Alternative Schooling Strategies and the Religious Lives of American Adolescents

Abstract

I analyze the effects of Catholic schooling, Protestant schooling, and homeschooling on adolescents’ religious lives and test three mechanisms through which these schooling strategies might influence religiosity: friendship networks, network closure, and adult mentors. Data from Wave 1 of the National Survey of Youth and Religion suggest that Catholic schoolers attend religious services more frequently and value their faith more highly than public schoolers, but attend religious education classes and youth group less often. Protestant schoolers’ involvement in their local congregation is similar to public schoolers’, but their faith plays a more salient role in their life and they are more active in private religious activities. Homeschoolers do not differ significantly from public schoolers on any outcome considered. Moreover, friendship networks, network closure, and adult mentors play a very limited role in mediating the relationships between schooling strategies and adolescent religiosity. Interpretations of these findings are presented and discussed.

Introduction

Recent social scientific research has begun to document the social contexts that shape adolescent religiosity. These studies tend to emphasize the family, which is typically considered the primary agent of religious socialization, along with friends and religious congregations (Cornwall 1988; Erickson 1992). Yet adolescents spend many of their waking hours in school, and the religiosity of their schoolmates may actually have a larger effect on their religious lives than does the religiosity of their friends (Regnerus, Smith, and Smith 2004). The religious effects of different types of schools—namely Catholic, Protestant, and homeschools—however, are rarely evaluated. Even rarer are nationally representative assessments of these alternative schooling strategies. So while proponents of these schools often cite the religious advantages of educating children in settings that promote the moral and spiritual values of the (religious) family (Parsons 1987; Princiotta, Bielick, and Chapman 2004), there remains a lack of compelling social scientific evidence to either support or refute their claim.

Influential studies on Catholic schooling date back to the 1960s and 70s (e.g., Greeley and Rossi 1966; Greeley, McCready, and McCourt 1976), but significant demographic shifts in these schools have taken place in the years since, and they might now be better understood as elite private schools rather than working-class institutions catering specifically to Catholics (Baker and Riordan 1998). Studies of Protestant schools (e.g., Erickson 1964; Johnstone 1966) are equally dated and do not account for the extraordinary growth and expansion of these schools during the 1970s and 80s. Even less is known about homeschoolers. They and their families are difficult to locate, and when found, they respond to surveys at notoriously low rates (Collom 2005). A sizable proportion of these students are educated at home for explicitly religious reasons, but the religious effects of the homeschooling approach are unknown to social scientists.

Conclusions about the religious influence of alternative schooling strategies to date are, according to one review, “muddy” (Spilka et al. 2003:118). This study uses data from Wave 1 of the National Survey of Youth and Religion (NSYR), collected in 2002–2003, to assess the effect of alternative schooling strategies on a variety of religious outcomes: religious service attendance, religious education class attendance, youth group attendance, religious salience, and private religious activities. This study also explores how friends’ religiosity, network closure, and adult mentors might mediate these relationships, as well as any cumulative effects of these schooling strategies that might be present. I also explore how schooling strategies may explain in part how parents influence their adolescents religiously.

Schooling Effects on Adolescent Religiosity

Despite noisy public discourse about the place of religion in education, the influence of different schooling environments on the religious lives of adolescents is seldom explored. The school environment may serve to reinforce the religious socialization of the parents or, alternatively, to break down the religious plausibility structures erected by the family (if the family is at all religious). Social scientists are becoming increasingly aware of the complex ecological influences adolescents encounter and navigate (Bronfenbrenner 1979; Regnerus, Smith, and Smith 2004), yet social scientific research rarely considers different schooling types, and their unique religious contexts, as meaningful predictors of adolescent religiousness. When schooling types are considered, Catholic and Protestant schoolers are typically combined into one homogeneous group of “religious schoolers” (e.g., Gunnoe and Moore 2002; Regnerus, Smith, and Smith 2004; Trinitapoli 2007). This strategy may mask important differences between Catholic and Protestant schools.

Catholic Schooling

Much of the research on Catholic schools evaluates long-term effectiveness, focusing on adult religious outcomes. Older research indicated that Catholic schooling had a strong positive religious effect in adulthood. Greeley and Rossi’s (1966) seminal work on Catholic schools showed positive religious benefits of Catholic school attendance, benefits that persisted after accounting for parental religiousness and social class. Schooling effects, however, were less important than these other two attributes. Greeley and his colleagues have returned to the question in subsequent years and found similar results (Greeley, McCready, and McCourt 1976; Fee et al. 1981). More recent research on Catholic schooling provides mixed findings. One study of Catholic adults who were raised Catholic (i.e., “cradle Catholics”) found that Catholic schooling is related to higher levels of (a) traditional Catholic beliefs and practices, (b) recent practices (i.e., Bible study and prayer groups), and (c) agreement with Catholic teachings on social and reproductive ethics, but only for Catholics with more than 12 years of parochial schooling, so attendance at Catholic universities, not secondary schools, may be the important predictor of adult religiosity (Davidson et al. 1997). On the other hand, Perl and Gray (2007) report that more than three years attendance in a Catholic high school protects against disaffiliation from Catholicism in adulthood.

Much has changed demographically within Catholic schools, however, since many of these studies’ respondents were educated. More than 2.5 million students are now enrolled in Catholic schools, including 600,000 secondary school students (Broughman and Pugh 2004). And according to Riordan (2000), 21 percent of the seniors are not actually Catholic, up from only two percent in 1972. More than 20 percent of them do not even consider themselves religious. The changes involve less religious matters as well. In 1992, 46 percent of Catholic secondary schoolers came from families in the top quartile of a socioeconomic scale, compared to just 27 percent of public schoolers (Riordan 2000). Given these shifts in the last few decades, many Catholic schools seem to be turning into elite academies, focused as much on academic achievement as religious development (Baker, Han, and Broughman 1996; Baker and Riordan 1998). Perhaps, as one Indiana parishioner put it, Catholic school students are just “[shooting] paper wads in religion class” (Davidson et al. 1997:99).

Protestant Schooling

We know more about the goals of Protestant schools than about whether they actually achieve those goals. Two ethnographic studies of fundamentalist schools indicate a major objective of Protestant schooling to be sheltering children from the evils of the outside world, to “keep them out of the hands of Satan” (Peshkin 1986; Rose 1988). But these studies do not speak to the vast diversity within Protestant schooling, diversity that springs from the religious differences within conservative Protestantism itself.¹ Fundamentalist schools tend to emphasize the development of character, morals, and strict doctrine, while evangelical schools focus on the “integration of faith and learning,” not separation from secular society. And denominationally affiliated schools, namely those run by the Lutheran Church–Missouri Synod, have another agenda altogether: They emphasize Lutheran distinctives such as music, German language, and confirmation classes (Sikkink 2001).

The demographics of Protestant schools are also fairly well understood. These schools enjoy their strongest support from Pentecostals and charismatics; evangelicals are more apt to support public schooling for their children as an opportunity to “witness” to non-Christians (Sikkink 1999; Smith 2000). Children attending Protestant schools tend to be white, from two-parent families, residents of the Southern US, from two-income families, and from parents with higher-than-average educations and incomes (Sikkink 2001). These demographics may vary among types of Protestant schools, however. Significant class differences exist between the working-class fundamentalist schools and the middle-class evangelical schools (Rose 1988).

Despite the notable differences among Protestant schools, significant commonalities bind them together. Protestant schools are overwhelmingly devoted to the religious development of their students as a top priority of their institution (Baker, Han, and Broughman 1996). They also tend to view the family, church, and school as partners in the development of their students, a sort of threefold sacred canopy that fosters religious development (Sikkink 2001). With this broader goal in mind, many Protestant schools play down doctrinal distinctions, offering a generic conservative Protestant education (Wagner 1997). So Protestant schools may differ in their methods and makeup, but they appear united in their focus on the religious development of their students.

Research assessing the success of Protestant schools toward this end has largely been denominationally based, presumably due to data-collection issues. Erickson’s (1964) study stands as one exception. When he compared students in five “Fundamentalist” day schools to those in public schools, there was no significant difference in religiosity after family and church background were controlled. A study of Lutheran-school students and Lutheran children in public schools reported similar findings, attributing the differences in religiosity to the family of origin (Johnstone 1966).

Studies of evangelical-school students are scarce. One study of Arkansas evangelical schoolers finds a positive cumulative effect of Protestant schooling on biblical literacy (Simpson 2002). Another study of students from seven evangelical schools, however, indicates that the number of years of evangelical schooling does not affect the biblical worldview of students (Meyer 2003). While these studies speak to the cumulative effect of Protestant schooling, they do not compare Protestant schoolers to public schoolers to determine any baseline effects. Considering the growth and change of Protestant schools over the past several decades,² an investigation into the impact of these schools on adolescent religiosity seems noticeably absent from the literature.

Homeschooling

The homeschooling movement in the United States has enjoyed phenomenal growth over the past four decades. In the late 1960s, only about 15,000 American children were homeschooled (Lines 2001); by 2003, that number had increased to 1.1 million (Princiotta et al. 2004). In just the four years from 1999 to 2003, the proportion of homeschoolers increased from 1.7 percent of the school-age population to 2.2 percent (Princiotta, Bielick, and Chapman 2004). Data from the National Center for Education Statistics indicate that these homeschoolers are typically white; from large, two-parent families; and from parents with high educational attainment (Bielick, Chandler, and Broughman 2001). These students are also difficult to study. Inadequate sampling frames and lack of parental cooperation have impeded thorough research (Collom 2005).

Still, we do know something about why parents choose to homeschool in the first place. When identifying their primary motivation, 31 percent of homeschooling parents cite concern about the environment of other schools and 30 percent indicate religious reasons. The rest note academic reasons or the special needs of their child (Princiotta, Bielick, and Chapman 2004). When able to pick more than one motivation, nearly half of parents think their child could receive a better education at home, while 38 percent cite a religious motivation (Bielick, Chandler, and Broughman 2001). Characteristics such as race and socioeconomic status are not particularly helpful in classifying different types of homeschoolers; they are indeed a diverse population (Collom 2005).

Though a large number of parents homeschool for religious reasons, little is known about whether these parents accomplish their goal of religious socialization (Cizek 1994). Instead, the scant attention paid to homeschoolers is largely devoted to their educational outcomes. As one exception, a study of Baptist youth in Texas finds homeschoolers score no differently than conventionally schooled students on a Faith Maturity Scale, but are less likely to behave in ways that conflict with the teachings of their religious tradition (McEntire 2003). Unfortunately, this study does not consider the religiosity of parents, so it is not clear if these differences (or lack thereof) are attributable to family or schooling effects.

Explaining Schooling Effects on Adolescent Religiosity

Selection Effects

One explanation for schooling effects on adolescent religiosity is selection. Adolescents clearly are not randomly distributed across schooling types; rather, a number of characteristics “select” adolescents into alternative schooling strategies, not the least of which is their parents’ religiosity—perhaps the most important determinant of adolescent religiosity (Smith and Denton 2005). A host of other characteristics, discussed above, may also confound the relationship between schooling types and adolescent religiosity and must be accounted for before any confident claims can be made about the influence of alternative schooling strategies on adolescents’ religious lives.

Mediating Factors

There are at least five ways alternative schooling strategies might influence the religious lives of adolescents. First, attendance at alternative schools—be they Catholic schools, Protestant schools, or homeschools—may situate adolescents in more religious friendship networks. Religious friends, in turn, may lead to higher adolescent religiosity (Erickson 1992; Gunnoe and Moore 2002; King, Furrow, and Roth 2002; Martin, White, and Perlman 2003; Regnerus, Smith, and Smith 2004). Second, adolescents attending these types of schools may have increased network closure (Coleman 1988). In these small, close-knit communities, friends’ parents may have more interaction with adolescents and their parents, which may lead to increased monitoring of adolescent behavior and reinforcement of parental values (Smith 2003). Third, alternative schooling strategies—especially Catholic and Protestant schools—may provide additional religious role models for adolescents in the form of teachers, coaches, and other staff members. As may be the case with the parents of adolescents’ friends, these extra-familial influences may serve to buttress the religious values of parents (Smith 2003). Fourth, adolescents in alternative schooling environments are typically subjected to a large amount of explicitly religious education, which is associated with heightened religiosity among adolescents (Benson, Donahue, and Erickson 1989). Fifth, enrollment in an alternative schooling type may situate adolescents in a moral community or context that values religion. An overt focus on religious and spiritual development among members of the community may foster an environment that is conducive to high levels of religiosity; such an environment may serve as a plausibility structure that upholds religion’s sacred canopy³ and gives credence to an adolescent’s faith (Berger 1967). Indeed, Regnerus, Smith, and Smith (2004) find that schoolmates’ religiosity—even net of friends’ religiosity—is an important predictor of adolescents’ religiousness. Unfortunately, the NSYR data do not allow these last two explanations to be tested directly.

Parent Religiosity as a Moderating Factor

Parent religiosity, in addition to selecting adolescents into alternative schooling strategies, may also moderate the effect of these strategies. Catholic and Protestant schooling may only be effective if the religious values promoted therein are emphasized and reinforced in the adolescents’ home. Alternatively, religious schooling may serve a compensatory function, such that its religious influence is limited only to those who are not from particularly religious families. The moderating effect of parent religiosity may be especially salient for homeschoolers; homeschooling by nonreligious parents may not influence adolescent religiosity at all, while homeschooling by religious parents may foster religious commitment among adolescents.

Schooling Types as Mediating Factors

Understanding the role of alternative schooling strategies in shaping adolescent religiosity may help shed light on how parents transmit their religiosity to their adolescents. Put another way, schooling types may be understood not just as an independent variable predicting adolescent religiosity, but also as a mediating variable between parent religiosity and adolescent religious outcomes. Indeed, although few would deny the important role parents play in their adolescents’ religious development, the mechanisms through which parents influence their adolescents’ religious lives are less clear. Much social scientific research on intergenerational transmission of religion utilizes social learning and social capital theories to explain the role of parents in their adolescents’ religious lives (e.g., Myers 1996; Lee, Rice, and Gillespie 1997; Bao et al. 1999; King, Furrow, and Roth 2002; Regnerus, Smith, and Smith 2004). These studies suggest a direct influence of parents on their adolescents through processes such as modeling and parent-child interaction. Another line of thinking suggests that parents affect adolescent religiosity through what has been termed “channeling” (Himmelfarb 1980). According to this hypothesis, parents indirectly influence adolescents’ religious outcomes by guiding them into more religious social settings, such as peer groups and schools. If the channeling hypothesis is correct, and if different schooling strategies do exert influence on adolescent religiosity, then the influence of parents’ religiosity should be reduced or eliminated once schooling types are considered. If the effect of parents’ religiosity is not reduced, this would suggest a direct-effects explanation for intergenerational transmission of religion.

Data Measures and Methods

Data

The data for this study come from Wave 1 of the National Survey of Youth and Religion (NSYR), a nationally representative telephone survey of 3,290 American adolescents ages 13–17. An oversample of 80 Jewish adolescents was also drawn, bringing the total number of respondents to 3,370. For this study I exclude respondents from the oversample, those not currently enrolled in school, and those without complete information on their schooling type, bringing the total N for this study to 3,217. Means were imputed for missing values on ordinal and continuous variables, and a dummy variable to indicate a missing value on that variable was created. Dummy variables to indicate missing values for binary variables were also created.

The NSYR was conducted by researchers at the University of North Carolina at Chapel Hill between July 2002 and April 2003, and it stands as the most extensive religion-focused survey of adolescents to date. A random-digit-dial (RDD) method was used to conduct the interviews, and the survey was administered to both a parent and the adolescent in the household with the most recent birthday. This sampling method allowed researchers to survey youth not enrolled in school, including dropouts and homeschoolers. It also provided freedom to ask sensitive questions about religion that might not be permitted by school officials in a school-based survey. When weights are applied, the NSYR can be treated as an accurate representation of American adolescents ages 13–17 and their parents. Additional details about the NSYR can be found in Smith and Denton (2005).

Measures

Dependent Variable

This study examines five outcomes, each meant to measure a different aspect of adolescents’ religious lives. The first three dependent variables measure adolescents’ involvement in a religious community: their religious service attendance, religious education class attendance, and youth group participation. These religious behaviors require a certain degree of commitment and action on an adolescent’s part and are thus good, objective measures of adolescent religiousness. The religious service attendance measure is created from two questions. Respondents were first asked if they attended religious services more than once or twice a year (not including wedding, baptisms, funerals, or religious services at their school). Those who said “no” were coded 1 for religious service attendance. Those who responded “yes” were asked a follow-up question about how often they attended. Response categories ranged from a few times a year (coded 2) to more than once a week (coded 7). The religious education class question asked respondents about their attendance at a religious Sunday school, CCD, or other religious education class over the past year (not including classes at their school). Response categories ranged from never (coded 1) to more than once a week (coded 7). Youth group participation was constructed similarly to the religious service attendance measure. Those who said they were not involved in a youth group (and those who did not attend church and were skipped out of the question) were coded 1, while those who say they were involved were asked a follow-up question about the frequency of their participation. Responses to this follow-up question ranged from almost never (coded 2) to more than once a week (coded 7).

The other two dependent variables measure more private aspects of religiosity. First, I examine adolescents’ self-reported importance of religion faith in their daily lives. This measure of religious salience gauges the extent to which the respondent has internalized the religious teachings to which they have been exposed. Adolescents were asked, “How important or unimportant is religious faith in shaping how you live your daily life?” Five response categories were given, ranging from not important at all (coded 1) to extremely important (coded 5). I also include a measure that gauges adolescents’ private religious activities. This is a two-item index of frequency of personal prayer and scripture reading. The NSYR asked, “How often, if ever, do you pray by yourself alone?” and “How often, if ever, do you read from [Scripture] to yourself alone?” Seven response categories for each question ranged from never (coded 1) to many times a day (coded 7). The two responses are summed, and this index’s coefficient of reliability is .75.⁴

Independent Variables

There are two key independent variables for this study: parent religiosity and schooling type. The parent religiosity measure is the sum of three responses to questions relating to the parent respondent’s religiosity. Parents were asked about the frequency of their attendance at religious services and the importance of religious faith in shaping their daily lives. Further, adolescent respondents were asked how frequently their family talked about religion in their home. When these three items are summed, the index ranges in value from 0 to 16, with higher values indicating higher religiosity, and the alpha coefficient of reliability is .81.

I include two types of schooling variables in my analysis: dummy variables indicating the type of school attended, and continuous variables that mark the number of years the adolescent has attended that type of school. Parents were asked what type of school their adolescent attended and could choose from a variety of options including public school, private school, and homeschool. Parents answering “private school” were then asked if that school was religious or not. If the school was religious, a follow-up question was posed to determine the type of religious school. The parent could choose between Catholic, Lutheran, Baptist, another type of Christian school, or something else. Since not enough respondents were enrolled in each of these types of schools to allow meaningful analysis, adolescents in Lutheran, Baptist, and other types of Christian schools were grouped into a Protestant school category. NSYR also asked parent respondents if their adolescent had ever attended a religious school. Although they did not ask what type of religious school the adolescent formerly attended, I grouped former religious schoolers with Catholic parents together, former religious schoolers with Protestant parents together, and former religious schoolers with another type of parent together. Then, dummies were created for Catholic-school students (N=122); Protestant-school students (N=71); homeschoolers (N=76); those attending another type of religious school (N=30); former Catholic-school students (N=148); former Protestant-school students (N=248); private, nonreligious-school students (N=60); and those formerly attending another type of religious school (N=40).⁵ In addition to the type of school their adolescent attended, parent respondents were asked how long they had attended that type of school. Using this information, I constructed variables for the number of years the adolescent had attended their type of school to measure the cumulative effect of different schooling strategies on religiosity.

Mediating Variables

In addition to these key independent variables, I also examine factors that might explain or mediate the effect of different schooling types on adolescent religiosity. Students at religious schools might be enmeshed in more religious friendship networks than their public school counterparts. Thus, I include a measure for the proportion of the adolescents’ friends who are religious. These students may also experience a closed intergenerational network, which has been linked to positive adolescent outcomes (Coleman 1988), so I include a measure of the proportion of friends with whom the adolescent has a closed social network.⁶ Lastly, I consider the effect of added religious mentors (i.e., the number of adults adolescents can turn to for support).

Control Variables

I include a number of control variables for both parent and adolescent characteristics that might covary with the dependent and independent variables. Following Steensland et al. (2000), I account for the religious affiliation of the parent respondent. Controls for the parental respondents’ marital status, education, gender, homeownership, assets, income, educational aspirations for their child, and quality of the parent-child relationship are included in each model. I also include controls for the adolescent’s gender, age, race, region of residence, urbanicity, level of autonomy, number of family transitions, number of school transitions, and educational aspirations. For descriptive statistics for all variables, see Table 1.

Table 1.

Means, Ranges, and Standard Deviations of Measures (N = 3,217)

Variables	Mean	SD	Range
Religious service attendance	4.21	2.19	1–7
Religious education class attendance	3.41	2.13	1–7
Youth group attendance	2.75	2.35	1–7
Religious salience	3.46	1.13	1–5
Private religious activities	6.94	3.27	2–14
Parent religiosity	9.73	4.25	0–16
Attends public school	.76	.43	0,1
Attends Catholic school	.04	.19	0,1
Attends Protestant school	.02	.15	0,1
Homeschooled	.02	.15	0,1
Formerly attended religious school, Catholic parent	.05	.21	0,1
Formerly attended religious school, Protestant parent	.07	.25	0,1
Attends private, nonreligious school	.02	.14	0,1
Years at Catholic school	.33	1.80	0–13
Years at Protestant school	.17	1.20	0–13
Years homeschooled	.13	1.02	0–13
Years formerly attended religious school, Catholic parent	.29	1.45	0–13
Years formerly attended religious school, Protestant parent	.30	1.31	0–11
Years at private, nonreligious school	.13	1.04	0–13
Proportion of friends who are religious	.80	.31	0–1
Closed social network	.47	.35	0–1
Number of adults adolescent can turn to for support	5.25	4.48	0–20
Evangelical Protestant parent	.30	.46	0,1
Black Protestant parent	.12	.33	0,1
Mainline Protestant parent	.15	.35	0,1
Catholic parent	.26	.44	0,1
Jewish parent	.02	.13	0,1
Mormon parent	.03	.17	0,1
Parent from other religion	.03	.17	0,1
Parent has no religious affiliation	.06	.24	0,1
Parents are married	.71	.46	0,1
Parents are divorced or separated	.17	.38	0,1
Parent is widowed	.02	.14	0,1
Parent is cohabiting	.04	.20	0,1
Parent is single, never married	.06	.24	0,1
Parents’ education	6.67	2.42	0–10
Quality of parent-child relationship	16.47	2.89	5–22
Parent respondent is female	.81	.39	0,1
Parent owns home	.73	.45	0,1
Parent has assets	.48	.50	0,1
Parent in debt	.22	.41	0,1
Parent breaking even	.28	.45	0,1
Family income	6.68	3.40	1–12
Parent educational aspirations for child	4.49	.74	1–5
Female	.49	.50	0,1
Age	15.48	1.41	12.91–18.49
White	.65	.48	0,1
African-American	.15	.36	0,1
Asian-American	.01	.11	0,1
Hispanic	.10	.30	0,1
Other race	.08	.27	0,1
Lives in the south	.37	.48	0,1
Lives in the northeast	.17	.37	0,1
Lives in the midwest	.22	.41	0,1
Lives in the west	.24	.43	0,1
Lives in an urban area	.26	.44	0,1
Lives in a suburban area	.51	.50	0,1
Lives in a rural area	.22	.41	0,1
Level of autonomy	5.18	2.13	2–10
Number of family transitions	.49	.74	0–3
Number of school transitions	.94	1.67	0–20
Educational aspirations	.20	.40	0,1

Notes: Ns for dependent variables vary slightly. Dummies for other types of religious schools and dummies for missing values not shown.

Methods

In Tables 2 and 3, I present a series of four regression models for each outcome of interest. Odds ratios from ordered logit regression models are displayed for religious service attendance, religious education class attendance, youth group attendance, and religious salience. Coefficients from ordinary least squares (OLS) regression models are shown for the private religious activities outcome. The first models include the parent religiosity variable and controls. The second models add the dichotomous schooling variables to test the effect of schooling after parent religiosity is considered. The third models consider the mediating effects of religious friends, a closed social network, and adult mentors. Finally, the fourth models add measures for the number of years of schooling in an alternative schooling strategy to identify any cumulative effect these strategies might have.

Table 2.

Odds Ratios from Ordered Logit Regression Models Predicting Different Aspects of Adolescent Religious Involvement

	Religious service attendance (N = 3,212)				Religious education class attendance (N = 3,204)				Youth Group Attendance (N = 3,211)
	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4
Parent Effects
Parent religiosity	1.36***	1.36***	1.34***	1.34***	1.23***	1.24***	1.22***	1.22***	1.26***	1.26***	1.24***	1.24***
Schooling Effects
Attends Catholic school		1.52+	1.41	.85		.19***	.17***	.60		.52*	.45**	.16*
Attends Protestant school		1.41	1.31	2.69		1.17	1.09	2.20		1.02	.93	.53
Homeschooled		.94	1.09	.87		.74	.80	.36*		.78	.80	.57
Formerly attended religious school, Catholic parent		.98	1.04	1.06		1.00	1.04	1.38		1.03	1.04	1.68
Formerly attended religious school, Protestant parent		1.00	1.00	1.64		.83	.83	.75		.92	.94	.72
Attends private, nonreligious school		.53*	.53*	.47		.56	.56	.23*		.80	.81	.23**
Years at Catholic school				1.06				.86*				1.12
Years at Protestant school				.91				.91				1.08
Years homeschooled				1.04				1.15				1.05
Years formerly attended religious school, Catholic parent				1.00				.95				.93
Years formerly attended religious school, Protestant parent				1.00				1.02				1.06
Years at private, nonreligious school				1.02				1.14+				1.19**
Mediating Effects
Proportion of friends who are religious			4.33***	4.36***			2.72***	2.71***			3.29***	3.36***
Closed social network			1.26+	1.26+			1.27*	1.25*			1.15	1.14
Number of adults adolescent can turn to for support			1.02**	1.02**			1.01	1.01			1.04**	1.03**
−2 log likelihood	10,366.39	10,351.06a	10,172.02b	10,166.40	10,528.93	10,454.61c	10,355.21b	10,336.10d	6,612.88	6,603.32	6,494.46a	6,480.44b
Pseudo R-square	.14	.14	.15	.15	.10	.10	.11	.11	.13	.13	.15	.15

⁺p < .10

^*p < .05

^**p < .01

^***p < .001

^aSignificantly improved model fit over Model 1 at p < .10

^bSignificantly improved model fit over Model 2 at p < .001

^cSignificantly improved model fit over Model 2 at p < .001

^dSignificantly improved model fit over Model 3 at p < .05

Notes: The suppressed reference category for schooling type is attends public school. All models contain, but do not display, controls for parent’s religious affiliation, marital status, gender, educational aspirations for their child, financial assets, homeownership, income, and educational attainment, as well as adolescents’ age, gender, race, region of residence, urbanicity, autonomy from parents, educational aspirations, number of family transitions, number of school transitions, attendance and former attendance at another type of religious school, and nomination of no friends. Models also contain dummies for missing values on all variables (except schooling type) when relevant. All continuous and ordinal variables are mean-centered except for parent religiosity.

Table 3.

Odds Ratios and Coefficients from Ordered Logit and Ordinary Least Squares (OLS) Regression Models Predicting Religious Salience and Private Religious Activities, Respectively

	Religious salience (N = 3,211)				Private religious activities (N = 3,199)
	Model 1	Model 2	Model 3	Model 4	Model 1	Model 2	Model 3	Model 4
Parent Effects
Parent religiosity	1.28***	1.27***	1.25***	1.25***	.34***	.33***	.29***	.30***
Schooling Effects
Attends Catholic school		1.47+	1.34	1.19		.44	.24	.21
Attends Protestant school		1.96*	1.83*	.61		1.34***	1.23***	.78
Homeschooled		1.09	1.30	1.25		−.02	.14	−.80+
Formerly attended religious school, Catholic parent		.87	.90	.70		−.27	−.26	.46
Formerly attended religious school, Protestant parent		1.16	1.11	.66		−.09	−.09	−.57
Attends private, nonreligious school		1.05	.82	.86		−.50	−.44	−1.61+
Years at Catholic school				1.01				.00
Years at Protestant school				1.16+				.06
Years homeschooled				1.01				.17*
Years formerly attended religious school, Catholic parent				1.04				−.12
Years formerly attended religious school, Protestant parent				1.12+				.11
Years at private, nonreligious school				1.03				.18
Mediating Effects
Proportion of friends who are religious			5.02***	5.00***			1.73***	1.73***
Closed social network			1.07	1.06			.25	.23
Number of adults adolescent can turn to for support			1.03**	1.03**			.06***	.06***
−2 log likelihood (Ordered logit)/Intercept (OLS)	8,035.32	8,019.39a	7,825.06b	7,812.39	3.69***	3.74***	4.08***	4.06***
Pseudo R-square (Ordered logit)/R-square (OLS)	.15	.15	.17	.17	.35	.36c	.40b	.40d

⁺p < .10

^*p < .05

^**p < .01

^***p < .001

^aSignificantly improved model fit over Model 1 at p < .05

^bSignificantly improved model fit over Model 2 at p < .001

^cSignificantly improved model fit over Model 1 at p < .001

^dSignificantly improved model fit over Model 3 at p < .05

Notes: The suppressed reference category for schooling type is attends public school. All models contain, but do not display, controls for parent’s religious affiliation, marital status, gender, educational aspirations for their child, financial assets, homeownership, income, and educational attainment, as well as adolescents’ age, gender, race, region of residence, urbanicity, autonomy from parents, educational aspirations, number of family transitions, number of school transitions, attendance and former attendance at another type of religious school, and nomination of no friends. Models also contain dummies for missing values on all variables (except schooling type) when relevant. All continuous and ordinal variables are mean-centered except for parent religiosity.

Table 4 displays coefficients and regions of significance for interaction effects between each alternative schooling strategy and parent religiosity from OLS regression models that include all main effects from the third models of Tables 2 and 3. Regions of significance for the interactions were calculated using an online program (Preacher, Curran, and Bauer 2004).

Table 4.

Coefficients and Regions of Significance for Interaction Effects between Schooling Type and Parent Religiosity, from Ordinary Least Squares Regression Models, Predicting Adolescent Religious Outcomes

	Religious service attendance	Religious education class attendance	Youth group attendance	Religious salience	Private religious activities
β, Intercept	1.72***	1.61***	1.26***	2.42***	4.02***
β, Parent religiosity	.26***	.20***	.20***	.10***	.30***
β, Attends Catholic school	.61	.15	.62	.11	.80
β, Attends Catholic school * Parent religiosity	−.03	−.15**	−.14**	.00	−.06
Region of significance, α = .05 (% in region)	NS	> 4.84 (87%)	> 7.20 (77%)	NS	NS
Region of significance, α = .10 (% in region)	NS	> 4.38 (87%)	> 6.82 (80%)	NS	NS
β, Attends Protestant school	.74	−1.25	.60	.38	1.14
β, Attends Protestant school * Parent religiosity	−.04	.10	−.04	−.01	.01
Region of significance, α = .05 (% in region)	NS	NS	NS	13.02–14.53 (23%)	> 8.94 (95%)
Region of significance, α = .10 (% in region)	NS	> 14.41 (51%)	NS	11.90–15.38 (76%)	> 7.83 (95%)
β, Homeschooled	−.43	−.39	.24	.31	−1.12*
β, Homeschooled * Parent religiosity	.04	.02	−.04	−.02	.12*
Region of significance, α = .05 (% in region)	NS	NS	NS	NS	< 1.86 (5%)
Region of significance, α = .10 (% in region)	NS	NS	NS	NS	< 2.76 (6%)
β, Formerly attended Catholic school	−.29	.43	.87	−.09	.34
β, Formerly attended Catholic school * Parent religiosity	.03	−.04	−.09	.00	−.06
Region of significance, α = .05 (% in region)	NS	NS	NS	NS	NS
Region of significance, α = .10 (% in region)	NS	NS	NS	NS	NS
β, Formerly attended Protestant school	−.44	−.60	−.26	.05	−.27
β, Formerly attended Protestant school * Parent religiosity	.04	.04	.03	.00	.02
Region of significance, α = .05 (% in region)	NS	NS	NS	NS	NS
Region of significance, α = .10 (% in region)	NS	2.38–10.31 (29%)	NS	NS	NS
β, Attends private, nonreligious school	−.05	−.14	.11	.28	1.76*
β, Attends private, nonreligious school * Parent religiosity	−.05	−.04	−.03	−.03	−.25*
Region(s) of significance, α = .05 (% in region)	7.03–13.69 (48%)	NS	> 11.11 (28%)	NS	< 3.37 (17%) and >12.81 (25%)
Region(s) of significance, α = .10 (% in region)	> 5.78 (70%)	7.27–8.85 (7%)	> 9.47 (48%)	NS	< 4.23 (27%) and > 10.76 (41%)

⁺p < .10

^*p < .05

^**p < .01

^***p < .001

Notes: Statistics derived from OLS models with identical covariates as Models 3 of Tables 2 and 3. All continuous and ordinal variables are mean-centered except for parent religiosity. Parent religiosity ranges from 0–16 and has an overall mean of 9.73. All continuous and ordinal variables are mean-centered except for parent religiosity.

Results

Table 2 displays odds ratios from ordered logit regression models predicting different aspects of adolescent religious involvement. The first series of models displays estimates for religious service attendance. Model 1 reveals that parents play a significant role in the religious involvement of their adolescents. Every unit increase in parent religiosity (which ranges from 0 to 16) results in a 36% increase in the odds that an adolescent will attend religious services more frequently. Given what is already known about the intergenerational transmission of religion (e.g., Myers 1996; Regnerus, Smith, and Smith 2004; Smith and Denton 2005), this is not particularly surprising. Model 1 does not reveal, however, whether these effects are direct or indirect.

Model 2 suggests that the strong effect of parent religiosity is not mediated at all by schooling type, though different schooling strategies may have independent effects on adolescent religiosity. Catholic-school students are more likely to attend religious services more frequently than public-school students (though the difference is only marginally significant), and private, non-religious school students are less likely to attend religious services more often. Protestant schoolers, homeschoolers, and former Catholic and Protestant schoolers do not differ from public school students with respect to their religious service attendance, net of controls and their parent’s religiosity.

The effect of parent religiosity on adolescent religious service attendance does not appear to operate through alternative schooling strategies, and neither does it appear to be mediated by peers, network closure, or mentors. Though each of these is associated with increased odds of more frequent religious service attendance, the effect of parent religiosity is barely altered when these measures are included in Model 3. These mediators eliminate the statistically significant odds ratio for Catholic schoolers, but not for private, non-religious school students. Lastly, Model 4 suggests that there is no significant cumulative effect of schooling type on religious service attendance, net of all other factors.

The results for religious education class attendance are somewhat different. Catholic school students are about 81% less likely to attend religious education classes than their public school counterparts, a difference that remains largely the same when friends’ religiosity, network closure, and mentors are considered. Indeed, there also appears to be a cumulative negative effect of Catholic schooling on religious education class attendance: Each additional year of Catholic schooling is associated with a 14% decrease in the odds of an adolescent reporting more frequent religious education class attendance. Though the average effect of private, non-religious schooling is not significant, each additional year of private schooling is associated with (marginally significant) increased odds of more frequent religious education class attendance. As with religious service attendance, however, parent religiosity is not attenuated much at all by either the schooling effects or the peer religiosity, network closure, and mentor variables.

Table 2 suggests that the effects of parent religiosity, schooling type, and the mediating variables for youth group attendance are similar to that of religious education class attendance. Again, Catholic school students are less likely than public schoolers to attend youth group more frequently—though the difference between the two groups is not as large as the difference in religious education class attendance—but there is no cumulative effect of Catholic schooling on adolescents’ youth group participation. Adolescents attending other types of schools do not differ from public schoolers in their youth group attendance. And as with the previous two outcomes, parent religiosity remains a powerful influence on adolescent youth group participation even after considering schooling type, peer religiousness, network closure, and number of adult mentors.

Table 3 displays odds ratios and coefficients for more private aspects of religiosity, religious salience and private religious activities. With respect to religious salience, a consistently strong, positive effect of parent religiosity is found across all four models, again suggesting that parents exert direct influence on their adolescents’ religious lives. Model 2 reveals that both Catholic and Protestant schoolers are more likely to report higher religious salience than their public school counterparts. The odds that a Catholic schooler will report higher religious salience are 47% greater than those for public schoolers (at p < .10), and the odds of reporting higher religious salience are 96% higher for Protestant schoolers than for public schoolers. The Catholic school effect is reduced to nonsignificance in Model 3 when friends’ religiosity, network closure, and adult mentors are considered. The Protestant school effect, however, is only slightly reduced by these mediators. Even after accounting for differences in friends’ religiousness, network closure, and mentors, Protestant schoolers are 83% more likely to report higher religious salience than public schoolers. There is also marginally significant evidence to suggest a cumulative effect of Protestant schooling on religious salience, and additional years of former Protestant schooling may also lead to increased odds of higher religious salience.

The second half of Table 3 reports coefficients from OLS regression models predicting adolescents’ private religious activities (i.e., private prayer and scripture reading). Protestant schoolers report more private religious activities than public schoolers, net of their parent’s religiosity and other controls. Attending Protestant school is associated with a 1.34 unit increase in private religious activities. As with religious salience, this increase is only slightly attenuated by the mediating variables. There is also an evident cumulative effect of homeschooling on private religious activities. Each additional year of homeschooling results in a .17 unit increase in private religious activities. Also, consistent with the other outcomes, parent religiosity is positively associated with more private religious activities and is not mediated much at all by either the schooling variables or friends’ religiosity, network closure, and adult mentors.

Table 4 displays coefficients from OLS models that are parallel to Models 3 in Tables 2 and 3 but include multiplicative interaction terms between each schooling type and parent religiosity. For Catholic schooling, significant interaction effects are found for religious education class attendance and youth group attendance. Although the effect of Catholic schooling on religious education class in not significant when parent religiosity is zero, each additional unit increase in parent religiosity is associated with a .15 unit decrease in religious education class attendance. In other words, the negative effect of Catholic schooling on religious education class attendance is strongest for Catholic school students with the most religious parents. A similar interaction is found for youth group attendance: Catholic schooling’s negative effect is strongest among Catholic schoolers with more religious parents.

There is some evidence that homeschooling effects are contingent on parent religiosity, but only with respect to private religious activities. Homeschoolers whose parents are not religious at all report less frequent participation in private religious activities than their public school counterparts (β=−1.12), but each unit increase in the religiosity of homeschoolers’ parents is associated with a .12 unit increase in private religious activities. Nevertheless, homeschoolers with extremely religious parents are not statistically different on this outcome than public schoolers with extremely religious parents. No other significant interaction effects were found between homeschoolers and their parents’ religiosity.

The opposite effect on private religious activities is found for private, nonreligious school students. Those with nonreligious parents report higher private religious activity, but increases in parent religiosity are associated with .25 unit decreases in this effect. Thus, private school students with nonreligious parents report higher private religious activities than their public school counterparts, but those with more religious parents (i.e., those with a religiosity score of 12.81 or higher) report lower levels of private religious activities. No significant interaction effects were found among Protestant schoolers, former Catholic schoolers, or former Protestant schoolers.

Discussion

So do alternative schooling strategies actually influence the religious lives of adolescents? And do friendship networks, network closure, and adult mentors explain this influence? These data suggest the answers to these questions depend on the type of school parents choose and the types of religiosity they wish to foster.

Given the demographic changes in Catholic schools and their shifting emphasis towards academic excellence, one might expect that the religious influence of contemporary Catholic schools is minimal. Nevertheless, the evidence presented here suggests that Catholic schools may encourage modestly higher levels of church attendance and religious salience among adolescents. These differences are only marginally significant, however, and are explained at least in part by friends’ religiosity, network closure, and adult mentors. Moreover, adolescents who formerly attended Catholic school are neither more nor less religious than public schoolers. Still, Catholic schools may help foster some forms of religious commitment among their students. At the same time, Catholic school students—especially those with religious parents—attend religious education classes and youth group less frequently than public schoolers. Although this could be interpreted as a negative religious influence, it is more likely that Catholic school students are substituting their attendance at Catholic secondary schools (and their concomitant enrollment in religion classes there) for these other religious activities. Taken together, these findings suggest that Catholic schools may contribute modestly to their students’ religious faith and practice.

Protestant schools appear to contribute more substantially to their students’ religious lives than do Catholic schools. Protestant schoolers are just as likely as public schoolers to attend religious services, religious education classes, and youth group. So while Catholic schoolers may be relying heavily on their Catholic school for their religious education, Protestant schoolers may reap the religious benefits of involvement in both their religious school community (and its Bible classes, chapels, etc.) and their local congregation (and its youth programming). These religious benefits are evident in Protestant schoolers’ private religiosity. Even after accounting for the religiousness of their friendship networks, the closure of their social network, and their available adult mentors, Protestant schoolers report higher levels of both religious salience and private religious activities. These findings suggest a change in the influence of Protestant schools since earlier studies (e.g., Erickson 1964; Johnstone 1966) as Protestant schools have become more numerous and more established. The data also point to a cumulative effect of Protestant schooling on religious salience, indicating that increased exposure to a Protestant schooling strategy may result in heightened private religiosity.

How and why Protestant schools matter for private religiosity is not easy to explain. Accounting for religious friends, adult mentors, and closed social networks does very little to explain Protestant schoolers’ heightened private religiosity. Recall, however, that schoolmates’ religiosity may have a more salient impact on adolescent religiosity than friends’ religiousness (Regnerus, Smith, and Smith 2004). I suggest that immersion in a religious culture can have powerful effects on adolescents, as has been explored elsewhere (e.g., Stark 1996). Protestant schoolers are surrounded by a community of religious peers and adults who place a high premium on religious faith and practice and who encourage religious and spiritual development in students. This religious community serves as a plausibility structure that helps to sustain religious commitment (Berger 1967; Smith 1998). Furthermore, Protestant-school students receive explicit religious instruction in school in the form of Bible classes and chapel services that may increase private religiosity (Benson, Donahue, and Erickson 1989); and although Catholic schools also provide this instruction, it is more valued among Protestant school administrators (Baker, Han, and Broughman 1996). This premium on religious development may be a key factor that distinguishes Catholic and Protestant schools and their students.

Nevertheless, potential enthusiasm regarding the increased private religiosity of Protestant schoolers should be tempered by the finding that those who formerly attended Protestant schools are no different than other public schoolers. There are several potential explanations for this. Most straightforward, perhaps, is the possibility that Protestant schooling does not exert any lasting impact on its students once they leave. If this is the case, the religious community of Protestant schools, rather than any religious education per se, is likely the driving factor behind the Protestant school-adolescent religiosity relationship. A second explanation is that Protestant school students who are less religious may select out of Protestant schooling. That is, those who are cold or tepid toward the religious emphasis of Protestant schools may convince their parents to remove them from that school and place them in public school. In reality, both of these processes are probably at work, but given limited data, they are difficult to parse out. Of course, there other possible explanations as well, though they are less compelling. For example, there may be a minimal threshold of Protestant schooling required to for religious influence, or Protestant schooling may be influential during secondary school but not during elementary and middle school.

Surprisingly, I find very little effect of homeschooling on any aspect of adolescents’ religious lives. Despite the fact that a significant minority of these students are homeschooled for explicitly religious reasons (Bielick, Chandler, and Broughman 2001), there appear to be scant religious benefits to this schooling strategy. On one hand, this may speak to the great diversity of motivations for homeschooling. There are a large number of nonreligious (or at least not extremely religious) homeschool parents, and religious benefits should not be expected from that type of homeschooling situation. But even when I tested for interaction effects between homeschooling and parent religiosity, no positive effects of homeschooling were apparent. These findings, together with the findings for Catholic-school and Protestant-school adolescents, highlight the importance of religious community for cultivating and maintaining adolescent religiosity. Furthermore, if the religiosity of homeschoolers, net of confounding factors, can be interpreted as the religiosity of adolescents absent of any schooling context (which it may not be), this study’s findings could suggest that public schooling is neither detrimental to nor beneficial for adolescent religiosity.

Though these schools were not the focus of this study, private, nonreligious schools tend to be associated with decreased adolescent religiosity, especially among private schoolers with religious parents (vis-à-vis public schoolers with religious parents). These adolescents may be more likely to encounter intellectual cultures, expectations, or ideas that undermine religious commitment (e.g., Sherkat 1998).

Although different schooling strategies influence adolescents’ religious lives in different ways, the overall contribution of alternative schooling strategies to adolescent religiosity is quite modest. The inclusion of schooling type variables does significantly improve model fit for all of the outcomes except youth group attendance, but the size of that improvement is small. Indeed, the pseudo R-square does not change for any of the outcomes between Models 1 and 2, and the R-square for private religious activities increases by only .01. Table 1 shows that only eight percent of adolescents are in Catholic schools, Protestant schools, or are homeschooled, so the potential overall explanatory power of these variables is rather limited from the start.

The findings here also shed light on the role parents play in transmitting their religion to their adolescents. Parents retain strong influence on their adolescents’ religiosity, even after accounting for adolescents’ schooling situation, friends’ religiosity, network closure, and extra-familial adult mentors. In fact, none of these factors seems to attenuate the role of parents at all. Thus, I find little support that parents are indirectly influencing their adolescents by “channeling” them into religious schooling environments or networks of religious friends and adults. Rather, parents directly influence their adolescents. Social learning, spiritual modeling, and spiritual capital explanations may all explain this direct influence, but this study has not directly tested these theories.

Several limitations of this study must be noted. Observational studies such as this one can never fully deal with the issue of selection. Although I have controlled for a number of variables that might select adolescents into different schooling types, I cannot fully discount the persistence of selection effects. Nevertheless, propensity score models using nearest-neighbor matching techniques—which reduces selection on the observables (Winship and Morgan 1999)—produced similar estimates to those provided here.

The cross-sectional data employed here are also not without their limitations. Of primary concern is the causal ordering of the relationships I have presented. I cannot reject the hypothesis that more religious adolescents talk their parents into enrolling them into religious schools, for instance, though some evidence suggests that parents are firmly in control of their adolescents’ schooling decisions (Irvine 2002). Moreover, it is certainly possible and probable that religious adolescents choose more religious friends, and this is driving part of the relationship between friends’ religiosity and adolescent religiosity. If the causal influence is moving in this direction and not the other, this could explain why the mediating effects of this variable are so limited. The causality here is likely bi-directional, however.

Though this study has sought to answer several important questions regarding alternative schooling strategies and the religious lives of adolescents, there remain several promising avenues for future research in this area. Future research on alternative schooling strategies should analyze the long-term effect of these approaches, and the effects of these strategies during different stages of childhood need to be evaluated. Additionally, the measures in NSYR can account for neither the great diversity in Protestant schooling nor the different motivations for homeschooling. Protestant schools are united in their commitment to adolescent religious and spiritual development, but their varying methods and philosophies may produce dissimilar results. A closer investigation into this matter is merited. With respect to homeschoolers, the NSYR did not ask parents why they choose this schooling method. An analysis of adolescents homeschooled for religious reasons may reveal differences between these students and public schoolers. Moreover, schooling effects may be contingent on a number of factors that are not explored here, most notably region and urbanicity. Finally, this study uses broadly applicable measures of religiosity in order to contribute to a general understanding of adolescent religiosity. Researchers interested, for instance, in the “Catholic-ness” of parochial school students compared to other Catholics may wish to incorporate different measures.