Does Peer Socialization Within Cohorts Foster Political Attitudes? A Longitudinal Study of Elite Business Students

Getting a university education is a defining experience in a person’s life and scholars suggest that it can foster distinct political attitudes and electoral behaviors (Stubager, 2008). Yet, despite the considerable interest in the impact of education on political attitudes, surprisingly little is known about the underlying mechanism that drives the association. It is still an open question whether the correlation is causal, and if so, what the underlying mechanism is (see e.g., d'Hombres & Nunziata, 2016; Campbell & Horowitz, 2016). Previous research into the association between education and political attitudes suggests that informal peer socialization in comparatively elite milieus is one of the most probable mechanisms in the case of a 'higher education effect’ (Apfeld et al., 2022; Stubager, 2008; Surridge, 2016). Students may uphold and transmit social norms about how to think and act politically (Mendelberg et al., 2017). The characteristics of the peers would therefore shape both the intensity of the socialization and the salient issues within the cohort. Accordingly, variations in student compositions might account for why certain studies identify a causal effect of attending university on political attitudes (e.g., Scott, 2022), whereas others do not (e.g., Simon, 2022).

The idea of an inter-group transmission of attitudes within university cohorts has, however, rarely been systematically investigated due to limitations in data availability on students’ social interactions. Instead, the assumption of a peer socialization effect is the outcome of an empirical exclusion of other potential explanations (e.g., Scott, 2022; Stubager, 2008; Surridge, 2016) or is based in data from over 30 years ago (Dey, 1996; Pascarella and Terenzini, 1991). Even if some studies investigate political socialization within smaller friendship groups (Algan et al., 2023; Hjerm et al., 2018) or between college roommates (Strother et al., 2021), there is a need for research on broader political socialization processes in higher education. Indeed, if peer socialization is a partial driver of the education divide, where highly educated tend to hold more liberal preferences, it is probable that socialization at this phase of an individual’s education and at this group level would answer to such attitudinal differences. Only Mendelberg et al. (2017) succeed in identifying a causal effect of social norms in university classes on students’ political attitudes in their novel U.S. study while focusing on students’ economic preconditions. This paper studies the peer socialization mechanism with a different research design and context, utilizing more detailed measurements of social behavior as well as an objective reduction of socialization due to the Covid-19 pandemic’s social restrictions. It further theorizes and investigates attitudinal distributions within university cohorts rather than solely looking at a directional attitudinal change to more closely align with research on socialization effects (see Dey, 1996). This paper examines the question: Does peer socialization in higher education influence political attitudes?

We employ a longitudinal network study of four top-ranked business schools in Sweden and Finland. These prestigious educational institutions encourage and facilitate social interactions among the students (Tyllström et al., 2022), increasing the possibility of detecting a socialization effect in this environment. The empirical analysis of the data consists of three parts. First, the study descriptively examines the distribution of attitudes within cohorts over time cross-sectionally, where convergence indeed is an expected outcome given the hypothesis of socialization effects. Second, an ordinary least square (OLS) regression examines the association between students’ social life and their deviation from the average attitude in their cohort. Through this analysis, it is possible to directly examine whether students with various extents of social interactions are closer to the majority norms in class. Third, the study adopts the Covid-19 pandemic as an exogenous shock to political socialization among peer networks in a two-wave panel, allowing for a difference-in-difference design to test the influence of peers on attitude formation causally. The rationale behind this ‘treatment’ is the fact that some cohorts randomly experienced their education online during the pandemic and were recommended social distancing. In this study, we thus extend previous research in several ways. We use a unique longitudinal network dataset of undergraduate business students studying at prestigious universities, covering both their extent of social interactions and their experience of social life within the cohort. We also expand the methodological approach to study socialization effects by examining the impact of social interactions through two avenues: initially, by analyzing students’ self-reported social behaviors, and subsequently, by leveraging an external shock affecting such socialization.

Contrary to expectations of convergence, this study does not yield support for the presence of peer socialization effects on attitude formation. There are no signs of attitude convergence over time and little evidence to support that different social behaviors are associated with more conforming political attitudes. Even worse for the possibility to support a peer socialization mechanism, the results from the difference-in-difference design show that there are no significant differences between cohorts studying on campus or online. While the paper’s design only enables us to make causal claims about the socialization mechanism, the descriptive findings are consistent with the notion of an overall attitudinal change resulting from university attendance. Consequently, we conclude that if participation in higher education indeed has an overall causal effect on political attitudes, peer socialization at the cohort level is unlikely the driving force behind it. The study thereby nuances the notion of the underlying mechanism of the 'higher education effect’, which is of critical interest given the increasing role of education as a predictor for political preferences in advanced democracies (Bornschier et al., 2021). In addition, these insights contribute to an even broader question of how, when and for whom political socialization processes operate. We close the paper by discussing the general implications of the results and suggest future studies to move the field forward.

Despite the robust and consistent association between higher education and political attitudes (Marshall, 2016; Stubager, 2008), a large body of research questions the notion of a causal effect of university attendance. The relative importance of self-selection into higher education is the subject of considerable discussion. Pre-adult factors such as parental socialization, family socioeconomic status, genetics, and intelligence may cause individuals to develop certain attitudes as well as attend higher education (Campbell & Horowitz, 2016; Persson, 2015; Weinschenk et al., 2021). However, while recent research backs up the ‘null hypothesis’ of the education-attitudes link (e.g., Finseraas et al., 2018; Kunst et al., 2020; Lancee & Sarrasin, 2015; Simon, 2022; Miragaya et al., 2023), several studies with equally strong research designs for identifying a causal effect of higher education do support the idea of attitude formation through university attendance (e.g., Scott, 2022; Marshall, 2016; Cavaille & Marshall, 2019; d'Hombres & Nunziata, 2016; Mendelberg et al., 2017).

If the first question occupying the literature on education effects concerns the causality of the effect, the second question concerns the underlying mechanism that may change individuals’ political attitudes while in education. Although such attitudinal change could result from an increased cognitive sophistication that occurs in parallel with an individual’s studies (Nie et al., 1996), the model of explanation with the most considerable support in the literature relates to universities as a site for interpersonal socialization (see e.g., Stubager, 2008). This socialization mechanism is based on one central idea: through education, individuals internalize the values they encounter in the university milieu through interpersonal interactions (Dey, 1996; Mendelberg et al., 2017; Nickerson, 2009). The social influence on attitude formation has proven central in other realms of an individual’s life. Discussion partners may influence political attitudes (Huckfeldt & Sprague, 1991; however, see Farrar et al., 2009), and family socialization during childhood has long-lived effects on individuals’ political views (Rekker et al., 2017). If university attendance indeed changes political attitudes, it is thus not unlikely that socialization is the main driver behind it.

Empirically, the works of Dey (1996), Paterson (2009), Stubager (2008), Scott (2022) and Surridge (2016) all suggest (informal) processes of socialization on educational campuses as the main driver of the linkage between education and political attitudes. While suggestive evidence indicates that professors may also serve as socializing agents (Apfeld et al., 2022; Gross & Fosse, 2012), peers can be the most important attitudinal influences. Students transmit norms on campuses about what are socially accepted political preferences.¹ As such, Simon (2022) posits that the divergent conclusions from her and Scott’s (2022) studies, both studying if there is a causal effect of education on British students, could be attributed to the varying student group compositions across the time periods they examined. In the realm of peer socialization, both the direction of attitudinal change and the convergence of attitudes are thus of interest. While much of the extensive literature in political science primarily focuses on the liberalizing effect of education and posits peer socialization as a possible driver of such change (Stubager, 2008; Surridge, 2016), socialization literature suggests that peer transmission of attitudes rather tends to make attitudes more similar (Dey, 1996).

The socialization mechanism is mainly confirmed by the exclusion of other potential explanations and does not directly consider students’ social interactions and behaviors (Scott, 2022; Stubager, 2008; Surridge, 2016). Due to the inability to control for all possibly important confounders, and the risk of measurement errors in the data collected through surveys (as noted by Simon, 2022, and Campbell & Horowitz, 2016), these studies could overestimate the influence of peer socialization. Only a few studies have attempted to identify the impact of socialization processes in the university milieu using sophisticated quantitative techniques. As such, research indicates that students randomly assigned to live together as roommates, an intense socializing experience, tend to become more politically alike rather than necessarily more liberal throughout their studies (Strother et al., 2021). Interestingly, Strother et al. (2021) did not identify a causal effect of university attendance when examining a general shift in attitudes among the students. In addition, Mendelberg et al. (2017) show that a student’s values may be influenced by their peer group’s economic views and consumption behavior (see also Newcomb, 1943; Pascarella and Terenzini, 1991), and may even move students in a conservative direction.

While the influence of roommates and smaller friendship groups highlights the impact of close relationships on attitude formation, the effects of class norms on attitudes do not need to be limited to close ties (Newcomb, 1943). In fact, if peer socialization in universities would explain parts of the ‘higher education effect’, cohorts should be the units of interest. Only Mendelberg et al. (2017) have collected data on ideological class composition and political attitudes providing unbiased estimates on the impact of class norms. Their data structure allows for examinations of school class characteristics and attitude formation, but detailed information about the social experiences within school classes is lacking. It further remains uncertain whether the conclusions drawn from their research apply to students from other countries with less politically polarized climates (Boxell et al., 2022). Colleges in the United States are further known for their high admission fees (Garritzman, 2016), which are likely to impact the characteristics of the higher education entrants and the overall campus experience. We thus aim to investigate a distinct case of student sample in the form of business students in Sweden and Finland. Furthermore, we aim to bridge these two bodies of literature by placing a primary emphasis on the convergence of attitudes while also acknowledging the significance of the direction of possible attitudinal change.

According to the assumptions of a peer socialization effect, attitudinal change during an educational trajectory occur as students internalize the political attitudes they encounter in order to fit in with the peer group (Stubager, 2008). However, even though the correlation between education and liberal attitudes is widespread and the socialization thus assumed to be directional (Scott, 2022; Simon, 2022; Stubager, 2008), there are no theoretical reasons based on the socialization literature for the attitudinal change to necessarily be liberalizing (Mendelberg et al., 2017; Newcomb, 1943). As discussed, the underlying assumption of socialization effects is rather that attitudes are more likely to align with the specific cohort’s social norms (cf. Strother et al., 2021). The composition of students will thus influence the values that prevail in the classroom (Mendelberg et al., 2017; Ricart-Huguet & Paluck, 2023). It is therefore reasonable to expect a decrease in attitudinal differences in university cohorts over time, as students who hold non-conforming attitudes are more likely to shift their beliefs to conform with the majority, resulting in less variability between students. Students who already lie close to the majority views may not experience attitudinal change, while students who hold more deviating attitudes could be expected to experience attitudinal change and converge toward the majority norms.

Furthermore, central to the transmittance of attitudes is social embeddedness since it directly or indirectly exposes students to the social norms in class (Mendelberg et al., 2017; Newcomb, 1943). Social groups can require specific attitudes for social acceptance and sanction ‘illegitimate’ behaviors and attitudes (Newcomb, 1943). Additionally, indirect social comparison processes can also influence attitudes. For instance, students may view others at the same level and within the same field of education as their reference group and adjust their opinions on salient issues to conform to those of the prototypical group member (Westholm, 1999). This process does not necessarily involve direct social sanctions, nor does it assume close ties to individuals in that particular group. The effect of education may thereby solely appear for students who are frequently exposed to the norms prevailing within the cohort. Students who are not engaging with their peers would accordingly be less likely to experience an attitudinal change, while students who socialize with their peers would be more likely to conform to the majority views. This paper will therefore operate under the following hypotheses:

Identifying the influence of peer socialization on opinion formation processes is exceedingly difficult as several omitted variables could drive potential outcomes. Therefore, we propose a research design to investigate the transmission mechanism that is based on three studies, each of which contributes independently to examining socialization effects. The studies have a complementary nature due to their distinct strengths and limitations as explained below.

First, we examine differences in attitude variance between cohorts who have studied for varying durations, ranging from 1 to 3 years. Test 1 is conducted both in a cross-sectional form (as shown in the main manuscript) and in a panel (as shown in the appendix). This analysis allows us to descriptively observe attitudinal differences between the cohorts and preserves a full sample size. This stage of the analysis thereby offers an initial overview of whether attitudes tend to be more alike among students in more advanced stages of their educational trajectory. However, it does not examine the impact of their diverse socializing behaviors directly. Therefore, we leverage survey data with information about the students’ political attitudes and social life in an OLS regression analysis in the next step. Through this analysis, it is possible to observe the relationship between students’ self-reported social behaviors and their attitudes. Still, causal conclusions cannot be inferred from ‘conventional’ regression-based studies, which only identify associations (Simon, 2022). In addition, since the way one maintains and creates social contacts could be considered a sensitive topic, it is not unlikely that social desirability bias will interfere with the results. The study, therefore, takes advantage of the restrictions on social interactions amongst peer networks during the Covid-19 pandemic. Test 3 compares changes in political attitudes between the cohort that started studying during the pandemic and whose social networks were seriously affected to the cohort that started studying before the pandemic. Test 3 thereby utilizes the panel structure of the data and an externally imposed reduction of peer socialization, although with a smaller sample. Taken together, the first study tests H (1) concerning convergence in attitudes within cohorts over time, whilst the second and third studies test H(2) whether the social behaviors of students influence their attitudinal similarity.

One of the core research areas in political science is understanding how citizens develop their political attitudes. This study contributes to the field by being among the first to directly test whether peer socialization fosters political attitudes within university cohorts. We leverage a unique dataset that tracks students from Swedish and Finnish top-ranked business schools and examine the link between their social behaviors and attitudes. The study further considers the political norms in their respective cohorts and thereby bridges the large literature in political science focused on directional attitudinal change from university attendance with a closer socialization perspective. Additionally, we leverage the Covid-19 pandemic as an exogenous shock that reduced the degree of contact between peers in certain cohorts, and thereby rely on both self-reported and objective measures of the students’ social interactions.

Contrary to the results of the study conducted in the U.S. by Mendelberg et al. (2017), but in line with the results by Strother et al. (2021), the findings did not support the idea of a general socialization mechanism within the university cohorts. There were no signs of convergence in attitudes over time, and we therefore do not have support for the first hypothesis, The second hypothesis was also rejected, as there were no significant differences between students who varied in their social behaviors. In fact, the models only explained a very limited part of the variability in attitudes, which suggests that background factors and earlier socialization are more important determinants for attitudes. While the design of the study did not allow us to make claims of the causality of attending university on attitude formation, suggestive results from the descriptive study largely aligned with the idea of education generating attitudinal change. Consequently, we underscore that if there indeed is a causal effect of attending university, the underlying mechanism is unlikely to be peer socialization.

These findings further contribute to our understanding of broader socialization processes, beyond a focus on the ‘higher-education effect’. The socialization effect may only apply to individuals from low socioeconomic backgrounds (see e.g., Lindgren et al., 2019), or to earlier stages in an individual’s life. Since the results diverge from the findings by Mendelberg et al. (2017), and considering that Strother et al. (2021) identified socialization effects only for roommates, the socialization may need to be even more intense or occur in smaller groups in order to hold a formative impact on students’ attitudes in this phase of a students’ educational journey.¹³ The effects of peer socialization may also be conditioned by the political environment through its impact on the saliency of political issues within the peer group (cf. Mendelberg et al., 2017). As such, there are higher levels of political polarization in the U.S, where both cited studies were fielded, which may increase the intensity and frequency of political discussions and thus the strength of the socialization. Furthermore, it is sometimes claimed that conflict avoidance is a characteristic of Nordic culture (cf. Daun & Helgesen, 2006) which might make students less prone to discuss sensitive political issues with each other. At any rate, inter-personal cultural differences have to our knowledge not been studied as a mediating factor in political socialization processes in educational institutions.

While students at elite business schools may be particularly likely to experience formative peer socialization (Schleef, 2005; Tyllström et al., 2022), the null result could also be attributed to the homogenous background of the students. These business students are predominantly from privileged family backgrounds, often more assertive than other student groups, and may possess strong pre-existing dispositions (Vedel & Thomsen, 2017). They could therefore be less susceptible to attitudinal change through peer interactions. Homophily in political values within close peer groups, as similar students are likely to be drawn to each other, may further cancel out any potential socialization effects (cf. McPherson et al., 2001), as would the fact that students who deviate from the political norms may interact less with their peers. While this possibility cannot fully be ruled out by this study, it would not explain the null result from the restrictions in social interactions imposed by the pandemic. Even if the study focused on business students, the data included different types of business schools in two different countries which might point to some of our findings being relevant also in other contexts. The fact that we find no consistent differences between schools also points to the fact that schools’ organization per se might have little to do with students’ values and homophily (similarly to the findings of Mayer & Puller, 2008). The most significant school difference–that Hanken differs on the issue of welfare dividends–is expected as Finland, unlike Sweden, already prohibits them.

One of the challenges of studies concentrating on peer socialization is to fully separate the possible effects of socialization from cognitive development resulting from the educational content. When students begin their studies in higher education, several processes coincide: they learn new things, develop skills, and become embedded in a new social context. To disentangle the mechanisms, one way forward could be to examine potential shocks to the school class composition, such as political reforms that increase school segregation or compare differences between students who stay on campus and those who commute. This study partially overcame these issues by leveraging data on students’ social behavior and political attitudes and concentrating on the cohorts’ attitude distribution. It thereby contributes both to the literature on the general ‘higher education effect’ as well as to the specifics of business education on students’ attitudes and values (Fischer et al., 2017; Miragaya et al., 2023). Naturally, this study calls for further research to test the results in different contexts and within more heterogeneous student groups in order to unequivocally challenge the peer socialization hypothesis. Nonetheless, this study reveals that the most prevailing mechanism to explain an ‘higher education effect’ is not as robust nor uniform as previously hypothesized.