IntroductionAs students transition into college, the influence of family diminishes, while peer relationships increasingly become a pivotal factor widely believed to shape academic performance1. Traditional research on peer relationships has predominantly focused on the impact of physical networks—such as those formed within dormitories, classes, and academic majors—on academic outcomes2,3,4. Yet, these studies fail to grasp the more substantive aspects of peer integration among college students. Notably, in most Chinese universities, for example, dormitory assignments are determined through a random sampling process, which may not necessarily foster genuine peer relationships5. The degree of genuine peer acceptance within these relationships, particularly among roommates, is a crucial yet underexplored factor related to academic performance.Research conducted on primary and secondary school students has consistently shown that peer acceptance contributes to academic progress6,7. Yet, few empirical studies have extended this inquiry to college students, primarily because traditional methods of measuring peer acceptance—such as the “nomination” method8—are more suited to younger students and present challenges when applied to a university context.The concept of “commensality”, the practice of eating together, while often considered cultural or even religious, provides a novel perspective of empirically measuring peer relationships. This study posits that commensality serves as an innovative and effective technique for studying peer acceptance among college students. By integrating commensality data with rich demographic and academic information, this research aims to open new avenues for understanding the intricate relationship between peer relationships and academic performance in a university context, utilizing big data methodologies.When students enter university, they become separated from their families, and peers emerge as their primary social contacts. This is a period where the concept of “peer effects,” which encompasses the externalities of human capital, psychological state, and social behavior that arise from interactions within peer groups—is considered to reach its peak9. This concept posits that adolescents, in particular, are more susceptible to peer influences than adults, impacting their emotional10, behavioral11, and academic adaptation12. This susceptibility is especially pronounced among Chinese college students, where the reduction of CEE (College Entrance Examinations) pressures and increased peer interaction amplify peer influences. The “college friend circle,” comprising dormitories, classes, colleges, and clubs, plays a significant role in the socialization process, shaping students’ psychological characteristics and swaying their major choices13, career planning14, academic performance15, and behaviors such as smoking16 and alcohol use17.The psychological community has long investigated the role of peer effects on academic performance. As early as 1966, the Coleman Report indicated that students’ grades are closely linked to the educational backgrounds and aspirations of their peers. The report suggested that youth subcultures, distinct from the adult world, exert a stronger influence among peer groups than family or school18. Subsequent studies have confirmed the presence of peer effects in primary and secondary schools, highlighting their impact on academic performance19. Yet, research findings on peer effects in higher education present a divided picture.Some early studies suggest that peer effects positively influence academic performance in college. For instance, Hoel’s study at Reed College found a significant peer effect on academic outcomes20. Similarly, Hasan and Bagde’s research in an Indian engineering college further specified this causal mechanism. They found that the pre-college academic skills of randomly assigned roommates significantly influenced a student’s performance. Crucially, this effect was most potent when there was a match between the roommate’s skills and the student’s courses, e.g., a roommate skilled in math improved a student’s math grades, but not English21. Sacerdote’s study at Dartmouth College utilized the random assignment of roommates to isolate peer effects. His findings revealed that it was not the mere presence of a roommate, but rather the specific characteristics of that roommate, that mattered. Specifically, being randomly assigned a roommate with a higher GPA causally led to a significant increase in a student’s own GPA2. In China, studies on top Chinese universities, such as Tsinghua, also support these findings, showing a positive impact of dormitory peers on academic performance, with some evidence suggesting this impact grows over time22,23.Yet, there are dissenting voices in academia that argue against the positive impact of peer effects on college students’ academic performance. For example, Zimmerman found that college students’ academic outcomes were not substantially influenced by their peers in a study of roommates’ grades and SAT scores at Williams College24. In a panel study at the University of Maryland, Foster similarly found that peer effects on academic performance do not exist. Furthermore, Arum and Roksa argued that peer relationships might negatively affect undergraduates’ achievements in critical thinking, complex reasoning, and writing skills25,26.These conflicting findings have compounded the uncertainty in college policy and decision-making. Take China as an example, only a few universities have attempted to assign dormitories based on student characteristics, such as smoking habits, while most continue to use random assignments. The inconsistency in research outcomes stems from two primary issues: the lack of scientific rigor in defining and measuring core peer relationships and the failure to account for the heterogeneity of these dynamics. Studies that focus solely on physical connections, such as shared dormitories or majors, overlook core aspects of peer relationships, including genuine peer acceptance, potentially leading to distorted conclusions about cause and effect. Additionally, the dynamics of peer relationships likely vary across different disciplines, majors, genders, ethnicities, and CEE scores. Although studies have recently begun to recognize the importance of these heterogeneous factors27, there remains a need for comprehensive research that accounts for these variables using reliable, substantial empirical data.Introducing peer acceptance as a dimension of analysis is pivotal for understanding the link between real peer relationships and academic performance among college students. Previous research on primary and secondary school students has demonstrated that peer acceptance is a positive and protective factor in peer relationships, which can mitigate internalizing and externalizing problems and reduce students’ propensity for maladaptive outcomes28. Consequently, peer acceptance generally exerts a positive influence on academic performance.Yet, despite the established importance of peer acceptance in earlier educational stages, this concept has been underexplored in the context of college students. A significant challenge lies in the applicability of the traditional “nomination” method, commonly used to measure peer acceptance. This approach typically requires younger students to list their most and least favored classmates, which allows researchers to categorize peer relationships into distinct types, such as popular, rejected, or neglected8,29. Yet, this method is less suitable for college students, who are generally more mature and for whom such questionnaires cannot reliably capture the nuances of peer acceptance.To address this gap, we introduce the concept of “commensality” as a novel measure of peer acceptance among college students. Commensality is an interdisciplinary research area that explores the social, psychological, and cultural aspects of shared meals. From a psychological and behavioral perspective, commensality is believed to foster social connections and enhance a sense of belonging, which in turn can positively influence academic performance. Research by Gregory suggests that sharing meals promotes social interaction, enabling individuals to express emotions, share experiences, and build close relationships, thereby enhancing overall well-being30. Additionally, commensality has been shown to alleviate stress and anxiety levels, with shared meals among friends or family reducing psychological stress and facilitating emotional regulation31. Furthermore, commensality plays a key role in shaping individual self-identity32.Regarding the second issue, the conflicting findings on peer relationships and academic performance are not only a matter of heterogeneity; they are symptomatic of a fragmented and unsystematic body of research. The limitations of research samples or the absence of important variables often prevent a comprehensive understanding and lead to contradictory results. To improve the scientific rigor, it is essential to conduct a systematic examination of these factors. Based on the existing literature, this study identifies several key aspects of heterogeneity—first, physical space. Research has shown that different dormitory allocation methods can result in varying peer effects. Globally, dormitory assignments generally fall into three categories: no on-campus dormitory provision, non-random dormitory assignments, and random dormitory assignments. In China, most universities assign dormitories randomly based on students’ majors, creating a natural random sampling of dormitory, floor, and building assignments. In contrast, universities in the United States and other countries often use multi-criteria evaluations for dormitory assignments, resulting in greater variability in standardized test scores among roommates. There is also a higher prevalence of single rooms. The nature of peer relationships is quite different in multi-person dormitories compared to single rooms. Furthermore, Chinese students often bring a competitive mindset into dormitory life, where roommate interactions can be influenced by competition for awards, party membership quotas, and employment opportunities, leading to “utilitarian tendencies and small group phenomena”. In addition, compared to primary and secondary students, college students experience greater physical and social distance, making it necessary to consider dormitory buildings and other spatial elements in peer relationship research33.Second, gender. Existing studies indicate that the link between peer relationships and academic performance varies by gender. Some social psychology research suggests that women are more susceptible to peer influence in competitive environments34, a view supported by empirical studies35. Yet, in the context of college students, most research finds that male students are more influenced by their peers36.Third, family backgrounds. According to Harris, family background provides children with opportunities to build different types of peer networks. Although college students are physically distant from their families, family factors continue to influence peer effects to a great extent. For instance, in China, urban-rural and household registration factors can affect peer acceptance; children of migrant workers are often marginalized in public schools and less likely to be accepted by peers28. Family economic conditions and consumption levels also play a role. Students whose parents share similar educational backgrounds, occupations, and socioeconomic status are more likely to form friendships. For example, Mayer and Puller found that students of the same race, income level, and with at least one parent having higher education are more likely to become friends37. Conversely, students from disadvantaged socioeconomic backgrounds likely face reduced peer acceptance or occupy a marginalized position within peer groups.Fourth, individual social identity and status. Students’ specific social identities and roles within their communities are also powerfully correlated to peer relationships. Political orientation, for example, influences peer relationships, as students with similar political views are more likely to connect37. Additionally, student leadership experience can affect peer dynamics; those who have held leadership positions tend to have broader peer networks and receive more positive peer evaluations28. Ethnicity is another fundamental aspect of social identity, with studies indicating that it can also shape peer effects38,39.Therefore, in this study, we first investigate the relationship between peer interactions among college roommates and academic performance, focusing on how commensality variables that we defined to innovatively measure peer acceptance are linked to academic performance. We then extend the analysis by broadening the categories of peers. In addition to roommates (Category 1), we examine Category 2 peers—classmates living on the same dormitory floor but not sharing a room—and Category 3 peers, who are classmates but live on different floors or in different buildings. This expansion allows us to explore how peer relationships at varying physical distances are associated with academic performance. Finally, we investigate the heterogeneity of these associations across various subgroups, considering how factors including gender, academic background, and family influence moderate the relationship between peer acceptance and academic performance. While this study refrains from making premature causal claims, by robustly mapping the abovementioned relationships, it provides a solid empirical groundwork for future inquiries.ResultsThe short-term and sustained links between roommate-peer relationships and academic performanceTo begin, we analyzed the short-term association between peer acceptance, as quantified by “commensality” among roommates, and their semester average scores. “Commensality” is operationalized through this: If two students make their first card swipe in the same canteen, on the same day, and during the same meal period, and the time difference between these first swipes is within a specified threshold, this is considered a single instance of them dining together. As a result, each student’s “commensality value” is incremented by 1. As shown in Table 1, there is a significant positive correlation between the value of roommate commensality in the first semester and the students’ average scores at the end of that semester. Specifically, the standardized coefficient was \(\beta\)= 0.092, with a p value of less than 0.001 and a 95% confidence interval of [0.004, 0.008]. This finding indicates a positive and significant association, where higher peer acceptance among roommates is linked to higher semester average scores in the short term.Table 1 Regression results of standardized coefficients for roommates’ commensality and average scoresFull size tableTo validate these findings, we replaced the first-semester data with second-semester data and conducted another round of standardized coefficient regression analysis. The results remained consistent, showing that the short-term association remained significant. The commensality value for the second semester was significantly and positively linked to the course average scores at the semester’s end (\(\beta\)= 0.142, p