Falling Behind: Lingering Costs of the High School Transition for Youth Friendships and Grades

Academic achievement and social integration

Academic outcomes in high school remain critical indicators of educational success for young people. Having low grades in high school is linked to early school dropout rates (Battin-Pearson et al. 2000), whereas high grade point average (GPA) is associated with a higher college GPA and elevated rates of college degree completion (Zwick and Sklar 2005). Furthermore, scholarship dating back to the early social network research of Coleman (1961) in The Adolescent Society points to adolescent social life as a key element relating to academic outcomes. Youth embeddedness with their peers and school represents one element of the social capital that can substantially shape success in school (Coleman 1988). Academically oriented social ties, in particular, provide social capital in the form of protection against problems at school (Crosnoe, Cavanagh, and Elder 2003), and reciprocated friendships link to enhanced school belonging and higher grades (Vaquera and Kao 2008). In addition, having more close friends in high school reduces the likelihood of dropping out of school (Carbonaro and Workman 2013). A variety of friendship measures, such as friends’ GPA, affect changes in school performance and social behavior (Cook, Deng, and Morgano 2007), and classroom-level peer effects influence math and reading outcomes for students at multiple education levels (Burke and Sass 2008). The courses that adolescents take also shape their friendships and access to social capital, which can have consequences for their academic grades (Frank, Muller, and Mueller 2013). At the same time, the academic climate in a school or classroom has direct implications for adolescent friendship formation (McFarland et al. 2014). Given the importance of both social life and academics for young people’s future success, and considering that these two aspects of young people’s lives are entwined, we examine students’ friendship networks and their academic grades.

Life Course Perspective

In addition to using a social network framework, we borrow from a life course perspective on social processes (e.g., Alwin 2012; Elder 1994) to better understand the ways in which young people’s relationships and grades evolve over the high school years. A life course perspective helps contextualize structural transitions in the U.S. educational system, and it connects individual patterns to the broader social framework (Benner 2011; Langenkamp 2011). Two life course concepts are especially pertinent to our topic: linked lives and life transitions. According to the fundamental life course principle of linked lives (Elder 1994), individuals’ experiences are lived interdependently within a set of shared relationships that are embedded in personal networks. One major normative life event of early adolescence is puberty, and a primary goal during that life stage is to establish social network ties outside the family and, in particular, form friendships with peers from school (Wrzus et al. 2013).

According to Kahn and Antonucci (1980), the concept of linked lives can be thought of as a convoy that evolves over time and contains valued resources for people’s well-being and functioning. This notion of a network convoy is particularly informative for theories of social network change. Research suggests that the convoy’s inner core is likely to remain relatively consistent (e.g., family and close friends), whereas more peripheral relationships are less stable (Wrzus et al. 2013). As individuals mature through adolescence, they gain and lose ties from their network convoy, and this variation influences the social, psychological, and material resources to which they have access. For instance, if adolescents reconfigure their network ties by befriending peers who are enrolled in their same academic courses, the new friends could provide academic resources, such as assistance with homework, in addition to social connections.

A second core concept of the life course paradigm that is relevant to our research is that of life transitions. Transitions, or specific life-changing events, such as marriage, employment, and school changes, occur over relatively short time spans and are located within life course trajectories (Elder 1985). The life course does not simply proceed in a linear fashion along with aging; rather, it includes turning points that represent abrupt alterations in individuals’ experiences and outcomes.

Challenging transitions disturb established routines and social roles, and they also tend to disrupt social networks. These types of social upheavals can trigger subsequent changes and result in distress (Almeida and Wong 2009; Caspi and Moffitt 1993). Furthermore, a single life course transition has the potential to shape long-term life course trajectories (Alexander and Entwisle 1988; Hayward and Gorman 2004), which makes these turning points important to study. Considering these two core life course concepts, we will examine whether the transition to high school disrupts the process of building a convoy of linked lives, resulting in fewer friendships and greater social isolation, and whether transition costs extend to academic grades.

The Transition from Eighth to Ninth Grade

Normative school transitions between levels of the U.S. educational system represent noteworthy and often challenging phases of childhood and adolescence. The process of changing schools from elementary to middle school or from middle to high school necessitates adapting to a novel environment and different expectations. A person’s experiences during this process can have broad and lasting consequences. Transitions to another level of schooling, and in particular, the shift to high school, typically entail increasingly difficult academic challenges (Roderick and Camburn 1999). In most cases, moving school levels means young people are placed in classrooms with many new classmates, in addition to encountering different teachers, guidance counselors, coaches, club mentors, and support staff. Furthermore, given that extracurricular activities promote friendship formation (Schaefer et al. 2011), changes in sports and other school activities necessitated by compulsory school changes may offer opportunities for new acquaintances, but they also are apt to result in the termination of older, more established friendships. More generally, school transitions inevitably involve changes from familiar and routine environments to unfamiliar environments that may seem unpredictable (Caspi and Moffitt 1993). Note, too, that these changes occur at the same time adolescents face an array of developmental life course transitions at the physical, cognitive, psychological, and social levels, all of which could contribute to stress during this period.

Yet, life course theorists note that life transitions may not require major adjustments for individuals in all cases. Normative transitions, for example, pose less of a problem than unexpected major changes (Pearlin 2010), and institutionalized matriculation from middle school to high school represents a normative undertaking. Furthermore, the level of problems induced by a life transition is related to the degree of preexisting stress in the situation preceding the change (Wheaton 1990). In other words, transitions that remove a person from a troublesome circumstance can lead to less costly, or even beneficial, outcomes. Adolescents who struggle in the earlier grades, either socially or academically, are the most likely candidates for possible school transition benefits. Transferring to a new environment could represent an opportunity for such youth to begin afresh and expand their social ties, adding acquaintances with whom they interact for the first time in class, clubs, or sports teams. Some youth may take advantage of their shift in surroundings to tackle previously unmet educational challenges (Schiller 1999; Weiss and Bearman 2007), perhaps with the assistance of unique learning aids available in secondary schools.

Research Hypotheses

This study thus aims to examine the effects of school transitions on young people’s friendship networks and reported grades by comparing students who change schools when matriculating to high school with students who remain in the same institution. Our sample includes a substantial number of schools with and without transitions and thus provides a relatively strong empirical base. Studies of school transitions and youth friendships tend to focus primarily on cross-sectional data and ignore the evolution of such relationships over the course of the school year and beyond (Benner 2011; Newcomb and Bagwell 1995). Our research design goes beyond prior work by focusing on a lengthy period that begins in sixth grade and continues through the senior year of high school. Studying this span allows us to compare students’ social and academic progress before and after a shift to ninth grade, and we can investigate whether the possible effects of a transition persist past this initial year. Based on theory regarding life transitions, as well as evidence of the disruptive nature of life transitions on peer networks (e.g., Entwisle 1990), we hypothesize that structural school transitions from eighth to ninth grade will have negative ramifications for social integration, as measured by friendship indegree and social isolation. We expect the negative consequences to occur directly after the transition. Given the harmful effects of early school transitions on academic outcomes (Schwerdt and West 2013; Seidman et al. 1994; Wigfield et al. 1991), we also expect students’ grades will be negatively influenced by structural school transitions immediately following the shift.

Little research considers whether transition effects are likely to persist over the high school years. A move to a high school undoubtedly poses an initial shock to adolescents’ social and academic experiences, but they might recover relatively quickly, adapt to the new environment, and return to previous levels of social engagement and school performance. Several scholars (Barber and Olsen 2004; Blyth et al. 1983; Wigfield et al. 1991) argue along these lines, maintaining that most adolescents’ social declines are limited to the period directly after the transition and that ill effects are generally gone a year later. On the other hand, ninth grade is often viewed as the most critical high school year, one that “makes or breaks” students over the long run (McCallumore and Sparapani 2010). A new school context that contributes to dwindling friendship centrality and dropping grades may continue to pose challenges throughout high school. Once set in motion, social interaction styles and studying habits, for instance, may become resistant to change. If so, initial negative school transition effects could persist for longer than a year or two. Yet despite calls for lengthier studies on this issue, research seldom follows students long enough to examine such a scenario (Benner 2011). For these reasons, the current study examines the degree to which the influence of normative, structural shifts to high school tend to be long-lasting on adolescents’ social integration and academic outcomes by considering a particularly lengthy period of study. Given the potential problems arising from school changes, we expect the costs will continue past the first year of high school.

Finally, we consider whether there is further variation with regard to the types of structural transitions students experience. In our sample, we are able to differentiate between two types of structural transitions to high school. In some communities, students transition from a single, lower-level school to one high school; in other communities, students experience transitions that involve multiple feeder schools merging into a high school. We define the former as “single-feeder” districts and the latter as “multiple-feeder” districts. Both single- and multiple-feeder educational moves require students to confront novel learning environments, classrooms that fail to include some or all of their previous friends, and alterations in school protocol and expectations. Thus, we expect both types of school moves will entail more social and academic costs than remaining within the same school between eighth and ninth grades (see Hypotheses 1 and 2). Compared with single-feeder districts, however, multiple-feeder districts likely pose particular challenges, given the exacerbated levels of change in the student body as well as the need to assimilate students from diverse sets of educational backgrounds. With some exceptions (e.g., Langenkamp 2010), previous studies tend to find that, compared with school matriculation in single-feeder districts, school moves in multiple-feeder districts are particularly damaging to students (Blyth et al. 1983; Crockett et al. 1989; Temkin et al. 2015). We thus hypothesize that the effects of transitions involving multiple-feeder schools merging into one high school will be particularly costly compared with transitions from a single, lower-level school to one high school.

Sample

We use data on 14,462 students who attended middle or high school within one of 28 public school districts participating in the PROSPER study. Half the school districts are located in Iowa, and the other half are in Pennsylvania. Participating districts were required to have between 1,300 and 5,200 students, with at least 15 percent eligible for free or reduced lunch. All school districts were located in rural or small-city communities with populations ranging from 7,000 to 45,000.

Although our sample of communities is not nationally representative, it does have some valuable advantages. First, a substantial portion of the U.S. population lives in such communities, but these settings receive relatively little attention in educational research (White and Corbett 2014). In the 2010 Census, 19.3 percent of U.S. residents lived in rural areas, and 9.5 percent lived in communities with populations of 2,500 to 50,000 (U.S. Census Bureau 2010). Second, the relatively limited school choice in smaller communities has advantages for our research design. Friendship networks within grade cohorts should capture a larger share of the pool of likely friends, and these networks may be more stable over time than in communities with greater school choice. Also, each of our communities possesses only a single high school, which reduces the chances that apparent transition effects are due to individual-level selection, such as parents preferring schools with or without a transition to high school between eighth and ninth grades. In these communities, selecting a school district to avoid a secondary transition for a child would necessitate the major disruption of moving to an entirely new city or town.

Data were collected for two cohorts: students entering the sixth grade in 2002 and those entering in 2003. Self-administered surveys were distributed to students during the fall and spring semesters of their 6th-grade year and during the spring semesters of their 7th- through 12th-grade years. This yields eight waves of data that each includes an average of roughly 9,000 students. The target sample for each wave was all currently enrolled students in the grade level of interest. Students who moved out of the school district or were retained a grade were dropped from the sample; students who moved into the district or skipped a grade were added. Response rates were generally high over all eight waves, varying from 86 to 90 percent, as were participant retention rates; students who participated in the first wave of the survey completed an average of 5.90 waves. To simplify our analysis and ensure that our outcome variables are measured at consistent time points during the course of the school year, we omit the first wave of data that was collected during the fall of respondents’ sixth-grade years. ¹ We also exclude three district-cohort networks from our analysis due to irregularities (e.g., one district was affected by a fire), resulting in a final sample of 14,462 total students in 51 friendship networks from 26 school districts.

Definition of Variables

We measure individual students’ positions in their social networks by examining their individual indegree centrality, or the number of friendship nominations students receive from their peers (Wasserman and Faust 1994). This measure is often interpreted as popularity (Moody et al. 2011; Valente, Unger, and Johnson 2005). In each study wave, students were asked, “Who are your best and closest friends in your grade?” They were permitted to nominate up to two “best friends” and as many as five “other close friends,” allowing each student to nominate a maximum of seven peers. Although students were limited to seven friendship choices, there was no limit to the number of nominations a student could receive, resulting in a discrete measure of indegree that varies from 0 to 20 nominations.

Of the students who participated at any wave, 93.9 percent nominated at least one peer as a friend. Approximately 83.0 percent of the nominations were successfully matched to other students participating in the study; 14.7 percent of the nominations appeared to be students who were in another grade level or attended a different school. Coders were unable to match 1.9 percent of names because multiple possible matches existed, and they deemed .4 percent of names to be implausible (i.e., names of celebrities). The data include only within-community and within-grade friendship nominations, from which we constructed a total of 51 global friendship networks. In additional analyses, we also consider social isolation, which we define as receiving zero friendship nominations, as our dependent variable.

At each survey wave, students were asked, “What grades do you generally get in school?” Responses varied from “mostly As (90 to 100)” to “mostly lower than Ds (<60).” ² Following previous work (Fröjd, Kaltiala-Heino, and Marttunen 2011; Nelson, Nelson, and Malone 2004), we dichotomized this variable so that a score of 1 indicates the student reported mostly As or Bs and a score of 0 indicates the student received mainly Cs, Ds, or lower. ³

Structural school transitions are the main independent variables of interest in our study. Specifically, we focus on transitions that occurred between eighth and ninth grades. Youth from all school districts in our sample experienced at least one structural school transition during the course of the study. However, the timing of these transitions varied across the 26 communities. The vast majority of districts have students change schools between eighth and ninth grades (n = 22, or 84.3 percent of schools in the sample). Among the sample, 19 districts were single-feeder and the remaining three were multiple-feeder districts. For the purpose of our analysis, districts that do not have students physically transition between eighth and ninth grades are defined as control districts (n = 4). All but one of the districts in our sample have two separate grade cohorts. Other research that analyzes the PROSPER Peers data examines the less common transition from sixth to seventh grade (experienced by 35.3 percent) and its effect on multiple measures of social network centrality (Felmlee et al. 2018), and Temkin and colleagues (2015) studied additional aspects of friendship networks over a shorter time span. These studies reached similar conclusions to those obtained here for social integration. ⁴

To test for the effect of school transitions between eighth and ninth grades on students’ indegree and academic grade trajectories, we include two binary variables. The first is a time-invariant variable that indicates whether a student attends a school district that structurally transitions between eighth and ninth grades (1 = experienced eighth-/ninth-grade transition). The second, which varies over time, indicates the timing of an observation relative to the school transition of interest. For this second variable, students who attend a school that experiences an eighth-/ninth-grade transition will receive a score of 0 during their sixth- through eighth-grade years and a score of 1 for ninth grade and beyond; students who attend schools in control districts receive a score of 0 for each study wave. We include both of these dummy variables in our analysis. The coefficient for the first will reflect the pretransition difference between transition and comparison groups; the coefficient for the second will reflect increases or decreases in that difference following the transition. To test our final hypothesis, we replace these binary variables with new variables that further differentiate between types of transition, that is, whether a district merges multiple-feeder schools together (i.e., a multiple-feeder district) or moves students from a single middle school to a high school (i.e., a single-feeder district). The new variables allow us to directly consider the differential effect of transition type (no transition, multiple feeder, or single feeder) on friendship popularity and academic performance.

Other than having varying transition patterns, the school districts have relatively similar characteristics. As confirmed by two-sample t tests, there is no significant difference between transition districts and control districts in community population size, median household income, proportion of students on free/reduced lunch, average test scores, or dropout rates. Transition districts spend significantly less money per student, however, and are more likely to be in Iowa, compared with control districts. We account for these differences, as well as for variation in the size of the student body, through the inclusion of three district-level controls. We include a binary indicator for geographic location (0 = Pennsylvania, 1 = Iowa), a continuous measure for average money spent per student (in dollars per year), and a continuous measure for average school grade size (in number of students). ⁵

We include several other control variables that could be associated with students’ tendency to receive nominations from their peers or their academic performance. First, we include three time-invariant binary variables to control for students’ cohort (1 = cohort 1), gender (1 = girl), and race (1 = white). We also include several time-variant measures, including two binary controls for free/reduced lunch status (1 = receives free/reduced lunch) and living with both biological parents (1 = lives with both) and two continuous measures of school adjustment/bonding and delinquency. ⁶ We treat time as a continuous measure (as wave of measurement) and also include a squared term for time to account for curvilinear trends. ⁷ In our statistical models, all continuous variables, except for the wave polynomials, are centered by their grand mean to ease interpretation of their coefficients. Data were occasionally missing from specific wave- and individual-level variables, with missing data varying from .23 to 3.85 percent of cases. To account for missing data, we used multiple imputation (using chained equations) to estimate 10 imputed data sets.

Plan of Analysis

To test our research questions regarding the relationship between school transitions, individual centrality, and grades, we use multilevel models (MLMs) to account for the nested structure of our data. Respondents in our sample belong to one of two successive cohorts that were nested within 26 school districts. Thus, we use three-level longitudinal MLMs, where the first level represents the wave of the study, the second represents the student, and the third represents the school district. Three-level MLMs are similar to standard regression models except they take into account dependence among observations by allowing intercepts and slopes to vary by the groupings at higher levels (Hox 2010). The design of our models allows us to test not only how a transition affects indegree centrality and grades in the year immediately following a move, but it also enables us to examine effects for the remainder of students’ high school careers while controlling for individual- and district-level factors.

For each dependent variable, we estimate three different multilevel models. The first model, which tests whether we expect to see different outcomes for students in the years following the transition, uses indegree as the dependent variable:

Y i j k | λ i j k = Poisson ( m i j k , λ i j k )

η i j k = ln ( λ i j k )

η i j k = γ 000 + γ 001 T k + ∑ γ 00 x X k + γ 010 C j k + ∑ γ 0 y 0 P j k + γ 100 P S T i j k + γ 200 W i j k + γ 300 W i j k 2 + ∑ γ z 00 Z i j k + v 0 k + u 0 j k + u 02 k C j k

Because this outcome is a count variable, Equation 1 specifies a Poisson distribution for indegree at wave i for respondent j in district k, Y_ijk, and includes an overdispersion term, m _ijk . Equation 2 is a link function that translates the response variable, η _ijk , to a metric appropriate for a linear model. Equation 3 represents our combined model, where γ₀₀₀ is the district-level intercept and T_k is the binary variable that indicates whether district k experiences a transition between eighth and ninth grades. In other words, by interpreting γ₀₀₁, we can distinguish whether a significant difference exists in the outcome variable for different district types before the transition occurs. The variable C_jk indicates individual j’s cohort, and PST_ijk is the time-varying post-transition variable. We account for time by including two polynomials for wave, W_ijk and W_ijk², and we include controls at the district (X_k), individual (P_jk), and wave (Z_ijk) levels. The variables ν_0k and u_0jk are unexplained residual terms at the district and individual level, respectively. Finally, the nesting of grade cohorts within districts plays a unique role in our study because students were asked to name friends only from their district-specific cohort, that is, peers in their same school and grade. Therefore, it is important to address any statistical dependence arising from cohort differences that are idiosyncratic to particular districts. We do so by allowing cohort difference to vary across districts through the random effect u_01k. Note that the distribution in Equation 1 and the link function in Equation 2 can be adjusted to account for our binary dependent variables.

Model 2 builds on the first model by including cross-level interactions between each of the wave polynomials, W_ijk and W_ijk², and the binary variable that indicates whether students in a district experience an eighth-/ninth-grade transition, T_k. This model tests whether students in transition districts are able to “catch up” to the control districts as they progress through high school. Model 3 differentiates between single- and multiple-transition districts and allows us to test whether differences between the two transition types are statistically significant.

Descriptive Statistics

Our sample is composed of slightly more girls (51.5 percent) than boys, and there is relatively little racial diversity: 84.9 percent of students are white (see Table 1). Over the course of the investigation, the average number of friendship nominations received by each student was 3.34, with this measure varying from 0 to 20. The students in the sample tended to report relatively high grades; 73.9 percent of students reported receiving mainly As or Bs.

OPEN IN VIEWER

Table 1.

Individual-level Descriptive Statistics Averaged over All Study Waves.

Variable	M	SD	Min.	Max.
Indegree	3.344	2.682	0	20
Proportion of isolates	0.052	—	0	1
Transition after sixth grade	0.345	—	0	1
Transition after eighth grade	0.840	—	0	1
Transition after ninth grade	0.115	—	0	1
Cohort 1	0.487	—	0	1
Iowa	0.501	—	0	1
Female	0.515	—	0	1
White	0.847	—	0	1
Free/reduced lunch	0.260	—	0	1
School adjustment and bonding	3.673	0.746	1	5
Lives with both biological parents	0.601	0.489	0	1
Mostly As or Bs	0.739	—	0	1
Delinquency	1.560	2.512	0	12

Average individual indegree centrality, or number of friendship nominations, slightly increases until it reaches its peak during early adolescence at a maximum average of 4.05 nominations in seventh grade. After attaining this maximum level, indegree declines steadily until 12th grade, when it reaches its minimum at 2.23 nominations. Over time, a greater proportion of students become friendship isolates, that is, no one selects them as a friend. In 7th grade, only 8.1 percent of respondents are isolates, but by 12th grade this proportion almost triples (2.75 times) to 22.1 percent of respondents.

Over the course of our study, students’ lowest academic performance tended to occur when they were in ninth grade. In sixth grade, at the start of our investigation, 78.8 percent of respondents reported that they receive either As or Bs in their classes. This percentage steadily declines until it reaches a minimum of 67.2 percent in ninth grade. After ninth grade, the proportion of students who receive mostly As or Bs increases until it peaks at 79.3 percent during senior year.

School Transitions and Centrality

Network graphs

We present visualizations of friendship network graphs for two schools in our sample and plot them at two points in time, one for eighth grade and the other for ninth, as shown in Figure 1. The first network is from a district in which students transition from middle school to a new high school after they complete their eighth-grade year. The second is located in a control district, where students do not structurally transition following eighth grade but instead remain within the same location. The friendship networks for both schools display more ties in eighth grade than in ninth grade, reflecting a tendency for an overall decline in friendship ties over this period. However, the winnowing of friendships is particularly noticeable for the students who underwent a structural transition (see the top row of Figure 1). The graphs for the transition school show a 13 percent decrease in density, or the proportion of potential ties present in the observed network, between the two waves of interest. Specifically, several of the densely interconnected friendship groups in the eighth-grade network become increasingly sparse in ninth grade. For the school located in the control district, network density decreases by only 7.1 percent between the two grades. This decline is only about half the size of the decrease experienced by students at the school in the compulsory transition district.

OPEN IN VIEWER

Figure 1.

Friendship networks for eighth-/ninth-grade transition and eighth-/ninth-grade nontransition schools at eighth and ninth grades.

Multi-level Models

Next, we use several three-level Poisson models to statistically analyze how the transition from eighth to ninth grade affects indegree centrality for friendships over a lengthy period of time. ⁸ In our first model (Table 2, Model 1), we see evidence of trends over time in friendship indegree as well as effects from the high school transition. Initially, there is a slight increase in indegree (wave, linear trend: b = .059, p < .001), but this trend quickly reverses as students progress through high school and tend to receive decreasing numbers of friendship nominations (wave-squared, quadratic trend: b = –.024, p < .001). In addition to these trends over time, we see evidence that the transition from middle to high school significantly relates to the number of times a student is chosen as a friend. In the years prior to the high school transition, we see no significant difference, on average, in the number of friendship nominations received by students enrolled in transition versus control districts (eighth-/ninth-grade transition, pretransition: b = .001, p = .962). This result is noteworthy because it demonstrates that sixth graders in transition districts do not differ significantly on indegree from sixth graders in control districts. However, in the years following the transition (9th through 12th grades), students who structurally transition to a secondary school start to receive significantly fewer friendship nominations than students in the reference group (eighth-/ninth-grade transition, change after transition: b = –.110, p < .001), lending support to Hypothesis 1. In other words, all respondents’ average friendship popularity declines from 9th through 12th grades, but students who transition to a new school after 8th grade incur an additional post-transition penalty. In the years following the transition, these students are expected to receive 10.42 percent fewer nominations than their peers who attend control schools. ⁹

OPEN IN VIEWER

Table 2.

Three-level Poisson Models for Number of Friendship Nominations Received (Indegree).

Variable	Model 1		Model 2		Model 3
Difference between any transition and control districts
Pretransition	.001	(.023)	−.008	(.031)
Change after transition	−.110	(.019)***	−.111	(.017)***
Linear trend			.019	(.031)
Quadratic trend			−.002	(.007)
Difference between single-feeder transition and control districts
Pretransition					−.027	(.031)
Change after transition					−.104	(.018)***
Linear trend					.020	(.031)
Quadratic trend					−.003	(.007)
Difference between multiple-feeder transition and control districts
Pretransition					.102	(.058)
Change after transition					−.134	(.041)**
Linear trend					−.012	(.029) †
Quadratic trend					−.000	(.008)
Wave (linear trend)	.059	(.009)***	.050	(.028)	.050	(.028)
Wave-squared (quadratic trend)	−.024	(.002)***	−.023	(.006)**	−.023	(.006)**
Mostly As/Bs	.094	(.010)***	.094	(.010)***	.094	(.010)***
Lives with both biological parents	.124	(.013)***	.124	(.013)***	.124	(.013)***
Free/reduced lunch	−.124	(.011)***	−.124	(.012)***	−.126	(.011)***
Delinquency	.005	(.002)**	.005	(.002)**	.005	(.002)**
School bonding	.016	(.006)*	.016	(.006)*	.016	(.006)*
Female	.236	(.015)***	.236	(.015)***	.236	(.015)***
White	.174	(.015)***	.174	(.015)***	.174	(.015)***
Cohort 1	−.002	(.043)	−.002	(.043)	−.001	(.043)
Iowa	−.018	(.032)	−.018	(.032)	−.037	(.034)
School grade size a	−.755	(.171)***	−.755	(.171)***	−.904	(.163)***
Average money spent/student a	.002	(.008)	.002	(.008)	.002	(.008)
Intercept	.825	(.044)***	.833	(.044)***	.836	(.045)***
Random-effect variance components
Individual-level intercept	.279***		.279***		.279***
District-level intercept	.006***		.006***		.006***
Cohort slope	.007***		.007***		.007***

Note: Control group is schools that experience no transition between grades 8 and 9. Robust standard errors are reported in parentheses.

a

Coefficients and standard errors have been multiplied by 1,000.

†

Significantly different from single-feeder districts at p < .05 (Wald test).

*

p < .05. **p < .01. ***p < .001.

Isolates

We next examine whether the high school transition influences the particularly troublesome situation in which an adolescent develops into a friendship isolate. Using Bernoulli MLMs, we test whether a school transition between eighth and ninth grades is associated with an increase in a student’s odds of receiving no friendship nominations (i.e., being a friendship isolate). Complementing our previous findings on overall indegree, experiencing an eighth-/ninth-grade transition is associated with a 50.9 percent increase in the odds of becoming a friendship isolate (b = .412, p < .001), as hypothesized. Transitions not only diminish the number of friendship nominations received by the most socially integrated students, but they further isolate students at the bottom of the social hierarchy (see Figure 2; MLM analyses available upon request).

OPEN IN VIEWER

Figure 2.

Change in individual odds of being an isolate over time and by transition patterns.

School Transitions and Self-reported Grades: MLMs

We now examine whether the structural transition to high school relates to students’ chances of reporting mostly As or Bs, using three-level Bernoulli models. ¹¹ As indicated by the first model in Table 3, the relationship between time and high grades is curvilinear. Initially, students experience a decline in their odds of reporting mostly As or Bs (wave, linear trend: b = –.18, p < .01); later, this trend reverses and they see a modest increase in their odds of high grades (wave-squared, quadratic trend: b = .047, p < .001). Structural transitions also significantly shape academic performance in our sample. During the waves that precede the transition, we find no difference, on average, in self-reported grades between students enrolled in school districts that transition between eighth and ninth grades and those who are not (eighth-/ninth-grade transition, pretransition: b = .163, p = .277). However, similar to indegree, students who change schools experience a significant post-transition penalty in their grades (b = –.421, p < .01), a finding that supports Hypothesis 2. In the years following the transition, students who switched schools have 34.4 percent lower odds of reporting mostly As and Bs than their peers who did not experience a school transition.

OPEN IN VIEWER

Table 3.

Three-level Bernoulli Models for Attaining Mostly As and Bs.

Variable	Model 1		Model 2		Model 3
Difference between any transition and control districts
Pretransition	0.163	(0.151)	−0.028	(0.148)
Change after transition	−0.422	(0.123)**	−0.415	(0.139)**
Linear trend			0.255	(0.077)**
Quadratic trend			−0.047	(0.139)***
Difference between single-feeder transition and control districts
Pretransition					−0.080	(0.191)
Change after transition					−0.249	(0.067)***
Linear trend					0.187	(0.074)*
Quadratic trend					−0.034	(0.011)**
Difference between multiple-feeder transition and control districts
Pretransition					−0.075	(0.284)
Change after transition					−0.945	(0.121)*** †
Linear trend					0.468	(0.090)*** †
Quadratic trend					−0.085	(0.013)*** †
Wave (linear trend)	−0.184	(0.058)**	−0.402	(0.032)***	−0.403	(0.063)***
Wave-squared (quadratic trend)	0.047	(0.009)***	0.087	(0.006)***	0.087	(0.010)***
Indegree	0.089	(0.006)***	0.089	(0.006)***	0.088	(0.010)***
Lives with both biological parents	0.715	(0.032)***	0.715	(0.032)***	0.718	(0.036)
Free/reduced lunch	−0.565	(0.051)***	−0.569	(0.051)***	−0.585	(0.038)***
Delinquency	−0.058	(0.005)***	−0.057	(0.005)***	−0.059	(0.006)***
School bonding	1.175	(0.027)***	1.177	(0.027)***	1.180	(0.024)***
Female	0.410	(0.042)***	0.410	(0.042)***	0.415	(0.040)***
White	0.362	(0.063)***	0.363	(0.063)***	0.364	(0.051)***
Cohort 1	−0.111	(0.050)*	−0.112	(0.050)*	−0.113	(0.052)*
Iowa	−0.076	(0.151)	−0.075	(0.152)	−0.014	(0.149)
School grade size a	0.502	(0.543)	0.512	(0.547)	0.956	(0.837)
Average money spent/student a	−0.008	(0.036)	−0.008	(0.036)	−0.007	(0.035)
Intercept	0.768	(0.124)***	0.933	(0.098)***	0.937	(0.178)***
Random-effect variance components
Individual-level intercept	2.161***		2.161***		2.186***
District-level intercept	0.064***		0.064***		0.065***
Cohort slope	0.012		0.012		0.013

Note: Control group is schools that experience no transition between grades 8 and 9. Robust standard errors are reported in parentheses.

a

Coefficients and standard errors have been multiplied by 1,000.

†

Significantly different from single-feeder districts at p < .05 (Wald test).

*

p < .05. **p < .01. ***p < .001.

Single- versus multiple-feeder school districts

Next, we consider the difference between districts in which students transition to high school from single- versus multiple-feeder schools. In support of Hypothesis 4, students’ self-reported grades tend to be more negatively affected by multiple-feeder transitions to ninth grade compared to single-feeder transitions (see Table 3, Model 3). Although we find no statistically significant difference between the single-feeder, multiple-feeder, and reference group districts during the study’s first wave (single feeder, pretransition: b = –.080, p = .679; multiple feeder, pretransition: b = –.075, p = .796), students’ odds of obtaining mostly As or Bs follow significantly different trajectories over time, depending on the district in which they are enrolled. For example, students in multiple-feeder districts suffer the greatest post-transition penalty with regard to odds of achieving high grades. Compared with youth who attend control schools, students in multiple-feeder transition districts experience a 61.1 percent greater decrease in their odds of obtaining mostly As or Bs upon transitioning from eighth to ninth grade. Students who attend school in single-feeder districts also see a decrease in their odds, but it is significantly less than the decrease that occurs in multiple-feeder districts, as confirmed by a Wald test comparing the two coefficients (p < .001). After transitioning from eighth to ninth grade, respondents in single-feeder districts see an average decrease of 22.0 percent in their odds of reporting mostly As or Bs. Students in multiple- and single-feeder districts continue to be at a significant academic disadvantage compared with their peers in control districts through their senior year of high school, as confirmed by additional Wald tests (p < .001). When compared with students in control districts, students who attend schools in multiple-feeder districts have 71.9 percent lower odds of reporting mostly As and Bs in their senior year, and the odds for students at schools in single-feeder districts are 34.7 percent lower (see Figure 3). Expected academic performance is also significantly lower for 12th graders in multiple-feeder districts, compared with their peers attending single-feeder districts.

OPEN IN VIEWER

Figure 3.

Change in individual odds of receiving mostly As or Bs over time and by transition patterns.

During the pretransition years, the initial decline in the odds of obtaining high grades is not as steep for students in single-feeder transition districts compared with students who will not have a location transition (single feeder, linear trend: b = .187, p < .05). For multiple-feeder districts, we find no initial decline in academic performance in the pretransition period. Instead, students in these districts actually see a rise in their odds of obtaining mostly As and Bs during the study’s earlier years (multiple feeder, linear trend: b = .468, p < .001). In the latter half of the investigation, adolescents in all types of schools see an increase in their likelihood of getting high grades. However, this uptick is significantly smaller for students in single-feeder districts (single feeder, quadratic trend: b = –.034, p < .01) and almost nonexistent for students in multiple-feeder districts (multiple feeder, quadratic trend: b = –.085, p < .001) (see Figure 3). These findings highlight the additional costs to academic performance for students in school districts where a high school transition takes pupils from multiple middle schools instead of a single-feeder school, and they provide partial support for Hypothesis 4.

Finally, regardless of the type of school district, we find that an adolescent’s social network indegree is positively and significantly associated with the likelihood of reporting high grades (b = .088, p < .001; Table 3, Model 3). Greater social integration in a school, as captured by indegree centrality, appears to occur alongside improved grades.

Transitions and Social Integration

Our results demonstrate the key role that structural school transitions play in shaping young people’s friendship networks. All students tend to see a decrease in social integration during the later years of adolescence, but students who experienced locational transitions between eighth and ninth grades received fewer friendship nominations during this decline than did students who experienced no structural change. Our analysis allowed us to assess whether a structural transition represents an initial shock to the system whose effects dissipate quickly. We found, however, that the negative association between school changes and friendship was not limited to the first year in a new school. The effect lingers and does not attenuate over subsequent high school years. Teens who transition to high school continue to be nominated less frequently as friends during their junior and senior years of high school, compared with students who do not make a structural transition.

For some students, the impact of changing to a new secondary school can be severe, resulting in social isolation from their school peers. Losing one or two friends might not be too worrisome for an adolescent’s well-being, but the rise in social isolates following the compulsory change to a new high school constitutes a particularly troubling finding. As noted earlier, social integration is important for a number of reasons, including better grades and overall positive health behaviors (Umberson et al. 2010; Vaquera and Kao 2008). Isolation, in contrast, portends a host of negative outcomes, such as lower self-worth, increased risk of suicidal ideation, and reduced income as an adult (Bagwell et al. 1998; Bearman and Moody 2004; Shi and Moody 2017).

Several processes likely contribute to adolescents’ declining social outcomes. One such mechanism could be a lack of opportunity and availability for social interaction. Sharing similar courses is one key factor guiding adolescent friendship formation (Frank et al. 2013), but upon entering a new high school, a student’s classes and activities will be populated with many unfamiliar faces. Students will likely have much less daily classroom contact with former friends from middle or elementary school. This decline in regular contact may lead students to nominate fewer former friends, and at the same time, new peers may not know each other well enough to consider one another friends. Note, too, that young people move from being the oldest, and often the leaders, in their crowds, clubs, and sports to being the youngest school members and thus possibly the lowest in the high school social hierarchy (Blyth et al. 1983). A decline in social status may also contribute to losses in network centrality.

Transitions and Grades

Patterns regarding students’ grades display complex forms over time, but they appear to be shaped, in part, by compulsory school transitions. From sixth through eighth grades, among respondents who did not change schools between eighth and ninth grades, their odds of receiving high grades steadily declined over each subsequent observation point. After ninth grade, this trend reversed, and by senior year these odds rebounded to rates higher than they were at the study’s inception. Students who structurally transitioned to ninth grade, however, incurred a post-transition grade penalty, and the later increase in their odds of obtaining high grades was not as pronounced. This highlights the long-term negative impact of moving to a new high school.

All structural transitions entail costs, but we find that the type of school transition matters for students’ academic achievement but not significantly for friendship indegree. ¹⁴ Districts with multiple-feeder transitions, as opposed to single-feeder districts, yield particularly adverse outcomes for students’ grades. Not only do students in such schools suffer from a large post-transition penalty in their odds of obtaining high grades, but experiencing a multiple-feeder transition dramatically changes the relationship between transitions and grades. Prior to a structural shift, students attending school in multiple-feeder districts tend to have the greatest odds of obtaining high grades; after a transition, however, they have the lowest. These complex effects on students’ grades would not be apparent in a study that failed to compare students who move to a new secondary school with those who remain within the same school.

We believe a variety of factors make structural school transitions particularly taxing for students’ academic performance. New secondary schools come with more intimidating environments, larger class sizes, difficult academics, and heavy workloads (Roderick and Camburn 1999), all of which may interfere with the time, energy, and motivation necessary to succeed academically. Such an environment could also make it difficult to develop an extensive friendship network, thus curtailing the social capital that contributes to academic outcomes. For instance, youth with relatively few friends tend to have lower grades (Vaquera and Kao 2008). Previous research on school transitions also suggests that adolescents are likely to face significant declines in self-esteem (Barber and Olsen 2004), which in turn could influence their ability to make friends and to obtain high grades.

What about the opportunities provided by a school transition to start afresh, make new friends, and perhaps rejuvenate slipping grades? For some young people, a high school transition indeed proves beneficial. A minority of our respondents experienced small increases in their indegree centrality following a transition, and a very few improved their grades. In line with prior research, students who profited from the transition tended to find eighth grade particularly challenging, socially or academically (Kinney 1993; Schiller 1999; Weiss and Bearman 2007). This trend is consistent with the argument that life transitions are smoother when individuals faced stressful situations prior to the change (Wheaton 1990). Additionally, the transition to high school may be an opportunity for some youth to leave behind disruptive or otherwise problematic friendship groups and make new friends, which, in turn, may promote greater academic success (Langenkamp 2009). For students who struggle in their middle school environment, a change in schools may prove beneficial, although additional research is needed to distinguish substantive patterns from possible statistical shifts in extreme values. Nevertheless, for the bulk of our sample, and as implied by extant literature (e.g., Barber and Olsen 2004; Benner 2011; Temkin et al. 2015), the exogenous shock of relocating one’s place of education between eighth and ninth grades proves costly, with consequences that do not readily dissipate.

Our general findings reinforce several tenets of a life course perspective on adolescents’ social development. According to this theory, life transitions can alter individuals’ developmental paths in substantial ways, and here we see considerable evidence of such effects for an important life transition, that of a structural change in education levels. Our focus on the linked lives of young people’s friendship networks demonstrates how certain friendship ties, presumably the most peripheral, fade with time, although the inner core likely remains more stable (Wrzus et al. 2013). Our results signal that this period in the life course tends to be associated with a winnowing of friendships, but structural high school transitions appear to jumpstart, or exacerbate, the process. Moreover, our finding that high school transitions are associated with a lower likelihood of high grades demonstrates the power of life transitions to disrupt individuals’ developmental processes. Finally, our results have implications for social network research, which should be attuned to possible effects of structural aspects of the educational system, such as school transitions, on the over-time patterns of adolescent friendship networks.

Future Research

Our investigation has a number of strengths, but there remain certain limitations. Our large sample of schools with and without structural transitions to high school, the extended longitudinal follow-up, and our data for full friendship networks provide a study design that is strong for this area of research. Even so, larger urban communities and minority populations are not well represented. We maintain, though, that research in rural and small-city settings is important in its own right. A substantial share of the U.S. population lives in such areas, but these locations receive relatively little attention in educational research (White and Corbett 2014). In addition, previous scholarship claims that urban students are particularly vulnerable during high school transitions. Students in Chicago, for example, experience a variety of associated negative consequences, with ninth-grade course failure rates as high as 40 percent (Roderick and Camburn 1999). Our findings’ generalizability to other settings, however, can be established only through further research.

Future research could also augment our work by using alternative measures of achievement, such as GPAs taken directly from student transcripts, when examining the effects of school transitions on academic trends over time. Information regarding friends in other grades and other schools could also supplement our study, which was limited to within-grade/school social ties. Finally, greater details are needed regarding the underlying mechanisms and processes that combine—at the individual, peer, teacher, school, and district levels—to produce the outcomes documented here.