From Classroom to Streets: The Effect of Schools on Political Participation Among Chilean Youth

Abstract

During the past 20 years, students in Chile have taken center stage in several political demonstrations. In this study, we explore the effects of school-related variables on youth political participation using administrative data on school attendance during protests in October 2019 and survey data from the 2016 International Civic and Citizenship Education Study (ICCS). We use the loss of school attendance as a proxy for youth participation in political mobilization. Our results show that openness in classroom discussions at school has a positive association on conventional and non-conventional (legal) political participation. On the other hand, a respectful school environment and parental participation in school activities are associated with a lower attendance loss. Overall, our findings suggest that schools can help prevent illegal forms of political participation, provided that there is a good school environment. Our results highlight the importance of considering the formal and hidden curricula when studying the effect of schools on political participation.

1. Introduction

What impact do schools, as agents of political socialization, have on young people’s participation in democracies? In many societies, public perception assumes that formal education shapes positive civic attitudes, fostering conventional and unconventional forms of involvement in the political system. However, the empirical literature, which focuses mainly on the US and Western European contexts, has been inconclusive due to methodological challenges such as the endogeneity of sociodemographic and psychosocial factors that influence education and political participation, as well as the particularities of the contexts under study, including high levels of institutional trust and an emphasis on formal civic education (Ehman, 1980; Sapiro, 2004). In a different context, such as Latin America, where youth mobilizations have challenged the status quo and trust in institutions is low, do schools still have an effect on youth political participation in the same way as described in the literature? Which features of the school environment are associated with which forms of political participation, and do schools stand in the same relation to forms of political participation channeled through institutions as they do to non-institutionalized forms of participation?

The reorientation toward the Global South is already underway in youth studies. An emerging body of work has argued that knowledge production about young people remains concentrated in the Global North, and has called for conceptual and empirical space beyond universalized Northern frameworks (Cuervo & Miranda, 2019). Cooper et al. (2019) propose that youth studies in the Global South should be organized around three commitments, namely, disentangling, decentering, and democratizing the field. Cooper et al. (2021) consolidate this research agenda across a wide range of Southern contexts. We situate our study within this body of literature. Our contribution consists of raising a question that has been examined almost exclusively using data collected in the Global North: how do schools relate to the political participation of young people?

In this analysis, we examine how different school-related variables relate to political participation among Chilean youth, using administrative data on school attendance during the massive protests throughout Chile in October 2019 and survey data from the 2016 International Civic and Citizenship Education Study (ICCS). The results reveal that schools with higher levels of classroom politicization show greater non-attendance, which we interpret as an indication of higher political mobilization. Parental participation in school activities and a respectful school climate are associated with the opposite pattern. The same results hold for expected participation, where a respectful school climate is associated with lower expected illegal political participation. Taken together, our research associates a respectful school environment with lower levels of observed and expected illegal political participation. These results show the importance of considering the formal and hidden curricula when studying the effect of schools on youth political participation.

Our study design addresses several methodological problems that have limited previous studies. Common methodological approaches—such as regressions that correlate formal education with political participation without controlling for endogeneity—run the risk of bias, as shared ideological affinities can explain educational exposure and participatory outcomes (Brady et al., 1995). Our focus is on the specific features of the school environment that students encounter and how these relate to political participation. This shifts the question from how much schooling students receive to what kind of school they attend. The study design remains observational, and we describe our estimates as associations. In the Chilean context, where political socialization is influenced by historical mobilizations (Donoso, 2017; Somma & Mackenna, 2020), our findings provide new evidence about the effects of local dynamics while engaging in broader debates on how schools shape participation in democracies. This extends previous arguments, such as those of Denver and Hands (1990), on the hidden curriculum and its role in fostering political efficacy (Koskimaa & Rapeli, 2015).

This article reports two studies that address the research question from different directions. Study 1 uses the loss of school attendance during the 2019 protests as a proxy for observed participation in these demonstrations. Data constraints place the study at the school level, with all individual-level variables aggregated by school. Study 2 uses data from the 2016 International Civic and Citizenship Education Study and is carried out at the student level. Although we acknowledge that each study has limitations—the first because it relies on a proxy variable and the second because it relies on self-reports—we believe that a strength of our research lies in the fact that the two studies complement each other, as they examined the same school cohort at two different moments, capturing expected political participation in 2016 (Study 2) and observed participation during the 2019 protests (Study 1).

This article is structured as follows: Section 2 offers a brief literature review on research related to political participation and the effects of socializing agents such as schools. Section 3 provides relevant information on the Chilean context, while Section 4 presents the data of Studies 1 and 2 and the methods. Section 5 describes the main results, and finally, Section 6 provides an integrative discussion of both studies.

2. Political Participation/Socialization and the Effect of Schools: An Unresolved Debate

Political behavior encompasses the various ways in which citizens interact with the political system they inhabit (Almond & Verba, 1963). Political participation, as a central manifestation of this behavior, has traditionally been categorized into three main forms. Conventional participation includes electoral and formal activities such as voting, party affiliation, and attending political forums. Legal unconventional participation encompasses demonstrations, peaceful marches, petition signatures, and boycotts. Illegal unconventional participation comprises actions such as sit-ins, traffic blockades, and painting slogans on walls (Stockemer, 2014). Much of the empirical literature has used this classification to examine how people, particularly young people, channel their political concerns in different sociopolitical contexts.

As Kaim (2021) argues, the conventional/unconventional distinction conceals at least six separate dualisms (legal/illegal, institutional/non-institutional, public/private, and collective/individual, among others). Legality is one of the six dualisms. In this work, the three items that make up the illegal participation measure, i.e., spray painting slogans on walls, blocking traffic, and occupying public buildings, are illegal under Chilean law but do not involve physical harm to people. We therefore reserve the term ’illegal’ for its strict legal sense and not as a terminological preference. In the Chilean context, the occupation of a school has been a routine and largely non-violent tactic used by students since at least 2006, and treating it as equivalent to political violence would misdescribe the phenomenon our research sets out to understand.

A second qualification concerns the standpoint from which legality is defined. To classify an action as illegal is to take the state’s legal order as the reference, which is unproblematic where that order is broadly uncontested. Celikates (2016) notes that the conditions typically placed on legitimate contestation (be non-violent and accept legal sanction) are themselves stipulated by the order under contestation. de Sousa Santos (2009) argues more broadly that categories produced by social sciences research in the Global North render the Southern experience unintelligible so that a classification such as ’illegal participation’ may register the boundary of an official legal order while telling us little about the logic of the action it names. This point is particularly relevant in Chile, where the constitutional framework was drafted under the military regime and remained politically contested throughout the period examined in our research, including during the constitutional process that followed the 2019 protests. We therefore treat expected illegal participation as a measure against a legal boundary that a substantial part of the population does not regard as settled and avoid interpreting it as a deficit of civic disposition.

Patterns of political participation vary, first of all, by sociodemographic factors. Brady et al. (1995) developed a resource model that explains how time, money, and civic skills differentially impact different types of participation. Educational attainment has proven to be one of the most powerful predictors, with studies confirming that higher formal education is associated with a greater likelihood of participating in conventional and unconventional forms of political participation (Alvarez et al., 2017). Socioeconomic status also exerts a considerable influence, as reported by Castillo et al. (2014) that social origin maintains a strong influence on attitudes toward political participation among Chilean students. Age configures distinctive participation patterns, with young people exhibiting different participatory preferences than adults.

In addition to sociodemographic factors, political participation is also conditioned by various psychosocial factors. Political knowledge provides the basic cognitive tools to understand the political system and generates a higher likelihood of involvement (Denver & Hands, 1990). Political efficacy, both internal (perception about one’s own ability to influence) and external (confidence in the responsiveness of the political system), is also a determinant, especially for conventional forms of participation (Koskimaa & Rapeli, 2015). Recruitment networks, for their part, operate as channels that connect citizens to concrete opportunities for participation. These networks are closely related to political socialization processes, as socializing agents (family, school, and peers) function as potential recruiters that encourage participation within their spheres of influence (Andolina et al., 2003).

Political socialization constitutes the set of developmental processes through which people acquire political orientations and behavioral patterns (Sapiro, 2004). This concept was formally introduced by Hayman (1959), who identified a gap in the study of the formation of political behavior. Since then, this field has gained relevance within political science (Almond & Verba, 1963; Niemi & Hepburn, 1995), establishing itself as fundamental to understanding how political attitudes and identities are configured. Political socialization influences the formation of political ideologies, values, and attitudes that determine the political behavior of citizens throughout their lives (Dukuen, 2018; Gidengil et al., 2016). This process is influenced by multiple factors that can be classified into three main categories: individual characteristics, political–institutional context, and socializing agents (Hatemi & Ojeda, 2021; Jaime-Castillo, 2000).

The process of political socialization is particularly affected by the historical–cultural environment in which individuals develop, especially during the ‘impressionable years’ (14–25 years), a period during which relatively stable political attitudes are formed (Dinas, 2013). Socializing agents, defined as the instruments or means through which political values are transmitted to the individual (Sapiro, 2004), constitute the direct channels of influence, with family and school being the main ones in this process. The timing of this process matters as much as its content. Sears and Valentino (1997) propose that preadult socialization is catalyzed by exogenous political events, that such events socialize attitudes selectively and only in the domains they make salient, and that stable predispositions are therefore acquired episodically. Their evidence comes from a three-wave panel of adolescents aged ten to seventeen, an age range close to the cohort studied here.

Family represents the first and, possibly, the most influential political socializing agent. As Dowse and Hughes (1971) point out, the political opinions and attitudes of parents tend to be transmitted to their children, establishing the fundamental bases for their children’s future political orientation. This socializing agent exerts its influence from an early age when children begin to develop basic understandings of authority, norms, and values, which they will later transfer to the political sphere (Jaime-Castillo, 2000). Research shows that parents influence not only their children’s political ideology but also their levels of political interest and participation (Koskimaa & Rapeli, 2015). The family transmission model is strengthened by trust and emotional closeness, producing effects that tend to endure over time, although such effects can be modified by other socializing experiences (Sandoval & Hatibovic, 2010).

Besides family, school constitutes a critical socializing agent that complements or, at times, counterbalances family influence. During the ’impressionable years,’ the school environment occupies a considerable part of young people’s time (OECD, 2021), becoming a fundamental space for civic formation. The preadult socialization model argues that education shapes young people’s democratic character by instilling habits, feelings, and skills necessary for critical thinking and understanding of the political system (Gutmann, 1999; Willeck & Mendelberg, 2022).

The specific mechanisms through which school exerts its socializing influence remain under debate. On the one hand, the formal curriculum, particularly civic education, has significant effects according to previous studies such as those by Denver and Hands (1990), which link exposure to courses on political topics with higher levels of political interest and participation. Cantoni et al. (2017) confirm this perspective, demonstrating that curricular changes can generate lasting transformations in political attitudes. However, the study by Ehman (1980) emphasizes the importance of the ’hidden curriculum’ and the classroom environment, noting that openness to respectful discussions about current issues increases interpersonal trust, political interest, and critical awareness. Large-scale comparative research has generally treated the two mechanisms as complementary. Analyzing data from the IEA Civic Education Study across twenty-eight countries, Torney-Purta (2002) reported that an open classroom climate for discussion predicted civic knowledge in roughly two thirds of the participating countries, and concluded that schools obtain their best results when rigorous instruction in civic content, an open climate for discussion, and a participative school culture operate together. The relative weight of each of these components may nonetheless vary with the socio-political context and the pedagogical practices implemented, and it remains an open question whether they hold together in settings where the formal curriculum devotes little attention to civic and electoral matters.

Despite advances in the study of political socialization, important research gaps persist, especially regarding contexts outside the United States and Western Europe (Sapiro, 2004). The Chilean case is particularly relevant due to its recent history of student mobilizations and democratic transformations. The studies by Saavedra (2013) on the role of civil society during the dictatorship and by Sandoval and Hatibovic (2010) on university political socialization offer valuable insights, but they lack the necessary depth to fully understand the role of schools as a socializing agent in Chile. The research by Castillo et al. (2014) shows how social inequalities influence the expectations of Chilean students regarding political participation, but it does not explore in detail the specific mechanisms of this influence. A related line of work has examined the school-level mechanisms behind differences in political participation. Castillo et al. (2015) demonstrate the roles of civic knowledge and classroom climate in mitigating the socioeconomic gap in expected political participation. Disi Pavlic and Mardones (2021) model how the curricular implementation of citizenship education relates to participatory dispositions, showing that a cross-cutting curricular implementation is associated with more favorable attitudes toward electoral participation, while other implementation approaches show negligible or negative associations with other forms of participation.

The contribution of our research differs from this body of work in two ways. All of the aforementioned studies measured participation as reported by students using a survey, while Study 1 draws on administrative data of school attendance during an actual political mobilization. Moreover, none of the previous studies addressed the period following October 2019, which had reshaped the participatory context in which students act.

3. The Chilean Context

As illustrated above, in the puzzle of political participation, some pieces relate to political socialization and others to the political and social context under study. To understand how these pieces are interrelated in the Chilean context, it is necessary to briefly review the features of the Chilean educational system, the political values of Chilean youth, and the latest youth social movements. The Chilean educational system is characterized by a standardized curriculum that does not adequately address civic education issues. As in other parts of the world, young people in Chile have lost interest in traditional forms of political participation and have turned to non-institutional forms. In the Chilean context of student mobilizations that began more than two decades ago, the non-institutional political participation of young people has taken the forms of joining demonstrations and occupying educational institutions. Here, schools have played a crucial role in political formation, serving as spaces for organization and the construction of collective identities.

3.1. School and Family

Chile has a free and mandatory education system covering elementary to high school. According to their administrative dependency, we can distinguish three types of schools: (i) private schools, which receive no public funding; (ii) private subsidized schools, which are privately administered but receive public funding; and (iii) municipal-dependent schools, which receive public funding and are administered by municipalities. These three types represent 7%, 52%, and 37% of high schools, respectively (Ministerio de Desarrollo Social y Familia, 2017).1 The Ministry of Education (MINEDUC) establishes minimum common learning objectives for all schools in the country, offering syllabi for different subjects, including social sciences whose syllabi cover topics of civic and political education (Ministerio de Educación, 2016).

However, electoral participation is one of the least discussed contents in the Chilean school curriculum, according to the United Nations Development Program (PNUD, 2017). This would imply that educational institutions do not adequately educate their students on crucial topics for good citizenship, positioning Chilean students below the international average on civic knowledge (IEA, 2010). According to the National Youth Survey (INJUV, 2018), the quality of civic education is the worst aspect evaluated by Chilean students.

On the other hand, in recent decades, the Chilean family structure has undergone significant transformations, with an increase in single-parent households and increased female labor participation (Calvo et al., 2011). The decline of parental presence in the formative process of their children due to work demands and modern family dynamics may have reconfigured the traditional patterns of political socialization in the family environment.

3.2. Values and Political Attitudes

Since 1990, Chile has maintained low electoral participation rates, with a drop of 36% compared with 1990, ranking among the countries with the greatest decreases in electoral participation (PNUD, 2017). Although there are institutional factors that drive this decline2, changes in the political attitudes of youth and the absence of a transversal civic education policy have also been pointed out as causes of this decline (PNUD, 2017).

According to the National Youth Survey, only 16% of Chilean youth are very interested in political issues, with men and people of high socioeconomic levels reporting the highest interest (INJUV, 2018). In addition, 38% of youth state that they are unhappy with the democracy in Chile, only 6% show interest in participating in a political party, and 65% declare that they do not have a political affiliation (INJUV, 2018). People between 18 and 24 years of age constitute the age group that is least likely to consider voting as an effective tool for political participation (PNUD, 2017), leading them to seek alternative ways to participate in politics.

All of the above reveals the skepticism that this age group presents toward more traditional methods of political participation. However, this skepticism does not indicate political disinterest; rather, it reflects a transformation in their mechanisms of expression, with contemporary youth showing a shift in preference from conventional to unconventional forms of political participation. Sandoval and Hatibovic (2010) show that young Chileans prefer unconventional forms of participation, with universities and schools being the spaces where these preferences are reinforced among peers. According to the National Youth Survey, 24% of young people have participated in some social demonstrations, such as marches and strikes (INJUV, 2018). Likewise, young people are more likely than adults to justify and legitimize alternative forms of political action, such as participation in marches, occupations of private properties and educational establishments, hunger strikes, and graffitiing facades with political slogans and demands (INJUV, 2018).

3.3. The Uprise of Chilean Social Movements

Since the 2000s, Chilean students have participated in demonstrations under the influence of various groups. In 2001, the Coordinating Assembly of High School Students (ACES, for its initials in Spanish) organized several protests with the participation of high school students, a phenomenon called the ‘mochilazo’ by the media3. These protests helped change the administration of student transportation passes, setting a precedent for the use of protests to generate political change (Borri, 2016; Donoso, 2017). The ACES, influenced by the Communist Youth and the Revolutionary Coordination of Autonomous Students (Borri, 2016), aimed to politicize the daily experiences of high school students, pushing for very specific demands related to infrastructure, dress code, and participatory decision-making procedures in schools (Donoso, 2017).

In 2006, protests by high school students grew massively and became known as the ‘penguin revolution’ (for the black-and-white school uniforms). In this case, the demands of students were related to issues of educational inequality and changes in the educational system (Donoso, 2017). Although the government reacted proactively, the official proposal did not meet the expectations of these students. There were some protests in 2007–2008 that were not significant, representing the end of the ‘penguin’ movement, with general dissatisfaction among the protesters regarding the policies and solutions of the government and Congress (Von Bülow & Bidegain Ponte, 2015).

For the first time since Chile’s return to democracy, in 2011, students were confronted by a government that was ideologically opposed to them. Demonstrations increased, alongside the occupations of various schools and universities4. Although one cannot speak of continuity, the leaders of this wave of protests were socialized in previous student movements, particularly that of 2006. Once again, the 2011 movement placed education at the center of political debate, demanding for a lowering of interest rates charged on student loans. The protests continued in a fragmented way after 2011 by different political organizations and collectives, mostly left-wing, until 2013 (Von Bülow & Bidegain Ponte, 2015).

In October 2019, a social movement marked by massive and often violent protests took place, which became known as the ‘social revolt’. Although people of all ages participated in the demonstrations, young people (between ages 18 and 30) participated the most (Somma & Mackenna, 2020). Furthermore, high school students began mass fare evasions on the Santiago subway system at the beginning of the movement. After these initial demonstrations, people were mobilized throughout the country under a variety of slogans, and students, once again, were one of the groups that mobilized the most people in collective protests (Somma & Mackenna, 2020). According to Rivera-Aguilera et al. (2021), these young protesters learned forms of political action from student demonstrations that began in 2006.

The literature reviewed above, when interpreted against the Chilean context, yields five hypotheses. We state these hypotheses as directional associations, since our study design is observational. (H1) Formal curriculum: Exposure to civic learning at school is positively associated with all three forms of expected political participation. This is the hypothesis derived from research linking educational instruction in political content to political interest and involvement. (H2) Deliberative climate: Openness in classroom discussion is positively associated with conventional and unconventional legal forms of political participation. Open discussion of contested public issues under conditions of mutual respect is expected to build student dispositions through which political participation can be institutionally channeled. (H3) Relational climate: A respectful school climate is negatively associated with forms of political participation that depart from institutional channels. In Study 2, this implies lower expected illegal participation, and in Study 1, lower attendance loss during the October 2019 protests. (H4) School politicization: School-level politicization, measured by the frequency of debates on public issues and by the presence of an active class council, is positively associated with observed political participation. (H5) Family presence: Parental participation in school activities is negatively associated with observed political participation, and parental political interest is positively associated with conventional political participation.

H2 and H4 concern the same underlying phenomenon, which is the discussion of politics within the school environment, but they point to different outcomes. The former anticipates a shift toward political participation channeled through institutional means, and the latter anticipates greater participation in street mobilization. The two are compatible if the mobilization in question is itself legal. Distinguishing them is one of the reasons for running the two studies together.

4. Materials and Methods

4.1. Study 1: Participation in October 2019 Protests

Study 1 was conducted at the school level, with schools as the unit of analysis. The analytic sample comprises 2422 schools with 11th-grade classes in the public and private subsidized sectors.

There are no official records of citizen participation in the October 2019 protests, so we cannot directly measure student participation in these political demonstrations. However, the Chilean Ministry of Education maintains records on student attendance, at least for public and private subsidized schools. Thus, we used the difference in school attendance between November 2019 and November 2018 as a proxy for participation in these protests.5 This definition assumes that students who skipped classes were likely to have participated in these protests. Of course, there were other reasons that could lead to lower school attendance even if students did not participate in these protests, such as problems with public transportation or parental fear of sending their children to school in a violent context. However, evidence from Chile’s National Youth Surveys underscores the scale of such disruptions during the 2019 social revolt. The 2022 survey revealed that 9.4% of youth participated in occupations of schools and 20.2% in strikes, with 57.3% engaging in at least one protest-related activity (INJUV, 2022), highlighting widespread involvement but also leaving a substantial proportion of youth vulnerable to secondary effects such as fear and mobility issues. On the other hand, school attendance may be overreported, as private subsidized schools receive funding based on recorded attendance, potentially incentivizing inflated figures. However, since this bias would lead to an underestimation of the true attendance loss, our estimates represent a conservative lower bound for the protest-related effects. These issues will be revisited later in the Section 5.

To measure the relationship between school environment and student participation in protests, we relied on three official data sources: (i) school attendance, provided by the Center for Studies of the Chilean Ministry of Education through its Open Data platform6; (ii) school performance on the national standardized test of education quality (SIMCE, by its Spanish initials), provided by the Education Quality Agency7; (iii) the private database of the SIMCE 2018 surveys of students and parents (Agencia de Calidad de la Educación, 2018). Since these data are not publicly available, they were requested directly from the Education Quality Agency, who evaluated the research project and subsequently granted access. This private database allowed us to construct variables related to a respectful school environment and parental involvement.

The SIMCE is a test administered annually to selected grades in all Chilean schools. As the SIMCE was not administered in 2019 due to protests, we used the 2018 applications from students in the 10th grade (second year of high school). Thus, the loss of school attendance during the 2019 protests was estimated for the 11th grade. In this way, we examined the same cohort: those who were in the 11th grade in 2019 and took the SIMCE in 2018. Since private schools are not legally bound to report attendance, our database is restricted to the 11th-grade classes of all public and private subsidized schools. Of the total number of schools, 22% were public (municipal) and 78% were private subsidized schools; 15% belonged to the low socioeconomic group, 39% to the lower-middle group, 30% to the middle group, 15% to the upper-middle group, and 0.5% to the high socioeconomic group. Regarding education quality, 14% of the schools performed poorly, while 26% showed a lower-middle performance, 49% a middle-level performance, and 12% a high level of academic performance.

In this research, we attempted to explain the difference in average school attendance using several variables related to school climate and political socialization. We developed these explanatory variables based on the study by Ehman (1980) on the effects of the school environment on American students’ political attitudes, and the study by Andolina et al. (2003) on the influence of family practices on youth’s political development. Since we used multiple data sources, and school identifier (RBD, for the Spanish initials of Rol Base de Datos) was the common variable, we worked at the school level. Thus, all individual-level variables were aggregated at the school level.

4.1.1. Dependent Variable: Average Monthly Difference in School Attendance

The difference in average school attendance was estimated as follows: (i) individual monthly attendance was measured as the proportion of days a student attended classes compared with the total working days in that month; (ii) average school monthly attendance was calculated by aggregating individual monthly attendance for all students in the target class of a given school; (iii) the difference in average school attendance was obtained by subtracting, for each school, the average attendance in November 2019 from the average attendance in November 20188. This variable ranged between −1 and 1, with positive numbers representing an increase in non-attendance in November 2019 compared with November 2018, indicating a possible increase in student participation in mobilization.

4.1.2. Independent Variables

Based on the private database of the SIMCE 2018 surveys of students and parents, we constructed the following variables related to school environment and parental involvement: (i) parental participation in school activities, such as recreational, sports, academic, or cultural activities; (ii) parental involvement in student routine, such as establishing a schedule for homework and recreation, or talking to their children about how they feel at school; (iii) parental perception of discrimination in their children’s school because of gender, sexual orientation, and beliefs, among others; (iv) parental perception of school climate, measured as respectful interaction among students, parents, and teachers; (v) student perception of respect in their school; (vi) student perception of discrimination in their school; (vii) student participation in school activities, particularly in debates; and (viii) the existence of class councils in school. The last variable was used to represent the level of student organization within the school. The details of these and other study variables are presented in Table 1 and in the Supplementary Materials.

Table 1. Variables used in Study 1. All the Item Response Theory (IRT) variables were created using the package mirt for the R programming language (Chalmers, 2012).

Type of Variable	Variable	Operationalization
Dependent
Continuous	Diff. in attendance	Difference in the proportion of students attending school between November 2018 and November 2019. This variable serves a proxy of students joining political and social protests.
Independent
Continuous	Single-parent households	Proportion of single-parent households, based on Q 4 of the parental questionnaire
Continuous	Parental involvement in student routine	IRT from Qs 33–34 of the parental questionnaire
Continuous	Parent participation in school	IRT from Q 26 of the parental questionnaire
Continuous	Parental perception of school climate	IRT from Q 15–16 of the parental questionnaire
Continuous	Parental perception of discrimination in school	IRT from Q 17 of the parental questionnaire
Continuous	Student perception of respect in school	IRT from Qs 6 and 9 of the student questionnaire
Continuous	Student perception of discrimination in school	IRT from Q 7 of the student questionnaire
Continuous	Debates in school	Proportion of schools that answered ‘many times’ or ‘always’ to Q 16.3 of the student questionnaire
Continuous	Class council in school	IRT from Qs 17–18.1 of the student questionnaire
Control
Continuous	Socioeconomic level	IRT from Qs 6, 7, and 10 of the parental questionnaire
Categorical	Socioeconomic level	1 (low), 2 (medium-low), 3 (medium), 4 (medium-high), 5 (high)
Categorical	Administrative dependency	1 (municipal school), 0 (private subsidized school)
Continuous	Quality of education	Average score on SIMCE test (math)
Categorical	Quality of education	Performance level on SIMCE test: insufficient, medium-low, medium, high
Categorical	Rurality	1 (urban), 0 (rural)
Categorical	Region	1 to 16, according to the Chilean political administrative organization

4.1.3. Control Variables

The following sociodemographic indicators were included as control variables: the socioeconomic level of each school, the administrative dependency of each school, and the quality of education that each school offered. We also controlled for the urban/rural location of schools and added a fixed effect by region. To account for family presence, we built an indicator of the proportion of single-parent households in each school using data from the SIMCE 2018 survey for parents.

4.2. Study 2: International Civic and Citizenship Education Study

Study 2 was conducted at the student level, with analytic samples of 4635, 4630, and 4627 students across the three outcomes. These figures are derived from the full Chilean ICCS 2016 sample of 5081 students by list-wise deletion of cases with missing values on any variable entering the model. The three samples differ slightly because the three outcome scales have different patterns of non-response.

The International Civic and Citizenship Education Study is an international assessment of students’ perceptions, attitudes, and activities related to civics and citizenship (International Association for the Evaluation of Educational Achievement, 2018). It also includes questions on individual characteristics and family background, as well as a student test of civic knowledge. In addition to direct responses to the questionnaire items, the database includes Item Response Theory (IRT) variables that aim to capture different dimensions of social participation and school climate, among others. The survey is administered to grade 8 (or equivalent) students in more than 5300 schools from 38 countries. In this study, we used the 2016 wave carried out in Chile, since it was the closest in time to Study 1.

Students in the 8th grade in 2016 would have reached the 10th grade in 2018, when the SIMCE questionnaires were administered, and the 11th grade in 2019, when school attendance was recorded. The two studies therefore examined the same school cohort, although they did not observe the same individuals, since the ICCS and SIMCE recruited independent samples. The time interval was also substantial, since Study 2 examined student dispositions three years before the October 2019 protests while Study 1 looked at student behavior during these protests. Study 2 describes the participatory dispositions this cohort held before a mass mobilization, and Study 1 describes how the cohort behaved during it. We make no claim that the former predicts the latter at the individual level, which our data cannot support.

4.2.1. Dependent Variable: Intention to Participate in Politics

The student questionnaire incorporated several IRT questions labeled as ‘Participating in society’, which were intended to measure different aspects of social and political participation. Here, we examined three aspects: (i) students’ expected electoral participation, encompassing questions such as intention to vote in local or national elections or to join a political party as an adult; (ii) students’ expected participation in legal protest activities, referring to activities such as participating in a peaceful march or rally, collecting signatures for a petition, or buying products in support of social justice; and (iii) students’ expected participation in illegal protest activities, such as spray painting slogans on walls and protesting by occupying public buildings or blocking traffic. The specific questions that make up these variables are described in Table 2, and can be found in the ICPSR 37147 user guide and codebook.

Table 2. Variables used in Study 2.

Type of Variable	Variable	Operationalization
Dependent
Continuous	Expected electoral participation (S_ELECPART)	IRT from Q IS3G31A-C
Continuous	Expected participation in illegal protest activities (S_ILLACT)	IRT from Q IS3G30I-K
Continuous	Expected participation in legal protest activities (S_LEGACT)	IRT from Q IS3G30A-G
Independent
Continuous	Abuse at school (S_ABUSE)	IRT from Q IS3G20A-F
Continuous	Importance of conventional citizenship (S_CITCON)	IRT from Q IS3G23A-F
Continuous	Self-efficacy (S_CITEFF)	IRT from Q IS3G29A-G
Continuous	Civic learning in school (S_CIVLRN)	IRT from Q IS3G18A-G
Continuous	Student interaction at school (S_INTACT)	IRT from Q IS3G19G-I
Continuous	Trust in civic institutions (S_INTRUST)	IRT from Q IS3G26A-F
Continuous	Openness in classroom discussions (S_OPDISC)	IRT from Q IS3G17A-F
Continuous	Political discussions outside school (S_POLDISC)	IRT from Q IS3G14A-F
Continuous	Participation in school activities (S_SCACT)	IRT from Q IS3G32A-E
Control
Categorical	Highest parental interest in politics (S_HINT)	Binary variable based on IS3G10B and IS3G10C, originally measured on a 4-level Likert scale
Categorical	Student religious affiliation (S_RELIG)	Binary variable based on IS3G33
Categorical	Student gender (S_GENDER)	Binary variable based on IS3G02
Categorical	Student interest in politics (S_SINT)	Binary variable based on IS3G10A, originally measured on a 4-level Likert scale
Continuous	Socioeconomic background (S_NISB)	IRT from Q IS3G11, IS3G08A, IS3G08B, IS3G06A, IS3G06B, IS3G09, and IS3G07

Note: The variable codes as appeared in the original database are shown in parentheses.

4.2.2. Independent Variables

We used different explanatory variables considering the effects of internal and external political efficacy, political socialization with parents and friends, and civic education in school. We also included a set of questions intended to measure the school climate, such as perception of respect and openness to classroom discussions. Other questions related to student participation in school activities and perception of the importance of conventional citizenship. The specific questions that make up these variables are listed in Table 2, and can be found in the ICPSR 37147 user guide and codebook.

4.2.3. Control Variables

The ICPSR 37147 database includes the following sociodemographic variables: (i) a national index of socioeconomic background, which is an IRT variable constructed based on home literacy resources and the highest parental educational level and socioeconomic status; (ii) a binary gender variable (boy/girl); and (iii) a binary religion variable (religion/no religion). We also controlled for student interest in politics and the highest parental interest in politics, based on the maximum value between father’s and mother’s interest in politics. Student interest in politics and the highest parental interest in politics were originally measured on 4-level Likert scales and were binarized for this study.

4.3. Model

Most of the variables in this study were constructed from a set of items intended to measure a certain attribute. Therefore, to build our models, we relied on the Item Response Theory (IRT), which attempts to explain the relationship between latent traits and their observed responses (Cai et al., 2016). For Study 1, we constructed the IRT variables using the package mirt for the R programming language (Chalmers, 2012), generating continuous and standardized latent variables from groups of Likert-scale questions. In particular, we used a Graded Response Model (Samejima, 2013), which is an extension of the two-parameter logistic model (2PL), allowing us to handle ordered polytomous categories. This model enabled the prediction of the probability of response to an item using two key parameters: item discrimination, which measures the item’s ability to differentiate subjects according to their latency level, and item difficulty, where a higher level of difficulty is associated with a higher latency (Montesinos & French, 2016). For the ICPSR questionnaire (Study 2), we used the IRT variables that come with the dataset. In this case, the latent score estimation also considered the sample weights of the data.

For each study, ordinary least squares (OLS) models were used. For Study 1, we ran two models, one with the control variables only and the other with the variables related to socializing agents. For Study 2, we ran three different models, one for each type of political participation (conventional, legal unconventional, and illegal). Because the students were sampled from a set of schools, the observations might not be independent. Thus, we used clustered standard errors, with the ID of each school as the unit of clustering. The models included IRT variables and binary/categorical variables. The correlations between the variables can be found in Figures S1 and S2 of the Supplementary Materials. Neither study estimates a multilevel model. The clustered standard errors used in Study 2 correct inference for the non-independence of students within schools. They do not estimate variation between schools, and no school-level predictor enters those models. Nevertheless, the interclass correlation coefficient (ICC) is close to 0 for each of the three models in Study 2, which indicates that little of the variance in expected participation lies between schools and that a random intercept specification would add little in this case. Study 1 addressed the school level directly, as school was the unit of analysis, and correspondingly, no implication was made about variation among students within a school.

Two features of the study design deserve attention. In Study 2, the predictors and the outcomes were derived from the same student questionnaire, so the covariance between them could arise from the shared source alone (Podsakoff et al., 2003). Study 1 was not exposed to this risk, since its outcome was derived from administrative attendance records and its predictors from a separate instrument. Responses to sensitive items are also subject to misreporting (Tourangeau & Yan, 2007), and expected electoral participation and expected illegal political participation are unlikely to be affected in the same direction. We therefore interpret the differences in magnitude across the three models with caution.

5. Results

5.1. Study 1

Table 3 shows the results of the OLS model for Study 1. The first model (left panel) includes only the control variables, and we can see that being a rural school or a private subsidized school is negatively associated with attendance loss. On the other hand, the proportion of single-parent households is positively associated with attendance loss, as well as a higher socioeconomic level. The quality of education, measured by the SIMCE score, does not have a significant effect. These results remain under different variable specifications. Table S1 presents alternative models using categorical variables of socioeconomic status and quality of education instead of their continuous formulations. Table S2 displays the same models as those in Table S1, excluding SLEP schools from the ‘Municipal’ category of administrative dependency.9

Table 3. OLS model results for Study 1.

	Diff. in Attendance (Loss of Attendance)
	Control Variables	Full Model
Socioeconomic level (continuous)	0.029 ** (0.010)	0.031 ** (0.011)
Admin. dependency (private subsidized)	−0.088 ** (0.010)	−0.083 ** (0.010)
Average SIMCE score (math)	0.0001 (0.0002)	0.0003 (0.0002)
Rurality (rural)	−0.047 ** (0.015)	−0.034 * (0.015)
Single-parent households	0.176 ** (0.038)	0.160 ** (0.038)
Parental involvement in student routine		0.005 (0.019)
Parental participation in school		−0.056 ** (0.016)
Parental perception of school climate		0.024 (0.017)
Parental perception of discrimination in school		0.062 (0.033)
Student perception of respect in school		−0.070 ** (0.016)
Students perception of discrimination in school		−0.046 (0.029)
Debates in school		0.070 * (0.028)
Class council in school		0.024 (0.016)
Region	Yes	Yes
Constant	0.110 ** (0.039)	0.036 (0.046)
Observations	2422	2422
R2	0.293	0.309
Adjusted R2	0.287	0.301
Residual Std. Error	0.174 (df = 2401)	0.172 (df = 2393)
F Statistic	49.788 ** (df = 20; 2401)	38.158 ** (df = 28; 2393)

Note: * p < 0.05; ** p < 0.01.

The right panel of Table 3 shows the complete OLS model results for Study 1. According to these results, parental participation in school activities is negatively associated with attendance loss. This suggests that a greater presence of parents in school activities is accompanied by lower levels of student mobilization, suggesting a tension between school and family as socializing agents. The school climate is also relevant, as shown in Table 3. Student perception of respect in school, which reflects relationships between students, teachers, and school staff in general, is negatively associated with attendance loss.

Regarding variables with a positive association with attendance loss, only ‘debates in school’ is significant at the 5% level. This variable is the closest approximation to an indicator of classroom politicization within the school environment, as it refers directly to debates on issues of public interest. Thus, this result indicates that schools where a larger share of students report frequent debates on public issues show greater attendance loss. It should be noted that our dataset does not allow us to distinguish whether this association is the result of politicization imposed by the school (or certain teachers) on students, or whether it corresponds to a demand from already politicized students to organize debates. Given that the assignment of students to schools was not random until 2016, both explanations are probably true, and to some extent, students chose schools with certain attributes related to political participation.

As noted above, private subsidized schools have an incentive to overreport attendance, since their funding depends on recorded attendance. Because our dependent variable is a difference between two years, reporting inflation that is stable over time cancels out. The relevant threat is inflation that grew between November 2018 and November 2019, which is plausible, since the incentive to inflate is strongest when real attendance falls. We therefore cannot determine whether the negative association between attendance loss and private subsidized schools, which is −0.083 in the full model, reflects lower participation or a larger increase in reporting inflation.

The coefficients we interpret are affected only if reporting behavior also varies with other variables within a given administrative dependency. To examine this, we estimated the full model on private subsidized schools alone, which holds the reporting incentive constant across the sample. The association for students’ perception of respect holds, as does the one for the proportion of single-parent households. The associations for parental participation and socioeconomic level lose statistical significance (see Table S7 in the Supplementary Materials). We read the first result as evidence against the school climate findings being produced by differences in reporting behavior across administrative dependency. We are more cautious about parental participation, since restricting the sample also compresses the variance in socioeconomic level and reduces the power available to detect these associations.

An additional concern relates to the validity of attendance loss as a proxy for political mobilization. As we have noted in the Introduction, during the 2019 protests, there was damage to public transport, which might have caused students to miss school due to the difficulty of getting there. The Santiago subway system was severely damaged, and the bus system to a lesser extent. However, regional public transport did not suffer major damage from vandalism10. Therefore, as a robustness check, we re-ran the full model while excluding all schools within the Metropolitan Region (Santiago). Although this model has a lower coefficient, parental participation in school activities and student perception of respect in school remain significant predictors (see Table S5 in the Supplementary Materials). As an additional test, we re-ran the full model for a year without protests, using the difference in attendance between 2018 and 2017 as a dependent variable. The results are reported in Table S6. None of the parental or school-related variables are significant in this model.

Finally, when running alternatives models for attendance loss among students in the 9th and 10th grades, the variable—class council in school—is significant at the 10% level. This variable indicates the presence of an organized class council where students can opine and are encouraged to participate. Thus, if there is a functional class council, it is more likely that the class can be organized to participate in various forms of political participation. This is a relatively unexplored aspect of the effect of schools on political participation, which can be analyzed in light of collective action problems.

5.2. Study 2

Table 4 shows the results of the OLS model for Study 2. Regarding sociodemographic variables, being female increases legal political participation in conventional and unconventional forms. Religious affiliation and a higher socioeconomic status have a positive and significant association with conventional political participation and a negative association with illegal political participation.

Table 4. OLS model results for Study 2.

	Dependent Variable
	Conventional	Legal Unconv.	Illegal
	(1)	(2)	(3)
Student interest in politics (high)	1.375 ** (0.377)	1.641 ** (0.402)	−0.651 (0.436)
Student gender (girl)	1.130 ** (0.318)	1.064 ** (0.315)	−0.606 (0.353)
Student religious affiliation (yes)	1.470 ** (0.396)	−0.355 (0.459)	−1.536 ** (0.367)
Highest parental interest in politics (high)	2.089 ** (0.368)	1.420 ** (0.405)	0.607 (0.350)
Socioeconomic background	1.237 ** (0.147)	−0.288 (0.186)	−1.316 ** (0.203)
Student interaction at school	0.015 (0.016)	−0.007 (0.017)	−0.076 ** (0.018)
Openness in classroom discussions	0.034 * (0.015)	0.042 ** (0.016)	−0.009 (0.016)
Civic learning at school	0.032 (0.016)	0.039 (0.020)	−0.030 (0.016)
Self-efficacy	0.115 ** (0.016)	0.376 ** (0.018)	0.213 ** (0.015)
Importance of conventional citizenship	0.190 ** (0.018)	0.133 ** (0.019)	0.022 (0.016)
Trust in civic institutions	0.147 ** (0.017)	0.083 ** (0.020)	−0.021 (0.021)
Political discussions outside school	0.075 ** (0.020)	0.109 ** (0.023)	−0.063 ** (0.022)
Constant	16.504 ** (1.232)	10.396 ** (1.508)	52.847 ** (1.671)
Observations	4635	4630	4627
R2	0.287	0.351	0.078
Adjusted R2	0.285	0.349	0.076
F statistic	155.193 **	207.833 **	32.644 **
	(df = 12; 4622)	(df = 12; 4617)	(df = 12; 4614)

Note: * p < 0.05; ** p < 0.01.

The pattern of results differs sharply across the three models, with the divergence following a consistent structure. Variables that index attachment to the institutional system lose their predictive power for illegal political participation. Student interest in politics is a strong positive predictor of conventional and legal unconventional political participation, but it is not significant in model (3). The same results are observed for trust in civic institutions and the importance attributed to conventional citizenship. Citizenship self-efficacy is the only exception among the attitudinal predictors, remaining positive and significant across all three models.

The low explanatory power of model (3) therefore has a specific source. The predictors assembled for this model were drawn from a body of literature concerned largely with political participation channeled through institutions, and they cease to discriminate precisely when the outcome falls outside the legal boundary of such institutions. We will return to this issue in the Discussion.

The association between parental interest and participation is much stronger for conventional participation than for legal unconventional participation. This could reflect a generational effect, as older people rely more on conventional electoral participation methods11. Regarding political efficacy, internal efficacy is a positive and significant factor for all forms of political participation; however, the coefficient is higher for unconventional legal participation. In contrast, external efficacy (measured as trust in institutions), although positive and significant for both forms of legal participation, has a higher coefficient for conventional participation. These results are consistent with the expected effect of political efficacy on political participation: as conventional political participation is based on elected representatives, people who rely on these methods should show greater trust in political institutions. Unconventional political participation, on the other hand, is often associated with direct democracy, in which citizens believe in their own ability to make decisions.

School-related variables provide relevant information regarding how this socializing agent affects the expected political participation of students. A respectful school environment is associated with lower illegal participation, as demonstrated by the negative coefficient of the variable ‘student interaction at school’. Abusive situations show the opposite association (see Table S3 in the Supplementary Materials)12. Furthermore, the perception of openness in classroom discussions is associated with higher legal participation in conventional and unconventional forms. Civic learning at school shows no significant association with any of the three outcomes.

We are cautious about reading the contrast between these two school variables as a difference in magnitude. The coefficients for openness in classroom discussions and for civic learning are close in size and carry similar standard errors in models (1) and (2), and our data do not establish that they differ from one another. What our results support is narrower, and it holds across both studies. Formal civic instruction shows no detectable association with any form of expected participation, while classroom climate is associated with outcomes in Study 2 and with observed attendance loss in Study 1.

Because openness in classroom discussions reaches conventional significance thresholds while civic learning does not, and because the two coefficients are close in size, we examined how robust each estimate is to resampling. We performed a 1000 repetition bootstrap of the conventional and legal unconventional models, estimating the bootstrapped coefficients with the R package car and the bootstrapped clustered variance covariance matrix with package sandwich (R version 4.5.1). For conventional participation the bootstrapped coefficient is 0.034 (SE 0.015) for openness in classroom discussions and 0.032 (SE 0.017) for civic learning at school. For legal unconventional participation the values are 0.042 (SE 0.016) and 0.038 (SE 0.020). Civic learning falls close to the conventional threshold in both models. Neither test evaluates whether the two coefficients differ from one another, and we make no claim that they do. Our interpretation rests on the consistency of the pattern across the three outcomes of Study 2 and across both studies.

Interestingly, it seems that none of the school-related variables has a higher positive effect on expected conventional (electoral) participation than on unconventional legal political participation. The only exception is parental interest in politics: its effect on electoral participation is much greater than its effect on unconventional legal political participation. It would seem that family, rather than school, has the greatest capacity to influence electoral participation, although a respectful school climate is associated with lower expected illegal political participation.

The final result that deserves attention concerns political conversations outside the school. This variable refers to having conversations about political and social issues with family and friends, as well as obtaining related information from television and newspapers. In particular, its association is positive for legal political participation and negative for illegal political participation. It would seem that talking about politics favors legal forms of political participation over illegal participation, but only if these conversations take place outside of the school.

To test this, we ran an alternative model (see Table S4 in the Supplementary Materials) using the variable ‘participation in school activities’ instead of ‘self-efficacy’13. The variable ‘participation in school activities’ measures the willingness to participate in school elections or in discussions in student assemblies. In this model (Table S4 in the Supplementary Materials), the effect of political conversations outside the school environment still holds, while the effect of participation in school activities is positive for all forms of political participation. This result shows that while political activities within the school environment encourage all forms of political participation, including illegal ones, political conversations outside the school is positively associated only with legal forms of participation.

6. Discussion

In this work, we have delved into the relationship between the school environment and youth political participation. Our approach consists of two studies: the first used a school-level data source for loss attendance as a proxy of political mobilization, and the second used an individual survey of expected political participation. When interpreted against the five hypotheses set out in Section 3, our results support three while leaving two in question. Openness in classroom discussion is associated with both legal forms of political participation, as anticipated in H2. A respectful school climate is associated with lower levels of political participation outside institutional channels in both studies, as anticipated in H3. Parental presence in school follows the pattern set out in H5. School politicization is associated with observed political participation for debates in school but not for class councils, partially supporting H4. Exposure to civic learning shows no detectable association with any of the three outcomes; thus, H1 is not supported.

Study 1 links a respectful school climate with a lower likelihood of joining protests, measured by loss of school attendance during the 2019 protests. These results complement those of Study 2, which was conducted at the individual level and analyzed expected conventional, unconventional but legal, and illegal political participation. According to this study, openness in classroom discussions about political issues has a positive effect on conventional and unconventional legal political participation. Regarding the school climate, a respectful school environment is associated with lower expected illegal political participation, in line with the results of Study 1.

The two studies measured school climate differently, and the correspondence between them should be interpreted with that in mind. In Study 1, the measure is based on the mean of student responses within each school, so it represents a property of the school, which is defensible only to the extent that students in a school agree in their assessments. In Study 2, the measure is based on each individual student’s perception, which varies between students attending the same school. The two are distinct though closely related. We therefore interpret the correspondence of the results between our studies as indicating the convergence across two different operationalizations of school climate. Measuring a single construct twice would offer a weaker basis for the argument than measuring two distinct constructs pointing in the same direction.

Overall, this work distinguishes the contribution of curricular variables, such as civic learning classes, from that of variables related to the hidden curriculum, such as school climate. Formal instruction shows no detectable association with expected political participation in our data, while the relational and deliberative dimensions of school climate are associated with the outcomes in the two studies. Comparative work has generally expected these components to reinforce one another (Torney-Purta, 2002). Our results suggest that they can come apart in a context where the formal curriculum gives limited attention to civic and electoral matters and trust in institutions is low. We also note that the comparative evidence concerns civic knowledge, while our outcome is expected political participation, and formal instruction may build the former without translating into the latter.

Regarding family influence, Study 1 suggests that students attending schools with greater parental presence are less likely to mobilize. Consistently, Study 2 shows that parental influence favors conventional (electoral) forms of political participation. In sum, our results suggest that schools can help prevent illegal forms of political participation, provided that there is a respectful school climate. School could also strengthen parental influence toward conventional forms of political participation. In this sense, school could act as a complementary or counteracting force to family influences in a context of sociopolitical transformations. This highlights the importance of school, which goes beyond being an agent conveying political values to being a facilitator for later public life. The interaction between school and family as socializing agents, with an effect extending beyond the unique effect of each, poses an interesting topic for further exploration.

Our findings are specific to a school system that combines a large publicly funded private sector, low institutional trust among the young, and two decades of secondary student mobilization. We expect them to travel to settings that share these features, and to travel less well where student representation is guaranteed by law or where protest is not part of the ordinary repertoire of adolescents. The question is directly answerable, since the ICCS 2016 instrument was administered in five Latin American countries and the school variables used in Study 2 can therefore be compared across cases that differ in exactly these respects. The behavioral component of our design is harder to export, since it depends on administrative attendance records held at the school level and on a mobilization episode intense enough to register in them.

Finally, this work leaves some questions open for further investigation, which can offer other perspectives on the topic. The positive effect of political debates in schools on student mobilization is particularly intriguing, as it suggests that the school environment may be politicizing young people. However, at the same time, openness in classroom discussions about political issues has a positive association with legal forms of political participation (Study 2). Thus, it is worth asking what kinds of classroom discussions or school debates encourage legal or illegal forms of political participation.

The second open question concerns the mechanisms underlying expected illegal political participation. The literature on collective action has consistently found that normative and non-normative actions display distinct motivational routes (Tausch et al., 2011). Anger at social injustice and group efficacy drive the former, while contempt for the institutional system drives the latter. On that account, a model assembled from variables related to political interest, trust in institutions, and adherence to conventional citizenship should perform well in predicting political participation channeled through institutional means but poorly for political participation outside institutional boundaries, as observed in our research. In this regard, the political efficacy measure available in the ICCS captures individual citizenship competence, but it does not extend to the belief that collective action can produce change, so our results do not speak to the group efficacy component of that body of literature.

A third question concerns the categorization of forms of political participation itself. As noted in Section 2, classifying an action as illegal takes the state’s legal order as the measurement boundary. Writing from within the decolonial tradition, Prada Alcoreza (2014) argues that communal rationality in the Andean world stands against the individualistic legalism of the colonial state so that the collective reclamation of space asserts a community right over what the state administers. Applied to the school environment, this recasts our finding that greater parental presence is associated with lower student mobilization, since it would follow that families integrated into institutional channels for political participation orient students toward those channels. Platzky Miller (2023) documents a comparable logic in the Brazilian secondary school occupations in 2015 and 2016, where students treated the occupied schools as a site of political learning in its own right. We cannot test this interpretation with the present data, since neither the ICCS nor the SIMCE measures perceived state legitimacy, historical memory of earlier mobilizations, or communal identification. We regard this as the most promising direction for constructing instruments designed for the Latin American contexts.