boarding school research paper

ORIGINAL RESEARCH article

Boarding and day school students: a large-scale multilevel investigation of academic outcomes among students and classrooms.

• 1 School of Education/School of Psychology, University of New South Wales, Sydney, NSW, Australia
• 2 Department of Educational Studies, Macquarie University, Macquarie Park, NSW, Australia
• 3 The Future Project, The King’s School, Sydney, NSW, Australia

Boarding school is a major educational option for many students (e.g., students living in remote areas, or whose parents are working interstate or overseas, etc.). This study explored the motivation, engagement, and achievement of boarding and day students who are educated in the same classrooms and receive the same syllabus and instruction from the same teachers (thus a powerful research design to enable unique comparisons). Among 2,803 students (boarding n = 481; day n = 2,322) from 6 Australian high schools and controlling for background attributes and personality, we found predominant parity between boarding and day students in their motivation, engagement, and achievement. We also found that classroom-average motivation, engagement, and achievement was not significantly affected by the number of boarders (relative to day students) in the classroom. In addition, the effects of boarding were generally not moderated by students’ background or personality attributes. We conclude that boarders have academic opportunities and outcomes that are comparable to their day student counterparts. Implications for students, teachers, and parents are discussed.

Introduction

Boarding schools 1 constitute a major mode of education in many countries. For example, in Australia (the site of the present study) there are an estimated 170 schools with boarding students, and 470 schools in the United Kingdom and 340 schools in North America that accommodate boarding students ( Martin et al., 2014 ). There has been a growing body of research into boarding school, particularly in Australia (the site of the present study). This research has been quantitative and qualitative and contributed to increasing understanding of boarders, their academic and social-emotional wellbeing outcomes, and the factors contributing to these. Research in this area is important because boarding (and other residential education settings) is often a necessary educational pathway for many students for a variety of reasons (e.g., living in remote areas, parents working overseas, choosing education outside home country, etc.). Indeed, investigating boarding school effects involves quite a unique research design in that boarding and day students are educated in the same classrooms, taught by the same teachers, and receive the same instruction and syllabus. Thus, boarders may be considered something of a “treatment” group and day students something of a “comparison” group, with most curricular classroom and instructional features held constant.

In numerous ways, the present study adds to research findings about boarding school students. First, it explores in a large-scale sample, the role of boarding in students’ domain-general academic motivation and engagement (i.e., motivation and engagement in school generally). Second, it extends the domain-general motivation and engagement research by also investigating the role of boarding in students’ domain-specific (science) motivation, engagement, and achievement. Third, it augments prior multilevel research (that focused on students nested within boarding houses and schools; Martin et al., 2016 ) by conducting multilevel research investigating student- and classroom-level effects of boarding status on academic outcomes—e.g., whether the number of boarders relative to the number of day students in a class affects classroom-average motivation, engagement, and achievement.

Figure 1 presents the multilevel model we apply to address these three issues. At Level 1 of this figure are the student-level associations to be tested. Here boarding status (no/yes; or, day/boarding) predicts science motivation, engagement, and achievement and also predicts domain-general motivation and engagement. Importantly, boarding status is a predictor of these outcomes alongside students’ background attributes (e.g., age, gender, Indigenous status, etc.) and their personality in order to ascertain the role of boarding beyond the role of background attributes and personality. At Level 1 also, interaction effects are tested that explore whether boarding status effects vary as a function of background attributes and personality (e.g., whether boarding status effects vary as a function of different age groups). Level 2 explores boarding effects on science motivation, engagement, and achievement at the classroom level—that is, whether the proportion of boarders in a science classroom predicts class-average science outcomes. Importantly, multilevel modeling disentangles Level 2 from Level 1 effects; thus, Level 2 findings shed light on class-average effects beyond individual student effects.

Figure 1. Multilevel path model to be tested.

Theoretical Perspectives

There are numerous theoretical frameworks that can inform thinking about the effects of boarding. Relevant to this study’s substantive foci are ecological systems, human capital, critical race, social identity, and extracurricular activity theories.

Ecological Systems, Human Capital, and Critical Race Frameworks

Ecological systems theory emphasizes the ongoing person-environment interactions that shape human development ( Bronfenbrenner, 2001 , 2005 ). Under this theory:

… human development takes place through processes of progressively more complex reciprocal interaction between an active, evolving biopsychological human organism and the persons, objects, and symbols in its immediate external environment. To be effective, the interaction must occur on a fairly regular basis over extended periods of time ( Bronfenbrenner and Morris, 1998 , p. 996).

We contend that boarding represents a somewhat intensive and on-going process of interactions between student and environment—interactions that have potential to shape students’ academic outcomes ( Martin et al., 2014 , 2016 ). In fact, given the salient contextual aspects of boarding, it is perhaps not surprising that Bronfenbrenner (1970) conducted one of the earliest formal investigations of boarding effects. According to Bronfenbrenner, because the boarding context plays a different role in shaping children’s academic development, it is conceivable boarding students’ academic outcomes may differ from those of day students.

Bass (2013) draws on Bourdieu’s (1986) ideas around social and educational capital to explore the potential that boarding may (or may not) hold for improving life chances for disadvantaged youth through opportunities for meeting their social and educational needs. At the same time, however, these capital theories and their positive contentions do not always apply to some groups of boarding students, e.g., due to a lack of supporting data ( Guenther and Fogarty, 2018 ). Also, human capital theory has been connected with other pertinent theories such as critical race theory ( Aleman, 2013 ) that might suggest potentially challenging perspectives on boarding effects, particularly for some student groups. For example, critical race theory has been applied as a lens to understand boarding school for Indigenous students ( Benveniste et al., 2019 ). For these students, boarding school may reproduce dominant cultural values through their daily practices, policies and procedures that are not appropriately sensitive or supportive of Indigenous students and their cultural and social-emotional needs.

Social Identity Theory

Social identity theory is also relevant. Individuals’ self-concepts are based on their membership to their social group ( Tajfel, 1978 ; Tajfel and Turner, 1986 ). Social identities are most influential when the individual has strong emotional connections to a group and when membership in a particular group is considered by the individual to be central to their self-concept. The individual garners self-esteem through affiliation with the group, typically through influential processes such as within-group assimilation (pressure to conform to the group’s norms) and intergroup bias (favorably appraising one’s own group relative to other groups). These processes are particularly powerful in peer groups ( Leaper, 2011 ). This being the case, there have been applications of social identity theory to the educational context. Mavor et al. (2017) , for example, have described how the “self” is not a fixed entity among students, but amenable to variation as a function of change in experience, including formal and informal learning at school. These ideas are particularly relevant when considering students who experience a boarding context for their residential experience and who are taught within specific classrooms for their educational experience. The present study and its multilevel design are ideally placed to investigate these processes in terms of the development of boarding and day students’ academic outcomes at school generally and also in science classrooms.

Extracurricular Perspectives

Extracurricular activity is any out-of-class involvement that absorbs students’ energy, time, and attention ( Marsh and Kleitman, 2002 ), and as such, boarding can be considered a form of extra-curricular activity. The “identification/commitment” model of extracurricular activity ( Marsh, 1992 ) proposes that school-based extracurricular activities can “improve school identification, involvement, and commitment in a way that enhances narrowly defined academic outcomes” ( Marsh and Kleitman, 2002 , p. 471). It has been found that school-based extracurricular activities are more likely to increase students’ affiliation with the school ( Fredricks and Eccles, 2005 ). Following from this, Martin et al. (2014) proposed that context-specific affiliation (e.g., school affiliation) boosts students’ identification with and commitment to that context, resulting in positive academic outcomes. They further proposed that boarding may afford greater student activity at and with the school—indeed, being resident at school may also involve a greater requirement or expectation to be involved in extra-curricular activities. Thus, it is possible that one’s boarding status is linked with adaptive academic outcomes, consistent with what might be hypothesized for school-based extracurricular activity under the “identification/commitment” perspective ( Marsh and Kleitman, 2002 ). However, the counterpoint to this is that time spent in one activity comes at the expense of potential development in other parts of life ( Marsh and Kleitman, 2002 ); for example, boarding may deprive students of necessary development opportunities (such as what they might gain at home), and potentially have negative effects.

Research Relevant to the Boarding Experience

To date, research into boarding has revealed a somewhat mixed body of results, finding positive, negative, or generally null (or equivocal) effects in boarders’ academic and social and emotional outcomes. It is also the case there are different student groups for whom boarding is a more salient educational option and research has identified effects particular to these students as well. This research is briefly reviewed.

Positive and Negative Effects of Boarding

The Association of Boarding Schools ( The Association of Boarding Schools (TABS), 2003 , 2013 ) compared the experiences of U.S boarding students and day students. Findings showed that boarding students were more likely to report they were satisfied with their academic experience and were more likely to report that school prepared them for college. In a qualitative investigation, White’s (2004) study of Anglo-Australian and overseas students suggested that boarding instills independence and acceptance of cultural diversity. Also, in qualitative work, Bass (2013) found that boarding for disadvantaged students enhanced their exposure to social and educational capital. An Australian study by Martin et al. (2014) found predominant parity between boarding and day students (described below), but where small effects were identified, they favored boarders. These studies thus suggest potentially positive outcomes for boarding students.

There is also research documenting negative effects of boarding for some children. Lester and Mander (2020) investigated mental health and wellbeing among high school boarders (boys) as they transitioned to and into boarding school. They found increases in emotional problems among boarding students over time. They also found that academic motivation declined over time; however, this was the case for both boarders and day students. In longitudinal research, Mander and Lester (2017) found that boarding and day students reported increases in depression, anxiety, and emotional symptoms between Grades 7 and 9, but that boarding students reported higher levels of anxiety and stress than day students at the end of Grade 8. It has also been documented how some boarding schools are contexts perpetuating institutional and societal power structures and problematic ideologies—such as those around gender ( Khan, 2010 ; Finn, 2012 ; also see Duffell, 2000 ; Schaverien, 2011 for other analyses of negative boarding effects).

Null Boarding Effects

There is also research showing there is not a major difference in educational outcomes when comparing boarding and day school students. As noted above, Martin et al. (2014) conducted a large-scale Australian study and found relatively few differences (with small effect sizes) in academic wellbeing (e.g., domain-general academic motivation and engagement) and personal wellbeing (e.g., peer relations, mental health, etc.) when comparing boarding and day students in the same school. In a similar vein, in a longitudinal study of students transitioning from day to boarding status, Downs (2002) found no major changes in self-concept through this transition. Behaghel et al. (2017) found that disadvantaged students in boarding initially experienced low levels of wellbeing, but their wellbeing adjusted during their boarding experience. They also found boarders experienced academic gains 2 years after commencing boarding, but this effect did not generalize across students (it was stronger for students higher in initial academic ability).

Insights From Particular Student Groups

It is also the case that particular student groups have a more long-established history of attending boarding school. On the international stage, overseas students are one such group (usually because their parents are working in another country). In the Australian context (the site of the present study), boarding has been a major educational pathway for Indigenous students, with most research identifying mixed yields in the boarding school experience for these students. For example, in a study of Indigenous girls in a residential college it was found they enjoyed their residential experience and the new friendships developed, but also found that homesickness and lifestyle restrictions were a challenge for the girls ( English and Guerin, 2017 ). These findings were similar to a study by MacDonald et al.(2018 ; see also Guerin and Pertl, 2017 ) where school leaders and Indigenous students reported that boarding allowed enhanced career opportunities and health outcomes, but that there were issues to navigate to attain these outcomes such as homesickness, racism, and post-school transition difficulties. Guenther and Fogarty(2018 ; see also Guenther et al., 2020 ) identified the positive possibilities of boarding school for Indigenous students, but also noted the evidence does not always support the positive potential. They suggested that when interpreting Indigenous students’ development through cultural and human capital lenses, there emerge potential problems and difficulties in boarding for Indigenous students that have significant implications for educational policy. Indeed, quantitative research among high school Indigenous boarders supports this, finding lower scores on resilience and psycho-social wellbeing. Also, when these students transitioned back to their community, they reported less connectedness with family and community and even lower levels of resilience and psycho-social wellbeing ( Redman-MacLaren et al., 2019 ).

Summary and Focus of This Study

Taken together, it is evident the diversity of research methodologies that have examined the experiences and outcomes of boarding, has yielded varied findings. Each has informed a distinct aspect of the boarding phenomenon, both positive and negative. The present study adds to what is known by addressing two novel dimensions in this space. First, given that boarding students are typically taught in the same classes as day students, what is the effect of the relative proportion of boarders in a class on class-average academic outcomes? For example, does the presence of relatively more (or fewer) boarders in a class affect class-average outcomes? Second, prior research has investigated the effects of boarding on domain-general academic outcomes (e.g., motivation in school generally), but we do not know if such findings generalize to specific school subjects. We therefore investigate the effects of boarding on domain-general (in school, generally) and domain-specific (in science) academic outcomes. Figure 1 shows the multilevel processes we investigate (described above).

Domain-General and Domain-Specific Outcomes, Background Attributes, and Multilevel Considerations

Target domains and outcomes under focus.

As noted, we focus on domain-general (i.e., in school, generally) academic outcomes and domain-specific academic outcomes. Our domain-specific focus is science—specifically motivation, engagement, and achievement in science. Exploring these issues in science is somewhat topical because there are concerning trends in science achievement and science pathways (especially among “Western” nations). In Australia, for example, achievement in science declined in the 2015 Trends in International Mathematics and Science Study (TIMSS; Thomson et al., 2016 ). In the Programme for International Student Assessment (PISA), the long-term change in Australia’s mean performance in science over the period of its participation demonstrates one of the largest decreases among PISA-participating countries ( OECD, 2020 ). Also, science participation and enrollments among senior school students demonstrate long-term decline ( Office of the Chief Scientist, 2014 ) and there is concern about students’ declining interest in science in high school ( Tröbst et al., 2016 ). Thus, motivation, engagement, and achievement have been identified as outcome targets for improvement in science and there have been recommendations for researchers to explore factors that may be implicated in these outcomes ( Ross and Poronnik, 2013 ; Abraham and Barker, 2015 ). Our study therefore investigates the role of educational context (specifically, boarding vs. day status) as one potential factor. Importantly, to ascertain if potential boarding status effects are distinct to science or not, we also assess the role of boarding status in motivation and engagement for school in general. In operationalizing motivation and engagement as “outcomes” in this study, we do recognize that they can also be considered as “input” or predictor factors for achievement and other academic outcomes. We herein position them as outcomes because it is more feasible that boarding status and background attributes such as gender, SES, etc. predict motivation and engagement, than vice versa. Thus, motivation and engagement can be either a means to desirable outcomes, or desirable outcomes in their own right—and it is the conceptualizing, research questions, and research design that determine where in the educational process they are modeled ( Marsh and Martin, 2011 ; Martin, 2012 )—viz. “outcomes” in the case of the present study.

Because we seek to systematically build on the recent large-scale quantitative study by Martin et al. (2014) , we adopt the main motivation and engagement measures employed by them; namely, positive motivation (e.g., self-efficacy), positive engagement (e.g., persistence), negative motivation (e.g., anxiety), and negative engagement (e.g., self-handicapping). These are all operationalized through the Motivation and Engagement Scale that has domain-general ( Martin, 2007 ) and domain-specific (including in science; Green et al., 2007 ) forms. For achievement, we administer an in-survey science test that assesses students on the extent to which they have acquired core information from the state science syllabus.

Background Attributes Important to Consider

It is possible that boarding status may be associated with various student background attributes that are also linked with motivation, engagement, and achievement. To understand the unique effects of boarding, it is thus important to include such attributes in modeling in order to partial out their potential influence. Martin et al. (2014) identified numerous such factors, including age, gender, socio-economic status, language background, Indigenous status, parent education, prior achievement, and personality. For example, they found that alongside boarding status, parents’ education, prior achievement, conscientiousness, agreeableness, and openness all positively predicted positive motivation—while prior achievement, conscientiousness, and agreeableness negatively predicted negative motivation (and neuroticism positively predicted negative motivation). Furthermore, if boarding represents a distinct educational ecology and socializing environment ( Bronfenbrenner, 1970 ; Holden et al., 2010 ), then time spent in that environment (i.e., years in boarding school) may affect one’s identification with and internalization of that environment, including academic effects of the experience. Thus, background attributes do explain variance in this study’s academic outcomes beyond the effects of boarding status. Accordingly, alongside the predictive role of boarding status, these background attributes are also included in the present study as predictors of motivation, engagement, and achievement (i.e., shared variance is controlled for; see Figure 1 ).

Furthermore, according to Martin et al. (2014) , it is also possible that background attributes may moderate the effects of boarding. For instance, perhaps boarding effects vary as a function of students’ age, Indigenous status, personality, etc. In Australia, boarding is identified as one means by which distant students (e.g., Indigenous, rural, or remote) can access education (e.g., Curto and Fryer, 2011 ; MacDonald et al., 2018 ; Osborne et al., 2019 ; Guenther et al., 2020 ). Also, we earlier identified research revealing a negative history of boarding school for some students and in part this has been attributed to the commencement of boarding at a young age ( Duffell, 2000 ). Although our study is conducted in high schools, we can test if age moderates the effects of boarding on academic outcomes. Or, it may be that the somewhat social nature of residential education is better suited to students high in extraversion. Thus, we include interaction terms (e.g., boarding status × Indigenous status, boarding status × age, boarding status × extraversion, etc.) to test for the potential moderating role of the study’s background attributes (see Figure 1 ).

Multilevel Considerations

There is widespread recognition of how important it is to analyze hierarchical data in appropriate ways ( Marsh et al., 2012 ). In our study we have students nested within classrooms and therefore conduct multilevel modeling to account for this and to understand variance attributable to student- and classroom-levels. There are known statistical biases associated with single-level research designs (e.g., dependencies within groups; confounding of within- and between-group variables) and multilevel approaches aim to resolve these biases (for discussions see Goldstein, 2003 ; Marsh et al., 2008 ; Raudenbush and Bryk, 2002 ). To our knowledge only one study has investigated boarding from a multilevel perspective— Martin et al. (2016) investigated motivation and social climates among students nested within boarding houses that were nested within schools (thus, student-, house-, and school-level effects).

Our study differs from the Martin et al. (2016) work by exploring student- and classroom-level effects. Specifically, we investigate whether the proportion of boarding students (relative to day students) in a class has a significant bearing on class-average motivation, engagement, and achievement. Multilevel modelers have established the reciprocity of individual and group dynamics: individuals can affect the group to which they belong and in turn the group can affect these individuals ( Raudenbush and Bryk, 2002 ; Goldstein, 2003 ; Marsh et al., 2008 ). This raises the question as to whether a critical mass of boarding students in a classroom affects overall class-average outcomes. For example, does the distinct socialization experience of boarding ( Holden et al., 2010 ) lead to a cohesion or collective identity among boarders in a classroom such that they evince distinct effects relative to day student counterparts in the same classroom? By capturing data on science motivation, engagement, and achievement in science classrooms, our research design could address this question.

When conducting multilevel modeling it is also important to establish whether climate or context effects are being investigated. Climate refers to shared perception of a characteristic of the group (e.g., classroom) that is common to members (e.g., students) in that group. For climate variables, the group referent is usually explicit in the item, indicator, or question (e.g., “… students in this classroom try hard”; Marsh et al., 2012 ). However, when the item referent is the individual (e.g., “I try hard”) and the item is aggregated “up” to also create a classroom-level variable, it is known as a context effect ( Marsh et al., 2012 ). As is evident in Materials below, in our study all variables at student- and classroom-level are context factors.

Aims of the Present Study

There were three main aims of the present investigation. The first aim was to explore, in a large-scale sample, the role of boarding in students’ domain-general academic motivation and engagement. The second aim was to also explore the role of boarding in students’ domain-specific (science) motivation, engagement, and achievement. The third aim was to investigate the association between the proportion of boarders in a classroom (relative to day students) and classroom-average motivation, engagement, and achievement—beyond the student-level motivation, engagement, and achievement relevant to the first two aims. Figure 1 presents the multilevel path model addressing these three aims.

Participants were 2,803 high school students from 6 Australian schools that comprised both boarding and day students. Students were surveyed in 224 science classrooms (mean class size = 11.68 students; sufficient to estimate classroom effects and not unduly disproportionate to the staff-to-student ratio for high schools in the independent school sector, taking into account non-teaching staff numbers, non-participation, student absences, and any students not receiving parental participation consent; Australian Bureau of Statistics, 2019 ). Seventeen percent ( n = 481) of students were boarders; 83% ( n = 2,322) were day students. Thirty-five percent of boarders had been boarding for less than 1 year, 31% for 1–2 years, and 34% for 3 years and over.

All schools were independent schools (i.e., not government or systemic) and located in Sydney, New South Wales (Australia’s most populous state). The average school size was 1,801 total students enrolled. Regarding the socio-demographics of the school, in 2018 (the year data were collected), 23% of the students enrolled within the 6 schools spoke a language other than English at home and 1% of students enrolled within the 6 schools identified as Aboriginal/Torres Strait Islander ( Australian Curriculum Assessment Reporting Authority (ACARA), 2020a ). For school socio-economic status, the average Index of Community Socio-educational Advantage (ICSEA) score for the schools sampled is 1,145 (compared to the national M = 1,000; Australian Curriculum Assessment Reporting Authority (ACARA), 2020a ). Regarding numeracy achievement in the National Assessment Program—Literacy and Numeracy (NAPLAN), the mean numeracy score of the 6 schools sampled was M = 626 (compared to the national M = 572; Australian Curriculum Assessment Reporting Authority (ACARA), 2020a,b ). Regarding literacy in NAPLAN, the mean literacy score of the 6 schools sampled was M = 593 (compared to the national M = 553; Australian Curriculum Assessment Reporting Authority (ACARA), 2020a,b ). Taken together, these trends indicate that the 6 schools perform above the national average.

Of the 6 schools, 1 school was co-educational, 1 school was a single-sex girls’ school, and 4 schools were single-sex boys’ schools. This being the case, the majority of students were boys (92%). This is disproportionate, but we point out that: (a) multigroup (male vs. female) confirmatory factor analysis of the motivation and engagement measures suggested scalar invariance (the minimum criterion for invariance; Van de Schoot et al., 2012 ) as a function of gender, with no change greater than 0.01 for CFI or greater than 0.015 for RMSEA ( Chen, 2007 ; Cheung and Rensvold, 2002 ), (b) in preliminary analyses (see Table 2 ) there were no gender differences in the proportion of boarders to day students, (c) there were no correlations ( Table 2 ), predictive main effects ( Table 3B ), or moderating effects (viz. boarding/day status × gender) between gender and outcome variables that attained our minimum benchmark for interpretability, (d) as we show below our findings align with those of Martin et al. (2014) whose research design we followed and which comprised relatively equal numbers of boys and girls, and (e) we selected a random sample of 8% boys to match the 8% girls and re-ran the final Step 3 model (see section “Data Analysis,” below), also finding that the only three boarding effects approaching our minimum benchmark for interpretability were the same three boarding effects that approached or attained our minimum benchmark for interpretability in the full sample ( Table 3B ). We thus tentatively conclude that our gender composition did not unreasonably impact factors and empirical associations in this study.

The average age of students was 14.14 years ( SD = 1.29; boarding students M = 14.47, SD = 1.25; day students M = 14.07, SD = 1.29). Eleven percent of the sample were from a non-English speaking background (boarding students 10%; day students 11%). Six percent were Indigenous students (boarding students 9%; day students 5%). Students rated their mother’s and father’s level of education from 1 (“no formal qualifications”) to 6 (“university undergraduate or higher degree”) (sample M = 5.14, SD = 1.28; boarding students M = 4.71, SD = 1.45; day students M = 5.23, SD = 1.22). Students’ socio-economic status (SES) based on the Australian Bureau of Statistics Index of Relative Socio-Economic Advantage and Disadvantage classification (sample M = 1120, SD = 65; boarding students M = 1035, SD = 93; day students M = 1137, SD = 41) was higher than the national average ( M = 1000, SD = 100). As shown in Figure 1 , each of these background factors was included in formal modeling to control for their influence in effects.

Human ethics approval was provided by the lead researcher’s institution. Approval was then received from each school principal agreeing to their school’s participation. Parents/carers and participating students then both provided consent. The online survey of motivation and engagement (as well as a science test) was administered to students during a science lesson in the second term (of four school terms) of 2018. Students were instructed to respond to the survey and test on their own. They were also informed that teachers could provide assistance with any procedural aspects of the process, but that teachers could not help students in answering specific items.

Science Motivation and Engagement

Science motivation and engagement were assessed using the Motivation and Engagement Scale—High School (MES-HS; Martin, 2015 ), adapted to science ( Green et al., 2007 ). Positive motivation in science comprised mastery orientation (e.g., “I feel very pleased with myself when I do well in this science class by working hard”; 4 items), self-efficacy (e.g., “If I try hard, I believe I can do well in this science class”; 4 items), and valuing (e.g., “Learning in this science class is important”; 4 items). Positive science engagement comprised task management (e.g., “When I study for this science class, I usually try to find a place where I can study well”; 4 items), planning behavior (e.g., “I try to plan things out before I start working on homework or assignments for this science class”; 4 items), and persistence (e.g., “If I don’t give up, I believe I can do difficult schoolwork in this science class”; 4 items). Negative science motivation was measured with anxiety (e.g., “When exams and assignments are coming up in this science class, I worry a lot”; 4 items), failure avoidance (e.g., “Often the main reason I work in this science class is because I don’t want to disappoint my parents”; 4 items), and uncertain control (e.g., “I’m often unsure how I can avoid doing poorly in this science class”; 4 items). Negative science engagement was assessed via disengagement (e.g., “I’ve pretty much given up being involved in things in this science class”; 4 items), and self-handicapping (e.g., “I sometimes put assignments and study off until the last moment, so I have an excuse if I don’t do so well in this science class”; 4 items). Students rated items on a scale of 1 (Strongly Disagree) to 7 (Strongly Agree). In previous research, these measures are shown to be normally distributed, reliable, and validated with educational outcomes (for review see Liem and Martin, 2012 ), including in science ( Green et al., 2007 ). Because the science motivation and engagement items in this study were directly relevant to the classrooms in which students were responding to the survey (i.e., their science lesson/class), we modeled the science motivation and engagement factors at Level 1 (L1, student-level) and at Level 2 (L2, class-level).

For this study we focused on the 4 higher order MES factors (positive motivation, negative motivation, positive engagement, negative engagement) that were estimated by (a) aggregating (mean scoring) the items of each first order MES factor (e.g., the 4 items for self-efficacy) to create 11 MES scale scores (e.g., a self-efficacy scale score) and (b) using these scale scores to create an error-adjusted mean score for each of the 4 higher order factors. Error adjusted scores were derived using the following formula: σ h 2 ∗ (1 −ω h ), where σ h 2 is the estimated variance of and ω h is the reliability estimate of the motivation and engagement factor (h) at either L1 (student) or L2 (class; Hayduk, 1987 ; see also Cole and Preacher, 2014 ). Error-adjusted scores were used because they help avoid unreliable standard errors and reduce the risk of inflated parameter estimates ( Cole and Preacher, 2014 ). This yielded standardized loadings as follows: positive science motivation, L1 = 0.96 and L2 = 0.98; positive science engagement, L1 = 0.94 and L2 = 0.95; negative science motivation, L1 = 0.93 and L2 = 0.87; and negative science engagement, L1 = 0.92 and L2 = 0.93. As shown in Table 1 , these factors were normally distributed. Table 1 also shows acceptable reliability ( McNeish, 2018 ) at L1 and L2 for positive motivation (L1ω h = 0.83; L2ω h = 0.98), positive engagement (L1ω h = 0.84; L2ω h = 0.96), negative motivation (L1ω h = 0.69; L2ω h = 0.87), and negative engagement (L1ω h = 0.72; L2ω h = 0.95).

Table 1. Descriptive and measurement properties for outcome variables.

Table 2. Multilevel correlations: students and classrooms.

Domain-General Motivation and Engagement

Domain-general academic motivation and engagement were assessed using the short form of the Motivation and Engagement Scale—High School (MES-HS; Martin, 2015 ). This measures all 11 factors represented in the Motivation and Engagement Scale, but does so via one item per factor. Single-item measures can present issues with reliability; however, because the larger research program from which these data emanate is focused on science, the full MES (44-items; see science motivation and engagement, above) was deemed too long. Therefore, the validated short form ( Martin et al., 2015 ) was used. Also to note is research suggesting single-item scales have merit in cases where long scales are not able to be used (e.g., see Gogol et al., 2014 ).

Positive domain-general academic motivation constituted self-efficacy (“I believe I can do well in my schoolwork”), valuing (“What I learn in my schoolwork is important and useful”), and mastery orientation (“In my schoolwork, I am focused on learning and improving more than competing and being the best”). Positive engagement comprised planning behavior (“I plan out how I will do my schoolwork and study”), task management (“I use my study/homework time well and try to study and do homework under conditions that bring out my best”), and persistence (“I persist at schoolwork even when it is challenging or difficult”). Negative motivation comprised anxiety (“I get quite anxious about schoolwork and tests”), failure avoidance (“I mainly do my schoolwork to avoid failing or disapproval from parents or the teacher/s”), and uncertain control (“I don’t think I have much control over how well I do in my schoolwork”). Finally, negative engagement comprised self-handicapping (“In my schoolwork, I sometimes reduce my chances of doing well [e.g., waste time, disrupt others, procrastinate]”) and disengagement (“I often feel like giving up in my schoolwork”). Students rated items on a scale of 1 (Strongly Disagree) to 7 (Strongly Agree).

Because domain-general motivation and engagement refer to general academics (not just science or science lessons/classes), we modeled it at L1 (student-level), not at L2 (science class-level). As with science motivation and engagement, we focused on the 4 higher order MES factors (positive motivation, negative motivation, positive engagement, negative engagement) that were estimated by aggregating (mean scoring) the items of each higher order MES factor (e.g., the 3 items for positive motivation) to create 4 domain-general motivation and engagement scores and then deriving an error-adjusted mean score for each of these 4 scores. This was done using the same formula as was used for science motivation and engagement (see above; Hayduk, 1987 ; Cole and Preacher, 2014 ). This yielded standardized loadings as follows: positive domain-general motivation, L1 = 0.94; positive domain-general engagement, L1 = 0.94; negative domain-general motivation, L1 = 0.95; and negative domain-general engagement, L1 = 0.91. We found generally acceptable reliability for positive motivation (L1ω h = 0.76), positive engagement (L1ω h = 0.83), negative motivation (L1ω h = 0.61), and negative engagement (L1ω h = 0.57; to note is that this factor comprised only 2 items and fewer items attenuate reliability). Table 1 shows descriptive and reliability statistics for these factors.

Science Achievement

Science achievement was assessed using an online test. It comprised 12 questions developed by the science department head of a large Sydney school. Following preliminary item development, language accessibility was assessed by the languages department head (at the same school). To accommodate the different year-levels of participating students, two forms were developed, one based on the Stage 4 (years 7 and 8) state science syllabus and the other based on the Stage 5 (years 9 and 10) state science syllabus. Questions were set within the contexts of content strands Earth and Space, Physical World, Chemical World, and Living World (NSW Science Syllabus; Nsw Education Standards Authority, 2019 ). Thus, the questions aligned with students’ skill level and what they had been taught—and considered a snapshot of their scientific literacy. The two forms were considered by five experienced science teachers who reviewed each test item in terms of: (a) alignment with the state science syllabus, (b) language and cultural accessibility of item text/graphics, and (c) the envisaged percentage of students likely to correctly answer an item (response options: 25, 50, or 75% of students). All answers were recoded as 0 = incorrect and 1 = correct. The number of correct responses was summed to a total score (as a continuous scale), reflecting something of a formative construct, not a latent construct. Scores were standardized by year level ( M = 0; SD = 1) so that students’ achievement scores were appropriately adjusted for different levels of science education and experience and for the fact two tests were administered (one test for years 7 and 8, raw M = 55 and 60%, respectively, one test for years 9 and 10, raw M = 52 and 57%, respectively). The science achievement factor was approximately normally distributed ( Table 1 ), with acceptable reliability at L1 and L2 (L1ω h = 0.69; L2ω h = 0.98).

Background Attributes

Numerous background attributes were used as covariates and also as potential moderators of boarding effects. Participants reported their boarding status (0 = day student; 1 = boarding student), prior achievement (relative year-group standing on science tests and assignments; 1 = “in the lower third of my year group,” 2 = “in the middle third of my year group,” 3 = “in the upper third of my year group”), age (a continuous measure), gender (0 = male, 1 = female), language background (0 = English speaking, 1 = non-English speaking), Indigenous status (0 = non-Indigenous; 1 = Indigenous), parents’ education (scale from 1 “No formal qualifications” to 6 “university undergraduate or higher degree”), and SES based on home postcode which was then matched to Australian Bureau of Statistics SES values (a continuous score, ranging from relatively greater socio-economic disadvantage to relatively greater socio-economic advantage, national M = 1000, SD = 100). As described in section “Data Analysis” below, in the boarding sample we also examined the association between years as a boarding student (a continuous variable) and academic outcomes.

Personality

We were also interested in the extent to which boarding status accounted for variance beyond existing personality traits (in line with recommendations by Martin et al., 2014 ). A brief personality scale ( Gosling et al., 2003 ) was administered, consisting in our study of a single item measure for each personality factor. Gosling et al. (2003) found adequate levels of validity and alignment between self and observer ratings. On a scale of 1 (Strongly Disagree) to 7 (Strongly Agree) students rated themselves on each of the “Big 5” personality traits, as follows: “I see myself as”… “sympathetic, warm” (agreeableness), “dependable, self-disciplined” (conscientiousness), “extraverted, enthusiastic” (extraversion), “anxious, easily upset” (neuroticism), “open to new experiences” (openness to experience).

Data Analysis

The analyses were conducted with M plus 7.31 ( Muthén and Muthén, 2015 ). Maximum likelihood with robustness to non-normality (MLR) was employed as the method of estimation ( Muthén and Muthén, 2015 ). Missing data (7.75% missing data points) were dealt with via the M plus full information maximum likelihood defaults (FIML; Muthén and Muthén, 2015 ). To account for the fact that students (L1) and classrooms (L2) are clustered within schools, we also adjusted standard errors for school using the “cluster” and “complex” commands in M plus (we did not conduct a 3-level multilevel model—students nested within classrooms within schools—because there was not a sufficient number of schools to justify this).

In the first instance, variance components analyses were conducted to determine between-class variation in boarding status, and science motivation, engagement, and achievement. Intraclass correlations (ICCs) were of interest here, identifying the percentage of between-class variance for each measure (reported in Table 1 ). Then, multilevel correlation analyses were conducted, where, in a single model, student-level (L1) associations among all variables (domain-general and -specific) were examined, as were all relevant class-level (L2) associations (domain-specific only).

Following this, analyses centered on multilevel path analysis. This proceeded through three stages. For Step 1 at L1, student boarding status was entered as a predictor of student-level science motivation, engagement, and achievement and also student-level domain-general motivation and engagement. For Step 1 at L2, class-level boarding status predicted class-level science motivation, engagement, and achievement. For Step 2 at L1, student boarding status, background attributes, and personality factors were entered as predictors of student-level science motivation, engagement, and achievement and also student-level domain-general motivation and engagement. Step 2 at L2 was the same as Step 1 at L2. Step 3 at L1 added to Step 2 by also assessing the extent to which student-level background and personality attributes moderated the effects of student-level boarding status—by way of interaction terms (e.g., boarding × age, etc.; calculated by zero-centering the main effects and finding their product; Aiken et al., 1991 ). Step 3 at L2 (classroom-level) was the same as L2 in Steps 1 and 2. Boarding status was modeled using the doubly latent format in M plus , with L2 effects disentangled from L1 effects; however, for completeness, in Table 3 notes we present findings for a model in which boarding status was modeled as a raw score at L1 and a cluster (class) aggregate at L2—the same pattern of findings was derived. Figure 1 presents the complete model at Step 3.

Table 3A. Multilevel path model: Step 1 (boarding as predictor) and Step 2 (boarding, background attributes, and personality as predictors).

Table 3B. Multilevel path model: Step 3—Boarding, background attributes, personality, and interactions predicting motivation, engagement, and achievement.

In a supplementary analysis among boarding students only, we also investigated the association between years in boarding school and academic outcomes. At L1, years in boarding (alongside background attributes, and personality factors) was entered as a predictor of student-level science motivation, engagement, and achievement and also student-level domain-general motivation and engagement. At L2, class-level years boarding (i.e., average years boarding in a class) predicted class-level science motivation, engagement, and achievement.

In our study, the sample is large and there is a risk that effects are disproportionately biased toward statistical significance. Thus, to avoid giving undue weight to effect sizes that are statistically significant but small (given the large sample size), we applied Keith’s (2006) guidelines and a more stringent p -value ( p < 0.001) to help us determine if a finding was interpretable. As per Keith (2006) , effect sizes (β) of 0.05 and above are considered small, β of 0.10 and above are moderate, and β of 0.25 and above are large. Accordingly, effects that are significant at p < 0.001 and β ≥ 0.05 are taken as interpretable.

Descriptive Statistics, Classroom Variation, and Multilevel Correlations

Table 1 presents means, standard deviations, skewness, kurtosis, and reliability (coefficient omega) for all substantive measures (motivation, engagement, achievement) in the study. Socio-demographic descriptive statistics were presented in Participants, above. The distributional properties demonstrated that the measures were approximately normally distributed, with standard deviations appropriately proportional to means, and skewness and kurtosis values within acceptable ranges ( Kline, 2011 ). Omega coefficients ranged between 0.61 and 0.84 at student-level (L1) and between 0.87 and 0.98 at classroom-level (L2), suggesting generally acceptable reliability ( McNeish, 2018 ).

Variance components analyses identified the between-class variation (i.e., difference between science classrooms) in boarding status, science motivation, science engagement, and science achievement. Findings are shown in Table 1 which presents intraclass correlations (ICCs) and indicate the percentage variance for these variables from class-to-class (i.e., the percentage of how much variation there is between science classrooms, relative to residual and student-to-student variation). ICCs for the study’s L2 variables were: boarding status = 0.15 (15%), positive science motivation = 0.14 (14%), positive science engagement = 0.09 (9%), negative science motivation = 0.08 (8%), negative science engagement = 0.14 (14%), and science achievement = 0.31 (31%). There is thus notable variation between classrooms on each of the L2 factors—and with more than 10% of the variance on most factors explained at Level 2, multilevel modeling was justified ( Byrne, 2012 ).

We proceeded to test multilevel correlations underlying the hypothesized multilevel path model. This generates bivariate correlations that are the first insight into the relationships tested in Figure 1 . Correlations are presented in Table 2 . Here we summarize only significant correlations with L1 and L2 boarding factors (all non-significant correlations and all correlations among background attributes, personality, and outcomes are in Table 2 ). For L1 we examine the association between students’ boarding status and their motivation, engagement, and achievement; with positive (or negative) correlations indicating boarders scoring higher (or lower) on motivation, engagement, and/or achievement. For L2 we examine the association between the proportion of boarding students in a classroom and class-average motivation, engagement, and achievement; with positive (or negative) correlations indicating classrooms with a higher (or lower) proportion of boarders scoring higher (or lower) on class-average motivation, engagement, and/or achievement. As noted in Data Analysis, given the number of participants and the many parameters tested, we here focus on effects attaining p < 0.001. At the student-level (L1), boarding status was significantly and positively correlated with age ( r = 0.14, p < 0.001; boarders older), SES ( r = −0.59, p < 0.001; boarders lower), parent education ( r = −0.12, p < 0.001; boarders lower), prior achievement ( r = −0.14, p < 0.001; boarders lower), agreeableness ( r = −0.05, p < 0.001; boarders lower), and science achievement ( r = −0.10, p < 0.001; boarders lower). At L2 (classroom-level), boarding status was not significantly correlated with any outcome factors.

Multilevel Path Analyses

Step 1 main effects.

In Step 1 at student-level (L1), boarding status was the sole predictor of domain-general motivation and engagement and science motivation, engagement, and achievement. At classroom-level (L2), boarding status (proportion of boarders in a classroom) was the predictor of class-average science motivation, engagement, and achievement. In all cases, positive (or negative) standardized beta values indicate that boarding is associated with higher (or lower) scores on academic outcomes. Multilevel path analysis showed that student-level (L1) boarding status predicted science achievement (β = −0.07, p < 0.001; boarders lower) and negative science engagement (β = 0.05, p < 0.01; boarders higher). However, only the effect for science achievement attained the dual criteria for interpretability [β ≥ 0.05 (as per Keith, 2006 ) and p < 0.001—see section “Data Analysis” above]. For Step 1 at the class-level (L2), boarding status did not significantly predict any L2 science motivation, engagement, or achievement factors. Thus, the number of boarding students in the class (relative to day students) was not differentially associated with academic motivation, engagement, and achievement. In keeping with these generally non-significant boarding effects, the variance explained ( R 2 ) in Step 1 is also low. All Step 1 findings (significant and non-significant) are presented in Table 3A .

Step 2 Main Effects

In Step 2 at student-level (L1), boarding status, background attributes, and personality were predictors of domain-general motivation and engagement and science motivation, engagement, and achievement. At classroom-level (L2), boarding status (proportion of boarders in a classroom) was the predictor of class-average science motivation, engagement, and achievement. In all cases, positive (or negative) standardized beta values indicate that boarding is associated with higher (or lower) scores on academic outcomes. All (significant and non-significant) findings are presented in Table 3A . Here we focus on boarding effects; effects for all other predictors are shown in Table 3A . These analyses showed that student-level (L1) boarding status predicted positive domain-general motivation (β = 0.05, p < 0.01; boarders higher), positive domain-general engagement (β = 0.06, p < 0.001; boarders higher), and negative science engagement (β = 0.03, p < 0.01; boarders higher). However, only the effect for positive domain-general engagement attained the dual criteria for interpretability; and, the interpretable Step 1 effect for achievement dropped out. Class-level (L2) boarding status did not significantly predict any L2 science motivation, engagement, or achievement factors. Thus, the proportion of boarding students in the class was not significantly associated with class-average academic motivation, engagement, and achievement.

Inclusion of Step 2 background and personality attributes yielded a significant increase (at p < 0.001) in explained variance for L1. Thus, at L1 for domain-general outcomes, beyond the role of boarding status these student attributes explained significant variance in positive motivation ( R 2 = 0.37), negative motivation ( R 2 = 0.30), positive engagement ( R 2 = 0.37), and negative engagement ( R 2 = 0.22). At L1 for science outcomes, beyond the role of boarding status the student attributes explained significant variance in positive motivation ( R 2 = 0.26), negative motivation ( R 2 = 0.25), positive engagement ( R 2 = 0.28), negative engagement ( R 2 = 0.25), and achievement ( R 2 = 0.20).

Step 3 Main and Interaction Effects

In Step 3 at student-level (L1), boarding status, background attributes, personality (as main effects) and the cross-products of boarding × background/personality attributes (interaction effects; e.g., boarding × age, etc.) were predictors of domain-general motivation and engagement and science motivation, engagement, and achievement. At classroom-level (L2), boarding status (proportion of boarders in a classroom) was the predictor of class-average science motivation, engagement, and achievement. In all main effects, positive (or negative) standardized beta values indicate that boarding is associated with higher (or lower) scores on academic outcomes. Interaction effects are unpacked as appropriate and described below. All (significant and non-significant) findings are presented in Table 3B .

For Step 3 main effects , multilevel path analysis showed that student-level (L1) boarding status predicted positive domain-general motivation (β = 0.05, p < 0.01; boarders higher), domain-general positive engagement (β = 0.08, p < 0.001; boarders higher), and negative domain-general engagement (β = −0.09, p < 0.01; boarders lower). However, only the effect for positive domain-general engagement attained the dual criteria for interpretability. Class-level (L2) boarding status did not significantly predict any L2 science motivation, engagement, or achievement factors. In this final model, other L1 main effects attaining the dual criteria for interpretability were age, prior achievement, neuroticism, openness, conscientiousness, and agreeableness (see Table 3B for strength and direction of standardized beta coefficients).

For Step 3 interaction effects , three effects attained Keith’s (2006) benchmark (β ≥ 0.05) and significance at p < 0.001. The first was boarding × conscientiousness for negative domain-general engagement (β = 0.10, p < 0.001). In follow-up simple effects tests, we found that for students low in conscientiousness there was a larger effect of boarding status on negative domain-general engagement (β = −0.09) than for students high in conscientiousness (β = −0.01). The second was boarding × gender for science achievement (β = −0.06, p < 0.001). For females, there was a larger effect of boarding status on science achievement (β = −0.18) than for boys (β = −0.04). The third was boarding × SES for negative science engagement (β = -0.07, p < 0.001). For low SES students, there was a larger effect of boarding status on negative science engagement (β = 0.04) than for high SES students (β < 0.01).

Supplementary Analyses: Years as a Boarding Student

In a supplementary analysis among boarding students only, we also investigated the association between years as a boarding student and academic outcomes. Controlling for background attributes and personality factors at L1, we found that years as a boarding student positively predicted science test achievement (β = 0.08, p < 0.01; more time in boarding associated with higher achievement); however, this effect did not attain our dual criteria for interpretability (β ≥ 0.05 and p < 0.001). Class-level (L2) years in boarding did not significantly predict any L2 science motivation, engagement, or achievement factors. Taken together, then, time spent in boarding was not a salient factor in students’ academic outcomes.

After controlling for background and personality attributes, we found predominant parity between boarding and day students in their motivation, engagement, and achievement. We also found that motivation, engagement, and achievement at the class-level were not significantly affected by the number of boarders in the classroom. In addition, the effects of boarding were generally not moderated by students’ background or personality attributes. Thus, we conclude that boarders have academic opportunities and outcomes that are comparable to day students.

Student-Level and Classroom-Level Effects: Boarding vs. Day Status

Schools comprising boarding and day students represent a unique research design. In our study, many students constituted what we might consider a “treatment” group (boarding students) and many others constituted a “comparison” group (day students). The two groups were educated in the same classrooms and received the same syllabus and instruction from the same teachers. In fact, the clustering of students in the same classrooms enabled us to extend prior research by investigating the extent to which the number of boarders in a (science) classroom had an impact on class-average motivation, engagement, and achievement outcomes (in science).

Prior multivariate research into boarding effects had only considered domain-general motivation and engagement ( Martin et al., 2014 ), and found predominant parity between boarding and day students on these outcome factors. However, that earlier research had been conducted at the individual student level, looking at an individual’s boarding (or day) status and its relationship with an individual’s motivation and engagement; it did not take into account the possibility that a critical mass of boarding students in a classroom may affect class-level outcomes. It is known that individuals can affect the group to which they belong and in turn the group can affect these individuals ( Raudenbush and Bryk, 2002 ; Goldstein, 2003 ; Marsh et al., 2008 ). This raises the question: given that boarding status represents a unique educational experience (see section “Introduction”), does that experience converge in a classroom where other boarders are present to affect overall class-average outcomes? Because we collected data on science motivation, engagement, and achievement in science classrooms, we were able answer this question. Our findings showed that the proportion of boarders in the classroom (relative to day students) was not significantly associated with class-average science outcomes.

The study’s multilevel design was important to help better ascertain the nature of boarding effects. With this design we could disentangle student-level (Level 1) variance from class-level (Level 2) variance. In doing so it was evident that boarding status was not associated with science outcomes at either level ( Table 3B ). Given this, it was interesting to note that there were boarding effects for domain-general motivation and engagement (boarders higher in positive domain-general motivation and engagement and lower in domain-general negative engagement)—though only one (for domain-general positive engagement) attained our dual criteria of interpretability (β ≥ 0.05 and p < 0.001). In our study, positive engagement comprised task management, planning and monitoring, and persistence factors. These are behavioral dimensions that may be quite responsive to the structured nature of study conditions in boarding contexts ( Lee and Barth, 2009 ). In these contexts, there are typically well-organized and well-planned study times and routines that students work to. These activities are also overseen and supported by teachers or other boarding house staff. Over time and relative to day students, these factors may have the effect of promoting a general disposition to better task management, persistence, etc. (i.e., positive domain-general engagement). Moreover, over time, boarding students may come to internalize these behaviors as their own capacities, further contributing to higher self-reported domain-general positive engagement. Interestingly, this was not the case for the domain-specific counterpart (positive science engagement) and this may be because there are key aspects of science engagement that are class-specific and applicable to both day and boarding students (e.g., science practicums, experiments, predicting, observing, etc.) and not linked to boarding study regimes.

Background and Personality Attributes

Findings in this study were also notable because they represented effects after controlling for background and personality attributes. As shown in correlations in Table 2 , there were significant bivariate associations among background attributes, boarding status, and academic outcomes—suggesting a need in our multivariate modeling to account for variance attributable to background attributes when assessing the unique relationship between boarding status and academic outcomes. In fact, Martin et al. (2014) emphasized the need for this and so our findings continue to underscore the fact that boarding effects cannot be fully interpreted without considering students’ background and personality attributes. Future research investigating boarding effects might thus consider these as particularly important to include and control for. Taken together, findings suggest that researchers ought not confuse or conflate boarding effects with effects due to some key background and personality attributes of boarding students. Relatedly, researchers ought to avoid raw comparisons of boarding and day students. Without adjusting for relevant background and personality attributes, raw comparisons may lead to biased results.

It has also been suggested that there may be some students for whom boarding may offer particular benefit. For example, it may provide Indigenous, rural, or remote students with educational opportunities not available to them in their distant residential communities. Or, being an inherently social residential context, perhaps boarding is better suited to students high in extraversion. We were able to test these possibilities through interaction effects (e.g., boarding × Indigenous status; boarding × extraversion, etc.; Table 3 ). Our findings suggested that for the most part boarding effects were not moderated by students’ background and personality attributes. Of the 108 possible interaction effects, only three attained our benchmark for interpretability (β ≥ 0.05 and p < 0.001): boarding × conscientiousness, boarding × gender, and boarding × SES. Taking our main and interaction effects together, then, it appears that including background and personality factors as main covariate effects is important when understanding boarding status, but it may not be necessary to model these background and personality factors as moderators of boarding status. Future research may seek to confirm this.

Practice Implications

We suggest that in the context of commentary and research documenting adversity for students in boarding schools, our finding of educational parity between boarding and day students is notable and has implications for educators and parents. For parents, one of their main concerns is that their child has educational opportunities and access on par with other students in a school. Indeed Lawrence (2005) identified that parents choose to board their child for various opportunities (e.g., extracurricular activity) and a structured and stable learning environment. Our findings suggest they receive such support—at least, to the extent that they evince academic outcomes comparable with day students. Many parents also send their child to board because, for one reason or another, their child does not have optimal educational access (e.g., due to geographic distance, etc.). We found that boarders’ results on motivation, engagement, and achievement were comparable to that of day students and we infer that this reflects equal opportunity and access for boarding students.

It was also interesting to note that the proportion of boarders in a classroom did not seem to be associated with class-average motivation, engagement, and achievement. There were no significant differences in these academic outcomes as a function of whether there were fewer or more boarders in the class. Schools often wrestle with classroom composition and how to collect students together to optimize academic and other outcomes. There has been a small body of research investigating classroom composition, finding some evidence that there are differences in motivation and engagement between classes taught by the same teacher ( Marsh et al., 2008 ). Our study adds to this work and would suggest that schools need not factor in the ratio of boarders to day students when deciding on class composition.

On the issue of access and opportunity, the general lack of moderation effects suggested no differences in academic outcomes between subgroups of boarders (and day students). For example, as we explained in the introduction, there have been questions about whether cultural identity may be unduly affected by the boarding experience or whether gender may play out in problematic ways in boarding contexts. In our study, there appeared to be no problematic patterns of interaction effects that would suggest issues along these lines: the general parity in academic outcomes between boarding and day students was found irrespective of a student’s background and personality attributes. From a practice perspective, the general lack of moderation effects suggests that efforts aimed at promoting motivation, engagement, and achievement among boarders need not be differentially directed at different sub-groups within the boarding community. Put another way, whether a student is a boarder or not, educational support to compensate low SES status or low prior achievement is required. Nevertheless, we did not assess some other potentially influential background attributes such as learning difficulties, which may require particular attention for some boarders (but conceivably not any more or less than among day students with learning difficulties, which would again suggest no interaction effect).

Importantly, however, although our study found predominant parity between boarders and day students and no interaction effects of note, there is no question that the transition from one’s community (and day school) to boarding school is a major one ( Martin et al., 2014 ). This being the case, it is prudent to consider educational support that can assist boarders in this transition and then through school. It is noteworthy that recent research in the Australian context (with particular focus on Indigenous students) has conducted quite a substantial body of work identifying supports that may be helpful. For example, research has shown that multidimensional intervention can be effectively administered to promote the resilience of Indigenous boarding students ( Benveniste et al., 2020 ). Likewise, a study of a social-emotional wellbeing program found that Indigenous boarders experienced an enhanced capacity to seek and provide help, work in groups, manage conflict, and discuss cultural issues ( Franck et al., 2020 ; see also Heyeres et al., 2018 ; Rutherford et al., 2019 ). Practices within the boarding school can also provide further opportunities to assist boarders’ academic and social-emotional wellbeing. For example, it has been shown that boarding staff can harness positive relationships with students to enhance students’ educational participation, mental health, and self-responsibility ( McCalman et al., 2020 ). Qualitative data from Indigenous boarding students and staff have also identified how boarding schools can be physically designed to optimize a sense of belonging. These include flexible spaces to foster relationships inside the boarding house, student voice in how spaces are designed and arranged, and spaces that provide “cultural relief” ( Whettingsteel et al., 2020 ). There are also culturally based strategies that can support boarding outcomes. For example, Lloyd and Duggie Pwerl (2020) showed how Indigenous students can achieve Western educational success in a boarding context through efforts by the school to maintain key aspects of their culture (also see Bobongie, 2017 ). Relatedly, Osborne et al. (2019) discuss this idea in terms of “both ways” capital where educators seek to affirm and strengthen Indigenous identity and help them develop positive Western academic codes.

Limitations, Future Directions, and Conclusions

There are limitations in our study important to consider when interpreting findings and which offer some direction for future research. First, we speculated that the boarding context comprises routines, structures, and interactions with educational support staff that are unique to that context—and that this may yield particular educational effects relative to day students. However, we did not have data on these factors in this study. We also did not have data on where boarders were from, including cohorts within the boarding sub-sample (such as Indigenous students). We therefore do not know if boarders from different areas evince different motivation, engagement, and achievement patterns and we cannot as fully contextualize the findings derived in this study. In future, collecting such data and ascertaining their impacts will identify more distinct effects particular to the boarding context and to whom these effects apply. Second, our study focused on high school students, not younger students in elementary school settings. Theories of attachment (e.g., Ainsworth and Bowlby, 1991 ) have emphasized the influential role of parents in children’s lives and it is possible that boarding reduces these important influences and may stunt personal development for younger students (also see Jack, 2020 ). It is thus important to test the generalizability of our findings to students in a range of other educational and residential contexts.

Third, given our cross-sectional data, there are questions about factor and causal ordering which may be answered by collecting data that monitors students who move from day status to boarding status and vice versa, as well as collecting data over time (i.e., longitudinal). This would provide a unique perspective on what changes (if any) occur as a result of changing status from one to the other. Relatedly, future research might look to include social-emotional wellbeing indicators as background attributes to disentangle the role of boarding status on academic outcomes from prior social-emotional wellbeing. These could include measures such as the Flourishing Scale ( Diener et al., 2009 ) and Kessler Psychological Distress Scale ( Kessler and Mroczek, 1994 ). Although our study did control for trait-like personality (including neuroticism as a mental health indicator), it is important to also control for more specific social-emotional wellbeing state-like measures. Indeed, the importance of considering a diversity of key factors and issues of modeling and causal ordering are increasingly being recognized by boarding school researchers (e.g., Guenther et al., 2020 ). Fourth, there is a need for more intensive real-time data. Recent research using mobile technology to capture real-time motivation and engagement has revealed in-the-moment variance in motivation and engagement ( Martin et al., 2015 , 2019b ) and it would be fascinating to explore motivation and engagement while students are doing study and homework in the boarding house. It is also important to recognize measurement issues for some groups of students in the boarding sector. For example, Langham et al. (2018) identified some challenges validating previously established tools measuring resilience among Indigenous boarders. It is therefore encouraging that the motivation and engagement measurement tool used in this study has been validated among Indigenous students ( Martin et al., 2019a ) and has previously been effective in assessing boarding effects among Indigenous students ( Martin et al., 2014 ). Fifth, despite modernization of the boarding sector ( Anderson, 2005 ), there are students for whom it is a negative experience. Future research might conduct person-centered analyses (e.g., latent profile analysis) to explore potential subgroups of boarders for whom it is a negative experience and examine the reasons why and the impact of this negative experience on their academic outcomes.

Sixth, as noted in Methods, the majority of students in this study were boys. Because of this, we conducted numerous additional statistical analyses leading us to tentatively conclude that gender composition did not disproportionately or unreasonably impact factors and empirical associations in this study. It is also the case that the sample was generally higher in SES than the national average; also, there were markedly more day students than boarding students. In some ways these imbalances are unavoidable in the Australian context as boarding schools tend to be higher SES independent schools and the ratio of boarding-to-day students is somewhat disproportionate given that most Australian boarding schools are in major urban areas or regional centers and enroll many “local” day students. We also point out that the average level of SES for our boarders was around the national average (see section “Methods”). Nevertheless, our investigation of interaction effects was important here because it allowed us to ascertain boarding and day status effects as a function of low and high SES. Another important feature of analyses was our modeling of SES as a covariate in analyses to control for variation attributable to it when seeking to identify unique boarding effects (beyond, for example, their lower SES relative to day students). This yielded a finding of predominant parity between boarding and day students. Thus, in the context of a history of negative effects of boarding on young people’s development ( Duffell, 2000 ), our finding of predominant parity is significant. This notwithstanding, future research should recruit more balanced samples to be further assured of the generalizability of the present results, as well as to look at potential gender differences. Finally, we did not have enough schools to model at the school-level; we could only do so at the student-level (for domain-general and domain-specific outcomes) and class-level (for domain-specific outcomes). We note prior research found variation between schools in their capacity to support boarders ( McCalman et al., 2020 ). Future research might recruit a sufficient number of schools to explore any variation in outcomes at the school-level. In all these ways we can better understand and assist boarding students as they navigate through their residential educational experience.

Data Availability Statement

The datasets presented in this article are not readily available because: Part of an industry research partnership; consent from participants to share dataset not available; summative data (e.g., correlation matrix with standard deviations) available to enable analyses. Requests to access the datasets should be directed to AM, [email protected] .

Ethics Statement

The studies involving human participants were reviewed and approved by the UNSW Human Ethics Committee. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

Author Contributions

AM led research design, led data analysis, and led report writing. EB and RK assisted with research design, assisted with interpretation of findings, and assisted with report writing. JP assisted with research design and assisted with report writing. VM-S assisted with report writing. All authors contributed to the article and approved the submitted version.

This study was funded by the Australian Research Council (Grant #LP170100253) and the Future Project at The King’s School.

Conflict of Interest

It is appropriate to note that one of the measures (the MES) in the study is a published instrument developed by this study’s first author attracting a small fee and royalty, part of which is put toward its ongoing development and administration and part of which is also donated to UNICEF. However, for this study, there was no fee involved for its use.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

• ^ Boarding students are situated in residential contexts, often called boarding houses, boarding dormitories, or boarding residences (referred to henceforth as boarding houses). Boarding schools comprise one or several boarding houses. The ratio of boarding-to-day students in a school varies. Some schools are mainly day schools, with relatively few boarding students; other schools have a larger boarding contingent.

Abraham, J., and Barker, K. (2015). An expectancy-value model for sustained enrolment intentions of senior secondary physics students. Res. Sci. Educ. 45, 509–526. doi: 10.1007/s11165-014-9434-x

CrossRef Full Text | Google Scholar

Aiken, L. S., West, S. G., and Reno, R. R. (1991). Multiple regression: Testing and interpreting interactions. London: Sage.

Google Scholar

Ainsworth, M. S., and Bowlby, J. (1991). An ethological approach to personality development. Am. Psychol. 46, 333–341. doi: 10.1037/0003-066X.46.4.333

Aleman, E. (2013). “Critical race theory and human capital theory,” in To what ends and by what means , eds G. M. Rodriguez and R. A. Rolle (Abingdon: Routledge), 47–70.

Anderson, E. W. (2005). Residential and boarding education and care for young people: A model for good practice. Abingdon: Routledge, doi: 10.4324/9780203694534

Australian Bureau of Statistics (2019). Schools Australia. Canberra: Australian Bureau of Statistics.

Australian Curriculum Assessment Reporting Authority (ACARA) (2020a). My school. Sydney: ACARA (accessed November 10, 2020).

Australian Curriculum Assessment Reporting Authority (ACARA) (2020b). National assessment program - literacy and numeracy: National report 2018. Sydney: ACARA. Available Online at: https://nap.edu.au/docs/default-source/resources/2018-naplan-national-report_file.pdf

Bass, L. R. (2013). Boarding schools and capital benefits: Implications for urban school reform. J.Educat. Res. 107, 16–35. doi: 10.1080/00220671.2012.753855

Behaghel, L., De Chaisemartin, C., and Gurgand, M. (2017). Ready for boarding? The effects of a boarding school for disadvantaged students. Am. Econom. J. Appl. Econom. 9, 140–164. doi: 10.1257/app.20150090

Benveniste, T., Guenther, J., Dawson, D., and King, L. (2019). Race, rules and relationships: What can critical race theory offer contemporary Aboriginal boarding schools? J. Intercult. Stud. 40, 32–48. doi: 10.1080/07256868.2018.1552573

Benveniste, T., Van Beek, A., McCalman, J., Langham, E., and Bainbridge, R. (2020). Can it be done? An evaluation of staff perceptions and affordability of a school-based multi-component integrated intervention for improving the resilience of Aboriginal and Torres Strait Islander boarding students. Aus. Int. J. Rural Educ. 30:33.

Bobongie, F. (2017). What is the key to a successful transition for Indigenous girls moving from Torres Strait Island communities to boarding colleges in regional Queensland, Australia? Int. J. Technol. Inclusive Educ. 6, 1144–1152.

Bourdieu, P. (1986). “The forms of capital,” in Handbook of theory and research for the sociology of education , ed. J. C. Richardson (Connecticut: Greenwood Press), 241–258.

Bronfenbrenner, U. (1970). Reaction to social pressure from adults versus peers among Soviet day school and boarding school pupils in the perspective of an American sample. J. Personal. Soc. Psychol. 15, 179–189. doi: 10.1037/h0029426

Bronfenbrenner, U. (2001). “The bioecological theory of human development,” in International encyclopedia of the social and behavioral sciences , Vol. 10, eds N. J. Smelser and P. B. Baltes (Amsterdam: Elsevier).

Bronfenbrenner U. (ed.) (2005). Making human beings human: Bioecological perspectives on human development. London: Sage.

Bronfenbrenner, U., and Morris, P. A. (1998). “The ecology of developmental processes,” in Handbook of child psychology: Theoretical models of human development , eds W. Damon and R. M. Lerner (New Jersey: John Wiley & Sons Inc), 993–1028.

Byrne, B. M. (2012). Structural equation modeling with Mplus: Basic concepts, applications, and programming. Abingdon: Routledge, doi: 10.4324/9780203807644

Chen, F. F. (2007). Sensitivity of goodness of fit indices to lack of measurement invariance. Struct. Equat. Model. 14, 464–504. doi: 10.1080/10705510701301834

Cheung, G. W., and Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Struct. Equat. Model. 9, 233–255. doi: 10.1207/S15328007SEM0902_5

Cole, D. A., and Preacher, K. J. (2014). Manifest variable path analysis: Potentially serious and misleading consequences due to uncorrected measurement error. Psychol. Methods 19, 300–315. doi: 10.1037/a0033805

PubMed Abstract | CrossRef Full Text | Google Scholar

Curto, V. E., and Fryer, R. G. (2011). Estimating the returns to urban boarding schools: Evidence from SEED. Natl. Bureau Econom. Res. 2011:w16746. doi: 10.3386/w16746

Diener, E., Wirtz, D., Tov, W., Kim-Prieto, C., Choi, D., Oishi, S., et al. (2009). New measures of well-being: Flourishing and positive and negative feelings. Soc. Indic. Res. 39, 247–266.

Downs, J. (2002). “Adapting to secondary and boarding school: self-concept, place identity and homesickness,” in Paper presented at the Self-concept Research: Driving International Research Agendas: Collected Papers of the Second Biennial Self-concept Enhancement and Learning Facilitation (SELF) Research Centre International Conference , (Sydney: SELF Research Centre).

Duffell, N. (2000). The making of them: The British attitude to children and the boarding school system. London: Lone Arrow Press.

English, S., and Guerin, B. (2017). What some female Indigenous secondary students say is important at an Indigenous residential college. J. Aus. Indigenous Iss. 20, 71–89.

Finn, P. J. (2012). Preparing for power in elite boarding schools and in working-class schools. Theory Pract. 51, 57–63. doi: 10.1080/00405841.2012.636339

Franck, L., Midford, R., Cahill, H., Buergelt, P. T., Robinson, G., Leckning, B., et al. (2020). Enhancing social and emotional wellbeing of Aboriginal boarding students: Evaluation of a social and emotional learning pilot program. Int. J. Environ. Res. Public Health 17:771. doi: 10.3390/ijerph17030771

Fredricks, J. A., and Eccles, J. S. (2005). Developmental benefits of extracurricular involvement: Do peer characteristics mediate the link between activities and youth outcomes? J. Youth Adolesc. 34:507. doi: 10.1007/s10964-005-8933-5

Gogol, K., Brunner, M., Goetz, T., Martin, R., Ugen, S., Keller, U., et al. (2014). My questionnaire is too long!” The assessments of motivational-affective constructs with three-item and single-item measures. Contemp. Educ. Psychol. 39, 188–205. doi: 10.1016/j.cedpsych.2014.04.002

Goldstein, H. (2003). Multilevel statistical models. 3rd ed. London: Hodder Arnold

Gosling, S. D., Rentfrow, P. J., and Swann, W. B. Jr. (2003). A very brief measure of the Big-Five personality domains. J. Res. Personal. 37, 504–528. doi: 10.1016/S0092-6566(03)00046-1

Green, J., Martin, A. J., and Marsh, H. W. (2007). Motivation and engagement in English, mathematics and science high school subjects: Towards an understanding of multidimensional domain specificity. Learning Individ. Differ. 17, 269–279. doi: 10.1016/j.lindif.2006.12.003

Guenther, J., and Fogarty, B. (2018). Examining remote Australian First Nations boarding through capital theory lenses. Crit. Stud. Educ. 9, 1–17. doi: 10.1080/17508487.2018.1543201

Guenther, J., Benveniste, T., Redman-MacLaren, M., Mander, D., McCalman, J., O’Bryan, M., et al. (2020). Thinking with theory as a policy evaluation tool: The case of boarding schools for remote First Nations students. Eval. J. Aus. 20, 34–52. doi: 10.1177/1035719X20905056

Guerin, B., and Pertl, J. (2017). The boarding school experiences of some Indigenous male secondary students: How the staff talk about the students. J. Aus. Indigen. Iss. 20, 54–70.

Hayduk, L. A. (1987). Structural equation modeling with LISREL: Essentials and advances. Baltimore, MD: Johns Hopkins University Press.

Heyeres, M., McCalman, J., Langham, E., Bainbridge, R., Redman-MacLaren, M., Britton, A., et al. (2018). Strengthening the capacity of education staff to support the wellbeing of Indigenous students in boarding schools: A participatory action research study. Aus. J. Indigen. Educ. 48, 1–14. doi: 10.1017/jie.2017.42

Holden, M. J., Io, C., Nunno, M., Smith, E. G., Endres, T., Holden, J. C., et al. (2010). Children and residential experiences: A comprehensive strategy for implementing a research-informed program model for residential care. Child Welfare 89, 131–149.

Jack, C. (2020). Recovering boarding school trauma narratives: Christopher Robin Milne as a psychological companion on the journey to healing. Oxfordshire: Taylor & Francis Group.

Keith, T. Z. (2006). Multiple regression and beyond. London: Pearson Education.

Kessler, R. C., and Mroczek, D. (1994). Final versions of our non-specific psychological distress scale. Michigan: University of Michigan.

Khan, S. R. (2010). Privilege: The making of an adolescent elite at St Paul’s School. Princeton: Princeton University Press.

Kline, R. B. (2011). Methodology in the social sciences. Principles and practice of structural equation modelling , 3rd Edn. New York: Guilford Press.

Langham, E., McCalman, J., Redman-MacLaren, M., Hunter, E., Wenitong, M., Britton, A., et al. (2018). validation and factor analysis of the child and youth resilience measure for indigenous Australian boarding school students. Front. Public Health 6:299–299. doi: 10.3389/fpubh.2018.00299

Lawrence, R. (2005). The sustainability of boarding: Choice and influencing factors of parents who elect to send their children to boarding school. Melbourne: Prospect Research and Marketing.

Leaper, C. (2011). More similarities than differences in contemporary theories of social development? A plea for theory bridging. Adv. Child Dev. Behav. 40, 337–378. doi: 10.1016/b978-0-12-386491-8.00009-8

Lee, B. R., and Barth, R. P. (2009). Residential education: An emerging resource for improving educational outcomes for youth in foster care? Child. Youth Ser. Rev. 31, 155–160. doi: 10.1016/j.childyouth.2008.07.007

Lester, L., and Mander, D. (2020). A longitudinal mental health and wellbeing survey of students transitioning to a boys’ only boarding school. Aus. Int. J. Rural Educ. 30, 67–83.

Liem, G. A., and Martin, A. J. (2012). The Motivation and Engagement Scale: Theoretical framework, psychometric properties, and applied yields. Aus. Psychol. 47, 3–13. doi: 10.1111/j.1742-9544.2011.00049.x

Lloyd, A., and Duggie Pwerl, T. (2020). Interschool partnerships: Remote Indigenous boarding students experiencing Western education whilst keeping culturally safe. Rural Soc. 29, 1–16. doi: 10.1080/10371656.2020.1809138

MacDonald, M., Gringart, E., Ngarritjan Kessaris, T., and Gray, J. (2018). A ‘better’ education: An examination of the utility of boarding school for Indigenous secondary students in Western Australia. Aus. J. Educ. 62, 192–196. doi: 10.1177/0004944118776762

Mander, D. J., and Lester, L. (2017). A longitudinal study into indicators of mental health, strengths and difficulties reported by boarding students as they transition from primary school to secondary boarding schools in Perth, Western Australia. J. Psychol. Counsellors Schools 27, 139–152. doi: 10.1017/jgc.2017.1

Marsh, H. W. (1992). Extracurricular activities: Beneficial extension of the traditional curriculum or subversion of academic goals? J. Educ. Psychol. 84, 553–562. doi: 10.1037/0022-0663.84.4.553

Marsh, H. W., and Kleitman, S. (2002). Extracurricular school activities: The good, the bad, and the nonlinear. Harvard Educ. Rev. 72, 464–514. doi: 10.17763/haer.72.4.051388703v7v7736

Marsh, H. W., and Martin, A. J. (2011). Academic self-concept and academic achievement: Relations and causal ordering. Br. J. Educ. Psychol. 81, 59–77. doi: 10.1348/000709910x503501

Marsh, H. W., Lüdtke, O., Nagengast, B., Trautwein, U., Morin, A. J., Abduljabbar, A. S., et al. (2012). Classroom climate and contextual effects: Conceptual and methodological issues in the evaluation of group-level effects. Educ. Psychol. 47, 106–124. doi: 10.1080/00461520.2012.670488

Marsh, H. W., Martin, A. J., and Cheng, J. (2008). A multilevel perspective on gender in classroom motivation and climate: Potential benefits of male teachers for boys? J. Educ. Psychol. 100, 78–95. doi: 10.1037/0022-0663.100.1.78

Martin, A. J. (2007). Examining a multidimensional model of student motivation and engagement using a construct validation approach. Br. J. Educ. Psychol. 77, 413–440. doi: 10.1348/000709906X118036

Martin, A. J. (2012). “Motivation and engagement: Conceptual, operational and empirical clarity. Section Commentary,” in Handbook of Research on Student Engagement , eds S. Christenson, A. Reschly, and C. Wylie (Berlin: Springer).

Martin, A. J. (2015). Motivation and Engagement Scale: User Manual. New South Wales: Lifelong Achievement Group.

Martin, A. J., Ginns, P., Anderson, M., Gibson, R., and Bishop, M. (2019a). Motivation and engagement among Indigenous (Aboriginal Australian) and non-Indigenous students. New South Wales: UNSW Sydney.

Martin, A. J., Mansour, M., and Malmberg, L.-E. (2019b). What factors influence students’ real-time motivation and engagement? An experience sampling study of high school students using mobile technology. Educ. Psychol. 40, 1113–1135. doi: 10.1080/01443410.2018.1545997

Martin, A. J., Papworth, B., Ginns, P., and Liem, G. A. D. (2014). Boarding school, motivation and engagement, and psychological well-being: A large-scale investigation. Am. Educ. Res. J. 51, 1007–1049. doi: 10.3102/0002831214532164

Martin, A. J., Papworth, B., Ginns, P., and Malmberg, L.-E. (2016). Motivation, engagement, and social ‘climate’: An international study of boarding schools. J. Educ. Psychol. 108, 772–787. doi: 10.1037/edu0000086

Martin, A. J., Papworth, B., Ginns, P., Malmberg, L.-E., Collie, R., and Calvo, R. A. (2015). Real-time motivation and engagement during a month at school: Every moment of every day for every student matters. Learning Individ. Differ. 38, 26–35. doi: 10.1016/j.lindif.2015.01.014

Mavor, K. I., Platow, M. J., and Bizumic, B. (eds) (2017). Self and social identity in educational contexts. Abingdon-on-Thames: Taylor & Francis.

McCalman, J., Benveniste, T., Wenitong, M., Saunders, V., and Hunter, E. (2020). It’s all about relationships”: The place of boarding schools in promoting and managing health and wellbeing of Aboriginal and Torres Strait Islander secondary school students. Child. Youth Ser. Rev. 20:104954. doi: 10.1016/j.childyouth.2020.104954

McNeish, D. M. (2018). Thanks coefficient alpha, we’ll take it from here. Psychol. Methods 23, 412–433. doi: 10.1037/met0000144

Muthén, L. K., and Muthén, B. O. (2015). Mplus user’s guide. Los Angeles: Muthén & Muthén.

Nsw Education Standards Authority (2019). Science Years 7–10 Syllabus (2018). New South Wales: NSW Education Standards Authority. Available Online at: https://educationstandards.nsw.edu.au/wps/portal/nesa/k-10/learning-areas/science/science-7-10-2018

OECD (2020). Education GPS. Paris: OECD.

Office of the Chief Scientist (2014). Benchmarking Australian science. Canberra: Australian Government.

Osborne, S., Rigney, L.-I., Benveniste, T., Guenther, J., and Disbray, S. (2019). Mapping boarding school opportunities for Aboriginal students from the Central Land Council region of Northern Territory. Aus. J. Indigen. Educ. 48, 162–178. doi: 10.1017/jie.2018.1

Raudenbush, S. W., and Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods , 2nd Edn. London: Sage.

Redman-MacLaren, M., Benveniste, T., McCalman, J., Rutherford, K., Britton, A., Langham, E., et al. (2019). Through the eyes of students: The satisfaction of remote Indigenous boarding students with a transition support service in Queensland, Australia. Aus. J. Indigen. Educ. 2019, 1–12. doi: 10.1017/jie.2019.3

Ross, P. M., and Poronnik, P. (2013). “Science education in Australia: Time of change,” in Grumpy scientists: The ecological conscience of a nation , eds D. Lunney, P. Hutchings, and H. F. Recher (New South Wales: Royal Zoological Society of New South Wales), 59–66. doi: 10.7882/FS.2013.013

Rutherford, K., McCalman, J., and Bainbridge, R. (2019). The post-schooling transitions of remote indigenous secondary school graduates: A systematic scoping review of support strategies. Aus. Int. J. Rural Educ. 29, 9–25.

Schaverien, J. (2011). Boarding school syndrome: Broken attachments a hidden trauma. Br. J. Psychother. 27, 138–155. doi: 10.1111/j.1752-0118.2011.01229.x

Tajfel, H. E. (1978). Differentiation between social groups: Studies in the social psychology of intergroup relations. Cambridge: Academic Press.

Tajfel, H., and Turner, J. C. (1986). “The social identity theory of intergroup behavior,” in Psychology of intergroup relation , eds S. Worchel and W. G. Austin (New Jersey, NJ: Hall Publishers), 7–24.

The Association of Boarding Schools (TABS) (2003). Boarding experience outcomes: High school, post-college, mid-career, and late-career groups. Asheville NC: TABS.

The Association of Boarding Schools (TABS) (2013). The truth about boarding school. Asheville NC: The Association of Boarding Schools.

Thomson, S., Wernert, N., O’Grady, E., and Rodrigues, S. (2016). TIMSS 2015: A first look at Australia’s results. Melbourne: ACER.

Tröbst, S., Kleickmann, T., Lange-Schubert, K., Rothkopf, A., and Möller, K. (2016). Instruction and students’ declining interest in science: An analysis of German fourth-and sixth-grade classrooms. Am. Educ. Res. J. 53, 162–193. doi: 10.3102/0002831215618662

Van de Schoot, R., Lugtig, P., and Hox, J. (2012). A checklist for testing measurement invariance. Eur. J. Dev. Psychol. 9, 486–492. doi: 10.1080/17405629.2012.686740

Whettingsteel, E., Oliver, R., and Tiwari, R. (2020). It would give you a space to be yourself’: Increasing a sense of belonging for Aboriginal students in boarding schools. Aus. Int. J. Rural Educ. 30, 84–110.

White, M. A. (2004). An Australian co-educational boarding school as a crucible for life. Ph.D. thesis, University of Adelaide, Adelaide.

Keywords : boarding, residential, motivation, engagement, achievement, science

Citation: Martin AJ, Burns EC, Kennett R, Pearson J and Munro-Smith V (2021) Boarding and Day School Students: A Large-Scale Multilevel Investigation of Academic Outcomes Among Students and Classrooms. Front. Psychol. 11:608949. doi: 10.3389/fpsyg.2020.608949

Received: 22 September 2020; Accepted: 02 December 2020; Published: 05 January 2021.

Reviewed by:

Copyright © 2021 Martin, Burns, Kennett, Pearson and Munro-Smith. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Andrew J. Martin, [email protected]

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.

• Open access
• Published: 23 September 2023

The impact of boarding schools on the development of cognitive and non-cognitive abilities in adolescents

• Fang Chang 1 ,
• Yanan Huo 1 ,
• Songyan Zhang 1 ,
• Hang Zeng 1 &
• Bin Tang 1  

BMC Public Health volume  23 , Article number:  1852 ( 2023 ) Cite this article

7571 Accesses

4 Altmetric

Metrics details

Since China adopted a policy to eliminate rural learning centers, boarding has become an important feature of the current rural student community. However, there is a lack of consensus on the impact of boarding schools on students' cognitive and non-cognitive development. This study investigates the effect of boarding schools on the development of cognitive and non-cognitive abilities of junior high school students in rural northwest China.

Using a sample of 5,660 seventh-grade students from 160 rural junior high schools across 19 counties, we identify a causal relationship between boarding and student abilities with the instrumental variables (IV) approach.

The results suggest that boarding positively influences memory and attention, while it has no significant effect on other cognitive abilities such as reasoning, transcription speed, and accuracy. Furthermore, we find no significant association between boarding and the development of non-cognitive skills.

Conclusions

Given the widespread prevalence of boarding schools in rural regions, our study highlights the growing importance of improving school management to promote the development of students’ cognitive abilities and integrating the development of non-cognitive or social-emotional abilities into students’ daily routines.

Peer Review reports

Introduction

Human ability is composed of cognitive and non-cognitive components, both of which are crucial to an individual's life [ 1 , 2 ]. The influence of cognitive and non-cognitive abilities has been observed in various aspects of life, including academic performance, educational choices, wages, labor market outcomes, employment decisions, health behaviors, and social integration [ 3 , 4 , 5 ].

Cognitive and non-cognitive abilities are core components of human capital [ 1 , 2 , 6 ]. Cognitive abilities are the endowments for extracting, storing and utilizing information from the objective world, This encompasses skills such as logical reasoning, abstract thinking and memory [ 7 ], while non-cognitive abilities have emerged as a concept distinct from cognitive abilities, aiming to distinguish factors beyond cognitive itself. These encompass qualities such as motivation, authority, work norms, self-control, perseverance, and more [ 8 ]. Numerous researches have shown that cognitive and non-cognitive abilities play an important role in academic performance, educational decisions, wages, labor market performance, employment choices, health behaviors and social integration [ 1 , 3 , 4 , 5 , 8 , 9 ]. Prior research has revealed a considerable disparity between the cognitive abilities of rural and urban students, with urban students scoring significantly higher on word and mathematics tests by 0.75 and 0.54 standard deviations, respectively [ 6 , 10 ]. Rural students also tend to exhibit lower levels of non-cognitive skills, including depression, self-esteem, and values, with left-behind children experiencing even greater disadvantages [ 11 ].

Beginning in 2001, China adopted a policy to eliminate rural learning centers, leading to the consolidation of educational resources and the growth of rural boarding schools. By the end of 2016, 26.08 million rural students were enrolled as compulsory boarders, comprising 27.5% of the total student population. Of these, 16.66 million were boarding students in rural junior high schools, amounting to a boarding rate of 58.6% [ 12 ]. Therefore, a comprehensive evaluation of the cognitive and non-cognitive development of boarding students in rural areas has become essential.

Studies have shown that the communal learning environment in boarding schools can increase learning time, optimize teaching resources, and provide more opportunities for boarders to communicate with their teachers and peers [ 13 , 14 ]. However, boarding students may also be exposed to at-risk peers, which can have negative effects on their development [ 15 , 16 ]. Boarding can also cause stress for students as they are separated from their familiar surroundings and parents, which can be particularly significant during critical growth stages [ 17 ]. Consequently, there is a lack of consensus on the impact of boarding schools on students' cognitive and non-cognitive development.

Extant research on boarding schools has primarily focused on elite schools in developed countries, which have generally been associated with positive academic performance [ 18 ]. However, public boarding schools have been set up in many developed countries for marginalized groups, such as the SEED public boarding schools in the US and the Internet Excellence program in France. Quasi-experimental studies have shown that boarding has had a significant positive impact on the academic performance of disadvantaged students in reading and mathematics [ 19 ]. Similarly, rural boarding schools in France have positively impacted academic performance, particularly for outstanding boarders, with significant improvements in French and mathematics scores two years after enrollment [ 14 ]. Nonetheless, studies in Turkey have reported a negative correlation between boarding and academic performance in Grades 5 to 9 [ 15 ]. Boarding also has a significantly negative impact on students' mental health, with boarders displaying more problem behaviors, such as anxiety, depression, hostility, substance abuse, alcohol dependency, and school bullying [ 20 , 21 ]. Notably, the impact of boarding varies at different stages of development. For instance, Mander et al. (2015) found no significant differences in social, emotional, and psychological well-being between boarders and non-boarders in elementary schools [ 22 ]. However, boarders in secondary schools exhibited a higher incidence of emotional difficulties, depression, anxiety, and stress compared to non-boarders. Given the mixed evidence, it is crucial to carefully consider the potential positive and negative impacts of boarding, especially for disadvantaged students attending public boarding schools.

As boarding school enrollment continues to rise in China, researchers have investigated the effects of boarding on students' cognitive and non-cognitive abilities and reported conflicting findings. Qiao and Di (2014) found that boarding significantly improved rural students' performance in mathematics [ 23 ], while Mo et al. (2012) reported a significant negative effect of boarding on primary school students' math scores [ 24 ]. Similarly, Wang et al. (2016), Li et al. (2018), and Zhu et al. (2019) found that boarding had no significant impact on students' standardized math scores or even reduced their standardized language scores [ 25 , 26 , 27 ]. Most studies indicate that boarding has a negative impact on students' non-cognitive skills. Rural boarders are more likely to experience bullying, loneliness, and depression in schools and have lower self-esteem, resilience, and emotional intelligence than non-boarders [ 27 , 28 , 29 , 30 ]. Taken together, these results suggest that the effects of boarding on students' academic and non-academic outcomes are complex and may vary depending on a range of factors, including the type of boarding school, the students' developmental stage, and their socio-economic background.

This paper aims to contribute to the existing literature on the impact of boarding on rural students' cognitive and non-cognitive abilities in three ways. Firstly, the literature has primarily measured cognitive abilities using subject-specific scores, which may not fully capture the breadth of cognitive abilities. There are numerous studies on cognitive abilities in different disciplines. psychologists commonly differentiate between fluid intelligence, which emphasizes more general capacities such as logical reasoning and abstract thinking, and crystallized intelligence, which is related to the accumulation of concrete knowledge and experience [ 31 , 32 ]. Academic performance, such as math and reading tests is often used to measure crystallized intelligence [ 33 ]. Conversely, fluid intelligence is frequently assessed through quotient tests (IQ tests), exemplified bu tools like the WISC-IV and Raven's Standard Progressive Matrices [ 34 ]. To improve accuracy and precision in measuring cognitive abilities, this paper utilizes the Wechsler Intelligence Scale for Children (WISC), Raven's Standard Progressive Matrices, and standardized mathematics scores. Secondly, the literature has relied on self-administered questionnaires to measure non-cognitive abilities, which may lack comprehensiveness and comparability. In contrast, this paper uses the Big Five Personality Inventory to measure non-cognitive abilities accurately [ 35 ]. Finally, prior studies have examined the effects of boarding on cognitive or non-cognitive abilities separately, which prevents a comprehensive assessment of the impact of boarding on students' human capital.

This study uses an instrumental variable approach to address endogeneity issues and analyzes data from 160 junior high schools in rural northwest China to illustrate the effects of boarding on students' cognitive and non-cognitive abilities. The results indicate a significant positive effect of boarding on the cognitive abilities of rural junior high school students, particularly in memory and attention, areas associated with fluid intelligence. However, boarding has no significant impact on the non-cognitive abilities of rural students. Furthermore, we provide evidence of heterogeneity in the impact of boarding on cognitive and non-cognitive abilities by gender. We also find a significant positive effect on the cognitive abilities of left-behind children and students from families with better socioeconomic status.

Participants

We conducted our study on first-year rural high school (seventh grade) students in three prefectures from two provinces in northwest China. These provinces were below the national median in terms of GDP, according to the National Bureau of Statistics of China (2015). Hence, the sample of rural students in these provinces can be considered representative of students in low-income areas in rural China.

We constructed our sample in two steps. First, we selected 23 counties from three prefectures, two counties with more developed economic status were excluded, and the remaining were included. Second, we obtained a list of all 496 junior high schools from the counties in Step 1. After excluding non-rural schools and schools with less than 20 students in the seventh grade (to address potential sample attrition or school merger issues), we obtained a final sample containing 5,660 seventh-grade students from 160 schools (see Table 1 ).

The sample was collected in two phases. The first phase was carried out in 2015, which involved administering tests to collect information on basic details of the sample students, mathematics teachers, and schools using questionnaires. Mathematics scores of students were also collected through tests (see Table 2 ). In the second phase, conducted in 2016, additional tests were administered, which included more Raven's tests, Wechsler tests, the Big Five Personality test, and the Perseverance Scale (see Table 2 ).

The data collection involved three steps: (1) recruiting and training researchers, (2) conducting questionnaire tests in schools, and (3) administering cognitive and non-cognitive ability tests. For (1), the project team recruited college students as researchers and provided uniform training and simulation exercises to ensure recruited researchers mastered standardized operations of the study, thus reducing measurement errors caused by inconsistent implementation by researchers. For (2), researchers organized students to take standardized math tests and questionnaires, which were developed by the project team in collaboration with the best secondary school teachers and calibrated to match the academic level appropriate for seventh-grade rural students. All sample schools used standardized math tests with identical questions assigned by the project team and proctored by researchers on-site. Researchers also conducted one-on-one questionnaire interviews with principals and mathematics teachers. In (3), cognitive ability tests included Raven's test and Wechsler's test. Raven's test was administered in a group and took approximately 45 min. The Wechsler test needed to be conducted one-on-one and required highly trained personnel, participants therefore received training in professional institutions. Additionally, the project team organized several practical exercises in non-sample schools to ensure the accuracy and consistency of the Wechsler test. Given the significant testing and time costs of the Wechsler test, three students from each sample class were randomly selected to take the Wechsler test individually. Students are selected based on their mathematics scores in the first research sample class, which were rank ordered into three groups: high, medium, and low; one student from each group was randomly selected for the Wechsler's test. The rest of the class took the Raven's test. The non-cognitive skills component primarily consisted of the Big Five personality test and the Perseverance Scale test, both of which were included in the student questionnaire.

• Cognitive ability

The objective of this research is to investigate students' cognitive abilities, measured using three tests: the Wechsler Intelligence Scale for Children-Fourth Edition (WISC-IV), Raven's Standard Progressive Matrices (Raven's IQ test), and a mathematics test. Cattell's (1987) suggested that cognitive abilities are divisible into two categories: crystallized intelligence and fluid intelligence [ 36 ]. The former pertains to skills attained through experience and knowledge, such as vocabulary, calculation, and verbal comprehension, whereas the latter refers to neural development, including perception, memory, and reasoning ability.

The WISC-IV is a tool for assessing intelligence in children aged 6 to 16 and comprises four indices: verbal comprehension, perceptual reasoning, working memory, and processing speed, along with the total IQ score [ 37 ]. The Chinese version of the WISC-IV short-form scale was employed in this research, which contains four subtests representing the four indices [ 38 ]. Footnote 1 The four subtests utilized for estimating the WISC IQ score were similarities, digit span, coding, and matrix reasoning. Similarities is designed to capture crystallized intelligence, while digit span, coding, and matrix reasoning are intended to measure fluid intelligence [ 39 ]. The aggregated WISC IQ score was used in the regression analysis.

The Raven's IQ test is a nonverbal test of intelligence that consists of pictorial questions related to spatial reasoning and pattern matching, which are designed to assess observational and thinking ability [ 40 ]. The test is culture-, language-, and age-neutral and consists of 60 questions that can be converted into IQ scores based on normative patterns. It is defined to capture fluid intelligence and was used for robustness testing in this study [ 31 ]. Footnote 2

The mathematics test, administered to all students in the sample, was developed by experienced secondary school mathematics teachers based on the standard high school syllabus. The test captures crystallized intelligence and was used for robustness testing in this study [ 33 ]. Several pre-studies of the questions were carried out by the research team to assess their suitability.

• Non-cognitive ability

Non-cognitive abilities represent a fundamental component of human capital and can be examined through various skills and traits, including self-control, self-esteem, self-confidence, due diligence, perseverance, self-awareness, and communication skills [ 45 ]. We employed the Big Five personality traits and the Short Grit Scales as measures of non-cognitive abilities.

DeYoung's Big Five personality traits consist of neuroticism, agreeableness, conscientiousness, extraversion, and openness, which capture diverse aspects of personality. Neuroticism assesses emotional instability and sensitivity, such as anxiety, hostility, depression, self-consciousness, impulsivity, and vulnerability. Extraversion captures interpersonal skills, positive affect, and energy levels. Openness refers to the imagination and intellectual curiosity as reflected in personal fantasy, aesthetics, feelings, actions, ideas, and values. Agreeableness evaluates how a person interacts with others through levels of trust, frankness, altruism, submissiveness, humility, and gentleness. Finally, conscientiousness assesses competence, order responsibility, effortful achievement, self-discipline, and thoughtfulness [ 46 ]. The Big Five personality traits have been widely studied and are recognized as being stable across different languages, disciplines, and raters [ 47 , 48 ].

The Short Grit Scale, developed by Duckworth et al. (2007), measures perseverance and passion for long-term goals [ 49 ]. This scale consists of eight questions that evaluate student attitudes and behaviors towards long-term goals, such as the tendency to prioritize new ideas over existing plans [ 50 ]. The Short Grit Scale has demonstrated strong internal consistency, test–retest stability, and high predictive validity [ 51 ]. Grit is considered a facet of Big Five conscientiousness and has gained recent attention in the literature on human achievement. In this study, we utilized it for robustness testing.

Model design

Consider a statistical model that links a student's cognitive and non-cognitive abilities, boarding status, and other determinants of ability as represented by:

where \({Y}_{jis}\) denotes the cognitive and non-cognitive abilities of student i in school s in county j; \(Boardin{g}_{jis}\) is an indicator of the student's boarding status (1 if boarding, 0 otherwise), and \({W}_{jis}\) is a set of exogenous covariates that includes student (e.g., age and gender), family (e.g., parental education), and school (e.g., teacher qualifications and school facilities) characteristics; \({\mu }_{j}\) is county fixed effect; and the error term \({\varepsilon }_{jis}\) captures the influence of all unobserved factors.

Equation (1) may be subject to endogeneity issues for two main reasons. First, reverse causality may arise where students with lower cognitive abilities or academic performance could be more likely to choose boarding [ 14 ]. This concern is particularly true in cases such as the French excellent boarding school program, which is designed to provide elite education for disadvantaged groups. In contrast, boarding schools in rural China aim to integrate educational resources and are more likely to be chosen because of the distance between the student's home and school [ 26 , 27 , 52 ]. Therefore, reverse causality may not be a problem in this study. Second, omitted variables may also pose a problem, given that factors that affect students' cognitive and non-cognitive abilities may exist at multiple levels, and crucial indicators such as genetic factors and parental emotional involvement may be difficult to measure [ 27 , 53 , 54 , 55 ].

To address these problems, we use the standard instrumental variables (IV) approach to identify an exogenous source of variation in one's boarding status. The proportion of boarders of all students in a particular school is used as an instrumental variable for boarding. This strategy is based on the assumption that the proportion of boarders is a strong predictor of one's boarding status, because a higher proportion of boarders within a school indicates a higher likelihood for students to become boarders in that school. We employ a two-stage least squares (2SLS) framework to estimate Eq. (1) and the following first-stage equation.

The first-stage equation:

The second stage equation:

where \(Boardin{g}_{jis}\) is the proportion of boarders. The definitions of other variables are the same as in Eq. (1).

Distribution of cognitive abilities and non-cognitive abilities

Figure  1 shows the distribution of WISC-IV scores in our sample. The density distribution of WISC-IV scores is right-skewed for both boarding and non-boarding students compared to the norm, indicating a relatively high proportion of students with cognitive delays in our sample. Boarding students exhibit a less right-skewed distribution of WISC-IV scores compared to non-boarding students, suggesting that boarding students have higher WISC-IV scores on average. Fig.  2 shows the density distribution of Raven's IQ scores for the sample students. The estimated IQ scores on Raven's test for both boarding and non-boarding samples are not significantly different from the norm. Moreover, there is no significant difference between the Raven's IQ scores of boarding and non-boarding students. Finally, Fig.  3 illustrates the density distribution of standardized math scores for the sample students, suggesting that there is no significant difference between the boarding and non-boarding students.

Distribution of WISC IQ scores for sample students and a healthy population. The WISC IQ scores density distribution in the healthy population is a normal distribution with a mean of 100 and a standard deviation of 15

Distribution of Raven’s IQ scores for sample students and a healthy population. The Raven’s IQ scores density distribution in the healthy population is a normal distribution with a mean of 100 and a standard deviation of 15

Distribution of standardized math scores for sample students. Math scores are standardized

Table 3 presents the differences in cognitive and non-cognitive abilities between boarding and non-boarding students. The results indicate that boarders' WISC-IV scores were 2.45 points higher than non-boarders, significant at the 5% level, and boarders' fluid intelligence scores were 0.363 points higher than non-boarder, similarly significant at the 5% level. Boarders also scored higher on the matrix reasoning scale by 0.82 points at the 1% significance level. Additionally, boarders' standardized math scores were statistically significantly higher than non-boarders. In terms of non-cognitive skills, boarders scored higher in extraversion by 0.042, but lower in agreeableness by 0.042 compared to non-boarders.

Table 4 verifies the representativeness of the WISC-IV-tested students in the sample. We examine the differences in student characteristics between those who took the test and those who did not. The results indicate no significant differences in individual characteristics, family characteristics, and baseline math scores between the two groups.

Impact of boarding on cognitive abilities

Table 5 presents an analysis of the impact of boarding on cognitive abilities among rural students, specifically focusing on WISC IQ scores, fluid intelligence, and crystal intelligence. Using ordinary least squares (OLS) estimates in columns 1, 3, and 5, the results show that boarding does not have a significant effect on students' cognitive abilities. To further examine the causal relationship between boarding and cognitive abilities, two-stage least squares (2SLS) estimates are presented in columns 2, 4, and 6, and the findings indicate that boarding has no significant impact on WISC IQ scores, fluid intelligence, and crystal intelligence. The first stage regression has a high F-statistic of 41.284, indicating the exclusion of weak instrumental variables. To better understand how boarding affects students' cognitive abilities, we also estimated the impacts of boarding on the four subdimensions of WISC IQ scores, which are similarities, digit span, coding, and matrix reasoning. The results presented in Table 6 . The 2SLS estimates for boarding on students’ scores in digit span has a parameter estimate of 2.024, significant at the 5% level. Since scores in digit span is a test of attention and memory, the result highlights the positive impact of boarding on students' performance in this particular cognitive dimension.

Impact of boarding on students' non-cognitive abilities

Table 7 presents the effects of boarding on the personality traits of rural students, encompassing extraversion, agreeableness, dutifulness, neuroticism, and openness. The OLS results in columns (1), (3), (5), (7), and (9) suggest that while there is a positive relationship between boarding and the extraversion of rural students, the IV results indicate that boarding does not have a statistically significant effect on any of the five personality traits examined. Therefore, we conclude that boarding does not have any significant effects on the non-cognitive abilities of rural students.

Robustness test

To enhance the robustness of the research findings, we conducted additional regression analyses. First, we added the bootstrap method to the original instrumental variables method to re-estimate the impact of boarding on students' cognitive and non-cognitive abilities. The bootstrap method involves treating the observed sample as the entire population, and repeatedly resampling with replacement from the original sample to estimate the sampling distribution. This approach can provide an estimate of the distribution without introducing bias. In this paper, we conducted 1000 bootstrap samples and then used the instrumental variables method for estimation, which can provide more robust standard errors. Tables 8 and 9 show the results, which indicate that boarding still has a significant positive effect at the 10% level on students' scores in digit span and no significant effect on students' non-cognitive abilities, which is consistent with the results above.

Second, we performed robustness tests using Raven's IQ, standardized math, and grit scores as additional measures of cognitive and non-cognitive abilities. Raven's IQ and standardized math scores are measures of fluid and crystal intelligence, respectively. The results in Table 10 suggest that boarding does not have a significant effect on students' Raven's IQ and standardized math scores, which are consistent with previous findings. Grit is closely related to conscientiousness in the Big Five personality traits, and although the two are not identical, they share strong similarities such as diligence and perseverance [ 56 ]. It has also been shown that grit is a more refined measure of conscientiousness [ 57 , 58 ]. However, columns (5) and (6) of Table 10 show that boarding does not have a significant effect on students' grit scores, which are consistent with the previous results.

Heterogeneity analysis

In addition to analyzing the effects of boarding, we investigate the heterogeneity of these effects along three dimensions: gender, whether the student is a left-behind child, and family asset status. Family asset status is defined by ranking students' family asset indices from smallest to largest, and students in the top 25% of the sample are classified as having bad family conditions, with a dummy variable indicating whether a student's family conditions are bad. Table 11 presents the results of the heterogeneity analysis for cognitive abilities. The estimates in Panels A and B show that boarding has a significant positive effect on boys' WISC IQ scores, particularly in fluid intelligence, as evidenced by improvements in digit span and matrix reasoning. These findings suggest that boarding enhances boys' memory, attention, and reasoning abilities. Panels C and D indicate that boarding also has a significant positive impact on left-behind children's WISC IQ scores, again largely reflected in fluid intelligence. Boarding increases the digit span score (memory and attention) of left-behind children by 2.952 points (p < 0.05). However, the coding score for non-left-behind children is negatively affected, indicating that boarding reduces their transcription speed, accuracy, general learning ability, and anti-distraction ability. Panels E and F demonstrate that boarding has a significant positive effect at the 5% level on the digit span scores (memory and attention) of students with good family conditions, while there is no significant effect on students' cognitive abilities from low-income families.

The heterogeneity analysis of non-cognitive abilities is presented in Table 12 . Panels A and B show gender differences in the effect of boarding on students' non-cognitive abilities. Specifically, there is a significant positive effect of boarding on girls' conscientiousness at the 10% statistical level, indicating that boarding enhances girls' abilities in areas such as responsibility, workability, and self-control. However, no significant differences were found in the effect of boarding on non-cognitive abilities in other aspects. These findings suggest that boarding schools have varying effects on different dimensions of students' cognitive and non-cognitive abilities, highlighting the importance of considering heterogeneity in understanding the overall impact of boarding school education.

Conclusion and discussion

This study examines the impact of boarding on the cognitive and non-cognitive abilities of 5,660 junior high school students from 160 schools in rural northwest China using an instrumental variable (IV) approach. Our findings suggest that boarding has a significant positive effect on the digit span scores of junior high school students, which is a key component of the Working Memory Index in WISC, and suggests that boarding improves students’ memory and attention. However, we did not find any significant effects on other aspects of cognitive ability, such as logical thinking, reasoning, and transcription speed and accuracy. Additionally, we found no effects on non-cognitive abilities. To ensure the robustness of our findings, we conducted bootstrap and alternative variable tests, which supported our main results.

We used comprehensive indicators of cognitive abilities, namely crystal and fluid intelligence, to investigate the impact of boarding on rural students' cognitive abilities. Our analysis shows that boarding has a significant impact only on the ability of students to perform digit-span tasks, which is an important component of the Working Memory Index in the WISC. This finding suggests that boarding can positively affect students' memory and attention skills. One possible explanation for this result is that intensive and continuous learning or training can enhance an individual's cognitive abilities [ 59 , 60 , 61 ]. Moreover, boarders have a more collaborative learning and living environment in comparison to commuters, and they spend more time on general studies and homework [ 29 ]. Thus, boarding can significantly improve students' cognitive abilities. Furthermore, previous research has shown that the breadth of children's working memory increases linearly between the ages of 4 and 15 [ 62 ]. As boarders spend more time at school than commuters, they may have more opportunities to exercise their cognitive abilities, particularly in the area of attention and memory. Therefore, the positive effect of boarding on their cognitive abilities is likely to be reflected in their performance on digit span tasks. Most previous studies on the relationship between boarding and students' cognitive ability have often used academic performance (such as grades in math, language, and reading) as a proxy variable for cognitive ability. The conclusions drawn from these studies have been mixed. For instance, some studies suggest a positive effect of boarding on students' math performance in elementary school [ 19 , 23 ], while others demonstrate a significant negative influence on reading scores [ 26 , 28 ]. On the other hand, boarding has been linked to a positive effect on math and language scores among junior high school students [ 29 , 63 ]. Nonetheless, academic performance merely measures a certain dimension of students' cognitive ability, often reflecting crystallized intelligence and not offering a comprehensive evaluation of cognitive ability. From this perspective, this study employs the Wechsler test to more precisely gauge the impact of boarding on students' cognitive ability. In the existing literature, China Education Tracking Survey (CEPS) data is the only source that gauged students' cognitive ability through a unified scale (but not the Wechsler and Raven tests utilized in this study). Their findings indicated that boarding does not significantly impact students' cognitive abilities [ 63 ], which broadly aligns with the results of our study.

Moreover, regarding boarding and noncognitive ability, the current literature has primarily focused on mental health indicators as measures of noncognitive ability (e.g., depression, bullying, etc.). For instance, studies have pointed to potential psychological risks associated with boarding for students [ 27 , 28 , 64 , 65 , 66 ]. However, the Big Five personality traits measurement is an internationally recognized and widely employed tool for non-cognitive abilities. Notably, there is a scarcity of literature investigates the impact of boarding on students' non-cognitive skills within the framework of these five dimensions.

We conducted further analysis on the heterogeneity of the effects of boarding on the cognitive and non-cognitive abilities of rural junior high school students by gender, stay-at-home status, and family conditions. Our results indicate that the effect of boarding on the cognitive abilities of boys is greater than that of girls, particularly in fluid intelligence, as measured by digit span and matrix reasoning scales that show improved memory, attention, and reasoning abilities. This could be due to differences in time management skills between boys and girls, with boys benefitting from the external discipline and communal learning atmosphere in boarding schools [ 67 ]. For left-behind students, boarding has a greater impact on their cognitive abilities, particularly in the areas of memory and attention, as they lack parental engagement and receive more support from teachers [ 13 ]. Additionally, boarding can have a positive impact on the cognitive abilities of students from better-off families due to improved nutritional intake in school, which is associated with better cognitive development [ 68 ]. Existing research often uses academic performance as a proxy for cognitive abilities, which may not accurately capture the full range of cognitive skills. Furthermore, the positive effects of boarding on non-cognitive abilities were not observed in our study, suggesting the need for further investigation into the impact of boarding on non-cognitive development.

Our analysis of the heterogeneous effects of boarding on students' cognitive abilities also extends to their non-cognitive abilities. We find that boarding has a more significant impact on girls' conscientiousness of the Big Five personality traits, which is consistent with previous research on gender differences in conscientiousness [ 69 ]. This may be attributed to female personality traits and the role of gender as a moderating variable affecting individual commitment [ 70 ]. Social role theory also suggests that individuals of different genders develop different senses of group identity, causing them to behave differently in different social situations [ 71 ]. Therefore, greater commitment by girls to group characteristics may explain why they are more likely to be influenced by the communal learning atmosphere and external discipline that boarding schools provide. Interestingly, we did not find a significant effect of boarding on non-cognitive abilities among left-behind children or those from different family backgrounds. This finding supports our main estimation results and suggests that boarding may not have a substantial impact on the non-cognitive abilities of rural middle school students. Furthermore, we ruled out the possibility of positive and negative effects of boarding in different subsamples cancelling each other out, which strengthens this conclusion.

In conclusion, our study provides evidence that boarding schools is not detrimental to the development of new human capital, including cognitive and non-cognitive abilities, among rural junior high school students. Moreover, boarding positively affects some dimensions of students' cognitive abilities, particularly in the areas of memory and attention, and has a greater effect on certain subgroups such as boys, left-behind children, and students from better-off families. The growing prevalence of boarding schools in rural areas underscores the need to explore ways to further enhance students' cognitive skills and foster the development of non-cognitive or socio-emotional abilities in their daily boarding life. This issue deserves ongoing attention and efforts from educators and policymakers.

We do acknowledge one limitation of this study. While our analysis and findings indicate that boarding has not significantly impacted the cognitive and non-cognitive abilities of rural junior high school students, it remains possible that potential adverse effects are still present. It’s worth noting that the cognitive and non-cognitive skills evaluated in this paper using measurement scales may not fully encompass the nuances of various students’ behaviors tied to boarding. For instance, aspects like the dynamics between boarders and roommates, as well as the prolonged separation of boarders from parents, could potentially exert negative influences on their non-cognitive abilities, including interpersonal skills and emotional well-being [ 20 , 21 , 63 , 72 ]. This points to both the limitation of our current study and the necessity for further research.

Availability of data and materials

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Availability of methods

All methods were carried out in accordance with relevant guidelines and regulations.

WISC-IV is a clinical instrument created by David Wechsler in the United States. It is utilized to individually evaluate the intelligence of children between the aged 6 and 16 years and is internationally recognized as one of the most authoritative and valid personal intelligence tests. It was published in 1949 and updated to the fourth edition (WISC-IV) in 2003. In 2007, Prof. Houcan Zhang of the School of Psychology at Beijing Normal University, introduced the fourth edition to China. Leveraging Chinese census data, 1100 people were proportionally sampled from 12 provinces in mainland China, with 100 people in each age group, to establish the Chinese version of the norm (with a mean of 100 and a standard deviation of 15). Resulting in a calculated mean of 88.258 and a standard deviation of 10.948 for the students in our study’s samples. The average reliability coefficients of each synthetic score of the scale ranged from 0.87 to 0.97, meeting the standards set by the original American version. For this paper, the results of the Wechsler test were adjusted according to the Chinese norm [ 38 ].

Raven's Standard Progressive Matrices is designed by the British psychologist Raven in 1938 designed for people over age of 5.5 with normal intellectual development. The norms used in this paper are from the 1989 version (a sample of 5,108 students aged between 5 and 70 underwent testing to determine China’s Raven’s norm for the Raven test, with a mean of 100 and a standard deviation of 15) [ 41 ], as no updated norms have been established specifically for the Chinese population. Related studies in China have also used this test [ 42 , 43 , 44 ]. The split-half and retest reliabilities of the Raven test were 0.95 (p < 0.1) for the single-sample test and 0.82 (p < 0.1) for the retest. Concurrent validity was 0.71 (P < 0.1), and predictive validity was 0.45 (P < 0.1).

Heckman JJ, Stixrud J, Urzua S. The effects of cognitive and non-cognitive abilities on labor market outcomes and social behavior. J Law Econ. 2006;24:411–82.

Google Scholar  

Hanushek EA. Developing a Skills-Based Agenda for “New Human Capital” Research. SSRN Electron J. 2010. https://doi.org/10.2139/ssrn.1889200 .

Article   Google Scholar  

Farkas G. Cognitive skills and non-cognitive traits and behaviors in stratification processes. Ann Rev Sociol. 2003;29:541–62.

Cunha F, Heckman JJ, Lochner L, Masterov DV. Interpreting the evidence on life cycle skill formation. In Handbook of the Economics of Education, eds. E. Hanushek and F. Welch (Elsevier), 2006. https://doi.org/10.1016/S1574-0692(06)01012-9 .

Durlak JA, Weissberg RP, Dymnicki AB, Taylor RD, Schellinger KB. The impact of enhancing students’ social and emotional learning: A meta-analysis of school-based universal interventions. Child Dev. 2011;82:405–32.

Article   PubMed   Google Scholar  

Huang G, Xie Y, Xu H. Cognitive ability: Social correlates and consequences in contemporary China. Chinese Sociolog Rev. 2015;47:287–313.

Autor DH. Skills, education, and the rise of earnings inequality among the “other 99 percent.” Scie. 2014;344:843–51.

CAS   Google Scholar  

Bowles S, Gintis H. Schooling in Capitalist America Revisited. Sociol Educ. 2002;75:1–18.

Lleras Borghans L, Duckworth AL, Heckman JJ, ter Weel B. The economics and psychology of personality traits. J Human Res. 2008;43:972–1059.

Zhang D, Li X, Xue J. Education inequality between rural and urban areas of the People’s Republic of China, migrants’ children education, and some implications. Asian Dev Rev. 2015;32:196–224.

Xu H, Xie Y. The causal effects of rural-to-urban migration on children’s well-being in China. Eur Sociol Rev. 2015;31:502–19.

Ministry of Education. The Ministry of Education’s Response to Recommendation 3638 of the Fifth Session of the 12th National People’s Congress.2016. http://www.moe.edu.cn/jyb_xxgk/xxgk_jyta/jyta_jijiaosi/201801/t20180109_324186.html

Martin AJ, Papworth B, Ginns P. Liem GA Boarding school, academic motivation and engagement, and psychological well-being: a large-scale investigation. Am Educ Res J. 2014;51:1007–49.

Behaghel L, de Chaisemartin C, Gurgand M. Ready for boarding? The effects of a boarding school for disadvantaged students. Am Econ J Appl Econ. 2017;9:140–64.

Ak L, Sayil M. Three different types of elementary school students’ school achievements, are perceived social support, school attitudes, and behavior-adjustment problems. Kuram ve Uygulamada Egitim Bilimleri. 2006;6:293.

Zhang X, Li Y, Zhang Q, Lu F, Wang Y. Smoking and its risk factors in Chinese elementary and middle school students: A nationally representative sample study. Addict Behav. 2014;39:837–41.

Granot D, Mayseless O. Attachment security and adjustment to school in middle childhood. Int J Behav Dev. 2001;25:530–41.

Steel A, Erhardt R, Phelps R, Upham P. Estimates of enhanced outcomes in employment, income, health, and volunteerism for the association of boarding schools member school graduates. J Adv Acad. 2015;26:227–45.

Curto VE, Fryer RG. The potential of urban boarding schools for the poor: evidence from SEED. J Law Econ. 2014;32:65–93.

Evans-Campbell T, Walters KL, Pearson CR, Campbell CD. Indian boarding school experience, substance use, and mental health among urban Two-Spirit American Indian/Alaska Natives. Am J Drug Alcohol Abuse. 2012;38:421–7.

Article   PubMed   PubMed Central   Google Scholar  

Pfeiffer JP, Pinquart M. Bullying in German boarding schools: a pilot study. Sch Psychol Int. 2014;35:580–91.

Mander DJ, Lester L, Cross D. The social and emotional well-being and mental health implications for adolescents transitioning to secondary boarding school. Int J Child Adolesc Health. 2015;8:131.

Qiao T, Di L. The causal inference of boarding’s effect in rural primary and secondary school education. J Soc Develop. 2014;1:138–52+245 (in Chinese).

Mo D, Yi H, Zhang L, Shi Y, Rozelle S, Medina A. Transfer paths and academic performance: the primary school merger program in China. Int J Educ Dev. 2012;32:423–31.

Wang A, Medina A, Luo R, Shi Y, Yue A. To board or not to board: Evidence from nutrition, health and education outcomes of students in rural China. Chin World Econ. 2016;24:52–66.

Li X, Zhu Z, Song Y, Wu Y. The impact of boarding on the reading ability of rural children in poor areas in China: Evidence from 137 boarding schools in Hebei and Sichuan provinces. China Rural Surv. 2018;2:129–44.

Zhu Z, Li Y, Song Y. Boarding education and children’s development: evidence from 137 rural boarding schools. Educ Res. 2019;8:79–91 (in Chinese).

Hou H, Wu Y, Song Y. Low-age boarding and rural pupils’ human capital accumulation: evidence from the “Primary School Merger Program.” Chin Rural Econ. 2018;7:113–29.

Yao S, Gao Y. Can large scale construction of boarding schools promote the development of students in rural area better. Educ Econ. 2018;34:53–60 (in Chinese).

Tang B, Wang Y, Gao Y, Wu S, Li H, Chen Y, et al. The effect of boarding on the mental health of primary school students in Western Rural China. Int J Environ Res Public Health. 2020;17:8200.

Nisbett RE, Aronson J, Blair C, Dickens W, Flynn J, Halpern DF, et al. Intelligence: new findings and theoretical developments. Am Psychol. 2012;67:130.

Kautz T, Heckman JJ, Diris R, Weel B, Borghans L. December. Fostering and Measuring Skills: Improving Cognitive and Non-Cognitive Skills to Promote Lifetime Success. National Bureau of Economic Research.2014. https://doi.org/10.3386/w20749 .

Roberts RD, Goff GN, Anjoul F, Kyllonen PC, Pallier G, Stankov L. The armed services vocational aptitude battery (ASVAB): Little more than acculturated learning (Gc)!? Learn Individ Differ. 2000;12:81–103.

Raven J, Raven JC, Court J. Manual for Raven’s Progressive Matrices and Vocabulary Scales. San Antonio: Harcourt Assessment; 1988.

Costa PT, McCrae RR. Four ways five factors are basic. Personality Individ Differ. 1992;13:653–65.

Cattell RB. Intelligence: Its structure, growth and action. Oxford, England: North-Holland; 1987.

Wechsler D. Wechsler intelligence scale for children—4th edition (WISC-IV®). San Antonio, TX: The Psychological Corporation; 2003.

Zhang H. The revision of WISC-IV Chinese version. Psychol Sci. 2009;32:1177–9 (in Chinese).

Keith TZ, Fine JG, Taub GE, Reynolds MR, Kranzler JH. Higher order, multisample, confirmatory factor analysis of the Wechsler Intelligence Scale for Children—Fourth Edition: What does it measure. Sch Psychol Rev. 2006;35:108–27.

Borghans L, Golsteyn BH, Heckman JJ, Humphries JE. What grades and achievement tests measure. Proc Natl Acad Sci. 2016;113:13354–9.

Article   PubMed   PubMed Central   CAS   Google Scholar  

Zhang H, Wang X. Standardization research on Raven’s standard progressive matrices in China. Acta Psychologica Sinica. 1989;2:113–21 (in Chinese).

Fan S, Liu H. Analysis of potential types of responses in raven advanced reasoning test. Psycholog Explor. 2016;36:257–63 (in Chinese).

Li C, Luo X, Wang T, Xu F. Gender differences in Raven’s reasoning test. Adv Psychol Sci. 2011;19:1076–82 (in Chinese).

Li H, Zheng Y. Analysis of combined raven testing and intervention effects on children with learning difficulties. Western Med. 2009;21:2129–30 (in Chinese).

Gutman LM, Schoon I. A Synthesis of causal evidence linking non-cognitive skills to later outcomes for children and adolescents. Brill Available. 2016. https://doi.org/10.1007/978-94-6300-591-3_9 .

Patrick CL. Student evaluations of teaching: effects of the Big Five personality traits, grades and the validity hypothesis. Assess Eval High Educ. 2011;36:239–49.

John OP, Naumann LP, Soto CJ. Paradigm shift to the integrative Big Five trait taxonomy: history, measurement, and conceptual issues. The Guilford Press; 2008.

Wang M, Dai X, Yao SQ. Development of Chinese Big Five Personality Inventory (CBF-PI)—Theoretical framework and relability analysis. Chin J Clin Psychol. 2010;18:545–8.

Duckworth AL, Peterson C, Matthews MD, Kelly DR. Grit: Perseverance and passion for long-term goals. J Pers Soc Psychol. 2007;92:1087–101.

Harms PD. Grit: the power of passion and perseverance. Pers Psychol. 2016;69:1021–4.

Duckworth AL, Quinn PD. Development and validation of the short grit scale (Grit–S). J Pers Assess. 2009;91:166–74.

Chen Q, Chen Y, Zhao Q. Impacts of boarding on primary school students’ mental health outcomes – Instrumental-Variable evidence from rural northwestern China. Econ Hum Biol. 2020;39:100920.

Jensen AR. Raising the IQ: the Ramey and Haskins study. Intelligence. 1981;5:29–40.

Gronqvist E, Ockert B, Vlachos J. The Intergenerational Transmission of Cognitive and Non-Cognitive Abilities. Working Paper. 2010. https://doi.org/10.2139/ssrn.1627657 .

Rimfeld K, Kovas Y, Dale PS, Plomin R. True grit and genetics: predicting academic achievement from personality. J Pers Soc Psychol. 2016;2016(111):780.

Eskreis-Winkler L, Duckworth A, Shulman E, Beal S. The grit effect: predicting retention in the military, the workplace, school and marriage. Front Psychol. 2014. https://doi.org/10.3389/fpsyg.2014.00036 .

MacCann C, Duckworth AL, Roberts RD. Empirical identification of the major facets of conscientiousness. Learn Individ Differ. 2009;19:451–8.

Eisenberg N, Duckworth AL, Spinrad TL, Valiente C. Conscientiousness: Origins in childhood? Dev Psychol. 2014;50:1331–49.

Ceci SJ, Liker JK. A day at the races: A study of IQ, expertise, and cognitive complexity. J Exp Psychol Gen. 1986;115:255–66.

Jacobson MJ. Problem solving, cognition, and complex systems: differences between experts and novices. Complexity. 2001;6:41–9.

Chang Y. Reorganization and plastic changes of the human brain associated with skill learning and expertise. Front Hum Neurosci. 2014. https://doi.org/10.3389/fnhum.2014.00035 .

Gathercole SE, Pickering SJ, Ambridge B, Wearing H. The structure of working memory from 4 to 15 years of age. Dev Psychol. 2004;40:177–90.

Zhou J, Xu N. The impact of boarding on student's academic achievement, cognitive ability and non-cognitive ability in junior high school. Educ Sci Res. 2021;5:53–9 (in Chinese).

Liu W, Hou W. A study on the mental health of junior first-year high school students and its influencing factors ——An analysis based on China’s educational follow-up survey(CEPS). J Shanghai Educ Res. 2018;04:39–43+48 (in Chinese).

Kleinfeld J, Bloom J. Boarding schools: effects on the mental health of Eskimo adolescents. Am J Psychiatry. 1977;134:411–7.

Article   PubMed   CAS   Google Scholar  

Shen Z, Chen L, Cui J, et al. Influencing factors of mental health of boarding and non-boarding junior high school students. Educ Meas Eval. 2012;8:38–42 (in Chinese).

Li B, Yang J. Relations of time management disposition self-efficiency attribution of junior school student to academic achievement. Psychol Expl. 2004;4:67–71 (in Chinese).

Fang H, Ma G. Nutrition problems and countermeasures for students in rural boarding schools in China. Chin J School Health. 2010;31:1039–40 (in Chinese).

Kling KC, Noftle EE, Robins RW. Why do standardized tests underpredict women’s academic performance? The role of conscientiousness. Soc Psycholog Personal Scie. 2013;4:600–6.

Vianello M, Robusto E, Anselmi P. Implicit conscientiousness predicts academic performance. Personal Individ Differ. 2010;48:452–7.

Eagly AH, Wood W. Explaining sex differences in social behavior: a meta-analytic perspective. Pers Soc Psychol Bull. 1991;17:306–15.

Wang S, Mao Y. The impact of boarding on social-emotional competence of left-behind children: an empirical study in 11 provinces and autonomous region in western China. J Educ Studies. 2015;11:111–20 (in Chinese).

Download references

Acknowledgements

We thank the researchers from Center for Experimental Economics in Education, Shaanxi Normal University for help on developing the study protocol and training the enumerators. We are also grateful to all respondents who participated in this study and the enumerators for data collection efforts.

This work was supported by National Natural Science Foundation of China [Grant No.72103114], National Natural Science Foundation of China [Grant No.71933003; Grant No.72373085], and 111 Project [Grant No. B16031].

Author information

Authors and affiliations.

Center of Experimental Economics in Education, Shaanxi Normal University Shaanxi Province, No. 620 West Chang’an Street, Chang’an District, Xi’an, 710119, China

Fang Chang, Yanan Huo, Songyan Zhang, Hang Zeng & Bin Tang

You can also search for this author in PubMed   Google Scholar

Contributions

FC and YNH contributed to the study design, data analysis and manuscript draft. BT gave critical feedback and conducted quality control. SYZ and HZ revised the manuscript. All authors have read and approved the manuscript.

Corresponding author

Correspondence to Bin Tang .

Ethics declarations

Ethics approval and consent to participate.

The Stanford University Institutional Review Board (IRB Protocol ID #19748) approved this study. Permission was received from local Boards of Education in each region and the principals of all schools. Written informed consent was obtained from at least one parent and/or legal guardian for all child participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Chang, F., Huo, Y., Zhang, S. et al. The impact of boarding schools on the development of cognitive and non-cognitive abilities in adolescents. BMC Public Health 23 , 1852 (2023). https://doi.org/10.1186/s12889-023-16748-8

Download citation

Received : 18 May 2023

Accepted : 12 September 2023

Published : 23 September 2023

DOI : https://doi.org/10.1186/s12889-023-16748-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

BMC Public Health

ISSN: 1471-2458

• General enquiries: [email protected]

• DOI: 10.3389/fpsyg.2024.1359626
• Corpus ID: 268849515

The impact of boarding school on student development in primary and secondary schools: a meta-analysis

• Zhiyong Zhong , Yang Feng , Yongqi Xu
• Published in Frontiers in Psychology 28 March 2024
• Education, Economics

One Citation

Learning accumulation from unfamiliar environments: informal focus groups, 55 references, attending boarding school: a longitudinal study of its role in students’ academic and non-academic outcomes, boarding school, academic motivation and engagement, and psychological well-being.

• Highly Influential

An Examination of Secondary School Students' Academic Achievement in Science Course and Achievement Scores in Performance Assignments with Regard to Different Variables: A Boarding School Example.

A study on the academic achievement and school adaptability of rural elementary school boarding students from five provinces and autonomous regions in western china, the effect of layout adjustment of rural schools on student achievement——analysis based on the two-level value-added model, impacts of boarding on primary school students' mental health outcomes - instrumental-variable evidence from rural northwestern china., the impact of boarding on social-emotional competence of left-behind children:an empirical study in 11 provinces and autonomous region in western china, on the development of personality and sociality in rural elementary boarding schools, bullying in german boarding schools: a pilot study, ready for boarding the effects of a boarding school for disadvantaged students, related papers.

Showing 1 through 3 of 0 Related Papers

1st Edition

The Psychological Impact of Boarding School The Trunk in the Hall

• Taylor & Francis eBooks (Institutional Purchase) Opens in new tab or window

Description

The Psychological Impact of Boarding School is a collection of research-based essays answering a range of questions about boarding school and its long-term impact. Through a combination of original in-depth first-person narratives as well as larger scale surveys, this book aims to fill gaps in current boarding school research and present new findings. Topics addressed include gender differences, eating behaviours, loneliness, mental health and relationships, the differences between younger and older boarders, and ex-boarder experiences of therapy. The research results highlight a key role in the age that children start boarding, the way that long-term psychological influences of friendships formed at school, and the larger role that parent and family relationships play in the psychological lives of boarders. Through these findings, the book ultimately challenges the current understanding of  'boarding school syndrome', proposing a move beyond the term and its concept. The book will appeal to psychologists, psychoanalysts, counsellors, academics, teachers, current and ex-boarders as well as parents and guardians interested in the impact of boarding schools from either a professional or a personal perspective.

Table of Contents

Penny Cavenagh, PhD, is currently a professor of health research and enterprise at the University of Suffolk, visiting professor at the University of Essex and a chartered psychologist. She has published extensively in books and academic journals in the field of medical management, medical education and dysfluency. She boarded for four years at an all-girls school in the early 1970s. Susan McPherson, PhD, is currently a professor of psychology and sociology at the University of Essex and has been a researcher in the field of mental health for 25 years after studying at the London School of Economics, University College London and Kings College London. She was a full-time boarder from age 8 to 18 during the 1980s and 1990s including three years at a junior boarding school. Jane Ogden, PhD, is currently a professor of health psychology at the University of Surrey where she has been teaching for the last 18 years. She has written 8 books relating to health psychology, eating behaviour, weight management, parenting and critical thinking and over 230 research papers on many aspects of health psychology.

Critics' Reviews

‘A scholarly and nuanced study on the long-term psychological impacts of British boarding schools: while some people thrive in boarding schools, for too many others the adverse psychological impacts last a lifetime.’ Benedict Rattigan , author and ex-boarder at Eton 'Boarding is often misunderstood and misrepresented, but this invaluable book provides a wide, and evidence-informed, analysis of the sector. Importantly, the authors allow the voices of the former pupils themselves to speak and in doing so reveal a nuanced and complex series of personal narratives. This is a book that should be read not only by those who work in boarding schools, but by anyone involved in education and child psychology.' David James , co-editor of The State of Independence

About VitalSource eBooks

VitalSource is a leading provider of eBooks.

• Access your materials anywhere, at anytime.
• Customer preferences like text size, font type, page color and more.
• Take annotations in line as you read.

Multiple eBook Copies

This eBook is already in your shopping cart. If you would like to replace it with a different purchasing option please remove the current eBook option from your cart.

Book Preview

The country you have selected will result in the following:

• Product pricing will be adjusted to match the corresponding currency.
• The title Perception will be removed from your cart because it is not available in this region.

Articles on Boarding schools

Displaying all articles.

American Indians forced to attend boarding schools as children are more likely to be in poor health as adults

Ursula Running Bear , University of North Dakota

Christianity was a major part of Indigenous boarding schools – a historian whose family survived them explains

Brenda J. Child , University of Minnesota

Acts of violence or a cry for help? What fuels Kenya’s school fires

Teresa A. Wasonga , Northern Illinois University

‘Truth and Healing Commission’ could help Native American communities traumatized by government-run boarding schools that tried to destroy Indian culture

David R. M. Beck , University of Illinois at Urbana-Champaign

How Native students fought back against abuse and assimilation at US boarding schools

Sarah Klotz , College of the Holy Cross

Some private schools need to change their models — they were losing students even before COVID

Paul Kidson , University of Wollongong

We need to know the true cost of Indigenous boarding school scholarships on communities

Jessa Rogers , University of Canberra

A burning question: why are Kenyan students setting fire to their schools?

Elizabeth Cooper , Simon Fraser University

Send disadvantaged pupils to boarding school and only the brightest thrive

Clement de Chaisemartin , University of Warwick

Related Topics

• Colonization
• education inequity
• Federal Indian boarding school system
• Language loss
• Native Americans
• Religion and society

Top contributors

Assistant Professor in the Department of Economics, University of Warwick

Assistant Professor of International Studies, Simon Fraser University

Senior Lecturer in Educational Leadership, Australian Catholic University

Professor of American Studies, University of Minnesota

Assistant Professor of English, College of the Holy Cross

Professor of History, University of Illinois at Urbana-Champaign

Professor, Educational Administration, Northern Illinois University

Assistant Professor of Population Health, University of North Dakota

First Nations Senior Research Fellow, Queensland University of Technology

• X (Twitter)
• Unfollow topic Follow topic

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Int J Environ Res Public Health

The Effect of Boarding on the Mental Health of Primary School Students in Western Rural China

1 Center for Experimental Economics in Education (CEEE), Shaanxi Normal University, Xi’an 710119, China; nc.ude.unns@nibgnat (B.T.); moc.361@eeeceuygnaw (Y.W.); [email protected] (Y.G.); moc.361@eeecnijihsuw (S.W.); moc.361@eeecilgnayoah (H.L.)

2 School of Economics and Finance, Xi’an Jiaotong University, Xi’an 710061, China

3 Food and Resource Economics Department, University of Florida, Gainesville, FL 32611, USA

4 School of Foreign Studies, Xi’an Jiaotong University, Xi’an 710061, China; nc.ude.utjx.uts@020031477a

Yaojiang Shi

Based on the panel data of 20,594 fourth- and fifth-grade students in the western provinces A and B in China, this paper analyzed the effect of boarding at school on the mental health of students using a combination of the propensity score matching (PSM) and difference-in-differences (DID) methods. The results showed that boarding had no significant effect on the mental health of students, but the tendency of loneliness among boarding school students was increased. Heterogeneity analysis found that fifth-grade students whose parents had both left home to work were more likely to have poorer mental health when boarding. This paper has essential policy significance for guiding rural primary schools to improve the mental health of boarding school students, especially left-behind children.

1. Introduction

For individuals, receiving good education is conducive to the improvement of individual human capital and income level; For the country, the development of education is also conducive to the overall level of human capital and economic transformation and development [ 1 , 2 ]. Many developing countries in the world have attached importance to the construction of school institutions at the basic education stage. For example, Duflo (2001) conducted an empirical study on the large-scale establishment of schools in Indonesia in the last century showing that if there is one more primary school for every 1000 children on average, the average years of schooling will increase by 0.12–0.19 years [ 3 ]. Berlinski, Galiani, and Gertler (2009) based on Argentina’s study also found that the construction of pre-school education institutions significantly contributed to primary school students’ academic performance [ 4 ].

China’s rapid progress in education in recent decades has benefited from its long-term investment in basic education. In the 1990s, China implemented the national compulsory education project and realized the pattern of “one village, one school” nationwide. However, since the mid-1990s, the number of school-age children in rural areas continued to decline, “one village, one school” school scale becoming small, and high management costs, staffing difficulties, and other issues were increasingly prominent [ 5 ]. The change of population situation brought about increase of management cost and difficulty of rural primary schools, which became the motivation of rural school layout adjustment. In 2001, China’s government issued the “Decision of the State Council on the Reform and Development of Basic Education” and proposed “adjusting the layout of rural compulsory education schools in line with local conditions”, which started a new round of adjustment of the layout of primary and secondary schools in rural areas, also known as the “withdrawal and consolidation of schools” [ 6 ].

In 2001, the State Council’s Decision on Reform and Development of Basic Education pointed out that the layout of rural compulsory education schools should be adjusted according to local conditions [ 7 ]. Since then, the movement “withdrawal and consolidation of schools” began and swept across China’s rural primary and secondary schools. The implementation of this policy to some extent did integrate rural education resources, reduce the cost of education per student, but also increased the number of boarding schools [ 8 ]. Since 2001, with the continuous development and in-depth reform of China’s education, as well as the needs of a large number of left-behind children in rural areas whose parents leave to work in urban areas, boarding schools have developed rapidly in rural areas of China and thus the number of boarding school students has quickly increased [ 9 ]. By the end of 2015, the total number of boarding school students across rural primary and secondary schools reached 26.363 million. The boarding rates of primary school and junior high school students were 14.4% and 58.6%, respectively. In the western rural areas, the proportion of primary school boarders reached 21.1% [ 10 ].

Although it is relatively easy to monitor the physical health status of students through more frequent and regular school physical examinations, it is more difficult to detect levels and problems of mental health [ 11 ]. A series of devastating suicides cases due to mental health problems from students in the early 2000s have prompted the Chinese government to make considerable efforts to reduce mental health problems in schools at all levels [ 12 , 13 ]. China’s government attaches great importance to the development of mental health among adolescents. The “Guidance on Strengthening Mental Health Services” report issued in December 2016 pointed out that “in order to comprehensively strengthen mental health education for children and adolescents, primary and secondary schools should pay attention to students’ mental health education, cultivate positive, optimistic, and healthy psychological qualities, and promote the sustainable development of students’ mind and body” [ 14 ]. The rapid development of the boarding system in rural primary and secondary schools in China has increased concern and discussion regarding boarding school students’ mental health.

Theoretically, boarding has both beneficial and detrimental effects on adolescent growth. On the positive side, boarding, a kind of collective life, to some extent, may reduce the risk of psychological problems on students. For example, group living environment supervised by school-assigned student caregivers may help maintain the health and safety of boarding students [ 15 ]. For those students who do not have enough family care, they may be able to get better care and life at school than in their families, which may contribute to their healthy development [ 16 ]. Moreover, psychological counseling from teachers and communication with peers can help disadvantaged students to overcome the troubles and psychological problems and other common challenges experienced by students [ 17 ]. On the negative side, boarding implying the lack of care from family may worsen students’ mental health. Students in the primary school education stage are in a critical period of growth and development and boarding school students are separated from their parents for a long time, thus it is difficult for them to receive daily care from their parents and families during that period [ 18 ]. The collective living environment of boarding schools is also likely to resulting in students being more affected by the bad habits and behaviors of other students, and even makes them more susceptible to suffering from bullying, both of which would have a long-term negative impact on students’ physical and mental development [ 19 , 20 ]. Serious mental health problems may hurt students’ academic performance, which is not conducive to the improvement of China’s human capital [ 21 ].

In recent years, the number of boarding schools in China has increased rapidly and a few studies have paid attention to the effect of boarding on students’ mental health status. However, from empirical literature, research on the impact of boarding on students’ mental health is inconsistent. A couple of studies have shown that the impact of boarding on school students’ mental health is negative. Evans-Campbell et al. (2012) believed that individuals that have been enrolled in boarding schools or have been raised by a person attending a boarding school are more likely to develop significant anxiety disorders and post-traumatic stress disorder than other individuals, and are even more likely to develop suicidal ideations [ 22 ]. Another study has shown that boarding hurts students’ mental health, and that the negative effects do not disappear due to short-term [ 23 ]. Ma et al., (2013) found that the overall mental health of boarding school students was poor by analyzing the variances in fifth- and sixth-grade primary school students. Primary school students’ self-blame tendency, learning adaptation independence, learning anxiety, allergy tendency, physical symptoms, and mental health test (MHT) total scores were significantly higher than those of non-boarding primary school students [ 24 ]. Chen et al., (2020) adopted instrumental variable to examine the effect of boarding and found to have negative effects on a few dimensions of students’ mental health (i.e., study anxiety, social anxiety, self-punishment, physical anxiety symptoms, and fear) reaching at 0.455 SDs (standard deviations) [ 11 ]. In contrast, some studies have shown that boarding either has no significant effect or has uncertain effects on the mental health of students. For example, Shen et al., (2012) used first- and second-year students in junior high school as research samples. He used multiple regression analysis to determine that there was no significant difference in the average scores of the full-scale MHT(Mental Health Test) and the eight subscales between boarding and non-boarding school students indicating that there was no significant difference in the mental health status between boarding and non-boarding school students [ 25 ]. Liu and Villa (2020) detected that though students who board in schools have improved their academic scores, their mental health level changed little [ 26 ]. Liu et al. (2004) used the psychotic symptom self-assessment form (SLE-90) to evaluate boarding and non-boarding high school students. They found that although boarding school students have more psychological problems in the initial enrollment stage, the psychological problems in the upper grades gradually decrease and the psychological health increases [ 27 ].

Overall, the research literature in this area is quite rich, but there are still a few obvious problems that need to be further solved: First, the research methods used are mostly one-dimensional or multiple regression models and cross-sectional data are mostly used, which fail to solve the endogeneity problem caused by the self-selection bias and the omitting relevant variable bias. For example, Ma et al. (2013) and other researchers studied only 900 fifth- and sixth-grade students and adopted a simple variance analysis method. Second, the definition of boarding in most of the literature only examines whether students are boarding at a certain time and does not consider the change in boarding status of students across a period of time. For example, Shen et al. (2012) used multiple regression and distinguished between boarding and non-boarding school students purely based on each student’s boarding status during the survey. Since the time at which different rural children can board is different, it is reasonable to suspect that those students who board earlier are likely to experience other exogenous events in later school life. Without identification and divestiture, it is likely that the negative effects of boarding behavior will be overestimated [ 27 ]. Another important deficiency is that the existing research samples are mostly from either a single or a few schools in a single province, and the sample sizes are small and non-representative. This paper will try to cope with the deficiency list above using large, representative data and a combination of Propensity Score Matching (PSM) and the Difference-in-Differences (DID) model to overcome endogenous problems.

Based on the panel data of 20,594 fourth- and fifth-grade students in the western provinces A and B, this paper used a combination of Propensity Score Matching (PSM) and the Difference-in-Differences (DID) model to correct the endogenous problems that may exist in the model, and to provide evidence for the effect of boarding on the mental health of rural primary school students.

The rest of the paper is organized as follows: Section 2 introduces the data and empirical models used in this paper; Section 3 shows the results of the model, including the impact of boarding on the mental health of primary school students, and heterogeneity analysis of the results; Section 4 provides the conclusion and policy recommendations.

2. Data and Methodology

The tools and data used in this study were from a sample survey conducted by the Center for Experimental Economics in Education of Shaanxi Normal University in the western provinces A and B (We are using A and B instead of the provinces’ names for anonymity purposes). The survey used a stratified random sampling method. First, 11 counties in a certain city of province A were selected, 7 counties in a certain city of province B were selected, and then 252 rural primary schools were randomly selected from the sample counties. A total of 20,594 students in the fourth and fifth grades were randomly selected. The research team conducted a baseline survey and a follow-up survey of a sample of the 20,594 students in September 2012 and May 2013, respectively. Since some students had transferred to other schools or did not present on the day of the follow-up survey, part of the sample was lost, and finally, 16,685 fourth- and fifth-grade students from 251 schools participated in the follow-up survey. In order to make out the difference between retained samples and lost samples, we conducted a comparison check between two parts of sample. As shown in Table 1 the mean and SD of most students’ characteristics between retained samples and lost samples are similar though some characteristics show difference (gender, grade, family assets, baseline math score).

Comparison of the characteristics of students between retained samples and lost samples.

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. SD, standard deviation. Data source: Authors’ survey.

2.2. Data Description and Variables

A standardized questionnaire survey was conducted among students in the fourth and fifth grades. The questionnaire collected the following information from the students: (1) mental health, (2) boarding status, and (3) family background and socioeconomic status.

2.2.1. Students’ Mental Health

This paper used the MHT to examine students’ mental health. This test was devised by Bucheng Zhou, a professor of the psychology department at East China Normal University, based on the “Uneasy Propensity Diagnostic Test” compiled by Suzuki Kiyoshi et al, from Japan [ 28 ]. The internal consistency coefficient of the full scale is 0.91, which has been widely used in the measurement of mental health of school-aged children in China.

The test has a total of 100 “yes” or “no” questions; selecting “yes” corresponds to 1 point, while “no” to 0 points. The 100 test questions can be divided into two parts, namely, the lying scale and the content scale. The lying scale consists of 10 questions; the higher the score, the higher the probability that the student will provide false information when completing the test. In this study, questionnaires with a score higher than 7 on the lying scale were invalidated (The samples used for analysis later have eliminated the samples which were invalidated). The content scale includes 8 sub-scales: learning anxiety, anxiety about people, the loneliness tendency, the self-blame tendency, the allergic tendency, physical symptoms, the horror tendency, and the impulsive tendency. The higher the sub-scale score, the worse the student’s performance in this dimension. The total score of 8 eight sub-scales indicates the overall level of mental health of the students—a high score corresponds to higher risk for mental health problems. This paper used standardized MHT scores to measure the students’ level of mental health.

The mean of the MHT standardized scores of the students in the sample is shown in Figure 1 . It can be seen that in 2012 and 2013, levels of mental health between boarding students and non-boarding students are almost same. Figure 2 and Figure 3 show the distribution of the MHT standardized scores in the samples from 2012 and 2013.

Mean of standardized students’ mental health test (MHT) scores.

Distribution of students’ standardized MHT scores in samples from 2012.

Distribution of students’ standardized MHT scores in samples from 2013.

2.2.2. Students’ Boarding Status

This paper aimed to examine the impact of boarding on students’ mental health through the changes in the two-year data. Therefore, students who did not board in the baseline period were selected (13,638 samples). Students who were non-boarding in the baseline period (2012) would be in two states in the follow-up period (2013): boarding in 2013 (1020 samples) and non-boarding in 2013 (12,618 samples). This paper aimed to find out the impact of boarding through the difference between boarding students and non-boarding students in the second year. In this way, we can estimate the net effect of a year of boarding on students’ mental health, excluding other factors affecting mental health. For example, if we do not rule out the baseline period of boarding samples, it will be not clear that the change of mental health is stemmed from one-year boarding or from other aspects since students have been boarding before.

To ensure that all students at baseline had the same boarding status (non-boarding), this paper removed 3047 boarding school students from the baseline and finally collected 13,638 samples from the remaining 200 schools for research. According to the boarding status of the remaining sample in the follow-up period, in this paper, the students who were not boarding in the baseline period (2012) but were boarding in the follow-up period (2013) were defined as the treatment group. In addition, students who were not boarding in the baseline (2012) or follow-up (2013) periods were defined as the control group. As can be seen in Table 2 , 81.74% of the students in the sample were not boarding in the baseline period, among which 75.62% were still not boarding during the follow-up period, and 6.11% of the students started boarding during the follow-up period.

Boarding status percentages.

Data source: Authors’ survey.

2.2.3. Students’ Family Background and Socio-Economic Status

In addition to boarding status, the student questionnaire collected basic personal and family information. The students’ basic personal information included gender, age, grade, whether or not they had myopia, their mathematics scores at baseline, and the distance from the school to the county. Before the questionnaire survey, the research team tested the students with internationally used mathematics test questions, and the investigators controlled the time and order on the spot to prevent cheating. The family information mainly included the family’s economic situation, the education level of parents, and the parents’ migration status for work. In the questionnaire, the students were asked whether their family had a business, whether they had computers, whether they had internet access, and whether they had electric cars, automobiles, running water, microwave ovens, refrigerators, cameras, video cameras, washing machines, and flushing toilets. Family assets were measured by the factor score, which was calculated by factor analysis according to the students’ answers to the above questions.

There were some significant differences in individual and family characteristics between the treatment and control groups ( Table 3 ). By directly comparing the individual and family characteristics of the treatment and control groups, it was found that, first, in terms of the individual characteristics of the students, the treatment group was mostly male and their mathematics scores at the baseline were lower. Second, as far as the family background is concerned, the students in the treatment group had a better family financial situation, the mothers were less educated, and the proportion of parents migrating to urban areas for work was low. In addition, the school where the treatment group was located was far from the county.

Comparison of the characteristics of students with different boarding status at baseline.

In summary, there were some significant differences between the treatment and the control groups at the individual and family levels. These differences may not only be related to students’ boarding status, but also to students’ mental health level. Therefore, it is necessary to control these factors and the endogenous problems generated by them when analyzing the impact of boarding on students’ mental health level.

3.1. Propensity Score Matching (PSM)–Difference-in-Differences (DID) Method

According to Table 3 , there were significant differences in individual characteristics and family economic background between the students in the treatment and control groups. These differences led to endogenous problems in the study, such as missing variable errors caused by unobservable factors and self-selection errors caused by observable factors. Therefore, this paper attempted to solve the above two problems by using a combination of propensity score matching (PSM) and the difference-in-differences (DID) model.

The DID model is a quantitative statistical method, which is applied to at least two periods of data. It simulates the samples with changed research behavior between two periods of data as the “experimental group” in the real experiment and the samples without change as the “control group” in the real experiment to explore the impact of the changes in research behavior on the outcome variables [ 29 , 30 ]. Specifically, in this study, when exploring the effect of boarding on students’ mental health, we chose students who did not board in the baseline period as the analysis sample (N = 13,638), and divided this sample into two groups: one group consisted of students who changed from non-boarding to boarding, as the “experimental group” (N = 1020); the other group consisted of the students who maintained their non-boarding status, as the “control group” (N = 12,618). The idea of applying DID method is to compare the difference in the change of an output variable between the experimental group and the control group before and after the implementation of a policy or before and after a certain behavior change.

The change in boarding behavior may be self-selection rather than random occurrence. If it is not treated and the sample is directly regressed, the parameter estimation will be biased. Economists proposed the propensity score matching method to reduce the error problem in the observation data set [ 31 ]. To avoid possible selection bias, propensity score matching was used in this paper. The basic idea of the PSM is to find one or several control group students with similar or even the same endowment characteristics for each experimental group student through PSM, and then to compare the mean values of the result variables of the experimental and control groups under by controlling the other variables to obtain the estimation of the impact of boarding on the students’ mental health [ 32 ]. The estimation steps involve: (1) the propensity scores of the students were estimated to establish the experimental group; (2) according to the common support of the propensity scores, the students in the experimental and control groups were matched by the closest matching method; and (3) a balance test was carried out on the two matched groups of samples.

In this paper, the DID method controlled the influence of unobservable variables through the difference in mental health before and after the change of boarding between the experimental group and the control group, especially the influence of factors that do not change over time and synchronously change over time, so as to effectively evaluate the net effect of boarding on students’ mental health. However, since the experimental group and the control group had different student and family characteristics, which does not meet the common trend premise assumption of using DID method, it was necessary to select the samples with similar characteristics in the control group and the experimental group as the counterfactual before performing DID. PSM can weaken selection bias and can obtain comparable samples of treatment and control groups. However, only observable factors can be controlled, and there are still missing variables for unobservable variables [ 33 ]. Therefore, this paper combined the PSM and DID methods to reduce endogeneity, since the DID method can control unobservable individual effects that are invariant over time. DID was used to estimate the average treatment effect of boarding on students’ mental health based on the PSM results, with the differences in the standardized MHT scores as the explained variable. Therefore, this paper used PSM–DID to estimate the average effect of boarding on students’ mental health.

3.2. Models

Based on the above principles, in order to analyze the impact of boarding on students’ mental health, the model in this paper was set as follows:

where i indicates the student; s indicates the school; Δ M H T i s refers to the change in students’ MHT standardized scores in the follow-up period compared to the baseline period; B o a r d i n g i s is the intervention variable. B o a r d i n g i s = 1 indicates that the students are in the experimental group, that is, the students were not boarding in the baseline period but were boarding in the follow-up period; meanwhile, B o a r d i n g i s   = 0 indicates that the students are in control group, that is, the students were not boarding in either the baseline or the follow-up period.

In this study, the important variables that may affect whether students board or not and the students’ mental health level were controlled. X i s in the model indicates the control variables: (1) basic personal information: Gender, age, grade, whether the students are myopic, standardized mathematics scores of the students at baseline, and the distance from the school where the students are located to their home county; (2) family background and socio-economic status: family income, parents’ education level, and whether both parents are migrant workers. M H T i s , b a s e l i n e indicates the standardized mental health level of students at baseline. Through the value of the coefficient β and its significance, we were able to analyze the impact of boarding on students’ mental health.

In order to further analyze the impact of boarding on students’ mental health, we also analyzed the impact of boarding on eight dimensions of mental health (learning anxiety, anxiety about people, loneliness tendency, self-blame tendency, allergy tendency, physical symptoms, horror tendency, and impulse tendency). The specific model settings were as follows:

where Δ y i s represents the scores of the standardized sub-scales for students in the follow-up period, and y i s , b a s e l i n e represents the scores of the standardized sub-scales for students at baseline.

Considering that for students with different characteristics, the impact of boarding on their mental health level may be different, we also analyzed the heterogeneity of the results from the perspective of students’ personal characteristics, family background, and socio-economic status. The specific model was set as follows:

where H i s represents the basic characteristic variables of each individual (i.e., the gender, age, grade of the students, whether the students are myopic, and the distance from the school where the students are located to the county), as well as various family background and socioeconomic status variables (i.e., family income, parents’ education level, and whether both parents are migrant workers).

4. Results and Discussion

4.1. the effect of boarding on students’ mental health.

After matching based on the personal and family characteristics of the students, the changes in the MHT standardized scores of the students in the experimental and control groups were found to differ by a standard deviation of 0.02 ( Table 4 , Row 1, Column 1), but the difference was not significant. This shows that boarding has no significant effect on students’ mental health. According to the various dimensions, there was a significant difference between the students in the experimental group and the students in the control group only in the standardized scores of loneliness tendency, differing by a standard deviation of 0.32 ( Table 4 , Row 4, Column 1). This shows that after boarding, there is no significant change in the other mental health dimensions except that the loneliness tendency increased after boarding.

The effect of boarding on students’ mental health.

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors (SEs) were calibrated by bootstrap (100 times). The table shows the regression results of boarding on mental health level and its different dimensions. Each regression equation controlled the personal characteristics of students (including gender, age, grade, myopia, standardized mathematics scores of students at baseline, and distance from school to county town) and family background (including natural logarithm of family finance, education level of parents, and whether parents migrate for work). a ATT, average treatment effect on treated, indicating the real effect of boarding on students’ mental health. Data source: Authors’ survey.

As for this result, there are three possible explanations.

First, in recent years (before or around 2012), China has attached great importance to the development of the mental health of primary and secondary school students and has increased its investment in boarding schools. For example, a few policies have been implemented to help students and schools to build up mental health education, such as “Guidelines for Mental Health Education in Primary and Secondary Schools” issued by Ministry of Education in 2002 and 2012. This policy helps teachers to help students with mental health disorders recover and adjust, especially for boarding students. Schools often play a substitute role for parents in terms of supervision and psychological counseling of students to some extent. However, due to students’ natural psychological attachment to their parents, schools cannot completely replace the role of parents. Studies such as that of Li et al., (2015) have shown that parent–child attachment has a direct impact on boarding school students’ adaptation to school [ 34 ]. Parent–child attachment refers to an intimate and lasting emotional connection with parents, which can provide support, security, and self-confidence for individuals. Boarding school students are far away from their parents and family for a long time and lack security, which makes it easy for them to feel lonely, which is a kind of explanation why loneliness tendency is significant higher.

Second, left-behind children accounted for a high proportion. In 2012, there were about 165 million migrants and 44.76% of Chinese left-behind children (LBC) aged 12 to17 lived with their grandparents [ 35 ]. If we analyze the impact of boarding and left-behind children, as shown in Table 5 last row, the mental health of boarding LBC is worse than that of non-borders, which has an impact of 0.29 standard deviation and is significant at the confidence interval of 5%. The result implies if one parent goes out for working and students are boarding, their mental health is worse than who are not boarding, which also shows that if parents are not at home, students are prone to psychological problems such as loneliness. The above result is consistent with a latest study published in 2020. Chen et al. (2020), using data of 7606 rural students, proved that the effect of boarding is significantly higher among disadvantageous students. This study uses Instrumental Variable to examine the casual effect.

The effect of boarding on the mental health of students from different groups.

Note: *** p < 0.01, ** p < 0.05, * p < 0.1 Standard errors were calibrated by bootstrap (100 times). The table shows the regression results of the heterogeneity analysis for different groups. Each regression equation controlled the personal characteristics of the students (including gender, age, grade, myopia, standardized mathematics scores of the students at baseline, and the distance from the school to the county) and family background (including the natural logarithm of family finance, the educational level of parents, and whether parents migrate for work). a ATT, average treatment effect for the treatment, indicating the real effect of boarding on students’ mental health. Data source: Authors’ survey.

Third, in order to accurately identify the causal effect of boarding, this paper regards boarding as an intervention, which merely lasts one year. Short boarding period (one-year boarding) may not be enough for students’ mental health change.

Though the above conclusions are contrary to the research results of the existing literature, there might be some deficiency of the previous research. Specially, most research on the relationship between boarding and students’ mental health states that boarding has a negative effect on students’ mental health, but such research has some problems, such as small sample sizes, unrepresentative samples, and unresolved endogeneity. For example, Zhang et al. (2009) used the mental health scale for middle school students as a test tool to compare the mental health status of 274 junior high school students in boarding schools and 300 junior high school students in non-boarding schools in Ningxia, China [ 36 ]. The study found that the overall psychological problems of junior high school students in boarding schools were significantly higher than those of non-boarding schools, and the detection rate of various psychological problems in the former was also higher than that of the latter. However, the sample size of this study was too small, and the method adopted was a simple t -test, which failed to solve the problems of endogeneity. Moreover, the research object of this paper was junior high school students, and there may be some differences between their mental health and that of primary school students. Another study from Wang and Mao (2015) analyzed the survey data of 8047 primary school students in grades 4, 5, 7, and 8 in 11 western regions [ 37 ]. They believed that boarding did not play a substitute role for family supervision for left-behind children, but rather became a negative factor affecting the development of left-behind children’s social–emotional ability. Although the sample size was large and representative, the method used in this study was simple multiple regression analysis, and the mental health level of the students was tested through the self-compiled “Primary and Secondary School Students’ Social Emotional Ability Questionnaire”. Although the scale passed reliability and validity tests, the reliability of the scale needs further examination. This may be the deviation caused by different measuring tools. However, this paper has adopted a more convincing method PSM–DID with large and representative data, which can clearly estimate the real impact of boarding. We assume the results have been similar with most recent updated studies, such as Shi et al. (2016), Chen et al. (2020), and Liu and Villa (2020) [ 11 , 26 , 35 ].

4.2. Heterogeneity Analysis of the Effect of Boarding on Students’ Mental Health

There was no significant change in the mental health level of the students after boarding, but considering the different characteristics of the students, the changes in the mental health level after boarding may be different. This paper analyzed the heterogeneity of the impact of boarding on the mental health of the students, and the results are shown in Table 6 . The results show that there is heterogeneity in the impact of boarding on students’ mental health in terms of grade and whether their parents are migrant workers. However, for the other individual and family characteristics, there is no heterogeneity in the impact of boarding. In order to further analyze the different effects of boarding on the mental health of different groups of students, this paper used model (1) to study the changes in the MHT standardized scores of the students after boarding. The results are shown in Table 7 .

Heterogeneity analysis of the effect of boarding on students’ mental health.

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. Standard errors were calibrated by using the bootstrap method (100 times). The table shows the regression results of the heterogeneity analysis for different groups. Each regression equation controlled the personal characteristics of the students (including gender, age, grade, myopia, standardized mathematics scores of the students at baseline, and the distance from the school to the county) and family background (including the natural logarithm of family finance, the educational level of parents, and whether parents migrate for work). a ATT, average treatment effect for the treatment, indicating the real effect of boarding on students’ mental health. b Fifty percent of the students in the sample had a distance of more than 32 km from their school to the county, while the other 50% had a distance of less than 32 km from their school to the county. Data source: Authors’ survey.

The effect of boarding on students’ mental health from heterogeneity of grade 4 and grade 5.

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. The table shows the regression results of boarding on mental health level and its different dimensions with difference in grade 4 and grade 5. Each regression equation controlled the personal characteristics of students (including gender, age, grade, myopia, standardized mathematics scores of students at baseline, and distance from school to county town) and family background (including natural logarithm of family finance, education level of parents, and whether parents migrate for work). a ATT, average treatment effect on treated, indicating the real effect of boarding on students’ mental health. Data source: Authors’ survey.

First of all, according to the results in rows 1 and 2 of Table 5 , the students in the experimental group of grade 4 had a better mental health level than those in the control group, while the students in the experimental group of grade 5 had a significantly poorer mental health level than those in the control group. That is, the mental health level of fourth-grade students improves after boarding, while the mental health level of fifth-grade students deteriorates. To explore the possible explanation, we made regression of the boarding effect on sub-dimensions of mental health between grade 4 and grade 5 as shown in Table 7 . It can be seen that among students who have been boarding, in the dimension of learning anxiety and anxiety about people, the anxiety level of the fifth grade students is obviously higher than that of the fourth grade students, amounting to 0.39 SD and 0.27 SD. This may be because as they reach higher classes, the students find it more difficult to learn the course; and the fifth-grade students are facing greater pressure than the fourth-grade students. Studies have proven that in rural schools students do not have a solid learning foundation. With the increase in grade, the learning content becomes more and more complex, the knowledge that students are not able to fully master increases, and their learning anxiety worsens [ 38 ]. Additionally, fifth-grade students are facing great pressure to enter junior high school, and their mental health level is poorer [ 35 ]. Therefore, the mental health level of fifth-grade students deteriorates significantly after boarding.

Second, according to the results in rows 5 and 4 of Table 5 , for families where at least one parent has not left to work in an urban area, there was no significant change in the mental health level of the students after boarding. However, for families whose parents are both migrant workers, the mental health level of students after boarding was significantly worse (SD = 0.30) at a significance level of 5%. If one parent is at home, boarders can avoid psychological problems by communicating with their parents when they return home on weekends. However, for families where both parents migrate to urban areas for work, students are not well cared for by their parents. After boarding, students are more likely to have psychological problems due to their reduced contact with their families, which is in line with literature in left-behind children, such as, Ge et al. (2015), Fellmeth et al. (2018), Jia and Tian (2010), and Zhao and Yu (2016) [ 39 , 40 , 41 , 42 ]. There is also possibility that families where both parents need to go out have unobserved differences to other families (i.e., financial issues, opportunity structure). Overall, it might be that boarding can make up for the lack of family supervision of left-behind children to some extent, but it cannot completely replace the emotional communication of families.

5. Conclusions

Based on the panel data from 2012 to 2013, this paper used the DID and PSM methods to estimate the impact of boarding on the mental health of rural primary school students. It was found that boarding had no significant effect on the mental health of rural primary school students, but the tendency of loneliness after boarding increased. There were significant differences in the mental health of students between grades. There was also a significant difference in the mental health of students depending on whether or not their parents migrated for work.

Unlike most existing studies, this study concluded that there was no significant difference in the mental health of boarding school students. There are three possible reasons. One possible reason is that in recent years, China has attached great importance to the development of adolescents’ mental health and has increased its investment in the country’s boarding schools. An example of the importance attached to boarding schools can be found in the “Guidelines of the State Council on Strengthening the Care and Protection of Left-behind Children in Rural Areas” released in 2016, which proposed that the administrative department of education should support and guide small- and medium-sized schools to strengthen mental health education, promote the positive and healthy development of adolescents’ mental health and personalities, and both detect and correct any psychological problems and bad behaviors early. The document also proposed to “strengthen the construction of rural boarding schools, promote the rational distribution of boarding schools, and meet the schooling needs of left-behind children in rural areas.” This work was listed as the main point of work for the Ministry of Education for 2016 and 2017 [ 43 ]. Therefore, schools could be developed to replace the supervisory role of the family to a certain extent and can provide a certain degree of psychological diversion. However, due to students’ natural attachment to their parents and the lack of family communication for boarding school students, school cannot completely replace the role of parents. Therefore, in addition to focusing on cultivating the independent living ability of boarding school students, schools should also pay attention to mental health issues such as loneliness as a result of boarding. The second reason that cannot be ignored is that left-behind children still account for a high proportion, so it is reasonable that loneliness tendency has been significantly higher among eight dimensions of MHT. In addition, it is also possible that the impact of one year’s boarding on students has not yet appeared. It is recommended to shorten the holiday period as appropriate to meet the students’ desire to get warmth and emotional connection from family ties. At the same time, schools should pay attention to the development of students’ mental health, implement relevant courses, and hire professional mental health teachers to provide psychological counseling to students.

Compared with the existing research, the advantage of this paper is that it received follow-up survey data from large and representative samples of rural students. Furthermore, the panel data were used to solve the endogeneity problem caused by the self-selection bias and missing variables of the sample through the methods of PSM–DID. However, there are still drawbacks in this paper: (1) the problem of self-selection bias caused by unobservable factors could not be completely solved by PSM; (2) it was difficult to solve the mutual causation problem between boarding decisions and students’ mental health levels. However, this paper still supplements the existing literature and provides a new dimension of evidence for understanding the impact of boarding on students.

Acknowledgments

The authors gratefully thank all participants in collecting the data and show appreciation to reviewers who have given many useful feedbacks on revision.

Author Contributions

Data curation, Y.W. and S.W.; formal analysis, Y.G.; project administration, Y.S.; resources, H.L. and Y.C.; writing—review and editing, B.T. All authors have read and agreed to the published version of the manuscript.

The authors would like to acknowledge the funding support from the Fundamental Research Funds for the Central Universities (grant No. 2017CBY017), the National Natural Science Foundation of China (grant No. 71933003), the 111 Project (grant No. {"type":"entrez-nucleotide","attrs":{"text":"B16031","term_id":"2123780","term_text":"B16031"}} B16031 ), and the China Scholarship Council.

Conflicts of Interest

The authors declare no conflict of interest. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Academia.edu no longer supports Internet Explorer.

To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to  upgrade your browser .

•  We're Hiring!
•  Help Center

Boarding Schools

• Most Cited Papers
• Most Downloaded Papers
• Newest Papers
• Last »
• Food habits Follow Following
• Nutritional Status Follow Following
• Community Capacity Building Follow Following
• Rural Schools Follow Following
• Malnutrition Follow Following
• Evidence-based policy Follow Following
• Pondok Pesantren Follow Following
• Crafts and Technology Follow Following
• Aboriginal education Follow Following
• School Uniforms Follow Following

Enter the email address you signed up with and we'll email you a reset link.

• Academia.edu Journals
•   We're Hiring!
•   Help Center
• Find new research papers in:
• Health Sciences
• Earth Sciences
• Cognitive Science
• Mathematics
• Computer Science
• Academia ©2024

The Kolmogorov Reform of Mathematics Education in the USSR

• First Online: 09 March 2023

Cite this chapter

• Alexandre Borovik 4  

Part of the book series: History of Mathematics Education ((HME))

814 Accesses

1 Citations

In the Soviet Union a reform movement in mathematics education was triggered by Andrey Kolmogorov in the 1970s, and it was followed by a counter-reform. This movement was rooted in the very different socioeconomic conditions of that time and place and followed a strategy with significant contrasts to similar programs in the United States, England, and France. This provides an interesting case study that may illuminate the way such movements arise and succeed or fail, and, at the social level, certain fundamental commonalities of constraints as well as significant differences according to local conditions. We shall show that the principal reasons for the failure of the Kolmogorov reform were political: (a) The reform ignored the reality of the socioeconomic conditions of the country; (b) The human factor was ignored, and very little attention was given to professional development and retraining of, and methodological help to, the whole army of teachers; and (c) An attempt to transfer mathematical content and methods from the highly successful advanced extension stream for mathematically strong and highly engaged children to mainstream education was an especially grievous error.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or Ebook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime
• Available as PDF
• Read on any device
• Instant download
• Own it forever
• Available as EPUB and PDF
• Compact, lightweight edition
• Dispatched in 3 to 5 business days
• Free shipping worldwide - see info
• Durable hardcover edition

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Similar content being viewed by others

Acknowledgments

The author thanks Glen Aikehead, Szabó Csaba, Gregory Cherlin, Michael Grinfeld, Roman Kossak, Dmitrii Pasechnik, Nikolai Vavilov, Alexander Veselov, and Theodore Voronov for their feedback and advice. The author also thanks Dirk De Bock for his patient and detailed editing of this chapter.

Author information

Authors and affiliations.

University of Manchester, Manchester, UK

Alexandre Borovik

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Alexandre Borovik .

Editor information

Editors and affiliations.

KU Leuven, Leuven, Belgium

Dirk De Bock

Rights and permissions

Reprints and permissions

Copyright information

© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG

About this chapter

Borovik, A. (2023). The Kolmogorov Reform of Mathematics Education in the USSR. In: De Bock, D. (eds) Modern Mathematics. History of Mathematics Education. Springer, Cham. https://doi.org/10.1007/978-3-031-11166-2_16

Download citation

DOI : https://doi.org/10.1007/978-3-031-11166-2_16

Published : 09 March 2023

Publisher Name : Springer, Cham

Print ISBN : 978-3-031-11165-5

Online ISBN : 978-3-031-11166-2

eBook Packages : Mathematics and Statistics Mathematics and Statistics (R0)

Share this chapter

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Publish with us

Policies and ethics

• Find a journal
• Track your research