intakes could help to improve decision-making, but mainly in boys.