<<

ABSTRACT

The Effects of on Participation in Drinking Alcohol among Adolescents and Emerging

Robert A. Thomson, Jr., M.A.

Thesis Chairperson: Sung Joon Jang, Ph.D.

Parents often choose to homeschool their children for reasons that, if their goals are accomplished, should contribute to relatively low levels of substance use among their children. Specifically, many parents desire to foster family- and religiously-centered values, while they are also concerned about the potential of negative peer pressure in public . In this paper, I use data from three waves of the National Study of and Religion (NSYR) to study the relationship between homeschooling and alcohol use.

As hypothesized, I find that homeschool students do tend to drink less than public and private students. Religiosity was found to be the strongest explanation of the observed group difference in cross-sectional analyses, while prior drinking was the dominant predictor in longitudinal analyses. In addition, concepts related to social bonding theory partially explain group differences, whereas those related to social learning and general strain tended to suppress them.

Copyright © 2013 by Robert A. Thomson, Jr.

All rights reserved

TABLE OF CONTENTS

I. List of Figures v

II. List of Tables vi

III. Acknowledgments viii

III. Chapter One: Introduction 1

IV. Chapter Two: Literature Review 3 Homeschooling and Delinquency The Modern Rise of American Homeschooling Homeschooling, Parents/Family, and Conventional Adults Homeschooling and Friendship Networks Religion and Homeschooling Hypotheses

V. Chapter Three: Data and Methods 17 Dependent Variables Secondary School History Independent Variables Moderating Variables Controls Analytical Strategy

VI. Chapter Four: Results 28

VII. Chapter Five: Discussion 48

VIII. Chapter Six: Conclusion 53

IX. Appendix A 59

X. Appendix B 62

XI. Bibliography 77

iv

LIST OF FIGURES

Figure 1: Predicted Probability of Drinking at Time 3 (Cross-Sectional) 33

Figure B.1: Predicted Probability of Drinking at Time 3 (3-Wave Longitudinal) 64

v

LIST OF TABLES

Table 1: Descriptive Statistics by Wave: NSYR 38

Table 2: Group Means of Dependent Variables and Covariates 39

Table 3: Binary Logistic Regression Odds Ratios for Drinking at Time 1 41

Table 4: Binary Logistic Regression Odds Ratios for Drinking at Time 2 43

Table 5: Binary Logistic Regression Odds Ratios for Drinking at Time 3 44

Table 6: Binary Logistic Regression Odds Ratios for Drinking at Time 2 (2-Wave Longitudinal) 45

Table 7: Binary Logistic Regression Odds Ratios for Drinking at Time 3 (2-Wave Longitudinal) 46

Table B.1: Binary Logistic Regression Odds Ratios for Drinking at Time 3 (3-Wave Longitudinal) 63

Table B.2: Binary Logistic Regression Odds Ratios for Getting Drunk at Time 1 65

Table B.3: Binary Logistic Regression Odds Ratios for Getting Drunk at Time 2 66

Table B.4: Binary Logistic Regression Odds Ratios for Getting Drunk at Time 3 67

Table B.5: Binary Logistic Regression Odds Ratios for Getting Drunk at Time 2 (2- Wave Longitudinal) 68

Table B.6: Binary Logistic Regression Odds Ratios for Getting Drunk at Time 3 (2- Wave Longitudinal) 69

Table B.7: Binary Logistic Regression Odds Ratios for Getting Drunk at Time 3 (3- Wave Longitudinal) 71

Table B.8: Binary Logistic Regression Odds Ratios for DUI at Time 2 72

Table B.9: Binary Logistic Regression Odds Ratios for DUI at Time 2 (2-Wave Longitudinal) 73

Table B.10: Binary Logistic Regression Odds Ratios for Binge Drinking at Time 3 74

vi

Table B.11: Binary Logistic Regression Odds Ratios for Binge Drinking at Time 3 (2- Wave Longitudinal) 75

Table B.12: Binary Logistic Regression Odds Ratios for Binge Drinking at Time 3 (3- Wave Longitudinal) 76

vii

ACKNOWLEDGMENTS

I would like to thank my thesis chairperson, Dr. Sung Joon Jang, for his tireless and patient mentorship throughout the process. I would also like to thank my committee members, Dr. Paul Froese and Dr. Jon Singletary for their assistance and service.

viii

CHAPTER ONE

Introduction

The family and school are widely recognized as critical arenas of in criminological theory, due to their importance in the development of both social capital and cultural resources in (Agnew, 2006; Akers, 2009; Gottfredson & Hirschi,

1990; Hirschi, 1969). The practice of homeschooling, a social structure that intersects family and , thus raises interesting questions to both sociology and .

Are parents who homeschool in order to instill family and religious values successful in doing so? Are teens who have been homeschooled less likely to offend? To be sure, some parents might be motivated to homeschool by a desire to provide a superior education through individualized attention that can result in better economic opportunities for their children, but many others are motivated by deep convictions regarding the preeminence of family and religion over secularized societal institutions

(c.f. Cizek, 1994; Collom, 2005; Jeub, 1994; Lines, 1994; Murphy, 2012).

While the homeschooling movement is relatively small, currently representing less than four percent of school-aged children in the United States, it has gained legal and popular legitimacy over the last few decades. Despite a growth rate that “has been nothing short of remarkable” since the 1970’s (Murphy, 2012, p. 10), however, research on homeschooling has been limited not only in number but also in scope as it has focused on pro-social outcomes, such as academic achievement and socialization of homeschoolers (see Cizek, 1994; e.g., Cordner, 2012; Ray, 2004; Uecker, 2008). Among

1

many gaps that remain (see Murphy, 2012, pp. 125–126), grossly lacking is criminological research on delinquency and substance use among homeschool students.

The present study intends to fill this gap in research and thus contribute not only to the homeschooling but also criminological literatures.

Specifically, I hypothesize that adolescents and emerging adults who have been homeschooled will participate in drinking alcohol less than those who attended public and private (religious or otherwise) high schools, and that the negative association between homeschooling and substance use is explained by variables related to , social learning, and general strain, as well as independent effects related to religion. Furthermore, I predict interaction effects between homeschooling and both parent and teen religiosity, such that participation in drinking alcohol declines more with increasing religiosity among homeschooled teens than non-homeschooled teens. These hypotheses will be tested using data from the National Study of Youth and Religion

(NSYR), a nationally representative three-wave panel study conducted from 2002 to 2008 that included both parent (wave 1 only) and teen questionnaires.

2

CHAPTER TWO

Literature Review

Homeschooling and Delinquency

When a family decides to homeschool rather than send their children to public or private school, they fundamentally alter the structure in which the children relate not only to their parents (and potentially other adults) but also their peers. Homeschooling thus has potential implications for three micro-level criminological theories: (1) , which posits that individuals are restrained from and delinquency by virtue of their attachment to conventional others, commitment to conventional goals, involvement in conventional activities, and pro-social beliefs (Hirschi, 1969); (2) social learning theory, which proposes that criminal behavior is both learned by imitating or associating with deviant others and conditioned through reward/ schedules

(Akers, Krohn, Lanza-Kaduce, & Radosevich, 1979; Akers, 1985, 2009; Sutherland &

Cressey, 1970); and (3) general strain theory, which conceptualizes crime as an illegal behavioral or emotional coping response to negative emotions caused either by the experience of negatively-valued stimuli or by the loss of positively-valued stimuli

(Agnew, 1992, 2006). It should be noted from the onset that while each of these distinct theories are expected to have some application to the homeschooling movement, the present work does not represent a theoretical integration, in that no attempt to reconcile the logical incompatibility of underlying assumptions regarding motivations and human proclivity toward criminal behavior will be made (see Thornberry, 1989).

3

As will be discussed below, religion has not only played a prominent role in the historical emergence of homeschooling, but it also serves as an important motivation for homeschooling today. Therefore, the impact of religion on criminality is also relevant to the homeschooling movement. Of course, the effect of religion on crime has been a matter of debate since Hirschi and Stark’s (1969) report that, while both moral values and attitudes regarding the law were highly predictive of delinquency, church attendance was not a strong predictor of either, at least not among their sample of students from Western

Contra Costa County, California. These findings have since stirred debate both about research methodology (e.g.,Higgins & Albrecht, 1977; Johnson, Jang, Larson, & De Li,

2001) as well as theoretical framing. As an example of the latter, Stark (1996) himself eventually developed a “moral communities” thesis, later elaborated by Regnerus (2003), which understands the effects of religion in social, rather than purely , terms.

The deterrent effects of religiosity, he explains, needs to be activated by social contexts where religion is influential in constructing group norms.

All told, the preventative effect of religion on crime and delinquency has been well-demonstrated. In a meta-analysis of 60 criminological studies, Baier and Wright

(2001) found that religion has a moderate but significant effect on crime, and a systematic review of 270 studies by Johnson and Jang (2011) reveals that about 90 percent of studies that have included relevant variables reported pro-social effects of religiosity on crime and delinquency. Of particular note, Smith and Denton (2005, pp.

232–233) report that religion consistently influences life outcomes for teenagers

(especially the highly devoted) over a broad spectrum of attitudes and behaviors, including those relevant to crime and delinquency (e.g., risky behavior, emotional well-

4

being, quality of family relations, moral reasoning, religious participation, etc.), even despite the teens’ inability to recognize or articulate differences attributable to religiosity.

Still, some researchers have suggested that the effects of religion on crime are spurious, fully explainable by, for example, concepts related to social bonding, self- control, or arousal (e.g., Cochran, Wood, & Arneklev, 1994; Ellis, 1987). And to be sure, it can be argued that religion is (at least partially) exogenous to the variables related to the established micro-level theories of crime, even though criminological theorists have tended to neglect religion. For example, religious individuals might be expected to commit less crime as a result of (1) increased bonding to conventional others, (2) increased self-control through the internalization of religious norms and encouragement of self-monitoring through religious socialization, (3) increased association with more conventional peers and decreased association with delinquent peers, and (4) reinforcement for conventional coping with and amelioration of negative emotions resulting from strain (Jang & Johnson, 2003; Johnson & Jang, 2011, p. 122; Ulmer,

2012).

On the other hand, some evidence suggests that the effect of religion is not merely indirect, let alone spurious. To the contrary, Longest and Vaisey (2008) have demonstrated that the effects of social bonding variables become non-significant when multidimensional concepts of religiosity are properly taken into account, rather than vice versa. Such a finding is consistent with many other studies that have revealed unexplained “direct” effects of religion on crime and delinquency, even when other established criminological variables are included as controls (Johnson & Jang, 2011, p.

123). The mechanisms by which religion directly reduces criminality might include

5

specifically religious or spiritual (as opposed to merely conventional) coping strategies, as well as doctrines of forgiveness by which vengeance is discouraged (Johnson & Jang,

2011). Francis T. Cullen (2011, p. 155) also suggests that religion increases social support, and that faith in a loving God might reduce punitive sentiments. All said, the prevalence of both indirect and direct effects of religion on criminality is expected, according to Jeffrey T. Ulmer (2012), because of the holistic nature of religion, as it tends to influence nearly every dimension of an individual’s social world. Given the pervasiveness of religious households among homeschool families (Murphy, 2012; Ray,

2004), such direct and indirect effects are expected in this study as well.

In the sections that follow, I present a brief summary of the historical context in which the modern homeschooling movement has arisen, followed by more detailed reviews of how particular motivations for homeschooling related to the family, friendship networks, and religion suggest an expectation that homeschooling should result in lower levels of delinquency. Finally, I present four hypotheses that derive from this discussion and will be tested herein.

The Modern Rise of American Homeschooling

Homeschooling emerged as a countercultural movement that was “largely a reaction against the mass of the modern liberal state” (Gaither, 2008, p. 85), and it arose from two fronts, one pedagogical and largely secular, the other ideological and mostly religious (see also Aurini & Davies, 2005; Basham, Merrifield, & Hepburn, 2007;

Gaither, 2009; Lyman, 1998). These unlikely allies were motivated to seek educational alternatives for reasons ranging from alienation in public schools due to traditional values of the families to concerns regarding dehumanization resulting from the bureaucratic

6

organization of the public school system. Today, the decision to homeschool is often complex; while most families who homeschool are religious, and typically conservative

Protestant (Ray, 2004), they decide to homeschool for a variety of academic, familial, religious, and social reasons (Jeub, 1994). But if there is one underlying theme, the decision to homeschool is generally about parents reclaiming control (Murphy, 2012, pp.

76–77). In a context shaped by the intertwining and sometimes competing interests of parents, children, and the state (Kunzman, 2012), homeschooling might be interpreted as an expression of parents trying to exert power over the educational system and provide their children with both traditional values and relevant preparation for emerging economic patterns and workplace demands (Collom, 2005; Lines, 1994).

As social movements go, homeschooling is relatively small, but it is also relatively new. When the contemporary homeschooling movement first emerged in the early 1970’s, only about 10,000 to 15,000 children were homeschooled in the United

States, but it quickly burgeoned to between 120,000 and 240,000 children by the mid-

1980’s (Gaither, 2009, p. 341; Murphy, 2012, pp. 7–8) during a period in which it began to gain social and legal legitimacy through a series of court battles (Cooper & Sureau,

2007). And it continues to grow. By 2003, surveys were estimating that the number of homeschooled children had eclipsed one million (Dan Princiotta, Bielick, & Chapman,

2004), and today, over two million school-age students are estimated to be homeschooling (Murphy, 2012, p. 10). Though this represents less than four percent of the total school-age population in the US, growth rates are estimated to be between 7 and

20 percent per year (Basham et al., 2007; Bauman, 2001; Murphy, 2012, p. 10). Statistics regarding both the size and growth rate are prone to variation because, generally

7

speaking, data on home education is difficult to collect. Procedures for tracking homeschool enrollment vary by state, and due to transience between school types, homeschoolers are likely to be underreported in official statistics (Aurini & Davies, 2005, pp. 462–463; Kunzman, 2010, p. 19; Murphy, 2012, p. 12).

Homeschooling, Parents/Family, and Conventional Adults

Certainly, the costs of homeschooling can be quite high, as it typically requires extensive investment of time and resources on behalf of the parents and usually the forgoing of at least one potential income (Barfield, 2002, p. 184; Lubienski, 2000, p.

209), but families often do not make school choices based on cold, rational, or economic evaluations of value. Rather, school choice is quite personal and sentimental, leading

Davies and Quirke (2005, p. 544) to conclude that it represents “a unique species of consumerism.” One of the driving factors in the origin of contemporary homeschooling, according to Milton Gaither (2008, p. 113), was the “American cult of the ,” experienced by progressives and conservatives alike, whose proponents sought to

“liberate the kids from what they took to be the deadening effects of institutionalization by keeping them at home.” Today, the primacy of the family as a key societal institution is a central tenant for most homeschool parents (Jeub, 1994; Murphy, 2012, p. 103).

Indeed, the top reason given for homeschooling by parents responding to the 1999

National Household Education Survey (NHES) was a belief that they could provide a better education in the home (Bielick, Chandler, & Broughman, 2001, p. 10).

Additionally, unspecified “family reasons” was the fourth most common response.

Homeschooling, according to proponents, can facilitate increased closeness between children and their parents and (e.g. Ballmann, 1987, p. 82; also see

8

Barfield, 2002, p. 21). To the extent that it does, pro-social consequences should result, as each of the criminological theories considered here recognizes the importance of the family in preventing delinquency. According to Hirschi’s (1969) social bonding theory, for example, children are less likely to be delinquent by virtue of attachment to, as well as increased involvement with, conventional parents. Similarly, social learning theory highlights the importance of the family, which “usually serves as a conventional socializer against delinquency and crime” (Akers, 2009, pp. 113–114). Conventional parents and siblings facilitate the learning of conventional definitions and development of self-control, and they also provide conforming models and differential reinforcement favoring non-delinquent behavior. Furthermore, close attachment to parents should influence children’s behavior even when they are not in their parent’s direct presence due to what Hirschi (1969, p. 89) called “virtual supervision.” If homeschooling does promote family cohesion, it should therefore lead to decreased adolescent delinquency.

A common concern for homeschooling is whether students are prone to becoming socially isolated (Kunzman, 2009b, p. 98; Medlin, 2000), but to the contrary, proponents argue that homeschooling has a number of socializing advantages over other school settings, and some of these advantages may also foster decreased delinquency. For instance, teens who homeschool are often involved in a number of extracurricular activities (Medlin, 2000, pp. 110–111), including activities with their parents (Barfield,

2002, pp. 46–47; Kunzman, 2009b, p. 18). A characteristic of homeschooling that is especially consistent with the alternative teaching philosophies of the progressive

“unschoolers” is the use of more “hands-on” learning experiences outside the home, as opposed to formal teaching sessions in a learning environment that is functionally

9

disconnected from the subject of instruction (Gaither, 2008, pp. 125–126; Lyman, 1998, p. 3). Based on social control theory, teens who are homeschooled should thus be less delinquent because they are involved in a greater amount of conventional activities

(Hirschi, 1969, p. 22), especially activities with their parents.

Furthermore, homeschooling proponents suggest that the age segregation of the public school system is inferior to the age integration that is possible in home education

(Kunzman, 2009b, p. 98; Medlin, 2000, p. 119). Homeschooled children learn in family environments with siblings, and sometimes other homeschooled children in multi-family co-ops (Barfield, 2002, pp. 9–10; Kunzman, 2009b, p. 19), and they also tend to have regular encounters with adults outside the home (Medlin, 2000, p. 112). To the degree that homeschooling results in increased attachment to conventional others, it should also decrease delinquency.

In sum, because of the emphasis on the role of the family among homeschooling parents, teens who are homeschooled are expected to be less delinquent because they are closer to their parents, parents are involved in more activities with their teens, and teens are expected to have more support aside from their parents.

Homeschooling and Friendship Networks

Sutherland’s original formulation of theory proposed that criminal behavior, like all behavior, is learned in interaction with intimate personal groups whose members provide “definitions” either favorable or unfavorable to violating the law to an individual, as well as both techniques and motivations for committing criminal acts (Bernard, Snipes, & Gerould, 2010, p. 180; Sutherland & Cressey, 1970, pp.

75–76). Both the ratio of conventional to deviant peers and the strength of friendships

10

with these peers help to form an individual’s orientation to the law. Ronald L. Akers

(2009, pp. 163–164) suggests that association with deviant peers is, next to an individual’s own prior delinquency, “the single best predictor of the onset, continuance, or desistance of delinquency” not only because illicit behavior is often conducted (and typically first becomes available) in peer-group contexts, but also because interactions with deviant peers serve to normalize deviant behavior. And notably, in fact, differential association measures have been shown to be a powerful determinant of adolescent drug and alcohol use (Akers et al., 1979).

An implicit desire to limit exposure of teens to sex, drugs, violence, and negative peer pressure appears to underlie the motivation of many parents who decide to homeschool. According to proponents, educating children in the home allows for better control of social interactions, especially those that involve contact with deviant peers

(Ballmann, 1987, pp. 71–74; Jeub, 1994). And, in fact, the third most common reason for homeschooling on the 1999 National Household Education Survey (NHES) was a “poor learning environment at school” (Bielick et al., 2001, p. 10). By the 2003 NHES, concerns about the environment of other schools, including issues regarding safety, drugs, or negative peer pressure, was the most commonly cited motivation for homeschooling (Daniel Princiotta, Bielick, & Chapman, 2006, p. 14). More than 85 percent mentioned it as a reason for homeschooling, and for nearly a third of all respondents, it was the single most important reason.

The decision to homeschool for these parents is thus an intentional effort to alter the differential association of their children, presumably in favor of anti-criminal peers.

It is therefore expected that children who homeschool will have fewer deviant close

11

friends and, especially if they are being homeschooled for religious reasons, more religious and fewer non-religious friends. This differential association skewed towards religious and conventional peers is consequently expected to decrease a teen’s deviant behavior. Furthermore, by limiting the teen’s exposure to deviant peers, they might also be exposed to less bullying, which is a type of strain that can lead to criminal behavior

(Agnew, 2006).

Religion and Homeschooling

Religion has played an important role in the emergence and growth of the contemporary homeschooling movement. Among its most important pioneers, for example, are Dr. Raymond and Dorothy Moore, Seventh Day Adventists who helped to rally support from the Christian Right through their message about the dangers of early school enrollment and the importance of the family’s role in educating children (Gaither,

2008, pp. 128–134, 2009, p. 338). In contrast with the countercultural left, for whom homeschooling served to provide a more “natural” environment to educate children

(Gaither, 2008, p. 100), those on the Religious Right tend to see homeschooling as a solution to perceived embattlement and a desire for autonomy. Families might homeschool their children for religious rather than strictly pedagogical or developmental reasons (1) if they feel their worldview, including beliefs, values, and morality, are threatened by the public school system, and/or (2) to express their authority, against that of the state, to shape the education of their children.

Ideologically, many conservative Christians became disaffected with the public education system in the wake of the 1925 Scopes “Monkey” Trial over the teaching of

Darwinian evolution (Marsden, 1980, pp. 184–195). Fundamentalism emerged as a

12

separatist movement in the trial’s aftermath because anti-evolution politics lost the support of moderates, and a divide between inclusivists and premillennial exclusivists widened. Homeschooling, according to Robert Kunzman (2010), is an ideal setting in which core principles of fundamentalism can be reinforced because it provides an environment in which parents can resist secular culture, instill particular family values, and endorse a cosmology more in line with a strict literalism of biblical creation accounts. It also caters to the fundamentalist suspicions of government institutions.

A decision to homeschooling is more complicated for evangelicals, however.

While similarly experiencing hostility over the teaching of evolution and sex education, as well as the 1962 and 1963 Supreme Court decisions outlawing prayer and school- sponsored Bible readings in public schools (Gaither, 2008, p. 107, 2009, pp. 337–338), they do not share the fundamentalist compulsion to separate from . On the contrary, they generally experience an “evangelical burden” to engage the world in order to transform it, and this might include keeping their children enrolled in public schools as a witness to an alternative way of life (Smith, Emerson, Gallagher, Kennedy, & Sikkink,

1998, p. 136). Conservative Christian response due to perceived embattlement in the public schools is, then, by no means uniform, varying at least partially in response to particular religious histories and traditions (Sikkink, 1999).

That said, religion remains an important motivator for homeschooling, especially for parents who feel a moral imperative to nurture their children’s moral and spiritual development (Ballmann, 1987, pp. 77–79; Cizek, 1994, pp. 45–46; Kunzman, 2009a, p.

315, 2009b, p. 75; Murphy, 2012, p. 88). According to the 1999 NHES, religion and the development of character or morality were two of the top five reasons cited by parents

13

who homeschool (Bielick et al., 2001, p. 10). On the 2003 NHES, nearly 72.3 percent cited their ability to provide religious or moral instruction as a reason for homeschooling, including nearly a third of all respondents that said it was their top reason, which is second only to concerns about school environment (Daniel Princiotta et al., 2006, p. 14).

While there is evidence that religiously-based motivations are shifting (Murphy, 2012, p.

79) or becoming intertwined with other reasons, conservative Christians who strongly emphasize biblical orthodoxy represent the majority of homeschool families, and a growing proportion observe Buddhism, Judaism, Mormonism, Islam, and New Age faiths as well (Gaither, 2009, p. 342; Murphy, 2012, p. 23; Ray, 2004, p. 6).

Prior research has demonstrated that teen outcomes can depend, to some degree, on parental characteristics (Collom, 2005). For parents who are religiously motivated, it might be expected that teenagers who are homeschooled will be less deviant through several mechanisms related to their religiosity. For instance, it is possible that teens who are homeschooled by religious parents will be more religious themselves (e.g. Kunzman,

2009b, p. 34). This, however, has not been demonstrated convincingly, as Jeremy E.

Uecker (2008) has found only very limited evidence in a cross-sectional analysis. In fact, parent religiosity was only found to interact with homeschooling in predicting increased participation in private religious activities such as prayer and scripture reading. Parent religiosity did not significantly predict differential attendance at religious services, religious education classes, or youth group, and notably no differential religious salience among homeschoolers as compared to public school students was observed. That said, it is possible that parent religiosity could moderate other predictor variables related to social bonding, self-control, and differential association. Further, because of the

14

hypothesized increases in family cohesion and conventional peers, religiosity may be more persistent among homeschoolers as they mature into emerging adulthood.

Other mechanisms by which homeschooling could limit relate to differential association and self-control theories. First, it is likely that homeschooled teens will have more religious and religiously-involved friends and fewer non-religious friends, particularly if the parents are religious, because of parents’ increased ability to monitor their teens’ social network. For similar reasons, religiously homeschooled teens should also have more adult religious support. Both religious peers and religious adults would be expected to encourage non-delinquent behavior by reinforcing conventional definitions and behaviors and modeling morally virtuous behavior (Johnson & Jang,

2011, p. 122). This effect should be moderated, however, by the teen’s own religiosity, since Longest and Vaisey (2008) have demonstrated that content, rather than mere contact, with religious networks matters in suppressing delinquency. Second, to the degree that religious homeschooling fosters teen religiosity, homeschooling should facilitate the development of self-control due to increased internalization of behavioral standards (Johnson & Jang, 2011, p. 122; Ulmer, 2012, p. 167). If this is so, teens who are homeschooled by religious parents should also have stronger convictions regarding moral absolutism.

Hypotheses

To summarize, I expect homeschooling to decrease a student’s use of substance, including alcohol, because it provides a differential social organization in which he or she is more likely to be related to conventional others (e.g., parents and religious friends)

15

than unconventional ones (e.g., delinquent friends). The following baseline hypothesis thus follows:

H1: Homeschooling is negatively associated with participation in drinking alcoholic beverages during and emerging adulthood.

The negative association is expected to be explained by variables of social bonding and social learning theories of delinquency as well as religiosity. The following are thus to be tested:

H2: The negative association between homeschooling and drinking alcohol is explained by differences between homeschooled and non-homeschooled individuals in (a) attachment and involvement with parents and conventional adults, (b) commitment to conventional goals and conventional beliefs, (c) association with nonreligious and deviant peers, and (d) religiosity, religious support, and religious friends.

Finally, I also tested whether the influence of homeschooling on an individual’s alcohol use is conditioned by religiosity:

H3: The negative association between homeschooling and drinking is strengthened by both (a) parent religiosity and (b) teen (or emerging adult) religiosity.

16

CHAPTER THREE

Data and Methods

Data for this study come from the National Study of Youth and Religion (NSYR), a nationally representative longitudinal telephone survey collected by researchers at the

University of at Chapel Hill. The first wave of the study was conducted between July 2002 and April 2003 (Time 1) using a random-digit-dial (RDD) method, and researchers collected interviews for 3,290 English- and Spanish-speaking teenagers aged 13 to 18.5, as well as one parent of each respondent. An additional oversample of

80 Jewish households was also collected, but these cases are excluded here as recommended by Smith and Denton (2005, p. 292) because they could distort proper proportions when generalizing descriptive results to the national population, as well as introduce potential sampling bias net of other independent variables controlled in multivariate analyses.

Two follow-up surveys were conducted on the teens (but not the parents), the first of which was collected from June to November 2005 (Time 2), and the second from

September 2007 to April 2008 (Time 3). During the final wave of the survey, respondents were between the ages of 17 and 24. A total of 2,604 of the original respondents participated in the second wave (retention rate of 78.6 percent), and 2,532 participated in the third wave for an overall retention rate of 77.1 percent. Diagnostic analyses comparing the NSYR with U.S. Census data and other comparable adolescent

17

surveys confirm that it provides a nationally representative sample when constructed raw weights are used (for more details, see Smith & Denton, 2005, p. 292).

While the number of panel studies on delinquency with a national probability sample of adolescents in the United States is small (Akers, 2009), even smaller, by extension, is the number of longitudinal studies for homeschooling students, due to the relatively short history of the movement and the difficulty on collecting data on home education. The NSYR is an unusual data source for the analyses proposed here, as it represents one of the first panel studies that allow for differentiation of homeschooling from other forms of secondary schooling.

Dependent Variables

The primary focus of the current study is on drinking alcohol. On each of the three waves of the survey, respondents were asked “How often, if at all, do you drink alcohol, such as beer, wine or mixed drinks, not including at religious services?”

Possible response options ranged from (1) “Almost every day” or “Once a day or more” to (7) “Never,” though responses were recoded dichotomously for this study as a participation measure. It should be recognized that the alcohol use would not be illegal for many of the respondents who may have matured past the age of legalization by the time the third wave of the study was collected. That said, alcohol use, especially when excessive and/or in highly religious family and peer contexts, may still be “frowned upon,” and at the least, they represent unhealthy or risky behaviors (Perkins, 1987) that are “analogous” to crime and often predictive of more serious criminality (Akers, 2009, p. 141; Gottfredson & Hirschi, 1990, p. 90).

18

Secondary School History

Categorizing secondary school type is challenging because it is not necessarily static for individual respondents across the duration of a panel study. Murphy (2012, pp.

26–27) reports that students enter and leave homeschooling situations at different points in their primary and secondary education careers, and the decision to homeschool is often considered on a year-to-year basis. That said, categorization of initial school type is straight-forward in this study, since data are not available regarding school type prior to

Time 1, and all categories of school type are mutually exclusive. Therefore, Time 1 school history is coded into a system of dummy variables including 1) homeschool

(reference), 2) public school, and 3) private school. Respondents who reported attending any other type of school or reported not attending school are excluded from the current study1.

While respondents might change school types between waves of the survey, they might also mature out of secondary school. School type for the follow-up surveys is thus coded into two separate systems of dummy variables, one reflecting respondents’ current school status, in addition to school history. Specifically, current school status options at both Times 2 and 3 include 1) secondary, 2) college/postsecondary, 3) not enrolled, no diploma or degree awarded, and 4) not enrolled, diploma or degree awarded. School history at Time 2 includes options of 1) homeschool (reference), 2) private school only,

1 Excluded school type categories include “Stopped going to school/dropped out,” “Magnet or charter school,” “Part high school and part college/postsecondary,” “Only college or postsecondary,” “Other,” “Don’t know,” and “Refused.” The total count of excluded responses at Time 1 is 84.

19

3) public school only, and 4) public and private school,2 while those reporting attendance at other types of school at Time 2 are excluded3. The third wave of the survey does not distinguish between public and private school students, so school history at Time 3 includes only homeschool and non-homeschool categories (excluding those who reported attendance at other types of school4 at Time 3). As homeschoolers are the primary group of interest, and because homeschooling N-counts are relatively low, the homeschool category at Times 2 and 3 includes respondents who reported attending either a public or private school at some other point during the study. Inclusion of these cases in the homeschool category should have a conservative effect, however, as relationships formed during attendance at public or private school are hypothesized to have a positive effect on alcohol use.

Independent Variables

Homeschooling is expected to provide a structure in which children and teens are exposed to anti-criminogenic influences, as predicted by social bonding, self-control, social learning, and general strain theories, and also including independent effects related to religion. To test these hypotheses, five groups of independent variables will be used as

2 Respondents in the “public and private school” category presumably switched between public and private schools in the time between the first and second waves of the survey, though the order (public to private, or private to public) is not differentiated here. While the order of these transitions may have consequences for juvenile delinquency, the issue is beyond the scope of the current study.

3 Excluded school types at Time 2 include “Other,” “K-12,” “Adult education,” “Magnet or charter school,” and “Other,” as well as those excluded at Time 1.

4 Excluded school types at Time 3 include “Other (specify),” “School that includes all grade levels (K through 12),” and “Adult education [classes outside of an official program],” as well as those excluded at Times 1.

20

explanatory variables in intermediate regression analyses, while all groups available for given waves will be used in full models.

The first group of variables relates to the expectation from social bonding theory that attachment to and involvement with parents and conventional adults tend to be negatively associated with delinquency (Hirschi, 1969). These concepts are operationalized as follows:

 Teen affective ties, or an index of three items (see Appendix A) related to how

close teens feel toward their mothers and/or fathers (responses are averaged when

both parents are present in the household), measured on all three waves of the

survey. Response values range on six-point ordinal scales, with each measure

coded to range from least- to most-attached to parent(s). Responses are then

standardized and summed into a single index (=.76, .79, and .83 for Times 1, 2,

and 3, respectively).

 Parent affective ties, or how close parents feel toward their teens. On the first

wave of the survey only, parents are asked how close they and their spouse (when

available) feel to their teen, with response values ranging from (1) “not close at

all” to (6) “extremely close” (responses are averaged when both parents are

present in the household).

 Adult support, or the total number of adults, excluding parents, to which a teen

can turn for support or advice, measured on the first and second waves of the

survey only and ranging from 0 to more than 15.

 Parent-teen activities, which represents a total number of conventional activities

in which parents and teens have participated together over the last 6 months

21

(measured at Time 1 only), with the following activities possible: 1) visited a

museum, art gallery, or historical site; 2) attended a play, concert, or other show;

3) visited a library; 4) worked on a project such as making a craft, building or

fixing something; and 5) played a game, sport, or exercised.

Social bonding theory also informs the second group of variables, which test whether commitment to conventional lines of action and increased belief in shared behavioral norms tend to be negatively associated with delinquency (Hirschi, 1969).

While more typical measures of these concepts, such as grade point average, may not be applicable to homeschool students, and others like time spent on homework are not available, I intend to use the following serve as proxies for commitment and belief:

 Future, specifically how often the teen plans for the future, measured on all three

waves with response options ranging from (0) “never” to (4) “very often.”

 School importance, or the degree to which doing really well in school-work is

important to the teen, measured on the first two waves of the survey only, with

response options ranging from (0) “not important at all” to (4) “extremely

important.”

 Purpose, the sum of two standardized items (see Appendix A) related to the teens’

sense of direction and purpose in life, measured on the third wave only and coded

from least to most purpose (=.55).

 Moral absolute, or whether the teen agrees that there are no definite rights and

wrongs for everybody, which is available on all three waves of the survey and

coded dichotomously (0=agreement, 1=disagreement) for the first wave of the

survey, and from (0) “strongly agree” to (4) “strongly disagree” for the second

22

and third waves. Responses from the second two waves are not dichotomized for

consistency because a small percent of responses are coded as an intermediate

response of “unsure.”

Respondents who are committed to conventional lines of action would be expected to plan more for the future, feel that doing well in school is important, and feel that they have direction or a purpose in life. A belief that morality is absolute is interpreted here as representing an objective belief in behavioral norms, though it is recognized that moral norms and legal norms are not necessarily synonymous to individual respondents.

A third group of variables operationalizes social learning predications that differential associations with conventional and deviant peers influence delinquency

(Akers, 2009; Sutherland & Cressey, 1970). These measures are derived from a series of measures following a prompt on the second and third waves of the survey for the respondent to think of up to five of their closest friends. Respondents were asked how many of these friends are not religious (non-religious peers), and how many do drugs or drink a lot of alcohol (drug-using peers), on both the second and third waves, and they were asked how many of these friends have been in trouble for cheating, fighting, or skipping classes (deviant peers) on only the second wave of the study.

The fourth group of independent variables relate to strain theory, which suggests that strain tends to generate negative emotions, which in turn leads to delinquency

(Agnew, 2006). Specifically, two measures of strain that were included on the first wave of the survey are neighborhood strain, which ranges from (1) “very safe” to (4) “not very safe,” and the frequency at which the teen is teased, which ranges from (0) “never” to (4)

“almost every day.” The gap between a student’s educational aspiration and educational

23

expectation was also measured on the first wave only as the difference between how far in school the teen would like to go and realistically how far in school the teen thinks they will go, both ranging from (1) “no farther in school” to (7) “postgraduate or professional school.”5

Two strain items included on the second and third waves of the survey are whether they experienced trauma, coded dichotomously (0=no, 1=yes), and their perceived health problems, ranging from (1) “excellent” to (5) “poor.” Further, two additional items were included on all three waves of the survey, specifically satisfaction with one’s body, ranging from (0) “very happy” to (4) “very unhappy,” and the number of parental breakups experienced by the teen, ranging from 0 to more than 10 (note that the time frame of reference indicated by the question differs on the second wave of the survey; see Appendix A for more information). In addition to types of strain, negative emotions that can be assessed include the parent’s perception at Time 1 of their teen’s temper, ranging from (1) “not bad” to (3) “very bad,” and the teen’s self-reported frequency of sadness or depression, measured on all three waves and ranging (1) “never” to (5) “always.”

Finally, to the extent that religion is expected to influence deviance independently from social bonding, social learning, and general strain theories, the effect of homeschooling on teen religiosity is also relevant to crime and delinquency. Variables related to religiosity thus comprise a fifth group of independent variables, including religious salience, which is measured by asking teen (or emerging adult) respondents on

5 Homeschoolers are not given options 1 (“no farther in school”) or 2 (“some high school”) to the aspiration measure, and they were not offered option 1 (“no farther in school) to the expectation measure.

24

each wave of the survey how important religious faith is in shaping how they live their daily life, and ranges from (0) “not important at all” to (4) “extremely important.”

Similarly, parent religious salience is measured at Time 1 by asking parental respondents how important religious faith is in guiding day-to-day living, with possible responses ranging from (0) “not important at all” to (5) “extremely important.” Both religious salience and parent religious salience were standardized by converting to Z-scores.

Religious support represents the number of adults to whom a teen can turn for support or advice that belong to a religious congregation, measured on the first and second waves of the survey only and ranging from 0 to more than 15. Of the five friends respondents were asked to think about, as mentioned above, they are asked how many were involved in a religious youth group (religiously-involved peers) at Time 2, and how many are religious

(religious peers) at Time 3.

Moderating Variables

It is likely that parent religiosity moderates the effect of homeschooling on endogenous predictors and teen deviance. Interaction terms are created by multiplying the value of parent religious salience at Time 1 with the school history dummies created for each wave (e.g., Parent Rel X Public). It is also likely that teen (or emerging adult) religious salience is a more important moderator of homeschooling, since it is more reflective of actual religious internalization by the teen, as opposed to that desired by their parents. A separate interaction term is thus created by multiplying the value of teen religious salience at each wave by the respective school type dummy (e.g., Rel X Public).

Interaction terms are included among the fifth group of independent variables (those related to religiosity).

25

Controls

Adolescent control variables include age, (female=1), race/ethnicity (a system of dummy variables including black, Hispanic, and other race, with white being the reference category), and religious affiliation (based on Steensland et al., 2000,

“Evangelical Protestant” as reference category). Additional family-level controls include region (a system of dummies including South (reference), Northeast, Midwest, and West) and household income, which ranges from (1) “less than $10k” to (11) “more than

$100k.” Furthermore, because homeschool families are more likely than non- homeschool families to be two-parent households (Bielick et al., 2001, p. 8; Daniel

Princiotta et al., 2006, p. 11), and because single-parent households have been shown to contribute to adolescent drinking and drug use (Gottfredson & Hirschi, 1990, p. 103;

Longest & Shanahan, 2007; Longest & Vaisey, 2008), a dummy variable for two-adult care-taker households is also included to remove any potential bias.

Analytic Strategy

Participation in drinking will be regressed in cross-sectional binary logistic analyses for each of the three waves of the study. For each set of analyses, a baseline model will include only demographics and school type (and current school status from

Times 2 and 3) to ascertain the effect of school type on the deviance of alcohol use. A series of intermediate models will then be analyzed, each of which includes one of the five groups of independent variables described above.6 A full model will then be regressed which includes all variables available for a given wave of the survey. This

6 Note that only four intermediate models are tested in analyses utilizing Time 1 independent variables, since questions about friends were not asked on the initial wave of the study.

26

strategy will allow for cross-sectional consideration of social bonding and general strain variables on all three waves, while social learning will be assessed only at Times 2 and 3.

Additionally, two sets of 2-wave longitudinal regressions will also be performed, including (1) Time 2 deviance variables regressed on Time 1 criminological variables and

(2) Time 3 deviance variables regressed on Time 2 criminological variables. Similar to the cross-sectional regressions, the baseline model will include only demographic, current school status (when applicable), and school type history variables. Because the strongest predictor of contemporary deviance tends to be prior deviance (Akers, 2009; Gottfredson

& Hirschi, 1990), the second longitudinal model will add a variable representing earlier participation in drinking. Then, a series of intermediate models will add one of the groups of independent variables each, and a full model will include all variables available from a given wave of the survey. The use of longitudinal analyses to compliment the cross-sectional analyses will provide evidence for causality in the homeschool-deviance relationship.

27

CHAPTER FOUR

Results

Table 1 presents descriptive statistics for the sample by wave of the survey, specifically those related to school categorization and socio-demographic variables. The proportion of homeschoolers represented in the sample ranges from 2.37 to 3.40 percent, which is consistent with the national estimates contemporary to this period (Murphy,

2012, p. 9). Teens who attended private school through the survey represented about

8.86 and 7.94 percent of the respondents during the first two waves, respectively.

Naturally, average age increased steadily through the study, and current school status reflected this maturation process. While all students included in the sample were enrolled in secondary education at Time 1, only about 63 percent were still in high school at Time

2, and by Time 3, those enrolled in secondary school dropped to 7 percent. Those enrolled in postsecondary education, however, rose to about 24 and 54 percent at Times 2 and 3 respectively, and the number of respondents not enrolled in any school rose through the course of the study as well.

The sample was split nearly evenly in terms of gender. A majority of the sample was white (about 65 to 69 percent), while about 15 to 18 percent were black, another 9 to

12 percent Hispanic, and other races represented about 5 percent. Evangelicals were the modal religious affiliation, comprising about a third of the sample, and Catholics represented about a quarter of the sample. Mainline Protestants, black Protestants, and those with no affiliation comprised roughly 10 to 12 percent of the sample, while Jewish,

28

Mormon, and other religious groups, as well as those whose religious affiliation was not known or determinate, were represented by less than 3 percent of the sample each.

Southerners were the largest regional group, with nearly 42 percent of the sample, and

Northeasterners were the least represented, comprising about 15 percent of the sample.1

Mean household income was close to $50,000 across all three waves of the survey, and nearly three quarters of the sample lived in two-adult households.

Table 2 presents group means by school type for dependent variables, as well as the religious and criminological covariates, with t-tests performed to determine if differences between homeschool students and those from other categories were significant. Consistent with Hypothesis 1, homeschooled adolescents and emerging adults were less likely to report participating in drinking alcohol than non-homeschooled teens and emerging adults at Times 2 and 3, while group differences were not significant at Time 1. Specifically, while about two thirds or more of those who were attending or had attended public and/or private schools at Times 2 (64 to 66 percent) and 3 (79 percent) reported alcohol use, the proportion of homeschoolers that reported the same was about 20 percent smaller at Times 2 (44 percent) and 3 (60 percent).

Very little differences were found in terms of social bonding, social learning, or general strain variables with several notable exceptions. First, homeschool teens had more religious or religiously-involved friends than public school students at Time 2 (1.96 vs. 1.30) or non-homeschool students at Time 3 (2.63 vs. 2.15) as anticipated, although

1 Despite apparent disproportionality in regional representation, diagnostics performed by others (Smith & Denton, 2005, p. 292) confirm that the sample is nationally representative when using raw weights. Raw weights were used in conformance with instructions by Smith and Denton (2005), specifically in conjunction with controlling for region and income, and excluding the Jewish oversample.

29

private school students (.90) reported fewer deviant peers than homeschool students

(1.35) at Time 2. Second, homeschoolers (1.75) tended to feel less safe in their neighborhoods at Time 1 than private school students (1.49), and they were in poorer health than private school students at Time 2 (2.34 vs. 1.91). Third, homeschooled teens

(3.08) had more religiously-involved adults, aside from their parents, to whom they could turn for support than public school students (2.04) at Time 2, and they consistently had more absolute beliefs regarding morality than their public school peers at Times 1 (.64 vs.

.52), 2 (2.36 vs. 1.82), and 3 (2.47 vs. 2.11), but not their private school peers (.59 and

2.41; no separate category of private school at Time 3). Interestingly, homeschooled teens were also more religious than those who did not homeschool (e.g., 2.70 vs. 2.44 at

Time 1), though their parents were not (4.13 vs. 3.98 at Time 1). While these bivariate results are informative, testing hypothesis requires multivariate analysis, whose results I now turn to.

Odds ratios for cross-sectional binary logistic regressions are presented on Tables

3 through 5. First, Table 3 shows that differences between homeschooled teens and those that attend public schools are not significant in the baseline model (.669), and the odds that private school students drink are .544 times less than, or 45.6% (= 1 - .544) lower than the odds that homeschool students drink, which is opposite to what was hypothesized. While this unexpected group difference persists in Model 2, which includes bonding to family and conventional adult variables, it becomes non-significant in all other intermediate models and the full model as well. This indicates that the smaller odds of private school students drinking relative to their homeschool counterparts at Time 1 were due in part to factors like teen-parent involvement and school importance,

30

which were found to be higher among private school than homeschool students, although the differences were found to be non-significant (see Table 2): that is, private school students reported lower levels of alcohol use partly because they had higher levels of activities with their parents and belief in the importance of school than homeschool students. While the observed group difference is opposite to Hypothesis 1, the way the difference was explained is consistent with Hypothesis 2.

On the other hand, explanatory variables included in the intermediate models tend to be associated with alcohol use in the expected direction. Specifically, the bonding variables of teen affective ties and teen-parent involvement (.855 and .826 in Model 2) as well as school importance and moral absolute (.660 and .715 in Model 3) are found to decrease the odds of drinking, whereas the strain variables of parental breakup, temper, and sad increase the odds (1.267, 1.444, and 1.382 in Model 4). Although the strain of being teased was found to decrease, not increase, the odds (.884) in the intermediate model, it became non-significant in the full model (Model 6). However, the social bonding and strain variables found to be significant in the intermediate model remained so in the final model, where an anomaly was observed: planning for the future, not significant in the intermediate Model 3 (1.056), was positively, not negatively, associated with drinking in the full model (1.131).

Unlike the results at Time 1, group differences in drinking at Time 2 were consistent with Hypothesis 1 (see Table 4). Specifically, the odds of drinking were 2.065 and 2.766 times higher among public school and public/private school students, respectively, while differences between private school students and homeschoolers

(1.553) were not significantly different (see Model 1). Interestingly, group differences

31

mostly increased in most intermediate models (Models 2 through 5) and the full model

(Model 7), suggesting possible suppressor effects; that is, while absolute moral belief

(0.764 in Model 3; 0.840 in Model 7) and social learning variables including both drug- using peers (1.810 in Model 4; 1.701 in Model 7) and deviant peers (1.248 in Model 4;

1.259 in Model 7) were robust and in the predicted direction, the unexpected change in group differences suggests that homeschool students are relatively disadvantaged in these variables. The lone exception proved to be the intermediate model that included religious variables (Model 6), which alone was able to reduce both the public and private school variables (though not public/private) to non-significance. Consistent with Hypothesis 2d, the odds of drinking decreased for each additional religiously-involved friend, by 17.6% in Model 6 and by 8.1% in Model 7.

Similar patterns were noted in the cross-sectional model for Time 3 (Table 5). In the baseline model, the odds that non-homeschoolers drank were 80% higher than those of respondents who had been homeschooled prior to or during the third wave of the survey. As found at Time 2, group differences between homeschoolers and non- homeschoolers were non-significant only in models that included religiosity variables

(including the full model in this case), though a variable different than religiously- involved peers was found to be the explanation: religious salience. As religiosity incrementally increased, the odds of drinking decreased by between 70 and 80 percent in the Models 6 and 7.

Furthermore, the interaction between religiosity and non-homeschoolers (2.513) was also significant in Model 6, which indicates that religiosity has a stronger protective effect against drinking among homeschoolers than it does among non-homeschoolers, as

32

predicted in Hypothesis 3b. As illustrated in Figure 12, among respondents with below- average religious salience, homeschool students were more likely to drink than non- homeschool respondents. As religious salience increases, the predicted probability of drinking decreases among all respondents, but more so for homeschoolers than for non- homeschoolers; among respondents with above-average religious salience, therefore, homeschoolers are less likely to drink than non-homeschoolers, by as much as nearly .3 at the highest levels of religiosity. This indicates that the homeschool students’ lower levels of alcohol use tend to depend on their religiosity more than homeschooling per se.

1

0.8

0.6

homeschool 0.4 non-homeschool

0.2

0 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 Religious Salience

Figure 1. Predicted Probability of Drinking at Time 3 (Cross-Sectional)

2 Predicted probabilities presented in Figure 1 assume respondents that are white, male, evangelicals in two adult care-taker households with average household income and average age at Time 3.

33

Longitudinal results, presented in Tables 6 and 7, tend to be consistent with the cross-sectional findings. First, when regressing drinking at Time 2 on Time 1 independent variables, the odds of drinking among those who attended public and private schools at Time 1 were 2.509 and 1.980 times higher, respectively, than those who were homeschooled at Time 1 (see Model 1 of Table 6). Odds ratios generally became higher when drinking at Time 1 was added (Models 2 through 7), which is consistent with what was observed at Time 1: that is, the smaller odds of public and private students drinking than their homeschool peers. As a result, drinking at Time 1 is found to suppress group differences in this longitudinal analysis. Given the relative strength of prior drinking (as it effectively serves as a proxy for all causal antecedents to prior drinking), this measure was a consistent and dominant predictor of drinking at Time 2, with odds ratios ranged from nearly 6.8 to more than 7.6 in each of model in which it was controlled. Thus, it is not unexpected that very few of the criminological and religious explanatory variables were significant in intermediate models, as only teen-parent involvement (.802 in Model

3; .756 in Model 7) had a robust and predicted effect. In the full model, group differences between homeschoolers and private school students (2.333) became non- significant, while differences between homeschoolers and public school students (2.549) did not. Here, the effects of parental breakup (1.150) and religious support (.947) at

Time 1 were also associated with drinking at Time 2 in expected directions.

Next, when Time 3 drinking was regressed on Time 2 independent variables, the odds that public school students drank increased by 81.7% compared to those of homeschoolers in the baseline model (see Table 7). Prior drinking, which was added to analyses starting at Model 2, was again a strong and consistent predictor of drinking in

34

wave 3, but in contrast to the Time 1 to Time 2 longitudinal analyses, drinking at Time 2 serves as an explanation for drinking at Time 3 rather than a suppressor, as group differences were found to be non-significant in all subsequent intermediate models and the full model as well. Like the prior analyses, though, very few of the criminological and religiosity explanatory variables were significant, again attesting to the power of prior drinking as a proxy for the presence of earlier causal antecedents to drinking.

Notable exceptions include the commitment/belief variables of planning for the future

(.852 in Model 4; .840 in Model 8) and absolute moral beliefs (.876 in Model 4; .824 in

Model 8), as well as the peer composition variable of drug-using friends (1.329 in Model

5; 1.183 in Model 8). Among religiosity variables, the effect of religiously-involved peers (.908) was significant in the intermediate model (Model 7), while religious salience

(.321) was negatively associated with drinking in the full model (Model 8).

In addition, I conducted supplemental analyses by regressing Time 3 drinking on

Time 1 independent variables, presented in Appendix B (see Table B.1). While they should be interpreted with caution given the long interval between the first and last waves, results show that group differences based on school history are generally non- significant with prior drinking at Time 1 (as well as Time 2) being a key predictor of drinking at Time 3. In addition, similarly to Time 3 cross-sectional analyses (see Table

5), religiosity at Time 1 was found to decrease the odds of drinking at Time 3 Model 6).

Furthermore, religiosity at Time 1 was found to interact with the influence of having attended public school (compared to being homeschooled) on drinking at Time 3, such that religiosity is a stronger protective factor of drinking alcohol for homeschool students than for public school students (see Figure B.1).

35

Supplemental analysis was extended to examine other survey items related to drinking alcohol, while they are limited in the scope of measurement compared to the one focused in the present study. Specifically, variables measured whether respondents have gotten drunk, either in the last year (as measured at Times 1 and 2) or in the last two weeks (Time 3), whether teen respondents had ever driven while under the influence

(DUI) within 3 years of Time 2, and whether they binge drank within two weeks of Time

3. Odds ratios for binary logistic regressions of participation in these activities are presented in Tables B.2 through B.12 of Appendix B. First, results for getting drunk largely mirrored those of drinking, as homeschoolers were less likely to get drunk than non-homeschoolers at Time 2 in both cross-sectional (Table B.3) and longitudinal analyses (Table B.5), though they were more likely to get drunk at Time 1 (Table B.2), and group differences were not significant at Time 3 (Tables B.4, B.6 and B.7). Second, for driving under the influence (Tables B.8 and B.9) and binge drinking (Tables B.10 to

B.12), group differences based on school history were not significant, though these results should be interpreted with caution; because they represent behaviors that are more serious than drinking, meaning they are likely to be relatively rare among all groups, lesser group differences are expected unless a larger sample is employed. For each of these behaviors, teen affective ties and teen-parent involvement typically had negative effects, as did religious support and religious peers, while alcohol- and drug-using friends typically had a positive effect on getting drunk.

Finally, socio-demographic variables were controlled in each model described herein, though they were reported in Table 3 only, and omitted from presentation in

Tables 4 through 7 and the Supplemental Tables. Results are generally consistent.

36

Typically, the odds of alcohol use are positively associated with age, which is consistent with developmental theories of delinquency (Moffitt, 1993; Sampson & Laub, 1993).

Gender is not typically significant when all other factors are considered, except in cross- sectional and longitudinal models for Time 3, when females are less likely to drink than males. Females are also less likely to drive under the influence of alcohol, as measured at

Time 2. Race and ethnicity tend to have a consistent and robust effect, as black and

Hispanic teens and emerging adults tend to participate in drinking less than whites, in accordance with prior studies (Bachman et al., 1991).

Religious affiliation tends to have some effect: all other things being equal,

Catholics and nones (and to a lesser extent, Mainline Protestants) tend to drink more than evangelicals, while Mormons tend to drink less than evangelicals. The most robust of these effects tend to be found among Catholics and Mormons; other religious affiliation differences tend to become non-significant in full models. Region is mostly non- significant, with some sporadic exceptions, though no general pattern was observed in direction or magnitude of significant regional effects. Income is typically positively associated with drinking, while having two adult care-givers in the household tends to be negatively associated with drinking. Current school status has some effect in that those currently enrolled in high school tend to drink less than those in college or not enrolled, though these effects tend to lose significance in full models.

37

Table 1

Descriptive Statistics by Wave: NSYR Time 1 Time 2 Time 3 Mean Mean Mean Variable N or % SD N or % SD N or % SD School History Homeschool 76 2.37 81 3.33 81 3.40 Non-Homeschool 2298 96.60 Private School 284 8.86 193 7.94 Public School 2846 88.77 2104 86.51 Public / Private 54 2.22 School Current School Status Secondary 1547 63.61 167 7.02 College / 576 23.68 1293 54.35 Postsecondary Not enrolled, no 78 3.21 384 16.14 diploma or degree awarded Not enrolled, diploma 231 9.50 535 22.49 or degree awarded Demographics Age 3205 15.47 1.42 2432 17.69 1.36 2379 20.01 1.44 Female 3206 49.50 2432 50.66 2379 51.28 Race White 3188 65.15 2417 69.22 2365 69.05 Black 3188 17.82 2417 15.93 2365 16.03 Hispanic 3188 11.76 2417 9.81 2365 9.94 Other 3188 5.27 2417 5.05 2365 4.99 Religious Tradition Evangelical Protestant 3206 31.69 2432 32.52 2379 32.16 Mainline Protestant 3206 10.67 2432 11.84 2379 12.02 Black Protestant 3206 12.23 2432 11.10 2379 10.97 Catholic 3206 24.92 2432 24.34 2379 24.97 Jewish 3206 1.53 2432 1.69 2379 1.77 Mormon 3206 2.25 2432 2.55 2379 2.61 Other 3206 2.68 2432 2.55 2379 2.52 None 3206 11.82 2432 11.35 2379 10.97 Not Determinate 3206 2.21 2432 2.06 2379 2.02 Region South 3206 41.73 2432 41.45 2379 41.11 Northeast 3206 15.25 2432 15.34 2379 15.22 Midwest 3206 22.86 2432 23.60 2379 24.51 West 3206 20.15 2432 19.61 2379 19.17 Household Income 3017 5.89 2.90 2295 6.15 2.89 2239 6.18 2.89 Two Adults 3206 72.24 2432 75.53 2379 75.41 Note: All statistics unweighted.

38

Table 2

Group Means of Dependent Variables and Covariates Homeschool Public Private Public/Private Variable (reference) Schoola School School Drink at Time 1 0.41 0.37 0.37 Drink at Time 2 0.44 0.66 *** 0.65 ** 0.64 * Drink at Time 3 0.60 0.79 **

Bonding to Family and Conventional Adults Time 1 Teen Affective Ties 0.14 -0.02 0.08 Parent Affective Ties -0.03 -0.01 0.04 Adult Support 5.04 4.99 5.73 Parent-Teen Involvement 3.34 3.14 3.46 Time 2 Teen Affective Ties 0.50 -0.01 0.15 -0.19 Adult Support 5.53 4.96 5.83 4.83 Time 3 Teen Affective Ties 0.12 -0.02

Commitment and Belief Time 1 Future 3.08 3.16 3.10 School Importance 3.31 3.35 3.36 Moral Absolute 0.64 0.52 * 0.59 Time 2 Future 3.03 3.19 3.13 3.34 * School Importance 3.20 3.25 3.32 3.21 Moral Absolute 2.36 1.82 *** 2.41 2.02 Time 3 Future 3.28 3.37 Purpose -0.08 -0.06 Moral Absolute 2.47 2.11 *

Peer Group Composition Time 2 Non-religious Peers 1.19 1.18 0.88 1.02 Drug-using Peers 1.46 1.76 1.79 1.66 Deviant Peers 1.35 1.36 0.90 * 1.44 Time 3 Non-religious Peers 1.23 1.36 Drug-using Peers 1.51 1.70

Strain and Negative Emotions Time 1 Aspiration/Expectation Gap 0.24 0.24 0.09 Neighborhood Strain 1.75 1.60 1.49 ** Teased 0.87 0.90 1.08 Body 1.01 0.97 1.00 (Table 2 continues)

39

Homeschool Public Private Public/Private Variable (reference) Schoola School School

Parental Breakup 0.55 0.57 0.40 Temper 1.54 1.48 1.42 Sad 2.45 2.26 2.24 Time 2 Trauma 0.55 0.55 0.46 0.60 Health Problems 2.34 2.19 1.91 ** 2.28 Body 1.31 1.21 1.24 1.09 Parental Breakup 0.33 0.24 0.13 0.19 Sad 2.43 2.31 2.28 2.43 Time 3 Trauma 0.56 0.48 Health Problems 2.35 2.24 Body 1.19 1.12 Parental Breakup 0.68 0.76 Sad 2.42 2.28

Religious Networks Time 1 Religious Salience 2.70 2.44 * 2.68 Parent Religious Salience 4.13 3.98 4.13 Religious Support 2.90 2.25 3.16 Time 2 Religious Salience 2.78 2.27 *** 2.42 * 2.30 * Religious Support 3.08 2.04 * 2.95 2.33 Religiously Involved Peers 1.96 1.30 ** 1.69 1.45 Time 3 Religious Salience 2.54 2.21 * Religious Peers 2.63 2.15 * Note: All statistics are unweighted; T-tests compare respective means to those of homeschoolers. aFor variables measured at time 3, values in the "public school" column include all individuals who did not report attending homeschool at any time during the study. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

40

Table 3

Binary Logistic Regression Odds Ratios for Drinking at Time 1 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Age 1.599 *** 1.582 *** 1.567 *** 1.579 *** 1.596 *** 1.532 ***

Female 1.004 0.997 1.111 0.920 1.048 1.047 Race (White=ref) Black 0.521 *** 0.478 *** 0.628 * 0.565 ** 0.658 * 0.632 * Hispanic 0.996 0.979 1.083 0.885 1.098 1.010 Other 0.869 0.810 0.861 0.940 0.934 0.842 Religious Tradition (Evangelical=ref) Mainline Protestant 1.326 1.285 1.210 1.314 1.233 1.141 Black Protestant 0.976 0.972 0.863 0.979 0.812 0.801 Catholic 1.460 *** 1.345 * 1.406 ** 1.521 *** 1.204 1.228 Jewish 1.612 1.752 1.569 1.649 1.228 1.369 Mormon 0.178 *** 0.181 *** 0.128 *** 0.161 *** 0.185 *** 0.127 *** Other 1.127 1.014 1.004 1.039 0.890 0.810 None 1.787 *** 1.646 *** 1.631 *** 1.625 ** 0.961 0.920 Not Determinate 1.570 1.331 1.463 1.660 1.371 1.135 Region (South=ref) Northeast 0.856 0.880 0.806 0.854 0.722 * 0.747 Midwest 0.966 1.030 0.966 1.012 0.875 0.988 West 1.123 1.092 1.086 1.091 1.033 1.024 Household Income 1.028 1.040 * 1.035 * 1.053 ** 1.009 1.049 * Two Adults 0.770 * 0.658 *** 0.789 * 0.890 0.819 0.841 School History (Homeschool=ref) Public 0.669 0.615 0.643 0.670 0.699 0.657 Private 0.544 * 0.511 * 0.549 0.592 0.684 0.705 Teen Affective Ties 0.855 *** 0.903 *** Parent Affective Ties 0.965 1.016 (Table 3 continues)

41

Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Adult Support 1.003 1.017 Teen-Parent Involvement 0.826 *** 0.848 *** Future 1.056 1.131 * School Importance 0.660 *** 0.767 *** Moral Absolute 0.715 *** 0.746 ** Aspiration/Expectation Gap 1.056 1.048 Neighborhood Strain 0.936 0.951 Teased 0.884 ** 0.917 Body 1.077 1.037 Parental Breakup 1.267 *** 1.232 *** Temper 1.444 *** 1.334 *** Sad 1.382 *** 1.226 *** Religious Salience 0.729 0.825 Parent Religious Salience 0.803 0.801 Rel X Public 0.899 0.893 Rel X Private 0.534 0.590 Parent Rel X Public 1.234 1.249 Parent Rel X Private 1.547 1.484 Religious Support 0.986 0.985

Max -rescaled R-Square 0.17 0.22 0.19 0.21 0.21 0.27 N 2996 2908 2878 2797 2856 2564 *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

42

Table 4

Binary Logistic Regression Odds Ratios for Drinking at Time 2 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 2.065 ** 1.983 ** 2.225 ** 2.560 ** 2.210 ** 1.743 2.844 ** Private 1.553 1.543 1.954 * 2.199 * 1.730 1.472 2.891 * Public/Private 2.766 * 2.877 * 3.117 ** 4.284 ** 2.820 * 3.067 * 5.324 ** Teen Affective Ties 0.902 *** 0.974 Adult Support 0.986 0.999 Future 1.116 * 1.207 ** School Importance 0.775 *** 0.899 Moral Absolute 0.764 *** 0.840 *** Non-religious Peers 0.998 0.941 Drug-using Peers 1.810 *** 1.701 *** Deviant Peers 1.248 *** 1.259 *** Trauma 1.294 ** 1.117 Health Problems 1.164 ** 1.011 Body 1.033 0.962 Parental Breakup 1.036 0.903 Sad 1.107 1.008 Religious Salience 0.641 0.719 Parent Religious Salience 0.945 1.876 Rel X Public 0.983 1.022 Rel X Private 0.833 0.919 Rel X Public/Private 0.741 0.884 Parent Rel X Public 1.145 0.574 Parent Rel X Private 1.410 0.739 Parent Rel X Public/Private 1.338 0.601 Religious Support 0.978 0.969 Religiously-involved Peers 0.824 *** 0.919 * Max-rescaled R-Square 0.14 0.16 0.18 0.35 0.15 0.22 0.38 N 2261 2239 2078 2146 2229 2156 1902 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

43

Table 5

Binary Logistic Regression Odds Ratios for Drinking at Time 3 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Non-Homeschool 1.800 * 1.703 * 1.757 * 2.060 * 1.843 * 1.657 1.468 Teen Affective Ties 0.980 1.021 Future 1.146 * 1.088 Purpose 1.003 1.029 Moral Absolute 0.806 *** 0.903 * Non-religious Peers 1.094 * 0.956 Drug-using Peers 1.500 *** 1.437 *** Trauma 1.301 * 1.190 Health Problems 1.157 * 1.042 Body 1.038 1.011 Parental Breakup 1.169 ** 1.137 * Sad 0.981 0.945 Religious Salience 0.242 *** 0.298 ** Parent Religious Salience 1.542 1.317 Rel X Non-Homeschool 2.513 * 2.177 Parent Rel X Non-Homeschool 0.620 0.716 Religious Peers 0.960 0.957

Max -rescaled R-Square 0.16 0.16 0.17 0.22 0.17 0.22 0.27 N 2214 2182 2201 2138 2160 2194 2046 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

44

Table 6

Binary Logistic Regression Odds Ratios for Drinking at Time 2 (2-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 2.509 ** 3.343 ** * 3.151 *** 3.243 *** 2.869 ** 3.421 *** 2.549 * Private 1.980 * 2.758 ** 2.685 ** 2.819 ** 2.530 * 2.796 ** 2.333 Teen Affective Ties 0.958 1.007 Parent Affective Ties 1.013 0.986 Adult Support 1.022 1.065 ** Teen-Parent Involvement 0.802 *** 0.756 *** Future 1.033 1.059 School Importance 0.809 ** 0.932 Moral Absolute 0.837 0.919 Aspiration/Expectation Gap 0.926 0.865 * Neighborhood Strain 0.937 0.831 * Teased 0.955 1.020 Body 1.089 1.064 Parental Breakup 1.121 1.150 * Temper 0.901 0.882 Sad 0.921 0.925 Religious Salience 0.905 0.780 Parent Religious Salience 0.951 0.970 Rel X Public 0.817 0.972 Rel X Private 1.035 1.319 Parent Rel X Public 1.080 1.119 Parent Rel X Private 1.115 1.115 Religious Support 0.987 0.947 * Drink at Time 1 7.577 *** 6.785 *** 7.584 *** 7.611 *** 7.072 *** 7.566 *** Max-rescaled R-Square 0.14 0.29 0.30 0.30 0.30 0.31 0.34 N 2310 2308 2239 2223 2158 2210 1984 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

45

Table 7

Binary Logistic Regression Odds Ratios for Drinking at Time 3 (2-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 School History (Homeschool=ref) Public 1.817 * 1.357 1.299 1.297 1.777 1.325 0.892 0.625 Private 0.973 0.796 0.773 0.816 1.005 0.776 0.639 0.438 Public/Private 2.179 1.756 1.537 1.832 3.068 1.892 1.219 1.266 Teen Affective Ties 0.999 1.034 Adult Support 0.990 1.017 Future 0.852 * 0.840 * School Importance 0.994 0.925 Moral Absolute 0.876 * 0.915 Non-religious Peers 0.910 0.824 ** Drug-using Peers 1.329 *** 1.183 ** Deviant Peers 0.955 1.036 Trauma 0.959 0.990 Health Problems 0.867 0.786 ** Body 1.027 1.020 Parental Breakup 1.099 1.039 Sad 0.875 0.838 (Ta ble 7 continues)

46

Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Religious Salience 0.393 0.321 * Parent Religious Salience 0.654 0.522 Rel X Public 2.206 2.692 Rel X Private 2.247 2.647 Rel X Public/Private 2.128 2.480 Parent Rel X Public 1.462 1.788 Parent Rel X Private 1.047 1.265 Parent Rel X Public/Private 1.550 2.364 Religious Support 0.963 0.938 Religiously-involved Peers 0.908 * 0.905 Drink Time 2 9.670 *** 9.567 *** 8.581 *** 7.536 *** 10.524 *** 8.543 *** 7.312 ***

Max -rescaled R-Square 0.16 0.36 0.36 0.36 0.39 0.37 0.39 0.41 N 1966 1960 1945 1820 1862 1935 1876 1668 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

47

CHAPTER FIVE

Discussion

Are homeschoolers less likely to drink alcohol than non-homeschoolers? Both bivariate and multivariate analysis results tend to indicate that they are, especially as respondents began to enter emerging adulthood, thus generally supporting Hypothesis 1.

Furthermore, in accordance with Hypotheses 2a through 2d, the group differences observed between homeschoolers and others were reduced to non-significance in full multivariate regression models when religious and criminological covariates were controlled in cross-sectional models at Times 1 and 3, and partially in longitudinal models as well. That said, intermediate models that independently tested the effects of groups of variables specified in Hypotheses 2a through 2d suggest more complex patterns of covariates.

For instance, Hypotheses 2a and 2b were empirically supported in cross-sectional analyses, as school history coefficients were lower than baseline models when controlling for bonding to parents and conventional adults at Times 2 and 3, and commitment/belief at Time 3. Notably, school history coefficients remained significant in these models, suggesting that social bonding variables provide a limited and partial explanation for group differences. However, while homeschool proponents suggest that homeschooling allows for the development of closer familial bonds, and many homeschool parents decide to homeschool for “family reasons,” homeschoolers were found to be no different

48

than non-homeschoolers on each of the most salient measures in these models, such as teen affective ties and parent-teen involvement.

That said, two important exceptions are of interest. First, homeschoolers had more absolute beliefs about morality, a pattern consistent with the frequently cited motivation of parents to instill religious or moral values, and moral absolutism was found to be a relatively consistent predictor of decreased drinking.1 The second exception relates to attachment to conventional adults. Proponents of homeschool suggest that home education allows for age-integration and greater socialization with adults, and that socialization with adults is superior to socialization with peers (Farris & Woodruff, 2000;

Medlin, 2000). Consistent with bonding theory, data used in this study partially support these arguments, but with a caveat. First, the number of adults to whom teen respondents could turn for support, aside from their parents, was statistically no different for homeschoolers than public and private school students in bivariate analyses, though homeschoolers did have more religious support at Time 2. Unexpectedly, both adult and religious support typically had no effect in multivariate regressions, except for longitudinal analyses at Time 2, where adult support at Time 1 had a positive effect, but religious support at Time 1 had a negative effect on drinking at Time 2. Similar results were encountered for the more serious delinquent behavior of driving while under the influence alcohol (Table B.9). Thus, it appears that it is specifically religious adult

1 As an aside, the effect of planning for the future was not as consistent; those who spent more time planning for the future were found to drink more in cross-sectional analyses, but less in longitudinal analyses. On the one hand, this finding may suggest that it is a poor proxy for the commitment concept, since planning for the future does not necessarily imply planning for conventional lines of action. On the other hand, since planning for the future only has the expected effect on future outcomes, the data suggest that drinking is less likely among those who have planned for the future than it is for those who are currently planning for the future.

49

support that decreases the odds that teens and emerging adults participate in substance use, as they are probably less likely to drink themselves.

In accordance with prior studies (e.g. Curran, Stice, & Chassin, 1997), the data here indicate that increased association with alcohol- and drug-using friends, as well as friends who have gotten into trouble for cheating, lying, or skipping class (though to a lesser extent), was a consistent risk factor for drinking. However, group differences increased from baseline levels in multivariate regressions that included social learning controls, suggesting that peer network composition might suppress group differences rather than explain them (counter to Hypothesis 2c). Even though proponents of homeschooling believe that the practice allows parents to have better control of their teen’s social contacts (Ballmann, 1987; Jeub, 1994), and many parents homeschool out of a concern for the learning environment of public schools, bivariate analyses suggested that homeschoolers had as many deviant peers as non-homeschoolers, and the protective effect of homeschooling is more evident when considered net of the presence of deviant peers (MacKinnon, Krull, & Lockwood, 2000).

Similarly, concepts related to general strain might also contribute suppressor effects at Times 2 and 3. While some parents are motivated to homeschool in order to prevent their children from experiencing strains such as teasing or bullying (Farris &

Woodruff, 2000), the homeschooling environment seemingly does little to either shield teens from negative experiences or cope with the resultant negative emotions in comparison to public and private high schools, as evidenced by non-significant group differences in bivariate analyses. In fact, homeschool teens were found to perceive less safety in neighborhoods at Time 1 and have poorer health at Time 2 than private school

50

students. Further, homeschooled teens experienced similar levels of temper and sadness as non-homeschooled teens, two negative emotions that were often found to predict higher odds of drinking in multivariate analyses. As indicated by the increase in school history coefficients in models that include strain and negative emotions, neglecting these concepts might also underestimate group differences in alcohol use.

While bonding to parents and conventional adults partially explains group differences by school type, and though variables related to composition and general strain were potentially suppressing group differences, intermediate models that include religiosity variables explained group differences more fully, thus supporting

Hypothesis 2d. This finding is consistent with prior literature that argue for the importance of accounting for religiosity in criminological studies (Johnson & Jang, 2011;

Ulmer, 2012) and consistent with those that demonstrate the strength of religious explanations of substance use relative to social bonding variables (Longest & Vaisey,

2008). Specifically, homeschooled respondents had more religious friends at Time 2 and were more religious themselves at Time 3, and both variables had protective effects on drinking. Moreover, Hypothesis 3b received some support (though Hypothesis 3a did not), as religiosity was found to moderate the effect of school history in both cross- sectional and longitudinal analyses of drinking at Time 3, such that the protective effect of religiosity was stronger among homeschoolers than non-homeschoolers.

Finally, the strongest predictor of drinking in longitudinal analyses was found to be prior drinking, which is not surprising (Akers, 2009; Gottfredson & Hirschi, 1990), as prior drinking serves as a proxy for any causal antecedent already experienced by individual respondents, including both those represented in the model (explaining the

51

lack of significance for most variables in intermediate and full models) as well as those that are not (explaining their relatively high odds ratios). What was unexpected, is that controlling for drinking at Time 1 increased, instead of decreased, the odds of non- homeschool school students drinking at Times 2 and 3, which shows the prior drinking suppressed, rather than of explained, group differences at Time 2 and, to a lesser extent,

Time 3. However, drinking at Time 2 explained the group difference at Time 3 as anticipated. These findings illustrate the importance of controlling for prior measure of dependent variable in longitudinal research on substance use because it may be either a suppressor or explanation of group differences.

52

CHAPTER SIX

Conclusions

The present study empirically demonstrates that teens and emerging adults who have been homeschooled are generally less likely to drink alcohol than those who attended public and/or private high schools. But just as motivations to homeschool are complex, mechanisms explaining propensity to drink in this population are likely to be complex as well. In short, surveys have shown that many families tend to homeschool either out of a desire to foster family-centered values or out of a concern for the learning environment at public schools, including factors associated with negative peer pressure

(Bielick et al., 2001; Daniel Princiotta et al., 2006). Results reported in this thesis tend to be consistent with social bonding and social learning theories, in that increased attachment to and involvement with parents, and decreased association with deviant peers, tends to be negatively associated with drinking. But results indicate that homeschoolers generally do not differ with non-homeschoolers in family relationships, whether parents or siblings, unlike what proponents of homeschooling tend to suggest

(Ballmann, 1987; Barfield, 2002; Jeub, 1994; Medlin, 2000). Accordingly, bonding variables tend to provide only a limited or partial explanation of group differences in drinking, while social learning (as well as general strain) variables tend to suppress, rather than explain, group differences, which is inconsistent with this study’s hypothesis.

As expected, however, homeschoolers do tend to differ from non-homeschoolers in religion, which has played a critical role in the historical emergence of modern

53

homeschooling in the United States (Gaither, 2008, 2009). Today, conservative

Protestants and fundamentalists represent large proportions of homeschool families

(Kunzman, 2009a; Murphy, 2012; Ray, 2004), and many choose home education out of an explicit desire to instill religious values (Bielick et al., 2001; Daniel Princiotta et al.,

2006) because of a felt moral imperative to nurture their children’s moral and spiritual development (Ballmann, 1987; Cizek, 1994; Kunzman, 2009a, 2009b; Murphy, 2012).

And indeed, homeschooled teens and emerging adults were found to be more religious

(especially in later waves of the survey), have more religious friends and supportive religious adults in their social networks, and have more objective beliefs about morality than their non-homeschool counterparts, especially those at public schools. Moreover, each of these measures tended to be negatively associated with drinking in multivariate regressions, and intermediate models that controlled for religiosity tended to explain more variance in cross-sectional analyses than those that controlled for bonding, learning, and strain variables.

As it became apparent in the course of this study, the quality of proxy measures of the criminological concepts are limited. For example, planning for the future as a proxy of Hirschi’s (1969) bonding element of commitment had inconsistent and sometimes unexpected effects on drinking, while more conventional measures were either not applicable to homeschoolers (e.g., grade point average) or not available (e.g., time spent on homework). Also unavailable were parental use of substances, which relates insofar as parental modeling of deviant behavior is likely to be associated with teen deviance (Jang

& Johnson, 2011).

54

Another limitation of the current study relates to the small number of homeschool students included in the present analysis, specifically, the small percentage of them in the nationally representative sample, though it was expected as modern homeschooling remains a fairly new movement in the United States. While significant group differences were found for drinking and getting drunk, group differences in more serious behaviors such as binge drinking and driving under the influence might have been masked by low standard errors due to sample size, as might the effect of interaction terms in all models.

That said, if the remarkable growth rate of homeschooling continues, further study of these individuals will become not only more feasible, but also more important.

Time use patterns among homeschoolers is one particularly interesting area for future study, since interaction results indicated that, while homeschoolers with above- average religiosity were less likely to drink than their non-homeschool counterparts, they were more likely to drink among those whose religiosity is below average. Literature on routine activities theory has demonstrated that unstructured socializing can result in increased substance use (Osgood, Wilson, O’Malley, Bachman, & Johnston, 1996), so whether homeschoolers have more opportunity for drinking in the absence of capable guardians because they spend less time in an institutional setting may provide an explanation. Changing patterns of homeschool motivation might also be relevant here.

Given that much of the protective effect of homeschooling was found to be related mainly to religiosity, and prior research has demonstrated that parental motivation effects student academic achievement (Collom, 2005), it is not unreasonable to expect that motivational shifts away from ideological and toward more pedagogical concerns (Aurini

55

& Davies, 2005; Collom, 2005; Murphy, 2012) could also affect substance use patterns among homeschoolers.

Topics related to health care and health education are also of particular interest.

Prior research has found that homeschooled children have less access to and utilization of health care than public school children (Cordner, 2012). Given that homeschoolers were just as likely to drink as public school students at Time 1 (and more likely to drink than private school students), whether a lack of health care exacerbates poor health outcomes due to early drinking would be of critical interest. Furthermore, future studies should investigate the level of health education typically received by homeschooled children of all ages.

On the other hand, increased religiosity at Time 3, a period corresponding to entry into early adulthood, is negatively associated with drinking, which is consistent with prior studies that have established a link between religiosity and reduced problem drinking in college, as well as benefits associated with physical, social, emotional, and academic outcomes (Bryant & Astin, 2008; Mayrl & Oeur, 2009; Nelms, Hutchins, Hutchins, &

Pursley, 2007; Perkins, 1987). That said, transitions to emerging adulthood, and especially college, tend to be associated with declining religious behavior (though not religious belief) due to changes in social networks and the experience of greater freedom and less supervision (Uecker, Regnerus, & Vaaler, 2007). As homeschoolers are increasingly attending college (Lyman, 1998; Ray, 2004), their adaptation to campus culture likely conditions substance use. Further study should investigate the degree to which religiosity persists on college campuses among homeschoolers, and the effect this has on both drinking and problem drinking.

56

In conclusion, the present study contributes not only to criminological literature but also to the growing body of homeschool literature by focusing on an important non- academic outcome of home education, namely the propensity to drink alcohol. Results demonstrate that homeschoolers do tend to drink less than non-homeschoolers, as hypothesized. While religion tends to provide the strongest explanation of group differences, concepts related to social bonding also tend to partially explain group differences, whereas those related to social learning and general strain tend to suppress them. The effect of homeschooling on child outcomes will continue to be a critical avenue of future research as long as the movement continues to grow, and as those who have been homeschooled as children continue maturing into adulthood.

57

APPENDICES

58

APPENDIX A

Measures Included on the NSYR by Wave

59

Measures Included on the NSYR by Wave

Time 1 Time 2 Time 3 Teen Affective Ties First, how close or not close do you feel to your       [M/F]? (MOM/DADCLOSE) Generally, how well do you and your [M/F] get along?    (MOM/DADRELAT) How often do you talk with your [M/F] about personal    subjects, such as friendships, dating, or drinking? (MOM/DADTALK1)

Purpose  You don't have a good sense of what it is you're trying  to accomplish in life. (PURPOSE2)   Some people wander aimlessly through life, but you    are not one of them. (PURPOSE3)      Parent Breakups  How many times, if ever, has [your teen] experienced    the breakup of the or marriage-like relationship of the parents in the household where [your teen] was residing at the time? (PBREAKUP) Since we last interviewed you in [survmonth] of   [survyear], how many times have the people you considered to be your parents experienced a break-up of a marriage or marriage-like relationship? (PARBRKUP) How many times, if any, have the people you consider   to be your parents experienced a break-up of a marriage or marriage-life relationship? (PARBRKUP)

60

Supplemental Deviance Measures Time 1 Time 2 Time 3

[If teen ever drinks alcohol, not including at religious     services] How often, if ever, have you gotten drunk in the last year? (DRUNK; responses range from (1)   "never" to (6) "more than once a week") [If has drank alcohol, dk or ref] How many times, if at   all, over the past two weeks have you drunk at least [if female, “4”; if male, “5”] in the same night?  (DRUNK; responses range from (1) "never" to (4)  "five or more times") How many times, if at all, over the past two weeks  have you binge drank? [Includes non-drinkers] (BINGEDR; responses range from (1) "never" to (4)  "five or more times") In the past 3 years have you ever driven while under   the influence of alcohol? (DUI)

61

APPENDIX B

Supplemental Analyses

62

Table B.1

Binary Logistic Regression Odds Ratios for Drinking at Time 3 (3-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

School History (Homeschool=ref) Public 1.596 1.625 1.639 1.696 1.870 1.126 1.251 Private 1.054 1.119 1.132 1.246 1.349 0.971 1.203 Teen Affective Ties 0.966 1.009 Parent Affective Ties 0.976 0.922 Adult Support 0.977 1.035 Teen-Parent Involvement 0.979 0.969 Future 0.813 ** 0.753 *** School Importance 0.862 0.988 Moral Absolute 0.767 * 0.861 Aspiration/Expectation Gap 0.902 0.914 Neighborhood Strain 1.078 1.009 Teased 0.865 ** 0.888 * Body 1.135 * 1.118 Parental Breakup 1.346 *** 1.380 *** Temper 1.070 1.067 Sad 0.896 0.811 ** Religious Salience 0.400 * 0.487 Parent Religious Salience 0.665 0.661 Rel X Public 2.255 * 2.114 Rel X Private 2.010 1.908 Parent Rel X Public 1.438 1.518 Parent Rel X Private 1.015 0.901 Religious Support 0.953 ** 0.946 * Drink at Time 1 5.490 *** 5.751 *** 5.420 *** 5.384 *** 5.317 *** 5.760 ***

Max -rescaled R-Square 0.15 0.24 0.25 0.25 0.27 0.26 0.30 N 2250 2249 2187 2162 2110 2160 1945 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

63

1

0.8

homeschool / private school public school

0.6

0.4 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 Religious Salience

Figure B.1: Predicted Probability of Drinking at Time 3 (3-wave Longitudinal)

64

Table B.2

Binary Logistic Regression Odds Ratios for Getting Drunk at Time 1 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 School History (Homeschool=ref) Public 0.545 * 0.450 ** 0.534 * 0.608 0.565 0.518 Private 0.364 ** 0.314 *** 0.386 ** 0.408 * 0.431 * 0.435 * Teen Affective Ties 0.862 *** 0.904 *** Parent Affective Ties 0.949 1.004 Adult Support 0.981 0.991 Teen-Parent Involvement 0.790 *** 0.782 *** Future 0.948 1.001 School Importance 0.674 *** 0.823 ** Moral Absolute 0.683 *** 0.738 ** Aspiration/Expectation Gap 1.093 1.073 Neighborhood Strain 0.931 0.965 Teased 0.888 * 0.935 Body 1.058 1.011 Parental Breakup 1.286 *** 1.255 *** Temper 1.356 *** 1.213 * Sad 1.411 *** 1.222 ** Religious Salience 0.798 0.844 Parent Religious Salience 0.714 0.719 Rel X Public 0.822 0.886 Rel X Private 0.643 0.714 Parent Rel X Public 1.380 1.323 Parent Rel X Private 1.596 1.511 Religious Support 0.960 * 0.973

Max -rescaled R-Square 0.18 0.23 0.21 0.22 0.21 0.29 N 2994 2905 2877 2795 2854 2561 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

65

Table B.3

Binary Logistic Regression Odds Ratios for Getting Drunk at Time 2 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 1.712 * 1.672 * 1.550 2.106 * 1.850 * 1.657 1.790 Private 1.452 1.439 1.488 2.006 1.601 1.575 2.143 Public/Private 3.289 ** 3.586 ** 3.416 ** 5.885 *** 3.579 ** 4.208 ** 5.652 ** Teen Affective Ties 0.917 *** 0.992 Adult Support 0.991 1.001 Future 1.083 1.174 * School Importance 0.747 *** 0.874 Moral Absolute 0.776 *** 0.817 *** Non-religious Peers 1.040 0.963 Drug-using Peers 1.787 *** 1.765 *** Deviant Peers 1.219 *** 1.204 *** Trauma 1.161 0.883 Health Problems 1.165 ** 1.074 Body 0.976 0.905 Parental Breakup 1.110 1.003 Sad 1.106 0.958 Religious Salience 0.801 0.756 Parent Religious Salience 1.064 1.892 Rel X Public 0.805 1.057 Rel X Private 0.627 0.764 Rel X Public/Private 0.629 0.928 Parent Rel X Public 1.012 0.564 Parent Rel X Private 1.202 0.723 Parent Rel X Public/Private 1.216 0.610 Religious Support 0.983 0.970 Religiously-involved Peers 0.809 *** 0.890 ** Max-rescaled R-Square 0.15 0.16 0.18 0.39 0.16 0.23 0.42 N 2278 2238 2077 2146 2228 2156 1902 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

66

Table B.4

Binary Logistic Regression Odds Ratios for Getting Drunk at Time 3 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Non-Homeschool 1.151 1.133 1.144 1.369 1.141 1.320 1.286 Teen Affective Ties 1.012 1.051 * Future 1.140 * 1.126 Purpose 1.015 1.082 Moral Absolute 0.884 *** 0.945 Non-religious Peers 0.952 0.836 *** Drug-using Peers 1.612 *** 1.601 *** Trauma 1.032 0.905 Health Problems 1.175 ** 1.101 Body 0.906 * 0.944 Parental Breakup 1.054 1.045 Sad 1.058 1.005 Religious Salience 0.539 0.753 Parent Religious Salience 1.159 1.041 Rel X Non-Homeschool 1.373 1.059 Parent Rel X Non- Homeschool 0.847 0.928 Religious Peers 0.954 0.880 **

Max -rescaled R-Square 0.12 0.12 0.13 0.25 0.12 0.15 0.28 N 2220 2187 2201 2137 2162 2192 2046 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

67

Table B.5

Binary Logistic Regression Odds Ratios for Getting Drunk at Time 2 (2-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 2.096 * 2.957 ** 2.815 ** 2.897 ** 2.683 * 3.021 ** 2.571 * Private 1.969 3.038 ** 2.899 ** 2.977 ** 2.967 ** 3.498 ** 3.016 * Teen Affective Ties 0.941 ** 0.976 Parent Affective Ties 1.058 1.044 Adult Support 1.016 1.041 * Teen-Parent Involvement 0.848 *** 0.831 *** Future 1.012 1.044 School Importance 0.801 *** 0.882 Moral Absolute 0.831 0.905 Aspiration/Expectation Gap 0.909 0.871 * Neighborhood Strain 0.911 0.840 * Teased 0.896 * 0.952 Body 1.025 0.985 Parental Breakup 1.116 1.114 Temper 0.992 0.938 Sad 0.991 0.975 Religious Salience 1.103 1.104 Parent Religious Salience 0.985 1.036 Rel X Public 0.653 0.674 Rel X Private 0.606 0.726 Parent Rel X Public 1.085 1.096 Parent Rel X Private 1.175 1.068 Religious Support 0.990 0.961 Drunk at Time 1 6.096 *** 5.590 *** 6.187 *** 5.723 *** 5.596 *** 5.493 ***

Ma x-rescaled R-Square 0.15 0.26 0.27 0.27 0.27 0.28 0.30 N 2327 2325 2254 2240 2173 2224 1997 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

68

Table B.6

Binary Logistic Regression Odds Ratios for Getting Drunk at Time 3 (2-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 School History (Homeschool=ref) Public 1.060 0.827 0.810 0.754 1.236 0.808 0.950 0.856 Private 0.913 0.760 0.758 0.685 1.161 0.771 1.067 0.936 Public/Private 0.920 0.523 0.520 0.470 0.862 0.519 0.638 0.574 Teen Affective Ties 1.009 1.034 Adult Support 0.976 0.966 Future 0.934 0.926 School Importance 1.072 1.113 Moral Absolute 0.999 1.049 Non-religious Peers 0.951 0.886 * Drug-using Peers 1.256 *** 1.270 *** Deviant Peers 0.999 0.983 Trauma 1.068 1.129 Health Problems 1.002 0.952 Body 1.046 1.058 Parental Breakup 1.055 1.021 Sad 0.889 0.942 (Table B.6 continues)

69

Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Parent Religious Salience 0.459 0.362 Rel X Public 0.811 0.690 Rel X Private 1.098 0.932 Rel X Public/Private 1.631 1.702 Parent Rel X Public 2.126 2.712 Parent Rel X Private 1.046 1.287 Parent Rel X Public/Private 2.368 2.709 Religious Support 0.972 0.982 Religiously-involved Peers 0.937 0.974 Drunk at Time 2 6.850 *** 7.003 *** 7.283 *** 4.999 *** 7.289 *** 6.685 *** 5.663 ***

Max -rescaled R-Square 0.10 0.30 0.30 0.31 0.32 0.31 0.32 0.36 N 1970 1968 1947 1823 1865 1937 1879 1672 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

70

Table B.7

Binary Logistic Regression Odds Ratios for Getting Drunk at Time 3 (3-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 1.104 1.223 1.209 1.193 1.342 1.151 1.342 Private 0.950 1.124 1.122 1.138 1.240 1.159 1.456 Teen Affective Ties 0.966 0.993 Parent Affective Ties 1.002 0.993 Adult Support 1.013 1.026 Teen-Parent Involvement 0.978 0.964 Future 0.993 0.940 School Importance 0.821 ** 0.921 Moral Absolute 0.777 ** 0.826 Aspiration/Expectation Gap 0.940 0.988 Neighborhood Strain 0.890 0.906 Teased 0.956 0.974 Body 0.998 0.992 Parental Breakup 1.104 1.133 * Temper 0.952 0.957 Sad 0.937 0.870 * Religious Salience 0.678 1.221 Parent Religious Salience 0.612 0.668 Rel X Public 1.198 0.709 Rel X Private 1.406 0.805 Parent Rel X Public 1.588 1.523 Parent Rel X Private 1.203 1.031 Religious Support 1.002 0.993 Drunk at Time 1 3.019 *** 3.049 *** 2.719 *** 2.901 *** 2.775 *** 2.576 ***

Max -rescaled R-Square 0.10 0.15 0.16 0.16 0.17 0.16 0.18 N 2256 2254 2191 2167 2115 2165 1950 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

71

Table B.8

Binary Logistic Regression Odds Ratios for DUI at Time 2 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 1.468 1.671 1.184 1.663 1.840 1.311 1.261 Private 1.672 1.906 1.342 2.036 2.166 1.421 1.257 Public/Private 2.310 2.705 2.110 3.413 2.742 1.654 1.789 Teen Affective Ties 0.940 * 1.042 Adult Support 0.984 0.977 Future 1.172 * 1.132 School Importance 0.681 *** 0.824 Moral Absolute 0.850 ** 0.887 Non-religious Peers 1.094 * 1.009 Drug-using Peers 1.554 *** 1.483 *** Deviant Peers 1.275 *** 1.315 *** Trauma 1.273 1.041 Health Problems 1.125 1.058 Body 0.980 0.958 Parental Breakup 1.042 1.041 Sad 1.155 1.022 Religious Salience 0.821 1.186 Parent Religious Salience 1.489 2.099 Rel X Public 1.010 0.983 Rel X Private 0.890 0.567 Rel X Public/Private 0.418 0.425 Parent Rel X Public 0.811 0.588 Parent Rel X Private 0.796 0.665 Parent Rel X Public/Private 1.096 0.630 Religious Support 0.985 0.964 Religiously-involved Peers 0.729 *** 0.774 *** Max-rescaled R-Square 0.17 0.17 0.18 0.34 0.18 0.21 0.35 N 2280 2240 2079 2147 2230 2157 1903 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

72

Table B.9

Binary Logistic Regression Odds Ratios for DUI at Time 2 (2-Wave Longitudinal)

Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 School History (Homeschool=ref) Public 1.785 1.755 1.822 2.541 1.962 2.832 Private 1.973 2.003 2.097 2.862 2.366 3.608 Teen Affective Ties 0.892 *** 0.951 Parent Affective Ties 0.942 0.989 Adult Support 1.024 1.066 ** Teen-Parent Involvement 0.812 *** 0.800 *** Future 1.009 1.056 School Importance 0.634 *** 0.737 *** Moral Absolute 0.776 * 0.836 Aspiration/Expectation Gap 1.039 1.030 Neighborhood Strain 0.894 0.880 Teased 0.815 ** 0.877 Body 1.024 0.998 Parental Breakup 1.142 1.131 Temper 1.432 ** 1.287 * Sad 1.147 1.052 Religious Salience 1.662 2.385 Parent Religious Salience 0.634 0.684 Rel X Public 0.425 0.341 Rel X Private 0.415 0.309 Parent Rel X Public 1.898 1.674 Parent Rel X Private 1.892 1.771 Religious Support 0.967 0.918 ** Max-rescaled R-Square 0.16 0.20 0.19 0.18 0.19 0.24 N 2329 2258 2244 2177 2228 2001 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

73

Table B.10

Binary Logistic Regression Odds Ratios for Binge Drinking at Time 3 Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Non-Homeschool 1.163 1.146 1.145 1.369 1.142 1.320 1.286 Teen Affective Ties 1.014 1.051 * Future 1.138 * 1.126 Purpose 1.013 1.082 Moral Absolute 0.884 *** 0.945 Non-religious Peers 0.952 0.836 *** Drug-using Peers 1.611 *** 1.601 *** Trauma 1.028 0.905 Health Problems 1.178 ** 1.101 Body 0.906 * 0.944 Parental Breakup 1.054 1.045 Sad 1.058 1.005 Religious Salience 0.539 0.753 Parent Religious Salience 1.159 1.041 Rel X Non-Homeschool 1.373 1.059 Parent Rel X Non- Homeschool 0.847 0.928 Religious Peers 0.954 0.880 **

Max -rescaled R-Square 0.12 0.12 0.13 0.25 0.12 0.15 0.28 N 2212 2180 2199 2136 2160 2192 2046 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

74

Table B.11

Binary Logistic Regression Odds Ratios for Binge Drinking at Time 3 (2-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 School History (Homeschool=ref) Public 1.066 1.034 0.918 1.530 1.074 1.225 1.127 Private 0.916 0.908 0.838 1.386 0.956 1.317 1.254 Public/Private 0.923 0.920 0.834 1.448 0.932 1.147 1.023 Teen Affective Ties 0.982 1.032 Adult Support 0.978 0.970 Future 0.996 0.981 School Importance 0.939 1.055 Moral Absolute 0.895 ** 0.981 Non-religious Peers 0.965 0.886 ** Drug-using Peers 1.475 1.489 *** Deviant Peers 1.046 *** 1.030 Trauma 1.112 1.071 Health Problems 1.071 0.996 Body 1.016 1.018 Parental Breakup 1.073 1.015 Sad 0.954 0.925 Religious Salience 0.998 1.083 Parent Religious Salience 0.574 0.563 Rel X Public 0.815 0.773 Rel X Private 0.970 0.936 Rel X Public/Private 1.525 1.938 Parent Rel X Public 1.727 1.771 Parent Rel X Private 0.968 0.977 Parent Rel X Public/Private 1.876 1.597 Religious Support 0.971 0.977 Religiously-involved Peers 0.870 *** 0.944 Max-rescaled R-Square 0.10 0.10 0.11 0.22 0.10 0.15 0.25 N 1964 1944 1820 1861 1934 1875 1668 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

75

Table B.12

Binary Logistic Regression Odds Ratios for Binge Drinking at Time 3 (3-Wave Longitudinal) Independent Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 School History (Homeschool=ref) Public 1.113 1.089 1.086 1.165 1.031 1.113 Private 0.955 0.950 0.977 1.018 0.999 1.163 Teen Affective Ties 0.943 ** 0.979 Parent Affective Ties 0.996 0.990 Adult Support 1.007 1.021 Teen-Parent Involvement 0.933 0.928 Future 0.980 0.937 School Importance 0.780 *** 0.902 Moral Absolute 0.752 ** 0.813 * Aspiration/Expectation Gap 0.955 0.998 Neighborhood Strain 0.882 0.904 Teased 0.930 0.961 Body 1.005 0.991 Parental Breakup 1.140 * 1.161 ** Temper 1.005 0.980 Sad 0.997 0.900 Religious Salience 0.591 0.988 Parent Religious Salience 0.681 0.752 Rel X Public 1.312 0.850 Rel X Private 1.510 0.961 Parent Rel X Public 1.414 1.340 Parent Rel X Private 1.085 0.914 Religious Support 0.995 0.994 Max-rescaled R-Square 0.10 0.11 0.12 0.12 0.12 0.15 N 2248 2185 2161 2110 2159 1945 Note: All models control for demographics; results are available upon request. *p<0.05 **p<0.01 ***p<0.001 (two-tailed tests)

76

BIBLIOGRAPHY

Agnew, R. (1992). Foundation for a General Strain Theory of Crime and Delinquency. Criminology, 30(1), 47–88.

Agnew, R. (2006). Pressured into Crime: An Overview of General Strain Theory. Los Angeles, CA: Roxbury Publishing Company.

Akers, R. L. (1985). Deviant behavior: a social learning approach. Wadsworth Pub. Co.

Akers, R. L. (2009). Social Learning and Social Structure: A General Theory of Crime and Deviance. Transaction Publishers.

Akers, R. L., Krohn, M. D., Lanza-Kaduce, L., & Radosevich, M. (1979). Social Learning and Deviant Behavior: A Specific Test of a General Theory. American Sociological Review, 44(4), 636–655.

Aurini, J., & Davies, S. (2005). Choice without Markets: Homeschooling in the Context of Private Education. British Journal of Sociology of Education, 26(4), 461–474.

Bachman, J. G., Wallace Jr., J. M., O’Malley, P. M., Johnston, L. D., Kurth, C. L., & Neighbors, H. W. (1991). Racial/ethnic Differences in Smoking, Drinking, and Illicit Drug Use among American High School Seniors, 1976-1989. American Journal of Public Health, 81(3), 372–377.

Baier, C. J., & Wright, B. R. E. (2001). “If You Love Me, Keep My Commandments”: A Meta- Analysis of the Effect of Religion on Crime. Journal of Research in Crime and Delinquency, 38(1), 3–21.

Ballmann, R. E. (1987). The how and why of Home Schooling. Westchester, Illinois: Crossway.

Barfield, R. (2002). Real-Life Homeschooling: The Stories of 21 Families Who Teach Their Children at Home. Simon and Schuster.

Basham, P., Merrifield, J., & Hepburn, C. (2007). Home Schooling: From the Extreme to the Mainstream (pp. 1–24). Vancouver, BC: The Fraser Institute.

Bauman, K. J. (2001). Home Schooling in the United States: Trends and Characteristics (No. 53). Washington, D.C.: U.S. Census Bureau. Retrieved from http://epaa.asu.edu/ojs/article/view/305

77

Bernard, T. J., Snipes, J. B., & Gerould, A. L. (2010). Vold’s Theoretical Criminology (Vol. 6th). New York: Oxford University Press.

Bielick, S., Chandler, K., & Broughman, S. P. (2001). Homeschooling in the United States: 1999. Washington, DC: United States Department of Education, National Center for Education Statistics.

Bryant, A. N., & Astin, H. S. (2008). The Correlates of Spiritual Struggle During the College Years. The Journal of Higher Education, 79(1), 1–27.

Cizek, G. J. (1994). Religious Education in Home Schools: Goals/Outcomes Mismatch? Religious Education, 89(1), 43–51.

Cochran, J. K., Wood, P. B., & Arneklev, B. J. (1994). Is the Religiosity-Delinquency Relationship Spurious - a Test of Arousal and Social-Control Theories. Journal of Research in Crime and Delinquency, 31(1), 92–123.

Collom, E. (2005). The Ins and Outs of Homeschooling: The Determinants of Parental Motivations and Student Achievement. Education and Urban Society, 37 (3), 307–335.

Cooper, B. S., & Sureau, J. (2007). The Politics of Homeschooling: New Developments, New Challenges. Educational Policy, 21(1), 110–131.

Cordner, A. (2012). The health care access and utilization of homeschooled children in the United States. Social Science & Medicine, 75(2), 269–273.

Cullen, F. T. (2011). Toward a Criminology of Religion: Comment on Johnson and Jang. In Contemporary Issues in Criminological Theory and Research: The Role of Social Institutions (pp. 151–161). Belmont, CA: Wadsworth.

Curran, P. J., Stice, E., & Chassin, L. (1997). The relation between adolescent alcohol use and peer alcohol use: A longitudinal random coefficients model. Journal of Consulting and Clinical Psychology, 65(1), 130–140.

Davies, S., & Quirke, L. (2005). Providing for the Priceless Student: Ideologies of Choice in an Emerging Educational Market. American Journal of Education, 111(4), 523–547.

Ellis, L. (1987). Religiosity and Criminality from the Perspective of Arousal Theory. Journal of Research in Crime and Delinquency, 24(3), 215–232.

Farris, M. P., & Woodruff, S. A. (2000). The Future of Home Schooling. Peabody Journal of Education, 75(1/2), 233–255.

Gaither, M. (2008). Homeschool: An American History. Palgrave Macmillan.

78

Gaither, M. (2009). Homeschooling in the USA: Past, present and future. Theory and Research in Education, 7(3), 331–346.

Gottfredson, M. R., & Hirschi, T. (1990). A General theory of Crime. Stanford, CA: Stanford University Press.

Higgins, P. C., & Albrecht, G. L. (1977). Hellfire and Delinquency Revisited. Social Forces, 55(4), 952–958.

Hirschi, T. (1969). Causes of Delinquency. Berkeley, CA: University of California Press.

Hirschi, T., & Stark, R. (1969). Hellfire and Delinquency. Social Problems, 17(2), 202–213.

Jang, S. J., & Johnson, B. R. (2003). Strain, Negative Emotions, and Deviant Coping Among African Americans: A Test of General Strain Theory. Journal of Quantitative Criminology, 19(1), 79–105.

Jang, S. J., & Johnson, B. R. (2011). The Effects of Childhood Exposure to Drug Users and Religion on Drug Use in Adolescence and Young Adulthood. Youth & Society, 43(4), 1220–1245.

Jeub, C. (1994). Why Parents Choose Home Schooling. Educational Leadership, 52(1), 50–52.

Johnson, B. R., & Jang, S. J. (2011). Crime and Religion: Assessing the Role of the Faith Factor. In Contemporary Issues in Criminological Theory and Research: The Role of Social Institutions (pp. 117–149). Belmont, CA: Wadsworth.

Johnson, B. R., Jang, S. J., Larson, D. B., & De Li, S. (2001). Does Adolescent Religious Commitment Matter? A Reexamination of the Effects of Religiosity on Delinquency. Journal of Research in Crime and Delinquency, 38(1), 22.

Kunzman, R. (2009a). Understanding homeschooling: A better approach to regulation. Theory and Research in Education, 7(3), 311–330.

Kunzman, R. (2009b). Write These Laws on Your Children: Inside the World of Conservative Christian Homeschooling. Beacon Press.

Kunzman, R. (2010). Homeschooling and religious fundamentalism. International Journal of Elementary Education, 3(1), 17–28.

Kunzman, R. (2012). Education, Schooling, and Children’s Rights: The Complexity of Homeschooling. Educational Theory, 62(1), 75–89.

Lines, P. M. (1994). Homeschooling: Private Choices and Public Obligations (pp. 1–25). Washington, D.C.: U.S. Department of Education, Office of Educational Research and Improvement, Office of Research.

79

Longest, K. C., & Shanahan, M. J. (2007). Adolescent Work Intensity and Substance Use: The Mediational and Moderational Roles of Parenting. Journal of Marriage and Family, 69(3), 703–720.

Longest, K. C., & Vaisey, S. (2008). Control or Conviction: Religion and Adolescent Initiation of Marijuana Use. Journal of Drug Issues, 38(3), 689–715.

Lubienski, C. (2000). Whither the Common Good? A Critique of Home Schooling. Peabody Journal of Education, 75(1/2), 207–232.

Lyman, I. (1998). Homeschooling: Back to the Future? The Cato Institute. Retrieved from http://www.cato.org/publications/policy-analysis/homeschooling-back-future

MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of the Mediation, Confounding and Suppression Effect. Prevention Science : The Official Journal of the Society for Prevention Research, 1(4), 173.

Marsden, G. M. (1980). Fundamentalism and American culture: The Shaping of Twentieth Century Evangelicalism, 1870-1925. New York: Oxford University Press.

Mayrl, D., & Oeur, F. (2009). Religion and Higher Education: Current Knowledge and Directions for Future Research. Journal for the Scientific Study of Religion, 48(2), 260– 275.

Medlin, R. G. (2000). Home Schooling and the Question of Socialization. Peabody Journal of Education, 75(1/2), 107–123.

Moffitt, T. E. (1993). Adolescence-Limited and Life-Course-Persistent Antisocial-Behavior - a Developmental Taxonomy. Psychological Review, 100(4), 674–701.

Murphy, J. (2012). Homeschooling in America: Capturing and Assessing the Movement. SAGE.

Nelms, L. W., Hutchins, E., Hutchins, D., & Pursley, R. J. (2007). Spirituality and the Health of College Students. Journal of Religion and Health, 46(2), 249–265.

Osgood, D. W., Wilson, J. K., O’Malley, P. M., Bachman, J. G., & Johnston, L. D. (1996). Routine Activities and Individual Deviant Behavior. American Sociological Review, 61(4), 635–655.

Perkins, H. W. (1987). Parental Religion and Alcohol Use Problems as Intergenerational Predictors of Problem Drinking among College Youth. Journal for the Scientific Study of Religion, 26(3), 340–357.

80

Princiotta, D., Bielick, S., & Chapman, C. (2004). 1.1 Million Homeschooled Students in the United States in 2003. Washington, DC: United States Department of Education, National Center for Education Statistics. Retrieved from http://nces.ed.gov/pubs2004/2004115.pdf

Princiotta, D., Bielick, S., & Chapman, C. (2006). Homeschooling in the United States: 2003. Washington, DC: United States Department of Education, National Center for Education Statistics. Retrieved from http://nces.ed.gov/pubs2006/2006042.pdf

Ray, B. D. (2004). Homeschoolers on to College: What Research Shows Us. The Journal of College Admission, 185, 5–11.

Regnerus, M. D. (2003). Moral Communities and Adolescent Delinquency: Religious Contexts and Community Social Control. The Sociological Quarterly, 44(4), 523–554.

Sampson, R. J., & Laub, J. H. (1993). Crime in the Making: Pathways and Turning points through Life. Cambridge, MA: Harvard University Press.

Sikkink, D. (1999). The Social Sources of Alienation from Public Schools. Social Forces, 78(1), 51–86.

Smith, C., & Denton, M. L. (2005). Soul Searching: The Religious and Spiritual Lives of American Teenagers. New York, NY: Oxford University Press.

Smith, C., Emerson, M., Gallagher, S., Kennedy, P., & Sikkink, D. (1998). American Evangelicalism: Embattled and Thriving. Chicago, IL: University of Chicago Press.

Stark, R. (1996). Religion as Context: Hellfire and Delinquency One More Time. Sociology of Religion, 57(2), 163–173.

Steensland, B., Park, J., Regnerus, M., Robinson, L., Wilcox, W. B., & Woodberry, R. (2000). The Measure of American Religion: Toward Improving the State of the Art. Social Forces, 79(1), 291–318.

Sutherland, E. H., & Cressey, D. R. (1970). Criminology (Vol. Eighth). Philadelphia, PA: J. B. Lippincott Company.

Thornberry, T. P. (1989). Reflections on the Advantages and Disadvantages of Theoretical Integration. In S. F. Messner, M. D. Krohn, & A. E. Liska (Eds.), Theoretical Integration in the Study of Deviance and Crime: Problems and Prospects (pp. 51–60). Albany, NY: State University of New York Press.

Uecker, J. E. (2008). Alternative Schooling Strategies and the Religious Lives of American Adolescents. Journal for the Scientific Study of Religion, 47(4), 563–584.

81

Uecker, J. E., Regnerus, M. D., & Vaaler, M. L. (2007). Losing My Religion: The Social Sources of Religious Decline in Early Adulthood. Social Forces, 85(4), 1667–1692.

Ulmer, J. T. (2012). Religion as a Unique Cultural Influence on Crime and Delinquency: Expanding on Jang and Johnson’s Agenda. In Contemporary Issues in Criminological Theory and Research: The Role of Social Institutions (pp. 163–171). Belmont, CA: Wadsworth, Cengage Learning.

82