Abstract

Using data from an original two-wave panel survey of California high school students and a two-wave panel survey of high school students in Chicago, we find that different pedagogical approaches influence different forms of civic and political engagement. Specifically, controlling for prior levels of engagement and demographic factors, we find that open discussion of societal issues promotes engagement with political issues and elections. In contrast, service learning opportunities increase community-based and expressive actions. Both kinds of opportunities promoted commitments to participatory citizenship. These patterns can teach us about the kinds of opportunities (both in school and out) that can shape adolescents' civic and political development. KEY WORDS: civic education, youth political development, political socialization "The qualifications of self-government are not innate," wrote Thomas Jefferson (1824). "They are the result of habit and long training" (p. 22). Few doubt this, but how best to train and educate for self-governance remains a matter of much uncertainty. Indeed, at the end of their influential assessment of high school civic education, Concern regarding this inattention grew due to a spate of studies detailing the low levels of civic and political engagement among youth Such findings, especially when aggregated for policy maker and practitioner audiences, may obscure differences in the relationships between varied educational approaches and varied outcomes. Civic learning opportunities are a collection of diverse approaches ranging from discussion of controversial issues, to service learning, to simulations, to learning how a bill becomes a law. Civic and political outcomes range from knowledge, to tolerance, to voting, to volunteering, to engaging in protests. Practices that advance one goal may not advance others. It is therefore important to compare the impact of varied approaches on a range of different outcomes. Recent evidence that norms regarding the form and focus of politics may be changing and that these changes may be most pronounced among youth makes the need for such analysis even clearer. Specifically, analysis suggests that a broad shift is occurring in the form and focus of politics. Scholars have found that youth are less likely to focus on what we refer to as "big P" politics-on elections and on the influence of elites and state institutions. Youth appear to be turned off by the conflictual and seemingly ineffectual nature of the political process and express, for example, less interest in elections (voting and working on a campaign) and in the traditional political debates regarding state institutions engaged in by politicians, interest groups, and elites We relate our analysis of two prominent civic education strategies to this distinction. Indeed, as will be discussed below, some prominent approaches to civic education (discussion of controversial political issues, for example) focus attention on "big P" politics and aim to make this focus more compelling, while other prominent approaches (service learning, for example) often focus attention more on what individuals can do to help others and on the compelling possibilities of "little p" politics. Understanding the relationship between varied pedagogical practices and these different kinds of engagement can help us assess the alignment of varied curriculum with these broader societal trends. In addition, studies of civic education's effect on political engagement can provide a valuable complement to other models of factors that promote civic and political participation. For example, rational models of political engagement predict that individuals will act politically when the benefits outweigh the costs of doing so. Such models have a long pedigree in studies of electoral turnout and vote preference (see, e.g., Kahne et al. 420 Focusing on the development of social movements, While valuable in many respects, these models often pay scant attention to the potential significance of nonpartisan civic educational interventions. Models of educational intervention focus on schools' role in developing participatory identities, commitments, knowledge, skills, and habits Two Strategies for Promoting Civic and Political Engagement We focus on two prominent civic learning strategies that scholars, policy makers, and practitioners have identified as promising (see Open Discussion of Societal Issues Portraits of effective strategies for promoting civic engagement often highlight the value of open and informed discussions of societal issues. There are several rationales for this emphasis. Interestingly, these rationales respond to several of the dynamics that may be undermining interest in "big P" politics. First, it appears that the conflictual nature of politics makes many youth (and adults) hesitant to engage. There is hope that normalizing this conflict will increase engagement with politics At their best, when engaging in such discussions, youth consider diverse perspectives, are respectful, and are informed by research and careful analysis Service Learning It is also common for educators, policy makers, and scholars to promote service learning as a means of fostering civic and political engagement Research Questions There is much we can do to strengthen our understanding of how these two prominent civic learning opportunities relate to civic and political engagement. First, aside from a few studies of particular interventions, there have been very few studies of open discussions of societal issues or service learning that assess the impacts of these activities on civic and political outcomes while also controlling for prior levels of commitment and activity. Indeed, since much research in this area is cross-sectional in nature, claims regarding causality are difficult to make. Moreover, because studies of specific interventions focus on a particular curriculum, a particular school or context, or a particular population, the risk of overgeneralizing is substantial. For these reasons, we examine Kahne et al. 422 whether open discussion of societal issues and service learning influence students' civic and political interests, commitments, and behaviors as young adults when controls are in place for demographic factors and prior levels of activity. H1: Open discussion of societal issues and service learning will promote civic and political engagement. Second, and of more central importance in this study, is consideration of whether differing pedagogical approaches influence different civic and political outcomes. Many studies of civic education examine a specific civic or political outcome. Voting, volunteering, protesting, writing a blog about an issue, and joining a community organization are all different kinds of civic and political activity; it is not clear that the same learning experiences will impact these different outcomes in the same ways. Illustrating just how careful one must be, a recent study of a curriculum focusing on constitutional rights and civil liberties found that this type of instruction increased the knowledge of those topics but not support for civil liberties This service-politics split and its relationship to civic education practices may be particularly important given the recent changes in youth perspectives on civic and political engagement. With these dynamics in mind, we examine whether different civic learning opportunities promote differing forms of civic and political engagement. H2: Service learning will promote "little p" politics (community-based and expressive action). H3: Open discussion of societal issues will promote "big P" politics (engagement with political issues and elections). Methods Drawn from two two-wave panel surveys of high school students, one in California and one in Chicago, the data on which we base this study offer a unique vantage point from which to assess the effects of civic education on youth civic and political involvement. Other data sets are crosssectional; they focus on a particular curricular practice, a particular intervention, or a specific population; or they lack controls for prior levels of civic and political engagement. We begin by detailing the methods used in and the results of our California study. In addition to the panel design, two strengths of that study are that it includes a very diverse student sample and had a wide range of outcomes. One limitation of the California sample is that it does not include a measure of student access to civic learning opportunities at T1, the first wave of our panel study. For Different Pedagogy, Different Politics 423 this reason, we then turn our attention to the Chicago sample. This sample includes a measure of civic learning opportunities at T1 and thus enables us to examine reciprocal effects and address some concerns regarding unobserved heterogeneity in preexisting dispositions to civic participation that might be raised with the California sample. Study 1: The California Civic Survey As a part of a larger initiative called the California Civic Survey (CCS), we surveyed students from seven high schools, each in a different school district in California. The districts and schools were purposely selected to ensure a diverse range of demographic and academic characteristics. The percentages of students receiving a free or reduced-price lunch varied widely across the schools we examined, from 1% to 83%. In addition, the sampled schools reported average Academic Performance Index (API) scores ranging from the bottom 20% to the top 10% of students from all of the high schools in California. Of those students who identified their ethnicity, 36.7% were European American, 30.8% were Asian American, 18.2% were Latino, and 7.6% were African American. T1 was conducted with 1,203 California high school juniors in 2006. The following year (T2), in 2007, we revisited the same schools and administered follow-up surveys to 502 of the students who had taken the survey the year before. This represents a panel retention rate of 41.7%. To minimize selection bias, we surveyed entire classes of juniors and seniors. Selection of these classes was based on class schedules and the availability of the computer lab, with absolutely no attention paid to or knowledge of students' exposure to civic learning opportunities or involvement with civic or political life. Since we surveyed whole classrooms of students during school time (in both T1 and T2), the size of the panel was not meaningfully influenced by students' willingness to take the survey. Attrition was due to the complexities associated with getting the T2 survey to those who had been surveyed in 2006. Students who were in classes in which the survey was administered in 2006 were spread out in different classes in 2007. We conducted the surveys in as many classrooms as we could, and thus reached many students, but since we were not able to pull individuals out of classes to take the survey, we were not able to resurvey everyone. We conducted regression analyses (not reported here) and found that the demographic characteristics of both samples were statistically similar, with the exception that there were more Latinos in T1 than T2. Those who took the T2 surveys scored higher than those who took the T1 but not the T2 surveys, however, on measures of voluntary activity, intention to vote, participatory citizenship, and political interest (p < .05). They scored lower on our measure of expressive and youth-centered action (p < .05). In response to this potential problem, our models control for all initial demographic and outcome differences and we focus on the impacts of practices on only those who took both the T1 and the T2 surveys. Neither sample proportionately mirrors the diversity of California or the nation. Thus, we do not use these data to make claims regarding the frequency of varied educational opportunities or outcomes. We focus instead on the relationships between civic learning opportunities and outcomes. Our subsequent analysis is based on those students who completed both waves of our data collection. Measurements Three groups of variables were created from the two waves of panel data: (a) indicators of civic and political engagement (outcome variables), (b) measures of opportunities for open discussion of societal issues and for service learning, and (c) control variables. Multi-item batteries for the dependent and independent variables allowed us to probe more accurately and deeply for the specific components of civic education and political engagement. For wording of the survey questions, see Appendix A. Kahne et al. 424 Measuring Civic and Political Engagement In an effort to sample diverse forms of civic and political engagement, we collected information on six different dependent variables. These measures drew from recent studies of and conceptual schemes for youth civic and political engagement (see We also explored the effects of civic education on three outcomes that are associated with "big P" politics. We assessed intention to vote, interest in politics, and interest in diverse perspectives. Choosing between candidates competing for elected office is perhaps the quintessential "big P" political act. The phrasing of the political-interest item ("I am interested in political issues") is suggestive of "big P" politics as contestation of policy issues in the formal sphere. And interest in differing perspectives may signal willingness to engage in the conflicts and disagreements inherent in "big P" politics. Finally, one of our measures, participatory citizenship (the belief that being concerned about and actively involved in community, state, and national issues is everybody's responsibility), straddles the line between "little p" and "big P" politics. Some items' emphasis on involvement in community organizations clearly embraces concerns at the core of "little p" politics, but other items' mention of national and state issues evokes the contentious, conflictual politics that typically characterize the "big P" variety. Thus, we place the participatory citizenship construct in a hybrid "little p"/"big P" category in our results tables. It should also be noted that while the items for this measure in the Chicago survey are similar to those in the California survey, they place more emphasis on community action and do not mention state or national engagement and therefore might be seen as fitting more fully within the realm of "little p" politics. Each of these six dependent variables was measured on two occasions, once during a student's junior year (T1) and once during senior year (T2). Appendix A contains the precise wording of items that constitute the dependent variables, as well as the response sets for each. Although we combined multi-item indicators into latent variables, using structural equation models (which removes measurement error), we nonetheless conducted reliability tests to determine whether items that we classified together a priori on the basis of conceptual similarity do, in fact, belong together. The items hang together well, both conceptually and empirically. The mean Cronbach's alpha coefficient for our outcome variables is 0.78, and all but one individual value was greater than 0.72. Given that the rationale for providing civic education often stresses long-term civic and political engagement, it is worth highlighting that our dependent variables are not only useful indicators of current behaviors and commitments but also appear to be solid predictors of future activity. Specifically, in a related panel study, we used the same measures of civic and political outcomes and surveyed individuals when they were high school seniors and then one-and-a-half to three-and-a-half years later. We found that the civic and political behavior and commitments that youth reported as high school seniors were strong predictors of behavior and commitments in early adulthood (Kahne, Crow, & Lee, forthcoming). In short, youth who reported frequently volunteering as high school seniors were more likely than others to report continuing to volunteer over the next few years. Youth who frequently participated in acts of expressive and youth-centered action were more likely than others to continue doing so. Youth who as high school seniors expressed a strong intention to vote regularly were more likely to have voted, and so on. This result is fully consistent with the broader literature, which finds that the behavior and expressed commitment of adolescents to civic and Different Pedagogy, Different Politics 425 political engagement are strong predictors of future activity Measuring Civic Learning Opportunities This study examines the impacts of opportunities for open-classroom discussion of societal issues and opportunities for service learning. Our measure of open discussion comprised five items, all of which were measured at T2 (in the spring of the students' senior year). These items asked about the extent to which, during the past year, students learned about, researched, and discussed contemporary social problems and the extent to which teachers encouraged students to make up their own minds on social and political topics. As with our multi-item dependent variables, discussion of societal issues appears in our analyses, purged of measurement error, as a latent construct in the structural equation models. Nonetheless, to ensure that the items did indeed belong together, we conducted reliability and factor analyses. Cronbach's alpha for the five discussion-centered learning items is a = 0.85. A single-factor confirmatory analysis reveals that the five items are, in fact, manifest indicators of a single underlying concept. Our measure of service learning, which was also taken at T2, consisted of a single item that gauged the frequency that students, during the past year, participated in "community service projects at school" as (1) "never," (2) "sometimes," (3) "often," and (4) "very often." Control Variables To avoid spurious findings and to help ensure that civic engagement gains are attributable to civic education, we included a number of control variables that scholars have found influence high school students' civic engagement. These include demographic factors, such as ethnicity and gender (see Analytic Strategy We analyzed the data using structural equation models (SEMs) that included the prior (T1) values of the latent-construct outcome variables as independent control variables. That is, latent constructs for each of the six forms of civic and political engagement at T2 were regressed on the corresponding T1 latent constructs (an autoregressive component), the two civic education independent variables (i.e., the latent construct for discussion-oriented learning and the single-item service learning variable), and the control variables. Kahne et al. 426 Combining structural equation models with dynamic, autoregressive models has two main advantages. First, SEM eliminates measurement error as a possible cause of explained variation in the dependent variables. Second, the autoregressive component in our models controls for students' prior propensity toward civic engagement. These models provide unbiased estimates of the effects of civic education on civic and political engagement by adjusting for any initial differences in the outcome variables that might exist between those who were already active as high school juniors and those who were not Results Engagement in "Little p" or "Big P" Politics, or Both. In general, service learning was more strongly related to our measures of "little p" politics than was discussion of societal issues (see For its part (as Finally, open discussion and service learning both promoted participatory citizenship, a hybrid "little p"/"big P" civic engagement outcome. The coefficient associated with open discussion was b = 0.11, and with service learning, b = 0.06 (see Other Determinants of Engagement Of course, factors other than open discussions and service learning affect the civic engagement of high school students during their senior year. First, in all cases, students' T1 measures of civic or political engagement (the dependent variable) were highly related to T2 measures of the dependent variables. 1 In addition, students whose parents discussed politics with them reported greater engagement in expressive and youth-centered action (b = 0.09), greater interest in politics (b = 0.25), and a greater intention to vote (b = 0.10) than did students from homes where political discussion was less frequent. Students whose parents were politically active participated themselves in expressive and youth-centered action more (b = 0.10) than students with less politically engaged parents. Curiously, our findings indicated that mother's education (as a proxy for SES) is negatively associated with expressive and youth-centered action (b = -0.06) and participatory citizenship (b = -0.04). Both the direction and intensity of students' ideology affected some of the civic engagement outcomes. Conservative students scored lower than their more liberal counterparts on the scale of interest in diverse perspectives (b = -0.07). Ideological extremism, whether on the left or the right, was associated with a lower commitment to participatory citizenship (b = -0.08). Last, there were significant differences in engagement with political issues and elections across ethnic groups and gender and by educational achievement. Latino high school students reported less of a commitment to participatory citizenship (b = -0.13) than did those in the European American reference group, and African American students reported less interest in politics (b = -0.37); however, given the small number of African American survey respondents, we interpret this finding with caution. In contrast, the Asian American high school students in our sample were more likely to engage in expressive and youth-centered action (b = 0.16) and to volunteer (b = 0.27). Female students demonstrated a greater commitment to participatory citizenship than did their male counterparts (b = 0.09), and students with a higher GPA were more likely to say they intended to vote (b = 0.15). Do the Effects of Civic Learning Opportunities Vary Across Subgroups? Using interaction terms, we also assessed whether discussion or service learning opportunities had differential impacts on all six of our measures of civic and political engagement ("little p," "big P," and mixed), depending on race, ethnicity, gender, SES, or plans to attend college. We found no statistically significant interaction effects and therefore have no reason to believe that these practices are more effective for some groups than for others (contact authors for details). Concerns Regarding Endogeneity While our analytic design has several strengths, one concern might be that the initial observation of the dependent variable is not necessarily exogenous. It could be influenced by "unobserved heterogeneity" that exists prior to the first measurement, a situation econometricians refer to as the "problem of initial conditions" We also used propensity score matching to test for possible bias. First, we used cluster analysis to divide the sample into two groups, "high" and "low" exposure, on the basis of our measures of 1 Appendix B contains full regression tables with all coefficients for each of the six dependent variables, fit statistics, and sample size. Kahne et al. 428 civic learning opportunities. Then we used propensity score matching to couple each member of the "treatment" group (in our case, "high" exposure to civics education) to one or more members of the "control" group (low exposure) on the basis of demographic, socioeconomic, and ideological covariates. If civic engagement gains owed to different compositions of the "treatment" and "control" groups rather than to different levels of exposure to civics education, we would expect sample matching to eliminate T1 to T2 civic education gains in the "treatment" group. It did not. 2 Perhaps one of the best ways to respond to this set of concerns is to test for reciprocal effects by including measures of the independent variable at T1. In other words, it may be that service learning causes a desire to volunteer, but it may also be that individuals volunteer first and then pursue opportunities for service learning. Unfortunately, we do not have measures of our civic learning opportunities at T1 in the California data set. However, we were able to perform a similar analysis on a panel data set that was collected in Chicago, where measures of civic learning opportunities were collected at T1. This analysis, discussed below, allows us to examine whether reciprocal effects are in fact occurring. Study 2: The Chicago Survey To complement our analysis of the CCS data, we analyzed panel data from Chicago high schools that had been collected by the Consortium on Chicago School Research in 2003 and 2005. The surveys were given during an assigned period to all ninth-, tenth-, and eleventh-grade students at a given school. Our analysis focuses on 52 schools that gave the survey in both years, and it includes those students who took the survey as ninth graders in 2003 and then again two years later in 2005 (N = 4,314). The demographics of this sample differ somewhat from the California sample in that 39.7% of the students were African American, 39.5% were Latino, 13.2% were White, and 7.3% were Asian (for additional details, see Kahne & Middaugh, 2008). 3 While these data were analyzed in an earlier study (See Kahne & Middaugh, 2008), the current analysis is different in that it includes a measure of civic learning opportunities in T1 (that variable was not used in the earlier study) and a dependent variable (students' intention to vote) that was also not included in the analysis by Measurements To enable a comparison to the California analysis, to the extent possible, we included in the analysis of the Chicago data survey items that aligned with those used in the analysis of the California data (see Appendix A). Open discussion of societal issues was measured by two items asking students how often their teachers (a) encourage students to discuss political and social topics in which people have different opinions and (b) encourage students to make up their own minds about political and social topics (interitem r = 0.66 for the 2003 survey and r = 0.73 for the 2005 survey). 4 Similar to the 2 Due to space limitations, we do not present the results of the augmented SEM or the matching analyses, but we are happy to furnish them on request. 3 This sample has 257 more students than the sample used in Different Pedagogy, Different Politics 429 California survey, we used a single item to measure service learning, where students expressed whether they had "worked on a service learning project to improve my community." Whereas we used a total of six outcome variables in the California survey, only two outcome variables were available in the Chicago survey: commitment to participatory citizenship and intention to vote. Participatory citizenship was measured by four items asking students about their level of agreement with the following four statements, again on a 4-point scale: (a) "being actively involved in the community is my responsibility," (b) I have good ideas for programs or projects that would help solve problems in my community, (c) "In the next 3 years, I expect to work on at least one community project that involves a government agency," and (d) "In the next 3 years, I expect to be involved in improving my community" (Cronbach's a = 0.78 for the 2003 survey and 0.86 for the 2005 survey). Intention to vote was measured by a single item asking the extent to which students agreed with the statement "Once I am old enough, I expect to vote in every election." Consistent with the analysis of the CCS data, we used SEM, where each latent construct is represented by its measured indicators to correct for measurement error. Although not displayed in the diagram of our model Analytic Strategy To analyze the potential reciprocal relationship between civic learning opportunities and civic and political outcomes, we employed two-wave cross-lagged panel models where each T2 variable The paths from our exogenous variables (i.e., sex, race, and math and reading scores) to the endogenous variables were estimated but are not displayed. Kahne et al. 430 (both latent constructs and single-item indicators) was predicted by that variable's previous value at T1 as well as the values of other variables at T1. More specifically, this model examines not only whether experiencing civic learning opportunities in 2003 predicted the levels of civic engagement in 2005 but also whether civic and political engagement in 2003 predicted engagement in civic learning in 2005. That is, the model estimates these cross-lagged effects while controlling for the lagged values of each outcome variable (as well as the lagged values of each independent variable). By accounting not only for the temporal stability of the dependent variables (as in the lagged dependent variable SEM models for the CCS data) but also for that of the independent variables, this technique provides more robust estimates of the effects of civic learning opportunities on changes in civic and political engagement over time Results Figure 1 describes a cross-lagged panel model that we ran to examine the effects of civic learning opportunities on participatory citizenship and on voting intention. This model produced a Chi-square goodness-of-fit statistic that was large enough for us to reject the null hypothesis that the model fits (c 2 = 451.7, df = 122, p < .001), thus indicating model misfit. However, the model yielded large values of incremental fit indices, such as Bentler's Comparative Fit Indices (CFI) and TuckerLewis indices (TLI), which are all greater than .95, a widely accepted threshold value for goodmodel fit (CFI = .99 and TLI = .98). Additionally, root-mean-square error of approximation (RMSEA) and standardized root mean residual (SRMR) values that are much smaller than a conventional threshold value of .05 also indicate good model fit (RMSEA = .03 and SRMR = .01). Given our large sample size (N = 4,314) and the known high sensitivity of Chi-square-based fit indices to sample size, we believe that the model fits the data well. Not surprisingly, autoregressive effects (i.e., regressions of a variable to its prior value) are particularly strong, ranging from b = .08, p < .001 for service learning to b = .34, p < .001 for participatory citizenship. More important, estimates of cross-lagged effects showed a pattern that was consistent with the California data and that indicated reciprocal causality. Specifically, and consistent with the California data, service learning, which was measured in 2003, was found to predict participatory citizenship assessed in 2005, b = .06, p < .01. Also consistent with the California data, voting intention assessed in 2005 was predicted by open discussion (b = .04, p < .05) assessed in 2003, but the lagged effect of service learning (b = -.04, n.s.) on voting intention was not significant. One notable difference we found between the two datasets was that open discussion, which was found to predict participatory citizenship in the California data, turned out to be unassociated with participatory citizenship in the Chicago data, b = .04, n.s. Our Discussion and Implications Civic Learning Opportunities Can Promote Civic and Political Outcomes For scholars, policy makers, and educators, the clearest implication of these findings is that civic learning opportunities promote civic and political engagement. We found that discussions of Different Pedagogy, Different Politics 431 societal issues and service learning opportunities foster gains in behaviors such as voting and volunteering, commitment to civic participation, and interest in politics and divergent perspectives. The two data sets examined a highly diverse group of students, considered prominent civic learning opportunities, controlled for prior levels of the outcomes in question, and focused on a wide range of outcomes. That analysis of both data sets highlighted similar relationships, that the California data set indicated changes in commitments as well as in behavior and interest, and that the findings were present in the cross-lagged model from the Chicago data adds to our confidence. We know of no other studies that have combined all of these features. As a result, analysis of these two data sets provides some of the strongest evidence to date regarding the impact of high school civic learning opportunities. These findings also help clarify interpretation of earlier studies that called into question the value of civics instruction. For example, studies of course-taking practices found very limited impact on civic and political outcomes Different Civic Learning Opportunities Promote Different Types of Engagement While the links between particular civic learning opportunities and civic and political engagement are encouraging, focusing solely on the efficacy of these practices may obscure differences in their impact. With only one exception, we found that different civic learning opportunities promoted different types of engagement. The one exception was the commitment to participatory citizenship in the California sample. We found that both service learning and open discussion of societal issues promoted participatory citizenship. This was not the case in the Chicago sample where only service learning promoted participatory citizenship. As noted in the methods section, this may be due to the fact that the participatory citizenship measure in the California survey includes attention to both "big P" and "little p" politics. In all other cases, engagement in service learning and open discussion of societal issues were associated with differing outcomes. For example, in both the Chicago and the California samples, open discussion of societal issues promoted the intention to vote, while service learning did not. In the California sample, where additional outcomes were assessed, service learning promoted voluntary activity and varied forms of expressive and youth-centered action, while open discussion of societal issues did not. Open discussion, however, promoted interest in politics and interest in diverse opinions, while service learning did not (although it was marginally related to a gain in interest in politics). This variation reinforces the need to attend to differing forms of civic learning opportunities and distinct dimensions of youth civic and political engagement. Indications, for example, that a given learning opportunity or context influences a particular outcome, such as voting, do not imply that that opportunity or context will foster volunteering. Conceptualizing the reasons for varied relationships requires assessment of the substantive focus of these learning opportunities, appreciation of the ways in which youth cluster varied forms of civic and political engagement, and consideration of the mechanisms that would lead curricular opportunities to influence varied outcomes. Below we describe one possible explanation for the differential impact of these civic learning opportunities. This explanation highlights a distinction noted earlier, between "big P" Politics and "little p" political and civic activities. We suspect that open-classroom discussions of societal issues Kahne et al. 432 generally focus on and promote "big P" Politics, while service learning opportunities more commonly focus on and therefore promote "little p" politics. We discuss this possibility below. A "Big P" and "Little p" Politics Split As noted in the introduction, scholars have argued that many youth are turned off by traditional ("big P") Politics (which they associate with electoral politics, disagreements, large interest groups, elites, and institutions), viewing such activities as corrupt, often ineffectual, filled with conflict, and highly partisan The distinction many youth make between formal politics on the one hand and informal, nongovernmental, community-based activism on the other provides a perspective from which to interpret our results. The two curricular approaches we examined can be seen as addressing different sides of the "big P"/"little p" politics split. For example, most discussions of societal issues focus on traditional "big P" Politics (e.g., policies Congress is debating) and highlight their importance In contrast, service learning opportunities often align well with forms of "little p" civic and political life, such as volunteerism and expressive activities. In addition, when youth are engaged in a service learning curriculum, often little attention is paid to politics, the importance of structural factors, or divergent perspectives; the focus is on ways youth can make a direct difference through such activities as environmental clean-up, tutoring youth, or helping out at a senior center (Walker, 2000; That both discussion and service learning opportunities were related to gains in commitments to participatory citizenship may have been due to the fact that the measure of participatory citizenship contains items that focus on both "little p" politics and more formal "big P" Politics. For example, respondents were asked if they agree with the statement "Being concerned with national, state, and local issues is an important responsibility for everybody." In short, different civic learning opportunities appear to influence civic and political engagement in differing ways. The alignment of varied civic learning opportunities with the "big P"/"little p" politics split provides an explanation for the relationships we found and adds to growing recognition of the importance of this distinction when thinking about youth politics. More studies are needed to clarify the nature of these relationships and examine alternative explanations. Such studies might also help educators to identify ways to tailor their curriculum so as to promote a broader range of outcomes simultaneously. Colby, Beaumont, Ehrlich, and Corngold's study of higher education curriculum (2007), for example, demonstrates that service learning and other experience-centered approaches can be structures to promote engagement with formal politics. Different Pedagogy, Different Politics 433 Similarly, we can imagine many ways that discussion of societal issues might be structured to support interest in community-based forms of engagement. More fine-grained analysis of particular curriculum and studies employing a range of mediating variables are needed to deepen our understanding of the factors that shape varied results. The Efficacy of Civic Learning Opportunities Highlights the Need for Equitable Distribution of Civic Learning Opportunities While the discussion above has focused on whether civic learning opportunities influence civic and political outcomes and on which opportunities influence which outcomes, these findings also have implications for those concerned about civic and political inequality. Given the indications from this study that discussion and service learning can promote civic and political outcomes, the possible impact of these opportunities on civic and political inequality deserve attention. Unfortunately, studies indicate that these learning opportunities are inequitably distributed. Drawing on a nationally representative sample of ninth graders, The good news from this study is that discussion and service learning opportunities appear to work equally well for all demographic groups. We found no significant interaction effects resulting from demographic variables. Policies that monitor whether the distribution of these civic learning opportunities is equitable and policies that provide equitable access to these opportunities are needed. Studies Should Attend to Reciprocal Causation To our knowledge, this is the first study to test for reciprocal causation between classroom-based civic learning opportunities and civic outcomes. It found statistically significant relationships in both directions. Reciprocal causation is not surprising: Students who are interested in service or politics might well be expected to seek out learning opportunities related to their interests. Due to these reciprocal relationships, two-wave, cross-lagged panel designs appear to be important. Crosssectional designs or designs that do not control for the impact of student commitments on their exposure to civic learning opportunities appear likely to overstate the relationships between civic learning opportunities and civic outcomes. Limitations This study has several limitations. First, in neither case was the sample representative of the nation. For that reason, we do not use these data to characterize the frequency of civic learning opportunities in schools in the United States. In addition, while a main strength of the California data set (attention to multiple outcomes) complements a strength of the Chicago data set (the ability to control for civic learning opportunities in T1), it would be better if both data sets had both features. Also, while we know from a related study that the behaviors measured in the California study are predictors of the same civic and political behaviors in early adulthood, it would be ideal to follow these particular students into adulthood to see how changes in activities and intentions during high school were related to behaviors in early adulthood. Kahne et al. 434 It would also be useful to complement panel studies with experiments that examine specified practices in more controlled environments and with random assignment. For these reasons, while we believe that our results are compelling and are a significant addition to research in this area, we certainly believe that additional research in this area is needed. Conclusion "The death of democracy," wrote Robert Maynard Hutchins (1952) "is not likely to be an assassination from ambush. It will be a slow extinction from apathy, indifference, and undernourishment" (p. 80). Our study suggests that open discussion of societal issues and service learning support civic and political engagement. More specifically, it appears that service learning tends to promote expressive, youth-centered, and community-based engagement, while open discussion of societal issues tends to promote more formal and electoral forms of political engagement. These practices appear effective for a broad cross-section of youth. Attending to the varied ways in which classroom experiences can foster youth civic and political engagement seems to be well worth the effort. ACKNOWLEDGMENTS