Sage Journals: Discover world-class research

Abstract

The author examines the role of affiliation networks in shaping the political identities of students in college, using panel survey data from a case study of a predominantly liberal institution, tracking students’ political identities and affiliation memberships throughout the course of college. Although there was some self-selection into politically homophilous student organizations and majors, the extent of political sorting was relatively low, which resulted in considerable political heterogeneity in the affiliation networks. During the course of college, students’ political identities shifted in both liberal and conservative directions. Results from hierarchical multinomial logistic regressions suggest that identity transitions were driven by both the political composition of peer networks and influences outside the educational institution, such as family and prior socialization. This research underscores the importance of considering network stratification and individual contexts for understanding heterogeneous influences of seemingly uniform institutional settings.

Keywords

political socialization political identity peer effects affiliation networks hierarchical multinomial models

A large body of research documents an association between college attendance and liberal political attitudes (Campbell and Horowitz 2016; Chickering and McCormick 1973; Funk and Willits 1987; Glaser 2001; Kane 1995; Phelan et al. 1995; Vedlitz 1983). This liberal bent among college-educated individuals, in addition to the fact that professors tend to lean left (Gross and Fosse 2012; Lipset 1982), now and again raises concerns that colleges may be sites of liberal “indoctrination” (Horowitz 2007; Shapiro 2010; Smith, Mayer, and Fritschler 2008).

Empirical research largely dismisses such claims, showing that political influence in college operates not through institutional characteristics but peers (Dey 1997; Hanson et al. 2012; Milem 1998). While this scholarship has been critical for recognizing the role of peers, two key gaps remain. First, prior scholarship has defined peer contexts broadly as the average political ideology of the entire student body or cohort, overlooking potential diversity across social networks within college. Large social systems often embed subgroups that are the loci of social interactions (McFarland and Pals 2005; McPherson and Smith-Lovin 1987; Roethlisberger and Dickson 1939), with effects of structural influences being contingent upon interactions within these groups (Friedkin and Johnsen 2011:15). Thus, even within the same college, the direction of peer influence can vary. Second, past research does not always capture “extrainstitutional” influences, such as parental attitudes and precollege socialization. These omissions could be consequential if there are systematic differences in extrainstitutional characteristics between students with different political leanings. For instance, if students from liberal families have more political exposure at home than those from conservative families, the former may be less susceptible to political influence in college. Without controlling for such relevant factors, we risk misestimating the effect of peers.

In this article I draw on a unique case study of a selective private university at which students were surveyed annually from 2013 to 2017, starting the summer before college—and hence before they began interacting with one another—until the year they graduated.¹ Using information on students’ academic and social memberships, I construct affiliation matrices to proximate their peer networks during each year of college. Data on precollege identities allows estimation of the extent of political sorting into memberships, while repeated observations allow tracking of changes within the same individuals over time. Results from hierarchical multinomial models suggest that differences in peer contexts propelled both liberal and conservative shifts in identity and that extrainstitutional factors remained relevant for ideological development in college.

Background

Affiliation Networks as Social Environments

A central idea in sociological theory is that social environments shape individuals (Berger and Luckmann 1966; Blumer 1969; Durkheim [1893] 1984; Goffman 1968; Mead 1934; Simmel 1955).

Defining the boundaries of social environments, however, is challenging. At any given time, individuals are members of multiple social groups (Feld 1980; Simmel 1955). These groups can result in varying degrees of closeness, ranging from “anchored relations” in which members have direct and intimate connections to “anonymous relations” in which members barely recognize one another (Goffman 1968:189–91). Even within the same group, closeness among group members varies (McPherson and Smith-Lovin 1987). The social environments individuals come to inhabit are often nebulous, and individual experiences within those environments may be heterogeneous.

Past research has used various approaches to approximate the diffuse nature of social environments in order to examine their influence on identity and attitudes. Some common approaches include analyzing entire organizations (Levine 1972), crowd networks (McFarland and Pals 2005), self-reported strong and weak ties (Walker 2015; Walker and Lynn 2013), and affiliation networks (Fujimoto, Unger, and Valente 2012). Self-reported strong ties may be the strongest sources of influence (Verbrugge 1977), but analyzing influence via close ties is rife with challenges. First, prior homophily predicts friendship in egocentric networks, making it difficult to parse out the role of influence versus self-selection (Kandel 1978; Shalizi and Thomas 2011). This is relevant for the study of political identity because egocentric networks tend to be politically homogeneous (Knoke 1990; Verbrugge 1977). Second, self-reported ties can be subject to recall bias (Borgatti and Halgin 2011; Marsden 1990). In these respects, networks based on affiliations are analytically attractive: they provide a fuller picture of individuals’ broader social milieu and they may not be equally driven by self-selection, especially if organizations direct or place limitations on affiliation memberships (Kossinets and Watts 2009; McFarland 2006; McPherson and Smith-Lovin 1987; Verba 1961).

Networks based on affiliations versus friendship are by no means substitutes for each other. Although affiliation memberships can give rise to close ties, organizations vary in their level of constraints on individuals’ time and opportunities for interpersonal interactions (Feld 1980; McPherson and Smith-Lovin 1987). As such, the nature of influence through affiliation memberships may be distinct from that of close ties. Nevertheless, affiliation-based ties are consequential in various contexts. For example, voluntary organization memberships predict elite influence (Cornwell and Dokshin 2014), and cigarette exposure in sports teams increases smoking among adolescents (Fujimoto et al. 2012). Even generic affiliation networks, such as union membership or church attendance, have been shown to have implications for political attitudes (Light 2015). Most relevantly, prior research finds that political discussion networks in college are shaped by students’ voluntary memberships (Eveland and Kleinman 2013).

Peer Influence

The two primary mechanisms of influence in peer groups are conformity to group norms and social comparison with role-based others. Social groups can enforce conformity by imposing penalties for deviant behavior (Simmel 1950:92–93; Verba 1961). For instance, individuals may adjust their political views to conform to the majority because they face a “pressure for consensus in order to avoid overt conflicts within the group . . . that will be disturbing both to the group and the individual member” (Verba 1961:27). A second process of influence is social comparison: as groups create shared role identities, individuals come to view others in similar role positions as their reference group and adjust their beliefs and opinions to conform with those of the prototypical group member (Asch 1955; Festinger 1954; McCall and Simmons 1978; Sherif and Sherif 1964; Shibutani 1955).

Homogeneous groups can provide “mutual verification contexts” that stabilize identities because people seek and desire to maintain such contexts (Burke and Stets 2009). In contrast, changes in social settings that deprive individuals of the means for verification of their identities cause psychological distress (Burke 1991; Burke and Harrod 2005; Trettevik 2016), and provide impetus for changing toward the normative orientations of the new group (Long and Hadden 1985). Consistent with this theory, empirical research on adolescents and college students suggests that peer group homogeneity has a stabilizing effect. For instance, observational studies find that having friends who share one’s religious beliefs decreases the probability of religious liberalization in college (Maryl and Uecker 2011), and group homogeneity in adolescent networks discourages identity change (McFarland and Pals 2005). Similarly, experimental research suggests that when presented with new information, college students embedded in politically homogeneous networks resist changing political attitudes more than those in relatively diverse networks (Levitan and Visser 2009).

Network density, or how cohesive or closely knit a network is, could regulate the pressure to conform that individuals face in social networks (Bott 1955; Coleman 1988). Dense groups can deter individuals from changing identities because people develop reputations in close-knit networks (Coleman 1988), and reputational concerns in turn affect behaviors (Fernandez and Fernandez-Mateo 2006; Gould 1999). It is possible that once political identities are established in a cohesive network, it becomes harder for individuals to change, because their decisions are likely to be scrutinized by friends and entail the embarrassing admission that one was wrong or ignorant. In a high school context, McFarland and Pals (2005) found that network cohesiveness indeed decreases the probability of identity change.

Extrainstitutional Influence

After peers, parents are a key source of social influence on college students. Prior research shows that parents play an important role in the development of political attitudes and behaviors (Glass, Bengtson, and Dunham 1986; Iyengar, Konitzer, and Tedin 2018; Rico and Jennings 2016; Tedin 1974; Westby and Braungart 1966). In addition, children’s politically relevant behaviors, such as civic engagement (McFarland and Thomas 2006) and news consumption (Edgerly et al. 2018), are influenced by parents. Although adults do not always continue to reflect their parents’ ideologies (Bucx, Raaijmakers, and Van Wel 2010; Jennings, Stoker, and Bowers 2009), parting ways is costly. For example, discussing clashing political views with family may be psychologically taxing (Bradford, Vaughn, and Barber 2008; Pruchno, Burant, and Peters 1994), while evading such arguments may lead to cognitive dissonance (Festinger 1957; Festinger and Carlsmith 1959), another form of psychological distress. Balance theories suggest that avoiding such conflict may become an incentive for maintaining political consensus with parents (Taylor 1970).

Finally, individuals vary in their susceptibility to social influence. In the standard model of network effects, variation in susceptibility is often captured by an idiosyncratic term, estimated as a residual (Friedkin and Johnsen 1999:57, 75). In the context of political influence, past experiences and exposures may determine how open individuals are to social influences. One mechanism of this influence could be informational. For example, a variant of the Asch experiment finds that people are more likely to observe and follow others under conditions of uncertainty (Deutsch and Gerard 1955), a finding affirmed by more recent studies (Melamed and Savage 2013; Melamed, Savage, and Munn 2019). Similarly, students with relatively little political exposure prior to entering college may actively seek cues regarding “appropriate” political values and behaviors from their peers. In addition, political participation itself may strengthen future political attitudes (Quintelier and Hooghe 2012). As such, individuals who enter social settings with low levels of political exposure may be more malleable than their peers.

Data and Methods

Observing the effects of networks on political identities requires longitudinal data on identities, information about social environments within college, and some measures of extrainstitutional contexts. Ideally, one would study these processes in a nationally representative sample to capture the full diversity of higher educational settings. Unfortunately, such rich and comprehensive data are not readily available and are costly to collect. I use the next best approach of analyzing a case study in which all measures of interest have been recorded. I use data from a longitudinal survey of a cohort of college students at Stanford University, a private, nonprofit institution ranked among the top 10 undergraduate colleges by U.S. News & World Report in 2018. The data were collected annually through online surveys from 2013 to 2017, and respondents were offered a small monetary incentive for participation.²

Stanford University is generally considered a politically liberal institution, but it also houses conservative research centers and student groups. University employees and organizations’ financial support for political parties provides a useful glimpse into the political milieu: in 2014 and 2016, while the majority of university-affiliated individuals provided political contributions to the Democratic Party, some Stanford-affiliated organizations largely supported the Republican Party.³ According to a survey conducted in 2004, the ratio of Stanford faculty members registered as Democrat versus Republican voters was 7.6 to 1 (Klein and Western 2004).

During the summer of 2013, the entire incoming class of 2017, comprising 1,674 students, was invited to participate in the study. In the first wave, 733 usable responses were collected, yielding an effective response rate of 44 percent. Subsequent surveys were sent out at the end of each academic year to those who responded to wave I. Respondents moved in and out of the study in subsequent waves, as is typical with longitudinal data (Hillygus and Snell 2015). Ancillary analyses show that the probability of attrition does not systematically vary by the observed characteristics of the respondents. This analysis is based on 621 respondents who participated in wave I, have available data for at least two consecutive survey years, and have no missing data on key variables. The analytical data set contains 2,006 person-year observations from surveys conducted between 2014 and 2017. On average, respondents are observed for three survey waves. Restricting the sample only to students observed in all survey waves does not change the substantive findings (Appendix A).

Table 1 compares the analytical sample with cross-sections of undergraduate students in the 2016–2017 academic year at Stanford University and selective four-year U.S. colleges. Demographically, Stanford University students are comparable with those at other selective U.S. colleges. However, the analytical sample overrepresents women and Asians and substantially underrepresents first-generation students, compared with the average Stanford undergraduate in academic year 2016–2017. Ancillary analysis shows that results are stable when probability weights are used to adjust for representativeness (Appendix B). The average age of entry in the analytical sample was recorded during the summer before college and is thus lower than Stanford’s average entry age.

Table 1.

Comparison of Analytical Sample with Stanford University Class of 2017 and Comparable Institutions.

	Analytical Sample	Stanford Class of 2017	Top 10 Four-Year Institutions^a
Average age of entry (years)	17.92	18.00	20.48
Percentage female	.55	.46	.49
Percentage white, non-Hispanic	.38	.35	.40
Percentage black, non-Hispanic	.04	.06	.06
Percentage Asian, non-Hispanic	.28	.19	.23
Percentage Hispanic	.18	.16	.12
Percentage first-generation	.14	.30	.23
Shares of degrees awarded by major
Natural sciences	.09	.09	.16
Social sciences	.16	.16	.21
Engineering	.20	.24	.20
Computer sciences	.19	.15	.10
Interdisciplinary major	.20	.17	.04
n	619	7,032	54,474

Source: College Scorecard 2016–2017, Stanford class of 2017 survey, and Stanford Common Data Set (https://ucomm.stanford.edu/cds/).

Includes Princeton University, Harvard University, the University of Chicago, Yale University, Columbia University, the Massachusetts Institute of Technology, Stanford University, the University of Pennsylvania, Duke University, and the California Institute of Technology. Based on U.S. News & World Report 2018 rankings.

Memberships in a total of 95 affiliations were recorded each year, including up to two actual or intended majors (63 fields of study) and extracurricular activities (32 organizations).⁴ Data on extracurricular activities are based primarily on self-reported sponsoring organizations of students’ top three commitments on campus each year. These responses were entered as free text in the survey, and I manually coded them to identify 100 groups. Of these, I retained affiliation with at least 20 observations over the entire panel.⁵ For some affiliations, information provided by the sponsoring organizations was not sufficiently consistent or detailed across all respondents.⁶ In these cases, I drew on another survey question that provided reliable, albeit more aggregated, information on memberships.⁷ Examples of extracurricular affiliations include racial/ethnic student organizations (separate affiliations for black, Asian, and Latinx), employment (separate affiliations for residential and dining enterprises, residential staff, university alumni/visitor center, etc.), lab research (separate affiliations for psychology, biology, physics, etc.), medicine and public health research or service, mental health service, university journalism, arts (separate affiliations for theater, dance, music, etc.), student government, athletics (separate affiliations for club and varsity), and so on. Examples of majors include engineering (separate affiliations for biomechanical, computer, electrical, etc.), language (separate affiliations for German, Latin, Italian, etc.), archeology, film, mathematics, urban studies, and so forth.

Four bimodal affiliation matrices were constructed for each year of college, on the basis of students’ memberships during the past 12 months. I converted these two-mode matrices to unimodal matrices in which ties were defined as the sums of the cross-products of associations (Borgatti and Everett 1997). All network measures were estimated in UCINET (Borgatti, Everett, and Freeman 2002) and merged to the main dataset. Statistical analyses were conducted in Stata.

Measures

Dependent Variable

The outcome is a categorical variable indicating the direction of change in political identity over the previous year.⁸ Respondents annually reported their political orientation as one of the following: very liberal, liberal, moderate, conservative, very conservative, apolitical, and unsure. Although transitions to and from all political positions are substantively interesting, only a few transitions can be meaningfully analyzed because of data limitations. I first collapsed the responses into four categories: liberal, conservative, moderate, and apolitical or unsure.⁹ I then aggregate the transitions into four categories that are of the greatest substantive interest: (1) liberal identity transitions, entailing any change from conservative, moderate, or apolitical/unsure to liberal and any change from conservative to moderate or apolitical; (2) conservative identity transitions, constituting any change from liberal to any other identity and any change to conservative from any other identity; (3) nondirectional transitions constituting changes from moderate to apolitical and vice versa; and (4) no identity change, which is the reference group. In the analytical sample, 74 percent of the observations reported no change in identity over the previous year, 12.5 percent transitioned toward liberal identities, 8.6 percent transitioned toward conservatism, and 4.9 percent reported nondirectional changes.

Peer Network Homogeneity

Homogeneity of the peer group is defined as the proportion of alters who share a particular attribute (McFarland and Pals 2005; McPherson and Smith-Lovin 1987). I calculate political homogeneity as follows:

H o m o g e n e i t y_{i t} = \frac{\begin{array}{l} t i e s w i t h a l t e r s t h a t s h a r e \\ p o l i t i c a l i d e n t i t y_{i t} \end{array}}{t o t a l t i e s_{i t}},

where a tie indicates having at least one common affiliation. This variable is calculated with respect to the focal person i at time t and treats the ties as dichotomous (in other words, ties are unweighted). I use political identities reported in the last survey wave and affiliations reported at the end of the year, pertaining to the prior 12 months. As such, this variable captures lagged social influence.

Peer Network Density

Density is measured as the ratio of the number of alter-alter ties in one’s network to the total number of possible ties among alters (Prell 2012). If L is the number of total alter-alter ties, and n is the total number of alters an ego i is connected to, then ego network density is calculated as

D e n s i t y_{i t} = \frac{L_{i t}}{n_{i t} (n_{i t} - 1) / 2} .

I estimated density on the unimodal matrix using weighted ties, where the ties were weighted inversely by affiliation size, or “column totals” (Borgatti and Everett 1997). I standardized this variable for ease of interpretation.

Political Congruence with Parents

Congruence with parents is a binary measure indicating whether students shared their political views with at least one parent at the beginning of the academic year. Parents’ political ideology is based on student reports from the first survey wave.¹⁰ Although most respondents provided data on both parents, 43 students reported a single parent, and 21 mentioned only nonparental guardians. I use a dichotomous measure to include students who did not have two-parent families. Ancillary analysis disaggregating consensus with one or both parents in two-parent families is presented in Appendix C.

Table 2 provides summary statistics of the three network measures.

Table 2.

Descriptive Statistics of Network Measures.

	Mean	Standard Deviation	Minimum	Maximum
Political homogeneity with peers (percentage)	39.95	20.40	.00	100.00
Density of peer network (unstandardized, percentage)	.07	.37	.01	6.25
Political homogeneity with parents (dichotomous, proportion)	.54	NA	0	1

Source: Stanford class of 2017 survey.

Note: NA = not applicable.

Political Exposure

I measure political exposure before college as an index derived from six items reported in the precollege survey pertaining to frequency of activities in the last two years of high school. These include (1) frequency of political discussions with friends, (2) frequency of political discussions with family, (3) frequency of participation in election campaigns, (4) frequency of involvement in organized political activities, (5) frequency of participation in social movements, and (6) frequency of participation in political protests. Because these measures are substantially correlated, I used factor analysis to derive a single measure of latent political exposure (Fabrigar and Wegener 2012). The scale reliability coefficient is .67. Appendix D describes the detailed methodology used to construct this index.

Control Variables

Political frames can often invoke identities such as race, gender, and class, that might make certain political ideologies more attractive to some demographic groups. As such, I include controls for gender, race, and first-generation status. First-generation status is used as an indicator of socioeconomic status (SES). Two other measures of SES available in the survey (family income and subjective social class) are strongly correlated with first-generation status but have more missing observations. The substantive findings are robust to alternative specifications of SES (Appendix E).

I include period fixed effects to account for historical events (such as the national political atmosphere, elections, social movements, and year in college) that potentially affected all students at a given point in time. Because age and period are nearly collinear, I do not control for age. This means that period effects also capture age effects.

Analytical Approach

I present the empirical analysis in two parts. The first part deals with a central problem that complicates inferences about social influence: namely, that individuals can select into affiliations on the basis of their political leanings. Homophilous selection of social ties complicates attribution of subsequent changes in attitudes to social influence (Kandel 1978; Shalizi and Thomas 2011). In affiliation networks, organizational features (such as the nature and variety of extracurricular activities, required courses, and course sequences) can limit the choice of affiliation memberships (McFarland 2006; McPherson and Smith-Lovin 1987; Verba 1961), which could decrease the extent of self-selection. However, homophilous memberships are not implausible in affiliation-based educational networks; for instance, one study revealed that members of a university who shared foci were significantly more alike than others in terms of attributes like gender and age (Kossinets and Watts 2009). Homophily within foci might also extend to other characteristics, including political views. As such, the extent to which students politically sorted in this setting is very much an empirical question.

To gauge the extent of political sorting in this setting, I compare the observed political diversity among affiliations with a hypothetical distribution that would be obtained had students been randomly assigned to affiliations in the frosh year.¹¹ To calculate the level of political diversity, I use the information theory index (H), which is a commonly used measure of diversity in settings with multiple population subgroups (Reardon and Firebaugh 2002; Theil 1971). The index is calculated as follows:

H = \sum_{P = 1}^{4} \sum_{A = 1}^{95} \frac{t_{A}}{T E} π_{A P} . l n (\frac{π_{A P}}{π_{P}}),

where H is an aggregate measure of segregation within affiliations (A) by political identity (P), π denotes proportions, t_A is a count of members within the affiliation and TE is the total sample size. The value of H varies between 0 (complete diversity) and 1 (complete sorting). The simulations randomly sorted students into the 95 affiliations, keeping fixed students’ political identities before college and the size of the affiliations. Comparing the diversity index observed in the data to the distribution that would have resulted from random assignment and the value that would be observed if there was complete political sorting provides an assessment of the extent of political selection.

Next, I examine the putative effects of social influences on political identity using a hierarchical multinomial logistic regression. The dependent variable is an unordered choice outcome and is appropriately modeled as a multinomial logit (Borooah 2011; Osborne 2017). In panel data, observations for the same individual are not independent of one another over time, which means that errors within individuals are correlated across survey waves (Hedeker 2003; Skrondal and Rabe-Hesketh 2003). To account for this interdependence, I use a hierarchical model that takes into account the nesting of observations within individuals (Anderson, Kim, and Keller 2013; Skrondal and Rabe-Hesketh 2003). The hierarchical multinomial model can be expressed as

l n \frac{P r (y_{i t} = m | x)}{P r (y_{i t} = b | x)} = β_{m} X_{i t} + u_{i} + ε_{i t},

where i indexes individuals and t indexes time. The reference category is b (no change in identity), and m depicts the other three outcome categories (liberal change, conservative change, or nondirectional change). The vector X represents time-varying and fixed independent variables, u is an idiosyncratic error term that varies across individuals, and ε_it is the residual error term. I lag all time-varying covariates by one year so that political identity in a given year does not influence the value of covariates in the prior year.

Results

I begin with a descriptive analysis of the extent and direction of political identity change during college, followed by a discussion of the diversity index. I then present findings from the hierarchical multinomial logistic regression.

Political Identity Transitions in College

Figure 1 shows the sequence plot of political identity transitions during the course of five years, beginning with the summer before college. Each horizontal line represents an individual, with white spaces indicating missing data. Comparing the precollege and graduating years suggests that college had an overall liberalizing influence: although just over half of the entering students identified as liberal, the share of liberal students rose to 67 percent in the senior year, a substantial increase of 13 percentage points. However, examining students’ trajectories in the intermediate years suggests that the path to political liberalization was often indirect. For instance, a majority of students changed toward nonliberal identities by the end of the first year, and many underwent multiple transitions during the course of college.

Figure 1.

Sequence index plot showing political identity transitions over the course of college.

Selection into Networks

Figure 2 shows the distribution of diversity scores on the basis of 1,000 simulations. The simulated distribution centers on an H score of .23, which is lower (closer to diversity) than the observed score of .31. The observed score is outside the range of the distribution generated by the simulations, suggesting that there was significantly more political sorting into affiliations than would be expected by chance. Nevertheless, the observed diversity score is much closer to random sorting (i.e., 0.23) than complete sorting (1.0). As such, although students’ political leanings did influence their selection into college majors and student organizations, it is not inaccurate to say that political ideology was not the primary driver of affiliation memberships. Importantly, as the affiliations were far from completely politically sorted, these data suggest that students gained substantial exposure to diverse political views through their affiliations during the first year of college.

Figure 2.

Frequency distribution of simulated political diversity in college affiliations among first-year students.

Social Influence and Political Identity Change

I now move to the results of the hierarchical multinomial model. Figure 3 plots the predicted probabilities of the outcome against varying levels of peer group homogeneity, calculated at the average values of covariates (full results in Table 3). Overall, political identities were more changeable when homogeneity levels were below the sample mean and substantially stabilized once homogeneity exceeded that threshold. This suggests that peer groups exercised influence on political identities as expected. However, the direction of this influenced varied considerably. At lower levels of homogeneity, identity changes are directed mostly toward liberalism, while a smaller proportion of predicted changes is nondirectional (i.e., to and from moderate and apolitical). This is not surprising given that the majority of students in this setting are liberal, and therefore students with less homogeneous networks are more likely to be nonliberal. However, interestingly, at the average level of homogeneity, the probability of a liberal change is negligible, whereas there is a 10 percent chance for a student to become more conservative. As homogeneity increases, the chance of rightward movement steadily grows. In other words, in this setting, although being a minority among their peers typically attracted students toward political liberalism or induced nondirectional changes in identity, being part of a majority group drew students toward conservatism.

Figure 3.

Predicted probability of direction of political identity change with affiliation network homogeneity.

Table 3.

Relative Risk Ratios from Hierarchical Multinomial Logistic Regression Predicting Direction of Change in Political Identity over Previous Year.

Independent Variable	Liberal Change vs. No Change	Conservative Change vs. No Change	Nondirectional Change vs. No Change
Network characteristics (t – 1)
Peer group homogeneity (percentage)	.897***	1.021***	.897***
Peer group homogeneity (percentage)	(.007)	(.007)	(.008)
Peer network density (standardized)	.949	1.286**	.385
Peer network density (standardized)	(.127)	(.146)	(.291)
Homogeneity × Density	1.003	.997	1.013
Homogeneity × Density	(.005)	(.002)	(.012)
Homogeneity with parents (binary)	.605**	.382***	.446***
Homogeneity with parents (binary)	(.127)	(.091)	(.111)
Precollege characteristics (t₀)
Political exposure	.894	.720**	.725
Political exposure	(.123)	(.094)	(.199)
Female (reference: male)	.997	.911	1.171
Female (reference: male)	(.210)	(.196)	(.338)
Asian (reference: white, non-Hispanic)	1.463	1.345	2.180**
Asian (reference: white, non-Hispanic)	(.379)	(.380)	(.760)
Other race/ethnicity (reference: white, non-Hispanic)	1.397	1.564*	.979
Other race/ethnicity (reference: white, non-Hispanic)	(.370)	(.389)	(.373)
First-generation college student (binary)	1.553	.689	1.458
First-generation college student (binary)	(.430)	(.210)	(.544)
Constant	1.901*	.078***	.822
Constant	(.661)	(.029)	(.352)

Source: Stanford class of 2017 survey.

Note: Clustered standard errors are in parentheses. N (observations) = 2,006. Model includes year fixed effects.

p < .05, **p < .01, and ***p < .001 (two-tailed tests).

Network density had a negligible effect on the relative risk for identity change for both liberal and nondirectional changes, and the interaction of density and homogeneity failed to reach significance across all outcomes. However, we do observe a significant and positive effect of density on the risk for conservative change. This effect is in the opposite direction of what was theoretically expected. Specifically, we had hypothesized that higher network density would deter identity change because of reputational concerns. These findings suggest that such concerns may not be as salient when it comes to political identity, and instead higher network density may even encourage identity change in certain contexts. However, given the low level of density in this network, the marginal effect of a standard deviation change in density is substantively small (Figure 4).

Figure 4.

Predicted probability of direction of political identity change with affiliation network density

Extrainstitutional Influences

Net of peer influence and demographic characteristics, both political congruence with parents and prior exposure deterred from ideological shifts. Overall, students whose identities were congruent with one or both of their parents had nearly half the risk of changing their identities compared with students who disagreed with their parents. It is relevant to recall that parental homogeneity is measured as a dichotomous variable in order to include students with nontraditional family structures (i.e., single parents and nonparental guardians). For students with two-parent families, a significant proportion of parents had conflicting political identities between themselves. Ancillary analyses restricted to two-parent families show that matching with both parents had a much stronger effect on identity stability compared to matching with just one parent (Appendix C).

Precollege exposure had a relatively weaker constraining effect on political identity. Although the coefficients of this measure are not statistically significant for all outcomes, the coefficient magnitudes are comparable across outcomes. Overall, it seems that after accounting for other sources of influence, political exposure prior to college has a marginal stabilizing influence on political identity.

Discussion

In this article I have examined the role of social factors in shaping political identity in college by drawing on the experiences of a class of students at one institution through the entire course of college. The results showed that instead of fully sorting into politically homophilous affiliations, entering students joined groups with a good deal of political heterogeneity and opportunity for exposure to diverse perspectives. Perhaps because of this broad exposure, students politically shifted in both liberal and conservative directions during college, even as at the macro-level, this was a predominantly liberal setting.

Some findings of the study are in agreement with prior theoretical assumptions. For instance, an overarching finding that held across all outcomes was that parents continued to figure as key actors when it came to political identity development. Although we cannot rule out other possible reasons for why parental consensus is so crucial, the fact that congruence with both parents carried more weight than agreement with a single parent (Appendix C) lends credence to a balance theory explanation. Net of parental influence, exposure to politics in high school was a weaker deterrent to ideological change. Together, these results underscore the continuing role of extrainstitutional factors for political socialization in college. I also found that political homogeneity of peer groups stabilized identities, which is consistent with prior work on attitude and identity change (Baldassarri and Bearman 2007; Burke 2006; Maryl and Uecker 2011; McFarland and Pals 2005).

Finally, and most important, the effect of peer influence, as measured jointly by the political composition and density of affiliation networks, was strongly shaped by the college setting. Given that liberal students constituted the numerical majority, it is unsurprising that students in less homophilous networks had a higher likelihood of shifting leftward compared with other directions. However, what is surprising is that students in highly homophilous networks had a nonnegligible probability of shifting rightward and that higher network density marginally increased the likelihood of a conservative change. In other words, being member of a majority liberal and dense network did not necessarily deter students from moving in the opposite ideological direction. Overall, these findings caution against any broad generalizations regarding the effects of peers on the direction of ideological change.

A relevant question this study prompts is whether educational and extracurricular affiliations may exercise different kinds of influence. In ancillary analyses, I estimated models using network measures derived separately from these two types of memberships. Results of these analyses are similar to each other and to the main analysis presented in this article, although the coefficients of political homogeneity attenuate when network measures derived from subsets of affiliations are used (Appendix F). The similarity in these results is not surprising given that there is substantial overlap between extracurricular and educational affiliations in this setting (e.g., all music majors engaged in music-related extracurricular activities, and the majority of computer science majors participated in computer science organizations). However, in a different institutional context, one might observe heterogeneous effects by nature of affiliation.

Before delving into the implications, some caveats need restating. An important limitation of this analysis is that unobserved individual characteristics might have led to both selection into affiliations and subsequent changes in political identities. Although the analysis of political diversity suggests that the level of explicit political sorting was low, it does not rule out selection bias. Particularly, if students selected into affiliations on the basis of latent proclivities that are also correlated with their future political identities, we would not be able to capture such selection using this approach. Second, these findings are based on a single case study and may not directly inform us about the ideological direction of change across all settings. Instead, the contribution of this study is in elucidating how the processes of political influence within an organizational setting are tempered by the local context. A better understanding of these general sociological processes can help develop a more accurate view about how college attendance might “make” students liberal or conservative. Third, although network affiliations are useful for understanding social environments, the measures used here are nevertheless approximations. Affiliations may exercise varying levels of constraints (Feld 1980) and can differ in the extent to which they are relevant for political identity. In ancillary analyses, I tried to address some of these concerns by accounting for two additional affiliation characteristics: degree centrality and the nature of educational focus. (Full details of variable measurement and analysis are presented in Appendix G.) Although the main findings of the study are robust to the inclusion of these variables, this exercise does not make up for all possible shortcomings of these data.

How colleges shape students’ political identities is an important question for several reasons. First, college is a potential site of political socialization. The salience of politics increases during college as young adults reach the age to vote, making political attitudes more plastic (Visser and Krosnick 1998). For students in residential colleges, this age effect may be amplified because of simultaneous changes in their social networks and routines (Munson 2010). Second, colleges are pathways for achieving and intergenerationally reproducing wealth and political power. Elite college attendance contributes to higher income (Chetty et al. 2017), and selective higher education institutions in particular play a formative role in the development of political elites (Feeney, Hogan, and O’Rourke 2017). Unsurprisingly, most members of the U.S. Congress attended highly selective colleges (Dye and Ziegler 2009). College education also increases political and civic participation (Brand 2010; Hillygus 2005). Altogether, given that college graduates have disproportionate control over political and economic resources in society, it is important to understand how college attendance contributes to the development of political ideologies. In particular, understanding this process is important for addressing concerns that colleges may be sites of liberal political indoctrination.

This research offers two contributions to the extant literature on the political effects of college. First, this study suggests that social influence is not a simplistic function of the overall political climate at a college but instead is significantly shaped by the composition of peer networks. In turn, this composition depends on the relative sizes of political groups in the entering class and how students sort into affiliations. Second, this research highlights the continuing relevance of extrainstitutional factors in early adulthood. Overall, the making of both liberal and conservative identities may be the outcome of similar types of social influences interacting in different ways. Ultimately, apart from being the loci of student interactions, colleges themselves may play a minimal role in this process.

Supplemental Material

sj-docx-1-srd-10.1177_2378023120982435 – Supplemental material for How College Makes Liberals (or Conservatives)

Supplemental material, sj-docx-1-srd-10.1177_2378023120982435 for How College Makes Liberals (or Conservatives) by Tamkinat Rauf in Socius

Footnotes

Acknowledgements

The data set used in this research was collected by Douglas McAdam, Priya Fielding-Singh, and Jennifer Hill and has been used with their permission. I thank Douglas McAdam, Jeremy Freese, Robb Willer, Dan McFarland, G. Robin Gauthier, and Austin Van Loon for their extensive feedback and helpful conversations. I am extremely grateful to two anonymous reviewers and the editors of Socius for their thoughtful comments on the manuscript.

Research Ethics Statement

Stanford University’s institutional review board approved the original survey that collected primary data on human subjects as part of a larger and independent study conducted by Douglas McAdam, Priya Fielding-Singh, and Jennifer Hill. I requested additional institutional review board protocol approving the use of deidentified secondary data for this article (Stanford University IRB Protocol 46215 of 2018).

ORCID iD

Tamkinat Rauf

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biography

Tamkinat Rauf is a PhD candidate in sociology at Stanford University. Her research focuses on well-being, inequality, and political sociology.

References

Anderson

Carolyn J.

Kim

Jee-Seon

Keller

Bryan

. 2013. “Multilevel Modeling of Categorical Response Variables.” Pp. 481–520 in Handbook of International Large-Scale Assessment, edited by Rutkowski

von Davier

Rutkowski

Boca Raton, FL: Taylor & Francis.

Asch

Solomon E.

1955. “Opinions and Social Pressure.” Scientific American 193(5):31–35.

Baldassarri

Delia

Bearman

Peter

. 2007. “Dynamics of Political Polarization.” American Sociological Review 72(5):784–811.

Baldassarri

Delia

Gelman

Andrew

. 2008. “Partisans without Constraint: Political Polarization and Trends in American Public Opinion.” American Journal of Sociology 114(2):408–46.

Berger

Peter

Luckmann

Thomas

. 1966. The Social Construction of Reality: A Treatise in the Sociology of Knowledge. New York: Penguin.

Blumer

Herbert

. 1969. Symbolic Interactionism: Perspective and Method. Berkeley: University of California Press.

Borgatti

Halgin

Daniel S.

2011. “Analyzing Affiliation Networks.” Pp. 417–33 in The Sage Handbook of Social Network Analysis, edited by Scott

Carrington

Thousand Oaks, CA: Sage.

Borgatti

Stephen

Everett

Martin

. 1997. “Network Analysis of 2-Mode Data.” Social Networks 19:243–69.

Borgatti

Stephen P.

Everett

Martin G.

Freeman

L. C.

2002. “UCINET 6 for Windows. Software for Social Network Analysis.”

10.

Borooah

Vani

. 2011. “Multinomial Logit.” Pp. 45–76 in Logit and Probit. Thousand Oaks, CA: Sage.

11.

Bott

Elizabeth

. 1955. “Urban Families: Conjugal Roles and Social Networks.” Human Relations 8(4):345–84.

12.

Boutyline

Andrei

Vaisey

Stephen

. 2017. “Belief Network Analysis: A Relational Approach to Understanding the Structure of Attitudes.” American Journal of Sociology 122(5):1371–1447.

13.

Bradford

Kay

Vaughn

LaToya Burns

Barber

Brian K.

2008. “When There Is Conflict.” Journal of Family Issues 29(6):780–805.

14.

Brand

Jennie E.

2010. “Civic Returns to Higher Education: A Note on Heterogeneous Effects.” Social Forces 89(2):417–33.

15.

Bucx

Freek

Raaijmakers

Quinten

Van Wel

Frits

. 2010. “Life Course Stage in Young Adulthood and Intergenerational Congruence in Family Attitudes.” Journal of Marriage and Family 10(1):137–48.

16.

Burke

Peter J.

1991. “Identity Processes and Social Stress.” American Sociological Review 56(6):836–49.

17.

Burke

Peter J.

2006. “Identity Change.” Social Psychology Quarterly 69(1):81–96.

18.

Burke

Peter J.

Harrod

Michael M.

2005. “Too Much of Good Thing?” Social Psychology Quarterly 68(4):359–74.

19.

Burke

Peter J.

Stets

Jan E.

2009. Identity Theory. New York: Oxford University Press.

20.

Campbell

Colin

Horowitz

Jonathan

. 2016. “Does College Influence Sociopolitical Attitudes?” Sociology of Education 89(1):40–58.

21.

Chetty

Raj

Friedman

John

Saez

Emmanuel

Turner

Nicholas

Yagan

Danny

. 2017. “Mobility Report Cards: The Role of Colleges in Intergenerational Mobility.” Working Paper No. 23618. Cambridge, MA: National Bureau of Economic Research.

22.

Chickering

Arthur W.

McCormick

John

. 1973. “Personality Development and the College Experience.” Research in Higher Education 1(1):43–70.

23.

Coleman

James S.

1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology 94(Supplement):S95–120.

24.

Cornwell

Benjamin

Dokshin

Fedor A.

2014. “The Power of Integration: Affiliation and Cohesion in a Diverse Elite Network.” Social Forces 93(2):803–31.

25.

Deutsch

Morton

Gerard

Harold

. 1955. “A Study of Normative and Informational Social Influences upon Individual Judgment.” Pp. 571–604 in Interpersonal Behavior in Small Groups, edited by Ofshe

Englewood Cliffs, NJ: Prentice Hall.

26.

Dey

Eric L.

1997. “Undergraduate Political Attitudes: Peer Influence in Changing Social Contexts.” Journal of Higher Education 68(4):398–413.

27.

Durkheim

Émile

. [1893] 1984. The Division of Labor in Society. Edited by Halls

W. D.

London: Macmillan.

28.

Dye

Thomas R.

Ziegler

Harmon

. 2009. The Irony of Democracy: An Uncommon Introduction to American Politics. 14th ed. Boston: Wadsworth.

29.

Edgerly

Stephanie

Thorson

Kjerstin

Thorson

Esther

Vraga

Emily K.

Bode

Leticia

. 2018. “Do Parents Still Model News Consumption? Socializing News Use among Adolescents in a Multi-device World.” New Media and Society 20(4):1263–81.

30.

Eveland

William P.

Kleinman

Steven B.

2013. “Comparing General and Political Discussion Networks within Voluntary Organizations Using Social Network Analysis.” Political Behavior 35(1):65–87.

31.

Fabrigar

Leandre R.

Wegener

Duane T.

2012. “Introductory Factor Analysis Concepts.” Pp. 1–18 in Understanding Statistics: Exploratory Factor Analysis. Oxford, UK: Oxford University Press.

32.

Feeney

Sharon

Hogan

John

O’Rourke

Brendan K.

2017. “Elite Formation in the Higher Education Systems of Ireland and the UK: Measuring, Comparing and Decomposing Longitudinal Patterns of Cabinet Members.” British Educational Research Journal 43(4):720–42.

33.

Feld

Scott L.

1980. “The Focused Organization of Social Ties.” American Journal of Sociology 86(5):1015–35.

34.

Fernandez

Roberto

Fernandez-Mateo

Isabel

. 2006. “Networks, Race, and Hiring.” American Sociological Review 71(1):42–71.

35.

Festinger

Leon

. 1954. “A Theory of Social Comparison Processes.” Human Relations 7(2):117–40.

36.

Festinger

Leon

. 1957. A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.

37.

Festinger

Leon

Carlsmith

James M.

1959. “Cognitive Consequences of Forced Compliance.” Journal of Abnormal Social Psychology 58:203–10.

38.

Friedkin

Noah

Johnsen

Eugene

. 1999. “Social Influence Networks and Opinion Change.” Pp. 1–29 in Advances in Group Processes, Vol. 16, edited by Thye

S. R.

Lawler

New York: Emerald.

39.

Friedkin

Noah

Johnsen

Eugene

. 2011. Social Influence and Network Theory. New York: Cambridge University Press.

40.

Fujimoto

Kayo

Unger

Jennifer B.

Valente

Thomas W.

2012. “A Network Method of Measuring Affiliation-Based Peer Influence: Assessing the Influences of Teammates’ Smoking on Adolescent Smoking.” Child Development 83(2):442–51.

41.

Funk

Richard B.

Willits

Fern K.

1987. “College Attendance and Attitude Change: A Panel Study, 1970–81.” Sociology of Education 60(4):224–31.

42.

Glaser

James M.

2001. “The Preference Puzzle: Educational Differences in Racial-Political Attitudes.” Political Behavior 23(4):313–34.

43.

Glass

Jennifer

Bengtson

Vern L.

Dunham

Charlotte Chorn

. 1986. “Attitude Similarity in Three-Generation Families: Socialization, Status Inheritance, or Reciprocal Influence?” American Sociological Review 51(5):685–98.

44.

Goffman

Ervin

. 1968. Relations in Public: Microstudies of the Public Order. New York: Basic Books.

45.

Gould

Roger V.

1999. “Collective Violence and Group Solidarity: Evidence from a Feuding Society.” American Sociological Review 64(3):356–80.

46.

Gross

Neil

Fosse

Ethan

. 2012. “Why Are Professors Liberal?” Theory and Society 41(2):127–68.

47.

Hanson

Jana

Weeden

Dustin

Pascarella

Ernest

Blaich

Charles

. 2012. “Do Liberal Arts Colleges Make Students More Liberal? Some Initial Evidence.” Higher Education 64(3):355–69.

48.

Hedeker

Donald

. 2003. “A Mixed-Effects Multinomial Logistic Regression Model.” Statistics in Medicine 22(9):1433–46.

49.

Hillygus

D. Sunshine

. 2005. “The Missing Link: Exploring the Relationships Between Higher Education and Political Engagement.” Political Behavior 7(3):301–18.

50.

Hillygus

D. Sunshine

Snell

Steven A.

2015. “Longitudinal Surveys: Issues and Opportunities.” In The Oxford Handbook of Polling and Polling Methods, edited by Atkeson

L. R.

Alvarez

R. M.

Oxford, UK: Oxford Handbooks Online.

51.

Horowitz

David

. 2007. Indoctrination U: The Left’s War on Academic Freedom. New York: Encounter.

52.

Iyengar

Shanto

Konitzer

Tobias

Tedin

Kent

. 2018. “The Home as a Political Fortress: Family Agreement in an Era of Polarization.” Journal of Politics 80(4):1326–38.

53.

Jennings

M. Kent

Stoker

Laura

Bowers

Jake

. 2009. “Politics across Generations: Family Transmission Reexamined.” Journal of Politics 71(3):782–99.

54.

Judd

Charles

Krosnick

Jon

Milburn

Michael

. 1981. “Political Involvement and Attitude Structure in the General Public.” American Sociological Review 46(5):660–69.

55.

Kandel

Denise B.

1978. “Homophily, Selection, and Socialization in Adolescent Friendships.” American Journal of Sociology 84(2):427–36.

56.

Kane

Emily W.

1995. “Education and Beliefs about Gender Inequality.” Social Problems 42(1):74–90.

57.

Klein

Daniel B.

Western

Andrew

. 2004. “How Many Democrats per Republican at UC-Berkeley and Stanford? Voter Registration Data across 23 Academic Departments.” Retrieved December 11, 2020. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=664045.

58.

Knoke

David

. 1990. “Networks of Political Action: Toward Theory Construction.” Social Forces 68(4):1041–63.

59.

Kossinets

Gueorgi

Watts

Duncan J. J.

2009. “Origins of Homophily in an Evolving Social Network.” American Journal of Sociology 115(2):405–50.

60.

Krosnick

Jon A.

Milburn

Michael A.

1990. “Psychological Determinants of Political Opinionation.” Social Cognition 8(1):49–72.

61.

Leenders

Roger T.A.J.

2002. “Modeling Social Influence through Network Autocorrelation: Constructing the Weight Matrix.” Social Networks 24(1):21–47.

62.

Levine

Joel H.

1972. “The Sphere of Influence.” American Sociological Review 16(1):14–27.

63.

Levitan

Lindsey Clark

Visser

Penny S.

2009. “Social Network Composition and Attitude Strength: Exploring the Dynamics within Newly Formed Social Networks.” Journal of Experimental Social Psychology 45(5):1057–67.

64.

Light

Ryan

. 2015. “Like Strangers We Trust: Identity and Generic Affiliation Networks.” Social Science Research 51:132–44.

65.

Lipset

Seymor Martin

. 1982. “The Political Behavior and Values of Faculty Elites.” Public Opinion Quarterly 46(2):143–68.

66.

Long

Theodore E.

Hadden

Jeffrey K.

1985. “A Reconception of Socialization.” Sociological Theory 3(1):39–49.

67.

Marsden

Peter V.

1990. “Network Data and Measurement.” Annual Review of Sociology 16(May):435–63.

68.

Maryl

Damon

Uecker

Jeremy E.

2011. “Higher Education and Religious Liberalization among Young Adults.” Social Forces 90(1):181–208.

69.

McCall

George

Simmons

J. L.

1978. Identities and Interactions: An Examination of Human Associations in Everyday Life. New York: Free Press.

70.

McFarland

Daniel

Pals

Heili

. 2005. “Motives and Contexts of Identity Change: A Case for Network Effects.” Social Psychology Quarterly 68(4):289–315.

71.

McFarland

Daniel A.

2006. “Curricular Flows: Trajectories, Turning Points, and Assignment Criteria in High School Math Careers.” Sociology of Education 79(3):177–205.

72.

McFarland

Daniel A.

Thomas

Reuben J.

2006. “Bowling Young: How Youth Voluntary Associations Influence Adult Political Participation.” American Sociological Review 71(3):401–25.

73.

McPherson

J. Miller

Smith-Lovin

Lynn

. 1987. “Homophily in Voluntary Organizations: Status Distance and the Composition of Face-to-Face Groups.” American Sociological Review 52(3):370–79.

74.

Mead

George Herbert

. 1934. Mind, Self, and Society from the Standpoint of a Social Behaviorist. Chicago: University of Chicago Press.

75.

Melamed

David

Savage

Scott

. 2013. “Status, Numbers and Influence.” Social Forces 91(2):1085–1104.

76.

Melamed

David

Savage

Scott

Munn

Chris

. 2019. “Uncertainty and Social Influence.” Socius 5:1-9. Retrieved December 11, 2020. https://journals.sagepub.com/doi/full/10.1177/2378023119866971.

77.

Milem

Jeffrey F.

1998. “Attitude Change in College Students: Examining the Effect of College Peer Groups and Faculty Normative Groups.” Journal of Higher Education 69(2):117–40.

78.

Munson

Ziad

. 2010. “Mobilizing on Campus: Conservative Movements and Today’s College Students.” Sociological Forum 25(4):769–86.

79.

Ojeda

Christopher

Hatemi

Peter K.

2015. “Accounting for the Child in the Transmission of Party Identification.” American Sociological Review 80(6):1150–74.

80.

Osborne

Jason W.

2017. “Simple Linear Models with Polytomous Categorical Dependent Variables: Multinomial and Ordinal Logistic Regression.” Pp. 97–132 in Regression & Linear Modeling: Best Practices and Modern Methods. Thousand Oaks, CA: Sage.

81.

Phelan

Link

Bruce G.

Stueve

Ann

Moore

Robert E.

1995. “Education, Social Liberalism, and Economic Conservatism: Attitudes toward Homeless People.” American Sociological Review 60(1):126–40.

82.

Prell

Christina

. 2012. Social Network Analysis: History, Theory, and Methodology. Thousand Oaks, CA: Sage.

83.

Pruchno

Rachel

Burant

Christopher

Peters

Norah D.

1994. “Family Mental Health: Marital and Parent-Child Consensus as Predictors.” Journal of Marriage and Family 56(3):747–58.

84.

Quintelier

Ellen

Hooghe

Marc

. 2012. “Political Attitudes and Political Participation: A Panel Study on Socialization and Self-Selection Effects among Late Adolescents.” International Political Science Review 33(1):63–81.

85.

Reardon

Sean F.

Firebaugh

Glenn

. 2002. “Measures of Multigroup Segregation.” Sociological Methodology 32:33–67.

86.

Rico

Guillem

Jennings

M. Kent

. 2016. “The Formation of Left-Right Identification: Pathways and Correlates of Parental Influence.” Political Psychology 37(2):237–52.

87.

Roethlisberger

F. J.

Dickson

William J.

1939. “The Internal Organization of the Group in the Bank Wiring Observation Room.” Pp. 348–56 in Management and the Worker. Boston: Harvard University Press.

88.

Schaefer

David R.

Kreager

Derek A.

2020. “New on the Block: Analyzing Network Selection Trajectories in a Prison Treatment Program.” American Sociological Review 85(4):709–37.

89.

Shalizi

Cosma Rohilla

Thomas

Andrew C.

2011. “Homophily and Contagion Are Generically Confounded in Observational Social Network Studies.” Sociological Methods and Research 40(2):211–39.

90.

Shapiro

Ben

. 2010. Brainwashed: How Universities Indoctrinate America’s Youth. Nashville, TN: WND Books.

91.

Sherif

Muzafer

Sherif

Carolyn

. 1964. Reference Groups: Explorations into Conformity and Deviation of Adolescents. New York: Harper & Row.

92.

Shibutani

Tamotsu

. 1955. “Reference Groups as Perspectives.” American Journal of Sociology 60(6):562–69.

93.

Simmel

Georg

. 1950. “On the Significance of Numbers for Social Life.” Pp. 87–98 in The Sociology of Georg Simmel, edited by Wolff

K. H.

New York: Free Press.

94.

Simmel

Georg

. 1955. “The Web of Group-Affiliations.” Pp. 127–95 in Conflict and the Web of Group Affiliations, edited by Wolff

Bendix

New York: Free Press.

95.

Skrondal

Anders

Rabe-Hesketh

Sophia

. 2003. “Multilevel Logistic Regression for Polytomous Data and Rankings.” Psychometrika 68(2):267–87.

96.

Smith

Bruce L. R.

Mayer

Jeremy D.

Fritschler

Lee

. 2008. Closed Minds? Politics and Ideology in American Universities. Washington, DC: Brookings Institution Press.

97.

Taylor

Howard

. 1970. Balance in Small Groups. New York: Von Nostrand Reinhold.

98.

Tedin

Kent L.

1974. “The Influence of Parents on the Political Attitudes of Adolescents.” American Political Science Review 68(4):1579–92.

99.

Theil

Henry

. 1971. “A Note on the Measurement of Racial Integration of Schools by Means of Informational Concepts.” Journal of Mathematical Sociology 1:187–94.

100.

Trettevik

Ryan

. 2016. “Identities, Goals, and Emotions.” Social Psychology Quarterly 79(3):263–83.

101.

Vedlitz

Arnold

. 1983. “The Impact of College Education on Political Attitudes and Behaviors: A Reappraisal and Test of the Self-Selection Hypothesis.” Social Science Quarterly 64(1):145–53.

102.

Verba

Sydney

. 1961. Small Groups and Political Behavior. Princeton NJ: Princeton University Press.

103.

Verbrugge

Lois M.

1977. “The Structure of Adult Friendship Choices.” Social Forces 56(2):576–97.

104.

Visser

Penny S.

Krosnick

Jon A.

1998. “Development of Attitude Strength over the Life Cycle: Surge and Decline.” Journal of Personality and Social Psychology 75(6):1389–1410.

105.

Walker

Mark H.

2015. “The Contingent Value of Embeddedness: Self-Affirming Social Environments, Network Density, and Well-Being.” Society and Mental Health 5(2):128–44.

106.

Walker

Mark H.

Lynn

Freda B.

2013. “The Embedded Self: A Social Networks Approach to Identity Theory.” Social Psychology Quarterly 76(2):151–79.

107.

Westby

David L.

Braungart

Richard G.

1966. “Class and Politics in the Family Backgrounds of Student Political Activists.” American Sociological Review 31(5):690–92.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.06 MB